ALICE-TDR-016 CERN-LHCC-2013-020

Colour reproduction
March 3, 2014
CERN Graphic Charter: use of the outline version of the CERN logo
The badge version must only be reproduced on a
Clear space
A clear space must be respected around the logo:
Minimum size
Print: 10mm
plain white background using the correct blue: other graphical or text elements must be no closer Web: 60px
Pantone: 286 than 25% of the logo’s width.
CMYK: 100 75 0 0
RGB: 56 97 170 Placement on a document
Web: #3861AA Use of the logo at top-left or top-centre of a
document is reserved for official use.
Where colour reproduction is not faithful, or the
background is not plain white, the logo should be
reproduced in black or white – whichever provides
the greatest contrast. The outline version of the
logo may be reproduced in another colour in
instances of single-colour print.

Technical Design Report

This is an output file created in Illustrator CS3

for the

Upgrade of the
ALICE Time Projection Chamber
The ALICE Collaboration⇤
CERN-LHCC-2013-020 / ALICE-TDR-016
03/03/2014

Copyright CERN, for the benefit of the ALICE Collaboration.

This article is distributed under the terms of Creative Commence Attribution License (CC-BY-3.0), which
permits any use provided the original author(s) and source are credited.

⇤ See list of authors in App. C

Executive summary

This Technical Design Report describes the upgrade of the ALICE Time Projection Chamber (TPC),
which is an integral part of the ALICE upgrade strategy after LHC Long Shutdown 2 (LS2) [1]. The
main design considerations and technical specifications are summarized in the following.
In Chap. 1 the chief scientific goals of the future ALICE physics program are briefly reviewed and the
resulting requirements for the TPC upgrade are presented. The expected increase of the LHC luminosity
after LS2 to about 50 kHz in Pb–Pb implies that TPC operation with a gating grid is no longer possible.
This motivates the choice of GEMs for the new readout chambers, since they feature intrinsic ion block-
ing capabilities that avoid massive charge accumulation in the drift volume from back-drifting ions, and
prevent excessive space-charge distortions. However, GEMs do not feature the same opacity for ions as a
gating grid. The requirement to keep the distortions at a tolerable level leads to an upper limit of 1 % for
the fractional ion backflow (IBF) at a gas gain of 2000. The resulting space-charge distortions are less
than 10 cm in most of the TPC drift volume and can be calibrated with sufficient precision. The achieve-
ment of this goal is the result of a major R&D effort presented in this document. The replacement of the
existing MWPC-based readout chambers by GEMs implies also the necessity for new readout electronics
that accommodate the negative signal polarity and enable continuous data readout. Moreover, the high
data rate requires data compression by a factor of about 20 in order to match the anticipated bandwidth
to permanent storage. This implies that significant pattern recognition and data format optimization must
be performed online. The present particle identification (PID) capability via the measurement of the
specific ionization dE/dx and the combined momentum resolution of the central barrel tracking system
must be retained by the TPC upgrade.
Most of the main components of the existing TPC, including the field cage, the endplates, the gas system
and services will be reused after the upgrade. A short overview of those components is given in Chap. 2.
The choice of the detector gas is presented in Chap. 3. The requirements in terms of drift velocity,
diffusion, gas gain, and ion mobility lead to Ne-CO2 -N2 (90-10-5) as a baseline gas mixture for the TPC.
The technical solution that was chosen for the new readout chambers is presented in Chap. 4. Their design
is such that the segmentation in azimuth and the division into inner and outer readout chambers (IROCs
and OROCs) is identical to those of the existing detector. The new readout chambers will employ stacks
of four GEM foils for gas amplification and anode pad readout. Quadruple GEM stacks have proven to
provide sufficient ion blocking capabilities at the required gas gain of 2000 in Ne-CO2 -N2 (90-10-5), in
particular when GEM foils with large hole pitch are used. The size of the readout chambers, in particular
of the OROCs, favors the use of large-size GEM foils manufactured with the single-mask technology.
Such foils were recently implemented successfully in the new KLOE-2 tracking system. Application of
this technique allows production of the readout chambers such that a single large GEM foil per layer can
be used in the IROCs, and three large GEM foils per layer in the OROCs. The readout pad structure of the
existing TPC, with three different pad sizes that increase from small to large radii, will be modified only
slightly for the new readout chambers. The local position resolution of the GEM detectors, in particular
at short drift distances, is slighly worse than that of the present system due to the lack of a pad response

i
ii The ALICE Collaboration

function. However, this has no observable effect on the combined momentum resolution of the ALICE
central barrel system, as shown in Chap. 7. At the same time, the difference in coupling to the readout
plane will reduce the detector occupancy in GEMs compared to that in the existing MWPC. To ensure
safe operation and long-term reliability, a careful quality assurance procedure for GEM foils and readout
chambers at the various assembly stages is being developed and described in Chap. 4.
The main results from R&D with small and full-sized prototypes are discussed in Chap. 5. In a quadruple
GEM system including foils with large hole pitch, ion backflow values below 1 % at a gas gain of 2000
and an energy resolution s (55 Fe) = 12 % are observed. These operational conditions match safely the
requirements of the detector and leave room for further optimizations. The ion backflow results are well
described by a microscopic detector simulation based on the Garfield++ framework. Test beam results
from a full-size IROC prototype equipped with a triple GEM stack show that the same dE/dx resolution
as in MWPC-based readout chambers can be achieved. Some of the possible further R&D directions,
including COBRA GEMs and MicoMegas, are described in Chap. 9.
The new TPC front-end electronics and readout system is discussed in Chap. 6. The main specifications
remain unchanged with respect to the existing detector. However, new characteristics have to be incorpo-
rated: The front-end ASIC has to amplify and process signals with opposite polarity as compared to those
of an MWPC. The detector signals have to be sampled continuously while concurrently the aquired data
is transferred off-detector. Finally, the data throughput will be strongly increased with respect to the cur-
rent system. The SAMPA project aims at delivering an ASIC fulfilling these requirements. The SAMPA
ASICs connect to the common read-out unit (CRU), which provides the interface to the online comput-
ing system, the trigger system and the DCS, through optical fibers via the GBT link. Both, SAMPA and
CRU, are common solutions for different ALICE subsystems and are described in a separate Technical
Design Report on the upgrade of the readout and trigger system [2].
In Chap. 7 an evaluation of the performance of the upgraded TPC is presented. Using a microscopic
simulation of the TPC with GEM readout, the intrinsic momentum and dE/dx resolutions are found to
be the same as with the existing TPC, if the local energy resolution does not exceed s (55 Fe) = 12 %.
No significant deterioration of the tracking efficiency nor the momentum resolution is observed when
event pileup, corresponding to collision rates of 50 kHz, is introduced. The dE/dx resolution slightly
worsens with increasing occupancy from 5.5 % in isolated pp events without pileup to about 7.5 % in
central Pb–Pb at 50 kHz. This behaviour is similar when using MWPC or GEM and is understood in
terms of an increasing overlap of clusters.
Detailed calculations of the ion space-charge density in the TPC caused by back-drifting ions from the
amplification region are presented. The time it takes for ions to drift to the central electrode is close
to 160 ms, leading to an average pileup of ions from about 8000 collisions in the TPC drift volume at
an interaction rate of 50 kHz. At a gas gain of 2000 in Ne-CO2 -N2 (90-10-5) and an ion backflow of
1 %, this results in space-charge distortions that stay below 10 cm in most of the TPC volume, with the
exception of the innermost region (small r) and at the largest drift length, where radial distortions up to
20 cm are observed. In order to reach the intrinsic track resolution of the TPC of a few hundred µm,
distortion corrections with a precision on the level of 10 3 need to be performed. It is demonstrated
that statistical fluctuations of the collision rate and of the charged-particle multiplicity lead to temporal
variations of the space-charge density that are of the order of a few percent, i.e. significantly larger
than the required precision of the correction. This implies that the space-charge density and thus the
space-charge distortion corrections must be determined as a function of time during the data acquisition.
Detailed studies demonstrate that a given space-charge configuration can be considered as static over
time intervals of ⇠ 5 ms.
The strategy for online reconstruction and calibration is discussed in Chap. 8. Online reconstruction is
necessary in order to achieve data compression by a factor of 20 as compared to the raw data size, and
TPC Upgrade TDR iii

to allow for permanent storage of the data. Such compression factors can be achieved if the association
of clusters to tracks can be performed online, which implies also the necessity for sufficient online
correction of the space-charge distortions. We argue that the standard TPC tracking scheme including the
use of a coarse distortion correction map that can be determined online provides high tracking and cluster
association efficiency. Moreover, it provides sufficient spatial accuracy to allow efficient matching of the
TPC tracks to the Inner Tracking System (ITS), which is necessary to conduct the subsequent calibration
steps. The final space-point calibration is based on a residual distortion correction employing external
reference track information from the ITS and the Transition Radiation Detector (TRD). In this step, the
residual space-charge density fluctuations, which require an updated residual correction map every 5 ms,
are addressed. We demonstrate that the available track statistics accumulated over such time intervals
is adequate to determine the residual distortions in a grid of sufficient spatial granularity, and with a
precision that is consistent with the intrinsic resolution of the TPC.
Chapters 10 and 11 discuss the detector control system and installation, commissioning and services.
Most of the aspects presented here build on the existing system and profit from the experience gained
with it.
The project organization and considerations on budget and time schedule are discussed in Chap. 12.
The ALICE TPC collaboration has gained a significant number of new collaborators with considerable
experience and manpower to carry out the upgrade project. The overall CORE cost of the project of
9.2 MCHF is consistent with the ongoing funding requests in different countries. The overall time line
of the project matches the current LHC schedule and foresees that the upgraded TPC will be ready for
operation after LS2.
iv The ALICE Collaboration
Contents

1 Physics objectives and design considerations 1

1.1 Physics Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Upgrade concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Design considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Detector overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2 Mechanical structure, field cage, and gas system 7

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Field cage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Endplates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4 Last resistor and skirt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.5 Service Support Wheel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.6 Gliders and I-bars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.7 Gas system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

3 Gas choice 13

4 Readout chambers 15
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
4.2 Mechanical structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4.3 GEM planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3.1 General structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3.2 Inner readout chambers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.3.3 Outer readout chambers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.4 High voltage supply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4.1 System overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.4.2 Typical HV settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

v
vi The ALICE Collaboration

4.5 Readout pad plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

4.6 Interface to front-end electronics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.7 Quality assurance of GEM foils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.7.1 Electrical characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.7.2 Optical scanning characterization . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.7.3 GEM gain mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.8 Quality assurance of chambers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

5 R&D with prototypes 41

5.1 R&D with small prototypes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.1.1 Experimental setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.1.2 Gain stability measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.1.3 Results of ion backflow measurements . . . . . . . . . . . . . . . . . . . . . . . 44
5.1.4 Discharge probability studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.1.5 Comparison with simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2 Full-size IROC prototype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.2.1 Detector design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.2.2 Quality assurance (QA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2.3 Detector assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2.4 HV supply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.2.5 Prototype commissioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2.6 Test campaign at the CERN PS . . . . . . . . . . . . . . . . . . . . . . . . . . 58

6 Front-end electronics and readout 63

6.1 System overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.2 Pileup and occupancies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.3 Data rates and bandwidth considerations . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.4 Common front-end ASIC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.4.2 General requirements for the analog part . . . . . . . . . . . . . . . . . . . . . . 67
6.4.3 Signal shaping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.4.4 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.4.5 Further requirements for the analog part . . . . . . . . . . . . . . . . . . . . . . 72
6.4.6 Electrostatic discharge protection . . . . . . . . . . . . . . . . . . . . . . . . . 73
TPC Upgrade TDR vii

6.4.7 Analog-to-digital conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

6.4.8 Digital signal processor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
6.4.9 Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
6.5 Front-end card . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.5.1 Partitioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.5.2 PCB design and layout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.5.3 System level input protection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.5.4 Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.5.5 Irradiation campaign . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
6.6 Common Readout Unit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

7 Simulation and detector performance 81

7.1 Current performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.1.1 Tracking performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.1.2 Particle identification performance . . . . . . . . . . . . . . . . . . . . . . . . . 82
7.2 Intrinsic performance of the upgraded TPC . . . . . . . . . . . . . . . . . . . . . . . . 83
7.2.1 Microscopic GEM simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
7.2.2 Tracking performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
7.2.3 Particle identification performance . . . . . . . . . . . . . . . . . . . . . . . . . 84
7.3 Performance with event pileup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
7.4 Space-charge distortions and corrections . . . . . . . . . . . . . . . . . . . . . . . . . . 86
7.4.1 Space-charge sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
7.4.2 Magnitude of the distortions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
7.4.3 Simulation of the space-charge distortions . . . . . . . . . . . . . . . . . . . . . 90
7.4.4 Space-charge density fluctuations . . . . . . . . . . . . . . . . . . . . . . . . . 91
7.4.5 Impact of the fluctuations on the distortion corrections . . . . . . . . . . . . . . 94
7.5 Performance with residual space-charge distortions . . . . . . . . . . . . . . . . . . . . 95

8 Online reconstruction, calibration, and monitoring 97

8.1 Continuous TPC operation at high luminosities . . . . . . . . . . . . . . . . . . . . . . 97
8.1.1 LHC conditions in RUN 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
8.1.2 TPC reconstruction, calibration, and data compression in RUN 3 . . . . . . . . . 99
8.2 Space-charge distortion corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8.2.1 TPC coordinate transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
viii The ALICE Collaboration

8.2.2 Space point corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

8.2.3 Space-charge density maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
8.3 First reconstruction stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
8.3.1 Standard tracking approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
8.3.2 Performance using corrections from the scaled average map . . . . . . . . . . . 104
8.4 Second reconstruction stage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
8.4.1 ITS-TRD track interpolation approach . . . . . . . . . . . . . . . . . . . . . . . 106
8.4.2 Momentum resolution after residual correction . . . . . . . . . . . . . . . . . . 108
8.5 Further optimizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
8.5.1 TPC standalone tracking approach . . . . . . . . . . . . . . . . . . . . . . . . . 110
8.5.2 Space-charge calibration by current measurements . . . . . . . . . . . . . . . . 113
8.6 Additional calibration requirements, monitoring, and quality control . . . . . . . . . . . 114
8.6.1 Additional calibration requirements . . . . . . . . . . . . . . . . . . . . . . . . 114
8.6.2 Monitoring and quality control . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

9 Alternative R&D options 119

9.1 R&D with COBRA GEMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
9.1.1 Characterization of single COBRA GEMs . . . . . . . . . . . . . . . . . . . . . 121
9.1.2 Triple structures with COBRA and standard GEMs . . . . . . . . . . . . . . . . 123
9.1.3 Energy resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.1.4 Conclusion and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
9.2 Studies with fast gas mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
9.2.1 Conclusion and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
9.3 R&D with hybrid gain structures: 2 GEMs + MicroMegas . . . . . . . . . . . . . . . . . 128
9.3.1 Conclusion and outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

10 Detector control system 131

10.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
10.1.1 Hardware architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
10.1.2 Software architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
10.1.3 System implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
10.1.4 Interfaces to devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
10.1.5 Interlocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
10.2 Front-end electronics control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
TPC Upgrade TDR ix

10.2.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

10.2.2 Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
10.2.3 Configuration and control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
10.3 Parameter export for online calibration and reconstruction . . . . . . . . . . . . . . . . . 134

11 Installation, commissioning and services 135

11.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
11.2 Installation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
11.3 Commissioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
11.4 Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
11.4.1 High voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
11.4.2 Low voltage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
11.4.3 Cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
11.4.4 Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139

12 Project organization, cost estimate and time line 141

12.1 Participating institutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
12.2 Cost estimate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
12.3 Schedule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
12.4 TPC upgrade TDR editorial committee . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
12.5 TPC upgrade TDR task force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

A Coordinate systems 147

A.1 Global coordinate system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
A.2 Local coordinate system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

B TPC upgrade collaboration 149

C The ALICE Collaboration 153

References 161

List of Figures 169

List of Tables 175

x The ALICE Collaboration
Chapter 1

Physics objectives and design

considerations

Studies of heavy-ion collisions at the Large Hadron Collider (LHC) are ideally suited to probe funda-
mental properties of QCD, including its non-perturbative aspects related to color charge deconfinement,
and the restoration of chiral symmetry. In particular, heavy-ion collisions allow the detailed characteriza-
tion of the Quark-Gluon Plasma (QGP), and the nature of the phase transition between QGP and normal
hadronic matter.
ALICE1 at the LHC is dedicated to these studies [1]. Operation of the ALICE detector in collisions of
208 Pb-ions at ps 1
NN = 2.76 TeV in 2010 and 2011 (integrated luminosity int = 0.16 nb ) has demon-
strated its excellent tracking and particle identification (PID) capabilities in an environment of large
charged-particle densities. The lead-ion campaigns in RUN 2 after the LHC Long Shutdown 1 (LS1),
starting in 2015, will conclude the initial LHC heavy-ion programme with 1 nb 1 .
A significant increase of the LHC luminosity for heavy ions is expected in RUN 3 after Long Shutdown 2
(LS2), leading to collision rates of about 50 kHz and int = 10 nb 1 . This implies a substantial enhance-
ment of the sensitivity to a number of rare probes that are key observables for the characterization of
strongly interacting matter at high temperature.
In order to fully exploit the scientific potential of the LHC in RUN 3, ALICE plans to extend its physics
programme by improving its detector performance. As an integral part of the ALICE upgrade strat-
egy, this Technical Design Report describes the concept of a novel readout scheme, based on GEM2
technology, that will be implemented in the Time Projection Chamber (TPC) upgrade. The present
MWPC3 -based readout chambers will be replaced by a GEM system to match the TPC readout rate with
the increased Pb–Pb collision rate of the LHC in RUN 3. At the same time, the front-end electronics and
readout system will be replaced in order to match the new readout chamber technology and increased
data rates. After these upgrades, the data collection rate of the TPC will be increased by about a factor
100 in the high-luminosity environment of the LHC in RUN 3, while the tracking and PID capabilities of
the present TPC will be retained.

1.1 Physics Objectives

The scientific goals of the upgraded ALICE detector are described in a comprehensive Letter of Intent [2].
They are aimed at improving measurements for understanding heavy-quark production at low transverse
1A Large Ion Collider Experiment (ALICE)
2 GasElectron Multiplier (GEM)
3 Multi-Wire Proportional Chamber (MWPC)

1
2 The ALICE Collaboration

momentum (pT ), mechanisms of quarkonium production and interaction in the QGP, initial tempera-
ture and partonic equation of state, possible effects of chiral symmetry restoration, parton energy loss,
medium modification and its dependence on properties of the parton and the QGP, and exotic hadronic
states. To accomplish these, the following measurements will be undertaken in the central barrel of the
ALICE detector:

– Yields and azimuthal distributions of hadrons containing heavy quarks (c, b) to study the mecha-
nism of heavy-quark thermalization in the QGP.

– Production of quarkonia at low pT , in particular the study of their possible dissociation and regen-
eration mechanisms in the QGP.

– Low-mass dielectron production to extract information on early temperature and the partonic equa-
tion of state, and to characterize the chiral phase transition.

– Jets and jet correlations, in particular their structure and particle composition, to study the mech-
anism of partonic energy loss in medium and its dependence on parton color-charge, mass and
energy.

– The production of nuclei, anti-nuclei and hyper-nuclei as well as exotic hadronic states such as the
H-dibaryon.

These measurements require excellent charged-particle tracking capabilities as well as a variety of PID
techniques in the central barrel, down to the lowest possible pT . Measurements at low transverse mo-
menta typically imply small signal-to-background ratios, which limits the applicability of standard low-
level triggering schemes. As a consequence, the detectors and readout systems must allow to operate at
very high readout and data acquisition rates in order to collect sufficient statistics.

1.2 Upgrade concept

A significant increase in the sensitivity of ALICE to these observables is achieved by major upgrades of
its detectors and readout systems in the central barrel of the experiment. A new Inner Tracking System
(ITS) improves by a factor 3 the resolution for secondary vertices and extends its tracking capabilities to
lower transverse momenta [3]. Moreover, the material budget of the ITS is reduced from 1.1 % of X0 per
detector layer to to 0.3 % for the three inner layers and 0.8 % for the four outer layers. The performance
of the new ITS will significantly expand the physics reach of the ALICE central barrel, in particular in
the heavy-flavor and low-mass dielectron sector.
On the other hand, precision measurements of the key observables outlined above require tracking and
PID information from the TPC. In order to conduct the envisaged physics programme with optimum
precision, exploiting the full LHC luminosity, the TPC will be upgraded. This upgrade is intended
primarily to overcome the rate limitation of the present system, which arises from the gated operation of
the MWPC-based readout chambers.
The ALICE TPC is the largest detector of its type, with an overall active volume of about 90 m3 [4, 5].
The TPC employs a cylindrical field cage with a central high voltage electrode at z = 0 (for a definition
of the ALICE coordinate system see App. A.1) and a readout plane on each endplate. It covers full
azimuth in |h| < 0.9 and provides charged-particle tracking over a wide transverse momentum range.
The readout planes consist of 72 MWPC-based readout chambers, with a total of about 550,000 readout
cathode pads.
Charged-particle tracking and PID via ionization energy loss (dE/dx) in the TPC is accomplished by the
measurement of the ionization of up to 159 samples along a trajectory of ⇠160 cm. In pp and central
TPC Upgrade TDR 3

Pb–Pb collisions a resolution s (dE/dx)/hdE/dxi of about 5.5 % and 7 % is achieved, respectively. Fur-
ther PID capabilities arise from topological reconstruction of the weak decays of strange hadrons and
gamma conversions.
The readout chambers are operated with an active bipolar Gating Grid (GG) which, in the presence of
a trigger, switches to transparent mode to allow the ionization electrons to pass into the amplification
region. After the maximum drift time of ⇠100 µs the GG wires are biased with an alternating voltage
DV = ±90 V that renders the grid opaque to electrons and ions. This protects the amplification region
against unwanted ionization from the drift region, and prevents back-drifting ions from the amplification
region to enter the drift volume. In particular, the latter would lead to significant space-charge accumu-
lation and drift-field distortions. Due to the low mobility of ions (µion = (10 3 ) · µelectron ), efficient ion
blocking requires the GG to remain closed for ⇠180 µs after the end of the event readout, corresponding
to the typical time it takes the ions in a Ne-based gas mixture to drift from the anode wires to the GG.
This gating scheme leads to an intrinsic dead time of the TPC system of ⇠280 µs, implying a principal
rate limitation of the present TPC to about 3.5 kHz. It should be noted that due to the present TPC readout
system the data rate is limited to ⇠300 Hz for central Pb–Pb collisions.
Operation of the TPC at 50 kHz cannot be accomplished with an active ion gating scheme. On the other
hand, back-drifting ions from the amplification region of a MWPC without gate will lead to excessive
ion charge densities and drift distortions that render precise space-point measurements impossible. The
proposed scheme therefore entails replacement of the existing MWPC-based readout chambers by a
multi-stage GEM system. GEMs have proven to operate reliably in high-rate applications and provide
intrinsic ion blocking capabilities, therefore enabling the TPC to operate in a continuous, ungated readout
mode at collision rates of 50 kHz. The TPC upgrade increases the readout rate by about two orders of
magnitude as compared to the present system, thus giving access to previously inaccessible physics
observables. As an example, the low-mass dielectron invariant mass spectrum is shown in Fig. 1.1,
accumulated in a typical yearly heavy-ion run (⇠ 3 nb 1 ) with the current (left) and upgraded (right)
TPC.
dN/dMeedy (GeV-1)

dN/dMeedy (GeV-1)

PbPb @ sNN = 5.5 TeV Sum PbPb @ sNN = 5.5 TeV Sum
Rapp in-medium SF Rapp in-medium SF
0 - 10%, 2.5E7 events Rapp QGP 0 - 10%, 2.5E9 events Rapp QGP
10-1 |ye| < 0.84 cocktail w/o ρ (± 10%) 10-1 |ye| < 0.84 cocktail w/o ρ (± 10%)
peT > 0.2 GeV/c cc → ee (± 20%) peT > 0.2 GeV/c cc → ee (± 20%)
0.0 < p < 3.0 2.5E7 'measured' 0.0 < p < 3.0 2.5E9 'measured'
t,ee Syst. err. bkg. (± 0.25%) t,ee Syst. err. bkg. (± 0.25%)
10-2 10-2

10-3 10-3

10-4 10-4

10-5 10-5

10-6 10-6
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
Mee (GeV/c2) Mee (GeV/c2)
p
Figure 1.1: Inclusive e+ e invariant mass spectrum for 0 – 10 % most central Pb–Pb collisions at sNN = 5.5 TeV, assuming
7 9
2.5 · 10 events (left panel) and 2.5 · 10 events (right panel). The spectra include a set of tight primary track cuts
based on the new ITS system to suppress leptons from charm decays. Also shown are curves that represent the
contributions from light hadrons (blue), charm (magenta) and thermal radiation from a hadronic gas (red) and a
QGP (orange). The figures are from [2].
4 The ALICE Collaboration

1.3 Design considerations

The present document describes a proposal for a new readout scheme of the ALICE TPC, based on
GEM technology. The main considerations for the design of the system and choices of technologies are
summarized in the following.

– The new GEM-based readout chambers must provide sufficient ion blocking to keep the resulting
drift field distortions in Pb–Pb collisions at a tolerable level. The distortions must be correctable
without deterioration of the online reconstruction efficiency and the final momentum resolution of
the detector.

– The new readout scheme requires a complete redesign of the TPC front-end and readout electronics
system. The new electronics must accommodate the negative signal polarity of the GEM detectors
and the continuous readout scheme. Additionally, the minimization of the ion space-charge density
requires the operation of the readout chambers at the lowest possible gas gain, leading to a front-
end noise requirement of ENC4 of around 670 e.

– The limited bandwidth of the data acquisition system requires significant online data reduction.
The present scheme foresees that cluster finding and association to tracks is performed in an online
computing system to achieve the required data compression factor of ⇠ 20.

– The upgraded TPC must preserve the performance of the existing system in terms of momentum
and dE/dx resolution. This requires that the space-charge distortions must be corrected to the level
of the intrinsic spatial track resolution of the TPC, i.e. to a few hundred µm. The present dE/dx
performance of the TPC requires a precise equilibration and normalization of the ionization energy
loss throughout the entire TPC volume. Moreover, the local energy resolution of the readout
chambers must not exceed s (55 Fe) = 12 % at 5.9 keV. The latter is affected by the transparency of
the GEM system to primary electrons and demands a careful optimization of the operational point
with respect to electron collection efficiency and ion blocking.

– The upgrade of the TPC readout chambers and electronics must allow the re-use of the existing
hardware to the greatest possible extent. This applies in particular to the existing field cage, gas
system, cooling, and services.

1.4 Detector overview

The requirements for the TPC upgrade listed above have led to the technical design presented in this
document. In the following section we summarize briefly the main aspects of the proposed solution.
The overall dimensions of the TPC will remain unchanged. Also the segmentation of the readout planes
into Inner and Outer Readout Chambers (IROCs and OROCs), 18 each on either endplate, will be pre-
served. This permits to re-use most of the components of the existing field cage and endplate structures.
The powering scheme of the field cage will be adapted to match the higher terminating voltages deter-
mined by the GEM system.
The upgraded TPC will be operated with a Ne-CO2 -N2 (90-10-5) gas mixture. This choice is mainly
driven by the higher ion mobility in neon as compared to argon, which leads to less space-charge accu-
mulation in the drift field.
The new TPC readout chambers will be equipped with quadruple GEM stacks for gas amplification.
Comprehensive R&D studies have shown that conventional triple GEM stacks using standard geometry
GEM foils will not lead to sufficient ion blocking. In prototype measurements with quadruple GEM
4 Equivalent Noise Charge (ENC)
TPC Upgrade TDR 5

systems, ion backflow fractions < 1 % have been reached at a gas gain of 2000 and an energy resolution
of s (55 Fe) = 12 % at 5.9 keV. These operational conditions will preserve the intrinsic dE/dx resolution
and keep the space-charge distortions at a tolerable level.
A new front-end electronics and readout system is being developed. The design of the new electronics is
driven by the requirement of low-noise operation and the challenges of continuous readout and high data
rate.
The upgraded TPC will provide similar momentum resolution as the present system. However, the
operation in continuous readout at high luminosity demands for innovative calibration and correction
schemes, in particular with respect to space-charge distortions. Moreover, significant data reconstruction
will have to be performed online to allow efficient data compression. With the design presented in this
document, these requirements can be fulfilled.
A summary of the TPC parameters is given in Table 1.1.
6 The ALICE Collaboration

Detector gas Ne-CO2 -N2 (90-10-5)

Gas volume 90 m3
Drift voltage 100 kV
Drift field 400 V/cm
Maximal drift length 250 cm
Electron drift velocity 2.58 cm/µs
Maximum electron drift time 97 µs
wt (B = 0.5 T) 0.32 p p
Electron diffusion coefficients DT = 209 µm/ cm, DL = 221 µm/ cm
Ne+ drift velocity 1.632 cm/ms
Maximum Ne+ drift time 153 ms
Readout chambers
Total number 2 ⇥ 2 ⇥ 18 = 72
Readout technology 4-GEM stack, single mask, standard (140 µm) and
large (280 µm) hole pitch
Gas gain 2000
Ion back flow <1%
Energy resolution at 5.9 keV 12 %
Inner (IROC)
Total number 2 ⇥ 18 = 36
Active range 848 < r < 1321 mm
Pad size 4 ⇥ 7.5 mm2 (rf ⇥ r)
Pad rows 63
Total pads (IROC) 5504
S:N 20:1
Outer (OROC)
Total number 2 ⇥ 18 = 36
Active range 1346 < r < 2461 mm
Pad size (inner) 6 ⇥ 10 mm2 (rf ⇥ r) (1346 < r < 2066 mm)
Pad rows (inner) 70
Total pads (inner) 6656
Pad size (outer) 6 ⇥ 15 mm2 (rf ⇥ r) (2086 < r < 2461 mm)
Pad rows (outer) 25
Total pads (outer) 3200
Total pads (OROC) 9856
S:N 30:1
Readout electronics
Number of channels 552,960
Signal polarity negative
Dynamic range 30 ⇥ MIP
System noise (mean) 670 e
PASA conversion gain 20 (30) mV/fC
PASA peaking time 160 (80) ns
ADC number of bits 10
ADC sampling rate 10 (20) MHz
Power consumption < 35 mW/ch
Operating conditions and data rate
Collision rate in Pb-Pb 50 kHz
Raw event size (Pb-Pb min bias) 20 MByte
Online data compression factor 20
Table 1.1: Synopsis of parameters of the upgraded TPC.
Chapter 2

Mechanical structure, field cage, and gas

system

2.1 Introduction
The upgraded TPC consists of the current structures of the detector, where the new readout chambers
and electronics will replace the existing ones. Thus, the field cage, the endplates, the insulating volumes,
the service support wheels, the I-bars, and the gas system will remain the same. These structures and
systems are described elsewhere [1]. In this chapter a brief overview of the mechanical structure of the
TPC is given, with emphasis on aspects relevant for the integration of the new readout chambers.

2.2 Field cage

The field cage is composed of four concentric cylinders which define the drift volume and the inner
and outer insulating volumes. These cylinders, the field cage vessels and the containment vessels, are
held together by the endplates. A schematic view of this assembly is shown in Fig. 2.1. They are made
of composite material, namely a Nomex honeycomb structure sandwiched between layers of prepreg
composite and Tedlar. The Tedlar foils provide gas tightness. Field-defining strips are employed to
degrade the potential from the central electrode, an aluminized mylar foil at the center, to nearly ground
potential close to the readout chambers.
The inner and outer voltage-degrading strips are suspended on 18 Macrolon rods each. These rods are
also used to circulate the gas through tiny holes along their length, such that the gas flows radially
outwards and does not exert any force on the central electrode. Some of these rods are also used to house
services such as optical elements for the laser system, the high voltage cable to the central electrode, and
the four removable and water-cooled resistor rods for the voltage degrading (see Fig. 2.2).

2.3 Endplates
The endplates keep together the field cage and containment vessels, and also hold the readout chambers.
These are arranged in 18 sectors, each one covering 20o in azimuth. The cutouts of the endplates define
the dimensions of the part of the chamber bodies which form the exterior surface of the endplates. The
gas tightness is ensured by one O-ring on the endplate, one on the chamber body and a sealing foil in
between. The chambers are mounted such that the whole body is inserted into the field cage and then
moved back into the endplate, where it is fixed from the inside. This allows to maximize the active
area of the detectors, but requires that the whole chamber is brought into the drift volume for insertion.
The full dimensions of the chambers bodies are determined by the geometry of the endplates, and in

7
8 The ALICE Collaboration

Figure 2.1: Schematic view of the ALICE TPC.

Figure 2.2: View of one of the endplates of the TPC; the different types of rods are indicated.
TPC Upgrade TDR 9

particular by the fixations on the endplate. A specialized mounting tool is used to install the chambers in
place. Special care has to be taken of the small clearances that occur during this operation, as shown in
the detailed views in Fig. 2.3 for the IROC and Figs. 2.4 and 2.5 for the OROC.

Figure 2.3: During the insertion of an IROC through the endplate the minimum clearance is 9 mm.

Figure 2.4: Insertion detail of an OROC. In the last stage, while moving the OROC back into its final position on the endplate,
the clearance between the chamber and the bracket that holds the rod is only 2 mm.
10 The ALICE Collaboration

Figure 2.5: Side view of the clearance between the OROC and the rod bracket.

2.4 Last resistor and skirt

The field cage has provisions for matching the drift field at the interface between the drift volume and the
readout chambers. On one hand, the resistor chain in the voltage dividers are terminated on the ground
side by a last, external resistor. The value of this resistor is tuned such that the drift field matches the
ground equipotential plane defined at the wire chambers. The potential of the gating grid is then fine-
tuned to minimise field distortions in this region, resulting in a potential of some -70 V. On the other
hand, there is a gap of a few cm between the outer side of the OROCs and the outer field cage, where
the ground defined by the endplates would produce sizeable field distortions. In order to avoid this, a
so-called skirt electrode is placed in this gap, as illustrated in Fig. 2.6.

Figure 2.6: Detail of the ground side of the outer field cage showing the skirt electrodes elevated from the endplate in order to
homogenize the electric field in the gap between OROCs and field cage.

In the case of the upgraded TPC, the potential on the GEM electrode facing the drift volume will amount
to 3 – 4 kV. This means that the skirt electrode will be set to a comparable potential, for which the
feedthrough connector will be replaced. For the tuning of the potential of the last strip via the last re-
sistor, though, an extra power supply to provide the necessary voltage while allowing to sink the current
across the voltage dividers will be necessary.

2.5 Service Support Wheel

The Service Support Wheels (SSWs) are installed in front of, and mechanically decoupled from, the
endplates. Their function is to hold the Front-End Cards (FECs) and their cooling panels and associated
services. Currently, each SSW supports 2800 kg of weight. The FECs are connected to the pad connec-
tors at the chambers via flexible cables. The low voltage to the FECs is supplied via bus bars installed in
TPC Upgrade TDR 11

the spokes of the SSW. The wheels themselves can be reused for the upgraded detector, but the frames
that hold the FECs may have to be replaced if new pad planes or FECs result in a different configuration
of the readout partitions. The 18 cutouts of the SSW are finally covered by flat cooling panels in order
to thermally isolate the endplates from the surrounding environment.

2.6 Gliders and I-bars

The TPC is inserted into and extracted from its position in the space-frame of the experiment by gliding
it on two rails with four pairs of teflon gliders installed on opposite sides, front and rear, on the endplate
and on the SSW. The relative position of these gliders is monitored at all times during gliding operations.
Finally, in order to keep the two field cage vessels well aligned, a set of so-called I-bars are used to push
or pull the set of inner cylinders against the outer set. A view of the TPC with SSW, I-bars and gliders
on the rails is shown in Fig. 2.7.

Figure 55: 2.7:

Figure Overall view
Overall ofofthe
view theTPC withwith
full TPC theSSW,
SSW,I-bars
railsand
andrails.
I-bars.

2.7 Gas system

The gas system is a closed-loop system where most of the gas is recirculated through the detector at
about 15 m3 /h with a regulated compressor module. The pressure in the TPC relative to atmospheric
is kept constant with better than 0.1 mbar precision. Oxygen and water are removed from the gas with
cartridges filled with a copper catalyzer. Some 50 l/h of mixed gas is continuously added to the system.
A small amount of gas is exhausted through an analysis line which provides a measurement of the O2
and H2 O contents, as well as the gas composition. Periodic samples are taken to analyze the gas with
a gas chromatograph. The excess gas is exhausted through a regulated flowmeter. A high-pressure gas
storage is used as a buffer to accumulate or release gas such that the system can absorb atmospheric
fluctuations without interruptions. The system, schematically shown in Fig. 2.8, is arranged in functional
12 The ALICE Collaboration

modules distributed on the surface, in the shaft, and in the cavern. A PLC1 controls the system and a
SCADA2 system provides a suitable user interface. The mixer unit is designed to mix up to three gases.
Default gases are Ne, CO2 , and N2 , but the mixer could inject Ar and CF4 in the desired proportions
by simply recalibrating the mass flow controllers for these gases. In the case of neon as noble gas, a
filling procedure where CO2 is trapped in molecular sieve cartridges minimizes wasting of neon during
this operation. The technique does not work for argon, which is however a low cost gas. In summary, no
significant hardware nor software modifications are needed in the gas system for any foreseen scenario
for the upgraded TPC.

Figure 2.8: Schematic diagram of the TPC gas system.

1 Programmable Logic Controller (PLC)

2 Supervisory Control and Data Acquisition (SCADA)
Chapter 3

Gas choice

The current TPC, which employs MWPCs as readout chambers, uses Ne-CO2 (90-10) as operating gas
mixture. The addition of 5% N2 has also been successfully tried out. The transport properties of these two
mixtures, shown in Fig. 3.1, are very similar, while, due to the higher quencher contents, the operational
stability of the chambers is improved in the case of the mixture containing N2 . At the high rates foreseen
for RUN 2, 10 kHz Pb–Pb, further stability to the wire chambers will be provided by replacing the Ne by
Ar. Detailed simulations have shown that, with Ar-CO2 (90-10), similar resolutions in momentum and
dE/dx can be achieved as in Ne mixtures. With a triggered gating grid, space charge distortions remain
at the level of 1 cm with this gas, at a slight reduction of the event rate to tape.

4.0 300 4.0 300

Vd (cm/µs)

Vd (cm/µs)
Diffusion coefficient (cm )

Diffusion coefficient (cm )

3.5
Ne-CO2 (90-10) 3.5 Ne-CO2-N2 (90-10-5)
3.0 275 3.0 275

2.5 2.5

2.0 longitudinal 250 2.0 longitudinal 250

transverse transverse
1/2

1/2
1.5 1.5

1.0 225 1.0 225

0.5 0.5

0.0 200 0.0 200

200 250 300 350 400 450 500 200 250 300 350 400 450 500
E (V/cm) E (V/cm)

Figure 3.1: Drift velocity and diffusion at moderate electric fields. (Left) Ne-CO2 (90-10). (Right) Ne-CO2 -N2 (90-10-5).

The space charge distortions expected in the upgrade scenario are an important criterium in the gas
choice for RUN 3. An example of the distortions calculated for a few candidate gas mixtures is shown in
Fig. 3.2. The mobility of Ar+ ions in Ar is 1.52 cm2 V 1 s 1 , about three times lower than that of Ne+
ions in Ne (4.08 cm2 V 1 s 1 ). We disregard here the effect of drifting CO+2 ions in these mixtures. Since
the ion backflow is similar for these two noble gases (see Sec. 5.1.3), the different mobilities result in
larger space-charge distortions in argon, even at a factor of 2 lower gain.
Admixtures of CO2 or CF4 to neon are attractive for the upgraded TPC. Both perform similarly in terms
of space-point distortions. Due to the differences in the wt factor, CO2 performs better in rj, but CF4
does better in r. CF4 provides a very high drift velocity and thus results in a reduced event pileup, which
is however not a critical issue for this TPC (see Sec. 7.3). On the other hand, the use of CF4 in the current
TPC would have to be thoroughly validated for compatibility with all materials of the detector and the
gas system, before it could be regarded as a suitable operating gas.

13
14 The ALICE Collaboration

dr (cm) 10

d(r ) (cm)
30 SC
~ r -1.5 for 50 kHz
Ne-CO 2-N 2 (90-10-5) ( = 20) 5
Ar-CO2 (90-10) ( = 10)
20 Ne-CF4 (90-10) ( = 20)
Ne-CF4 (80-20) ( = 20) 0

10
-5

0
-10

-10
-15

-20 -20
100 120 140 160 180 200 220 240 100 120 140 160 180 200 220 240
r (cm) r (cm)

Figure 3.2: Radial and azimuthal distortions for four gas mixtures as a function of r at z = 0, where the distortions are largest.
The parameter e is defined in Eq. (4.2).

In the anticipated configuration of our GEM system rather large transfer fields are used. However, at
fields around 4 kV/cm amplification starts in Ne-CO2 , as shown in Fig. 3.3, which reduces the ion back-
flow performance and also reduces the stability. An increased concentration of CO2 rapidly decreases
the drift velocity unless the field cage voltage is increased beyond its certified limits. The addition of N2
alleviates both issues as shown in the figure and explained in [1].
60
-c (cm )
-1

Ne-CO2 (90-10)
Ne-CO2-N2 (90-10-5)
50
Ar-CO2 (90-10)

40

30

20

10

0
0 4 8 12 16
E (kV/cm)

Figure 3.3: Effective Townsend coefficient for three gas mixtures as a function of the electric field strength. The onset of gain
is shifted by 1 kV/cm by the admixture of N2 to the neon mixture. For the argon mixture it is substantially higher.

The base gas mixture is therefore Ne-CO2 -N2 (90-10-5). The basic properties of the most interesting gas
mixtures is summarized in Tab. 3.1.
Gas Drift velocity Diffusion coeff. Eff. ionization Number of electrons per MIP
vd pDL pDT wt energy Wi Np (primary) Nt (total)
(cm/µs) ( cm) ( cm) (eV) (e/cm) (e/cm)
Ne-CO2 -N2 (90-10-5) 2.58 0.0221 0.0209 0.32 37.3 14.0 36.1
Ne-CO2 (90-10) 2.73 0.0231 0.0208 0.34 38.1 13.3 36.8
Ar-CO2 (90-10) 3.31 0.0262 0.0221 0.43 28.8 26.4 74.8
Ne-CF4 (80-20) 8.41 0.0131 0.0111 1.84 37.3 20.5 54.1

Table 3.1: Properties of a few gas mixtures used in modern TPCs. The diffusion coefficients are evaluated at 400 V/cm.
Chapter 4

Readout chambers

4.1 Introduction
An interaction rate of Rint = 50 kHz for minimum bias Pb-Pb interactions is expected after the luminosity
upgrade of the LHC during LS2 [1]. At these rates particle tracks from Npileup = 5 events on average will
be superimposed in the drift volume of the TPC at any given time: Npileup = Rint ⇥td ⇡ 50 kHz ⇥ 100 µs =
5, where td is the maximum electron drift time in the TPC1 . A continuous, untriggered readout of the
TPC is the obvious mode of operation in such a scenario with overlapping events, precluding the use of
a gating grid. In the absence of the gating grid, however, ions created in the multiplication region must
be prevented from drifting back into the drift volume by other means.
This will be achieved by using Gas Electron Multiplier (GEM) [2] foils as charge amplifier instead of
conventional MWPCs. The GEM consists of a 50 µm thin insulating Polyimide foil with Cu-coated sur-
faces, typically 2 – 5 µm thick. The foil is perforated by photo-lithographic processing, forming a dense,
regular pattern of (double-conical) holes. In the standard geometry the holes have an inner diameter of
⇠ 50 µm, an outer diameter of ⇠ 70 µm, and a pitch of 140 µm. Other geometries, e.g. with a larger pitch
and thus a smaller optical transparency, are considered for the ALICE TPC readout chambers. Figure 4.1
shows an electron microscope photograph of a standard GEM foil.
The small dimensions of the amplification structures lead to very large electric field strengths (50 kV/cm),
sufficient for avalanche creation, inside the holes of the GEM foil when a moderate voltage difference of
typically 200 – 400 V (depending on the gas) is applied between the metal layers.
The dynamics of charge movement and avalanche creation inside the GEM holes is complicated. Fig-
ure 4.2 shows a simulation performed with the Garfield / Magboltz [3] packages, illustrating the sup-
pression of ion backflow from the amplification region. In this simulation the avalanche is started by two
electrons, that are guided into the GEM hole by the drift field. The ions created in the avalanches closely
follow the electric field lines because of their much smaller diffusion. Most of the ions are collected on
the top side of the GEM foil, because the field inside the GEM hole is much higher than the field above
the hole. Only a few ions drift back into the drift volume. The extraction of avalanche electrons from the
hole proceeds more efficiently by applying a higher transfer field below the GEM.
The electrons can then be transferred to another amplification stage or collected at the anode. Typically 3
or 4 GEM foils are combined in a stack, leading to effective gains (see Eq. (4.1)) of the order of 103 – 104
and at the same time guaranteeing a stable operation without the occurrence of discharges [5].
The effective gain of a GEM detector is determined by measuring the current at the readout anode Ianode
for a given rate R of incident X-rays, each X-ray conversion producing Nion ionization electrons:
1A discussion of pileup and occupancies is given is Sec. 6.2.

15
16 The ALICE Collaboration

Figure 4.1: Electron microscope photograph of standard GEM foil with hole pitch 140 µm.

Figure 4.2: Garfield / Magboltz simulation of charge dynamics for electrons (two in this simulation) entering into a GEM
hole [4]. Electron drift paths are shown as light lines, ion drift paths as dark lines. Dots mark places where ionization
(multiplication) processes have occurred. The paths have been projected onto the cross section plane.
TPC Upgrade TDR 17

Ianode
Geff = . (4.1)
eNion R

Defined in this way, Eq. (4.1) corresponds to the gain seen by the readout, and takes into account charge
losses in the GEM structures.
We define the ion backflow as2 the ratio of cathode to anode current,

Icathode 1 + e
IB = = , (4.2)
Ianode Geff

with e being the number of ions drifting back into the drift region from the amplification region per
incoming electron. Note that IB also includes a contribution from ions created during the ionization
process. Ion backflow values of IB = 0.25 % have been reached experimentally in conditions which
were more advantageous than the ones in ALICE, e.g. a high magnetic field (4 T) and a low drift field
(200 V/cm) [6].
Detectors based on GEM amplification were pioneered by the COMPASS experiment at CERN [7–10],
and are now routinely used in several high-rate particle physics experiments like LHCb [11], PHENIX [12],
and TOTEM [13]. New applications include the use of GEM-based detectors in KLOE-2 [14] and
CMS [15].
The usage of GEM detectors as readout chambers in the ALICE TPC, however, is a new domain of
application of these detectors with regard to several aspects:

– The ion backflow from the detector must be carefully optimized. A value of 1 % is necessary in
order to achieve the goal of limiting drift field distortions due to space charge to well below 10 cm
in most of the drift volume. This value has been established in the default gas mixture Ne-CO2 -N2
(90-10-5). At a gain of 2000, which is needed for a signal-to-noise-ratio of 20, this corresponds to
e = 20, i.e. 20 ions flowing back into the drift region per incoming electron.

– Special attention must be given to a high electron collection efficiency of the GEM system, in
particular when the operational conditions are optimized for small ion backflow (see Sec. 5.1.3).
Finite electron transparancy, in particular in the first layer, leads to a degradation of the local
energy resolution due to a loss of primary electrons, that will eventually compromise the dE/dx
performance of the detector. Simulations have shown that a local energy resolution equivalent to
12 % at 5.9 keV is sufficient to preserve the present dE/dx resolution of the TPC (see Sec. 7.2.3).

– For a Ne-based gas mixture the dependence of the Townsend coefficient on the electric field is
steeper compared to the standard Ar-based mixtures, with which most GEM detectors have been
operated until now. In order to be able to apply the very asymmetric fields above and below the
GEM, required for maximum ion backflow suppression, without entering a regime of avalanche
multiplication in the gaps between GEM foils, a small admixture of N2 is added to the detector gas
(see Chap. 3).

– At a rate of primary ionization clusters of 100 kHz/cm2 for an IROC at a radius of r = 85 cm

(for dNch /dh = 500) the current density at the readout anode at a gain of 2000 is ⇠ 1 nC/(cm2 s),
assuming that all primary particles are MIPs. With a safety factor of 10, which takes into account
the contribution of highly ionizing particles, background, secondaries, etc., the upper limit for the
rate and the current density is 10 kHz/mm2 and 0.1 nC/(mm2 s), respectively. Both values are far
2 There appear different definitions of the term ion backflow in the literature. We choose this definition since it can be easily

measured.
18 The ALICE Collaboration

below experimentally verified limits of rate capability (> 100 kHz/mm2 ) for GEM detectors [9,
16]. Nevertheless, the long-term operation in ALICE requires careful testing of all materials used
in the detector construction concerning their aging properties.

In the following we describe the concept of the GEM-based readout chambers and show the design
choices that are consistent with the above considerations.

4.2 Mechanical structure

The readout chambers are installed on both end plates of the cylindrical gas vessels of the TPC. The two
readout planes are azimuthally segmented into 18 sectors of trapezoidal shape, each sector covering 20 ,
as shown in Fig. A.2. Each sector is further divided into two different chambers, called Inner Readout
Chambers (IROCs) and Outer Readout Chambers (OROCs). The dimensions of the ALICE readout
chambers are shown in Fig. 4.3. These dimensions, as well as the segmentation, are taken over from the
original ALICE TPC design and will remain unchanged after the upgrade.

Figure 4.3: Dimensions (mm) of the ALICE TPC readout chambers.

Figure 4.4 shows an exploded view of a GEM IROC. It consists of the following components:

– a trapezoidal aluminum frame (alubody),

TPC Upgrade TDR 19

– a support plate made of fiberglass (strong back),

– the pad plane (multilayer PCB) and

– the GEM stack including a cover electrode.

The corresponding view of a GEM OROC is depicted in Fig. 4.5.

We intend to rebuild the complete readout chambers including the alubodies. This allows (i) to fully
adapt the mechanics to the GEM amplification scheme, (ii) to rearrange the connector layout to the new
front-end electronics and to optimize the accessibility of the front-end cards, and (iii) to start production
and testing of the chambers well in advance of the end of LHC RUN 2.
The overall design of the alubodies will be very similar to the existing ones. The mechanical stability
of the chambers will be more than sufficient to prevent deformations due to gravitational forces and the
tension on the GEM foils. Minor modifications will be introduced for the feedthroughs of the high-
voltage (HV) supplies, the mounting of the boxes with the resistor chains for the HV, the cut-outs for
connectors of the front-end cards, and the mounting of the GEM planes.
The pad plane consists of a single multi-layer PCB for the IROCs, and, for production reasons, between
two and four PCBs for the OROCs. The final segmentation depends on the choice of number of readout
channels per FEC. The size of the readout pads of the present TPC matches the expected occupancy
during RUN 3 and will therefore be kept. Some reshuffling of pads and traces to connectors is needed
in order to accommodate the GEM stacks and the new front-end cards. For improved stiffness and gas
tightness, the pad plane with a thickness of 2 mm will be glued to a strong back plate of 3 mm thickness
made of fiber-glass reinforced epoxy. This design has proven to provide extremely good gas tightness3
and stability in the original version of the readout chambers.
The technique for insertion and mounting the chambers into the TPC will be taken over from the original
ALICE TPC (see Sec. 11). Special care has to be taken in order not to expose the GEM foils to dust
during the mounting procedure.

4.3 GEM planes

4.3.1 General structure
The electrons created by ionizing particles traversing the active volume of the TPC drift towards the end
plates, where they are amplified in order to induce a detectable signal in the readout pads. In the new
readout chambers, the amplification will be provided by avalanche creation inside the GEM holes. In
order to achieve the required gain and at the same time provide a sufficiently high suppression of back
flowing ions, a stack of four GEM foils will be used for all chambers. A schematic drawing of a GEM
stack is shown in Fig. 4.6.
The mechanical layout of the GEM stage follows a modular design wherever possible to allow pre-
assembly and testing of individual components at every stage. The introduction of dead zones is kept to
a minimum, so as not to deteriorate the resolution of the chambers.
The size of GEM foils needed for the ALICE IROC and OROC exceeds that of most of the currently
operating GEM detectors. The conventional method of GEM foil patterning requires photo-lithographic
processes based on two masks with identical hole patterns, placed on the two sides of the copper-coated
base foil, and aligned with a precision of 1 µm with respect to each other in order for the holes to be
perpendicular to the surface. These alignment requirements limit the maximum size for GEMs produced
3 The overall oxygen contamination in the current TPC at a flow of 15 m3 /h is of the order of 1 ppm.
20 The ALICE Collaboration

Figure 4.4: Exploded view of a GEM IROC.

Figure 4.5: Exploded view of a GEM OROC.

TPC Upgrade TDR 21

Cover electrode
Edrift
GEM 1
ET1 2 mm
GEM 2
ET2 2 mm
GEM 3
ET3 2 mm
GEM 4
Eind readout anode 2 mm
Pad plane
Strong back

Figure 4.6: Schematic exploded cross section of the GEM stack. Each GEM foil is glued onto a 2 mm thick support frame
defining the gap. The designations of the GEM foils and electric fields used in this TDR are also given. Edrift
corresponds to the drift field, ETi denote the transfer fields between GEM foils, and Eind the induction field between
the fourth GEM and the pad plane. The readout anode (see Eq. (4.2)) is indicated as well. The drift cathode is
defined by the drift electrode not shown on this schematic.

with this technique. Another important constraint is the size of the industrially available base material
and of the machinery required for the processing, both being presently limited to a width of 600 mm.
The first limitation can be bypassed by employing a single-mask technique [17]. This technique has
proven to deliver comparable results with respect to homogeneity and gain performance of the GEM-
foils as the standard technique. A small decrease in gain by 25 % has been observed in comparison with
a standard GEM at the same conditions, which can easily be compensated for by a slight increase of the
operating voltage.
Large-size foils with single-mask GEM technique have been pioneered in the framework of R&D for the
cylindrical GEM tracker of the KLOE-2 detector by the RD51 collaboration [18]. For the construction
of the full-size KLOE-2 tracker, which has been completed recently [19], a total of 50 large-size single-
mask foils with active areas of up to 430 ⇥ 700 mm2 have been produced at CERN. After thorough testing
with QA criteria similar to the ones to be adapted for ALICE (see Sec. 4.7), only eight bad foils were
identified. Most of the problems were related to an over-etching of the polyimide, a problem which,
according to the CERN workshop, has been resolved in the meantime. GEM foils with even larger active
areas (990 ⇥ (220 – 455) mm2 ) are now routinely being produced in the framework of developments for
the CMS muon system [15, 20]. At the time of writing this TDR, six full-size triple-GEM detectors
with single-mask GEM foils have been built by the CMS GEM collaboration. This collaboration also
measured the uniformity of the gain of a final detector to be within 12 – 15 % (RMS). The GEM foils
needed for the ALICE TPC4 are of a similar size. Hence the single-mask technique can be considered
mature for application to the ALICE TPC.
In order to reduce the total charge stored in the GEM foil, one side of the foil is segmented into HV
sectors with a surface area of approximately 100 cm2 , as shown in Figs. 4.8 and 4.12. The inter-sector
distances are reduced to 200 µm. Each HV sector is powered separately through high-ohmic SMD5
loading resistors soldered directly onto the foil and connected to a HV distribution line implemented on
the boundary of the foil. This scheme has proven to reduce the probability of discharges propagating
between GEM foils and from the last GEM foil to the readout circuit [5]. Figure 4.11 shows a detailed
view of the segmented side of an IROC GEM foil with the loading resistors in place and the frame of the
next GEM layer on top of it. The voltages to each GEM foil are supplied by two external HV sources,
one for each side of the foil (see Sec. 4.4).
The GEM foils will be pre-stretched with a force of 10 N/cm on all four sides using a stretching technique
developed at GSI and TUM, making use of a pneumatical method. A frame originally designed for the
stretching of stencils for PCB assembly was modified to meet the stretching force needed for GEM foils.
4 See Tab. 4.1
5 Surface Mount Device (SMD)
22 The ALICE Collaboration

Figure 4.7: Photograph of an IROC GEM foil in the stretching frame.

The system is shown in Fig. 4.7. The stretching force is applied by springs integrated in the frame. For
mounting, the spring tension is released by the pneumatic system. When the pressure is taken away, the
foils are tensioned.
The foils in the stretching frames are then positioned on a custom-made alignment plate for gluing of the
support frames made of glass-fiber reinforced plastics. The width of the frames is chosen to be 10 mm
to match the outer edges of the pad plane and for sufficient stability. The total thickness of the frames
is constrained by the distance between the layers in the stack of 2 mm. The frames also include a grid
of very thin bridges of ⇠ 400 µm thickness intersecting the active area of the detector, which serves as
spacer in order to guarantee sufficiently small sagging of the foils due to electrostatic forces. The spacer
grid is aligned with the HV sector boundaries of the GEM foils in order to minimize the dead area.
Four framed GEM foils are stacked on top of each other above the pad plane and fixed to it by non-
metallic screws. Our prototype studies have shown that this can be achieved without wrinkles even for
the trapezoidal shape of foils needed for ALICE (see Sec. 5.2.3). The top side of GEM 1 facing the drift
electrode of the TPC is covered by one extra frame which is metallized on the top side (see Fig. 4.6). The
potential applied to this cover electrode is adjusted to homogenize the electric field in the drift region
above the first GEM.

4.3.2 Inner readout chambers

The first batches of large-area GEM foils for the ALICE IROC prototype (see Sec. 5.2) with a size
of 500 ⇥ 470 mm2 were produced at CERN during the last year. Quality assurance tests have been
implemented in order to identify GEM defects both at macroscopic and microscopic levels as early
during the assembly stage as possible (seeSec. 4.7 for a detailed discussion of QA procedures).
Figure 4.8 shows the top view of the design of an IROC GEM foil, including the segmentation into 18
HV sectors, powered from both sides of the foil. Figure 4.9 shows a photograph of a prototype foil used
in the first beam tests (see Sec. 5.2).
TPC Upgrade TDR 23

Figure 4.8: Dimensions of an IROC GEM foil.

4.3.3 Outer readout chambers

The size of an OROC as given in Fig. 4.3 would require GEM foils with a size of more than 1000 mm
length and 900 mm width, which is beyond the limits of today’s technology. The obvious solution is
to segment the active area into three independent detector modules, tiled in radial direction. With this
scenario we end up with foil sizes of 900 ⇥ 400 mm2 or smaller. This is well in accordance with what
has already been achieved at the CERN production site in recent developments for ALICE prototypes
and other experiments [14, 15, 21]. Figure 4.12 shows the dimensions of the three modules for one
ALICE OROC, each having a width of ⇠ 360 mm. The division in separate smaller detector modules
is also expected to increase the yield of foils and facilitate their handling. For stability of the frames,
however, a minimum width of 10 mm is required. This will create two insensitive zones of 20 mm width
in radial direction in each OROC. Corresponding gaps are also foreseen in the pad layout, making room
for fixation holes for the frames. The design of the foils allows to distribute the power to all HV sectors
from two sides of the detector, without the need for HV lines between sectors. Thus, the gaps between
HV sectors will also be as small as 200 µm as for the IROC case. The number of HV sectors is 20, 24,
and 30 for the three detector modules.
Table 4.1 summarizes the parameters of the four different types of GEM foils needed for the upgraded
TPC:

Detector Size Active area No. of HV sectors No. of foils

(cm2 ) (cm2 )
IROC 54 ⇥ 54 1678.0 18 144
OROC 1 70 ⇥ 54 1997.3 20 144
OROC 2 77 ⇥ 54 2240.5 24 144
OROC 3 91 ⇥ 54 2949.0 30 144
Table 4.1: Parameters of GEM foils for the ALICE TPC.
24 The ALICE Collaboration

Figure 4.9: Photograph of an IROC GEM foil. The outer part is required for stretching and alignment and is cut away after the
frame has been glued onto the foil.

4.4 High voltage supply

4.4.1 System overview

Extensive studies have shown that by reducing the capacitance between the two metal surfaces of a GEM
foil the probability of discharges in a GEM detector can be significantly reduced [5]. To this end, one
side of the GEM foil is segmented into individually powered HV sectors with a surface of ⇠ 100 cm2 ,
limiting the amount of charge which is involved in case of occasional sparks.
The potential at the segmented side of the GEM foil is defined through large (MW) bias resistors. The
potential at the unsegmented side is supplied directly, i.e. without bias resistor. This scheme has the
advantage that in case of a temporary or permanent short circuit across a GEM foil in one or several HV
sectors a voltage drop over the bias resistor occurs only for the affected sectors, while the rest of the foil
remains fully operational without change of potential.
In a standard GEM detector, e.g. as used for the COMPASS experiment [7], the GEM foils are oriented
such that the segmented sides face the drift electrode. In case of a discharge between the two sides of a
GEM foil a voltage drop then occurs only on the segmented side, whereas the unsegmented side remains
at its nominal potential. This prevents the propagation of the discharge to the next GEM foil or to the
readout circuit.
In the case of the ALICE TPC, however, this scheme cannot be applied to the first GEM foil in the stack,
because its potential on the side facing the drift electrode defines the drift field, which has to remain
constant under all circumstances. Therefore, GEM 1 will be mounted with the unsegmented side facing
the drift electrode, as shown in Fig. 4.13. In this way, its potential will be defined by the voltage UGEM
applied to the GEM stack. Considering the moderate voltage drop across GEM 1, this scheme is expected
to be safe in terms of discharge propagation despite the inverted orientation of the foil.
The potentials at the GEM electrodes are defined by an external voltage divider network, which is housed
in a box mounted to the backside of the alubodies. Therefore, nine HV-feedthroughs are needed for the
TPC Upgrade TDR 25

Figure 4.10: Detailed view of the HV distribution on the segmented side of an IROC GEM foil.

Figure 4.11: Bias resistors and HV supply of an IROC GEM foil.

26 The ALICE Collaboration

Figure 4.12: Dimensions (mm) of OROC GEM foils. The active area of the OROC is divided into three individual GEM stacks
of different sizes.

Figure 4.13: Schematics of the HV distribution scheme for a GEM detector module. The resistor values are given in units of
MW and correspond to the field settings of Tab. 4.2.
TPC Upgrade TDR 27

Figure 4.14: Potentials at all GEM electrodes of a quadruple GEM stack in case of a temporary short circuit between the two
electrodes of GEM 1 (top panel) and GEM 4 (bottom panel), occurring at time 0.1 s and lasting until 0.5 s.

four GEM foils and the cover electrode of each detector module. The values of the resistors in the
network are chosen such that an effective gain of 2000 and a minimum ion backflow of around 0.5 – 1 %
is provided. The voltage necessary to achieve a given gain will vary between modules due to variations
of GEM foil parameters like e.g. hole diameters. In order to have a homogeneous drift field in the drift
volume of the TPC, the potential at the top electrode of GEM 1 has to be the same for all modules. Thus,
the potential difference across GEM 4 will be adjusted with the use of a remote controlled regulation
circuit in parallel to the resistor defining the coarse potential drop.
Figure 4.14 (top panel) shows the absolute potentials at all four GEM foils of a stack during an event
where a temporary short circuit (due to a discharge) occurs between the two electrodes of GEM 1 (upper
panel). Due to the bias resistors at the bottom side in case of GEM 1, the potential at the bottom side
increases to the value of the top side (blue curve). The potential at the top side of GEM 1 does not
change, as required for a constant drift field. Figure 4.14 (bottom panel) displays the analogous situation
for GEM 4, which has the bias resistors on the top side. Here, the potential at the bottom side remains
approximately constant in order to minimize the probability of the discharge propagating to the readout
pads. The short circuits result in an increase of the overall current through the resistive divider of 1.4 µA
for the case of GEM 1, and of 25.4 µA for the case of GEM 4.
For online monitoring of space-charge effects it is essential to measure the currents on all GEM sectors
with a high frequency of the order of ms and a precision of ⇠ 100 pA. The current measurement will be
implemented in the HV divider box, and the data will be included in the TPC data stream (see Sec. 10.3).
28 The ALICE Collaboration

In this way, the data are readily available for online calibration. We also intend to include in the HV
distribution boxes the possibility to send a pulse to the lowest GEM electrode, which then induces sig-
nals on all electronic channels of a given module for calibration purposes. This will be accomplished
by a coupling capacitor, which will however be disconnected during normal operation for reasons of
operational safety. The voltage divider boxes have to be accessible from the outside in order to allow for
access during a shutdown period of the experiment.
The high voltage system for the GEM stacks requires good voltage stability (ripple and noise <50 mV),
high precision current measurement (resolution 1 nA), a fast trip mechanism, adjustable ramp speeds,
full remote controllability and output voltages up to 6 kV at a maximum current of 1 A. While a single
HV channel is required for each IROC, the three OROC modules will be powered separately but syn-
chronously for reasons of flexibility and stability. In total, therefore, 144 HV channels are needed for the
GEM stacks, plus the same number for the cover electrodes.

4.4.2 Typical HV settings

In order to minimize the backflow of ions produced in the amplification region, the electric field con-
figuration of the GEM stack as well as the sharing of the gain among the four amplification stages is
optimized, thus bringing the ALICE TPC to a novel mode of operation of a GEM detector. Our measure-
ments on small test detectors, reported in Sec. 5.1.3, show that an ion backflow of less than 1 % can be
achieved with a quadruple GEM system at typical voltage settings shown in Tab. 4.2. At an effective gain
of 2 · 103 in the GEM stack this corresponds to e < 20 back-drifting ions per electron reaching the GEM
stack. In order to achieve this result, a configuration was used, where the second and the third GEM foils
have a pitch of 280 µm in contrast to the standard 70 µm. Such large-pitch foils have a smaller optical
transparency and therefore block ions more effectively. Strategies for an even further reduction of the
ion backflow will be discussed in Sec. 9.

Drift Field = 0.4 kV/cm

Potential at top of GEM 1 = 3150 V
DUGEM1 = U1top U1bot = 270 V
Transfer Field 1 (ET1 ) = (U1bot U2top )/0.2 cm = 4.0 kV/cm
Potential at top of GEM 2 = 2080 V
DUGEM2 = U2top U2bot = 250 V
Transfer Field 2 (ET2 ) = (U2bot U3top )/0.2 cm = 2.0 kV/cm
Potential at top of GEM 3 = 1430 V
DUGEM3 = U3top U3bot = 270 V
Transfer Field 3 (ET3 ) = (U3bot U4top )/0.2 cm = 0.1 kV/cm
Potential at top of GEM 4 = 1140 V
DUGEM4 = U4top U4bot = 340 V
Collection/Induction Field (Eind ) = U4bot /0.2 cm = 4.0 kV/cm

Table 4.2: Typical high voltage settings for IB < 1 % in a quadruple GEM in Ne-CO2 -N2 (90-10-5) at an effective gain of 2·103 .
Note the high transfer field in the 1st and 2nd gap, whereas ET3 is very low. The potentials at the top of the 4 GEM
foils are given as well.

4.5 Readout pad plane

The electrons emerging from the last GEM stage induce a fast signal on the readout pads. A very precise
determination of the position in two-dimensional space of an arriving electron cluster can be achieved if
the signal is distributed over several adjacent pads. In this case a suitable weighting of these signals will
yield a resolution much better than the typical pad size. Charge sharing is achieved by the combination
TPC Upgrade TDR 29

(cm)
GEM 4x7.5 85<r<132 cm GEM 6x10 135<r<198 cm GEM 6x15 198<r<247 cm

r
0.12 MWPC 4x7.5 MWPC 6x10 MWPC 6x15

0.1

0.08

0.06

0.04

0.02

0
0 50 100 150 200 250 50 100 150 200 250 50 100 150 200 250
z(cm) z(cm) z(cm)

Figure 4.15: Point resolution in the magnetic bending plane rj as a function of z for (blue) GEM and (red) MWPC amplifica-
tion, with (left panel) 4 ⇥ 7.5 mm2 , (middle panel) 6 ⇥ 10 mm2 , and (right panel) 6 ⇥ 15 mm2 pads, respectively.

of several effects: spread of the drifting charge clouds due to diffusion and track inclination, intrinsic
response width of the amplification stage due to diffusion in the GEM stack, finite hole distance, and
signal induction in the induction gap. The fraction of the total charge in the avalanche which is induced
on a given pad as a function of the distance between the center of the pad and the arrival point of the
electron initiating the avalanche is described by the Pad Response Function (PRF). Amplification in a
multi-GEM detector results in a narrower PRF compared to that of an MWPC, approximately equal to
the width of the charge cloud emerging from the last GEM.
In a microscopic simulation of the TPC it was studied whether diffusion in the drift gas is sufficient
to provide adequate charge sharing despite the smaller PRF, when GEM detectors are combined with
the present readout pad geometry. For the following studies, single tracks with minimum energy loss,
a flat distribution in polar angle q and zero azimuthal inclination (i.e. very high momentum) have been
used. It should be noted that a finite azimuthal inclination of tracks gives a dominant contribution to the
point resolution, so that the differences between MWPC and GEM amplification shown in the following
would be washed out. In this sense, the simulation results correspond to a worst-case scenario. A purely
projective PRF was assumed for the GEM case, i.e. no broadening of the charge cloud in the GEM
stack was taken into account. Exponential gain fluctuations for single electrons in the GEM stack were
included. The average gain was adjusted such that a signal-to-noise ratio of 20 for the most probable
value was reached. A Gaussian shaping with s = 100 ns was applied and the resulting signal was sampled
at 10 MHz.
Figure 4.15 shows the position resolution in the azimuthal rj coordinate as a function of the z-coordinate
of the cluster (the drift distance increases towards smaller values of z) for the three different pad sizes
used in the current pad planes (see Tab. 4.3). For drift distances larger than about 30 cm for the small
pads and 80 cm for the larger pads, the resolution achieved with GEM readout is only about 10 % worse
than for the MWPC-based readout, despite the large pad size compared to the GEM PRF. In this region,
diffusion provides sufficient broadening of the charge cloud, such that clusters with more than a single
pad dominate, as shown in Fig. 4.16.
For shorter drift distances, however, the resolution gradually deteriorates due to the appearance of single-
pad clusters. As expected, the effect is strongest for the OROCs with large pads and z > 180 cm. It should
be noted, however, that short drift distances correspond partially to regions outside the nominal accep-
tance of the ALICE central barrel (|h| < 0.9). This implies that, even with the present pad granularity,
the position resolution with GEM detectors will decrease by no more than 20 % for tracks in |h| < 0.75.
It is shown in Chap. 7 that this does not affect the momentum resolution, especially in the high pileup
scenario.
30 The ALICE Collaboration

Fraction 1
GEM 4x7.5 GEM 6x10 GEM 6x15
1-pad
0.8 2-pad
3-pad
4-pad
0.6

0.4

0.2
Fraction

MWPC 4x7.5 MWPC 6x10 MWPC 6x15

0.8

0.6

0.4

0.2

0
0 25 50 75 100 125 150 175 200 225 25 50 75 100 125 150 175 200 225 25 50 75 100 125 150 175 200 225 250
z(cm) z(cm) z(cm)

Figure 4.16: Fraction of one-, two-, three- and four-pad clusters for a (top row) GEM-based readout and (bottom row) an
MWPC-based readout, with (left) 4 ⇥ 7.5 mm2 , (middle) 6 ⇥ 10 mm2 , and (right) 6 ⇥ 15 mm2 pads, respectively.

The effect of the different PRF for GEMs and MWPCs is also visible in Fig. 4.17, where the average size
of clusters for the different pad sizes as a function of radius r is presented. Since clustering is performed
in pad rows, i.e. a cluster extends over several pads within a given pad row and over several time bins,
the cluster size is given in units of its area in rj-z coordinate space, i.e.

Acl = Nbins ⇥ (pad width) ⇥ (time bin width) . (4.3)

Here, Nbins is the number of bins above threshold in pad and time units, pad width is 0.4 or 0.6 cm,
and time bin width is given by the drift velocity divided by the sampling frequency. For all radii, the
GEM readout gives an approximately 20 % smaller cluster size than the MWPC readout, mainly in rj-
direction, due to the smaller PRF. Within a given pad size the cluster size increases with decreasing r
because of the increasing average polar inclination of tracks, which extends the clusters in time direction.
The steps in cluster size at r ⇡ 140 cm and 200 cm appear because of the increasing pad size, which,
combined with the polar inclination of tracks, mostly affects the extension of a cluster in time.
On the one hand, the smaller cluster size with GEM readout reduces the occupancy of the detector.
However, the larger number of single-pad clusters will make it more difficult to disentangle overlapping
clusters. The number of single-pad clusters can be reduced using Chevron-shaped pads, which could
restore the spatial resolution of the MWPC-based readout, at the expense, however, of a larger occupancy.
It should be noted that occupancy prohibits a larger pad size in radial direction, at least for the IROCs,
which would allow to operate the TPC at a lower gain while maintaining a given signal-to-noise ratio.
This in turn would decrease the distortions due to ion backflow, as can be seen from Eq. (4.2). The use of
a low diffusion gas, like the CF4 option discussed in Chap. 3 and Sec. 9.2, would reduce the occupancy
and therefore allow larger pad sizes in radial direction, or Chevron-shaped pads.
TPC Upgrade TDR 31

cluster size (cm2)

MWPC 4x7.5
3.5
MWPC 6x10
MWPC 6x15
3

2.5

1.5
GEM 4x7.5
1 GEM 6x10
GEM 6x15
0.5

80 100 120 140 160 180 200 220 240

r(cm)
Figure 4.17: Cluster size as defined in Eq. (4.3), as a function of radius r and the corresponding pad sizes, for (blue) GEM and
(red) MWPC readout.

Pad size (mm2 ) Number of rows Number of pads

IROC (841 < r < 1321 mm) 4 ⇥ 7.5 63 5504
OROC (1346 < r < 1696 mm) 6 ⇥ 10 35 2944
OROC (1716 < r < 2066 mm) 6 ⇥ 10 35 3712
OROC (2086 < r < 2461 mm) 6 ⇥ 15 25 3200
TPC total (2 ⇥ 18 sectors) 158 552,960
Table 4.3: Dimensions and parameters of readout planes and pads.

In conclusion, the current choice of pad sizes, optimized for occupancy and resolution with an MWPC-
based readout, provides sufficient space-point resolution even for the case of GEM amplification without
decreasing the pad size. The good momentum resolution in the acceptance of the barrel detectors of
ALICE will be maintained after the upgrade, as shown in Chap. 7. It should be noted that the smaller
PRF for GEM detectors reduces the occupancy of the detector. Thus, it allows to operate the TPC with a
Ne-CO2 -N2 (90-10-5) gas mixture also at the luminosity foreseen after the upgrade of the LHC.
Figure 4.18 shows the design of the pad plane for the OROC, where the division of the GEM detectors
in three separate modules has been taken into account. There are two dead areas under the frames of the
foils. Moreover, the transition between medium and large pads has been moved up by 8 cm with respect
to the current pad plane. This results in 1 pad row less in the new pad plane, i.e. 1.3 % less pads as
compared to the current system.
Table 4.3 summarizes the pad sizes for the different regions of the readout chambers. In total, the pad
re-arrangement in the OROCs results in 0.8 % less pads in the TPC as compared to the currect system.

4.6 Interface to front-end electronics

The front-end electronics is described in Chap. 6. The signals induced on the readout pads are routed
to connectors on the backside of the pad plane. As in the present system, short flexible flat cables will
be used to decouple the weight of the FEE6 , which is supported by the service support wheel, from the
6 Front-End Electronics (FEE)
32 The ALICE Collaboration

Figure 4.18: Pad layout of an ALICE OROC. The gaps have been introduced to allow fixation of the GEM detector modules.
Units are in mm.
TPC Upgrade TDR 33

readout chambers and the field cage. The signal routing must be optimized in order to minimize the trace
length. The exact configuration of the connectors, flat cables and cards is still a matter of optimization
and has thus not been decided upon yet.

4.7 Quality assurance of GEM foils

The proper selection of the GEM foils to be assembled into the readout chambers is of major importance.
COMPASS was the first large-scale experiment to use GEM-based detectors. Quality assurance pro-
cedures for all detector components have been established in order to guarantee a uniform and stable
operation of the COMPASS readout chambers [7]. These procedures have successfully been applied and
extended to other projects like LHCb [22], PHENIX [12], TOTEM [13] and the SuperFRS [23, 24], and
will be adopted to a large extent for the ALICE TPC.
In this section we will go through the transportation requirements and the characterization process, in
particular optical and electrical uniformity checks, that are essential in order to minimize the risk of
initial failure and in order to ensure the long-term stable operation of the modules. A methodology has
been developed for some of the characterization steps. These will be treated separately in the following.

4.7.1 Electrical characterization

The first electrical quality assurance test consists of measuring the leakage current in each GEM HV sec-
tor in response to an applied voltage across the foil. The test is performed in a dry nitrogen environment
to avoid instabilities related to water or dust. In order to measure the leakage current, a setup as shown
in Fig. 4.19 is used.
The components of this setup are: a desiccator with a gas flowmeter, an electrometer, a high voltage
source and a computer to control the ramping of voltages and to read and store the current measurements.
A voltage is applied between the unsegmented side and one of the HV sectors on the other side of the foil,
with all the other sectors grounded. The voltage is ramped in steps up to a final potential difference of
550 V between the two sides of the GEM foil, which is slightly below the breakdown limit of GEM foils
in N2 gas. Once the GEM is fully charged, the current drops and saturates at the leakage current through
the GEM HV sector under study, which should normally be . 0.1 nA. Foils with leakage currents larger
than ⇠ 0.5 nA are rejected and returned for re-cleaning. The foil is kept at maximum voltage for up to
30 min. If the foil is stable during this time with a leakage current smaller than 0.5 nA, it is accepted and
given to the next characterization setup (optical scanning).
A typical leakage current evolution during the electrical characterization is shown in Fig. 4.20. Fig-
ure 4.21 shows a second example of a leakage current test, where a HV sector experienced several
discharges, but could be fully recovered after flushing with N2 . In cases where continuous discharges are
observed, the foil is rejected and returned for re-cleaning.

4.7.2 Optical scanning characterization

After a foil has passed the coarse visual inspection and the static electrical quality assurance testing, it
is subjected to precision optical inspection using a high resolution camera. In this way the geometrical
characteristics of the GEM foil can be studied. In particular, the distributions of inner and outer hole
diameters and the hole pitch for both surfaces of the foil are measured. It is well known that a variation
of the inner hole diameter causes a variation of the intrinsic gain of the foil. The dispersion of the
distributions can be taken as a quantitative indicator of the general foil uniformity. Foils not meeting
the specifications over the full surface (i.e. copper hole diameter typically 70 µm ± 5 µm, polyimide hole
diameter typically 55 µm ± 5µm) are rejected.
In addition, the method is suited to identify smaller defects which were not detected in the first visual
34 The ALICE Collaboration

Figure 4.19: Leakage current measurement setup.

inspection step. Defects in the form of under- or over-etched areas in the foils can occur during the
manufacturing process. Other defects may come in the form of chemical residues from the production
process, droplets of glue or dust attached to the foil, very large holes, missing holes, etc. All of them
may cause operational instabilities; such foils are rejected and possibly re-cleaned.
The setup used for this purpose is based on a back-illuminated light table with area 100 ⇥ 100 cm2 . It
accommodates a full foil used for an IROC or OROC. In Fig. 4.22 the scanner used for this optical
characterization is shown [25]. A 9 Megapixel camera with a single pixel size of 1.75 µm2 is mounted on
an x-y positioning system above the light table. The optical system has a resolution of 144 light points
per mm and a field of view (single image) of 6.1 ⇥ 4.6 mm2 . After compression, the total image size is
about 20 MByte. Up to 32,000 individual images are required to cover the full active area.
During the scanning procedure the diameters of inner and outer holes, the pitch between holes and their
shape are recorded. Distributions of the diameter, the width of the polyimide rim (the distance between
the border of the outer and inner holes), and the pitch are shown in Fig. 4.23. Parameters describing
the shape of the hole are obtained from an ellipse interpolation with a sigma of 0.52 µm. A suitable
criteruim to determine e.g. the maximum allowable rim width (see tail in middle panel of Fig. 4.23) will
be established.
A two-dimensional map of the GEM foil characteristics is used to visualize the uniformity of measured
parameters as a function of position on the foil. Figure 4.24 shows an example of the spatial variation
TPC Upgrade TDR 35

Figure 4.20: Leakage current evolution during electrical characterization of an accepted GEM foil. The spikes at early times
correspond to the five voltage steps scanned through during ramping up at the beginning of the test. After the
target voltage has been reached, the current quickly drops below 1 nA. It is then measured for as long as 30 min.

Figure 4.21: Leakage current results from a recovered GEM foil.

36 The ALICE Collaboration

Figure 4.22: Setup of the high resolution scanning system.

Figure 4.23: Distributions of inner and outer hole diameters (left panel), the rim width (middle panel), and the pitch between
holes (right panel).
TPC Upgrade TDR 37

Figure 4.24: Map of hole diameters of a 10 ⇥ 10 cm2 GEM foil. (Left) inner hole diameter, (right) outer hole diameter.

of the diameters of the inner and the outer holes. Here, the diameters of holes are averaged over an area
of 1 mm2 . Maps such as these can be used during assembly of the GEM stack to avoid accumulation of
unwanted features in similar positions over the stack area.

4.7.3 GEM gain mapping

To evaluate the uniformity of the gain distribution, a mapping of the gain of each GEM foil is performed.
An acceptance criterion can be specified for the gain variation across the surface of the foil, e.g. 10 %
(RMS). In order to determine the gain map of a GEM foil, a setup as shown schematically in Fig. 4.25

Figure 4.25: Schematics of the setup for the gain mapping measurement.

is used. It includes the GEM foil under test and a reference detector, typically a multiwire proportional
chamber or alternatively a Micromegas detector. The reference detector detects the primary ionization
that takes place either above or below the GEM foil. In addition, a pad readout geometry is used to
record a 2D map of the foil under test. The source on top of the detector is used to illuminate the entire
GEM foil, making it possible to measure the gain and its evolution in time without having to scan over
the detector surface. A prototype of this setup, shown in Fig. 4.26, has been built at Yale University.
By illuminating the whole active area with a source, a double spectrum is measured for each pad, as
shown in Fig. 4.27. The low-amplitude spectrum corresponds to X-rays absorbed below the GEM,
whereas the high-amplitude one corresponds to X-rays absorbed above the foil. The ratio between the
position of the two main peaks is therefore proportional to the gain of the GEM foil for this particular
pad area. The uniformity of the gain is then obtained from the variations of the gain of each pad with
38 The ALICE Collaboration

Figure 4.26: (Left) Gain map setup. (Right) Flange together with the multiwire proportional chamber and the GEM foil under
test.

Figure 4.27: Pulse height distribution for a single pad.

respect to the mean gain value. For the particular example shown in Fig. 4.28, the pad dimensions were
2 ⇥ 2 cm2 , and the total area of the GEM foil was 10 ⇥ 10 cm2 , resulting in the 5 ⇥ 5 bin gain map shown.
For the GEM foil test illustrated in this figure, the mean gain obtained was 23 with a dispersion of 4.8 %
(RMS). At the same time this setup also permits mapping of the energy resolution of a single GEM foil,
and thus a measurement of the uniformity of the expected energy resolution.

4.8 Quality assurance of chambers

Once the GEM stacks are mounted on the IROCs or OROCs, the completed chambers need to pass
further tests. They are installed in test boxes which provide a gas volume and a small field cage with a
drift electrode and a voltage divider. The box is flushed with the nominal gas at a reasonable flow.
The first test is the gas tightness test. The pad plane structure may exhibit leakage if not properly as-
sembled. For testing this, the oxygen contamination in the gas outlet of the test box is measured with
an appropriate oxygen sensor. In order to provide final oxygen levels compared to the current ones in
the TPC, levels of 1 2 ppm in the test box at flows of 20 l/h should be achievable. However, viscous
leaks, e.g. through tortuous paths between the pad plane and the alubody, can be difficult to detect by
this method, unless the gas pressure in the test box is kept lower, by a few mbar, than the ambient pres-
sure. Therefore, the gas system utilized for these tests will contain provisions for producing a controlled
underpressure in the test box.
TPC Upgrade TDR 39

Figure 4.28: Relative gain distribution (top panel) and map (bottom panel) of a single GEM foil.

The next test is to produce the gain curve as a function of the applied voltages in order to characterize
each chamber. All chambers should deliver similar gains at similar voltages, and should hold at least a
factor of 2 higher gain than the anticipated operational gain.
The measurement of the uniformity of the gain across the active area follows next, with a coarser gran-
ularity than in the previous QA of the foils. This test assures that the stack is properly mounted and
clean.
Finally, a long-term test is carried out. The chamber is operated at a gain higher than the nominal
value for one or two days, while being irradiated with e.g. an Fe55 source. During this test the relevant
parameters are periodically recorded: currents, main peak position for the Fe55 spectrum, FWHM of
the main peak, ambient conditions, oxygen and water concentrations. The chamber must hold this test
without any trips or current excursions.
Once a chamber passed all these tests (timescale ⇠ 1 week per chamber) it is validated and properly
stored in a sealed environment under controlled temperature and pressure until its installation in the TPC
field cage. About 80 chambers will be built and tested this way in several construction and testing sites.
A convenient split is two test sites for all IROCs and two for all OROCs.
40 The ALICE Collaboration
Chapter 5

R&D with prototypes

In this chapter the R&D carried out with small and large prototype chambers is described. The obtained
ion backflow performance and the comparison with detailed simulation studies is presented in Sec. 5.1.
The construction and in-beam operation of a full-size prototype of an IROC, assembled with three GEM
foils, is described in Sec. 5.2.

5.1 R&D with small prototypes

In this section the question of the minimization of the ion backflow of a GEM structure is addressed.
To this end, scans of field configurations are performed, and various GEM structures are investigated,
in particular triple and quadruple stacks assembled with standard and double-pitch foils. The resulting
performances are compared with simulations. The energy resolution and the stability of the emerging
solutions are also investigated or in the process of being demonstrated. Although various gas mixtures
have been tried out, we focus here on results obtained with the baseline mixture Ne-CO2 -N2 (90-10-
5). The detectors are irradiated with sufficiently low X-ray rates in order to avoid building up positive
space-charge near the foils, since as it was found out this severely affects the measured ion backflow.

5.1.1 Experimental setup

Several small detector systems were built and set up in various labs within the scope of this R&D effort.

Setup for gain stability measurements

Measurements are performed to monitor the gain stability of a GEM stack over time, including the
response to changes in high voltage and radiation levels. For this purpose, a GEM chamber under test is
continuously irradiated while its current is periodically measured. The ambient temperature and pressure
are also recorded, together with the water content of the gas, which plays an important role in charging-
up processes of GEM foils. A single-wire proportional counter is operated in the same gas stream and
its gain monitored via the position of the 55 Fe line for correction purposes. Both the GEM detector and
the wire counter are installed in a box which may be flushed with N2 for further moisture reduction.

CERN ion backflow setup

The detector housing used for ion backflow measurement at CERN comprises a GEM holder and a field
cage. The GEM box is equipped with HV feedthroughs for up to four 10 ⇥ 10 cm2 foils and a pad plane
subdivided into 16 ⇥ 16 pads. All pads are connected together to a DC voltmeter for measurement of the
current. The field cage, constructed from Rohacell to permit irradiation with X-rays, is 8 cm deep and
has a drift electrode and 10 strips made of copper. The drift electrode is powered with one channel of a

41
42 The ALICE Collaboration

55Fe A
source
A
iseg hv
Drift foil/ A module
Cathode
A and
25 mm
A
Mpod
A controller
GEM 1
1.8 mm
GEM 2 A
1.8 mm
GEM 3 A
1.8 mm
GEM4
1.4 mm A
Readout foil/ 1 pad
Anode 1 pad

LabVIEW
A
a) b)

10 M
(signal input)
ADC
Windows LIN
FAN (test input) c)
IN / A m p l i f i er P r eam p l i f i er
Oscilloscope OUT
Pulse
Generator
Windows
Counter Level Discriminator
Converter

Figure 5.1: Sketch of the Munich quadruple GEM setup.

HV power supply and its current is measured with a floating picoammeter with 100 pA resolution. The
field cage is powered with a second HV channel so that the current through its 10 ⇥ 10 MW resistors does
not add to the small current on the drift electrode. A last, external resistor ensures reasonable matching
of potentials of the last strip and the top GEM electrode. The gas is supplied from a premixed bottle
for the reference measurements with Ar-CO2 (70-30) and from a mixer for the other gas mixtures. The
HV is supplied with programmable power supplies for convenient scanning of gains and fields. A small,
silver-anode X-ray tube is used as source of radiation.

Munich ion backflow setup

The Munich detector has been specifically built to measure ion backflow. The geometrical configuration
is summarized in Tab. 5.1. It features a short drift length, individual control of each HV electrode, and
the possibility to irradiate the detector from the top and a side wall. Standard 10 ⇥ 10 cm2 GEM stacks
are used.

Drift gap 25 mm
Transfer gap 1 1.8 mm
Transfer gap 2 1.8 mm
Transfer gap 3 1.8 mm
Induction gap 1.5 mm
Thickness top window 150 µm
Thickness side window 50 µm
Table 5.1: Properties of the Munich test detector in the quadruple GEM configuration.

The readout consists of 512 strips, that can be either read out individually or connected via analog sum-
mation cards to a total of 16 channels. By setting the appropriate jumper one can select the corresponding
size of the readout area, while all other strips remain grounded.
TPC Upgrade TDR 43

Figure 5.2: Signal amplitude in the single wire counter (left panel) and current in the triple GEM (right panel) as a function of
time with a Ne-CO2 (90-10) gas mixture and 180 ppm of water.

Figure 5.3: (Left) GEM current corrected by the gain of the single wire counter as a function of time, during the few hours
following an increase in the GEM gain. (Right) Projected distribution fitted with a Gaussian. The relative variation
is 0.45 %.

The picoammeters used to measure all currents contain a wireless transmission unit which allows them
to be kept at a floating potential. They are able to measure currents down to 0.1 nA at up to ⇠5 kV.
These picoammeters are connected to each electrode as shown in Fig. 5.1. All currents are measured
simultaneously and therefore should sum up to zero. The readout of the pads is realized with conventional
spectroscopy amplifiers and a multichannel analyzer for recording pulse height spectra. Two small X-ray
tubes, with gold and silver targets, are used to irradiate the detector.
The ion backflow and e used in the following are defined according to Eq. (4.2). The gain is determined
by the usual method of recording the current at the pad plane and the rate of absorbed X-rays of known
energy.

5.1.2 Gain stability measurements

Several gain stability measurements have been carried out in order to demostrate that the gain in a GEM
system is stable with time, even under changing conditions of gain or irradiation rate.
As an example, it is shown in Figure 5.2 the long term behavior of a triple GEM setup. The measurements
were performed in Ne-CO2 (90-10). The right-hand panel shows the current of the triple GEM stack over
a period of more than 11 days, in the middle of which the gain was increased from 1200 to 1800. The
observed fluctuations are due to gas pressure, temperature, and perhaps composition fluctuations, as
44 The ALICE Collaboration

confirmed by the pulseheight data of a single-wire proportional counter used as reference (left panel).
The wire counter data is used to correct the gain variations of the GEM detector. In Fig. 5.3 the corrected
GEM gain is shown for a period of about 21 hours, just after the gain was increased. Within this time
the corrected gain remains very stable, within 0.45 %, as indicated by the fit of the right panel of the
figure. Thus, no settling time is observed after changing the operating conditions. It should be noted that
a humidity level of about 180 ppm of water was maintained for the entire period.

5.1.3 Results of ion backflow measurements

Baseline solution
A suitable working point in terms of ion backflow and local energy resolution was found by utilizing a
quadruple GEM system in which the foils in layer 1 and 4 have a standard hole pitch (Standard, 140 µm),
whereas the foils in layer 2 and 3 have a hole pitch that is two times larger (Large Pitch, 280 µm). This
arrangement, denoted S-LP-LP-S, allows to block ions efficiently by employing asymmetric transfer
fields and foils with low optical transparency. An increasing sequence of gas gains down the GEM stack
helps reducing the ion backflow since ions created in the inner two layers are blocked more efficiently.
On the other hand, the efficiency for electron transmission, in particular in the first two layers, is also
affected by this configuration. Therefore, a combined optimization with respect to both ion backflow and
energy resolution is mandatory.

20
σ (%)

UGEM3/UGEM4=0.8 UGEM3/UGEM4=0.95
18 UGEM2=235 V UGEM2=235 V
UGEM2=255 V UGEM2=255 V
16 UGEM2=285 V UGEM2=285 V

14

12

10

0.0 0.5 1.0 1.5 2.0 2.5 3.0

IBF (%)

Figure 5.4: Correlation between ion backflow and energy resolution at 5.9 keV in a quadruple S-LP-LP-S GEM in Ne-CO2 -N2
(90-10-5) for various settings of DUGEM2 . The voltage on GEM 1 increases for a given setting between 225 and
315 V from left to right. The voltages on GEM 3 and GEM 4 are adjusted to achieve a total effective gain of 2000,
while keeping their ratio fixed. The transfer and induction fields are 4, 2, 0.1 and 4 kV/cm, respectively.

In Fig. 5.4 the ion backflow and energy resolution at 5.9 keV obtained with a S-LP-LP-S arrangement are
summarised for various voltage settings, illustrating the competing mechanisms of electron transmission
and ion blocking. The results are obtained in a Ne-CO2 -N2 (90-10-5) gas mixture for different com-
binations of DUGEM1 and DUGEM2 , and at different ratios DUGEM3 /DUGEM4 . Clearly the ion backflow
improves for lower gains at GEM 1 and GEM 2, while the energy resolution deteriorates accordingly.
Typical values of ion backflow around 0.7 % at energy resolutions of ⇠12 % are reached. This per-
formance fulfills the requirements for maximum allowable space-charge distortions and proper dE/dx
TPC Upgrade TDR 45

resolution of the detector, and defines the baseline solution for the detector configuration presented in
this TDR: a quadruple GEM system of the type S-LP-LP-S. A full set of typical HV settings for optimal
operation is shown in Tab. 4.2.
In the next sections we describe the R&D evolution from the commonly used standard S-S-S triple GEM
stack to the final S-LP-LP-S configuration and the systematic optimizations by scanning various settings
of GEM voltages and transfer fields. Further understanding on the behaviour of the quadruple system is
also provided.

Ion backflow results for a triple GEM setup

Extensive studies were performed with triple GEM detectors aimed at minimizing the ion backflow of the
structure. In order to find the optimal settings, the transfer fields ET1 and ET2 were varied systematically
while the voltage across GEM 3 was used to keep the gain at the desired value. The effective gain was
kept at ⇠2000, the drift field at 400 V/cm and the induction field at 4 kV/cm, while the X-ray rate was
always lower than a few tens of nA/cm2 . Several gas mixtures have been studied, although results are
shown only for the baseline gas mixture Ne-CO2 -N2 (90-10-5).
As described in Sec. 4.1, ion backflow suppression in a GEM system is based on an asymmetric field
configuration within the stack. This concept was proposed by the International Linear Collider (ILC)
community and described in [1]. Following this concept, the values of ET1 and Eind are maximized,
whereas ET2 is set at a minimum value to achieve maximum blocking of the ions produced at GEM 3. In
addition, the potential differences across the three GEMs follow an increasing order, such that

DUGEM1 < DUGEM2 < DUGEM3 . (5.1)

In this way, most of the ions are produced on GEM 3 and must enter the blocking region.
As an illustrative result, in Fig. 5.5 the ion backflow is shown as a function of ET1 for several ET2 values
in Ne-CO2 -N2 (90-10-5). A decrease of the ion backflow as a function of ET1 is observed due to the
increasing extraction efficiency of GEM 1. The best performance is reached at the lowest applied ET2
(0.1 kV/cm).
IBF (E ) for Ne-CO -N2 (90-10-5)
T1 2
0.07
IBF

ET2 = 0.1 kV/cm ET2 = 0.3 kV/cm

0.065 ET2 = 0.4 kV/cm
ET2 = 0.15 kV/cm
ET2 = 0.2 kV/cm ET2 = 0.5 kV/cm
0.06
ET2 = 0.6 kV/cm
0.055

0.05

0.045

0.04

0.035

0.03

0.025

0.02
3 3.5 4 4.5 5 5.5
ET1 (kV/cm)

Figure 5.5: Ion backflow in a triple GEM detector as a function of ET1 for several values of ET2 .

In conclusion, the ion backflow in triple GEM systems can be reduced to ⇠2.5 % after careful optimiza-
tion of the voltage and field settings. This number falls short of the requirements by more than a factor of
46 The ALICE Collaboration

two. No strong dependence on the gas mixture has been observed in any measurement (not shown here).
Therefore, structures with four GEMs, including those with large hole pitch, are considered next.

Ion backflow results with quadruple GEMs

As the next step towards a minimization of ion backflow, systematic studies of systems with four GEMs
were performed. The selected results presented in this section were obtained for various configurations of
quadruple GEM stacks, measured with different experimental setups. These studies include also the use
of foils with large hole pitch (LP) [2], 280 µm, that are combined with foils that have the standard hole
pitch of 140 µm (S). The ion backflow was measured simultaneously with the energy resolution using an
55 Fe X-ray source. All measurements presented here are for the baseline gas mixture Ne-CO -N (90-
2 2
10-5). The gain was systematically adjusted to 2000 by tuning DUGEM3 and DUGEM4 while maintaining
a defined ratio between the two.
Several HV scans were carried out with a structure composed of four standard foils (S-S-S-S). As an
example, a two-dimensional scan of the ion backflow as a function of ET2 and ET3 is shown in Fig. 5.6,
where ion backflow values of 2 % are reached for low values of ET2 . The energy resolution, expressed as
the sigma of a gaussian fit to the 5.9 keV peak divided by the mean value, is about 11 %.

IBF over T2 and T3 for Ne-CO2-N2 (90-10-5)

4
ET3 [kV/cm]

1 1.93 2.29 2.66 2.92 3.25 3.52 3.82 4.07 4.28 4.54

0.9 1.96 2.33 2.63 2.89 3.19 3.56 3.74 4.01 4.27 4.50 3.5

0.8 1.97 2.28 2.62 2.93 3.26 3.48 3.77 4.01 4.23 4.39
3
0.7 1.99 2.33 2.64 2.95 3.21 3.46 3.65 3.92 4.10 4.28

2.5
0.6 1.99 2.31 2.62 2.90 3.15 3.40 3.65 3.84 3.98 4.15

0.5 2.01 2.32 2.60 2.86 3.10 3.32 3.51 3.64 3.78 3.88
2
0.4 1.98 2.33 2.57 2.81 3.03 3.19 3.32 3.43 3.56 3.65
1.5
0.3 1.98 2.27 2.52 2.69 2.86 2.99 3.09 3.21 3.30 3.37

0.2 1.93 2.22 2.37 2.55 2.67 2.78 2.87 2.92 2.99 3.05 1

0.1 2.00 2.16 2.29 2.34 2.49 2.58 2.62 2.65 2.71 2.81
0.5
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
ET2 (kV/cm)

Figure 5.6: Two-dimensional scan of the ion backflow in a quadruple stack of standard GEMs (S-S-S-S) as a function of ET2
and ET3 . ET1 and Eind are both 4 kV/cm, and the voltages across the GEMs are tuned, in an increasing sequence, to
achieve an effective gain of 2000.

In the next step, one large-pitch foil is introduced to form a configuration of the type S-S-LP-S, aiming
at improving the ion backflow by applying a low transfer field (ET2 ) just above the large-pitch GEM with
low optical transparency. The two-dimensional scan shown in Fig. 5.7 illustrates that the ion backflow for
this configuration indeed improves as ET2 decreases and, to a lesser extent, as ET3 increases. It should be
noted that the use of large-pitch foils efficiently prevents accidental alignment of the holes of consecutive
foils. Ion backflow values below 1.5 % are achieved in the S-S-LP-S configuration.
This mechanism of ion blocking can be reinforced by introducing a second large-pitch foil in the quadru-
ple GEM stack, forming a S-LP-LP-S structure, where the two large-pitch GEMs are mounted in the core
of the stack. A scan of the ET2 field for such a configuration is shown in Fig. 5.8. Ion backflow values as
low as 0.34 % are obtained for high ET2 and low ET3 . In this configuration ions are absorbed at the top
side of GEM 4 due to the low field above, and the remaining ones are focussed between two (typically
misaligned) large-pitch foils and partially absorbed at the bottom of GEM 2. On the other hand, the en-
ergy resolution observed in this configuration, also shown in Fig. 5.8, is insufficient and requires a more
detailed optimization of the voltage configurations.
TPC Upgrade TDR 47

IBF over T2 and T3 for Ne-CO2-N2 (90-10-5)

4

ET3 [kV/cm]
4 1.14 1.51 1.65 1.78 1.88 1.99 2.08 2.09 2.16 2.16 2.16 2.40 2.34 2.28
1.16 1.53 1.69 1.75 1.91 2.02 2.12 2.17 2.19 2.20 2.20 2.03 2.38 2.33
3.5 1.17 1.55 1.72 1.84 1.95 2.06 2.15 2.21 2.23 2.24 2.24 3.5
2.44 2.36
1.19 1.58 1.77 1.88 2.01 2.09 2.21 2.27 2.28 2.29 2.29 2.31 2.50 2.41
3 3
1.21 1.61 1.80 1.93 2.05 2.15 2.27 2.31 2.33 2.34 2.34 2.35 2.64 2.46
2.5 1.24 1.65 1.83 1.97 2.10 2.22 2.34 2.37 2.40 2.41 2.40 2.42 2.66 2.52
2.5
1.30 1.69 1.90 2.04 2.17 2.29 2.42 2.44 2.47 2.46 2.47 2.48 2.76 2.61
2 1.31 1.76 1.96 2.11 2.24 2.36 2.51 2.51 2.53 2.52 2.55 2.55 2.78 2.69
2
1.35 1.85 2.03 2.16 2.30 2.44 2.63 2.54 2.62 2.63 2.62 2.63 2.80 2.76
1.5
1.40 1.87 2.09 2.23 2.37 2.45 2.66 2.68 2.69 2.65 2.68 2.84 2.81
1.5
1 1.43 1.90 2.12 2.26 2.37 2.51 2.58 2.66 2.69 2.68 2.68 2.67 2.81 2.80
1.48 1.94 2.10 2.23 2.36 2.46 2.57 2.61 2.63 2.64 2.62 2.63 2.75 2.73
0.5 1.51 1.88 1.99 2.12 2.22 2.32 2.43 2.48 2.50 2.48 2.47 2.47 2.61 2.58 1

1.45 1.48 1.58 1.69 1.79 1.89 1.99 2.01 2.02 2.03 1.97 2.00 2.20 2.10
0 0.5
0 0.5 1 1.5 2 2.5 3 3.5 4
ET2 (kV/cm)

Figure 5.7: Two-dimensional ion backflow scan as a function of ET3 and ET2 for a quadruple GEM of the type S-S-LP-S (large
pitch GEM in 3rd position). ET1 and Eind are both 4 kV/cm, DUGEM1 = 233 V, DUGEM2 = 250 V, and DUGEM3 and
DUGEM4 are adjusted to achieve the nominal gain of 2000.

2.0 20
IBF (%)

σ (%)
1.8 18
1.6 16
1.4 14
1.2 12
1.0 10
IBF
0.8 8
σ
0.6 6
0.4 4
0.2 ET3 = 0.1 kV/cm 2
0.0 0
0.0 0.5 1.0 1.5 2.0 2.5
ET2 (kV/cm)

Figure 5.8: Ion backflow and energy resolution in a quadruple S-LP-LP-S GEM as a function of ET2 for ET3 = 0.1 kV/cm. ET1
and Eind are both 4 kV/cm, and the voltages across the GEMs are in increasing sequence to achieve an effective gain
of 2000.

The low ion backflow values achieved so far provide room for the optimization of the energy resolution
by tuning other settings, such as the voltages across the GEM foils. For the same S-LP-LP-S arrangement
as before, Fig. 5.9 (left panel) illustrates the dependence of the ion backflow and the energy resolution
on DUGEM1 for three settings of DUGEM2 . A significant improvement of the energy resolution can be
achieved by the appropriate choice of DUGEM1 and DUGEM2 , at the expense of only a moderate increase
of the ion backflow. As quoted above, values of ion backflow around 0.7 % at energy resolutions of
⇠12 % can be reached, which fulfils the detector requirements.
In conclusion, suitable ion backflow figures can be achieved in a S-LP-LP-S quadruple GEM system,
but a simultaneous optimization of ion backflow and energy resolution must be made. The correlation
between the energy resolution and the square root of the ion backflow as shown in the right panel of
Fig. 5.9 demonstrates that the transparency of the system to electrons andpthe fraction of backflowing
ions are closely related. In settings where the ion backflow is large i.e. 1/ IB ! 0, energy resolutions
well below 10 % at 5.9 keV are observed, which is consistent with the expectation for ideal electron
48 The ALICE Collaboration

2.2 20 20

(%)
IBF (%)

σ (%)
σ
2.0 18 UGEM3/UGEM4= 0.95
1.8 16 UGEM2= 235 V
1.6 UGEM2= 255 V
14
15 UGEM2= 285 V
1.4
12
1.2
10
1.0
8
0.8 UGEM3/UGEM4= 0.8
10
6
0.6 UGEM2= 235 V
IBF, UGEM2=235 V 4
0.4 UGEM2= 255 V
IBF, UGEM2=255 V UGEM2= 285 V
0.2 2
IBF, UGEM2=285 V
0.0 0 5
220 240 260 280 300 320 0.5 1.0 1.5 2.0
VGEM1 (V) 1/sqrt(IB)

Figure 5.9: (Left) Energy resolution at 5.9 keV and ion backflow in a quadruple S-LP-LP-S GEMpas a function of DUGEM1
and various settings of DUGEM2 . (Right) Correlation between energy resolution and 1/ IB for various settings of
DUGEM1 , DUGEM2 and DUGEM3 /DUGEM4 . The voltage on GEM 1 decreases for a given setting between 225 and
315 V from left to right. The voltages on GEM 3 and GEM 4 are adjusted to achieve a total effective gain of 2000,
while keeping their ratio fixed. The transfer and induction fields are 4, 2, 0.1 and 4 kV/cm, respectively.

transparency.
p At small ion backflow, there is an almost linear correlation between the energy resolution
and 1/ IB. Since the energy resolution is most sensitive to losses of primary electrons that occur in
the first GEM layers, this strong correlation suggests that most of the remaining backflowing ions are
produced in the first GEM layers as well.
More systematic studies remain to be performed, where further voltage settings and stack arrangements
will be tried out. Given the performance and the flexibility in the choice of operational parameters
demonstrated in this section, we have confidence that a working solution for the ALICE TPC upgrade
will be reached with an ion backflow  1 %, based on a quadruple GEM system.

5.1.4 Discharge probability studies

Extensive studies of the discharge behaviour of standard triple GEM systems were carried out previ-
ously [3]. However, the detector concept for the ALICE TPC upgrade proposed in this TDR differs
considerably from these well-studied configurations. Therefore, detailed discharge probability studies
of the S-LP-LP-S quadruple GEM system at the proper voltage settings in Ne-CO2 -N2 (90-10-5) are
presently being performed.
We have already confirmed earlier discharge rate measurements [3] obtained by irradiating a standard
triple GEM with highly ionising particles at very high gains. While evidence exists that the addition of
N2 to the Ne-CO2 mixture reduces the discharge probability, the operation of a triple stack in a configu-
ration typical for minimising the ion backflow (increasing gain sequence and low ET2 ) does increase the
discharge rate. The addition of a fourth foil is expected to improve the situation. In order to assess the
robustness of quadruple stacks with the inclusion of large-pitch foils and ion backflow settings several
discharge measurements are being carried out in various laboratories. The studies will be focussed on
operation at nominal or slightly higher than nominal gains with various types of radiation, including al-
pha, beta and X-ray radioactive sources, and minimum ionizing and heavy-ion beams. The latter include
tests with 150 GeV/c pion beams at the CERN SPS1 , and with proton and heavy-ion collisions in the
ALICE cavern at the LHC.
The experience of the GEM-based muon trigger detector of LHCb can serve as a reference to judge the
required discharge performance of the detectors [4]. In this case a discharge probability of 10 12 , as
1 Super Proton Synchrotron
TPC Upgrade TDR 49

measured with a close-to-MIP beam, was considered satisfactory since the detectors had survived during
this beam test the number of sparks expected in 10 years of operation in the LHC [5] at the quoted rate.
The determination of a reasonable discharge probability limit with highly ionising particles can be done
by scaling the energy deposition to MIPs. These results shall be confirmed by operating a well-certified
full-size prototype detector under LHC conditions.

5.1.5 Comparison with simulations

Simulation studies with Garfield
The ion backflow is studied with the Garfield++ simulation package [6] and compared with experimental
measurements presented in the previous section. The geometry of the detector, the material properties
of the GEM, the voltage configurations, and the boundary conditions are defined in ANSYS [7]. This
program calculates the electric field in the detector with a finite elements analysis method. For the cal-
culation of the properties of gas mixtures and transport properties of electrons in a given electric field,
the simulation packages Heed [8] and Magboltz [9] are used. In the simulation studies, the gas mix-
tures Ar-CO2 (90-10), Ar-CO2 (70-30), Ne-CO2 (90-10), and Ne-CO2 -N2 (90-10-5) are used, although
results shown here are for the latter mixture only. The drift of electrons and the avalanche inside a hole
are studied with a microscopic transport and avalanche method. In Garfield++, this method tracks the
electron path at the molecular level using the drift velocity and the diffusion, Townsend and attachment
coefficients as calculated by Magboltz from the various electron-gas cross sections. Penning effects are
introduced in the calculation of the gas gain. Because a microscopic transport calculation for ions is not
available in this package, transport parameters such as the ion mobility as a function of the field over the
gas density (E/N) are set pby hand according to [10–12]. The diffusion of ions is given by the thermal
diffusion of DL = DT = 2kB T /qE, where kB is the Boltzmann constant, T the temperature, E the elec-
tric field and q the electrical charge. A Monte Carlo integration technique is utilized for tracking of ions.
In these simulations, the Munich detector configuration, as described in Sec. 5.1.3, is used.
It has been found in the course of these studies that the alignment of the holes of GEM 1 and GEM 2
crucially affect the resulting ion backflow in a triple GEM system. Since the hole alignment between
standard GEMs cannot be controlled in the measurements, misalignments between GEM 1 and GEM 2
are introduced as a free parameter in the simulations in order to describe the experimental results.
Figure 5.10 shows the probability distribution of the minimum distance between holes for two randomly
positioned GEM layers of 140 µm pitch. The mean and most probable values of the distribution are
50 µm and 70 µm, respectively; therefore, a misalignment of 50 µm is a reasonable value.
Figure 5.11 shows ion backflow in a 3-GEM stack as functions of ET1 and ET2 in Ne-CO2 -N2 (90-10-5)
for a misalignment of 45.9 µm. Although some difference in IB is observed for ET2 0.7 kV/cm and
ET1  3.5 kV/cm, the measured ion backflow for ET2  0.6 kV/cm is well described by the simulations.
As ET2 decreases and ET1 increases, the number of ions drifting back into the drift region from GEM 3
is reduced and the contribution of ions from GEM 2 becomes relatively large. The consistency of the
misaligment values used in the simulations has been checked by comparing measurements of the ion
backflow performed with the same detector setup but with different gas mixtures: Ar-CO2 , Ne-CO2 , and
Ne-CF4 . The measurements are always reproduced best in the simulations when using misalignment
values between 40 and 50 µm. The uncertainties in the mobility, longitudinal and transverse diffusion of
ions at low electric field, where no measurements exist, might add an additional systematic uncertainty
to the comparison between simulations and measurements.
The measured ion backflow in Ne-CO2 -N2 (90-10-5) with two different quadruple GEM configurations
(with one and two large pitch foils) has been compared with the corresponding simulations. The results
are shown in Fig. 5.12. In the simulations various combinations of misalignment for GEM 2 and GEM 3
were considered, otherwise the voltage settings are identical. In general, good agreement is achieved
50 The ALICE Collaboration

minimum distance between holes

Probability
0.025

0.02

0.015

0.01

0.005

0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
dr (mm)

Figure 5.10: Probability distribution of the distance between closest holes in two randomly aligned GEM layers with pitch
140 µm.

hole distance = 0.65625*70 ( m) hole distance = 0.65625*70 ( m)

0.12 0.12
IBF

IBF

meas. sim. E T2 = 0.8 kV/cm meas. sim. E T2 = 0.7 kV/cm

meas. sim. E T2 = 0.6 kV/cm meas. sim. E T2 = 0.5 kV/cm
0.1 meas. sim. E T2 = 0.4 kV/cm 0.1 meas. sim. E T2 = 0.3 kV/cm
meas. sim. E T2 = 0.2 kV/cm meas. sim. E T2 = 0.1 kV/cm

0.08 0.08

0.06 0.06

0.04 0.04

0.02 0.02

0 0
3 3.5 4 4.5 5 5.5 3 3.5 4 4.5 5 5.5
ET1(kV/cm) ET1(kV/cm)

Figure 5.11: Ion backflow in Ne-CO2 -N2 (90-10-5) as function of ET1 for different settings of ET2 for a triple GEM config-
uration. The data from measurements (closed symbols) and simulations (open symbols) with a misalignment of
45.9 µm is distributed on the two panels for better visibility.
TPC Upgrade TDR 51

between simulations and measurements. In particular, the observed improvement from S-S-LP-S to S-
LP-LP-S is well described by the simulations. This indicates that the qualitative behavior of multi-GEM
stacks in terms of ion backflow is well understood, in spite of an incomplete knowledge of ion transport
parameters and the uncertainty expected from alignment.

0.050

IBF
0.045 Meas S-S-LP-S
Sim1 S-S-LP-S
0.040 Sim2 S-S-LP-S
Sim3 S-S-LP-S
0.035
0.030
0.025
Meas S-LP-LP-S
0.020 Sim1 S-LP-LP-S
0.015 Sim2 S-LP-LP-S
Sim3 S-LP-LP-S
0.010
0.005
0.000
0 1 2 3 4 5
ET1 (kV/cm)
Figure 5.12: Comparison of ion backflow simulations (open symbols) with measurements (closed symbols) as a function of ET1
in Ne-CO2 -N2 (90-10-5) for two quadruple GEM configurations. The circles represent an S-S-LP-S arrangement,
where the voltages across the GEMs are 220, 270, ⇠ 275 and ⇠ 280 V, respectively. The squares represent an
S-LP-LP-S arrangement, where the voltages across the GEMs are 230, 280, ⇠ 290 and ⇠ 320 V, respectively. For
both arrangements the comparison has been done for ET2 = 3.7 kV/cm, ET3 = 0.2 kV/cm and EInd = 4 kV/cm.
In the measurements the gain is adjusted to 2000, and the simulations follow the same settings. The different
simulation setups labeled Sim 1 to Sim 3 were performed with three different sets of foil misalignment.

5.2 Full-size IROC prototype

A full-size prototype of a TPC Inner Readout Chamber (IROC) equipped with a triple GEM readout stage
was built. The goal was to study aspects of GEM integration on a large-size chamber and to validate the
dE/dx capabilities of a GEM-based TPC. Moreover, issues of operational stability are studied under
LHC running conditions.

5.2.1 Detector design

The prototype is assembled on a spare MWPC IROC of the TPC [13]. It is a trapezoidal chamber with
dimensions 497 ⇥ (292 – 467) mm2 . The chamber, after removing the wires, was equipped with three
GEM foils. The mechanical structure of the chamber is described in Sec. 4.2.

GEM foils
Four large area GEM foils were produced at CERN using the single-mask technique. The top side of the
trapezoidal foil is segmented into 18 individually powered sectors with an area of ⇠100 cm2 each (see
Fig. 5.13). The inter-sector gap is 400 µm, the same as the thickness of the spacer grid of the mounting
frames (see next paragraph). An additional 100 µm of copper between the edges of the sector and its
active area (GEM holes) is added to account for possible misalignment during the production of the foil.
The diameter of the biconical holes is ⇠50 µm (inner) and ⇠80 µm (outer). The hole pitch is 140 µm.
52 The ALICE Collaboration

Figure 5.13: Layout of a prototype GEM foil; the top side is segmented into 18 sectors.

The HV distribution traces run along three sides of the foil. It consists of 2 mm copper paths and con-
nection flaps (5 for the top side, 4 for the bottom). Sectors are powered in parallel via loading resistors
soldered directly on the foil in the designed place between the distribution path and a sector (more on
HV supply in Sec. 5.2.4). The bottom (not segmented) side of the foil is connected directly to the HV,
therefore no distribution path is needed.

Support frames

The foils are glued on 2 mm fiberglass (G10) frames (see Fig. 5.14a) and then mounted in a stack. The
frames contain a 400 µm thick spacer grid whose role is to prevent the foils to approach each other due
to electrostatic forces. Each grid is aligned with the sector boundaries, so no additional dead area is
introduced.
When the stack is mounted, the loading resistors fit into grooves milled in the bottom side of the frame
placed above. Framed foils, therefore, lie flat on top of each other with 2 mm distance between them.
The sides of the frames include grooves for the HV flaps and wires. They are aligned with feed-throughs
machined in the alubody of the chamber (see Fig. 5.14b). The HV wire runs through this groove to the
flap where it is soldered, as shown in Fig. 5.14c so that all material is contained within the dimensions
of the alubody; this is necessary for mounting the readout chambers on the TPC endplates [13].
The frames, after production in the CERN workshop, were polished, cleaned in an ultrasonic bath and
dried in an oven, in order to remove remaining pieces of fiber and dust of e.g. G10 material, which may
spoil the electrostatic integrity of the foils.

5.2.2 Quality assurance (QA)

All four foils are tested before and after framing to validate them for the final assembly. These tests
include optical and high voltage tests, as foreseen for the large-scale production of the TPC chambers
(see Sec. 4.7 for more details). In the following subsection, we will focus on the main results of the QA
tests and its implications for the operation of the prototype.
TPC Upgrade TDR 53

Figure 5.14: (a) Layout of the support frame. (b) Detail of the provisions for connecting each sector to its HV wire. (c) HV
wire connected to the flap.

Optical check

The main tasks of this QA step are to check the hole size uniformity across the GEM active area and
to record the mechanical features of the foils. Large-size defects on the copper or Kapton layers may
increase the probability of electrical breakdown. Big mechanical defects, like cuts, may result in short
circuits between the top and bottom side of a foil.
The measured diameter of the inner (Kapton) holes varies between 40 and 50 µm. The outer (copper)
holes have a diameter of 70 – 80 µm. The variations of the hole sizes may come from the etching process
and are within reasonable limits. The pitch of ⇠140 µm is constant over the foils, since it is defined by
the printed mask.

HV tests

A HV test is performed at each step of the detector assembly. The testing procedure is similar to the one
which will be applied for QA of the final foils (see Sec. 4.7), therefore only a few particular aspects of
the procedure and the results are described in this subsection.
During the HV test each sector of the foil is ramped up twice to 550 V while the surrounding sectors
are kept at ground potential. During the first ramp, the leakage current (Ileak ) and the discharge rate are
measured at five voltage steps during the ramp. Sectors are connected via protection resistors with value
100 MW. Typical leakage currents vary between 0 and 0.3 nA. Sectors drawing currents higher than 5 nA
do not pass the test. Also, if the number of discharges at each ramping step exceeds 2 in about a minute,
the procedure is stopped and the sector fails the test. For the second test, the sectors are ramped up to
550 V directly. To pass the test, a sector should stand at this potential without sparking for a few minutes.
All foils (72 sectors) are tested before and after framing the foils.
Before assembling the detector, all sectors in all four foils were stable and had low and acceptable leakage
currents.
54 The ALICE Collaboration

5.2.3 Detector assembly

Gluing the foils
All foils were glued onto the fiberglass (G10) frames. Before gluing, the foils are stretched on a pneu-
matic stretching tool with a tension of 10 N/cm. Once the foil is stretched, it is positioned on its frame
and aligned with metal pins. A heavy aluminum plate (milled in a way to prevent it from touching the
active area of the foil) is used to press the foil onto the frame. The epoxy used is ARALDITE 2011 [14].
The full assembly is kept for 24 hours under a hood heated to 60 o C. Subsequent steps of the gluing
procedure are shown on Fig. 5.15.

Figure 5.15: Foil gluing procedure. The different mounting steps are described in the text.

GEM stack
After gluing, the remaining material surrounding the frames is cut off, and the framed foil (see Fig. 5.16a)
is ready for a new HV test. After the test, the loading resistors (1 and 10 MW SMD) are soldered as
shown in Fig. 5.16b. Next, the three foils are mounted in a stack on the alubody of the chamber, with
the unsectorized side facing the pad plane. Exceptionally, for this prototype an additional frame was
mounted between the last (bottom) GEM foil and a pad plane in order to increase the induction gap from
2 to 4 mm.
Six HV wires are soldered to the HV flaps on each foil. High voltage is applied to the wires via SHV
connectors on the other side of the alubody. The wires run through feed-throughs drilled in the aluminum
frame of the chamber, which are then sealed with epoxy. The GEM stack is finally screwed to the alubody
with nylon screws.

Test box with field cage

The chamber is mounted in a test box (see Fig. 5.16d), which contains a drift cathode and a rectangular
field cage with dimensions of 57 ⇥ 61 cm2 . The drift electrode is made of 50 µm aluminized Kapton
foil. The field cage has 8 field-defining strips (see Fig. 5.17) with a pitch of 15 mm. The strips are
interconnected with 1 MW resistors.
The maximum drift distance to GEM 1 is 10.6 cm. The last strip of the field cage is located 1 mm below
the position of that foil (see Fig. 5.17). Therefore, the potential of the last strip, which is grounded via a
TPC Upgrade TDR 55

Figure 5.16: (a) Framed GEM foil (b) HV connection to a GEM sector through the loading resistors. (c) GEM stack mounted
on the alubody. (d) chamber mounted in the test box.

3.33 MW resistor, is adjusted by applying a voltage in order to match the drift field at the top electrode of
GEM 1.

Figure 5.17: Cross section of test box with field cage and mounted IROC. The position of the first foil in the GEM stack is
marked by a red line and labeled.

Two walls of the test box, closest to the parallel sides of the chamber, were machined to install mylar
windows for measurements with beam and radioactive source.

5.2.4 HV supply
The detector is powered using an ISEG EHS 8060n 8-channel 6 kV high voltage module for the three
GEM foils and the last strip voltage of the field cage, and by an ISEG HPn300 30 kV module for the
drift electrode. The system features high precision current measurements (resolution 1 nA) on each HV
56 The ALICE Collaboration

channel, adjustable ramp speeds and full remote control, which allows a global shutdown in case any
channel trips. The powering scheme of the GEM-stack is displayed in Fig. 5.18.

Figure 5.18: Schematics of the HV distribution of the prototype, showing the loading and grounding resistors.

Loading resistors (RL ) are installed for each sector on the top side of each foil. In addition, each channel
is grounded through grounding resistors (RG ). Placing loading resistors on the top side of the foil assures
that in case of a discharge across a foil, the voltage drop occurs only on the top side, whereas the bottom
side stays at its nominal potential. This helps prevent the propagation of the discharge to the next foil or
to the pad plane and readout electronics. Moreover, in case of occasional sparks, the loading resistors
limit the current supplied by the power supply before it trips. Loading resistors also decrease the current
flowing through the sector in case of a short circuit between top and bottom sides of the foil.
The values of RL were chosen keeping in mind the current densities expected in the future Pb–Pb colli-
sions at 50 kHz at a gain of 2000, which is roughly 5 nA/cm2 (500 nA per GEM sector). Such a current
may result in a significant potential drop across a large loading resistor, thus reducing the gain. Therefore
10 MW resistors were chosen for GEM 1 and GEM 2, whereas RL = 1 MW for GEM 3 .
Large resistors to ground for each channel are chosen to assure safe discharge of the GEM foils after
a HV trip. Such connection will result in a constant DC current to ground which must not exceed the
current limit of each channel. On the other hand, in case of a short across a foil in one or several sectors,
the rest of the sectors should remain fully operational. Therefore, the value of RG on the top side should
be high enough to allow sufficient current supply through the shortened sectors, whereas the resistor to
ground for the bottom side should be as low as possible in order to avoid reverse currents into the HV
supply. The HV powering scheme with the chosen loading and grounding resistors is shown in Fig. 5.18.
Each sector can be treated as an individual capacitor in parallel with the other sectors on the foil. A time
constant RC for (dis-)charging the foil is given by the full capacitance to the foil and the sum of loading
and grounding resistances. In case the time constant is higher for the top side of the foil than for the
bottom one, the potential of the top electrode may decrease to zero slower than the bottom one. This
would result in a sudden increase of the potential difference between the electrodes which may lead to
breakdown or even damage of the foil.
The example described above is simplified and does not take into account other parasitic elements or the
influence of different elements in the HV circuit on each other. Nevertheless, it shows the importance
of a proper choice of all electronic elements in the HV circuit used for powering the GEM stack. The
final choice of the grounding resistors was done after performing a set of transient SPICE simulations
and measurements of the discharge decay times. For the measurements a model of a GEM foil was made
TPC Upgrade TDR 57

with equivalent capacitors and resistors corresponding to the GEM assembly. The voltage difference
across both sides of the equivalent GEM was recorded on an oscilloscope. The values of the grounding
resistors were 5 MW, 5 MW and 3.33 MW for the top side of GEM 1, GEM 2, GEM 3, and 10 MW, 10 MW
and 3.33 MW for the bottom sides, respectively. This configuration allows to run the detector with several
shortened sectors in GEM 1 or GEM 2 and up to 2 shortened sectors in GEM 3, using the HV settings
described in the next paragraph.

HV settings for the prototype

We work with two groups of HV settings: the so-called ”standard” settings, typically used for triple GEM
structures, and the so-called ”ion backflow (IBF) settings”, where the field configurations are aimed at
minimizing the ion backflow. Each setting can be scaled in order to vary the total gain. Both are defined
in Tab. 5.2.
Standard IBF
Drift Field 0.4 kV/cm 0.4 kV/cm
DUGEM1 276 V 225 V
Transfer Field 1 2.57 kV/cm 3.8 kV/cm
DUGEM2 252 V 235 V
Transfer Field 2 2.57 kV/cm 0.60 kV/cm
DUGEM3 221 V 285 V
Induction Field 2.57 kV/cm 3.8 kV/cm
Table 5.2: Standard and ion backflow high voltage settings for a gain of ⇠2000 in Ne-CO2 (90-10).

The standard settings are inherited from the COMPASS experiment [15] where GEM detectors are op-
erated in Ar-CO2 (70-30). In Ne-CO2 (90-10) however, since the amplification in this mixture starts at
lower electric fields, a scaling factor (SF) between 69 – 73 %, resulting in gains between 2000 and 6000,
is applied. The standard settings are optimized for maximum stability according to the principles of
stable operation described in [16]. The highest amplification takes place in the first GEM and decreases
in subsequent stages. These settings however are not optimized for minimum ion backflow. In the case
of the ion backflow configuration, where the voltages were optimised in the lab, only the GEM voltages
(DUGEMi ) were scaled with factors of 100, 103, 105 and 107 %. In addition four different values of ET2
were used: 200, 400, 600 and 800 V/cm. Combining different values of SF and ET2 together 16 ion back-
flow settings were tested. The value of the ion backflow for these settings, in Ne-CO2 (90-10), varies
between 3 % and 6.5 %. The approximate gains for all 16 combinations of ion backflow settings vary
between 1000 and 6000. The highest amplification in this configuration occurs in GEM 3.
The nominal drift field of the ALICE TPC, Edrift = 400 V/cm, is also applied to the prototype. The
potential on the last strip of the field cage is equal to the potential on GEM 1 top.

5.2.5 Prototype commissioning

The prototype in the test box was tested in the lab with Ar-CO2 (90-10) and Ne-CO2 (90-10) gas mixtures,
with standard HV settings, and irradiated with a 55 Fe X-ray source.
For the commissioning of the detector, ⇠250 pads (⇠75 cm2 ) were connected to a charge sensitive pream-
plifier and the signal was digitized with a multi-channel analyzer for measuring the X-ray spectrum. A
picoammeter is used to measure the gain by counting the rate of absorbed X-rays in the gas and measur-
ing the current in the pad plane. Figure 5.19 shows the measured effective gain of the chamber for both
gas mixtures.
Figure 5.20a shows the energy spectrum of 55 Fe obtained in Ar-CO2 (90-10) at 86 % of the standard
settings, which corresponds to a gain of around 6000 (see next paragraph for more details). Figure 5.20b
shows the spectrum obtained in Ne-CO2 (90-10) at 75 % of the standard settings, which corresponds to a
58 The ALICE Collaboration

Gain
104 Ne-CO2 (90-10)
Ar-CO2 (90-10)

3
3× 10

70 72 74 76 78 80 82 84 86 88 90
Scaling Factor (%)

Figure 5.19: Effective gain of the chamber as a function of HV, expressed in terms of the scaling factor of the standard settings.

gain of around 11000. The energy resolution of the main peak for both measurements is between 10 and
11 % (sigma).

Figure 5.20: 55 Fe spectra obtained in Ar-CO2 (90-10) (left panel) and Ne-CO2 (90-10) (right panel).

5.2.6 Test campaign at the CERN PS

Experimental setup
The dE/dx resolution of the GEM IROC prototype was evaluated in a test beam at the CERN PS with
beams of either e+ and p + or e and p with momenta ranging from 1 to 6 GeV/c.
For each type of beam both standard and ion backflow high voltage settings are used (see Sec. 5.2.4).
The drift field is 400 V/cm and the gas mixture Ne-CO2 (90-10) as in the current TPC. Two scintillators
are used for beam definition, and a Cherenkov counter and a Pb-Glass calorimeter are used for online
particle identification.
The prototype is equipped with 10 front-end cards, covering about 1200 pads (see Fig. 5.21a). The
readout electronics have been borrowed from the LCTPC (Linear Collider TPC) collaboration. The
width of the readout region is 6 – 7 cm (see Fig. 5.21b).
The system has an RMS noise of about 600 electrons. The zero suppression threshold is 2 ADC counts,
corresponding to about 2000 electrons (120 ns peaking time and 12 mV/fC conversion gain). The sam-
pling frequency is 20 MHz.
The FECs are read out using the current TPC readout system [13]: Two Readout Control Units (RCU)
TPC Upgrade TDR 59

send the data to a Local Data Concentrator PC, which contains the receiving ReadOut Receiver Card
(RORC) and runs the ALICE data acquisition system DATE. The corresponding trigger logic is handled
by a Local Trigger Unit (LTU) and a Busy Box [13]. Data transfer and trigger communication are based
on optical links. The beam detectors, scintillators, Cherenkov counter and Pb-Glass calorimeter, are
read out through a classic CAMAC system into a PC running a LabView acquisition system. In order
to synchronize the events from both systems, a busy logic is implemented in the CAMAC system which
incorporates the busy status of both systems. Thus, the CAMAC leads the acquisition. The two data
streams are subsequently merged into a single data file based on the proper synchronization of the trigger
and an event tag. This allows the online visualization of the detector response to different particle species,
as selected by the beam detectors.
The average DAQ rate was 500 events/spill (where the spill length was 0.5 s) for a beam intensity of
about 2000 particles/spill.

dE/dx measurements
For the measurement of the dE/dx resolution, only events with isolated tracks are selected. Additional
cuts are applied on the number of clusters per track and the cluster drift time.
Figure 5.22 (left) shows the distribution of cluster maxima as a function of the pad row number. The first
rows show low amplitudes due to a mapping error in the readout electronics. Moreover, a few pad rows
with low gain are observed. These are related to the HV sector boundaries and the position of the spacer
grid. After gain equilibration, the overall spread is reduced (right) but the lower gain in the first pad
rows and at the positions of the spacer is still visible. The calibration of these pad rows requires a more
sophisticated procedure. For the present analysis, 17 low-gain pad rows out of 63 are excluded. It should
be pointed out that the distance between the GEM sectors, 400 µm in this prototype, (see Sec. 5.2.1) will
be reduced by a factor of 2 in the final design (see Sec. 4.3.1). In addition, the number of spacer grid
bridges in the support frames will also be reduced. No correction for pressure and temperature variations
are applied because the runs are rather short.
The dE/dx of each track is defined as the truncated mean of the 5 – 75 % highest of up to 46 cluster
charges. An example dE/dx spectrum of 1 GeV/c electrons and pions recorded at a gain of about 5000 is
shown in Fig. 5.23. The dE/dx distributions are fitted with a Gaussian function to obtain the mean value
hdE/dxi and the width of the distribution s (dE/dx). The relative resolution, defined as

Figure 5.21: (a) Detector equipped with the front-end electronics. (b) Schematic view of the read out region of the chamber.
60 The ALICE Collaboration

300 300
cluster charge (arb. units)

cluster charge (arb. units)

250 250

200 200

150 150

100 100

50 50

0 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
pad row pad row

Figure 5.22: Gain profile before (left panel) and after (right panel) the gain equalization for the 63 pad rows. The structure of
the spacer grid can be clearly seen in both panels.

e- p = 1GeV/c
600
π- G = 5000
500

400

300

200

100

0
0 50 100 150 200 250 300 350 400 450
〈dE /dx 〉 (total charge)

Figure 5.23: dE/dx spectrum of 1 GeV/c electrons and pions at a gain of 5000.

s (dE/dx)
, (5.2)
hdE/dxi

is derived for pions and electrons, for different momenta and HV settings. Figure 5.24 shows the results
obtained within this analysis. All ion backflow settings yield similar dE/dx resolution at a given beam
momentum and particle type. No significant dependence on the gas gain is observed within the range
under study. The results for the standard settings are slightly better than those for the ion backflow
settings.
The corresponding separation power, defined as

2 hdE/dxiA hdE/dxiB
SAB = (5.3)
s (dE/dx)A + s (dE/dx)B

for two particle types A and B, is shown in Fig. 5.25, for electrons and pions at all measured particle
momenta and HV settings. At a given beam momentum, asymptotic separation powers are reached for
gains above ⇠1500.
This dE/dx performance can be compared with results of a MC2 simulation shown in Fig. 5.26. The
2 Monte Carlo (MC)
TPC Upgrade TDR 61

(dE /dx )/ dE /dx (%)

14 tf2: 200
1GeV e- tf2: 400 1GeV -

13 tf2: 600
tf2: 800
Std (gain/69%*100%)
12

11

10

9
100 102 104 106 108 100 102 104 106 108
14
2 GeV e- 2 GeV -

13

12

11

10

9
100 102 104 106 108 100 102 104 106 108
14
3 GeV e- 3 GeV -

13

12

11

10

9
100 102 104 106 108 100 102 104 106 108
14
6 GeV e+ 6 GeV +
13

12

11

10

9
100 102 104 106 108 100 102 104 106 108
gain scaling factor (%)
Figure 5.24: The relative dE/dx resolution measured for different HV settings for electrons and pions with momenta ranging
from 1 to 6 GeV/c. Only 46 pad rows are used for the analysis. In the standard settings (grey curve) the gain
ranges from 2000 to 6000; in the ion backflow settings (coloured curves) the gain ranges from 1000 to 6000.
62 The ALICE Collaboration

e/ separation ( )
7
1GeV (neg.) 2 GeV (neg.)
6

0
100 102 104 106 108 100 102 104 106 108
7
3 GeV (neg.) 6 GeV (pos.)
6
5

4
3
2
1
0
100 102 104 106 108 100 102 104 106 108
gain scaling factor (%)
Figure 5.25: Separation power between pions end electrons with 1 6 GeV/c momentum measured for different HV settings.
In the standard settings (grey curve) the gain ranges from 2000 to 6000; in the ion backflow settings (colourful
curves) the gain spans between 1000 and 6000 (see Sec. 5.2.4 for more details).

simulated dE/dx spectrum for 1 GeV/c pions obtained with 46 pad rows in a GEM stack yields a relative
energy resolution of ⇠10.5 %. This is slightly better than our test beam results. Nevertheless, the dE/dx
resolution observed in the GEM IROC is compatible with that of the MWPC IROCs. Therefore these
measurements corroborate that the dE/dx resolution of the upgraded TPC will be preserved.
Counts (normalized)

0.3

0.25 63 rows

46 rows
0.2

0.15

0.1

0.05

0
25 30 35 40 45 50 55 60 65 70
dE/dx (ADC ch)
Figure 5.26: Simulated dE/dx spectrum for 1 GeV/c pions using information from all 63 rows (red) and for 46 rows (blue) as
in the data from test beam. The resulting relative resolution is 10.5 % (9.0 %) for 46 (63) rows.
Chapter 6

Front-end electronics and readout

This section reviews the requirements for the front-end electronics (FEE) and describes the general ar-
chitecture and the basic building blocks of the readout chain.

6.1 System overview

The requirements for the front-end and readout electronics are derived from the detector performance
requirements. The argumentation for many items is based on the successful implementation of the current
TPC readout system (as described in [1, 2]). The three main changes for the front-end electronics with
respect to the existing system are related to the new concept of continuously reading data from an ungated
GEM TPC:

1. The GEM readout provides signals with opposite polarity as compared to those generated in an
MWPC.

2. The continuous readout scheme necessitates the development of electronics that can concurrently
sample the detector signals and transfer the aquired data off the detector.

3. The continuous readout in conjunction with the step-up in interaction rate (and thus detector occu-
pancy) leads to a strongly increased data throughput.

The readout system for the GEM TPC is shown schematically in Fig. 6.1. The current signals are passed
from the pads on the detector’s readout plane to the front-end cards (FEC), located a few cm away, via
flexible Kapton cables. In the FEC, a custom-made FE1 ASIC2 , the SAMPA chip, processes the data
from 32 individual front-end channels concurrently. The SAMPA is common among different ALICE
subsystems (see Sec. 6.4). The first stage of the SAMPA is a charge-sensitive preamplifier and shap-
ing amplifier, which transforms the currents induced in the pads into differential semi-Gaussian voltage
signals. These signals are continuously digitized and processed by a DSP3 . Concurrently, the acquired
data are transferred to the GBTx ASIC [3], which multiplexes them and transmits them via the versatile
optical link components [4] to a Common Readout Unit (CRU). Also the CRU is a common ALICE
development. It serves as interface to the online farm, trigger and detector control system, and is situ-
ated off-detector in a control room close to the online farm. The common ALICE electronics projects,
SAMPA and CRU, are described in more detail in a separate Technical Design Report [5].
1 Front-End (FE)
2 Application-Specific Integrated Circuit (ASIC)
3 Digital Signal Processor (DSP)

63
64 The ALICE Collaboration

ESD protection
SAMPA ASIC
GBTx ASIC
GBT readout links: data
and monitoring (unidir.)

VTTx

Trigger, timing and

FEC
VTRx clock distribution (TTS)

power

Figure 6.1: Schematic of the readout system of the GEM TPC. The two main building blocks of the FEE are shown: The front-
end ASIC SAMPA on the front-end cards (FECs). The FECs connect to a Common Readout Unit (CRU), located
off-detector in the control room, through radiation hard GBT links.

In addition to the continuous readout mode, also a triggered mode has to be supported for calibration
purposes and for running at lower interaction and/or trigger rates. Here, the processing of the continu-
ously sampled data stream starts upon arrival of a first-level trigger; only the data corresponding to the
detector drift time (td ⇡ 100 µs) is frozen in the data memory and read out.

6.2 Pileup and occupancies

p
At Pb–Pb collisions with sNN = 2.76 TeV densities of primary tracks of hdNch /dhicent ⇡ 1600 and
hdNch /dhimb ⇡ 400 were measured for the most central collisions (0-5 %) and for minimum bias col-
p
lisions, respectively [6]. By scaling with (sNN )0.15 [7] to the full LHC energy of sNN = 5.5 TeV this
translates to hdNch /dhicent ⇡ 2000 and hdNch /dhimb ⇡ 500 for central and minimum bias Pb–Pb colli-
sions, respectively.
At 50 kHz interaction rate, the average number of interactions within a time window of td ⇡ 100 µs is
Npileup = 5. The expected primary track densities can be expressed conveniently by introducing an equiv-
alent charged-particle pseudo-rapidity density dNch /dh, which has the mean value hdNch /dhiequiv =
2500 for Npileup = 5. For central events (embedded in 4 minimum bias events) the value is dNch /dh|equiv =
4000. This is still well below the primary charged particle multiplicity for central Pb–Pb collisions orig-
inally anticipated when the current TPC was designed (dNch /dh ⇡ 8000).
The relative fluctuation of the number of charged tracks inside the TPC drift volume sNch /hdNch /dhiequiv
can be written as

s ✓ ◆2
sNch 1 sNMB
=p 1+ . (6.1)
hdNch /dhiequiv Npileup µNMB

With Npileup = 5 and the relative fluctuation of the number of minimum bias events sNMB /µNMB ⇡ 1.15 this
leads to sNch /hdNch /dhiequiv ⇡ 0.68. The relative fluctuations of the equivalent multiplicity as a function
of Npileup are depicted in Fig. 6.2. For Npileup = 5 the equivalent multiplicity will stay below (1 + 3 ⇥
0.68)hdNch /dhiequiv ⇡ 7500 for 99.7 % (corresponding to 3 sNch ) of the time intervals (⇠ 100 µs).
TPC Upgrade TDR 65

equiv
14000
Mean
Median

dNch/d
12000 k=90 %
k=95 %
k=99 %
10000 k=99.9 %

8000

6000

4000

2000

0
1 2 3 4 5 6 7 8 9
<N ev>
Figure 6.2: Equivalent multiplicity as function of mean number of events including kth order statistics (the kth-smallest value).
E.g. at 50 kHz collision rate (for Npileup = 5) there is a < 1 % probability that the equivalent multiplicity is larger
than 7500.

6.3 Data rates and bandwidth considerations

The following estimates of data rates are based on measured sizes of isolated events (no pileup). We
assume unchanged shaping, sampling and zero suppression parameters of the FE ASIC. In RUN 1 the
p
size of minimum bias TPC events for sNN = 2.76 TeV Pb–Pb collisions is 13.6 MByte. Several factors
have to be considered when extrapolating the data volume to RUN 3. An increase of the pad occupancies5
p
by a factor 1.25 must be expected due to the larger charged particle multiplicities at sNN = 5 TeV.
Moreover, an additional factor of 1.25 accounting for the deterioration of the compression ratio of the
zero suppression and run-length encoding must be incorporated at these larger occupancies6 . On the other
hand, the cluster sizes are reduced by a factor 0.8 due to the new readout technology (see Fig. 4.17), and
overlap of clusters reduces the data volume by a small factor. With these considerations in mind we
conservatively assume an average event size of 20 MByte for an isolated event in the TPC.
An average data rate from the TPC of 50 kHz ⇥ 20 MB = 1 TByte/s is expected for RUN 3. The average
pad occupancies will be around 15 %, increasing to up to 27 % for the innermost pad row, as can be seen
in Fig. 6.3. For a central event at Rint = 50 kHz we find occupancies of up to 42 % for the innermost pad
row, but even more extreme variations should be considered: for dNch /dh|equiv = 7500 occupancies of
up to 80 % may be occasionally reached.
Table 6.1 summarizes the number of front-end channels, the bandwidth requirements (based on the aver-
age expected occupancies) and the resulting distribution of FECs, SAMPAs, optical components (VTTx7
and VTRx8 ) and readout and TTS9 fibers. Here a segmentation of each TPC sector into 5 readout par-
titions is used: 2 for the IROC and 3 for the OROC (following the GEM segmentation, Sec. 4.5). The
bandwidth requirement is calculated using a safety factor of ⇠ 2 on the expected average data rates to
incorporate also local occupancy variations due to geometry. This factor depends on the occupancy dis-
tribution on the given readout partition and on the buffer size in the SAMPA ASIC and is presently being
verified in detailed simulations of the full readout architecture.

5 Pad occupancy is the fraction of samples within a given time window exceeding the zero suppression threshold.
6 This is the worst case scenario. Overlap of the clusters will actually reduce the impact of this effect.
7 Versatile Twin Transmitter (VTTx)
8 Versatile Transceiver (VTRx)
9 Trigger and Timing distribution System (TTS)
66 The ALICE Collaboration

Average pad occupancy (%)

70 Occupancies at different equivalent
multiplicities (dNch/d equiv ),
60 interaction rate: 50 kHz

50
750
0
40

30
4000

20 average
(2500)
10

0
0 20 40 60 80 100 120 140 160
Pad row
Figure 6.3: Expected average occupancies within a given time window for equivalent multiplicities of dNch /dh|equiv = 2500,
4000 and 7500. The data is extrapolated using measured occupancies in isolated (no pileup) events recorded in
2010.

IROC 1 IROC 2 OROC 1 OROC 2 OROC 3 TPC

Avg. data rate / chan. (Mbit/s) 22 16 15 13 11
Req. bandwidth / chan. (Mbit/s) 40 30 30 25 20
Front-end channels 2304 3200 2944 3712 3200 552,960
Total data rate (Gbit/s) 50 50 45 50 35 8280
Total bandwidth (Gbit/s) 100 100 90 100 70
FECs (5 SAMPAs each) 15 20 19 24 20 3528
SAMPAs 75 100 95 120 100 17,640
GBTx ASICs 30 40 38 48 20 6336
Versatile link twin transmitter (VTTx) 15 20 19 24 0 2808
Versatile link transceiver (VTRx) 8 10 10 12 20 2160
GBT uni-directional data links (3.2 Gbit/s) 30 40 38 48 20 6336
GBT uni-directional TTS links 8 10 10 12 10 1800

Table 6.1: Data rates and bandwidth requirement and partitioning of SAMPAs, FECs, optical components, readout and TTS
fibers with 5 readout partitions per TPC sector. The acronyms are explained in the text. The numbers in the last
column show the sum over the 2 ⇥ 18 TPC sectors. The uni-directional versatile links [4] for the data to be sent to
the CRU and online farm are driven by bi-directional optical transceivers (VTRx) or uni-directional twin transmitters
(VTTx). Since the bandwidth needs in the up- and downstream directions are asymmetric, less VTRx are needed
to receive trigger and timing information via the ALICE TTS4 system. We aim at installing one TTS link on every
second FEC.
TPC Upgrade TDR 67

6.4 Common front-end ASIC

The readout of the detector signals is done by a 32 channel FE ASIC that is developed as a common
solution for different ALICE sub-detectors. The concept assumes the integration of low-noise analog
components and continuously operating, digital functionality on the same silicon die.
The SAMPA project at the University of São Paulo in Brazil targets the design, simulation, validation and
production of a signal acquisition and digital processing ASIC based on TSMC 0.13 µm mixed signal
technology [5]. This ASIC will comply with the requirements defined by the upgrade of the TPC, as well
as the ALICE Muon tracking detector.

6.4.1 Overview
A schematic of the SAMPA is shown in Fig. 6.4. The data fed into each of the 32 channels is processed
by a PreAmplifier/ShAper circuit (we reuse the name PASA from the current system for this block), a
SAR10 ADC11 and a DSP. Before being read out, the data are temporarily buffered in an event memory
and multiplexed. The PASA and DSP have configurable parameters that can be accessed via a common
logic and interface unit.

SAMPA
pad 32 channels Bias

Cd Rf
pad
Cf 320Mbs
Cd
Shaper 10b Buffer Elink
pad Elink
ADC DSP Buffer

Cd + 10MSPS Buffer Elink

Control & Trigger
CSA Buffer Elink

VREF+ VREF- IOs

FEC Shaping time control Gain control

Figure 6.4: Schematic of the SAMPA ASIC for the GEM TPC readout, showing the main building blocks.

6.4.2 General requirements for the analog part

The requirements for the SAMPA are summarized in Tab. 6.2 and discussed in the following:

– The requirement on the signal-to-noise ratio (S:N) is taken over from the current system [1, 2]. In
order to reach the required detector resolution, a S:N ratio of 20:1 and 30:1 for MIPs12 is required
for the IROCs and OROCs, respectively13 . At the same gas gain it is larger in the OROCs due to
the longer pads, which collect more ionization due to the longer track length seen.

– If the noise level of the current system (670 electrons) is retained, the required S:N can be achieved
by applying an effective gain of 2000 in the GEM stack. The maximum pad and time bin for each
charge cluster of a MIP track in this case corresponds to a charge of typically 2.1 to 3.2 fC (1.3 to
2 ⇥ 104 electrons).

10 Successive Approximation Register (SAR)

11 Analog-to-Digital Converter (ADC)
12 Minimum-Ionizing Particle (MIP)
13 The S:N ratio is calculated using the maximum pad and time bin for each charge cluster.
68 The ALICE Collaboration

RUN 1 RUN 3
(measured) (requirement)
Signal polarity Pos Neg
Detector capacitance (range) (pF) 12 33.5 12 33.5
S:N ratio for MIPs (IROC) 14:1 20:1
(OROC 6⇥10 mm2 pads) 20:1 30:1
(OROC 6⇥15 mm2 pads) 28:1 30:1
MIP signal (fC) 1.5 – 314 2.1 – 3.2
System noise (at 18.5 pF, incl. ADC) 670 e 670 e
PASA conversion gain (at 18 pF) (mV/fC) 12.74 20 (30)
PASA return to baseline (ns) < 550 < 500
PASA average baseline value (mV) 100 100
PASA channel-to-channel baseline variation (s ) (mV) 18 18
PASA shaping order 4 4
PASA peaking time (ns) 160 160 (80)
PASA crosstalk < 0.1 %15 < 0.2 %
PASA integrated non-linearity 0.2 % <1%
ENC (PASA only, at 12 pF) 385 e 385 e
ADC voltage range (differential) (V) 2 2
ADC linear range (differential) (fC) 160 100 (67)
ADC number of bits 10 10
ADC sampling rate (MHz) 10 (2.5, 5, 20) 10 (20)
Power consumption (analog & digital) (mW/ch) 35 < 35

Table 6.2: Measured PASA and ALTRO parameters for the current system (RUN 1) and the requirements for the upgraded
front-end electronics (SAMPA parameters for RUN 3). The parameters are explained in the text.

– A dynamic range of the electronics of 100 fC allows the measurement of the ionization signals of
low momentum particles, which may produce signals 30 times larger than those of a MIP. With
respect to the current system the linear range is reduced for the benefit of a better resolution around
the threshold level.

– To minimize the quantization error16 , the conversion to digital values should take place with a
precision of at least 10 bits.

– The conversion gain (20 mV/fC) is chosen such that the maximum expected output signal matches
the amplifier voltage swing and the input dynamic range of the ADC (2 V peak-to-peak). In order
to approximately match the signal amplitudes in IROCs and OROCs, a second conversion gain
setting of 30 mV/fC can be used in the IROCs. In this case the linear range is decreased from
100 fC to 67 fC.

– The large number of front-end electronics channels and a requirement for an overall power con-
sumption < 20 kW gives a limit of 35 mW per channel. The heat is removed from the readout
modules with the existing water cooling system (see Sec. 11.4.3).

– Special care has to be taken to protect the system against potential corruption of data and control
registers caused by radiation (Single Event Effects).

– The electronics will be located in an area with limited access. High reliability is thus a requirement.

14 Forthe higher gain anticipated in the original TDR the value was 4.8 fC for the OROCs.
15 The requirement was < 0.3 %. p
16 The RMS value of the quantization error is 1/ 12 ⇡ 0.29 LSB. It becomes smaller with larger bit depth.
TPC Upgrade TDR 69

6.4.3 Signal shaping

The shaping parameters determine the size and shape of the individual charge clusters. The signal from
an ionizing particle on an individual readout pad spreads in time due to the longitudinal diffusion DL and
due to the track inclination angle l . The mean width of the charge clusters in z direction in units of time
is given by:

!
2 1 tan2 l Lpad
2
stime = 2 D2L zd + 2
+ sPASA , (6.2)
vd 12

where zd is the drift length for the given cluster and sPASA is the approximate sigma of the semi-gaussian
signal output of the PASA. It is related to the shaping time tFWHM like sPASA ⇡ tFWHM /2.4. For Ne-CO2 -
N2 (90-10-5), the first term of Eq. (6.2) varies between 135 ns for tracks at mid-rapidity (h = 0, long drift
length) and 175 ns for tracks at h ⇡ 0.9 (see Fig. 6.5). A value of sPASA = 80 ns (160 ns peaking time,
190 ns FWHM), slightly shorter than the signal width, had been chosen for the current TPC readout [1]
in order to ensure that the cluster shape in time direction is not dominated by the PASA response. For
the upgraded system these considerations remain the same. Therefore, the same shaping parameters as
found in the current readout system can again be chosen.

200
Cluster width (ns)

180

160
an
140 gu total
lar
eff
120 ec
t Ne-CO2-N2
sion
100 diffu

80

60

40
Ne-CF4
20

0
0 50 100 150 200 250
Drift length (cm)
Figure 6.5: Length of clusters in time units for the baseline gas mixture Ne-CO2 -N2 (90-10-5) and for Ne-CF4 (90-10), cal-
culated using Eq. (6.2) at a radius in the center of the OROC medium pad region and without the contribution of
sPASA .

6.4.4 Noise
The Equivalent Noise Charge (ENC) requirement of 670 e (mean system noise) is discussed in this sec-
tion. Such a noise performance is possible to achieve, as the noise distributions measured with the
current system and test data taken with the S-ALTRO demonstrator ASIC show (discussed in detail in
the following).

Noise on the current system

For the current TPC the ENC specification was < 1000 e at a total capacitance of 25 pF. The measured
noise distributions for all readout chambers and separately for the three different pad sizes (small pads on
IROC and medium and long pads on OROC) are shown in Fig. 6.6 (left panel). On the installed system
70 The ALICE Collaboration

5
10
counts All pads

RMS noise (LSB)

1.2
4 7.5 mm 2 pads
6 10 mm 2 pads
104 6 15 mm 2 pads 1

3 0.8
10

0.6
102
0.4
10
0.2

1
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 20 40 60 80 100 120 140 160
noise (LSB) trace length (mm)
Figure 6.6: (Left) Noise distribution for all FEE channels on all readout chambers and separately for the different pad sizes for
the current TPC [8]. (Right) Correlation of noise and the trace length on the padplane PCB for the medium-sized
pads (6 ⇥ 10 mm2 ). A straight-line fit describes the data well (compare to Tab. 6.3).

a noise of 0.71 LSB17 was achieved. At an average PASA gain of 12.74 mV/fC (at 18.5 pF average
detector capacitance), with 2 V linear range and with a 10 bit ADC this corresponds to an ENC value of
670 e. Only a small fraction of channels exceeds the specified requirement of a noise lower than 1000 e
(corresponding to ⇠ 1 LSB): 0.3 % for the IROC region, 0.2 % for the OROC medium pad region and
4.9 % for the OROC large pad region [8].

Noise dependence on capacitance

The right panel in Fig. 6.6 shows how the noise correlates with the additional capacitances given by the
varying trace lengths on the padplane PCB18 . A straight-line fit describes the data well and suggests a
capacitance of about 1.3 pF per cm of trace length. This correlation can be done for the different pad size
regions on the IROC (4⇥7.5 mm2 ) and OROC (6⇥10 mm2 and 6⇥15 mm2 ). The extracted noise features
are given in Tab. 6.3.
Table 6.4 shows a breakdown of the capacitance contributions from individual elements for the current
system. The values range from 12 to 33.5 pF. For the upgraded system, a reduction of these values will
be very challenging, but could be possible by using wider flexible cables with larger distances between
the traces, a thicker pad plane PCB, a larger distance between the pad plane PCB and the metallic ROC
body (thicker strong back), as well as by possibly further minimizing the trace lengths.

Noise contributions
The theoretical noise of the amplifier alone19 is about 300 e at zero capacitance. Also the ADC con-
tributes to the total noise with both input-referred noise20 and quantization noise21 . The total contri-
bution of the ADC is ⇠ 300 e. The measured total noise values between 570 e (laboratory) and 670 e
(current TPC system noise, see Fig. 6.5) also include all other effects like the finite input capacitance,
contributions from board design, power and ground and pickup of different kinds.
17 LeastSignificant Bit (LSB)
18 PrintedCircuit Board (PCB)
19 The main amplifier noise sources are the series thermal noise from the ESD protection circuit, the (flicker) noise from the

input transistor and the parallel thermal noise from the feedback transistor.
20 Input-referred noise consists mainly of the sampling noise due to resistors and “kT /C” noise.
21 Quantization noise is the difference between the actual analog value and the quantized digital value of an ideal ADC
TPC Upgrade TDR 71

IROC OROC System

Pad size (mm2 ) 4 ⇥ 7.5 6 ⇥ 10 6 ⇥ 15
Trace length (range) (mm) 4.6 – 114.7 5.6 – 113.9 5.4 – 146.8
(mean) 39.2 45.5 60.2 47.1
System noise (mean) (LSB) 0.67 0.7 0.78 0.71
(at zero tr. length) 0.589 0.603 0.624
(vs trace length) (LSB/mm) 2.1 · 10 3 2.2 · 10 3 2.3 · 10 3 2.2 · 10 3

PASA noise slope (simulated) (e/pF) 17 17

ENC (most probable) (e) 634 653 672 650
(PASA) 475 500 525 500
(ADC) 420 420
Mean capacitance (extracted) (pF) 17.4 18.8 20.2 18.5

Table 6.3: System parameters and ENC values for the current TPC readout system (MWPCs and PASA) for the three different
pad size regions [9, 10].

Capacitance (pF)
IROC OROC 6⇥10 OROC 6⇥15 Mean
Chip packaging 2 2
Traces on FEC 1–3 2
Connector on FEC 0.5 0.5
Flexible cables (long) 7 (12) 7 7 7
Connector on padplane 0.5 0.5
Traces on padplane [range] 5 [0.5 – 14] 6 [1 – 15] 8 [0.5 – 20] 6
Pad to Ground 0.2 0.4 0.6 0.5
Sum [range] 17.2 [12 – 28] 18.4 [12.5 – 29] 20.6 [12 – 33.5] 18.5

Table 6.4: Contributions and sum (mean value and range in square brackets) of capacitances for the current TPC readout (MW-
PCs and PASA) for the three regions with different pad sizes [9, 10].

Table 6.5 shows that simulations of the analog part of the SAMPA ASIC show the same noise perfor-
mance as those performed for the current PASA. It can thus be concluded that the noise performance on
the system level can be retained.

Noise considerations for mixed analog and digital ASIC designs

The two major differences between the SAMPA architecture foreseen for the readout of the GEM TPC
(see Fig. 6.4) and the existing system based on the two ASICs PASA and ALTRO are:

– The analog signals are being processed at the input of the chip while concurrently the previously
aquired data are processed and shipped off through the serial interfaces (continuous readout).

– Low-noise analog components and complex digital functions are implemented on the same silicon
die.

The feasibility of the second item was demonstrated by the Super-ALTRO (S-ALTRO) chip [11]. The S-
ALTRO is a 16 channel demonstrator ASIC that was designed with the readout of the ILC-TPC in mind.
On a small area (3.07 mm2 /channel) sensitive analog components and massive digital functionalities are
integrated on the same chip. The S-ALTRO chip showed that by using careful design techniques the
effect of the integration on the noise performance is small.
22 Measurements with first SAMPA prototypes will be carried out during 2014.
72 The ALICE Collaboration

Current TPC PCA16 S-ALTRO SAMPA

(PASA)
Simulated (w/o ADC) 244e (@0 pF) 220e (@0 pF) 260e (@0 pF)
385e (@12 pF) 320e (@10 pF)
480e (@18 pF) 480e (@18 pF)
Measured (w/o ADC) 385e (@12 pF) 270e (@10 pF)
(Laboratory) 570e (@12 pF) 550e (@0 pF)
(System) 670e (@18.5 pF)

Table 6.5: Simulated and measured ENC values for the current readout (PASA, 160 ns peaking time and 70 W resistance in the
ESD network are fixed parameters) and for PCA16 (at 120 ns peaking time with no resistance in the ESD network)
and S-ALTRO (at 120 ns peaking time and with 62 W resistance in the ESD network). A first simulation result for
the SAMPA ASIC is also added22 for the same parameters used in the current PASA (at 160 ns peaking time and
with 70 W resistance in the ESD network). Note also that often slightly different values of the input capacitance were
used in the results presented here. They are specified in the table.

Each S-ALTRO channel contains a low noise PASA, a pipeline ADC and a DSP unit. The 10 bit ADC
samples the output of the PASA at a frequency of up to 40 MHz before providing the digitized signal to
the DSP which performs baseline subtraction, signal conditioning and zero suppression. The PASA used
in the S-ALTRO is based on the 16 channel Programmable Charge Amplifier (PCA16) prototype [12].
It is programmable in terms of gain and peaking time and can operate with both positive and negative
polarities of input charge. With respect to the PCA16 ASIC the input protection has been increased by
adding a larger series resistor.
Table 6.5 gives an overview of simulated and measured ENC figures, comparing the performance of
the original TPC PASA to values from tests using the PCA16 and the S-ALTRO in different configu-
rations. System measurements with the PCA16 have resulted in ENC values on-detector23 of 0.52 LSB
and 0.55 LSB, as measured by the ILC-TPC collaboration [13] and with the ALICE TPC GEM IROC
prototype in 2012, respectively24 . With the S-ALTRO demonstrator chip an ENC value of 547 e was
measured in a laboratory setup25 [11]. This value was achieved with a chip with the input pins unbonded.
It includes all other noise sources, in particular pickup from the DSP, but also random noise from PASA
and ADC and the quantization error.
It can be concluded that the integration of low-noise analog components and complex digital functions on
the same silicon die is possible. The design experiences from the S-ALTRO demonstrator are transferred
to the design of the SAMPA.

6.4.5 Further requirements for the analog part

Some further important requirements are given in the following.

– The DC level of the PASAs differential output (baseline) should be kept at a value sufficiently close
to the bottom of the circuits dynamic range, in order to preserve the maximum dynamic range (see
Tab. 6.2).

– The baseline values have to be very stable in time in order to guarantee that the zero suppression
can work efficiently.

– The signals have to return to the signal baseline within < 500 ns without undershoot.
23 The settings used were 120 ns peaking time and 12 mV/fC conversion gain.
24 A conversion to ENC is difficult, since the exact conversion gain is unknown. At 12 mV/fC the measured values would
correspond to 520 e and 550 e, respectively.
25 The settings used were 120 ns peaking time, 12 mV/fC conversion gain and 1 V bias decay voltage. The CDM input pad

was connected.
TPC Upgrade TDR 73

– Synchronous pulsing of all 32 channels must be possible without loss of resolution.

– Crosstalk between adjacent channels can deteriorate the energy and position resolution and, there-
fore, has to be kept below 0.2 %.

6.4.6 Electrostatic discharge protection

The input of the SAMPA may be exposed to ESD26 events. Semiconductor devices usually offer device-
level ESD protection based on the charge device model (CDM) or the human body model (HBM). How-
ever, device-level ESD specifications may not be sufficient to protect the SAMPA when connected to a
readout detector. Some experience with destructive discharges exists in the collaboration. The current
TPC readout system was suffering from such events in the first two years of operation at the LHC. The
destructive events could be traced to the energy stored in capacitors used in the HV network (4.7 nF). In
case of a discharge inside the MWPC the charge stored in these capacitors (6 µC) would be released and
injected through the readout pad into a FE channel, making it insensitive to further signals. In certain
cases, presumably for large discharge events, a low impedance path was produced between the power
(Vdd) and ground planes of one amplifier. As a consequence, the voltage regulator on the corresponding
FEC could not supply the nominal voltage any longer, making the whole FEC unusable. After removing
the HV capacitors in 2011 the amount of stored energy was reduced by a factor 10. After this intervention
no more destructive events have been observed.
Models for discharges in gaseous detectors are not available. However, laboratory tests have been carried
out using a capacitor connected to a needle with micrometric adjustment. The needle was lowered onto
pads on a test board connected to the inputs of a PASA chip. The capacitor was charged to different
voltages by a high voltage generator in order to simulate discharge events of different severity. Such
discharges are probably different from those in a gaseous detector, but help to understand the limitations
of the ESD protection. It was found that a charge of a few hundred nC is sufficient to destroy the ESD
network at the input of a PASA channel. Larger charges of a few µC may create a short between Vdd
and ground. By inserting resistors before the input to the PASA, the tolerated charge could be increased
to the required level. However, quite large resistors had to be used (several hundred W).
Problems with destructive discharges have appeared in a similar manner also during prototype GEM
tests both in the ALICE test at the PS and with ILC-TPC prototype tests. In GEM detectors discharges
generally begin with a sudden, radiation-induced breakdown of the gas rigidity in one GEM, normally
the last in a cascade of multipliers. The capacitance of one GEM segment is slightly smaller than 5 nF
and the charge stored is 1.5 µC. The discharge may propagate to other electrodes such that a fraction (up
to all) of the stored charge can be delivered to the readout pad.
Additional resistances and capacitances (protection diodes) at the input to the SAMPA will have a neg-
ative impact on the noise performance. Therefore, the best strategy is to implement device-level ESD
specifications based on the HBM and add system-level protection on the front-end card level after addi-
tional testing to derive a realistic estimation of the expected ESD energy, wuth keeping the best possible
noise performance in mind.

6.4.7 Analog-to-digital conversion

The analog signal output of the PASA is sampled, at a rate of 10 MHz, by an ADC with 10-bit dynamic
range. The requirement on the power consumption (below 35 mW per front-end channel) calls for low
power ADCs. This requirement can be met by SAR ADCs implemented with a low power switching
technique. They allow implementations with a low power consumption around 2 mW per channel.
26 ElectroStatic Discharge (ESD)
74 The ALICE Collaboration

Sampling frequency
A shaper peaking time of 160 ns preserves partially the spatial characteristics of the drifting electron
clusters (see Sec. 6.4.3). In RUN 1 it was found that this feature can be best taken advantage of in
the clusterization process when a sampling frequency of 10 MHz is chosen. Consequently, also for the
GEM TPC readout this sampling frequency represents the optimum. The maximum electron drift time
of ⇠ 100 µs is in this case divided into about 1000 time bins, each corresponding to a drift distance of
about 2.8 mm.

6.4.8 Digital signal processor

The digitized TPC signals are further processed by a set of digital circuits integrated into the DSP part
of the SAMPA, which are described below.

Pedestal subtraction
The DC level of the PASAs differential output (baseline or pedestal) has to be subtracted from the ADC
output for each channel in order to perform efficient zero suppression. A fixed pedestal subtraction mode,
where one value can be subtracted for each channel for all time bins, has shown to be sufficient for the
current system. The baseline values are actually extremely stable over time, requiring only very rarely
pedestal calibration runs; a feature that should be kept.

Digital filter
Since the GEM signals do not feature a tail, the necessity to minimize the pileup effect of subsequent
signal tails by a tail cancellation filter is removed.
A calculation and subtraction of the moving average may help to correct possible perturbations of the
baseline produced by non-systematic effects. The value to be subtracted from the current samples is
the moving average of the signal itself and of some previous samples. Fast variations in the signal, like
clusters, are excluded from the baseline calculation. Such a filter is implemented in the ALTRO ASIC
but could not yet be successfully used during data taking. The reason lies in the fact that in rare cases
the calculation of the value to be subtracted may fail, resulting in a pedestal shift and failure of the zero
suppression.

Data compression using zero suppression

If no data compression were applied in the front-end electronics, the 552,960 readout channels of the
GEM TPC would produce data at a rate of 7 TByte/s (100 Mbit/s per channel). To reduce this throughput,
it is foreseen to compress the data via zero suppression (ZS). In this mode all data words are dropped
inside the SAMPA if their value is so close to the reference level (baseline or pedestal) that they can be
assumed to not contain any useful information. In practice, a fixed threshold is applied for the data in
each channel, and only values above this threshold are kept. The resulting signal clusters are run-length
encoded, which requires the addition of two extra words, describing the start time and length of the
non-suppressed cluster.
To remove glitches, the minimum number of consecutive samples needed to define a valid signal can
be configured27 . If needed, a number of pre- and post-samples that should be kept before and after a
peak can be defined28 . In order to perform the zero suppression, the pedestal is subtracted. The zero
suppression threshold is naturally determined as a multiple of the width of the baseline distribution
(noise). Pedestal and noise values are measured in a dedicated pedestal run and the measured values for
27 A minimum of 2 consecutive time bins above threshold is required in the current readout system.
28 This feature is not used in the current readout system.
TPC Upgrade TDR 75

each channel are then used to configure the filter. The threshold must be configurable for each individual
front-end channel.
This technique is a lossy technique; small clusters or tails of clusters may be discarded. It achieves
compression ratios which are roughly inversely proportional to the TPC occupancy. For occupancies of
15 % and 42 % the data compression ratio is about 0.18 and 0.5, respectively. Table 6.6 compares the
compression ratio with a different running scenario (see below).

minimum bias central

Average occupancy 0.15 0.24
Compressed data size (ZS) 0.18 0.28
Compressed data size (Huffman) 0.21 0.25

Table 6.6: Expected average pad occupancies and approximate data compression ratios (compressed data size relative to raw
data size) for two different data compression modes.

Other compression scenarios

Instead of (or in addition to) zero suppression, lossless transformations like variable length codes (e.g.
Huffman coding) or lossy compression methods like vector quantization may be applied to compress the
data.
Small ADC values occur very often in the data stream of a gaseous detector like the TPC, but larger
ones are more rare. The distribution is approximately exponential. The expected size of the data can be
reduced if short words are used for frequent values and longer ones for rare values. The theoretical lower
bound on the average word size that can be achieved by this strategy is called the entropy of the data
source29 . Huffman coding [14] approaches this bound and is easy to implement. Only if the distribution
of input values is too extreme the method can run into problems. In such a case events can become larger
than uncompressed.
A TPC is obviously not a stochastic data source, as the ADC values found in adjacent pads are highly
correlated. Therefore, it could be possible to compress the data to an even lower bit rate than the entropy
of the ADC values. In other words, there could be representations of the TPC data that have a lower
entropy than the formats described above. Various methods like differentiation, prediction, etc. have
been evaluated, but in the end none of these approaches yield results that are much better than plain
Huffman coding (see [1] and references therein).
Huffman coding has been applied to TPC events before. The entropy analysis results in about 6 bits per
sample for cluster data, depending somewhat on the abundance of overlapping clusters due to high oc-
cupancy. A noise of 0.7 ADC counts requires about 1.3 bits per sample for coding the baseline/pedestal.
Applying Huffman coding to non-zero-suppressed data will result in a compression ratio of about 0.3 for
an occupancy of 27 %, which is expected for the inner pad rows.
Huffman coding can be implemented in the SAMPA for encoding and in an FPGA30 on the online farm
side for decoding. It will give comparable or even better compression factors than ZS for occupancies
larger than 25 % (see Tab. 6.6), while ZS reduces the event size by a large margin at low occupancies.
Huffman coding is lossless, while ZS implies typically a loss of 15 % of the total charge of a MIP cluster.
Both methods require similar resources for implementation and suffer from the same drawbacks, i.e. that
the data volume might increase in case of noisy channels (by a larger factor in case of Huffman coding)
29 The entropy of the data source can be computed as the sum over the probability of the occurrence of an ADC value
multiplied by the logarithm of the probability; the sum running over the set of all possible words that are output by the data
source.
30 Field Programmable Gate Array (FPGA)
76 The ALICE Collaboration

and that the output format is more prone to errors induced e.g. by SEUs31 than the raw ADC data. The
first problem has to be taken care of by the output module in the SAMPA; the latter can be solved by
applying error detection and correction techniques to the transmission protocol.

Event memory
The data read out from 32 channels is transferred into the SAMPA event memory. Before transferring
the data from each channel off-chip, it is temporarily buffered here. Naturally occurring bursts with
higher occupancy lead to temporarily increased data rates, which have to be smoothed properly by the
event memory. The event memory in this case serves as a de-randomizing buffer, ensuring that the data
throughput is always below the bandwidth of the serial output links from the SAMPA. The optimal size
of this buffer has to be found in simulations. In the undesirable case where the data rate is too high for
an extended period, the data will have to be truncated, which will be marked in the data stream.
The event memory can also be used as a lookup table to generate patterns to be read out through the
output links for test purposes.

Interfaces
The data are read from the event memory and transferred to four electrical links (e-links) with bandwidth
320 Mbit/s each.
For control and monitoring of the SAMPA the GBT-SCA32 , a dedicated ASIC implemented in CMOS
130 nm using radiation tolerant techniques is used. It implements the I2 C protocol33 towards the SAMPA
chips and is controlled via a dedicated slow control e-link from a GBTx chip.

6.4.9 Testing
The ASICs have to be thoroughly tested before assembly on the front-end cards. Even minimal mechan-
ical damage may lead to loss of assembly quality. Due to the large number of chips to be tested, an
automatized testing bench is required in order to guarantee reliable sorting and book-keeping.
In the manufacturing process for the original FECs for the ALICE TPC about 40,000 PASA and ALTRO
chips each were tested at Lund University. A semi-automatic test system had been developed, with the
test procedures described in [2] and on the web page [15]. A pick-and-place robot moved chips from the
trays (see Fig. 6.7) and placed them in the test socket which opened and closed by compressed air control.
The test boards with the socket carried buffer memories to allow the communication and operation of
the chip under test to go on at nominal speed. Extensive testing takes quite long time per chip. It was
not judged meaningful to automatize the exchange of chip trays as the robot could be loaded for about 4
hours of testing.
Each chip was tested for about 1 minute (the ALTRO slightly longer, the PASA slightly shorter). All
supply currents were measured, all DC levels checked. The ALTRO channels were excited by a sinus
waveform which was digitized and verified, thus checking all bits of the analog-to-digital conversion. All
memory cells were checked by downloading test patterns and reading back. The PASA channels were
excited by a step-voltage, feeding charge into the preamp input over a small capacitor. The waveform,
sampled with a 12 bit ADC at 40 MHz was compared to the expectation and the amplifier gain was
recorded. Chips were sorted in three categories depending on the test result: accepted, slightly outside
the acceptance limits and failure. A test protocol for each chip was stored and a summary of the test
results was published [9].
31 Singe Event Upset (SEU)
32 Slow Control Architecture (SCA)
33 Inter-Integrated Circuit (I2 C)
TPC Upgrade TDR 77

Figure 6.7: Image of the PASA and ALTRO testing robot at Lund University.

All equipment and software for testing is still available and can be used for the new SAMPA by modifying
the test card with a different socket, different output buffering and some other minor changes. The analog
test pulsing feature can remain unchanged, but the procedures for ALTRO and PASA testing will have to
be merged.

6.5 Front-end card

The front-end card (FEC) contains the complete readout chain for amplifying, shaping, digitizing, pro-
cessing, and buffering the TPC signals. The FEC must be designed as an integral part of the detector to
obtain the best performance for the experiment.

6.5.1 Partitioning
The FEC plays a crucial role in the optimization of the detector performance. Some boundary conditions
are given in the following:

1. Available space: the number of pads on the IROCs of the current TPC is as high as can be read
out with the present channel density of the electronics (16 channels per ASIC). In order to manage
this, the mechanical mounting of the FECs, as well as their cooling, is the result of a large engi-
neering and construction effort. The resulting density of hardware has led to many difficulties in
maintenance operations. However, for the OROCs the space situation is much more relaxed.

2. Reuse of mechanical structures: with a new FEC with equal size one may take advantage of
reusing the mechanical support and cooling structures.

3. Pad plane segmentation: due to the size of the OROC, the pad plane PCB will have to be seg-
mented. Since traces can not cross PCB segments, not all configurations for the traces are possible.
78 The ALICE Collaboration

4. Trace lengths: higher intergation of the readout system (more channels per FEC) could lead to
longer trace lengths on the pad plane PCB with a negative impact on the noise performance.

5. Flexible signal cables: the trace density on the Kapton signal cables connecting to the readout
chambers also directly relates to the noise behavior.

6. Readout link: the GBTx ASIC [3] can multiplex 10 (14) e-links with (without) forward error
correction enabled. This number together with the number of output e-links per SAMPA (4) makes
only certain combinations attractive. Not using all inpus to the GBTx ASIC effectively means a
waste of useful bandwidth.

7. Readout scheme: when the data are shipped off-detector, they are ordered in such a way that the
information from all pads with hits are sent in consecutive order following the pad row structure
and directed to the same cluster finder instance (see Sec. 8.1.2).

The 6th requirement basically defined the number of SAMPA chips per FEC to be 5 (or 7). Such a scheme
with 5 readout partitions per TPC sector is summarized in Tab. 6.1. The last requirement (data ordering)
can always be achieved in the CRU, which gathers the data from many FECs.

6.5.2 PCB design and layout

A simplified layout of the FEC is shown in Fig. 6.8. The FEC receives the analog signals from the
readout chamber through a set of flexible cables. To minimize the trace lengths and thus noise (but
also crosstalk), the SAMPAs have to be positioned very close to the input connectors. The signals are
processed by 5 SAMPAs, with an estimated maximum power consumption of less than about 1 W per
SAMPA. The FEC also contains additional protection for the input to all SAMPA channels (see next
section).
SAMPA

SAMPA

SAMPA

SAMPA

SAMPA
GBTX

GBTX

SCA
GBLD
GBLD

GBTIA
LD

PD

LD

Figure 6.8: Schematic of the front-end card (FEC) with SAMPA. The connectivity is realized through the GBT system via
optical links.

Control and monitoring of the SAMPA is implemented via the GBT-SCA ASIC. Physics and monitoring
data are multiplexed on the FEC by up to two GBTx ASICs. The uni-directional versatile links [4] for
the data to be sent to the CRU and online farm are driven by bi-directional optical transceivers (VTRx) or
TPC Upgrade TDR 79

uni-directional twin transmitters (VTTx). Since the bandwidth requirements in the up- and downstream
directions are asymmetric, less VTRx are needed to receive trigger and timing information via the ALICE
TTS34 system. We aim at installing one TTS link on every second FEC. Also configuration data and
control commands to the FECs are sent via the TTS links. A possible distribution of optical components
and links on the 36 TPC sectors is shown in Tab. 6.1, where we consider that less bandwidth is needed
for the outermost readout partitions due to lower occupancies in this region.

6.5.3 System level input protection

The protection strategy foresees the implementation of system-level protection on the front-end card in
the form of diode pairs, on top of the device-level protection in the SAMPA. Within the SRS project
and the RD 51 collaboration a hybrid readout board for MPGD detectors has been developed. The
module (RD51 APV25 HYBRID) is based on the APV25 chip and successfully uses an IC35 with diode
protection for 4 channels.
Broken channels may lead to a short to ground which in fact makes the whole FEC inoperable. This
could be avoided by separate voltage regulation for the analog power of each SAMPA (or possibly for
each pair of SAMPAs).

6.5.4 Testing
The assembled FECs for the current TPC readout system were tested on the bench at Frankfurt University.
This test step benefits from a final readout system being operational, which allows parallel testing of
many FECs as well as long term testing (stress tests). The purpose of the test is to find assembly errors,
chips which have broken during the mounting and handling, as well as any other malfunctioning of any
component. The test procedures have been described in [2] and on the web page [15] and are a good
starting point for the testing of the future FECs.

6.5.5 Irradiation campaign

The radiation levels expected for the high luminosity phase (50 kHz Pb–Pb collision rate) impose re-
quirements on the radiation hardness/tolerance of the innermost FECs of the ALICE TPC. Fast hadrons
can lead to a Single Event Upset (SEU) in digital structures, or to a Single Event Latch-up (SEL), a
type of short circuit that triggers parasitic effects which can disrupt proper functioning of the element, or
possibly even lead to its destruction.
The flux of fast hadrons (> 20 MeV) at the TPC inner (outer) layer is expected to reach 3.4 kHz/cm2
(0.7 kHz/cm2 ), including a safety factor of 2 [5]. The TPC electronics located at the inner radius of the
service support wheel has to stand a dose of 2.1 krad.
The SAMPA has to be designed carefully in order to avoid SELs and to protect critical logic against
SEUs. FEC prototypes and all components will be irradiated in test beams of fast protons or neutrons,
either individually in dedicated campaigns for critical components (e.g. FE ASIC) and/or finally after
being mounted on FEC prototypes.

6.6 Common Readout Unit

The Common Readout Unit (CRU) is described in [5]. It provides the interface between the on-detector
electronics and the trigger system, the online farm, and the DCS36 . The CRU units are located in the
34 Triggerand Timing distribution System (TTS)
35 NUP4114UPXV6T1G from ON semiconductor.
36 Detector Control System (DCS)
80 The ALICE Collaboration

control room outside the radiation area and will receive data from the detectors through optical fibers via
the GBT link. A schematic is shown in Fig. 6.9.

LTU

timing and trigger

distribution (TTS)

control and
configuration

DCS
monitoring data
VTTx Physics &
GBTx monitoring data detector
GBT
GBT data links
GBT (2x) CRU
VTRx Trigger, control (DDL3)
SAMPA (5x) and configuration
32 channels online
front-end Physics data
farm
160 input GBT-SCA
FEC links (GBT)
channels

Figure 6.9: Schematic of the TPC readout system with the CRU as central part interfacing the front-end electronics to the trigger
system, the DCS and the online farm.

The CRU steers and controls the configuration, readout and monitoring of the FEE and the trigger han-
dling. When the TPC data are forwarded from the CRU to the online system, the individual data frag-
ments are ordered by the geometrical position of the pad rows and pads on which the charge cluster
signals were read such that the data are sent consecutively pad-by-pad, pad-row-by-pad-row. Also a
cluster finder algorithm could be implemented on the CRU.
On each FEC the data from 5 SAMPAs are multiplexed by 2 GBTx ASICs [3] (only 1 for the outermost
readout partitions due to lower occupancies in this region) and sent to the CRU and online farm through
a double optical transmitter. The effective bandwidth for data readout is 2 ⇥ 3.2 Gbit/s using a special,
SEU robust transport protocol with forward error correction. The trigger, timing and clock distribution
and control and configuration data are transfered to the FEE on each TPC sector using fewer (e.g. one
per 2 FECs) uni-directional GBT links (TTS [5]).
The CRU is the physical and logical interface to the ALICE online farm, to the DCS and to the trigger
system. Different firmware modules are required, that implement these interfaces:

– The trigger module receives trigger and timing information, forwards a clock and a synchronisation
(or a trigger) signal to the FECs and prepares a timestamp header.

– The data acquisition module receives the data from the FECs and checks its integrity. This module
will need to be a TPC specific implementation due to special requirements for the continuous
readout of data.

– The TPC specific data formatter re-orders the channel data, adds header information into the data
stream and sends the data to the firmware module servicing the optical uplink.

– A TPC specific DCS module may perform monitoring (voltages, currents and temperatures) and
time critical operations like switching off FECs in case of overheating.

The configuration, control and monitoring of the TPC FEE (SAMPA chips, FECs and CRUs) is integrated
into the existing hierarchical control structure of the ALICE DCS. A software FeeServer37 at the CRU
configures the front-end electronics and monitors its status (see Sec. 10.2).

37 Front-End Server (FeeServer)

Chapter 7

Simulation and detector performance

The expected performance of the upgraded GEM TPC during RUN 3 is described in this chapter. It
starts with an overview on the performance of the current TPC, followed by a discussion of the intrinsic
resolution of the GEM TPC (Sec. 7.2) and of the performance with track pileup from different interac-
tions (Sec. 7.3). In Sec. 7.4 distortions due to space charge are discussed. In particular, the importance
of space-charge fluctuations is pointed out. Finally, in Sec. 7.5 the performance including also space-
charge distortions of the expected magnitude, and corrected with the online calibration tools discussed
in Chap. 8, is presented.

7.1 Current performance

The current ALICE TPC was installed in the ALICE experimental cavern in 2007 and successfully
operated since the LHC start-up in 2009. The observed performance is in agreement with or better than
the design specifications reported in [1, 2].

7.1.1 Tracking performance

In pp collisions the particle tracking efficiency within the TPC is better than 98 % for findable tracks1 .
In Pb–Pb collisions it decreases by about 1 – 3 % due to the higher occupancy [3] (dNch /dh ⇡ 1600 in
p
0 – 5 % central collisions at sNN = 2.76 TeV).
The achievable momentum resolution is driven by multiple scattering at low pT , and by the space-point
resolution and residual mis-calibration after all distortion corrections at high pT . Both contributions are
discussed in the following.

Space-point resolution The parameters dominating the space-point resolution of the TPC are diffu-
sion (which depends on drift length zd ), track inclination angle and total cluster charge on a given pad
row. For tracks with large transverse momentum pT , and thus vanishing deflection angle, the space-
point resolution in local-y (rj) direction2 is around 400 µm (800 µm) for short (long) drift length [4].
The current results are at the level of the intrinsic (statistical) limit. From the covariance matrix of the
tracks, the track resolution at different reference points can be obtained. For high-pT tracks the intrinsic
track resolution is approximately sintr = 200 µm at the entrance of the TPC inner field cage and 2 mm
extrapolated to the interaction point. This intrinsic track resolution determines the precision that is re-
quired for all calibrations and drift field distortion corrections, e.g. the drift velocity, alignment, E ⇥ B
and space-charge distortion corrections.
1 Findable tracks are defined by a minimum track length in the active zones of the TPC.
2 The space-point resolution in rj determines the transverse momentum (pT ) resolution. For the coordinate system see
App. A.

81
82 The ALICE Collaboration

Drift field distortion corrections For RUN 1 the largest systematic correction applied to the space-
point coordinates is due to the E ⇥ B effect, which reaches 1 cm (in very localized regions). The precision
of this correction is limited by the residual misalignment of the electric and magnetic fields.
So far, space-charge effects due to ions in the TPC drift volume have been small. In the 2011 Pb–Pb data
a space-charge induced track distortion of ⇠ 200 µm (when extrapolated to the interaction point) was
observed.

Momentum resolution Figure 7.1 shows the resolution s1/pT for the current status of the calibration
p
and for a data set from p–Pb collisions at sNN = 5.02 TeV. Expressing the momentum resolution using
s1/pT has the advantage that the distributions based on this variable are approximately Gaussian, even
at large pT . The performance shown in Fig. 7.1 translates into a pT resolution of s pT /pT . 3.5 % at
pT = 50 GeV/c and below 1 % at pT = 1 GeV/c.
(GeV/c)-1

0.016 p-Pb, sNN = 5.02 TeV, | |<0.8

TPC standalone tracks
0.014 TPC tracks constrained to vertex
T
1/p

TPC+ITS combined tracks

0.012 TPC+ITS constrained to vertex 18/11/2013

0.01

0.008

0.006

0.004

0.002

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
-1
1/pT (GeV/c)

Figure 7.1: Resolution in 1/pT as a function of 1/pT for the ALICE central barrel (current system). The plot shows the 1/pT
resolution for TPC standalone tracks and for global tracks combining tracking in ITS and TPC with and without
vertex constraint. The data is from p–Pb collisions collected in 2013.

7.1.2 Particle identification performance

The specific energy loss dE/dx is obtained for each track by calculating the truncated mean of the cluster
charges on currently up to 159 pad rows3 , which adds significant PID4 capabilities to the ALICE central
barrel. Identification of pions, kaons, and protons is possible also in the relativistic rise region of the
Bethe-Bloch energy-loss curve [5] at high pT . The pT reach is currently limited to 20 GeV/c by statistics.
With the event sample anticipated in RUN 2 and further improved understanding of the characteristics of
the energy-loss curve in the relativistic rise region, the identification of pions, kaons and protons up to
50 GeV/c is within reach. The usage of high-momentum cosmic muons to improve the understanding of
the energy loss in this pT region is under study.
In pp and central Pb–Pb collisions the dE/dx resolution is about 5.5 % and 7 %, respectively. The res-
olution in pp is consistent with the intrinsic dE/dx resolution dominated by the ionization fluctuations
(Landau tail), folded with gas gain fluctuations, the detector readout granularity and threshold effects. In
3 In the GEM TPC the number of pad rows is reduced to 158, see Sec. 4.5.
4 Particle IDentification (PID)
TPC Upgrade TDR 83

Pb–Pb collisions the performance deteriorates at large multiplicities mainly due to an increasing overlap
of clusters.

7.2 Intrinsic performance of the upgraded TPC

The expected performance of the TPC after upgrade with GEM-based readout is quantified using micro-
scopic simulations. For these studies, the current ITS design is used. Since the performance will improve
with the upgraded ITS [6, 7], the results on the global tracking performance reported in this chapter can
be considered as conservative.

7.2.1 Microscopic GEM simulations

The most important parameters for the microscopic simulation of the intrinsic performance are:

– Gas choice: The baseline gas mixture is Ne-CO2 -N2 (90-10-5), see Chap. 3.

– Pad geometry: The baseline pad geometry of the upgraded TPC is very close to that of the current
system, see Sec. 4.5.

– Pad response function: The shape of the induced charge distribution on the readout plane of a
GEM detector is very narrow, due to the parallel plate geometry of the induction gap. A purely
projective pad response function is used in the simulation, i.e. electrons create signals only on the
pad directly below their arrival point on the first GEM plane.

Table 7.1 shows the parameters used for the simulations. The microscopic simulation algorithm is de-
scribed in [1] and has been validated with test beam data [8]. Here, only a short summary of the simula-
tion steps is given.

Item Value
Mean number of primary electrons for MIP 14.0
First ionization potential I0 (eV) 20.77
Effective energy for e-ion pair creation Wi (eV) 37.3
Drift velocity at 400 V/cm (cm/µs)
p 2.58
Longitudinal diffusion constant (DL ) (µm/pcm) 221
Transverse diffusion constant (DT ) (µm/ cm) 209
wt (for magnetic field B = 0.5 T) 0.32
Oxygen content (ppm) 5
Effective GEM gain 2000
System noise (RMS) (e) 670
Dynamic range 2 V; 10 bits
Conversion gain (mV/fC) 20
Peaking time (ns) 160
Sampling frequency (MHz) 10
Zero suppression settings 3 LSB5 threshold; glitch filter

Table 7.1: Parameters used in the simulation: Gas properties, GEM effective gain (which includes the signal coupling to the
readout plane) and electronics parameters.

The simulation is based on a modified version of GEANT3 [1]. The ionization by charged particles
entering the active gas volume is simulated using a TPC-specific energy-loss method. Based on the
average number of primary electrons per cm of track length, the interaction length is calculated and a step
5 Least Significant Bit (LSB)
84 The ALICE Collaboration

length is randomly generated. A random energy-loss value is assigned to this step according to 1/E 2.2
(I0  E  10 keV, where I0 is the first ionization potential). The total number of ionization electrons
including secondaries is calculated. Each of these electrons is drifted to the readout chambers, taking
into account diffusion and residual distortions. Space-charge distortions are not taken into account at this
point; they are discussed separately in Secs. 7.4 and 7.5. At the GEM readout stack an exponential gain
variation is assumed for each electron. The charge is projected onto the pad plane using the described
pad response function in space and a semi-Gaussian shaping function in time. The reconstruction of
events simulated in this way does not differ from that of real data.
The described simulation algorithm is routinely used in the ALICE simulation and reconstruction frame-
work and was thoroughly validated in the past years.

7.2.2 Tracking performance

Figure 7.2 shows that the momentum resolution in the acceptance of the central barrel detectors of ALICE
will be maintained for the TPC with GEM-based readout. The plot indicates a slight deterioration of the
resolution in 1/pT for tracks using only TPC information. This deterioration is however fully recovered
for global tracks combining tracking in ITS and TPC. The good agreement of the simulated data shiown
in Fig. 7.2 (left panel) with the measured p–Pb data shown in Fig. 7.1 indicates a thorough understanding
of material budget and detector response.
-1
-1

(GeV/c)
(GeV/c)

0.014 0.014
MWPC GEM

TPC only tracks 0.012 TPC only tracks

0.012
TPC constrained tracks
1/pT

TPC constrained tracks

1/p T

TPC+ITS combined tracks TPC+ITS combined tracks

0.01 0.01

0.008 0.008

0.006 0.006

0.004 0.004

0.002 0.002

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
1/pT (GeV/c)-1 1/pT (GeV/c)-1

Figure 7.2: Resolution in 1/pT as a function of 1/pT for MWPC (left panel) and GEM readout (right panel). The open red
squares are for tracks based on TPC information only, while the closed blue squares show TPC track fits including
the vertex point. The open black squares show the result for combined fits to TPC and ITS track points.

The deterioration of the resolution for TPC-only tracks is consistent with the slightly worse position
resolution due to the narrower pad response function for the GEM-based readout chambers as compared
to the present MWPCs (see Sec. 4.5). However, in most of the acceptance region |h| < 0.9 clusters with
signals on more than one pad dominate, as shown in Fig. 4.16.

7.2.3 Particle identification performance

The relative dE/dx resolution sdE/dx /hdE/dxi for isolated electron and pion tracks (1 < p < 6 GeV/c)
was measured with a triple-GEM IROC prototype at the CERN PS test beam. The results reported in
Sec. 5.2.6 show that the dE/dx resolution observed with the GEM IROC is compatible with that of the
MWPC IROCs. This behavior is confirmed in the simulation: Figure 7.3 (left panel) shows the dE/dx
resolution as a function of the momentum for MWPC and GEM readout at low track multiplicity. Only
a small difference between MWPC and GEM is observed.
TPC Upgrade TDR 85

0.1
0.08

/dEdx

dEdx /dEdx
GEM 0.075 Pb-Pb
0.09
dEdx
MWPC pp
0.07
0.08
0.065

0.07 0.06

0.055
0.06

0.05
0.05
0.045

0.04 0.04
2 4 6 8 10 12 14 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
p (GeV/c) Electron transmission efficiency
Figure 7.3: (Left) dE/dx resolution as a function of track momentum for MWPC and GEM readout. The simulation is carried
out at low multiplicity (pp). The performance at high multiplicities and with event pileup is discussed in Sec. 7.3.
(Right) Simulated dE/dx resolution for MIP tracks crossing 158 pad rows as a function of the electron transmission
efficiency in the GEM stack for pp and Pb–Pb collisions.

An important effect to be considered for a GEM readout system is the transmission efficiency of the
readout stack for primary electrons. While for an MWPC a transmission efficiency of 100 % can be
safely assumed, for a GEM stack it may be reduced due to the finite electron collection efficiency at
the first GEM stage. As shown in Sec. 5.1.3 (e.g. Fig. 5.9), this effect is enhanced for settings which
minimize the ion backflow. In measurements with an 55 Fe source it manifests itself in a degradation of
the energy resolution at 5.9 keV relative to the optimal value of about s (55 Fe) = 8.5 %.
As shown in Fig. 7.3 (right), the dE/dx resolution, which depends on the energy resolution, degrades
significantly only for values of the electron transmission efficiency below 0.5, both for pp and Pb–Pb
collisions. If we consider the first GEM foil the main contributor to the energy resolution, aptransmission
efficiency of 0.5 corresponds to a degradation of the energy resolution to s (55 Fe) = 8.5 %/ 0.5 ⇡ 12 %.
Due to the results shown in Fig. 7.3 (right), we consider a transmission efficiency of 0.5 the lower limit
(and an energy resolution of 12 % the upper limit) for the operation of the GEM readout system. In the
simulations shown in the rest if this chapter the additional effect of the finite transmission efficiency is
not included.

7.3 Performance with event pileup

The GEM TPC will operate at a Pb–Pb interaction rate of 50 kHz, where particle tracks from Npileup = 5
events on average are superimposed in the drift volume of the TPC. The resulting pad occupancies lead
to an increased probability for clusters to overlap and thus to be wrongly assigned during the tracking
step. In this section we study the impact of event pileup on the tracking performance. For more details
on the LHC running conditions see Sec. 8.1.1.
The current TPC was designed for a charged particle multiplicity of dNch /dh = 8000 in central Pb–Pb
collisions. Comprehensive studies showed that the required performance can be achieved at such extreme
conditions [1]. For the anticipated data taking scenario at Rint = 50 kHz, corresponding to an equivalent
charged-particle multiplicity (see Sec. 6.2) of dNch /dh|equiv ⇡ 2500, the effect of event pileup is thus
expected to be minor. However, the track topology in overlapping events is different from a single high-
multiplicity collision due to the displacement of the vertex positions. The performance of the TPC under
such conditions was verified in simulated central (0 – 5 %) Pb–Pb events (hdNch /dhi scaled to 2000)
embedded in different pileup scenarios. A varying number of minimum bias events (hdNch /dhi scaled
to 500) are added at random distances within a time window of ±80 µs. Several jets are added to the
86 The ALICE Collaboration

central event (simulated with Pythia) in order to enhance the fraction of high-pT particles. Naturally, in
this simulation the actual tracks seen from the pileup events are constrained by the readout time window
defined by the central event.
Figure 7.4 shows that the TPC standalone tracking efficiency at the different interaction rates is essen-
tially the same for the current TPC and for the GEM TPC. In these data, the lower bound scenario (central
event at Rint = 20 kHz) corresponds to an equivalent multiplicity dNch /dh|equiv = 2500, while the most
extreme scenario (central event at Rint = 70 kHz) corresponds to dNch /dh|equiv = 5000. Figure 7.5 shows
the resolution in 1/pT for the GEM TPC for these different scenarios. The tracking performance is
essentially unaffected by the additional track load.

1
TPC tracking efficiency

TPC tracking efficiency

1

0.995 0.995

0.99 0.99

0.985 0.985

central event at 20kHz central event at 20kHz

central event at 30kHz central event at 30kHz
central event at 40kHz 0.98 central event at 40kHz
0.98
central event at 50kHz central event at 50kHz
central event at 60kHz GEM central event at 60kHz
MWPC central event at 70kHz central event at 70kHz
0.975 0.975
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
-1
1/pT (GeV/c) 1/pT (GeV/c)-1

Figure 7.4: TPC standalone tracking efficiency for MWPC (left panel) and GEM (right panel) readout. In the simulated data
central events (hdNch /dhi scaled to 2000) are embedded in the avereage background given by minimum bias events
(hdNch /dhi scaled to 500) at different interaction rates.

Figure 7.6 compares the dE/dx resolution for MWPC and GEM readout at different interaction rates. For
the MWPC readout the simulated performance is slightly better than the measured resolution of around
7 %, since in the current framework the deterioration due to the ion tail is not part of the simulation.
This does however not affect the results obtained for the GEM readout. We can conclude that the dE/dx
resolution is not compromised by the upgrade with GEM readout.

7.4 Space-charge distortions and corrections

At high collisions rates and large charged-particle multiplicities, the TPC drift volume contains a large
number of positive ions that pile up due to the slow ion drift velocity. The ion pileup effect can lead to a
significant accumulation of space charge. The resulting field distortions modify the electron drift lines,
introducing drift field distortions that have to be corrected. The correction of space-charge distortions
with sufficient precision is one of the major challenges within the TPC upgrade scheme. The goal is
to keep the residual distortions (after calibration) at a level not significantly larger than the intrinsic
resolution, i.e. a few hundred µm, as is achieved for the current TPC.

7.4.1 Space-charge sources

The ion pileup consists of the prompt contribution of the gas ionization by charged particles in the TPC
drift volume and the delayed contribution due to ion backflow from the GEM readout system. In case of
the upgraded TPC, the latter contribution is larger by a factor of 2e on average, where e is the number of
TPC Upgrade TDR 87

(GeV/c)-1
0.02 central event at 20 kHz central event at 30 kHz
TPC standalone

T
1/p
TPC constrained
0.015 TPC + ITS

0.01

0.005

0
(GeV/c)-1

0.02 central event at 60 kHz central event at 70 kHz

T
1/p

0.015

0.01

0.005

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1/p (GeV/c)-1 1/p (GeV/c)-1
T T

Figure 7.5: Resolution of 1/pT as a function of 1/pT for GEM readout. In the simulated data central events (hdNch /dhi scaled
to 2000) are embedded in a background of minimum bias events (hdNch /dhi scaled to 500) at different interaction
rates.

0.085 0.085
dE/dx

dE/dx

0.08 K 0.08 K
p p
0.075 0.075

0.07 0.07

0.065 0.065

0.06 0.06

0.055 0.055

0.05 dE/dx resolution for central events; 0.05 dE/dx resolution for central events;
MWPC Arrows indicate values for pp GEM Arrows indicate values for pp
0.045 0.045
20 30 40 50 60 70 20 30 40 50 60 70
Interaction rate (kHz) Interaction rate (kHz)

Figure 7.6: dE/dx resolution for central events as a function of the interaction rate for MWPC (left panel) and GEM (right
panel) readout. In the simulated data central events (hdNch /dhi scaled to 2000) are embedded in a background of
minimum bias events (hdNch /dhi scaled to 500). Only tracks in the momentum range from 4 to 5 GeV/c are used.
88 The ALICE Collaboration

ions drifting back into the TPC drift volume for each electron entering the multiplication region6 . For a
residual ion backflow of 1 % and an effective gain of 2000 the value of the e parameter is 20.
The average space-charge density is proportional to e, to the interaction rate Rint and to the ion drift time
tdion . For Rint = 50 kHz and tdion = 0.16 s the tracks of ⇠ 8000 interactions contribute to the ion pileup at
any given moment.
The magnitude of the space-charge distortions can be calculated using parametrized charged particle
density distributions under the assumption of certain symmetries:

a bz+ce
rsc (r, z) = . (7.1)
rd

The resulting space-charge density map for Rint = 50 kHz and e = 20 is shown in Fig. 7.7. The distri-
bution is symmetric in j and linear in z (due to the constant ion drift velocity7 ). From data we extract a
radial dependence with a value of d between 1.5 and 2 (see Sec. 7.4.4). The small step at z ⇡ 0 is due to
addditional background from the muon absorber on the C side.

Ne-CO2-N2 (90-10-5): 50 kHz, = 20

140
(fC/cm3)

120

100
SC

80

60

40
100
20 120 )
(cm
140
250 200 160 r
150 100 180
z (cm) 50 0 -50 200
-100-150 220
-200-250 240

Figure 7.7: Average space charge density for Ne-CO2 -N2 (90-10-5), Rint = 50 kHz and e = 20.

7.4.2 Magnitude of the distortions

The magnitude of the space-charge distortions (without fluctuations) is calculated using parametrized
space-charge density distributions as in Eq. (7.1). The resulting average distortions in r, rj and z direc-
tions are shown in Fig. 7.8.
The space charge has a focusing effect, forcing the drifting electrons to divert from the ideal drift path
towards larger (smaller) radii at the inner (outer) field cage of the TPC. Moreover, the emerging radial
electric field component in the presence of the magnetic field leads to distortions in rj due to the E ⇥ B
effect. In general, the effect is strongest close to the central electrode (|z| ⇡ 0) and at the inner and outer
field cage.
The magnitude of the distortions at the central electrode of the TPC (at z ⇡ 0 cm) is shown in Fig. 7.9. At
the inner field cage the space-point distortions reach up to 19 cm in r direction and 7 cm in rj direction.
However, in the largest part of the volume, the distortions are well below 10 cm. Distortions of similar
6 The factor of 2 describes the average drift length difference for ions from primary ionization and ions from ion backflow.
7 The average interaction rate is also constant on the short timescales considered here.
TPC Upgrade TDR 89

dr (cm) for Ne-CO -N2 (90-10-5), 50 kHz, = 20 dr (cm) for Ne-CO -N2 (90-10-5), 50 kHz, = 10
2 2
25 25
r (cm)

r (cm)
240 240
20 20
220 220

15 15
200 200

180 10 180 10

160 5 160 5

140 0 140 0

120 120
-5 -5

100 100
-10 -10
-250 -200 -150 -100 -50 0 50 100 150 200 250 -250 -200 -150 -100 -50 0 50 100 150 200 250
z (cm) z (cm)
d(r ) (cm) for Ne-CO2-N2 (90-10-5), 50 kHz, = 20 d(r ) (cm) for Ne-CO2-N2 (90-10-5), 50 kHz, = 10
r (cm)

r (cm)
240 4 240 4

220 220
2 2

200 200
0 0
180 180
-2 -2
160 160

-4 -4
140 140

120 -6 120 -6

100 -8 100 -8

-250 -200 -150 -100 -50 0 50 100 150 200 250 -250 -200 -150 -100 -50 0 50 100 150 200 250
z (cm) z (cm)
dz (cm) for Ne-CO2-N2 (90-10-5), 50 kHz, = 20 dz (cm) for Ne-CO2-N2 (90-10-5), 50 kHz, = 10
5 5
r (cm)

r (cm)

240 240
4 4

220 3 220 3

200 2 200 2

1 1
180 180
0 0
160 160
-1 -1
140 140
-2 -2

120 -3 120 -3

100 -4 100 -4

-5 -5
-250 -200 -150 -100 -50 0 50 100 150 200 250 -250 -200 -150 -100 -50 0 50 100 150 200 250
z (cm) z (cm)

Figure 7.8: Space-point distortions in r (top panels), rj (center panels) and z (bottom panels) for Ne-CO2 -N2 (90-10-5), Rint =
50 kHz and e = 20 (left panels) and 10 (right panels). The data shows the integrated distortions for an electron
originating at z, r as it drifts to the GEM readout at z = ±250 cm.
90 The ALICE Collaboration

magnitude are encountered or expected in the near future in the TPC of the STAR experiment at RHIC [9,
10]. The distortions expected for the TPC considered for the PANDA experiment at FAIR [11] are smaller
by an order of magnitude.
dr (cm)

d(r ) (cm)
20
Ne-CO2-N2 (90-10-5): 50 kHz 2
15
= 20
= 10 0
10

5 -2

0 -4

-5
-6

-10
-8
100 120 140 160 180 200 220 240 100 120 140 160 180 200 220 240
r (cm) r (cm)

Figure 7.9: Space-point distortions in r (left panel) and rj (right panel) as a function of the radial position r close to the central
electrode (z ⇡ 0 cm) for Ne-CO2 -N2 (90-10-5), Rint = 50 kHz, e = 10 and 20.

The space-charge distortions do not increase linearly with e. A drifting electron will be deflected by the
space-charge field towards regions with lower field gradients. The resulting saturation effect is shown in
Fig. 7.10.
d(r ) (cm)
dr (cm)

70 0
Ne-CO2-N2 (90-10-5)
60
R int = 50 kHz -5
50

40 -10

30
-15
20

10 -20
x = 85cm, y = 0cm, z = 10cm x = 85cm, y = 0cm, z = 10cm
00 10 20 30 40 50 0 10 20 30 40 50

Figure 7.10: Space-point distortions in r (left panel) and rj (right panel) close to the central electrode (z ⇡ 10 cm) and in the
center of a readout chamber (y = 0) as a function of e. The broken line is an eye guide representing linear increase
with e.

7.4.3 Simulation of the space-charge distortions

Equation (7.1) provides a straightforward solution to calculate the magnitude of the space-charge distor-
tions. However, fluctuations in the space-charge density distribution are not taken into account. Conse-
quently, it describes well a situation with the main contribution to the space-charge distribution coming
from primary ionization (e ! 0). In case of the upgraded ALICE TPC, however, the main contribution
to the space-charge distribution arises from ion backflow. Here, the non-active regions along the sector
boundaries together with local variations of gain and e break the azimuthal symmetry. As we will see,
also the fluctuations in the number of interactions per unit time, and in the charged track multiplicity
TPC Upgrade TDR 91

per event have to be considered, as they give rise to space-charge fluctuations along z. Each interaction
actually creates a disc of slowly moving ions within the TPC volume, and the space charge density dis-
tribution at 50 kHz interaction rate consists of the superposition of ⇠ 8000 such disks with fluctuating
density.
In the following, an detailed simulation procedure for space-charge distortions is described, that deals
with these features. It consists of the following steps:

1. Calculation of the space-charge density maps: Detailed 3-dimensional average space-charge

maps can be calculated without any symmetry assumptions utilizing real raw data from minimum
bias Pb–Pb collisons from RUN 1. Appoximately 130,000 events were used for this purpose. The
individual events are compressed along the z direction by a factor td /tdion and superimposed with
random positions along z. Moreover, space-charge maps for randomly selected groups of events are
used to mimic the fluctuations of the space-charge distribution in space and time. A subset of 8000
events is needed to simulate an interaction rate of 50 kHz. The individual events are superimposed
with random positions along the z direction with a spacing defined by the interaction rate and the
ion drift velocity.
2. Calculation of the space-charge field: Based on the space-charge maps obtained in the previous
step, the resulting electric field components are calculated. Taking into account the precision
requirement and the granularity of the space charge, it is clear that the electric field has to be
calculated in 3 dimensions and with high spatial granularity. In the following studies a granularity
of 360 bins in rj, 125 bins in z, and 158 bins in r is used. A 3D electric field calculation is done
using a customized implementation of the Poisson relaxation method [12]. The originally proposed
analytical approach does not provide a 3D solution with sufficient spatial granularity.
3. Calculation of the distortion maps: The electric field deviations D~E based on rsc can be com-
puted by solving the inhomogeneous Laplace equation. This task is usually performed by solving
the Poisson equation with the help of discrete numerical methods like Finite Elements as applied
in STAR [9] and PANDA [11] or by using an analytical approach based on the Green’s function
as proposed for ALICE [13]. However, these approaches turn out to be not able to cope with the
granularity requirements and instead the Poisson relaxation method is used.
To compute the distortion maps as a function of the starting position of the drifting primary elec-
tron, the Langevin equation of motion for electrons in electric and magnetic fields is integrated
along the drift line with a fourth-order Runge Kutta algorithm [14].

Repulsive electrostatic forces between the ions are not yet considered in this calculation.They will cause
distortions of the ion drift path as well, which will lead to smoother space-charge distributions than those
discussed in this TDR.

7.4.4 Space-charge density fluctuations

In this section we investigate the fluctuations of the space-charge density distribution using data obtained
with the simulation method described in Sec. 7.4.3 and with an analytical formula for the expected
fluctuations.

Features of the space-charge map distributions

Figure 7.11 shows the xy and rz projections of 2 example space-charge density maps and of the full
analysed event sample of 130,000 events.
An almost regular sector modulation of the space charge due to the dead zones between the readout
chambers is observed. Moreover, there is a step in the density distribution between the IROC and OROC
92 The ALICE Collaboration

9 9
10 10
250

(C/m /e0)
(C/m /e0)
y (cm)

50

y (cm)
240

3
3
200 50

220

sc
sc
150 40
40
100 200

50 30 180
30
0
160
-50 20 20
140
-100

-150 10 120
10
-200 100

-250 0 0
-250 -200 -150 -100 -50 0 50 100 150 200 250 -250 -200 -150 -100 -50 0 50 100 150 200 250
x (cm) z (cm)
9 9
10 10
250

(C/m /e0)
y (cm)

(C/m /e0)

80
y (cm)

240

3
3

200 100
70
220

sc
150
sc

60
100 200 80

50 50
180
60
0 40
160
-50
30 40
140
-100
20
-150 120
20
10
-200 100
-250 0 0
-250 -200 -150 -100 -50 0 50 100 150 200 250 -250 -200 -150 -100 -50 0 50 100 150 200 250
x (cm) z (cm)
9 9
10 10
250

(C/m /e0)
(C/m /e0)
y (cm)

y (cm)

50 240

3
3

200 50
220

sc
sc

150
40
100 200 40

50 180
30
30
0
160
-50
20
140 20
-100

-150 120
10 10
-200 100

-250 0 0
-250 -200 -150 -100 -50 0 50 100 150 200 250 -250 -200 -150 -100 -50 0 50 100 150 200 250
x (cm) z (cm)

Figure 7.11: Calculated space-charge density maps for an ion pileup of 2000 (top row), 8000, and 130,000 (bottom row). The
histograms are normalized to 10,000 ion pileup events. Left column: xy projection at z = 10 cm. Right column: rz
projection at j = 0.05. In these plots on the horizonal axis the full drift length |z| = zroc = 250 cm corresponds to
the ion drift time tdion = 0.16 s. The data is based on the superposition of minimum bias Pb–Pb collisions recorded
in RUN 1. The increased charge density in a few sectors at certain radial positions is explained by a few floating
wires in the MWPC readout system used in RUN 1, which increase the local gas gain.
TPC Upgrade TDR 93

regions, caused by the systematically different gas gain applied in order to keep the same signal-to-noise
ratio in these two regions with different pad and wire geometry. An r 1.5 -scaling is found for the radial
dependence in the full analysed event sample (130,000 events), in contrast to the r 2 scaling used by
STAR [15].
Strong local variations of the space-charge density are observed. The effect is most pronounced in the z
direction, where regions of high ion density corresponding to central events are clearly visible.
In Fig. 7.12 the expected rj distortion maps are shown without (left) and with (right) magnetic field.
The figure illustrates the effect of a sector modulation that can be explained by the lower space-charge
density close to the dead zones in between readout chambers. The positive ion charge attracts the drifting
electrons towards the center of the readout chamber. In the setup with magnetic field the symmetry is
broken due to the E ⇥ B effect. Consequently, the mean value of the distortion is shifted.

250 250

y (cm)

d(r ) (cm)
d(r ) (cm)
y (cm)

0.15 0.2
200 200

150 0.1 150 0

100 100
0.05
-0.2
50 50
0
0 0
-0.4

-50 -0.05 -50

-100 -100 -0.6

-0.1
-150 -150
-0.8
-200 -0.15 -200

-250 -250
-250 -200 -150 -100 -50 0 50 100 150 200 250 -250 -200 -150 -100 -50 0 50 100 150 200 250
x (cm) x (cm)

Figure 7.12: xy projection of the rj distortion map close to the TPC central electrode (at z = 10 cm). The data are based
on a detailed 3-dimensional space charge map normalized to e = 5 (in order to avoid complications due to non-
linearities). The figures illustrate the effect of a sector modulation that is modified by the magnetic field, which is
set to B = 0 T (left) and B = 0.5 T (right).

Contributions to the space-charge fluctuations

ion
The average number of interactions contributing to the ion pileup in the TPC drift volume is Npileup ⇡
8000. The actual number of interactions is described by a Poissonian distribution around this mean
value. The relative fluctuation of the space-charge density ssc /µsc , where µsc = hrsc i is the average
space-charge density, can then be written as:

v
u
u ✓ ◆2 ✓ ◆2 !
ssc 1 t1 + sNmult 1 sQtrack
=q + 1+ . (7.2)
µsc ion
Npileup µNmult F µNmult µQtrack

The relative fluctuation of the space-charge density depends on three contributions:

q
ion
1. 1/ Npileup ⇡ 1.1 % is the relative fluctuation of the number of ion pileup events. Already this
contribution is larger than the required precision of a few ‰.
sNmult
2. µNmult ⇡ 1.4 % is the relative RMS of the multiplicity distribution.
sQtrack
3. µQtrack ⇡ 1.7 % is the relative variation of the ionization of a single track.
94 The ALICE Collaboration

F is a geometrical factor describing the spatial range over which the space-charge fluctuations are rel-
evant for the distortions. To describe the fluctuation in the space-point distortions, the relevant scale is
determined by the range of the Coulomb interaction.
A fast Monte Carlo (MC) simulation has been developed in order to estimate the contributions to the
space-charge fluctuations (events, tracks and charge). The result is shown in Fig. 7.13. Here, sub-
volumes of 10 % of the size of a TPC sector are considered.
0.05
sc
Contributions to the fluctuations
sc
0.045
Number of events: Nev
0.04 Nev and multiplicity (mult)
Nev + mult, -region (1/180)
0.035 Nev + mult + charge per track,
-region (1/180)
0.03

0.025

0.02

0.015

0.01

0.005

0
5000 10000 15000 20000
Ion pileup events
Figure 7.13: Relative fluctuation of the number of events, tracks and energy deposit (charge) from a MC simulation compared
ion .
to expectations based on Eq. (7.2) as a function of Npileup

The RMS of the space-charge fluctuations as obtained from the fast MC algorithm agrees well with the
expectations from the analytical formula, Eq. (7.2), for fluctuations of the number of events, number
of tracks (a Poissonian convoluted
q with the multiplicity distribution) and energy deposit. The relative
ion . Depending on the spatial granularity, the relative fluctuation is around
fluctuation scales with 1/ Npileup
ion = 8000. This result agrees well with the observations for the fluctuations of the charge
2 – 3 % for Npileup
density reported in Sec. 7.4.3 and the fluctuations on the track level (see Sec. 7.4.5).

7.4.5 Impact of the fluctuations on the distortion corrections

To estimate the effect of fluctuations on the space-charge distortion corrections, 50 different space-charge
configurations are created. By random superposition of real Pb–Pb events, the number of ion pileup
ion
events Npileup is varied between 2000 and 8000 in order to study different magnitudes of the ion pileup.
For each space-charge configuration, the distortion maps are calculated. The distorted tracks are cor-
rected using an average map based on the full sample of 130,000 interactions, that is normalized to the
ion . The rj position of the tracks after correction is compared to that of undistorted
corresponding Npileup
tracks. While the correction works well on average, there is a considerable spread. The RMS of this
spread as shown in Fig. 7.14 is characteristic for the expected space-charge fluctuations. For tracks with
pT > 1 GeV/c the residual distortions yield up to ⇠ 2.8 mm, which is significantly larger than the intrinsic
resolution limit sintr .
Finally, we demonstrate the importance of the knowledge of the z distribution of the space charge and
estimate the frequency by which the space-charge distortion map needs to be updated. To this end, the
following procedure is carried out:

1. The space-points of the TPC tracks are modified using a given distortion map.
TPC Upgrade TDR 95

(cm)
0.4 Residual distortions in r after average correction
50 kHz, ion pileup = 8000
0.35

r
37.5 kHz, ion pileup = 6000
25 kHz, ion pileup = 4000
0.3 12.5 kHz, ion pileup = 2000

0.25

0.2

0.15

0.1

0.05

0
-4 -3 -2 -1 0 1 2 3 4
-1
q /p (GeV/c)
T

Figure 7.14: Residual rj distortions as function of rigidity q/pT after correction using an average space-charge distortion map
(full sample of 130,000 interactions to the corresponding multiplicity). The asymmetry between positive and
negative rigidity is related to the sector modulation effect, caused by the lower space-charge density close to the
dead zones between readout chambers, as shown in Fig. 7.12 (right).

2. A second distortion map using the same space-charge density distribution, but shifted in z direction,
is used for correction: r2 (x, y, z0 ) = r1 (x, y, z + Dz ).

The result of the correction is compared to the undistorted track position. The RMS of the observed
residual distributions is shown in Fig. 7.15 as a function of 1/pT . The residuals are increasing with Dz
and saturate at the random limit which is comparable with Fig. 7.14. The results shown in Fig. 7.15 reach
the intrinsic resolution limit of a few hundred µm already at effective displacements of Dz = 16 cm,
corresponding to an ion drift time of 10 ms. This defines the requirement to update the space-charge
correction maps during the online calibration procedure after about 5 ms, as described in Chap. 8.

7.5 Performance with residual space-charge distortions

The large space-charge densities expected during the high luminosity running in RUN 3 create distortions
which can locally exceed 10 cm, while being below that value in the largest part of the drift volume. In
order to preserve the performance of the detector, the space-charge distortions have to be corrected with
a precision comparable to the intrinsic track resolution, which is of the order of a few hundred µm. The
implementation of a correction framework in the general calibration scheme is discussed in detail in the
following Chap. 8.
Figure 7.16 shows the ITS-TPC matching effiency (left) as well as the transverse momentum resolution
(right) obtained from a full MC, implementing the remaining residual space charge distortions after
applying the calibration procedure described in Sec. 8.4 (“second reconstruction stage”).
The matching efficiency is above 95% for all momenta. For standalone TPC tracks (red), the pT resolu-
tion is slightly worse than the one obtained without distortions (see also Fig. 7.5), while for TPC tracks
constrained to the interaction vertex (blue) and tracks matched to the ITS (black) the performance is
practically the same as for the ideal case without space-charge distortions.
96 The ALICE Collaboration

(cm)
0.45
Distortion due to time delay for correction map Distortion due to time delay z =112 cm
12.5 kHz, ion pileup = 2000 for correction map
0.4
r 25 kHz, ion pileup = 4000 z =80 cm

0.35 z =48 cm

z =16 cm
0.3

0.25

0.2

0.15

0.1

0.05

0
(cm)

0.45
Distortion due to time delay for correction map Distortion due to time delay for correction map
0.4 37.5 kHz, ion pile-up = 6000 50 kHz, ion pile-up = 8000
r

0.35

0.3

0.25

0.2

0.15

0.1

0.05

0
-4 -3 -2 -1 0 1 2 3 -3 -2 -1 0 1 2 3 4
q/p T (GeV/c) -1 q/p T (GeV/c) -1

Figure 7.15: Residual rj distortions of the tracks. The space points are distorted using randomly selected raw data with ion
pileup numbers of 2000, 4000, 6000 and 8000 and corrected using the same space-charge map but shifted by an
offset Dz .

1 0.03
(GeV/c)-1
TPC - ITS matching efficiency

central event at 50 kHz central event at 50 kHz

0.98
0.025 TPC standalone
T
1/p

0.96
TPC constrained
0.94 TPC + ITS
0.02
0.92

0.9 0.015

0.88
0.01
0.86

0.84
0.005
0.82

0.8 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
1/p (GeV/c)-1 1/p (GeV/c)-1
T T

Figure 7.16: ITS-TPC track matching efficiency (left) and pT resolution (right) with residual distortions after the second recon-
struction stage.
Chapter 8

Online reconstruction, calibration, and

monitoring

Operation of the TPC in a continuous readout mode in the high-luminosity environment at the LHC in
RUN 3 will produce a vast amount of data. This necessitates a change of the ALICE computing model
towards massive online reconstruction, because only in this way it is possible to achieve sufficient data
reduction such that the data can be stored.
Efficient online cluster and track reconstruction implies that also a number of calibration steps need to
be performed online. In particular, an online correction scheme for the space-point distortions induced
by back-drifting ions must be applied to ensure efficient cluster-to-track association. Moreover, the
online corrections have to be sufficiently precise to match TPC tracks efficiently to the external detectors
ITS and TRD. If this can be achieved, a final calibration based on global tracks can be performed in a
subsequent processing step.

8.1 Continuous TPC operation at high luminosities

The online reconstruction and calibration strategy is determined by the expected running conditions,
defined by the LHC, as well as by the constraints given by the detector (continuous readout) and the
online systems (need for data compression). In this section, the relevant operational conditions in RUN 3
are summarized and an overview of the general online reconstruction and calibration scheme is given.

8.1.1 LHC conditions in RUN 3

The expected running conditions in RUN 3 are defined by beam energy, luminosity and interaction rates,
and interaction spacing.

Average interaction rates

During the first Pb–Pb run in November 2010 a peak luminosity of peak = 2.5 · 1025 cm 2 s 1 and
a hadronic interaction rate of ⇠ 200 Hz was reached [1]. In the second Pb–Pb run in 2011 the peak
luminosity reached peak = 4 · 1026 cm 2 s 1 with a hadronic interaction rate of ⇠ 3.5 kHz [2]. At the
same time the readout rate was limited to a few hundred Hz by the readout speed of the current TPC
electronics [3].
In RUN 3 the LHC will reach minimum bias hadronic interaction rates of Rint = 50 kHz in Pb–Pb
p
collisions at the design center-of-mass energy of sNN = 5.5 TeV, corresponding to a luminosity of
= 6 · 1027 cm 2 s 1 . For this TDR we assume a constant interaction rate of 50 kHz throughout a fill

97
98 The ALICE Collaboration

of the LHC. An effective data-taking time of 106 s per year is assumed. This allows to collect a total
integrated luminosity of at least int = 10 nb 1 during RUN 3.

Beam structure and interaction spacing

A possible (simplified) LHC filling scheme for Pb-ions in RUN 3 consists of 12 equally-spaced bunch
trains with 48 bunches each, leading to 576 bunches in each ring. The filling scheme and corresponding
bunch train structure are depicted in Fig. 8.1.

bunch
train 3
2 4

1 5 bunch train 1 bunch train 2

1 2 47 48 1 48
12 6 ..... .....
50ns
11 7
1 bunch train = 2.35 s = 48 bunches
10 8
9

Figure 8.1: Schematic LHC filling scheme and bunch train structure.

When two bunch trains are passing, the instantaneous bunch-crossing rate is 20 MHz, while the average
bunch-crossing rate is 6.3 MHz. The interaction rate of 50 kHz translates into a probability µ = 0.0079
for a collision to occur in a single bunch crossing1 . Therefore, the probability for at least one hadronic
interaction (for more than one interaction) within one bunch train crossing (2.35 µs) is ⇠ 32 % (resp.
⇠ 5.5 %).

Event pileup
Event pileup in the TPC arises if more than one collision occurs within a time interval that corresponds
to the maximum electron drift time, i.e. within td ⇡ 100 µs. Even though the effect increases the detector
occupancy, it does not complicate the pattern recognition significantly, as demonstrated in Sec. 7.3.
Not only the average number of pileup events is of importance, but also their spacing in time. At an
interaction rate of 50 kHz on average Npileup = 5 interactions are contained in a time interval td . The aver-
age spacing between two collisions is 20 µs or 52 cm of drift length. The distribution of time differences
between two collisions is Poissonian with the bunch train structure of the LHC beam superimposed, as
illustrated in Fig. 8.2. Following the argumentation of the previous section, in ⇠ 17 % of the cases a
second collision occurs within the same bunch train crossing, i.e. within 2.35 µs, corresponding to the
first peak in Fig. 8.2. Separation of the tracks from two such collisions will be possible using external
detector information, i.e. from the ITS.

Beam induced background

For colliding Pb beams in RUN 3 it is not expected that machine background contributes significantly
to the charged track rates in the TPC. The beam currents (⇠ 85 mA) will not be large enough to trigger
electron-cloud effects, which need at least 310 mA [4]. In the case of pp collisions, the beam currents will
exceed the ones achieved in RUN 1, which were large enough to produce a background rate comparable
to the interaction rate. However, several improvements are foreseen that should result in an improved
background situation. These improvements include the replacement of the existing collimator at the end
of the transfer line (TDI), which is now a major source of electron-cloud effects, the removal of bulky
vacuum equipment near the ALICE interaction region, and the improvement of the vacuum at and near
Interaction Point 2 (IP 2).
1µ = 50 kHz/11 kHz/576 = 0.0079, with the LHC orbit frequency 11 kHz.
TPC Upgrade TDR 99

Entries
2200
2000
1800
1600
1400
1200
1000
800
600
400
200
0
0 10 20 30 40 50 60 70 80 90 100
t ( s)
Figure 8.2: Distribution of time differences between two collisions at Rint = 50 kHz. A Poissonian distribution is superimposed
to the bunch train structure of the LHC beam.

8.1.2 TPC reconstruction, calibration, and data compression in RUN 3

The ALICE upgrade requires a major change of the computing concept in order to be able to process and
store the large amount of data produced in Pb–Pb collisions. The main contributor to the data volume
will be the TPC. Most of the data processing and reconstruction will be performed on the computing
cluster of the new online systems [5]. New requirements to calibration and reconstruction algorithms
emerge in terms of running stability, processing time, memory consumption, as well as parallelizability
on different levels. These requirements build strict constraints on the data reconstruction and calibration,
as well as on data compression. A massive use of hardware co-processors, such as FPGAs for the early
processing steps, as well GPGPUs2 for the later processing is foreseen.

Overview of the TPC reconstruction scheme

The choice of the reconstruction algorithms is strongly driven by the constraints of the online processing.
However, in the discussion below we will focus on demonstrating the overall strategy and the general
feasibility of online reconstruction under the operational conditions described above.
A two-stage process, as depicted in Fig. 8.3, is foreseen for the online data processing. The first stage
(see below) will focus on cluster finding and the association of clusters to tracks, which are needed in
order to perform the necessary data size reduction (see Sec. 8.3). The compressed data will be written to
permanent storage. The reconstructed tracks have sufficient precision to allow matching to the external
detectors, mainly ITS and TRD. This is needed to improve the quality of the subsequent calibration step
during the next reconstruction stage.
The second reconstruction stage (see Sec. 8.4) will also be performed on the online computing cluster,
but in an asynchronous mode, and can thus be repeated any time. It aims at a further improvement of the
data quality, in particular in terms of the space-charge distortion calibrations, and employs information
from external detectors as well as more refined calibration input.

2 General-Purpose Graphics Processing Units (GPGPUs)

100 The ALICE Collaboration

Reconstruction Stage 1
TPC
Cluster Finding Seeding / Tracking Data Compression
electronics
- Dead channel map - Scaled average space-charge distortion map
- Drift velocity
- Pedestals
- Pad-by-pad gain equalization
- Chamber-by-chamber
Update interval O(15 min)
gain equalization (HV)

Permanent Storage

Reconstruction Stage 2
Tracking / External Track Matching
Physics Ready
Data
- High-resolution space-charge distortion map
- ITS-TRD external track reference
Update interval O(5 ms)

Data taking Online Systems Data analysis

Figure 8.3: Schematic outline of the calibration flow during the data-taking and reconstruction process.

Data size and data compression

After zero suppression on the level of the front-end electronics, the average TPC raw data size in min-
imum bias Pb–Pb interactions is ⇠ 20 MByte (see Sec. 6.3). Without any further compression applied,
this would result in an input rate of about 1 TByte/s to the online systems at an interaction rate of 50 kHz,
and a total amount of 3 EByte (1018 Byte) of TPC data recorded in RUN 3. Such numbers exceed the pre-
dicted available bandwidth and storage space by a large factor. Thus, in order to permit permanent data
storage, additional compression on top of zero suppression to below 1 MByte per interaction is required.
This can be achieved by different levels of pattern recognition that are performed in the online systems.
The zero-suppressed raw data are decoded at the input to the online farm, where the raw data digits
(which consist of arrival time at the front-end electronics and signal amplitude) are associated with the
corresponding geometrical position (pad row and pad coordinates). Cluster finding is performed on the
digits, which produces three-dimensional charge clusters3 .
As the xro coordinate4 is fixed by the well-separated pad rows, cluster finding is reduced to a two-
dimensional problem in the pad–time plane. A two-dimensional algorithm, as used in the current TPC
offline cluster finder, scans the pad–time plane for charge maxima using a sliding window. It allows a
good separation of close-by clusters. A different method is based on a one-dimensional algorithm, as
used in the current High Level Trigger (HLT) system. It processes the pads sequentially, which allows
to find the maxima within neighboring pads on the same pad row. In this case the separation of close-
by clusters is done separately in both dimensions. The algorithm allows massive parallelization and,
therefore, an easy implementation into an FPGA5 . It is the baseline solution for RUN 3, as it fits the
necessity of an early data reduction already at the input to the online systems.
The cluster finding is accompanied by a further compression step based on intelligent Huffmann Cod-
ing [6] of selected parameters.
Such a compression scheme has already been applied to TPC data during the 2011 Pb–Pb data taking
in RUN 1, resulting in a data compression factor of ⇠ 4 as compared to zero-suppressed raw data (see
Fig. 8.4). Further optimizations of the cluster data format and of the compression algorithm for RUN 3
will allow a total data reduction by a factor five to seven during the cluster finding step [5].
3 Alternatively the cluster finding could also be performed on the Common Readout Unit (CRU) modules for the TPC.
4 Details on the local and global coordinate system can be found in App. A.
5 Field Programmable Gate Array (FPGA)
TPC Upgrade TDR 101

Compression factor
8

102
6

10
4

2 1
0 10000 20000 30000 40000 50000 60000 70000 80000
Event size (kByte)

Figure 8.4: Data compression factor versus raw data size achieved using cluster finding and data-format optimizations in the
High-Level Trigger during the Pb–Pb data taking in RUN 1. Adopted from [7].

A further reduction in data size can be performed by tracking, where clusters are assigned to particle
tracks. This step, which is discussed in Sec. 8.3, allows to remove clusters that are not associated to
physics tracks (e.g. from delta electrons or noise) from the data stream. The possible reduction factor,
based on the experience from the past data taking in RUN 1, is of the order of two. The tracking step
also enables more advanced transformation schemes to optimize the parameter distributions for entropy
encoding as well as the possible replacement of some of the individual cluster parameters by track-based
properties. We estimate the further reduction potential to be of the order of two to three.
In total, the envisaged compression factor in the online system is of the order 20, resulting in an average
data size per interaction of less than 1 MByte. This will lead to a storage rate of ⇠ 50 GByte/s and a
total amount of ⇠ 150 PByte of stored data during RUN 3. To store the cluster information associated
to the reconstructed tracks is an advantage, allowing a possible re-calibration of individual clusters at a
later stage and, therefore, an improvement of the TPC performance. A summary of the consecutive data
compression factors is presented in Tab. 8.1.

Data Format Data Compression Factor Event Size (MByte)

Zero Suppression (FEE) 20
Clusterization 5-7 3
Remove clusters not associated 2 1.5
to relevant tracks
Data format optimization 2-3 <1
Table 8.1: The TPC event size and data compression factors for the different data compression steps performed in the front-end
electronics and the online systems.

Online calibration
Online reconstruction implies that also calibration information must be available at run time. Since
some of the calibration parameters change with time, the calibration information has to be retrieved and
updated synchronously to the data collection process.
The calibration process is embedded in the two-stage reconstruction process, as schematically shown in
Fig. 8.3. The calibration flow is described in the following. In the first reconstruction stage, the calibra-
tion needs to be precise enough to associate clusters to tracks in order to allow the data compression (see
102 The ALICE Collaboration

Sec. 8.1.2). In the second reconstruction stage the calibration aims at providing the detector performance
required for physics analysis6 .

1. Before the start of data taking, the front-end electronics need to be configured with the correct
pedestal values and zero suppression thresholds for each readout channel (see also Sec. 6.4.8) and
with the current map of active channels. These values are extracted from data recorded in special
pedestal runs between LHC fills. The high voltage of the readout chambers is set such that the
average gain of the chambers is equalized throughout the TPC. The settings are based on data
obtained using the Krypton calibration method (see Sec. 8.6.1).

2. During data taking the cluster finder accesses the map of active channels in order to account for
broken or malfunctioning readout equipment.

3. For the first reconstruction stage the relevant calibrations include drift velocity, effective gain,
and space-charge distortions. The corresponding calibration parameters change with time and
have to be updated in intervals of (15 min), but also the static pad-by-pad equalization of the
gain extracted using the Krypton calibration method is applied at this stage. An average space-
charge map, updated in intervals (15 min) as well, is used for a coarse space-charge distortion
correction. It accounts for slow variations of the luminosity, pressure and temperature and for
malfunctioning sectors. To improve the precision for use during the first reconstruction stage, the
average space-charge map is scaled by the actual charged-particle multiplicity integrated over the
preceding 160 ms, i.e. the maximum ion drift time in the TPC. This scaling procedure is based on
a running integral of TPC current and amplitude information over the corresponding time window
and reduces significantly the error due to temporal ion density fluctuations in the TPC.

4. During the second reconstruction stage a high-resolution distortion correction map is derived to
achieve the required momentum resolution. The final distortion correction is based on external
reference tracks using information from ITS and TRD and combines them with the space-charge
information based on TPC clusters with high granularity in space and time. According to the
typical time scales of space-charge fluctuations (see Sec. 7.4.5) the final distortion correction map
must be determined in time intervals of a few ms.

8.2 Space-charge distortion corrections

The largest contribution to drift distortions in the GEM TPC are due to the accumulation of space charge
in the drift volume. The magnitude and implications of such distortions are described in detail in Sec. 7.4.
Below we describe how distortions due to space charge are treated when the TPC is operated in contin-
uous readout, discuss how distortion corrections are obtained, and describe different levels of precision
that can be reached depending on the available calibration input.

8.2.1 TPC coordinate transformation

Each TPC raw data digit consists of a geometrical position at the readout plane (pad row and pad co-
ordinates), the arrival time at the front-end electronics (tdigit ) and the corresponding signal amplitude.
Charge clusters usually spread over a few pads and time bins, and the center-of-mass of a charge cluster
corresponds to a three-dimensional space point ~rcls = (x, y, z). The proper assignment of ~rcls requires a
precise knowledge of the drift time td of the cluster:

Z td
~rcls =~rro + ~vd (x, y, z) dt , (8.1)
0
6 The achievable TPC performance is discussed in detail in Chap. 7.
TPC Upgrade TDR 103

where ~vd is the drift velocity vector of electrons in the TPC and~rro = (xro , yro , zroc ) is the position of the
center-of-mass of the charge cluster at the readout plane (zroc is the z-position of the read-out chamber).
Neglecting distortions, i.e. for ~vd = (0, 0, vd ), Eq. (8.1) translates into

~rcls = (xro , yro , zroc v d td ) . (8.2)

In a triggered readout mode there is a strict relation between tdigit and td ,

td = tdigit t0 , (8.3)

where t0 is the time of the interaction that triggers the readout. In data aquired with a TPC in continuous
readout mode the important parameter t0 is a priori unknown, but can be derived using the information
from external detectors. However, even in a TPC standalone tracking scheme without external detec-
tor information, t0 can be estimated by extrapolation of track segments to x = y = 0, as discussed in
Sec. 8.5.1.

8.2.2 Space point corrections

In practice, the drift velocity vector is approximated by ~vd = (0, 0, vd ) and drift-field distortions, i.e. the
effect of non-vanishing vx and vy are treated by effective corrections:

~rcls = (xro , yro , zro ) +~D(xro , yro , zro ) , (8.4)

with zro = zroc vdtd and the drift time td = tdigit t0 (see Eq. (8.3)).
Numerical as well as analytical calculations of ~D are based on the knowledge of the density distribution
of the space charge (rsc ), as described in 7.4.3. Such maps are available with different levels of precision
as described below.

8.2.3 Space-charge density maps

During the two reconstruction stages, estimates of the current space charge distribution with different
precision and granularity are used for distortion correction.

– The reference map rref can be obtained from simulations. It is based on the geometric acceptance
of the readout chambers (incorporating non-active regions at the sector boundaries), dead regions
of the GEM system and the known variations of gain and e, the number of back-drifting ions per
primary electron. This map can serve as an input at the startup of the data taking.

– The average map rav , updated several times per fill, accounts for slow variations of luminosity,
ambient conditions and readout chamber status. It can be derived for example from an external,
high-pT track sample ( (1 min) statistics).

For the first reconstruction stage, space point corrections at the level of the single cluster resolution
( (mm)) are sufficient in order to perform the cluster-to-track association. The second reconstruction
stage aims at reaching the intrinsic performance of the detector, therefore requiring effective residual
space-charge distortions at the level of the intrinsic track resolution of a few 100 µm in rj. At both
stages space-charge maps with better precision, including also the effect of fluctuations, are necessary.
104 The ALICE Collaboration

– The scaled average map rscaled is based on rav , but scaled by the instantaneous ion current dur-
ing the last tdion = 160 ms. This covers the largest part (⇠ 2 %) of the space-charge fluctuations
(see also Fig. 7.14), which are the contributions due to fluctuations in the number of ion pileup
events7 and the fluctuations in the track multiplicity (centrality). These parameters are accessible
online through external detectors (collision counter, centrality measure) or from the TPC (r and rj
averaged signal charges or GEM currents).

– The high-resolution map rhigh res also contains topological fluctuations in r and rj, which are
responsible for the remaining fluctuations of the space charge of ⇠ 1 %. It can be obtained by the
calibration of TPC cluster residuals with respect to a track interpolation from the ITS and TRD
detectors, as described in Sec. 8.4.1

We aim at using rscaled and rhigh res for distortion corrections in the first and second reconstruction stage,
respectively.

8.3 First reconstruction stage

Online reconstruction is necessary in order to achieve data compression by a factor of 20 as compared
to the raw data size, and to allow for permanent storage of the data. Such compression factors can be
achieved if the association of clusters to tracks can be performed online, which implies also the necessity
for sufficient online correction of the space-charge distortions. In order to correct for these distortions
using Eq. (8.4), an estimate for the time of the interaction t0 is needed, such that the cluster arrival time
tdigit can be related to the drift time td (see Sec. 8.2.1). The standard TPC offline tracking approach
employs t0 information from external trigger detectors and was used to study the expected performance
of the first reconstruction stage, as described below.

8.3.1 Standard tracking approach

In this section we discuss the application of the standard TPC tracking algorithm in the first online re-
construction stage, where t0 information from external trigger detectors is used. In this tracking scheme,
distortion corrections are applied to all clusters found within a time interval (t0,i , t0,i + td ), where t0,i is the
time of the interaction i, which is recorded in a trigger detector. Here, we assume that all clusters found
within this time interval emerge from the interaction with collision time t0,i . The distortion correction
is performed by employing the scaled average map rscaled . This implies that residual distortions due to
fluctuations on the level of a few percent of the initial distortions will remain. For those clusters which
belong to tracks from the interaction at time t0,i , such residual distortions may reach up to a few mm.
Clusters from interactions which occurred not at t0,i are not properly corrected due to the improper drift
time assumption and form background to this interaction. After tracking of interaction i is finished, the
procedure is repeated for interaction i + 1 at time t0,i+1 .

8.3.2 Performance using corrections from the scaled average map

The performance of the tracking algorithm in this approach is studied using the full microscopic simula-
tion chain described in Sec. 7.3. Residual distortions as expected after correction with rscaled are imposed
to the clusters in the simulation. The space charge is described using the data-driven description given in
Sec. 7.4, assuming e = 20.
Figure 8.5 shows a comparison of the TPC tracking efficiency (left) and the TPC-ITS matching efficiency
(right) without and with space charge distortions. Even with residual distortions applying corrections
7 The number of ion pileup events during t ion is described by a Poissonian with mean N ion ion
d pileup = Rint td ⇡ 8000 at Rint =
50 kHz.
TPC Upgrade TDR 105

from the scaled average map (red points), i.e. in the first reconstruction stage, the efficiency of the TPC
tracking is not affected compared to the ideal case without distortions (blue points). The TPC-ITS
matching efficiency is lower by about 5% at high pT and 2% at low pT (0.5 GeV). This slightly lower
efficiency does not affect the requirements of the calibration, though, and is recovered in the second
reconstruction stage (black points) described in Sec. 8.4.

1 1

TPC-ITS matching efficiency

TPC tracking efficiency

0.995 0.98

0.99 0.96

0.985 0.94

0.98 0.92

0.975 0.9

0.97 0.88

0.965 central event at 50 kHz 0.86 central event at 50 kHz

no space-charge distortions no space-charge distortions

0.96 0.84
space-charge distortions, after the 1st reconstruction stage space-charge distortions, after the 1st reconstruction stage

0.955 space-charge distortions, after the 2 nd reconstruction stage 0.82 space-charge distortions, after the 2 nd reconstruction stage

0.95 0.8
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
1/p (GeV/c)-1 1/p (GeV/c)-1
T T

Figure 8.5: TPC track reconstruction efficiency (left) and TPC-ITS track matching efficiency (right) in Pb–Pb collisions at
50 kHz without distortions and with residual distortions after the first and second reconstruction stage.

The fraction of assigned clusters, defined as the number of associated clusters divided by the maximum
number of assignable clusters in the active region, is shown in Fig. 8.6. It is compared for the ideal case
without distortions (blue points) and with residual distortions using the scaled average space-charge map
(red points) or a high-resolution map (black points). There is no significant modification in the cluster
association when residual distortions are present.
Fraction of assigned clusters

0.94
central event at 50 kHz
no space-charge distortions
0.92
space-charge distortions, after the 1st reconstruction stage

space-charge distortions, after the 2 nd reconstruction stage

0.9

0.88

0.86

0.84

0.82

0.8
-5 -4 -3 -2 -1 0 1 2 3 4 5
q/p (GeV/c)-1
T

Figure 8.6: Cluster-to-track association efficiency in Pb–Pb collisions at 50 kHz without and with residual distortions using the
scaled average map (first stage) and a high-granularity distortion correction (second stage).

This clearly demonstrates that space-charge corrections using time-averaged maps, scaled by the instan-
taneous ion density, provide sufficient precision for efficient online tracking and cluster-to-track asso-
106 The ALICE Collaboration

ciation. This allows for powerful data compression and matching to the external detectors to perform
high-granularity distortion corrections for the second reconstruction stage.
It should be noted that the procedure described above is not optimized in terms of computing speed.
It implies multiple corrections of each cluster in the case of event pile-up, i.e. on average five times at
50 kHz. As a possible improvement, an alternative TPC standalone approach was developed, where an
initial estimate of the collision time, t0seed , is derived from TPC track seeds. This procedure is described
in Sec. 8.5.1.

8.4 Second reconstruction stage

The second reconstruction stage aims to restore the intrinsic detector resolution. Therefore, the main
objective is to reduce the effective residual space-charge distortions to the level of the intrinsic track
resolution of a few 100 µm in rj. Essentially, this implies a correction of the remaining local space-
charge fluctuations, which were not accounted for by the usage of the scaled average map rscaled during
the first reconstruction stage.
Such a residual miscalibration can be directly determined by measuring the TPC cluster residuals with
respect to an external track reference obtained from interpolation between ITS and TRD track segments.
This straightforward calibration procedure relies on the availability and proper calibration of those de-
tectors. It is eventually limited by the available track statistics in a typical calibration interval. The
procedure is described below.

8.4.1 ITS-TRD track interpolation approach

In this approach the TPC volume is subdivided into a number of volume elements (‘voxels’). Each voxel
is aligned by minimizing the mean residual of the TPC clusters within this voxel with respect to external
reference tracks. The number of voxels must be sufficiently large to account for the local variations of the
residual distortions with sufficient granularity. On the other hand, the voxel size determines the statistical
precision that can be achieved within a typical calibration interval. For the present study a total number
of 72,000 voxels in the TPC is assumed, corresponding to a voxel size of 10 cm, 16 cm and 1/72 p in z,
r, and rj direction, respectively. The optimization of the granularity is based on a detailed study of the
residual distortion pattern after the first reconstruction stage. The required precision calls for a certain
cluster statistics within a voxel. The resolution achieved from a single track within a voxel, svxl , is given
by

s ✓ ◆2
s
svxl = str2 + p cl , (8.5)
Ncl

where str is the external track precision, scl the local single cluster resolution, and Ncl the number of
clusters of the track in the voxel.
Figure 8.7 shows the precision of external reference tracks in rj inside the TPC for momenta above
1 GeV/c as a function of r, assuming perfectly aligned detectors. Extrapolation from the ITS and in-
terpolation between ITS and TRD are shown on the upper and middle panel, respectively. The ITS
extrapolation uncertainty ranges from < 1 mm at the inner radius to ⇠ 1.5 cm at the outer radius of the
TPC. For high-momentum tracks the extrapolation uncertainty is  1 mm for all radii. Using the ITS-
TRD interpolation the precision is always better than ⇠ 0.8 mm for momenta above 1 GeV/c and falls
below 100 µm for high-momentum tracks.
The track interpolation approach was studied using a fast MC8 , assuming perfectly aligned detectors.
8 Monte Carlo (MC)
TPC Upgrade TDR 107

1.6

(cm)
1.4 ITS extrapolation
1/pT = 1 1/pT = 0.4

ITS
1.2

r
1/p = 0.9 1/pT = 0.3
T
1 1/pT = 0.8 1/pT = 0.2
0.8 1/pT = 0.7 1/pT = 0.1
0.6 1/pT = 0.6 1/pT = 0
1/pT = 0.5
0.4
0.2

(cm)
0.08 ITS-TRD interpolation
0.07
ITS+TRD
0.06
r
0.05
0.04
0.03
0.02
0.01
ITS

0.45
Ratio
r

0.4
/
ITS+TRD

0.35
0.3
r

0.25
0.2
0.15
0.1
0.05
0
80 100 120 140 160 180 200 220 240 260
r (cm)
Figure 8.7: Precision of external tracks as a function of the radius inside the TPC. The color scale represents different values of
1/pT . Top: Extrapolation error of ITS tracks; center: Interpolation error for ITS-TRD tracks; bottom: Ratio of the
two.

In this MC, tracks from interactions at different t0 are propagated through the detector, creating track
points along their trajectory. Those track points are distorted according to the expected space-charge
distortions for e = 20 and smeared with the intrinsic cluster resolution (⇠ 1 mm in local-y and z). Typical
space-charge density fluctuations are considered. In the reconstruction step, the distorted space points
were corrected using the scaled average map rscaled . A realistic parametrization of the charged-particle
momentum distribution based on measurements was used to generate the tracks. The simulated track
statistics corresponds to 5 ms of data taking at 50 kHz, i.e. tracks from 250 minimum bias Pb–Pb events.
This corresponds to a typical calibration interval over which the space-charge density can be considered
static (see Sec. 7.4.5). For the analysis only tracks reaching the TRD detector were used.
The measured local distortion drj 0 is correlated with the radial distortion dr, if the track crosses the
padrow under an inclination angle, see Fig. 8.8 (left). In this case, the true distortion drj and the radial
distortion can be extracted employing a linear relation:

drj 0 = drj + dr · tan a (8.6)

where a is the local track inclination angle. An example of such a fit is shown in Fig. 8.8 (right).
Figure 8.9 shows an example of the comparison between the measured residual distortions as determined
by the ITS-TRD interpolation method (points) and the real residual distortions from MC (line). The
comparison is shown in the region of smallest radius and largest drift length, i.e. 86 < r < 102 cm and
0 < z < 10 cm, where the residual distortions are largest. Each data point corresponds to the fit result
in a single voxel and is used as a local correction to all space points within this voxel. The pattern of
the residual distortions is well described by the interpolation method. The momentum resolution after
application of this correction will be presented in Sec. 8.4.2.
108 The ALICE Collaboration

dr ϕ ’
1.5

0.5

-0.5

-1

-1.5

-2
-0.3 -0.2 -0.1 0 0.1 0.2 0.3
tan(α)

Figure 8.8: (Left) Illustration of the measured rj distortions being composed of the real rj distorions and the radial distortions,
shown for three example tracks crossing a pad row (grey area) under different local track inclination angles a.
(Right) Measured correlation between drj 0 and tan(a) (see text).
〈dr ϕ〉 (cm)

〈dr 〉 (cm)
0.1 2
Measured residuals Measured residuals
Real residuals Real residuals
0 1.5

-0.1 1

-0.2 0.5

-0.3 0

-0.4 -0.5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
sector sector

Figure 8.9: Comparison of measured and real residual distortions in rj (left) and r (right) for the region with the largest residual
distortions (86 < r < 102 cm and 0 < z < 10 cm).

In Fig. 8.10 the resolution of the interpolation method is shown as the difference between the measured
and the real residual distortions. The results are integrated over the full TPC acceptance.
In rj the remaining residual distortions are ⇠ 300 µm, i.e. compatible with the intrinsic track resolution.
The remaining distortions in radial direction are ⇠ 1.3 mm.
It should be noted that the distortion vectors in nearby voxels are correlated. Exploiting the knowledge of
the functional dependence of the residual distortions on variations of the space-charge density will allow
to constrain the extracted distortion correction, and thus further improve the precision of the method and
relax the required statistics per voxel.
In addition, the statistics at low radii, where the distortions are largest, can be enhanced by adding tracks
that do not reach the TRD. By also using the ITS extrapolation method the precision of the measurement
will be improved, in particular in this region.

8.4.2 Momentum resolution after residual correction

In order to verify the final tracking performance after application of the high-granularity corrections
derived from the ITS-TRD interpolation method, the remaining residual distortions were mapped and
used as input for the microscopic MC. The results presented in the following are for central Pb–Pb
collisions at a minimum bias collision rate of 50 kHz.
Figure 8.5 (in Sec. 8.3.2) shows a comparison of the TPC tracking efficiency (left) and the TPC-ITS
matching efficiency (right). The results are shown for the ideal scenario without space-charge distor-
TPC Upgrade TDR 109

3 3
×10 ×10
200
Measured - Real residuals (RMS: 292µ m) 100 Measured - Real residuals (RMS: 1.30mm)
180

160
80
140

120
60
100

80 40
60

40 20
20

0 0
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
∆ 〈drϕ〉 (cm) ∆ 〈dr〉 (cm)

Figure 8.10: Distribution of measured minus real residual distortions for all fluctuation scenarios integrated over full acceptance
of the TPC. (Left) rj-distortions. (Right) r-distortions.

tions (blue points) and including the expected residual distortions after the first reconstruction stage
(red points) and after the second reconstruction stage (black points), i.e. after application of the high-
granularity corrections derived from the ITS-TRD interpolation method as described above.
The TPC tracking efficiency is not affected by the residual space-charge distortions, even when only the
scaled average correction map is applied (first reconstruction stage). The TPC-ITS matching efficiency
is about 1% lower compared to undistorted tracks, after application of the high-granularity space-charge
corrections using the interpolation method. A better tuning, taking into account the measured residual
distortions, will further improve the matching. For residual distortions as expected during the first recon-
struction stage, the matching efficiency is further reduced, however this does not significantly affect the
requirements for the subsequent calibration step in the second reconstruction stage.
Figure 8.11 shows a comparison of the transverse momentum resolution for undistorted tracks (left panel)
as well as for tracks with residual distortions expected after the first (middle panel) and second recon-
struction stage (right panel). After the first reconstruction stage, the residual distortions lead to a dete-
rioration of the pT resolution by about a factor ⇠ 1.5 – 2 as compared to the ideal case. A significant
improvement is achieved after application of the corrections derived in the second reconstruction stage.
While the TPC standalone resolution after the second stage is still somewhat lower than in the ideal case,
the detector resolution of the undistorted case is practically fully restored if combined TPC-ITS tracks
are considered.
(GeV/c)-1

central events at 50 kHz

0.02
TPC standalone
T

TPC constrained
1/p

TPC + ITS
0.015

0.01

0.005

no distortions residual distortions, after 1st stage residual distortions, after 2nd stage
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
1/p (GeV/c)-1 1/p (GeV/c)-1 1/p (GeV/c)-1
T T T

Figure 8.11: Comparison of the momentum resolution without distortions (left) and with residual distortions after the first
(middle) and second (right) reconstruction stage in Pb–Pb collisions at 50 kHz.

In summary, the voxel alignment approach employing interpolation of track segments from the ITS and
110 The ALICE Collaboration

TRD detectors provides a method to minimize the remaining residual distortions and to restore the de-
tector performance without distortions. It is expected that further optimizations of the described method
will improve the results. This includes further optimization of the voxel size to account for the radial
dependence of the reference track resolution, of the hit density and of the residual distortion magnitude
and variation. Further improvement is also expected by considering the correlations of the distortions in
consecutive calibration time intervals. The current measurement approach described in Sec. 8.5.2 will
lead to an improved description of the local space-charge density fluctuations. Application of such an
improved space-charge map before the ITS-TRD interpolation method will considerably reduce the local
variations of the distortions. This will allow for a larger voxel size and thus for an improved statistical
precision of the correction.

8.5 Further optimizations

In addition to the studies presented above, further improvements of the online pattern recognition as well
as the space-charge distortion correction can be achieved following the concepts described below.

8.5.1 TPC standalone tracking approach

In order to improve computational aspects of the online reconstruction, an alternative tracking approach
which does not rely on t0 information from external detectors was studied. To this end, a “fast” MC
simulation was developed, that allows to vary the essential parameters and to develop and validate the
reconstruction strategy. In this MC, tracks from interactions at different t0 are propagated through the
detector, creating track points along their trajectory. Those track points are distorted according to the
expected space-charge distributions assuming 50 kHz Pb–Pb collisions and e = 20 (see Sec. 7.4) and are
smeared with the expected intrinsic cluster resolution (⇠ 1 mm in local-y and z).

Seeding
In this approach, seeding is the first online tracking step after the cluster finding. A track seed is defined
as a set of several clusters being close in space. The seeding procedure developed for the current TPC,
described in detail in [8], is used for these studies. It starts with a pair of points in pad rows (k) and
(k n). Clusters that are found within a search road between the starting points are associated to the
seed. The seed is rejected if less than n/2 clusters are found.
In order to maximize the efficiency of the seeding step, a coarse space-charge distortion correction is
applied to each cluster. It is based on the scaled average space-charge map, rscaled , not taking into
account topological fluctuations. Moreover, the t0 of the corresponding interaction is not known at this
stage. It is thus assumed, that all clusters belong to a track that has about half of the maximum drift
length, i.e. |h| = 0.45. Under this assumption, a hypothetical zro -position can be assigned to each cluster
and allows an ad-hoc distortion correction following Eq. (8.4). In this way, the maximal distortions are
reduced to about one half which improves the seeding efficiency. It should be noted that seeding is
performed in regions where the distortions are not largest, and that distortions vary only slowly over the
length of the seed.

Initial t0 estimate
An initial estimate of the time of the interaction can be derived from a track seed by extrapolating it in the
x, y,t space to the interaction region (x = y = 0). This procedure is schematically depicted in Fig. 8.12.
The calculated time from the extrapolation, textrapol , can be associated with the time of the interaction,
t0seed , using the relation z = zroc vd td (from Eq. (8.2)):

zroc zvtx
t0seed = textrapol . (8.7)
vd
TPC Upgrade TDR 111

seed

drift

extrapolation

zroc
z=0 zvtx , t0 z, time
Interaction point
Figure 8.12: Schematic of the seeding procedure and t0 estimation.

Here, zvtx is the z-position of the interaction vertex which can be approximated by zvtx ⇡ 0. In this case
the extrapolated time at the interaction region textrapol is one full drift time (td = zroc /vd ) after the t0 of the
interaction:

zroc
t0seed ⇡ textrapol . (8.8)
vd

This assumption leads to an irreducible uncertainty in t0seed due to the spread of the collision vertices
around the nominal interaction point (sz = 7 cm corresponding to a drift time of 2.7 µs), to which the
finite precision of the distortion correction adds.
Figure 8.13 shows the track-by-track distribution of the difference between the estimated t0seed and the
true interaction time t0 , multiplied by the drift velocity. The distribution has a width of about 5.2 cm. In
the simulation only events with vertices within 10 cm of the nominal interaction point were kept. The
RMS of the simulated event vertices is 4.7 cm, which implies that the precision of t0seed is dominated by
the irreducible contribution from the vertex spread.
#tracks

12000 Mean -0.3443

RMS 5.187
10000

8000

6000

4000

2000

0
-40 -20 0 20 40
(t seed
0
-t 0) ⋅ v (cm)
d

Figure 8.13: Deviation of t0seed from the real t0 , multiplied by vd , for a space-charge scenario at e = 20 in the fast MC.

Cluster-to-track association
The cluster-to-track association is an integral part of the first stage of the reconstruction. Starting from
track seeds, the track parameters are extrapolated inwards (smaller radii) and outwards (larger radii).
Then, within a road in the z and local-y directions, clusters are searched that are close to the track
extrapolation.
The efficiency of this procedure suffers from the coarse distortion corrections in the first step, in particular
in regions where the distortions are large, i.e. at small r and z. Therefore, the search road of the tracking is
112 The ALICE Collaboration

modified according to the local distortions estimated from t0seed . The t0seed resolution presented in Fig. 8.13
implies that the uncertainty on the cluster drift time can be reduced from the maximum (about ± 50 µs to
only a few microseconds (corresponding to the width of Fig. 8.13). Consequently, the distortions can be
corrected to the level of a few percent using t0seed , assuming that the distortions scale approximately linear
with the drift time. On this level, (1 mm), the remaining distortions do not affect the cluster-to-track
association.
It should be noted that at this point of the reconstruction the remaining distortions could be improved by
matching the track t0seed to the closest t0 from the list of collision times from an external trigger detector.
Given the t0seed resolution of ⇠ 2.7 µs quoted above, unambiguous matching to the proper interaction t0
is possible if no other collision occurred in the same bunch train crossing. This is the case for ⇠ 83 %
of the collisions, see Fig. 8.2. In the remaining cases, the achievable precision will be limited to that of
t0seed , as long as no external tracking information from ITS is included.

Matching with external detectors

A key issue for the calibration and reconstruction is the matching of TPC tracks to external detectors,
mainly ITS and TRD. Here, we focus on the implications for ITS matching since the occupancies and
distortions at smaller radii are much larger, and therefore, more demanding for the internal track calibra-
tion.
For the matching with the ITS we assume that a standalone ITS tracking will be performed based on
the ITS hits. This will allow a matching on the track level, in terms of the following track parameters:
the local-y and z position, the sine of the inclination angle in the bending plane, the tangent of the dip
angle of the track, and the curvature expressed as 1/pT . In addition, by propagating the ITS track, a
comparison of the track parameters at the inner wall of the TPC rather than at the ITS can be performed.
This allows a better matching precision since the uncertainties on the track points of the ITS are much
smaller than for the TPC. Therefore, an extrapolation of the ITS track towards the TPC is more precise
than vice versa.
#tracks

#tracks

no distortions (ideal)
3000 Mean -5.133 Mean 6.92e-05
50000 distorted/corrected (t )
RMS 4.339 distorted/corrected (t
0
seed
)
0 RMS 0.02831
2500
40000 Mean -0.02731
2000 RMS 0.1599
30000 Mean -0.01423
1500
RMS 0.2137
20000
1000

500 10000

0 0
-15 -10 -5 0 5 -1 -0.5 0 0.5 1
y -y (cm) y -y (cm)
TPC ITS TPC ITS

Figure 8.14: Matching of the local-y coordinate of ITS and TPC tracks. (Left) For tracks from undistorted clusters at the
outermost ITS layer (red) and the inner wall of the TPC (black). (Right) For tracks from fully distorted clusters
without correction (at the inner wall of the TPC).

The challenge of matching TPC tracks distorted due to space-charge with ITS information is manifest
in Fig. 8.14 (left panel). It shows the local-y matching built from distorted clusters without performing
any correction on the distortions. The shift and asymmetry of the distribution arises from an interplay
between charge and curvature of the track and the E ⇥ B effect. The deviations are large, which makes
unambiguous track matching at high occupancies impossible.
Figure 8.14 (right panel) shows the local-y matching of ITS tracks and TPC tracks built from distorted
clusters and corrected with the z-position estimate obtained from t0seed (red histogram). The RMS of this
TPC Upgrade TDR 113

distribution is about 2.1 mm, being half the width of the readout pads of the inner readout chamber. In
comparison to Fig. 8.14 (left panel), a strong improvement of the matching precision is observed. This
leads to a large reduction of the combinatorial background in the TPC-ITS matching procedure. Note
that this resolution is a result of the first online reconstruction stage based entirely on TPC information.
Matching with the proper ITS track will provide t0 information and thus further improvement of the
space-point distortion corrections, as shown in the same figure (green histogram). In addition, the match-
ing is compared with the intrinsic resolution (black histogram). For performing the final matching, all
possible TPC-ITS track combinations within a matching window are formed, each time refitting the TPC
track parameters using the correct t0 information delivered by the ITS for the z-position estimate. By
using all five track parameters for the matching, the number of candidates can be reduced further.
After this step, the residual distortions of the TPC clusters are dominated by the space-charge fluctua-
tions, which are not accounted for by using the scaled space-charge map rscaled for distortion correction.
As discussed in Sec. 8.5.2, the quality of the employed space-charge distortion maps can be further im-
proved by making use of the measured currents with high temporal and spacial granularity. Consequently,
also a large improvement on the matching precision is expected.

8.5.2 Space-charge calibration by current measurements

For the first reconstruction stage it is foreseen to base the space-charge distortion correction on the scaled
average space-charge density map, where the total space charge integrated over the ion drift time, tdion ,
of 160 ms is used for scaling (see Sec. 8.3). This method can, however, be refined by using the detailed
ion density distributions in time and space. Using such a procedure would yield a much more precise
description of the space-charge distortions. The idea is described in the following.
The space-charge density is proportional to the ‘signal current’ Iroc at the readout chambers multiplied
by e:

rsc ⇠ Iroc · e . (8.9)

Two methods are considered to measure Iroc :

– Hardware-based method: The currents on the GEM HV sectors are proportional to the charge
locally produced by gas amplification. They can be measured with high temporal granularity (see
Secs. 4.4, 10.3, and 11.4.1). A similar current measurement is currently being prepared for RUN 2
for a subset of the HV channels, albeit for a different purpose (current spike detection for the data
taking with the TPC in gated mode).

– Software-based method: The cluster charges in the raw data are proportional to the current Iroc .
They offer the ultimate granularity in space and time required to derive a high-resolution space-
charge map. The data are available at the input to the online farm on the FPGAs which perform
the cluster finding.

From either, or a combination of these methods, it will be possible to follow the amount of ions created in
space and time with high granularity. An integration over the full drift of the ions allows the calculation
of the present space-charge density maps rscaled and rhigh res .
From any space-charge density map rsc , obtained e.g. from the current measurement approach, the distor-
tions can be derived analytically. However, with the methods currently available this will not be possible
on the required time scale of a few ms. To overcome this limitation, the following method is foreseen:
114 The ALICE Collaboration

The actual space-charge distortion map, ~D, is estimated by performing a Taylor expansion of a reference
distortion map, ~Dref , obtained for a reference space-charge distribution9 , rref , over volume cells i:

~D = ~Dref + Â ∂ Dref d rsc

~
i
, (8.10)
i ∂rsc
i

where ∂ /∂rsci is the partial derivative with respect to a change of the space-charge density in a volume
i = ri
cell i and d rsc i
ref rsc the variation of the space-charge density in cell i. The local derivatives can be
pre-calculated analytically and stored in lookup tables for fast access. Together with the measured actual
space charge density rsc , this will allow a fast calculation of the present space-charge distortion map ~D.
For the expected space-charge densities, the distortions do not scale linearly with the average space-
charge density. Therefore, if the deviation of hrsc i is large, (10 %), compared to hrref i, ~Dref needs to be
updated. This can happen e.g. due to a change in luminosity. Such updates are expected at most on the
level of (1 10 min), making feasible the analytic recalculation of ~Dref .

8.6 Additional calibration requirements, monitoring, and quality control

This section summarizes the additional calibration steps that have to be applied to the TPC data and
discusses the requirements for online monitoring and quality control.

8.6.1 Additional calibration requirements

As most of the TPC calibration methods have been already developed for RUN 1, they are only briefly
mentioned in this section. According to their characteristic dependence on ambient conditions over time,
they are divided in time-independent and time-dependent calibrations.

Time-independent calibrations
Time-independent calibrations are characterized by a stable behavior over long periods. The pedestal
values and zero suppression thresholds for the individual front-end electronics channels, as well as the
map of dead channels, only change on time scales of the order of weeks or longer.
A proper calibration of the effective gain of the GEM readout system is mandatory for providing particle
identification with best quality. The gain needs to be corrected for time-independent pad-by-pad and
chamber-by-chamber variations. The retrieval of these calibration parameters is done typically once
per year based on the Krypton calibration method, originally developed by the ALEPH [9, 10] and
DELPHI [11] collaborations. The decay clusters of radioactive 83 Kr, which is released into the TPC gas,
are analyzed and allow the extraction of the mean gain per readout chamber and the relative gain of each
readout pad with respect to the mean. The method was already successfully used for the gain calibration
of a GEM based TPC [12, 13].
An additional method for pad-by-pad gain calibration is based on charged-particle tracks. The relative
pad gain factors can be retrieved from large statistics samples of clusters from a selected sample of
tracks with well-defined mean energy loss, e.g. MIPs. This method allows a higher granularity in time,
but needs additional attention when running in an environment of high space-charge distortions, due to
the compression of tracks in radial direction and the merging of ionization clusters.

Time-dependent calibrations
The calibration of the drift velocity vd and the gas gain have a time-dependent component that is con-
nected to changes of the ambient conditions and of the gas composition. Moreover, the average space-
9 This could be e.g. be the long term average map, rav , see Sec. 8.2.3
TPC Upgrade TDR 115

charge map has to be updated periodically in order to account for slow variations of luminosity, ambient
conditions, and for malfunctioning sectors.
Different methods for determining the drift velocity and gain are described below. An update of the
corresponding calibration parameters every 15 min has proven sufficient in the calibration scheme for
RUN 1. Between updates, a simple scaling based on pressure and temperature changes is performed.
Such a scaling can be used also for RUN 3, but also other possibilities are investigated.

Drift velocity The variations of the drift velocity vd in time are mainly induced by changes of the
ambient conditions (pressure and temperature) and of the gas composition. Two methods for determining
vd have been routinely used in RUN 1.
Laser measurements yield a very robust estimate for the drift velocity, which is independent of reference
detectors. Laser events are triggered and, therefore, the t0 of the laser event is known. The arrival time
of photo-electrons emitted from the central electrode (CE), which are created by scattered laser light
inside the TPC, determines the drift time for the full drift length. The method allows to extract local drift
velocity gradients and thus to monitor local temperature variations in the gas [14].
Track matching with external detectors (in particular the matching with ITS tracks) allows the deter-
mination of the drift velocity with high granularity in time10 . The drift velocity can be determined
using a Kalman-filter approach, fitting the differences in the z-positions of the TPC and ITS track. The
method is also used for the other track parameters and allows to simultaneously fit the distortions (see
also Sec. 7.4.5).

Readout chamber gain Like the drift velocity, the gas amplification is influenced by the ambient
conditions. The variations of the effective gain in the readout chambers can be followed using tracks
with a constant dE/dx. Minimum ionizing (MIP) pions or electrons are used for this purpose. Pions
are abundant, the MIP region is easy to identify, and the signal is clean from contamination. Electrons
can be extracted using topological reconstruction of conversion photons in the detector material. The
advantage of electrons is that their energy-loss is about 50 % higher than that of MIPs, and therefore,
usually less prone to threshold effects. However, it requires the reconstruction of secondary vertices and
thus they need either longer integration times or yield lower precision. Methods using both particles will
be applied for gain calibration of the upgraded TPC on the readout chamber level, allowing to follow the
gain variations.

8.6.2 Monitoring and quality control

The concept of online Quality Control (QC) combines continuous monitoring and control of the data
in order to ensure prompt reaction on data quality issues already during the data taking. The specific
monitoring tasks are based on the solutions developed already in RUN 1. In addition, the monitoring of
the space-charge distortions and their correction must be an integral part of this strategy.
In case of problems with the quality of the reconstrucion, the basic strategy has been to restart the recon-
struction process (i.e. start a new reconstruction pass) for the data collected in RUN 1. This procedure
is very successful in order to utilize a maximum of recorded data for physics analysis, but can not be
transferred to RUN 3, due to the online reconstruction paradigm. Therefore, an optimized concept for
data quality monitoring is needed for RUN 3, which is discussed below.

10 In the RUN 1 calibration scheme laser measurements were used as an initial estimate for the reconstruction. However, track

matching was the main source of the drift velocity calibration.

116 The ALICE Collaboration

Monitoring
Different kinds of input data will be used for the QC in RUN 3. These data comprise not only different
levels of reconstructed data and global, environmental observables, but also the calibration parameters
produced online. The key feature of QC will be the time-wise trending of those observables. This has
been demonstrated to be very successful during RUN 1 in the offline QA, as well as HLT online QA on a
run-by-run basis.
The observables, which have been already defined, will be used also for RUN 3, as they have been proven
to allow a good and comprehensive judgment of the performance of the TPC. A list of observables to be
monitored is given here:

– External Parameters such as as instantaneous and integrated luminosity, beam-background, am-

bient pressure, and temperature, which slowly change in time.

– Global Event Properties such as data sizes per interaction or per time frame, track multiplicities,
and compression factors.

– Calibration Objects such as alignment stability, drift velocity, gain stability, and space-charge
distortions.

– Cluster Parameters such as cluster charge and width, as well as fraction of clusters associated to
tracks from interactions.

– Track Parameters such as number of clusters associated to tracks, pseudo-rapidity, angular, and
momentum distributions, as well as hdE/dxi signal and dE/dx resolution for MIPs. Furthermore,
also distances-of-closest-approach to the primary vertex, as well as the track matching efficiency
to the external detectors, such as ITS, TRD, and TOF belong to this category.

– Advanced Physics Observables such as fits to the invariant mass distributions of V0 particles (K0s
and L), fits to the transverse momentum distributions, as well as parameters of tracks from cosmic
particles within normal collisions events. Finally, also the particle identification performance can
immediately be verified.

Space-charge calibration monitoring

The TPC laser system is an ideal tool to monitor the quality of the space-charge calibration because it
provides reproducible straight ionization tracks at known positions. In the TPC laser system the laser
light is split in several beams through an optical system of semi-transparent beam splitters, micro mirrors
and prisms, and guided into the TPC at different entry points through quartz windows (see Sec. 11.4.4
and [15]). The individual position and inclination angle of each micro mirror has been measured and is
available in a database. Laser tracks can thus be easily associated to a micro mirror from the different
angles and z-positions of the reconstructed track parameters. Figure 8.15 shows simulated laser tracks
with the space-charge distortions expected for e = 20 at 50k̇Hz. Together with the different layers along
the z-axis such data allow to verify the quality of the extracted correction maps with sufficient precision.
Therefore, laser data are ideally suited for quality assurance purposes.

Quality control
The calibration parameters applied in the online reconstruction process have to be closely monitored and
validated. This will allow to identify wrongly calculated values or even outliers in a trend, in cases where
a fit to data might have failed. A similar validation step has been already applied during RUN 1 before
the created calibration objects have been uploaded to the Offline Conditions DataBase (OCDB).
TPC Upgrade TDR 117

y (cm)
250

200

150

100

50

0
0 50 100 150 200 250
x (cm)

Figure 8.15: Simulated laser event. Black points indicate undistorted clusters, red points show distorted clusters for e = 20 at
50 kHz Pb–Pb data taking. (Left) All laser tracks are shown. (Right) Laser tracks from one micro-mirror bundle
are shown.

In order to establish these monitoring and control layers, a high level of automation is required not only
to allow fast reaction, but also to minimize manual interventions by shift personnel. Automatic alarms on
basis of smaller subsets of key trending histograms will be constantly available, in order to inform about
potential problems. A more detailed set of observables has to be collected at the same time in order to
allow the experts to react quickly in case of issues with the quality of the data or the performance of the
online reconstruction. Predefined procedures for fast detection of error conditions are crucial to ensure
a prompt reaction. Actually, error conditions could be anticipated using predictions based on trending
information on changing detector behaviour and/or conditions.
118 The ALICE Collaboration
Chapter 9

Alternative R&D options

This chapter describes the R&D status of alternative options for the upgrade of the TPC readout.
The baseline proposal for the design of the readout chambers of the upgraded TPC foresees a quadruple
GEM stack with asymmetric field configurations, as described in Chap. 4. This approach fulfills the
requirements listed in Chap. 1. In parallel, alternative R&D efforts have been made to study and develop
options to further improve the detector performance. The following key issues have been addressed:

– minimization of the ion backflow,

– minimization of the pileup of tracks from different events and

– preservation or improvement of the space-point, momentum, and dE/dx resolutions.

In particular, COBRA GEMs, perspectives of using a different gas mixture containing CF4 in combina-
tion with chevron-shaped readout pads, and a gas amplification option based on 2 GEMs and one MMG1
were studied in detail.

9.1 R&D with COBRA GEMs

A COBRA GEM is a GEM with a patterned electrode on the surface, which helps to block back-flowing
ions very efficiently [1, 2]. To characterize the properties of COBRA GEMs in a TPC application, a
comprehensive R&D study was performed.
Three kinds of COBRA GEMs have been developed in collaboration with SciEnergy Co. Ltd [3]. Ta-
ble 9.1 summarizes their specifications. In addition, two types of standard GEMs were used for the
present measurements, which are also listed in Tab. 9.1. Figure 9.1 shows a photograph of COBRA 1.

thickness (µm) hole size (ø) (µm) pitch (µm) rim size (µm) insulator
COBRA 1 400 300 1000 100 FR5
COBRA 2 200 150 500 50 FR5
COBRA 3 100 100 400 0 LCP
GEM 50 50 70 140 0 LCP
GEM 100 100 70 140 0 LCP
Table 9.1: Geometries of COBRA GEMs and standard GEMs used for the measurements.

1 MicroMegas: Micro-Mesh Gaseous Structure (MMG)

119
120 The ALICE Collaboration

Figure 9.1: Photograph of COBRA 1 showing the GEM (around the holes) and COBRA electrodes.

Figure 9.2: (a) COBRA GEM unit cell built in the simulation program ANSYS (b) ion drift lines in a COBRA GEM with three
different potential differences DUAC between GEM electrode (A) and COBRA electrode (C). Image from [1].

COBRA 1 and COBRA 2 are based on glass epoxy laminate (FR5) as insulator and 6 µm Cu layers,
covering an active area of 3 ⇥ 3 cm2 . The holes are pierced by a mechanical drilling technique. The
additional pattern on the top and bottom surfaces of the GEMs are produced by wet etching. Clearances
from the edge of the hole (rim) of 100 µm and 50 µm, respectively, are introduced to protect from
discharges. However, charge-up of the FR5 insulator at the rim occurs, which negatively affects the gas
gain and long term stability. To avoid this effect, a new production method for a COBRA GEM with
standard thickness (100 µm) and standard insulator (LCP2 ) without rim was started3 (COBRA3).
Figure 9.2 shows a COBRA unit cell built in the simulation software ANSYS and drift lines of ions at
different potential differences (DUAC = UGEM UCOBRA of 0, 120, and 180 V) between GEM electrode
(A) and COBRA electrode (C). The reduction of the ion backflow as the potential DUAC increases is
evident.

2 Liquid Crystal Polymer (LCP)

3 The production of COBRA3 is still underway at the time of writing this TDR, such that no test results are available yet.
TPC Upgrade TDR 121

9.1.1 Characterization of single COBRA GEMs

Measurements of gas gain and ion backflow with a single COBRA 1 and COBRA 2 have been performed
at different voltages UGEM and DUAC . All measurements have been performed in Ne-CO2 (90-10). Fig-
ure 9.3 shows the schematics of the measurement setup. The length of the drift volume is 3 mm. X-rays
are emitted into the detector from the top of the chamber. Primary electrons created between shield and
mesh (cathode) do not contribute to the multiplication of electrons, whereas the electrons created be-
tween the mesh and the GEM (drift region) enter the multiplication region. The currents on the mesh and
on the pad anode are read by picoammeters (ADC R8240 x2 or KEITHLEY 6487), while the currents
on the GEM electrodes are read out by high voltage power supply modules (CAEN N1471) with 0.5 nA
resolution. The tube current of the X-ray generator is varied between 0.03 and 3 mA, which results in a
current on the readout pad anode of between around 1 and 100 nA at a gas gain of ⇠ 2000 in Ne-CO2
(90-10). Since the expected current density in 50 kHz Pb–Pb collisions is of the order of 1 nA/cm2 , a
tube current 0.3 mA was found to be best suited for the measurements.

Figure 9.3: Schematic setup for the measurement of gas gain and ion backflow with an X-ray source.

0
The ion backflow is measured as IB = (Icathode Icathode )/Ianode , where Icathode is the current at the mesh,
0
Icathode is the current from primary ions and Ianode is the current at the readout pad plane. The current
from primary ions is measured at a drift field Edrift = 0.4 kV/cm applied both between shield and mesh,
and between mesh and the top of the GEM electrode. No voltage difference is applied across the GEM
0
for the measurement of Icathode . Note that the sign of the electric field flips at the mesh. This means that
ions from both sides of the mesh contribute to the primary current, but not electrons.
0
The effective gain is estimated by the ratio Ianode /(Icathode /2): the primary electron current contributing
0
to the gas amplification is half the primary current Icathode measured at the mesh. The electric fields in
the drift and induction regions have been kept at 0.4 kV/cm and 3 kV/cm, respectively.

Dependence on DUGEM

Figure 9.4 shows the effective gas gain and ion backflow of GEM 50, GEM 100 and COBRA 2 with
DUAC =0 V. The electric fields in the drift and induction regions are kept at 0.4 kV/cm and 3 kV/cm,
respectively. The observed ion backflow for GEM 50 and GEM 100 is about 20 – 30 %. The ion backflow
with COBRA 2 is ⇠ 40 – 50 % and gradually decreases for gas gains larger than 200.
122 The ALICE Collaboration

Effective Gain 104 1

IBF
103

102 10-1

1 GEM50, d=3mm, Ne-CO2(90-10)

10 1 GEM100, d=3mm, Ne-CO2(90-10)

COBRA2, d=3mm, Ne-CO2(90-10)

1 10-2
0 100 200 300 400 500 600 700 0 100 200 300 400 500 600 700
UGEM (V) UGEM (V)

Figure 9.4: Effective gain (left panel) and ion backflow (right panel) in Ne-CO2 (90-10) as a function of the voltage across the
GEM for GEM 50 (open circles), GEM 100 (open squares) and COBRA 2 (closed circles). The voltage across the
GEM and COBRA electrodes DUAC is kept at 0 V.

Dependence on DUAC

The effective gain and ion backflow of COBRA 2 as a function of DUAC were measured in Ne-CO2
up bot tend
(90-10) for DUGEM = 575 V and 390 V. Figure 9.5 shows that positive DUAC and negative DUAC
to increase the effective gain, which may be related to an improved efficiency for the collection of the
primary electrons into the GEM holes and for the extraction of avalanche electrons into the induction
region. The ion backflow depends both on DUGEM and DUAC and is thus correlated with the effective
gain. While the dependence of the ion backflow on DUGEM is rather weak, a factor of 2 – 3 within the
up
range of this study, the ion backflow drops by more than an order of magnitude if DUAC is increased
to 200 V. It should be noted that an ion backflow of 1 % or less can be achieved with a single COBRA
GEM, albeit at rather large gas gains.

104 1
Effective Gain

IBF

103

10-1

102

10-2

10 up
COBRA2, UAC = U AC: UGEM=390
up
COBRA2, UAC = U AC: UGEM=575
bot
COBRA2, UAC = U AC : UGEM=575

1 10-3
-250 -200 -150 -100 -50 0 50 100 150 200 250 -250 -200 -150 -100 -50 0 50 100 150 200 250
UAC (V) UAC (V)
up
Figure 9.5: Effective gas gain (left panel) and ion backflow (right panel) of COBRA 2 in Ne-CO2 (90-10) as a function of DUAC
(circles) with DUGEM = 575 V (open circles), DUGEM = 390 V (closed circles) and DUAC bot with DU
GEM = 575 V
(open squares). The drift length is 3 mm, Edrift is 0.4 kV/cm in all measurements.
TPC Upgrade TDR 123

9.1.2 Triple structures with COBRA and standard GEMs

Studies of triple GEM systems including one or two COBRA GEMs were performed. It was found that
a COBRA GEM placed as the first GEM layer is not effective for ion backflow suppression. This result
was confirmed in electrostatic simulations and is the consequence of the large distance between the GEM
surface and the drift electrode. Therefore, only configurations with COBRA GEMs in the second or third
layer are considered in the following.
The effective gain and ion backflow are studied for the triple GEM configurations (GEM 1 – GEM 2 –
GEM 3): GEM 50 – COBRA 2 – GEM 50 and GEM 50 – COBRA 2 – COBRA 2. In these measurements,
up
the dependence on DUAC , on the second transfer field ET2 , and on the X-ray tube current were studied.
The drift length and the transfer gaps all measure 3 mm. The electric fields Edrift and ET1 are set to
0.4 kV/cm, while Eind is 3 kV/cm. The voltage across GEM 1 is set to 200 V. The X-ray tube current is
varied from 0.03 mA to 3 mA.

Standard GEM – COBRA GEM – Standard GEM

Figure 9.6 shows the gas gain and the ion backflow measured with the GEM 50 – COBRA 2 – GEM 50
up
setup as a function of DUAC on COBRA 2. The measurement is performed with DUCOBRA2 = 400 V,
DUGEM3 = 300 V and ET2 = 0.75 kV/cm. The results were obtained with X-ray tube currents of 3 mA,
0.3 mA and 0.03 mA. The dependence of the gas gain on DUAC and on the X-ray tube current is weak.
The ion backflow depends strongly on DUAC , as already observed in the single GEM setup. A signifi-
cant dependence on the X-ray tube current is also observed, indicating the importance of space-charge
effects for tube currents above 0.3 mA. At small tube currents, values of the ion backflow below 1 %
can be achieved. Note that the COBRA GEM in the triple GEM setup is operated at rather low gain
(DUCOBRA2 = 400 V, see also Fig. 9.5).

104 1
Effective Gain

IBF

103 10-1

102 10-2

X-ray tube current = 0.03 mA

X-ray tube current = 0.3 mA
X-ray tube current = 3 mA
10-3
0 50 100 150 200 250 300 0 50 100 150 200 250 300
UAC (V) UAC (V)

Figure 9.6: Gas gain (left panel) and ion backflow (right panel) measured with the GEM 50 – COBRA 2 – GEM 50 configuration
up
as a function of DUAC on COBRA 2 for different X-ray tube currents. The second transfer field is ET2 = 0.75 kV/cm.

up
For the same setup, the gas gain and ion backflow as a function of DUAC on COBRA 2 is shown in
Fig. 9.7 for different transfer fields ET2 in the range 0.25 – 1 kV/cm. The voltage on GEM 3 (DUGEM3 ) is
adjusted for each ET2 setup to provide an effective gas gain in the range 1000 – 2000. The tube current of
the X-ray generator used in these measurements is 0.03 mA. A significant reduction of the ion backflow
can be achieved by lowering ET2 , which matches the observations in a standard triple GEM system (see
Sec. 5.1). For the triple GEM system with one COBRA GEM, an ion backflow of 0.5 % can be achieved.
124 The ALICE Collaboration

104 1
Effective Gain

IBF
103 10-1

102 ET2=1kV/cm (V =300V) 10-2

GEM3

ET2=0.75kV/cm (V =330V)
GEM3

ET2=0.5kV/cm (V =350V)
GEM3

ET2=0.25kV/cm (V =370V)
GEM3

10-3
0 50 100 150 200 250 300 0 50 100 150 200 250 300
U AC (V) U AC (V)
up
Figure 9.7: Effective gas gain (left panel) and ion backflow (right panel) measured as a function of DUAC on COBRA 2 with
the GEM 50 – COBRA 2 – GEM 50 configuration. The X-ray tube current is 0.03 mA.

Standard GEM – COBRA GEM – COBRA GEM

In the following, we present measurements with a standard GEM in the first layer, and COBRA GEMs
in the second and third layer (GEM 50 – COBRA 2 – COBRA 2). Figure 9.8 shows the gas gain and the
up
ion backflow as a function of DUAC on the COBRA GEMs at different X-ray tube currents. Open sym-
up up
bols correspond to measurements where DUAC (GEM 2) = DUAC (GEM3), i.e. both voltages are varied
up
simultaneously, while the closed symbols correspond to results where one of the DUAC voltages are fixed.
The GEM voltage and the field configurations are DUGEM1 = 200 V, DUGEM2 = 430 V, DUGEM3 = 430 V,
Edrift = 0.4 kV/cm, ET1 = 0.4 kV/cm, ET2 = 0.4 kV/cm and Eind = 3 kV/cm. For DUAC ⇠ 250 V, an ion
backflow of ⇠ 0.5% is achieved at the lowest X-ray tube currents.

104 1
Effective Gain

IBF

10-1
3
10

10-2

Itube = 0.03 mA, UAC (GEM3) = UAC (GEM2)

102 Itube = 0.03 mA, UAC (GEM3) = 225 V
Itube = 0.3 mA, UAC (GEM3) = UAC (GEM2) 10-3
Itube = 0.3 mA, UAC (GEM3) = 225 V
Itube = 3 mA, UAC (GEM3) = UAC (GEM2)
Itube = 3 mA, UAC (GEM3) = 225 V

10-4
0 50 100 150 200 250 300 0 50 100 150 200 250 300
U AC (V) U AC (V)

Figure 9.8: Effective gas gain (left panel) and ion backflow (right panel) measured with the GEM 50 – COBRA 2 – COBRA 2
up
configuration as a function of DUAC on the COBRA GEMs for different X-ray tube currents. Open symbols corre-
up up up
spond to the settings with DUAC (GEM2) = DUAC (GEM3), while the closed symbols are for DUAC (GEM2) = 250 V
up
and DUAC (GEM3) = 225 V.

Further reduction of the ion backflow is possible by tuning the values of DUGEM2 and DUGEM3 . Figure 9.9
up
shows the gas gain and the ion backflow as a function of DUAC changed simultaneously on GEM 2 and
up up
GEM 3 (so that DUAC (GEM2) = DUAC (GEM3)) at different values of DUGEM2 and DUGEM3 . An ion
TPC Upgrade TDR 125

backflow of less than 0.25% is achieved with DUGEM2  390 V and DUGEM3 470 V.

104 1
Effective Gain

IBF
10-1
103

10-2
( U GEM2 , U GEM3 ) = (430, 430)
102
( U GEM2 , U GEM3 ) = (390, 470)
( U GEM2 , U GEM3 ) = (350, 510)
( U GEM2 , U GEM3 ) = (310, 550)
( U GEM2 , U GEM3 ) = (260, 590) 10-3

0 50 100 150 200 250 300 0 50 100 150 200 250 300
U AC (V) U AC (V)
up
Figure 9.9: Effective gas gain (left panel) and ion backflow (right panel) as a function of DUAC of the COBRA GEMs (with
up up
the condition DUAC (GEM2) = DUAC (GEM3)) measured with the GEM 50 – COBRA 2 – COBRA 2 configuration
at different DUGEM2 and DUGEM3 .

9.1.3 Energy resolution

up
The energy resolution as a function of DUAC has been measured in a GEM 50 – COBRA 2 – COBRA 2
configuration. Like in all other measurements reported here, we used the gas mixture Ne-CO2 (90-10).
The drift length is 30 mm and a 55 Fe source radiates from the top of the drift field. Induced signals on
the anode pad (3⇥3 cm2 ) are amplified by a pre-amplifier and shaper and read out by an MCA (KromeK
K102). In these measurements, DUGEM1 , DUGEM2 , and DUGEM3 are fixed to 200 V, 430 V, and 430 V,
up
respectively, while DUAC is varied between 0 and 200 V. DUGEM2 and DUGEM3 are increased to 445 V
up
at DUAC = 200 V in order to keep the gain constant. Further settings are: Edrift = 0.4 kV/cm, ET1 =
0.4 kV/cm, ET2 = 0.4 kV/cm, and Eind = 3 kV/cm. Figure 9.10 shows the energy spectra obtained at the
up
different settings of DUAC V.
Counts

600
Counts

600 Ne-CO2 (90-10) Ne-CO2 (90-10)

= 17.0 % 500
= 16.3 %
500 U GEM1 =200 V U GEM1 = 200 V
U GEM2 =430 V 400 U GEM2 = 430 V
400
U GEM3 =430 V U GEM3 = 430 V
300 300
U AC = 0 V U AC = 100 V
200 200

100 100
Counts
Counts

600
600 Ne-CO2 (90-10) Ne-CO2 (90-10)
= 17.9 % = 20.9 %
U GEM1 =200 V 500 U GEM1 = 200 V
500
U GEM2 =430 V 400 U GEM2 = 445 V
400
U GEM3 =430 V U GEM3 = 445 V
300 U AC = 150 V 300
U AC = 200 V
200 200

100 100

0 0
0 100 200 300 400 500 600 700 800 0 100 200 300 400 500 600 700 800
ADC (ch) ADC (ch)

Figure 9.10: 55 Feenergy spectra measured in Ne-CO2 (90-10) with a GEM 50 – COBRA 2 – COBRA 2 configuration for dif-
up
ferent values of DUAC .
126 The ALICE Collaboration

up
An energy resolution (s ) of 17%, 16%, 18%, and 21% is achieved with DUAC = 0 V, 100 V, 150 V, and
200 V, respectively. The values are worse by a factor of almost 2 as compared to the resolution achieved
with a prototype IROC (see Fig. 5.20). Most likely the deterioration can be explained by a decreased
electron collection efficiency at GEM 1 due to the large pitch between the holes of the COBRA GEM or
the non-uniform multiplication inside the hole due to the relatively small ratio of hole size to thickness.

9.1.4 Conclusion and outlook

Measurements with triple GEM systems including one or two COBRA GEMs show that an ion backflow
of 0.25 – 0.5% can be achieved by tuning the potentials on the COBRA electrodes and the fields between
the GEMs. However, the energy resolution obtained is not on the level required for the GEM TPC.
The characterization of COBRA GEM systems will be continued in the future. In particular, the perfor-
mance of COBRA3 (see Tab. 9.1) will be studied. Further plans include the characterisation of large-size
COBRA GEMs (50⇥50 cm2 ) in terms of uniformity of gas gain, ion backflow, energy resolution, and
long-term stability.

9.2 Studies with fast gas mixtures

In Ne-CO2 (90-10) and Ne-CO2 -N2 (90-10-5) the maximum electron drift time is ⇠ 100 µs, resulting in
an average overlap of Npileup = 5 minimum bias events at 50 kHz Pb–Pb operation. This motivates the
search for alternative gas mixtures with significantly larger drift velocities at similar electric fields. One
of these options implies the use of CF4 as a quencher gas. As shown in Tab. 3.1, a drift velocity of more
than 8 cm/µs can be achieved in Ne-CF4 (80-20), which exceeds the drift velocity in Ne-CO2 (90-10) by
a factor of ⇠ 3 at the same drift field. As a result, the event pileup can be reduced by a similar factor.
The high electron mobility in Ne-CF4 (90-10) implies also a large wt factor and small diffusion at
the nominal magnetic field of B = 0.5 T (see also Tab. 3.1). This may potentially lead to a significant
improvement of the position resolution, if the readout pads are accordingly reduced to achieve smaller
cluster sizes.
In the following, results from detailed simulations of the TPC response in Ne-CO2 (90-10) and Ne-CF4
(90-10) including a microscopic description of a triple-GEM readout system are presented.

1 1
Fraction

Fraction

Rectangular pads, 6x15mm2 , Ne-CF (90-10) Chevron pads, 6x15mm2 , Ne-CF (90-10)
0.9 4 0.9 4

One-Pad Cluster
0.8 Two-Pad Cluster 0.8
Three-Pad Cluster
0.7 0.7

0.6 0.6

0.5 0.5

0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0
0 50 100 150 200 250 0 50 100 150 200 250
Drift distance (cm) Drift distance (cm)

Figure 9.11: Fraction of one-, two-, and three-pad clusters for rectangular (left) and chevron-shaped (right) pads (6 ⇥ 15 mm2 )
in a triple GEM readout system with Ne-CF4 (90-10).

As discussed in Sec. 4.5, the present pad geometry of the TPC readout chambers leads to a limitation of
the space-point resolution. The narrower pad response function in GEMs, which is approximately equal
TPC Upgrade TDR 127

to the width of the charge cloud emerging from the last GEM, implies that charge sharing among adjacent
pads is insufficient for small drift lengths. This is illustrated in Fig. 9.11 (left panel) where the fraction of
clusters with n = 1, 2 and 3 pads is shown as a function of the drift distance. The calculation is performed
for Ne-CF4 (90-10), assuming rectangular readout pads of size 6 ⇥ 15 mm2 . The frequent occurrence of
one-pad clusters at small drift distance results in a deterioration of the space-point resolution in this
region. This can be overcome by choosing chevron-shaped pads [4] instead of rectangular ones. The
frequency of n-pad clusters for 6 ⇥ 15 mm2 pads with an n-fold chevron structure is shown in Fig. 9.11
(right panel). The majority of clusters have a signal on two or three adjacent pads, while the probability
for one-pad clusters is negligible.
The space-point resolution in rj direction as a function of the drift length is shown in Fig. 9.12. For the
gas mixture Ne-CO2 (90-10) and with rectangular pads (6 ⇥ 15 mm2 ) a deterioration of the space-point
resolution is seen at small drift length, which is due to the increasing fraction of one-pad clusters. In
Ne-CF4 (90-10) with rectangular pads this effect is even more pronounced, due to the smaller transverse
diffusion. The space-point resolution can be improved by a factor ⇠ 3 in Ne-CF4 (90-10) if chevron pads
are used.

0.25
(r ) (cm)

Ne-CO2 (90-10), 3 GEMs, standard pads

Ne-CF4 (90-10), 3 GEMs, standard pads
0.2
Ne-CF4 (90-10), 3 GEMs, chevron pads

0.15

0.1

0.05

Ne-CF4 -C4 H10 (90-10-5), 2 GEM + MMG, chevron pads

0
0 50 100 150 200 250
Drift distance (cm)

Figure 9.12: Space-point resolution in rj obtained with rectangular and chevron pads in different gas mixtures.

The combined ITS-TPC momentum resolution as a function of pT is shown in Fig. 9.13. In a triple GEM
system using Ne-CO2 (90-10) and the present rectangular pad shape the resolution is worse by 5 – 10 %
as compared with the present MWPC readout. When using Ne-CF4 (90-10) as gas mixture and with
chevron pad readout the resolution of the present MWPC readout is restored or even slightly improved.

9.2.1 Conclusion and outlook

As shown for a triple GEM system, the usage of Ne-CF4 (90-10) and chevron-shaped readout pads could
remove the deterioration of the pT resolution with the baseline gas mixture and rectangular readout pads
due to the reduced position resolution in certain parts of the drift volume (see Sec. 7.2.2). However,
it should be noted that the full charged particle tracking scheme in the ALICE central barrel includes
also the Transition Radiation Detector (TRD), which improves the overall resolution by about a factor
two, making the differences between the different design choices in the TPC smaller. Still, a slight
improvement using Ne-CF4 (90-10) can be achieved if chevron pads are employed.
The strongest argument in favor of Ne-CF4 (90-10) is the larger drift velocity that will lead to a significant
decrease of the event pileup. However, comprehensive studies are necessary to investigate the chemical
properties of CF4 in combination with the materials used in the TPC. A detailed R&D study is presently
being prepared.
128 The ALICE Collaboration

0.035

(pT )/pT
MWPC, Ne-CO2 (90-10)
3 GEMs, Ne-CO2 (90-10)

0.03 3 GEMs, Ne-CF4 (90-10)

3 GEMs, Ne-CF4 (90-10), chevron pads
2 GEMs + MMG, Ne-CF4-C4H10 (90-10-5), chevron pads

0.025

0.02

0.015

0.01
0 5 10 15 20 25 30 35 40 45 50
p T (GeV/c)

Figure 9.13: Combined ITS-TPC momentum resolution for MWPC (red open circles) and triple GEM system obtained with
rectangular and chevron pads in different gas mixtures.

9.3 R&D with hybrid gain structures: 2 GEMs + MicroMegas

MicroMegas (MMG) intrinsically have very low ion backflow due to the very high ratio of the values of
the electric fields in the small amplification gap to the drift field above the MMG. Ion backflow values of
less than 1 % have been measured [5, 6].
Stable operation of an MMG detector at gains of 1 – 10⇥103 , necessary to measure single-charged MIPs,
is challenging. To overcome this limitation, studies have been carried out with hybrid GEM + MMG
detectors, with the primary goal to achieve stable operation [7–9]. The addition of a GEM to the gain
structure has the advantage of ” preamplifying ” the signal such that the MMG can be operated at a
lower gain. Furthermore, the GEM spreads the electron cloud in space transverse to the drift direction,
distributing the signal over a larger spot on the MMG. Both effects lead to an increased stability of the
MMG. Discharge rates of < 10 5 at gains of 104 were measured in [10] for such hybrid structures. It was
also established that the MMG has very good energy resolution. Finally, the technology for large-scale
production of MMG detectors is in hand [11].
Based on existing measurements and experience, a hybrid gain structure using two GEM foils above
an MMG appears to be a promising candidate to be able to achieve very low IBF and stable operation,
while maintaining good spatial and energy resolution. A detailed response simulation was carried out
using 2 GEM + MMG with a chevron pad readout plane. The resulting rj and momentum resolutions
are shown in Figs. 9.12 and 9.13 (see Sec. 9.2). The addition of a second GEM should reduce further the
ion backflow and allow operation of all elements at modest gains.
A newly constructed hybrid 2 GEM + MMG chamber using standard 10 ⇥ 10 cm2 GEM foils and an
MMG with 400 LPI4 of the same size was operated to perform ion backflow and energy resolution mea-
surements. Figure 9.14 shows the setup. The measurements were performed using the two mixtures Ar-
CO2 (70-30) and Ar-CO2 (90-10) with GEM foil voltages set to achieve a constant total GEM + MMG
gain and the required drift, transfer and induction fields. Results are shown in Fig. 9.15. The voltage
on the MMG mesh was varied to change the ratio of the fields above and below the MMG mesh (see
Fig. 9.14). The ratios of the currents at the cathode and the at the anode have been measured using a
radioactive source. The dependence of the ion backflow on the MMG voltage is shown in the left panel
of Fig. 9.15 (red) along with the FWHM for the peak from 55 Fe X-rays (blue).
4 Lines-Per-Inch (LPI)
TPC Upgrade TDR 129

Figure 9.14: Setup used for preliminary ion backflow measurement for the hybrid 2 GEM + MMG system.

The following settings were used for these measurements: Edrift = 0.4 kV/cm, ET1 = 3.5 kV/cm, Eind =
0.125 kV/cm, total gain = 3.5 (± 0.5) ⇥ 103 . For the 620 V mesh setting four measurements were made
over a range of source intensities varying by a factor of 10. All 4 ion backflow measurements are in the
range 0.15 – 0.16 %, consistent with the measurement error. At each MMG voltage setting the total gain
was adjusted and the resolution checked by measuring the PH spectrum at the anode (read-out plane)
with a standard CSP and shaper and an ADC with the chamber illuminated by a 55 Fe source (right panel
in Fig. 9.15).
Encouraged by these measurements, a new MMG is being procured for further studies. Since also the
mesh pitch influences the ion backflow performance of an MMG system [6], it is expected that the ion

Figure 9.15: (Left) Preliminary ion backflow measurement for hybrid 2 GEM + MMG system. The horizontal scale is the
voltage on the MMG mesh. The vertical scale is the ratio of cathode to anode currents in % (red points) and
the FWHM of the 55 Fe peak (blue boxes). (Right) Typical PH spectrum from the hybrid 2 GEM + MMG system
irradiated with an 55 Fe source.
130 The ALICE Collaboration

backflow can be suppressed further by a factor of about 1.5 – 2 by using a finer mesh with 600 lines-
per-inch. Such an MMG is currently being sought for the continuation of the tests, which will include
also the use of Ne-CO2 and Ne-CF4 gas mixtures.

9.3.1 Conclusion and outlook

The above R&D program is undertaken to explore the 2 GEM + MMG gain structure as a possible alter-
native to the present baseline quadruple GEM structures. In addition to ion backflow, energy resolution
and discharge properties, we will also study the potential impact that the adoption of this alternative gain
structure would have on the overall project execution plan including cost and schedule. A final report on
this R&D is expected in spring 2015.
Chapter 10

Detector control system

The TPC detector control system (DCS) is part of the global ALICE DCS and will follow its evolu-
tion accordingly. Since a large part of the TPC hardware will not be replaced after RUN 2, also the
corresponding DCS subsystems will not be redesigned. The implementation of the control of new TPC
components will be adapted to the current one. This concerns the high voltage control for the new readout
chambers (GEM HV control) as well as the front-end electronics configuration and monitoring.

10.1 Overview

The upgrade of the DCS system and its interface with the online farm will be described in detail in the
Online Systems Technical Design Report, which will be available in the year 2014.

10.1.1 Hardware architecture

The hardware architecture of the TPC DCS can be divided into three functional layers. The field layer
contains the actual hardware to be controlled (power supplies, front-end electronics...). The control layer
consists of devices for collecting and processing information from the field layer and making it available
to the supervisory layer. At the same time the devices of the control layer receive commands from the
supervisory layer to be processed and distributed to the field layer. The equipment in the supervisory
layer consists of computers and servers, providing the user interfaces and connecting to central DCS
infrastructure consisting of fileservers, database servers managing the configuration and archival data,
etc. The three layers interface mainly through a local area network (LAN).

10.1.2 Software architecture

The software architecture is a tree structure that represents the structure of the TPC, its subsystems and
devices. The structure, as shown in Fig. 10.1, is composed of control and device units with a single top
node (TPC DCS). The control unit steers the sub-tree below it and the device unit drives a device. The
behavior and functionality of each control unit is implemented as a finite state machine.
The control system is built using a control framework that includes drivers for the different types of
hardware, communication protocols, and configurable components for commonly used applications such
as high or low voltage power supplies [1, 2].

131
132 The ALICE Collaboration

TPC

LASER TEMP RUN CORE

Run
TSen
Laser Beam Sync ....... Detector FERO C. Puls
~500x

Las. A Las. C Image Mirror Sync Cool Busy Pul. A Pul. C

GEM CMON FC GAS DVM Inhibit

ESC
ESC
…
Ammeter PLC Gas DVM Inhibit 36x
Crate

GEM A GEM C MISC A MISC C FED LV

Ch Ch Ch Ch CRU Ch
....... ....... ....... ....... ....... .......

Figure 10.1: Overview of the software architecture of the DCS. The tree structure is build out of Device Units (boxes) and
Control Units (ellipses).

10.1.3 System implementation

The core software of the control system is the commercial SCADA1 system SIMATIC WinCC Open
Architecture (OA), formerly known as PVSS II, from the company ETM [3]. WinCC OA is an object-
oriented process visualization and control system that is used in industry as well as by the four LHC
experiments. It is event-driven and has a highly distributed architecture. The SCADA System for the
TPC is currently distributed over 12 computers.

10.1.4 Interfaces to devices

Where possible, commercial servers using the OPC2 standard of process control are used to interface the
SCADA system to devices. OPC servers interface the field cage high voltage, the front-end electronics
low voltage and the temperature monitoring system. For the readout chamber HV control the usage of the
OPC standard is envisaged as well. For non-commercial hardware the communication can be based on
the communication framework Distributed Information Management (DIM [4]). In a similar approach
DIM is currently used in the laser system, in the drift velocity monitor and the calibration pulser control
and will be used for the front-end electronics control and monitoring.

10.1.5 Interlocks
The safety of the detector is based on three layers of interlocks:

– Internal interlock: The internal mechanism of devices (e.g. power supply trip) are used wherever
applicable. The threshold and status of these interlocks are controlled by the SCADA system, but
their function is independent of the communication between hardware and software.

– External interlock: The interlocks between different subsystems are realized using Programmable
Logical Controller (PLC) systems with the possibility to enable or disable them.
1 Supervisory Controls And Data Acquisition (SCADA)
2 Open Platform Communications (OPC)
TPC Upgrade TDR 133

– Software interlock. Software interlocks are realized in the supervisory layer. They rely on the
communication between the hardware and the SCADA system and are thus only used to prevent
the system from unwanted but not harmful events like switching off the power supplies under full
load. The safety of the equipment does not rely on the software interlocks.

Internal interlocks are used for the readout chamber high voltage, the field cage high voltage, the front-
end electronics low voltage, the cooling and the gas system. External interlocks are implemented for the
field cage high voltage, the front-end electronics low voltage and the cooling system. Software interlocks
are used for the readout chamber high voltage, the front-end electronics low voltage and the front-end
electronics. In addition to the interlocks, the alert system of the SCADA system is set up to inform the
shift crew of unusual or potentially dangerous situations.

10.2 Front-end electronics control

The new front-end electronics (see Chap. 6) requires an updated DCS subsystem. It should provide
monitoring of temperatures, voltages, currents and status information and the ability to configure the
front-end electronics for data taking with different run types.

10.2.1 Overview
A schematic of the front-end electronics control is shown in Fig. 10.2. Between the common readout
unit (CRU) system in the control room and the on-detector electronics (front-end cards), the readout
architecture foresees 6336 GBT unidirectional links for the readout of the physics data. Interleaved with
the physics data about 1 % or less of DCS monitoring data will be transmitted on the same links to the
CRU where it is extracted and sent through a dedicated DCS output link to a DCS front-end processor.
The DCS front-end processor is a computer that hosts a dedicated front-end server application, which
pre-processes and filters the monitoring data and forwards it to the SCADA system for further processing
(alarm handling) and display. It also handles configuration requests from the supervisory layer. The
configuration data from the configuration data base and other control data (commands) are sent from
the CRU to the FECs via 1584 unidirectional links for the timing, trigger and clock distribution system
(TTS).

DCS Front End

Configura)on* Processor
DB*
RORC3
DDL3
monitoring data

Control &
configuration
data detector First Level
data links Processor Event
physics & monitoring data (DDL3) building
Front and
End front-end link (GBT) CRU physics data RORC3 processing
Cards

Figure 10.2: Schematic of the front-end electronics control and monitoring.

10.2.2 Monitoring
The SCADA system implements archiving and automatic checking of the monitoring data that it receives
from the front-end server, mainly consisting of temperatures, voltages, currents and status information.
Graphical user interfaces allow to display the data. Experience has shown that such functionality is very
useful as it allows to identify problems such as voltage drops or locally reduced cooling performance.
134 The ALICE Collaboration

10.2.3 Configuration and control

Each configuration of the TPC front-end electronics includes about 5 million configurable parameters.
The parameters for the FE chips, FECs and CRUs are stored in a configuration database hosted on central
DCS ORACLE servers. Configurations may change over time due to disfunctional or replaced hardware
or due to changing hardware behavior.
Only simple configuration commands are passed from the supervisory layer to the FE server, which
handles the actual configuration process. The command contains only a parameter which describes the
configuration type. Based on this the FE server assembles queries to the configuration database which
retrieve the configuration data for each CRU and its associated equipment.

10.3 Parameter export for online calibration and reconstruction

Many parameters that are gathered by the DCS system are of relevance for the online calibration and
reconstruction in the online systems.

– Environmental conditions data: The temperature and pressure trends are of special importance
for the drift velocity and gain calibration of the TPC data. They have to be made available to the
online calibration and reconstruction algorithms with a time granularity of a few Hz.
– GEM currents and detector status: The currents measured in the different GEM HV segments
are directly related to the amount of space charge produced in the corresponding section of the
drift volume. In order to correct online for space-charge effects (see Sec. 7.4.3), the currents must
be measured with a precision of nA and must be included in the TPC data stream in order to
be immediately available (see Sec. 11.4.1). Moreover, the status of the HV channels has to be
available in order to identify tripped GEM sectors where the voltage is below nominal or ramping.
This can be achieved by reading the currents from digital current meters through ethernet into a
computer that is included in the online processing farm. In this way the GEM currents and HV
states can be continuously injected into the data stream. The same computer runs a server software
making the currents available also to the DCS system (see Fig. 10.3).

DCS
Network

Trigger
(for timing
information)
TPC current Event-
processor building
Software: and
TPC Processing
Current
Monitor HV server
Current
meters (CMON)

Figure 10.3: Schematic showing how the GEM current values can be read out for online calibration and DCS.

– Front-end configuration: The latest FE configuration parameters (e.g. inactive regions and ana-
log and DSP parameters) are also of importance to the online calibration and reconstruction algo-
rithms and are thus exported for each data taking session in a suitable format.
– Further parameters of relevance for the online calibration and reconstruction algorithms include
the status of the laser system to identify periods with laser activity during a data taking session.
The list of needed parameters will probably be extended.
Chapter 11

Installation, commissioning and services

11.1 General
In this chapter the installation of the GEM readout chambers into the TPC, the subsequent commissioning
of the new front-end electronics and readout system, and the re-installation of the upgraded TPC in the
ALICE cavern are described. In addition, all the required services to operate the TPC are discussed and
necessary upgrades and modifications are introduced. The total time estimated for the exchange of the
readout chambers and electronics, various upgrade activities and pre-commissioning on the surface is
estimated to be approximately 40 weeks.

11.2 Installation
For the installation of the GEM readout chambers, the TPC is removed from its position in the ALICE
cavern and moved into the so-called Delphi frame to give it the proper mechanical support. Then it can
be moved by crane and truck to the clean room in building SXL2. Due to the sensitivity of the GEM foils
to dust a clean room class ISO 7 (corresponding roughly to class 10,000) is envisaged.

Installation tool For the dismounting of the MWPCs and the installation of the GEM readout chambers
(see Sec. 2.3) a special tool has been developed and successfully used in the past (photos see [1]). For the
proper positioning of this tool, it is attached to a hydraulically controlled platform (the so-called Yellow
Platform) and allows access to one side of the TPC at a time (Fig. 11.1). Save operation of the installation
tool requires two trained persons.

11.3 Commissioning
After the installation of the GEM readout chambers they have to be aligned to ensure the planarity of the
readout plane. This procedure is described in [1]. After the two SSWs are mounted in front of the end
plates the front-end electronics can be mounted together with the connections to LV, HV, and the cooling
distribution system. To ensure the proper working of the new chambers a pre-commissioning phase in
the clean room follows. This includes pedestal and noise measurements as well as calibration pulser tests
to validate the full functionality of the readout. For this purpose, a cooling system for at least two sectors
at a time and the corresponding readout chain has to be available in the clean room. In the next step, the
TPC is connected to a gas system and measurements of cosmic rays, using a dedicated trigger system,
will be performed. In addition, a laser system generates tracks at well-defined positions for alignment
purposes.
In Tab. 11.1 an overview of the main installation steps is given together with the corresponding time

135
136 The ALICE Collaboration

Figure 11.1: TPC with the Yellow Platform and the mounting tool.

estimates. These are partly based on past experience.

Activity duration
(weeks)
Dismantling of FEE, SSWs. Removal of temperature sensors etc. 3
Replacement of ROCs 12
Modification of SSWs (4 weeks in parallel) -
Resistor Rod modifications (1 week in parallel) -
Survey and ROC alignment 4
Mounting of temperature sensors etc. (1 week in parallel) -
Sealing 1
Leak test with He 1
FEE installation 5
Leak test with operating gas mixture (1 week in parallel)
Installation and upgrade of cooling pipes, cables 2
Pulser, cosmics and laser tests (2 sectors at a time) 10
Contingency 2
Total 40

Table 11.1: Time required for the most relevant activities of the GEM readout installation and commissioning in the clean room
(SXL2), partly based on past pre-commissioning experience, partly estimated.

Installation in the ALICE cavern After all pre-commissioning measurements in the clean room are
performed, the TPC can be moved from the clean room into the ALICE cavern. To protect against
influences from ambient conditions i.e., rain and or large temperature gradients, the TPC will be properly
packed before being placed on a truck and transported to the ALICE building where it is lowered into the
pit by crane. Once installed, the final commissioning of the GEM TPC will be performed. Essentially all
measurements from the clean room will be repeated in the cavern, now involving the full detector.
TPC Upgrade TDR 137

11.4 Services
In this section the various services necessary to operate the TPC are described and necessary modifica-
tions and upgrades are discussed.

11.4.1 High voltage

Readout chamber high voltage
For the operation of the GEM stacks new high voltage power supplies with negative output voltage
are required. Due to the use of voltage dividers (see Sec. 4.4) the required current is much larger than
before. We estimate 2 mA to achieve the necessary voltage stability. The use of actively regulated voltage
dividers are an attractive option. They are presently under development [2] and would possibly allow to
operate the divider chains at lower currents while still keeping voltage variations (relevant for the dE/dx
resolution) sufficiently small under changing loads.
A new fast high-precision current monitoring system for the GEM currents is foreseen to allow the
estimation of space charge due to the positive ion backflow with high time granularity. It will monitor
the currents to the individual GEM foils, i.e. after the voltage divider. The precision should allow to
measure the GEM currents down to 100 pA and up to 10 µA at a sampling rate of 1 kHz. In total 8 x 72
channels are needed. A schematic drawing of the input stage of such a device is shown in Fig. 11.2. In
Sec. 10.3 the overall setup with emphasis on the connection to DCS and DAQ is described.

Timing information
HV
from trigger system
To
Transimpedance
Galvanic 1 kHz Control online
converter
Sampling and farm
I ADC Communi
U isolator 16 bits cation DCS
i

To chamber
Figure 11.2: Schematic diagram illustrating the working principle of the current monitoring system.

The chamber high-voltage setup mapping the 32 channels of a set of HV modules to the 18 sectors per
side of the TPC is schematically shown in Fig. 11.3. A similar scheme is used for the OROCs.

Field cage high voltage

The present system providing the drift high voltage of 100 kV for the field cage (Heinzinger power
supply) has proven to work well and will remain unmodified. Since the upper side of the first GEM layer
is on a potential of about 3 kV (see Sec. 2.4), depending on the gain settings, additional high voltage
supplies are needed to allow the adjustment of the voltage at the ground end of the four voltage dividers
(voltage of the last strip of the field cage) to different GEM gain settings. They are connected to the
bottom of the voltage divider replacing the static last resistor in the present setup. This is illustrated in
Fig. 11.4.
The total current across the four voltage dividers of the TPC is about 374 µA. Therefore, each of the four
power supplies has to be able to work as a current sink and accommodate up to 100 µA current.
138 The ALICE Collaboration

32 channel 32 channel
HV HV
module module

16 16 16

4
Patch 4 Patch Patch
box box box

20 20

18 18
IROCs IROCs
A-side C-side

Figure 11.3: Mapping scheme of the high voltage power supplies to the IROCs.

Figure 11.4: Schematic diagram of the ground sides of the four resistor rods indicating the connections to the current measuring
system (connector 1) and to the new HV power supply (connector 2).

11.4.2 Low voltage

The present setup of the low voltage system supplying the voltages to the front-end electronics has
worked well in the past and can be reused without modifications [1]. It provides separate voltages for the
analog and digital parts of the front-end electronics. The presently foreseen new readout electronics for
the GEM based chambers will most likely need less voltage and less overall power (see Tab. 6.2).
Also the LV cables connecting the power supply to the local distribution boards via bus bars running
along the spokes of the SSW within the sectors can be reused. The low voltage setup is schematically
shown in Fig. 11.5 indicating the supply of two sectors by one power supply module.
In Tab. 11.2 the parameters of the LV power supplies are listed.

11.4.3 Cooling
FEE cooling
The cooling system for the front-end electronics is an under-pressure leak-less water cooled system. It
has gone through several upgrades in the past and is considered to be well suited for the cooling of the
TPC Upgrade TDR 139

3U crate
Digital Analog Digital Analog
LAN
4.4V 5.1 V 4.4V 5.1 V
132 A 83 A 132 A 83 A

~40 m

Sense lines

FECs

bus bars

Figure 11.5: Schematic of the low voltage setup for two sectors indicating typical voltages and currents for the present system.

Supply channels/crate nom. power voltage max. current continuous current

(W) (V) (V) (A)
Analog 2 600 2-7 115 100
Digital 2 1200 2-7 230 200
Table 11.2: Specifications of the existing LV power supply system.

new front-end electronics. Due to the lower power consumption foreseen, also the presently available
cooling power (25 kW) will be sufficient. Nevertheless, some minor modifications mainly regarding the
leak tightness are envisaged.

Resistor rod cooling

For the cooling of the resistor rods a separate cooling plant is used. It is also an under-pressure leak-less
water cooled system, however, with well controlled conductivity, since the water is exposed to the very
high voltage (100 kV) of the field cage. For the future activities, only minor modifications are foreseen.
These include better control of the flow through the resistor rods and better heat exchangers close to the
TPC for better temperature control and stability.

Heat screens
To thermally separate the TPC from neighboring detectors two different heat screens are used. The outer
heat screen isolates the TPC from the TRD and is supplied by the cooling plant of the TRD due to the
use of aluminum cooling panels (unlike the FEE cooling system which uses copper based pipes and
components). No changes are foreseen here. For the shielding against thermal effects from the ITS the
inner heat screen is used. Due to the use of stainless steel cooling panels it can be supplied by the TPC
cooling plant. No major changes are foreseen at this time.

11.4.4 Calibration
Calibration pulser
For the monitoring and gain calibration of the readout electronics a calibration pulser system will be
installed. It is connected to the output side of the last GEM and injects charge into the pre-amplifiers
by sending a voltage step to the bottom side of GEM 4. Its working principle is schematically shown in
140 The ALICE Collaboration

time

Delay adjustable 0 ... 100 µs

(min. step size 1 ns)

voltage
Pulse shape
adjustable

Amplitude adjustable: 0 ...- 8 V

Readout time of TPC (~95 µs)

Figure 11.6: Working principle of the calibration pulser system.

Fig. 11.6. The potential of the bottom side of GEM 4 is determined by the induction field and is typically
around 1 kV. Therefore, a decoupling HV capacitor is needed. Its capacity should be as small as possible
in order not to store charge that could lead to damages of the front-end electronics in case of a discharge.
Depending on the necessary signal amplitudes it may be possible to reuse the existing calibration pulser
setup shown schematically in Fig. 11.7. The present design provides amplitudes of up to 3.5 V. Since it
is foreseen to modify the ALICE trigger hardware, a new connection to the trigger and clock distribution
needs to be developed. The remote control of the system by DCS will be modified to allow direct access
via network.
Analog
Interface Controller reference signal Drivers
To chambers
Driver 1
mem Driver 2
Driver 3
Gated fan out

Driver 4
DAC

FPGA control Driver 5

Driver 6
Driver 7
Driver 8
µ processor Driver 9
Board 1
Board 2
Board 3
clock , trigger Board 4
LAN

Figure 11.7: Schematic setup of the calibration pulser system.

Laser
The laser system provides tracks at well defined locations in the TPC and represents an important calibra-
tion tool [1]. The hardware may need some upgrades over time like the exchange of the frame grabbers
of the cameras due to obsolete components of the system. To accommodate new requirements we may
change to diode pumped lasers since it improves our capability to generate laser events at a higher rate
compared to the present flashlight pumped laser system. This would facilitate the monitoring of space-
charge distortion corrections (see Sec. 8.5.2). Like for the calibration pulser system, the connection to
the trigger system and the control via network has to be upgraded when changing to continuous readout.

Krypton calibration
Another important calibration tool is the Krypton calibration, described in Sec. 8.6.1. The container
with the Rubidium source decaying into the radioactive Krypton is connected to a bypass line in the gas
system and there is no change foreseen in the way it is used [1].
Chapter 12

Project organization, cost estimate and

time line

12.1 Participating institutions

The list of institutions participating in the TPC upgrade is shown in Tab. 12.1. About half of the groups
were involved in the construction and operation of the present TPC. New institutes have recently joined
the TPC collaboration, among them a large number of US groups, that bring in significant expertise in
GEM technology, detector construction, engineering, electronics, and computing. The full TPC upgrade
collaboration list is shown in App. B.
The TPC upgrade project is represented and coordinated by a project leader, two deputy project leaders
and a technical coordinator, as shown in Fig. 12.1. The TPC upgrade project is split into a number of
sub-projects, covering all relevant aspects of detector development, installation, software development,
simulation, online computing and calibration. The TPC upgrade project emerges from the existing TPC
project structure and partially overlaps with it. Items or components where only minor modifications to
the present system are involved (e.g. laser system, gas system, detector control system) are not shown in
Fig. 12.1.
The sharing of responsibilities for the TPC upgrade among the participating institutions is shown in
Tab. 12.2.

12.2 Cost estimate

The CORE cost estimate for the TPC upgrade is summarized in Tab. 12.3. CORE costs include detector
components and production cost as well as external manpower for production and installation. They do
not include cost for internal manpower, basic infrastructure, and R&D.
A provisional funding scheme includes the following major contributions: The possibility of funds of
the order of 50 % of the total cost is indicated by the German BMBF and HGF. A substantial contri-
bution to the TPC upgrade is foreseen by the groups from the US, where the construction, assembly
and test of the IROCs will be conducted. This will be part of a wider involvement of US-DOE into the
upgrade of the ALICE central barrel detectors. A provisional funding scheme is consistent with a US-
DOE CORE contribution to the TPC that corresponds to the IROC fraction of the total cost (⇠ 36 %).
The development of the SAMPA ASIC within a common ALICE project is conducted by the Electrical
Engineering-Polytechnical School, University of São Paulo. Funds from Brazil for the development,
production, and test of the TPC SAMPA chips are envisaged. The Common Readout Unit CRU is being
developed by groups from the Wigner Research Center for Physics, Budapest, and from VECC, Kolkata,

141
142 The ALICE Collaboration

Country
Funding Agency City Institute
Croatia Zagreb Department of Physics, University of Zagreb
Denmark Copenhagen Niels Bohr Institute, University of Copenhagen
Finland Helsinki Helsinki Institute of Physics
Germany BMBF Bonn Helmholtz-Institut für Kern- und Strahlenphysik, Rheinische Friedrich-
Wilhelms-Universität Bonn
Germany BMBF Frankfurt Institut für Kernphysik, Johann Wolfgang Goethe-Universität Frankfurt
Germany BMBF Heidelberg Physikalisches Institut, Ruprecht-Karls Universität Heidelberg
Germany BMBF Munich Physik Department, Technische Universität München
Germany BMBF Tübingen Physikalisches Institut, Eberhard Karls Universität Tübingen
Germany BMBF Worms FH Worms, Worms
Germany GSI Darmstadt Research Division and ExtreMe Matter Institute EMMI, GSI
Helmholtzzentrum für Schwerionenforschung
Hungary Budapest Wigner Research Center for Physics, Budapest
India Kolkata Bose Institute
India Bhubaneswar Institute of Physics
India Bhubaneswar National Institute of Science Education and Research
India Indore Indian Institute of Technology
India Mumbai Indian Institute of Technology
India Kolkata Variable Energy Cyclotron Centre
Japan Tokyo University of Tokyo
Mexico Mexico City Instituto de Ciencias Nucleares, Universidad Nacional Autónoma de
México
Norway Bergen / Tonsberg Department of Physics, University of Bergen, Vestfold University Col-
lege, Tonsberg
Norway Bergen Faculty of Engineering, Bergen University College
Pakistan Islamabad Department of Physics, COMSATS Institute of Information Technology
Islamabad
Poland Cracow The Henryk Niewodniczanski Institute of Nuclear Physics, Polish
Academy of Science
Romania Bucharest National Institute for Physics and Nuclear Engineering
Slovakia Bratislava Faculty of Mathematics, Physics and Informatics, Comenius University
Sweden Lund Division of Experimental High Energy Physics, University of Lund
USA DOE Omaha Creighton University, Omaha, Nebraska
USA DOE Houston University of Houston, Houston, Texas
USA DOE Berkeley Lawrence Berkeley National Laboratory, Berkeley, California
USA DOE Livermore Lawrence Livermore National Laboratory, Livermore, California
USA DOE Oak Ridge Oak Ridge National Laboratory, Oak Ridge, Tennessee
USA DOE West Lafayette Purdue University, West Lafayette, Indiana
USA DOE Knoxville University of Tennessee, Knoxville, Tennessee
USA DOE Austin The University of Texas at Austin, Austin, Texas
USA DOE Detroit Wayne State University, Detroit, Michigan
USA DOE New Haven Yale University, New Haven, Connecticut
USA NSF San Luis Obispo California Polytechnic State University, San Luis Obispo, California
USA NSF Chicago Chicago State University, Chicago, Illinois

Table 12.1: List of institutions participating in the TPC upgrade.

TPC Upgrade TDR 143

Project(Leader(
!H.!Appelshäuser!(U!Frankfurt)!
Deputy(Project(Leader(
!C.!Garabatos!(GSI)!
Deputy(Project(Leader(
T.!Cormier!(ORNL)!
Technical(Coordinator(
C.!Lippmann!(GSI)!

SAMPA(
FEE(and(Readout( ALICE!Common!project!
D.!Silvermyr!(ORNL)!
J.!Alme!(Bergen)!
Detector(and(Physics(Performance(
M.!Ivanov!(GSI)!
C.!Lippmann!(GSI)! CRU(
ALICE!Common!project!

GEM(R&D(
Online/Offline(Reconstruc6on( B.!Ketzer!(U!Bonn)!
J.!Thäder!(GSI)! D.!Majka!(Yale)!
M.!Ploskon!(LBNL)! H.!Hamagaki!(U!Tokyo)!

OROC(
Online(Calibra6on( B.!Ketzer!(U!Bonn)!
M.!Ivanov!(GSI)! L.!FabbieJ!(TUM)!
J.!Wiechula!(U!Tübingen)! C.!Garabatos!(GSI)! Coordina6ng(Project(Engineers((
B.!Windelband!(U!Heidelberg)!
J.!Rasson!(LBNL)!
Simula6on( IROC(
D.!Majka!(Yale)!
T.!Gunji!(U!Tokyo)!
N.!Smirnov!(Yale)!!!!!!
P.!ChrisUansen!(Lund)!

Installa6on(
B.!Windelband!
(U!Heidelberg)!

Figure 12.1: Structure of the TPC upgrade project.

Item Institution
IROC Yale, Detroit, Oak Ridge, Knoxville, Austin
OROC Munich, Frankfurt, GSI, Heidelberg, Budapest, Bucharest
GEM R&D and QA Helsinki, Munich, Tokyo, Yale, Zagreb, GSI
Frontend Card Lund, Oak Ridge
FEE integration and test Oak Ridge, Lund, Houston, Tokyo, Bergen, Oslo, GSI
HV, LV, cooling Mexico-City, GSI, Munich
Detector Control GSI, Worms
Installation and engineering Heidelberg, Berkeley
Gas system and field cage GSI
SAMPA ASIC São Paulo, Bergen, Oslo
CRU Budapest, Kolkata, Bergen

Table 12.2: Sharing of responsibilities for construction and installation of the TPC upgrade. Note that SAMPA ASIC and CRU
are parts of common ALICE projects.
144 The ALICE Collaboration

in close collaboration with CERN. India and Hungary have indicated the possibility of funding for the
production cost of the TPC CRU. Further requests for funding are presently being prepared by the groups
from Finland, Japan, Mexico, Norway, and Sweden.

Readout chambers Quantity Cost (MCHF)

(incl. spares)
GEM foils1 480 0.5
Frames and components 960 0.1
Pad planes 160 0.4
Chamber bodies 80 0.3
HV divider 80 0.1
Assembly and installation tooling 0.4
Total Readout Chambers 1.8
Services Cost (MCHF)
GEM HV system 0.2
Fast current monitoring 0.2
HV supply for last FC resistor 0.1
Other services 0.2
Total Services 0.7
FEE and Readout Quantity Cost (MCHF)
(incl. spares)
SAMPA ASIC 19,500 0.78
Front-end card 3900 0.35
GBTx ASIC 7000 0.38
Optical transmitters/receivers 5500 0.79
CRU (control room, AMC40) 2.00
Optical fibers 9000 1.32
TPC Event Processing Nodes (TPC-EPN) 1.00
Other 0.02
Total Electronics 6.64
Total IROC 40 3.3
Total OROC 40 5.84
Total 9.14

Table 12.3: CORE cost estimate for the TPC upgrade.

12.3 Schedule
The current LHC schedule foresees LS2 to start in summer 2018. This defines the time schedule for
the TPC upgrade, see Fig. 12.2. Major technological choices will have to be made by the middle of
2015 to allow finalization of the design, a timely procurement of detector materials and preparation of
the series production. Such decisions involve a definition of the GEM geometry and configuration, as
well as their operational point. This implies a sustained R&D effort in 2014 and early 2015. Particular
emphasis will be put on a set of detailed measurements to characterize the discharge properties of the
baseline system under exposure to different radiation sources. Besides proceeding the studies with small
prototypes, a further test campaign of full-size IROC prototypes equipped with a quadruple GEM stack
will be performed in autumn 2014. This includes a test beam time at the PS with mixed electron and pion
beams to validate the dE/dx performance of this configuration, and a stability test with hadron beams at
the SPS. Similar studies are foreseen to allow a final conclusion on possible technological alternatives.
1 This number assumes 4 GEM segments per IROC and 12 GEM segments per OROC, see Chap. 4. Two segments can be

processed on a single foil. In addition, 50% spares are included. This yields a total of 0.5 · 1.5 · (40 · 4 + 40 · 12) = 480 foils.
TPC Upgrade TDR
ID Task Name 2013 2014 2015 2016 2017 2018 2019 2020
Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1
1 LS1
2
3 Read out chambers
4 R&D
5 Design and prototyping
6 Signal cables choice 01/07
7 Detector 01/04
8 Pad plane design 01/04
9 Chamber body design 01/04
10 Chamber bodies
11 Pad planes
12 GEM foil
13 GEM foil QA
14 ROC assembly
15 ROC
16 End of ROC 31/12
17
Figure 12.2: Time line for the TPC upgrade.

18 FEE
19 Design and prototyping
20 SAMPA ve
21 SAMPA 01/05
22 SAMPA and
23 N m. of channels per FEC 01/07
24 FEE 01/04
25 FEC preseries
26 FEC and
27 CRU prototype
28 CRU
29 IROC prototype
30 End of FEE 31/12
31
32 Service support wheel
33 FEC frame design
34 FEC frame
35
36 HV system
37 HV system design
38 HV system
39
40
41
42 LS2
43
44 and
45 TPC on
46 ng FEE and services
47 ROC replacement
48 Alignment and sealing
49 FEE
50 Pre-commissioning on
51 in cavern
52 Service

145
146 The ALICE Collaboration

Moreover, the availability of GEM foils from different suppliers needs to be investigated. At present,
three possible suppliers exist: TechEtch (USA), SciEnergy (Japan) and CERN TS-DEM. Foils from all
three suppliers are being studied.
The assembly and test of the readout chambers will be distributed among various collaborating institu-
tions in Europe (OROCs) and the US (IROCs), making optimum use of their experience, resources and
facilities (see Tab. 12.2). Delivery of the tested readout chambers to CERN will happen at the beginning
of 2018. A similar timeline for R&D, production and test is imposed to the frontend electronics.
Since the GEM readout chamber installation will have to be carried out in a clean environment, the TPC
must be removed from the cavern and moved into the SXL2 cleanroom on the P2 surface. In order to
maximize the available time for the readout chamber and electronics installation, the TPC removal will
take place at the beginning of LS2. The total duration of LS2 is 18 months. This will leave approximately
40 weeks for replacing readout chambers, installation of the new electronics and pre-commissioning on
the surface before re-installation towards the end of the shutdown period (see also Sec. 11.3).

12.4 TPC upgrade TDR editorial committee

The editorial committee of this TDR was formed by the following persons:
H. Appelshäuser, M. Ball, P. Christiansen, C. Garabatos, P. Gasik, T. Gunji, J. Harris, M. Ivanov, B. Ket-
zer, C. Lippmann, A. Oskarsson, N. Smirnov, R. Renfordt, D. Röhrich, J. Thäder, J. Wiechula.

12.5 TPC upgrade TDR task force

The following persons have contributed to the work presented in this TDR:
H. Appelshäuser, M. Ball, G. Barnaföldi, E. Bartsch, J. Bloemer, P. Christiansen, T. Cormier, K. Eckstein,
L. Fabbietti, C. Garabatos, F. Garcia, P. Gasik, T. Gunji, H. Hamagaki, J. Harris, E. Hellbär, D. Heuchel,
A. Hönle, M. Ivanov, B. Ketzer, M. Kowalski, C. Lippmann, M. Ljunggren, J. Margutti, S. Masciocchi,
A. Mathis, A. Oskarsson, V. Peskov, R. Renfordt, D. Röhrich, R. Shahoyan, D. Silvermyr, N. Smirnov,
K. Terasaki, J. Thäder, D. Vranic, M. Weber, J. Wiechula, Y. Yamaguchi.
Appendix A

Coordinate systems

A.1 Global coordinate system

The global ALICE coordinate system [1] is a right-handed orthogonal cartesian system, which has its
origin at the beam interaction point. A sketch of the global coordinate system is given in Fig. A.1.

C−Side
Muon−Side
A−Side ✓
Shaft−Side
z
x

Figure A.1: ALICE global coordinate system.

Its z-axis is parallel to the mean beam direction, pointing towards the ‘A-side’, away from the muon arm.
This side is also called shaft- or RB24-side. The opposite side (negative z values) is called C-side, or
also Muon- or RB26-side. The x-axis is lying in the local horizontal accelerator plane, pointing towards
the centre of the LHC ring. The side with positive x values is also called I-side (inner), the opposite
side correspondingly O-side (outer). The y-axis is chosen to define a right-handed system, thus pointing
upwards. The azimuthal angle j is increasing counterclockwise, starting from the x-axis (j = 0) and
looking from the A-side towards the C-side. The polar angle q is increasing from the z-axis towards the
xy-plane.

A.2 Local coordinate system

To account for the azimuthal segmentation of the central barrel detectors, the reconstruction software
uses a local coordinate system [2] related to a given sub-detector (TPC sector, ITS module etc.). The
local coordinate system is a right-handed cartesian system as well. In case of the TPC, it has the same
origin and z-axis as the global coordinate system, which is perpendicular to the sensitive planes of the
TPC sectors. The local x-axis lies in the sensitive plane and is parallel to the pad rows. Therefore, the

147
148 The ALICE Collaboration

pads in each row are in the direction of the y-axis. Small variations in the direction of the pads can be
expressed as variations in r · j and are often quoted as rj. A sketch of the local coordinate system is
given in Fig. A.2.

yglobal

cal
xlo
yloc
al

z ↵
xglobal
z
Figure A.2: ALICE local coordinate systems.
Appendix B

TPC upgrade collaboration

Department of Physics, University of Zagreb, Zagreb, Croatia

M. Planicic, N. Poljak, G. Simatovic, A. Utrobicic

Niels Bohr Institute, University of Copenhagen, Copenhagen, Denmark

J.J. Gaardhøje, B. Nielsen

Helsinki Institute of Physics, Helsinki, Finland

E. Brucken, F. Garcia, T. Hilden, J. Rak

Helmholtz-Institut für Kern- und Strahlenphysik, Rheinische Friedrich-Wilhelms-Universität, Bonn,

Germany
F. Boehmer, B. Ketzer

Institut für Kernphysik, Johann Wolfgang Goethe-Universität, Frankfurt, Germany

W. Amend, H. Appelshäuser, M. Arslandok, E. Bartsch, T. Bröker, S. Heckel, E. Hellbär, D. Just,
P. Lüttig, V. Peskov, F. Pliquett, P. Reichelt, R. Renfordt, A. Tarantola Peloni

Physikalisches Institut, Ruprecht-Karls Universität, Heidelberg, Germany

P. Glässel, J. Stachel, D. Vranic, B. Windelband

Physik Department, Technische Universität München, Germany

M. Ball, J. Bloemer, K. Eckstein, P. Gasik, D. Heuchel, A. Hönle, L. Fabbietti, J. Margutti, A. Mathis,
S. Weber

Physikalisches Institut, Eberhard Karls Universität, Tübingen, Germany

H. Schmidt, J. Wiechula

FH Worms, Worms, Germany

R. Keidel

Research Division and ExtreMe Matter Institute EMMI, GSI Helmholtzzentrum für Schwerio-
nenforschung, Darmstadt, Germany
A. Andronic, R. Averbeck, P. Braun-Munzinger, U. Frankenfeld, C. Garabatos, M. Ivanov, M. Köhler,
M. Krzewicki, C. Lippmann, A. Marin, N. Martin, S. Masciocchi, D. Miskowiec, M. Nicassio, J. Otwinowski,
C. Schmidt, K. Schweda, I. Selyuzhenkov, J. Thäder, J. Wagner

149
150 The ALICE Collaboration

Wigner Research Center for Physics, Budapest, Hungary

G. Barnaföldi, G. Bencedi, G. Hamar, D. Varga

Bose Institute, Kolkata, India

S. Das, S.K. Ghosh, S.K. Prasad, S. Raha

Institute of Physics, Bhubaneswar, India

P. Sahu, N. Sharma

National Institute of Science Education and Research, Bhubaneswar, India

S. Biswas, L. Kumar, B. Mohanty

Indian Institute of Technology, Indore, India

A. Roy, R. Sahoo

Indian Institute of Technology, Mumbai, India

S. Dash, B.K. Nandi, R. Varma

Variable Energy Cyclotron Centre, Kolkata, India

Z. Ahammed, S. Chattopadhyay, A.K. Dubey, P. Ghosh, S.A. Khan, T.K. Nayak, S.K. Pal, J. Saini,
R.N. Singaraju, Y.P. Viyogi

University of Tokyo, Tokyo, Japan

H. Hamagaki, T. Gunji, K. Terasaki, Y. Yamaguchi, K. Yukawa

Instituto de Ciencias Nucleares, Universidad Nacional Autónoma de México, Mexico-City, Mex-

ico
G. Paic

Department of Physics and Technology, University of Bergen, Bergen, Norway

D. Röhrich, K. Ullaland, A. Velure

Faculty of Engineering, Bergen University College, Bergen, Norway

J. Alme, H. Helstrup

Department of Physics, University of Oslo, Oslo, Norway

K. Røed, S. Mahmood, C. Zhao

Department of Technology, Vestfold University College, Tonsberg, Norway

J. Lien, R. Langoey

Department of Physics, COMSATS Institute of Information Technology Islamabad, Islamabad,

Pakistan
A. Bhatti, A. Rehman

The Henryk Niewodniczanski Institute of Nuclear Physics, Polish Academy of Science, Cracow,
Poland
M. Kowalski, A. Matyja
TPC Upgrade TDR 151

National Institute for Physics and Nuclear Engineering, Bucharest, Romania

M. Petris, M. Petrovici

Faculty of Mathematics, Physics and Informatics, Comenius University, Bratislava, Slovakia

M. Pikna, B. Sitár, P. Strmeň, I. Szarka

Division of Particle Physics, University of Lund, Lund, Sweden

P. Christiansen, M. Ljunggren, L. Österman, E. Stenlund, A. Oskarsson, T. Richert

Creighton University, Omaha, Nebraska, USA

M. Cherney, M. Poghosyan, J. Seger

University of Houston, Houston, Texas, USA

R. Bellwied, S. Jena, D. McDonald, L. Pinsky, A. Timmins, M. Weber

Lawrence Berkeley National Laboratory, Berkeley, California, USA

D. Gangadharan, P.M. Jacobs, C. Loizides, M. Ploskon, J. Porter, J. Rasson, X. Zhang

Lawrence Livermore National Laboratory, Livermore, California, USA

R. Soltz

Oak Ridge National Laboratory, Oak Ridge, Tennessee, USA

T.M. Cormier, D.J. Dean, M. Middlebrook, K.F. Read, D. Silvermyr

Purdue University, West Lafayette, Indiana, USA

R.P. Scharenberg, B.K. Srivastava

University of Tennessee, Knoxville, Tennessee, USA

C. Nattrass, N. Sharma, S.P. Sorensen

The University of Texas at Austin, Austin, Texas, USA

A.G. Knospe, C. Markert, J. Schambach

Wayne State University, Detroit, Michigan, USA

R. Belmont, C.A. Pruneau, P. Pujahari, J. Putschke, M. Verweij, S. Voloshin

Yale University, New Haven, Connecticut, USA

H. Caines, M.E. Connors, J.W. Harris, R. Majka, R.J. Reed, T. Schuster, N. Smirnov

California Polytechnic State University, San Luis Obispo, California, USA

J. L. Klay

Chicago State University, Chicago, Illinois, USA

E. Garcia-Solis, A. Harton
152 The ALICE Collaboration
Appendix C

The ALICE Collaboration

B. Abelev72 , J. Adam37 , D. Adamová80 , M.M. Aggarwal84 , M. Agnello91,108 , A. Agostinelli26 ,

N. Agrawal44 , Z. Ahammed126 , N. Ahmad18 , I. Ahmed15 , S.U. Ahn65 , S.A. Ahn65 , I. Aimo91,108 ,
S. Aiola131 , M. Ajaz15 , A. Akindinov55 , S.N. Alam126 , D. Aleksandrov97 , B. Alessandro108 ,
D. Alexandre99 , A. Alici12,102 , A. Alkin3 , J. Alme35 , T. Alt39 , S. Altinpinar17 , I. Altsybeev125 ,
C. Alves Garcia Prado115 , C. Andrei75 , A. Andronic94 , V. Anguelov90 , J. Anielski50 , T. Antičić95 ,
F. Antinori105 , P. Antonioli102 , L. Aphecetche109 , H. Appelshäuser49 , N. Arbor68 , S. Arcelli26 ,
N. Armesto16 , R. Arnaldi108 , T. Aronsson131 , I.C. Arsene94,21 , M. Arslandok49 , A. Augustinus34 ,
R. Averbeck94 , T.C. Awes81 , M.D. Azmi18,86 , M. Bach39 , A. Badalà104 , Y.W. Baek40,67 , S. Bagnasco108 ,
R. Bailhache49 , R. Bala87 , A. Baldisseri14 , M. Ball89 , F. Baltasar Dos Santos Pedrosa34 , R.C. Baral58 ,
R. Barbera27 , F. Barile31 , G.G. Barnaföldi130 , L.S. Barnby99 , V. Barret67 , J. Bartke112 , M. Basile26 ,
N. Bastid67 , S. Basu126 , B. Bathen50 , G. Batigne109 , B. Batyunya63 , P.C. Batzing21 , C. Baumann49 ,
I.G. Bearden77 , H. Beck49 , C. Bedda91 , N.K. Behera44 , I. Belikov51 , R. Bellwied117 ,
E. Belmont-Moreno61 , G. Bencedi130 , S. Beole25 , I. Berceanu75 , A. Bercuci75 , Y. BerdnikovII,82 ,
D. Berenyi130 , M.E. Berger89 , R.A. Bertens54 , D. Berzano25 , L. Betev34 , A. Bhasin87 , A.K. Bhati84 ,
B. Bhattacharjee41 , J. Bhom122 , L. Bianchi25 , N. Bianchi69 , C. Bianchin54 , J. Bielčı́k37 , J. Bielčı́ková80 ,
A. Bilandzic77 , S. Bjelogrlic54 , F. Blanco10 , D. Blau97 , C. Blume49 , F. Bock90,71 , A. Bogdanov73 ,
H. Bøggild77 , M. Bogolyubsky52 , F.V. Böhmer89 , L. Boldizsár130 , M. Bombara38 , J. Book49 ,
H. Borel14 , A. Borissov93,129 , F. Bossú62 , M. Botje78 , E. Botta25 , S. Böttger48 , P. Braun-Munzinger94 ,
T. Breitner48 , T.A. Broker49 , T.A. Browning92 , E. Bruna108 , G.E. Bruno31 , D. Budnikov96 ,
H. Buesching49 , S. Bufalino108 , P. Buncic34 , O. Busch90 , Z. Buthelezi62 , D. Caffarri28 , X. Cai7 ,
H. Caines131 , L. Calero Diaz69 , A. Caliva54 , E. Calvo Villar100 , P. Camerini24 , F. Carena34 ,
W. Carena34 , J. Castillo Castellanos14 , E.A.R. Casula23 , V. Catanescu75 , C. Cavicchioli34 ,
C. Ceballos Sanchez9 , J. Cepila37 , P. Cerello108 , B. Chang118 , S. Chapeland34 , J.L. Charvet14 ,
S. Chattopadhyay126 , S. Chattopadhyay98 , M. Cherney83 , C. Cheshkov124 , B. Cheynis124 ,
V. Chibante Barroso34 , D.D. Chinellato117,116 , P. Chochula34 , M. Chojnacki77 , S. Choudhury126 ,
P. Christakoglou78 , C.H. Christensen77 , P. Christiansen32 , T. Chujo122 , S.U. Chung93 , C. Cicalo103 ,
L. Cifarelli12,26 , F. Cindolo102 , J. Cleymans86 , F. Colamaria31 , D. Colella31 , A. Collu23 , M. Colocci26 ,
G. Conesa Balbastre68 , Z. Conesa del Valle47 , M.E. Connors131 , J.G. Contreras11 , T.M. Cormier81,129 ,
Y. Corrales Morales25 , P. Cortese30 , I. Cortés Maldonado2 , M.R. Cosentino115 , F. Costa34 , P. Crochet67 ,
R. Cruz Albino11 , E. Cuautle60 , L. Cunqueiro69,34 , A. Dainese105 , R. Dang7 , D. Das98 , I. Das47 ,
K. Das98 , S. Das4 , A. Dash116 , S. Dash44 , S. De126 , H. DelagrangeI,109 , A. Deloff74 , E. Dénes130 ,
G. D’Erasmo31 , A. De Caro29,12 , G. de Cataldo101 , J. de Cuveland39 , A. De Falco23 ,
D. De Gruttola12,29 , N. De Marco108 , S. De Pasquale29 , R. de Rooij54 , M.A. Diaz Corchero10 ,
T. Dietel50,86 , R. Divià34 , D. Di Bari31 , S. Di Liberto106 , A. Di Mauro34 , P. Di Nezza69 , Ø. Djuvsland17 ,
A. Dobrin54 , T. Dobrowolski74 , D. Domenicis Gimenez115 , O. Dordic21 , S. Dørheim89 , A.K. Dubey126 ,

153
154 The ALICE Collaboration

A. Dubla54 , L. Ducroux124 , P. Dupieux67 , A.K. Dutta Majumdar98 , R.J. Ehlers131 , D. Elia101 ,

H. Engel48 , B. Erazmus34,109 , H.A. Erdal35 , D. Eschweiler39 , B. Espagnon47 , M. Esposito34 ,
M. Estienne109 , S. Esumi122 , D. Evans99 , S. Evdokimov52 , D. Fabris105 , J. Faivre68 , D. Falchieri26 ,
A. Fantoni69 , M. Fasel90 , D. Fehlker17 , L. Feldkamp50 , D. Felea59 , A. Feliciello108 , G. Feofilov125 ,
J. Ferencei80 , A. Fernández Téllez2 , E.G. Ferreiro16 , A. Ferretti25 , A. Festanti28 , J. Figiel112 ,
S. Filchagin96 , D. Finogeev53 , F.M. Fionda31,101 , E.M. Fiore31 , E. Floratos85 , M. Floris34 , S. Foertsch62 ,
P. Foka94 , S. Fokin97 , E. Fragiacomo107 , A. Francescon28,34 , U. Frankenfeld94 , U. Fuchs34 , C. Furget68 ,
M. Fusco Girard29 , J.J. Gaardhøje77 , M. Gagliardi25 , A.M. Gago100 , M. Gallio25 ,
D.R. Gangadharan19,71 , P. Ganoti85,81 , C. Garabatos94 , E. Garcia-Solis13 , C. Gargiulo34 , I. Garishvili72 ,
J. Gerhard39 , M. Germain109 , A. Gheata34 , M. Gheata59,34 , B. Ghidini31 , P. Ghosh126 , S.K. Ghosh4 ,
P. Gianotti69 , P. Giubellino34 , E. Gladysz-Dziadus112 , P. Glässel90 , A. Gomez Ramirez48 ,
P. González-Zamora10 , S. Gorbunov39 , L. Görlich112 , S. Gotovac111 , L.K. Graczykowski128 , A. Grelli54 ,
A. Grigoras34 , C. Grigoras34 , V. Grigoriev73 , A. Grigoryan1 , S. Grigoryan63 , B. Grinyov3 , N. Grion107 ,
J.F. Grosse-Oetringhaus34 , J.-Y. Grossiord124 , R. Grosso34 , F. Guber53 , R. Guernane68 , B. Guerzoni26 ,
M. Guilbaud124 , K. Gulbrandsen77 , H. Gulkanyan1 , M. Gumbo86 , T. Gunji121 , A. Gupta87 , R. Gupta87 ,
K. H. Khan15 , R. Haake50 , Ø. Haaland17 , C. Hadjidakis47 , M. Haiduc59 , H. Hamagaki121 , G. Hamar130 ,
L.D. Hanratty99 , A. Hansen77 , J.W. Harris131 , H. Hartmann39 , A. Harton13 , D. Hatzifotiadou102 ,
S. Hayashi121 , S.T. Heckel49 , M. Heide50 , H. Helstrup35 , A. Herghelegiu75 , G. Herrera Corral11 ,
B.A. Hess33 , K.F. Hetland35 , B. Hippolyte51 , J. Hladky57 , P. Hristov34 , M. Huang17 , T.J. Humanic19 ,
D. Hutter39 , D.S. Hwang20 , R. Ilkaev96 , I. Ilkiv74 , M. Inaba122 , G.M. Innocenti25 , C. Ionita34 ,
M. Ippolitov97 , M. Irfan18 , M. Ivanov94 , V. Ivanov82 , A. Jachołkowski27 , P.M. Jacobs71 , C. Jahnke115 ,
H.J. Jang65 , M.A. Janik128 , P.H.S.Y. Jayarathna117 , R.T. Jimenez Bustamante60 , P.G. Jones99 , H. Jung40 ,
A. Jusko99 , S. Kalcher39 , P. Kalinak56 , A. Kalweit34 , J. Kamin49 , J.H. Kang132 , V. Kaplin73 , S. Kar126 ,
A. Karasu Uysal66 , O. Karavichev53 , T. Karavicheva53 , E. Karpechev53 , U. Kebschull48 , R. Keidel133 ,
B. Ketzer89 , M.M. KhanIII,18 , P. Khan98 , S.A. Khan126 , A. Khanzadeev82 , Y. Kharlov52 , B. Kileng35 ,
B. Kim132 , D.W. Kim40,65 , D.J. Kim118 , J.S. Kim40 , M. Kim40 , M. Kim132 , S. Kim20 , T. Kim132 ,
S. Kirsch39 , I. Kisel39 , S. Kiselev55 , A. Kisiel128 , G. Kiss130 , J.L. Klay6 , J. Klein90 , C. Klein-Bösing50 ,
A. Kluge34 , M.L. Knichel94,90 , A.G. Knospe113 , C. Kobdaj110,34 , M. Kofarago34 , M.K. Köhler94 ,
T. Kollegger39 , A. Kolojvari125 , V. Kondratiev125 , N. Kondratyeva73 , A. Konevskikh53 ,
V. Kovalenko125 , M. Kowalski112,34 , S. Kox68 , G. Koyithatta Meethaleveedu44 , J. Kral118 , I. Králik56 ,
F. Kramer49 , A. Kravčáková38 , M. Krelina37 , M. Kretz39 , M. Krivda56,99 , F. Krizek80 , M. Krzewicki94 ,
V. Kučera80 , Y. KucheriaevI,97 , T. Kugathasan34 , C. Kuhn51 , P.G. Kuijer78 , I. Kulakov39,49 , J. Kumar44 ,
P. Kurashvili74 , A. Kurepin53 , A.B. Kurepin53 , A. Kuryakin96 , S. Kushpil80 , M.J. Kweon46,90 ,
Y. Kwon132 , P. Ladron de Guevara60 , C. Lagana Fernandes115 , I. Lakomov47 , R. Langoy127 , C. Lara48 ,
A. Lardeux109 , A. Lattuca25 , S.L. La Pointe54,108 , P. La Rocca27 , R. Lea24 , G.R. Lee99 , I. Legrand34 ,
J. Lehnert49 , R.C. Lemmon79 , V. Lenti101 , E. Leogrande54 , M. Leoncino25 , I. León Monzón114 ,
P. Lévai130 , S. Li7,67 , J. Lien127 , R. Lietava99 , S. Lindal21 , V. Lindenstruth39 , C. Lippmann94 ,
M.A. Lisa19 , H.M. Ljunggren32 , D.F. Lodato54 , P.I. Loenne17 , V.R. Loggins129 , V. Loginov73 ,
D. Lohner90 , C. Loizides71 , X. Lopez67 , E. López Torres9 , X.-G. Lu90 , P. Luettig49 , M. Lunardon28 ,
G. Luparello54 , C. Luzzi34 , R. Ma131 , A. Maevskaya53 , M. Mager34 , D.P. Mahapatra58 ,
S.M. Mahmood21 , A. Maire51,90 , R.D. Majka131 , M. Malaev82 , I. Maldonado Cervantes60 ,
L. MalininaIV,63 , D. Mal’Kevich55 , P. Malzacher94 , A. Mamonov96 , L. Manceau108 , V. Manko97 ,
F. Manso67 , V. Manzari101,34 , M. Marchisone25,67 , J. Mareš57 , G.V. Margagliotti24 , A. Margotti102 ,
A. Marı́n94 , C. Markert113,34 , M. Marquard49 , I. Martashvili120 , N.A. Martin94 , P. Martinengo34 ,
M.I. Martı́nez2 , G. Martı́nez Garcı́a109 , J. Martin Blanco109 , Y. Martynov3 , A. Mas109 , S. Masciocchi94 ,
M. Masera25 , A. Masoni103 , L. Massacrier109 , A. Mastroserio31 , A. Matyja112 , C. Mayer112 ,
J. Mazer120 , M.A. Mazzoni106 , F. Meddi22 , A. Menchaca-Rocha61 , E. Meninno29 , J. Mercado Pérez90 ,
M. Meres36 , Y. Miake122 , K. Mikhaylov63,55 , L. Milano34 , J. MilosevicV,21 , A. Mischke54 ,
A.N. Mishra45 , D. Miśkowiec94 , J. Mitra126 , C.M. Mitu59 , J. Mlynarz129 , B. Mohanty76,126 ,
L. Molnar51 , L. Montaño Zetina11 , E. Montes10 , M. Morando28 , D.A. Moreira De Godoy115 ,
TPC Upgrade TDR 155

S. Moretto28 , A. Morreale109,118 , A. Morsch34 , V. Muccifora69 , E. Mudnic111 , D. Mühlheim50 ,

S. Muhuri126 , M. Mukherjee126 , H. Müller34 , M.G. Munhoz115 , S. Murray86 , L. Musa34 , J. Musinsky56 ,
B.K. Nandi44 , R. Nania102 , E. Nappi101 , C. Nattrass120 , T.K. Nayak126 , S. Nazarenko96 ,
A. Nedosekin55 , M. Nicassio94 , M. Niculescu59,34 , B.S. Nielsen77 , S. Nikolaev97 , S. Nikulin97 ,
V. Nikulin82 , B.S. Nilsen83 , F. Noferini12,102 , P. Nomokonov63 , G. Nooren54 , A. Nyanin97 ,
J. Nystrand17 , H. Oeschler90 , S. Oh131 , S.K. OhVI,64,40 , A. Okatan66 , L. Olah130 , J. Oleniacz128 ,
A.C. Oliveira Da Silva115 , J. Onderwaater94 , C. Oppedisano108 , A. Ortiz Velasquez60,32 ,
A. Oskarsson32 , J. Otwinowski94 , K. Oyama90 , P. Sahoo45 , Y. Pachmayer90 , M. Pachr37 , P. Pagano29 ,
G. Paić60 , F. Painke39 , C. Pajares16 , S.K. Pal126 , A. Palmeri104 , D. Pant44 , V. Papikyan1 ,
G.S. Pappalardo104 , P. Pareek45 , W.J. Park94 , S. Parmar84 , A. Passfeld50 , D.I. Patalakha52 ,
V. Paticchio101 , B. Paul98 , T. Pawlak128 , T. Peitzmann54 , H. Pereira Da Costa14 ,
E. Pereira De Oliveira Filho115 , D. Peresunko97 , C.E. Pérez Lara78 , A. Pesci102 , Y. Pestov5 ,
V. Petráček37 , M. Petran37 , M. Petris75 , M. Petrovici75 , C. Petta27 , S. Piano107 , M. Pikna36 , P. Pillot109 ,
L. Pinsky117 , D.B. Piyarathna117 , M. Płoskoń71 , M. Planinic95,123 , J. Pluta128 , S. Pochybova130 ,
P.L.M. Podesta-Lerma114 , M.G. Poghosyan83,34 , E.H.O. Pohjoisaho42 , B. Polichtchouk52 ,
N. Poljak95,123 , A. Pop75 , S. Porteboeuf-Houssais67 , J. Porter71 , B. Potukuchi87 , S.K. Prasad129,4 ,
R. Preghenella12,102 , F. Prino108 , C.A. Pruneau129 , I. Pshenichnov53 , M. Puccio108 , G. Puddu23 ,
P. Pujahari129 , V. Punin96 , J. Putschke129 , H. Qvigstad21 , A. Rachevski107 , S. Raha4 , J. Rak118 ,
A. Rakotozafindrabe14 , L. Ramello30 , R. Raniwala88 , S. Raniwala88 , S.S. Räsänen42 , B.T. Rascanu49 ,
D. Rathee84 , A.W. Rauf15 , V. Razazi23 , K.F. Read120 , J.S. Real68 , K. RedlichVII,74 , R.J. Reed131,129 ,
A. Rehman17 , P. Reichelt49 , M. Reicher54 , F. Reidt34,90 , R. Renfordt49 , A.R. Reolon69 , A. Reshetin53 ,
F. Rettig39 , J.-P. Revol34 , K. Reygers90 , V. Riabov82 , R.A. Ricci70 , T. Richert32 , M. Richter21 ,
P. Riedler34 , W. Riegler34 , F. Riggi27 , A. Rivetti108 , E. Rocco54 , M. Rodrı́guez Cahuantzi2 ,
A. Rodriguez Manso78 , K. Røed21 , E. Rogochaya63 , S. Rohni87 , D. Rohr39 , D. Röhrich17 ,
R. Romita79,119 , F. Ronchetti69 , L. Ronflette109 , P. Rosnet67 , A. Rossi34 , F. Roukoutakis85 , A. Roy45 ,
C. Roy51 , P. Roy98 , A.J. Rubio Montero10 , R. Rui24 , R. Russo25 , E. Ryabinkin97 , Y. Ryabov82 ,
A. Rybicki112 , S. Sadovsky52 , K. Šafařı́k34 , B. Sahlmuller49 , R. Sahoo45 , P.K. Sahu58 , J. Saini126 ,
C.A. Salgado16 , J. Salzwedel19 , S. Sambyal87 , V. Samsonov82 , X. Sanchez Castro51 ,
F.J. Sánchez Rodrı́guez114 , L. Šándor56 , A. Sandoval61 , M. Sano122 , G. Santagati27 , D. Sarkar126 ,
E. Scapparone102 , F. Scarlassara28 , R.P. Scharenberg92 , C. Schiaua75 , R. Schicker90 , C. Schmidt94 ,
H.R. Schmidt33 , S. Schuchmann49 , J. Schukraft34 , M. Schulc37 , T. Schuster131 , Y. Schutz109,34 ,
K. Schwarz94 , K. Schweda94 , G. Scioli26 , E. Scomparin108 , R. Scott120 , G. Segato28 , J.E. Seger83 ,
I. Selyuzhenkov94 , J. Seo93 , E. Serradilla10,61 , A. Sevcenco59 , A. Shabetai109 , G. Shabratova63 ,
R. Shahoyan34 , A. Shangaraev52 , N. Sharma120,58 , S. Sharma87 , K. Shigaki43 , K. Shtejer25 ,
Y. Sibiriak97 , S. Siddhanta103 , T. Siemiarczuk74 , D. Silvermyr81 , C. Silvestre68 , G. Simatovic123 ,
R. Singaraju126 , R. Singh87 , S. Singha76,126 , V. Singhal126 , B.C. Sinha126 , T. Sinha98 , B. Sitar36 ,
M. Sitta30 , T.B. Skaali21 , K. Skjerdal17 , N. Smirnov131 , R.J.M. Snellings54 , C. Søgaard32 , R. Soltz72 ,
J. Song93 , M. Song132 , F. Soramel28 , S. Sorensen120 , M. Spacek37 , I. Sputowska112 ,
M. Spyropoulou-Stassinaki85 , B.K. Srivastava92 , J. Stachel90 , I. Stan59 , G. Stefanek74 , M. Steinpreis19 ,
E. Stenlund32 , G. Steyn62 , J.H. Stiller90 , D. Stocco109 , M. Stolpovskiy52 , P. Strmen36 , A.A.P. Suaide115 ,
T. Sugitate43 , C. Suire47 , M. Suleymanov15 , R. Sultanov55 , M. Šumbera80 , T. Susa95 , T.J.M. Symons71 ,
A. Szabo36 , A. Szanto de Toledo115 , I. Szarka36 , A. Szczepankiewicz34 , M. Szymanski128 ,
J. Takahashi116 , M.A. Tangaro31 , J.D. Tapia TakakiVIII,47 , A. Tarantola Peloni49 ,
A. Tarazona Martinez34 , M.G. Tarzila75 , A. Tauro34 , G. Tejeda Muñoz2 , A. Telesca34 , C. Terrevoli31 ,
J. Thäder94 , D. Thomas54 , R. Tieulent124 , A.R. Timmins117 , A. Toia105,49 , H. Torii121 , V. Trubnikov3 ,
W.H. Trzaska118 , T. Tsuji121 , A. Tumkin96 , R. Turrisi105 , T.S. Tveter21 , J. Ulery49 , K. Ullaland17 ,
A. Uras124 , G.L. Usai23 , M. Vajzer80 , M. Vala56,63 , L. Valencia Palomo47 , S. Vallero25,90 ,
P. Vande Vyvre34 , L. Vannucci70 , J. Van Der Maarel54 , J.W. Van Hoorne34 , M. van Leeuwen54 ,
A. Vargas2 , M. Vargyas118 , R. Varma44 , M. Vasileiou85 , A. Vasiliev97 , V. Vechernin125 , M. Veldhoen54 ,
A. Velure17 , M. Venaruzzo24 , E. Vercellin25 , S. Vergara Limón2 , R. Vernet8 , L. Vickovic111 , G. Viesti28 ,
156 The ALICE Collaboration

J. Viinikainen118 , Z. Vilakazi62 , O. Villalobos Baillie99 , A. Vinogradov97 , L. Vinogradov125 ,

Y. Vinogradov96 , T. Virgili29 , V. Vislavicius32 , Y.P. Viyogi126 , A. Vodopyanov63 , M.A. Völkl90 ,
K. Voloshin55 , S.A. Voloshin129 , G. Volpe34 , B. von Haller34 , I. Vorobyev125 , D. Vranic94,34 ,
J. Vrláková38 , B. Vulpescu67 , A. Vyushin96 , B. Wagner17 , J. Wagner94 , V. Wagner37 , M. Wang7,109 ,
Y. Wang90 , D. Watanabe122 , M. Weber34,117 , S.G. Weber94 , J.P. Wessels50 , U. Westerhoff50 ,
J. Wiechula33 , J. Wikne21 , M. Wilde50 , G. Wilk74 , J. Wilkinson90 , M.C.S. Williams102 ,
B. Windelband90 , M. Winn90 , C. Xiang7 , C.G. Yaldo129 , Y. Yamaguchi121 , H. Yang54 , P. Yang7 ,
S. Yang17 , S. Yano43 , S. Yasnopolskiy97 , J. Yi93 , Z. Yin7 , I.-K. Yoo93 , I. Yushmanov97 , V. Zaccolo77 ,
C. Zach37 , A. Zaman15 , C. Zampolli102 , S. Zaporozhets63 , A. Zarochentsev125 , P. Závada57 ,
N. Zaviyalov96 , H. Zbroszczyk128 , I.S. Zgura59 , M. Zhalov82 , H. Zhang7 , X. Zhang71,7 , Y. Zhang7 ,
C. Zhao21 , N. Zhigareva55 , D. Zhou7 , F. Zhou7 , Y. Zhou54 , Zhou, Zhuo17 , H. Zhu7 , J. Zhu109,7 , X. Zhu7 ,
A. Zichichi26,12 , A. Zimmermann90 , M.B. Zimmermann34,50 , G. Zinovjev3 , Y. Zoccarato124 ,
M. Zyzak49,39

Affiliation Notes
I Deceased
II Also at: St. Petersburg State Polytechnical University
III Also at: Department of Applied Physics, Aligarh Muslim University, Aligarh, India
IV Also at: M.V. Lomonosov Moscow State University, D.V. Skobeltsyn Institute of Nuclear Physics,

Moscow, Russia
V Also at: University of Belgrade, Faculty of Physics and ”Vinča” Institute of Nuclear Sciences,

Belgrade, Serbia
VI Permanent Address: Konkuk University, Seoul, Korea
VII Also at: Institute of Theoretical Physics, University of Wroclaw, Wroclaw, Poland
VIII Also at: University of Kansas, Lawrence, KS, United States

Collaboration Institutes
1 A.I. Alikhanyan National Science Laboratory (Yerevan Physics Institute) Foundation, Yerevan,
Armenia
2 Benemérita Universidad Autónoma de Puebla, Puebla, Mexico
3 Bogolyubov Institute for Theoretical Physics, Kiev, Ukraine
4 Bose Institute, Department of Physics and Centre for Astroparticle Physics and Space Science

(CAPSS), Kolkata, India

5 Budker Institute for Nuclear Physics, Novosibirsk, Russia
6 California Polytechnic State University, San Luis Obispo, CA, United States
7 Central China Normal University, Wuhan, China
8 Centre de Calcul de l’IN2P3, Villeurbanne, France
9 Centro de Aplicaciones Tecnológicas y Desarrollo Nuclear (CEADEN), Havana, Cuba
10 Centro de Investigaciones Energéticas Medioambientales y Tecnológicas (CIEMAT), Madrid, Spain
11 Centro de Investigación y de Estudios Avanzados (CINVESTAV), Mexico City and Mérida, Mexico
12 Centro Fermi - Museo Storico della Fisica e Centro Studi e Ricerche “Enrico Fermi”, Rome, Italy
13 Chicago State University, Chicago, USA
14 Commissariat à l’Energie Atomique, IRFU, Saclay, France
15 COMSATS Institute of Information Technology (CIIT), Islamabad, Pakistan
16 Departamento de Fı́sica de Partı́culas and IGFAE, Universidad de Santiago de Compostela, Santiago

de Compostela, Spain
TPC Upgrade TDR 157

17 Department of Physics and Technology, University of Bergen, Bergen, Norway

18 Department of Physics, Aligarh Muslim University, Aligarh, India
19 Department of Physics, Ohio State University, Columbus, OH, United States
20 Department of Physics, Sejong University, Seoul, South Korea
21 Department of Physics, University of Oslo, Oslo, Norway
22 Dipartimento di Fisica dell’Università ’La Sapienza’ and Sezione INFN Rome, Italy
23 Dipartimento di Fisica dell’Università and Sezione INFN, Cagliari, Italy
24 Dipartimento di Fisica dell’Università and Sezione INFN, Trieste, Italy
25 Dipartimento di Fisica dell’Università and Sezione INFN, Turin, Italy
26 Dipartimento di Fisica e Astronomia dell’Università and Sezione INFN, Bologna, Italy
27 Dipartimento di Fisica e Astronomia dell’Università and Sezione INFN, Catania, Italy
28 Dipartimento di Fisica e Astronomia dell’Università and Sezione INFN, Padova, Italy
29 Dipartimento di Fisica ‘E.R. Caianiello’ dell’Università and Gruppo Collegato INFN, Salerno, Italy
30 Dipartimento di Scienze e Innovazione Tecnologica dell’Università del Piemonte Orientale and

Gruppo Collegato INFN, Alessandria, Italy

31 Dipartimento Interateneo di Fisica ‘M. Merlin’ and Sezione INFN, Bari, Italy
32 Division of Experimental High Energy Physics, University of Lund, Lund, Sweden
33 Eberhard Karls Universität Tübingen, Tübingen, Germany
34 European Organization for Nuclear Research (CERN), Geneva, Switzerland
35 Faculty of Engineering, Bergen University College, Bergen, Norway
36 Faculty of Mathematics, Physics and Informatics, Comenius University, Bratislava, Slovakia
37 Faculty of Nuclear Sciences and Physical Engineering, Czech Technical University in Prague,

Prague, Czech Republic

38 Faculty of Science, P.J. Šafárik University, Košice, Slovakia
39 Frankfurt Institute for Advanced Studies, Johann Wolfgang Goethe-Universität Frankfurt, Frankfurt,

Germany
40 Gangneung-Wonju National University, Gangneung, South Korea
41 Gauhati University, Department of Physics, Guwahati, India
42 Helsinki Institute of Physics (HIP), Helsinki, Finland
43 Hiroshima University, Hiroshima, Japan
44 Indian Institute of Technology Bombay (IIT), Mumbai, India
45 Indian Institute of Technology Indore, Indore (IITI), India
46 Inha University, Incheon, South Korea
47 Institut de Physique Nucléaire d’Orsay (IPNO), Université Paris-Sud, CNRS-IN2P3, Orsay, France
48 Institut für Informatik, Johann Wolfgang Goethe-Universität Frankfurt, Frankfurt, Germany
49 Institut für Kernphysik, Johann Wolfgang Goethe-Universität Frankfurt, Frankfurt, Germany
50 Institut für Kernphysik, Westfälische Wilhelms-Universität Münster, Münster, Germany
51 Institut Pluridisciplinaire Hubert Curien (IPHC), Université de Strasbourg, CNRS-IN2P3,

Strasbourg, France
52 Institute for High Energy Physics, Protvino, Russia
53 Institute for Nuclear Research, Academy of Sciences, Moscow, Russia
54 Institute for Subatomic Physics of Utrecht University, Utrecht, Netherlands
55 Institute for Theoretical and Experimental Physics, Moscow, Russia
56 Institute of Experimental Physics, Slovak Academy of Sciences, Košice, Slovakia
57 Institute of Physics, Academy of Sciences of the Czech Republic, Prague, Czech Republic
58 Institute of Physics, Bhubaneswar, India
59 Institute of Space Science (ISS), Bucharest, Romania
60 Instituto de Ciencias Nucleares, Universidad Nacional Autónoma de México, Mexico City, Mexico
61 Instituto de Fı́sica, Universidad Nacional Autónoma de México, Mexico City, Mexico
62 iThemba LABS, National Research Foundation, Somerset West, South Africa
158 The ALICE Collaboration

63 Joint Institute for Nuclear Research (JINR), Dubna, Russia

64 Konkuk University, Seoul, South Korea
65 Korea Institute of Science and Technology Information, Daejeon, South Korea
66 KTO Karatay University, Konya, Turkey
67 Laboratoire de Physique Corpusculaire (LPC), Clermont Université, Université Blaise Pascal,

CNRS–IN2P3, Clermont-Ferrand, France

68 Laboratoire de Physique Subatomique et de Cosmologie (LPSC), Université Joseph Fourier,

CNRS-IN2P3, Institut Polytechnique de Grenoble, Grenoble, France

69 Laboratori Nazionali di Frascati, INFN, Frascati, Italy
70 Laboratori Nazionali di Legnaro, INFN, Legnaro, Italy
71 Lawrence Berkeley National Laboratory, Berkeley, CA, United States
72 Lawrence Livermore National Laboratory, Livermore, CA, United States
73 Moscow Engineering Physics Institute, Moscow, Russia
74 National Centre for Nuclear Studies, Warsaw, Poland
75 National Institute for Physics and Nuclear Engineering, Bucharest, Romania
76 National Institute of Science Education and Research, Bhubaneswar, India
77 Niels Bohr Institute, University of Copenhagen, Copenhagen, Denmark
78 Nikhef, National Institute for Subatomic Physics, Amsterdam, Netherlands
79 Nuclear Physics Group, STFC Daresbury Laboratory, Daresbury, United Kingdom
80 Nuclear Physics Institute, Academy of Sciences of the Czech Republic, Řež u Prahy, Czech Republic
81 Oak Ridge National Laboratory, Oak Ridge, TN, United States
82 Petersburg Nuclear Physics Institute, Gatchina, Russia
83 Physics Department, Creighton University, Omaha, NE, United States
84 Physics Department, Panjab University, Chandigarh, India
85 Physics Department, University of Athens, Athens, Greece
86 Physics Department, University of Cape Town, Cape Town, South Africa
87 Physics Department, University of Jammu, Jammu, India
88 Physics Department, University of Rajasthan, Jaipur, India
89 Physik Department, Technische Universität München, Munich, Germany
90 Physikalisches Institut, Ruprecht-Karls-Universität Heidelberg, Heidelberg, Germany
91 Politecnico di Torino, Turin, Italy
92 Purdue University, West Lafayette, IN, United States
93 Pusan National University, Pusan, South Korea
94 Research Division and ExtreMe Matter Institute EMMI, GSI Helmholtzzentrum für

Schwerionenforschung, Darmstadt, Germany

95 Rudjer Bošković Institute, Zagreb, Croatia
96 Russian Federal Nuclear Center (VNIIEF), Sarov, Russia
97 Russian Research Centre Kurchatov Institute, Moscow, Russia
98 Saha Institute of Nuclear Physics, Kolkata, India
99 School of Physics and Astronomy, University of Birmingham, Birmingham, United Kingdom
100 Sección Fı́sica, Departamento de Ciencias, Pontificia Universidad Católica del Perú, Lima, Peru
101 Sezione INFN, Bari, Italy
102 Sezione INFN, Bologna, Italy
103 Sezione INFN, Cagliari, Italy
104 Sezione INFN, Catania, Italy
105 Sezione INFN, Padova, Italy
106 Sezione INFN, Rome, Italy
107 Sezione INFN, Trieste, Italy
108 Sezione INFN, Turin, Italy
109 SUBATECH, Ecole des Mines de Nantes, Université de Nantes, CNRS-IN2P3, Nantes, France
TPC Upgrade TDR 159

110 Suranaree University of Technology, Nakhon Ratchasima, Thailand

111 Technical University of Split FESB, Split, Croatia
112 The Henryk Niewodniczanski Institute of Nuclear Physics, Polish Academy of Sciences, Cracow,

Poland
113 The University of Texas at Austin, Physics Department, Austin, TX, USA
114 Universidad Autónoma de Sinaloa, Culiacán, Mexico
115 Universidade de São Paulo (USP), São Paulo, Brazil
116 Universidade Estadual de Campinas (UNICAMP), Campinas, Brazil
117 University of Houston, Houston, TX, United States
118 University of Jyväskylä, Jyväskylä, Finland
119 University of Liverpool, Liverpool, United Kingdom
120 University of Tennessee, Knoxville, TN, United States
121 University of Tokyo, Tokyo, Japan
122 University of Tsukuba, Tsukuba, Japan
123 University of Zagreb, Zagreb, Croatia
124 Université de Lyon, Université Lyon 1, CNRS/IN2P3, IPN-Lyon, Villeurbanne, France
125 V. Fock Institute for Physics, St. Petersburg State University, St. Petersburg, Russia
126 Variable Energy Cyclotron Centre, Kolkata, India
127 Vestfold University College, Tonsberg, Norway
128 Warsaw University of Technology, Warsaw, Poland
129 Wayne State University, Detroit, MI, United States
130 Wigner Research Centre for Physics, Hungarian Academy of Sciences, Budapest, Hungary
131 Yale University, New Haven, CT, United States
132 Yonsei University, Seoul, South Korea
133 Zentrum für Technologietransfer und Telekommunikation (ZTT), Fachhochschule Worms, Worms,

Germany

Acknowledgements
The Collaboration wishes to thank the following persons for their contribution to the preparation of this
TDR:
A. Augustinus, M. Berger, F. Böhmer, F. Carena, F. Costa, S. Dørheim, J. Hehner, H. D. Hernández,
A. Jusko, M. Krivda, P. Martinengo, A. Di Mauro, T. Morhardt, R. de Oliveira, E. Oliveri, F. Raviel,
L. Ropelewski, G. Scioli, M. Van Stenis, A. Tauro and A. Wasem.
160 The ALICE Collaboration
References

Summary
[1] ALICE Collaboration. Upgrade of the ALICE Experiment: Letter Of Intent. CERN-LHCC-2012-
012 / LHCC-I-022. 2012. URL: http://cds.cern.ch/record/1475243/.
[2] ALICE Collaboration. Upgrade of the ALICE Read-Out and Trigger System. Technical Design Re-
port, CERN-LHCC-2013-019, ALICE-TDR-015. 2013. URL: http://cds.cern.ch/record/
1603472.

References for Chapter 1

[1] ALICE Collaboration. “The ALICE experiment at the CERN LHC”. In: JINST 3.08 (2008),
S08002. DOI: 10.1088/1748-0221/3/08/S08002.
[2] ALICE Collaboration. Upgrade of the ALICE Experiment: Letter Of Intent. CERN-LHCC-2012-
012 / LHCC-I-022. 2012. URL: http://cds.cern.ch/record/1475243/.
[3] ALICE Collaboration. Upgrade of the ALICE Inner Tracking System. Technical Design Report,
CERN-LHCC-2013-024, ALICE-TDR-017. 2013. URL: http : / / cds . cern . ch / record /
1625842.
[4] ALICE Collaboration. Technical Design Report of the Time Projection Chamber. CERN/LHCC
2000-001. 2000. URL: https://edms.cern.ch/document/398930/1.
[5] ALICE TPC Collaboration. “The ALICE TPC, a Large 3-Dimensional Tracking Device with Fast
Read-out for Ultra-high Multiplicity Events”. In: Nucl. Instr. Meth. A 622.1 (2010), pp. 316–367.
DOI : 10.1016/j.nima.2010.04.042.

References for Chapter 2

[1] ALICE Collaboration. Technical Design Report of the Time Projection Chamber. CERN/LHCC
2000-001. 2000. URL: https://edms.cern.ch/document/398930/1.

References for Chapter 3

[1] ALICE TPC Collaboration. “The ALICE TPC, a Large 3-Dimensional Tracking Device with Fast
Read-out for Ultra-high Multiplicity Events”. In: Nucl. Instr. Meth. A 622.1 (2010), pp. 316–367.
DOI : 10.1016/j.nima.2010.04.042.

References for Chapter 4

[1] ALICE Collaboration. Upgrade of the ALICE Experiment: Letter Of Intent. CERN-LHCC-2012-
012 / LHCC-I-022. 2012. URL: http://cds.cern.ch/record/1475243/.

161
162 The ALICE Collaboration

[2] F. Sauli. “GEM: A new concept for electron amplification in gas detectors”. In: Nucl. Instr. Meth. A
386.2-3 (1997), pp. 531–534. DOI: 10.1016/S0168-9002(96)01172-2.
[3] R. Veenhof. Garfield - simulation of gaseous detectors. 1984 - 2010. URL: http://garfield.
web.cern.ch.
[4] F. V. Böhmer et al. “Simulation of Space-Charge Effects in an Ungated GEM-based TPC”. In:
Nucl. Instr. Meth. A 719.0 (2013), pp. 101–108. DOI: 10.1016/j.nima.2013.04.020.
[5] S. Bachmann et al. “Discharge studies and prevention in the gas electron multiplier (GEM)”. In:
Nucl. Instr. Meth. A 479.2-3 (2002), pp. 294–308. DOI: 10.1016/S0168-9002(01)00931-7.
[6] S. Blatt et al. “Charge transfer of GEM structures in high magnetic fields”. In: Nucl. Phys. B (Proc.
Suppl.) 150.0 (2006), pp. 155–158. DOI: 10.1016/j.nuclphysbps.2004.07.005.
[7] C. Altunbas et al. “Construction, test and commissioning of the triple-GEM tracking detectors
for COMPASS”. In: Nucl. Instr. Meth. A 490.1-2 (2002), pp. 177–203. DOI: 10.1016/S0168-
9002(02)00910-5.
[8] B. Ketzer et al. “Performance of triple GEM tracking detectors in the COMPASS experiment”. In:
Nucl. Instr. Meth. A 535.1 (2004), pp. 314–318. DOI: 10.1016/j.nima.2011.06.028.
[9] B. Ketzer et al. “A triple-GEM detector with pixel readout for high-rate beam tracking in COM-
PASS”. In: Nuclear Science Symposium Conference Record, 2007. NSS ’07. IEEE. Vol. 1. Piscat-
away, NJ: IEEE, 2007, pp. 242–244. DOI: 10.1109/NSSMIC.2007.4436323.
[10] P. Abbon et al. “The COMPASS Experiment at CERN”. In: Nucl. Instr. Meth. A 577.3 (2007),
pp. 455–518. DOI: 10.1016/j.nima.2007.03.026.
[11] G. Bencivenni et al. “A triple GEM detector with pad readout for high rate charged particle trig-
gering”. In: Nucl. Instr. Meth. A 488.3 (2002), pp. 493–502. DOI: 10.1016/S0168- 9002(02)
00515-6.
[12] Z. Fraenkel et al. “A hadron blind detector for the PHENIX experiment at RHIC”. In: Nucl. Instr.
Meth. A 546.3 (2005), pp. 466–480. DOI: 10.1016/j.nima.2005.02.039.
[13] M. G. Bagliesi et al. “The TOTEM T2 telescope based on triple-GEM chambers”. In: Nucl. Instr.
Meth. A 617.1-3 (2010), pp. 134–137. DOI: 10.1016/j.nima.2009.07.006.
[14] G. Bencivenni and D. Domenici. “An ultra-light cylindrical GEM detector as inner tracker at
KLOE-2”. In: Nucl. Instr. Meth. A 581.1-2 (2007), pp. 221–224. DOI: 10.1016/j.nima.2007.
07.082.
[15] D. Abbaneo et al. “Characterization of GEM detectors for application in the CMS muon detec-
tion system”. In: Nuclear Science Symposium Conference Record (NSS/MIC), 2010 IEEE. 2010,
pp. 1416–1422. DOI: 10.1109/NSSMIC.2010.5874006.
[16] M. Alfonsi et al. “High-rate particle triggering with triple-GEM detector”. In: Nucl. Instr. Meth. A
518.1-2 (2004), pp. 106–112. DOI: 10.1016/j.nima.2003.10.035.
[17] S. Duarte Pinto et al. “Progress on large area GEMs”. In: JINST 4.12 (2009), P12009. DOI: 10.
1088/1748-0221/4/12/P12009.
[18] M. Alfonsi et al. “Activity of CERN and LNF groups on large area GEM detectors”. In: Nucl.
Instr. Meth. A 617.1-3 (2010), pp. 151–154. DOI: DOI:10.1016/j.nima.2009.06.063.
[19] A. Balla et al. “Construction and test of the cylindrical-GEM detectors for the KLOE-2 Inner
Tracker”. In: Nucl. Instr. Meth. A (2013). DOI: 10.1016/j.nima.2013.08.021.
[20] D. Abbaneo et al. “The status of the GEM project for CMS high-h muon system”. In: Nucl. Instr.
Meth. A (2013). DOI: 10.1016/j.nima.2013.08.015.
TPC Upgrade TDR 163

[21] B. Ketzer. “A Time Projection Chamber for High-Rate Experiments: Towards an Upgrade of the
ALICE TPC”. In: (2013). arXiv:1303.6694 [physics.ins-det].
[22] M. Alfonsi et al. “The triple-GEM detector for the M1R1 muon station at LHCB”. In: Nuclear
Science Symposium Conference Record (NSS/MIC), 2005, IEEE. 2005, pp. 811–815. DOI: 10.
1109/NSSMIC.2005.1596379.
[23] F. Garcia et al. “GEM-TPC Prototype for Beam Diagnostics of Super-FRS in NUSTAR Experi-
ment - FAIR”. In: Nuclear Science Symposium Conference Record (NSS/MIC), 2009, IEEE. 2009,
pp. 269–272. DOI: 10.1109/NSSMIC.2009.5401762.
[24] F. Garcia et al. “Prototype development of a GEM-TPC for the Super-FRS of the FAIR facility”.
In: Nuclear Science Symposium Conference Record (NSS/MIC), 2011, IEEE. 2011, pp. 1788–
1792. DOI: 10.1109/NSSMIC.2011.6154683.
[25] M. Kalliokoski et al. “Optical scanning system for quality control of GEM-foils”. In: Nucl. Instr.
Meth. A 664.1 (2012), pp. 223–230. DOI: 10.1016/j.nima.2011.10.058.

References for Chapter 5

[1] M. Killenberg et al. “Charge transfer and charge broadening of GEM structures in high magnetic
fields”. In: Nucl. Instr. Meth. A 530.3 (2004), pp. 251–257. DOI: 10.1016/j.nima.2004.04.
241.
[2] Leszek Ropelewski. Private communication.
[3] S Bachmann et al. “Discharge studies and prevention in the gas electron multiplier (GEM)”. In:
Nucl. Instrum. Methods Phys. Res., A 479.CERN-EP-2000-151. 2-3 (2000), 294–308. 25 p.
[4] Alessandro Cardini, Giovanni Bencivenni, and Patrizia De Simone. “The Operational Experience
of the Triple-GEM Detectors of the LHCb Muon System: Summary of 2 Years of Data Taking”.
In: (2012).
[5] M Alfonsi et al. “High-rate particle triggering with triple-GEM detector”. In: Nucl. Instrum. Meth-
ods Phys. Res., A 518 (2004), pp. 106–112.
[6] H. Schindler and R. Veenhof. Garfield - simulation of tracking detectors. 2014. URL: http://
garfieldpp.web.cern.ch.
[7] ANSYS®. Academic Research, Release 13.0. 2012.
[8] I. Smirnov. Interactions of particles with gases. URL: http://consult.cern.ch/writeup/
heed/.
[9] S.F. Biagi. “Monte Carlo simulation of electron drift and diffusion in counting gases under the
influence of electric and magnetic fields”. In: Nucl. Instr. Meth. A 421 (1999), pp. 234–240.
[10] H. W. Ellis et al. “Transport properties of gaseous ions over a wide energy range”. In: Atomic Data
and Nuclear Data Tables 17 (1976), pp. 177–210. DOI: 10.1016/0092-640X(76)90001-2.
[11] H. W. Ellis et al. “Transport properties of gaseous ions over a wide energy range II”. In: Atomic
Data and Nuclear Data Tables 22 (1978), pp. 179–217. DOI: 10.1016/0092-640X(78)90014-
1.
[12] H. W. Ellis et al. “Transport properties of gaseous ions over a wide energy range III”. In: Atomic
Data and Nuclear Data Tables 31 (1984), pp. 113–151. DOI: 10.1016/0092-640X(84)90018-
4.
[13] ALICE TPC Collaboration. “The ALICE TPC, a Large 3-Dimensional Tracking Device with Fast
Read-out for Ultra-high Multiplicity Events”. In: Nucl. Instr. Meth. A 622.1 (2010), pp. 316–367.
DOI : 10.1016/j.nima.2010.04.042.
164 The ALICE Collaboration

[14] Huntsman Advanced Materials (Schweiz) GmbH, Klybeckstrasse 200, 4057 Basel, Switzerland.
URL : http://www.huntsman.com/advanced_materials.

[15] C. Altunbas et al. “Construction, test and commissioning of the triple-GEM tracking detectors
for COMPASS”. In: Nucl. Instr. Meth. A 490.1-2 (2002), pp. 177–203. DOI: 10.1016/S0168-
9002(02)00910-5.
[16] S. Bachmann et al. “Discharge studies and prevention in the gas electron multiplier (GEM)”. In:
Nucl. Instr. Meth. A 479.2-3 (2002), pp. 294–308. DOI: 10.1016/S0168-9002(01)00931-7.

References for Chapter 6

[1] ALICE Collaboration. Technical Design Report of the Time Projection Chamber. CERN/LHCC
2000-001. 2000. URL: https://edms.cern.ch/document/398930/1.
[2] ALICE TPC Collaboration. “The ALICE TPC, a Large 3-Dimensional Tracking Device with Fast
Read-out for Ultra-high Multiplicity Events”. In: Nucl. Instr. Meth. A 622.1 (2010), pp. 316–367.
DOI : 10.1016/j.nima.2010.04.042.

[3] P. Moreira et al. “The GBT Project”. In: Proceedings of the Topical workshop on electronics for
particle physics in Paris, France, September 2125 (2009), pp. 342–346. DOI: 10.5170/CERN-
2009-006.
[4] J. Troska et al. “The Versatile Transceiver Proof of Concept”. In: Proceedings of the Topical
workshop on electronics for particle physics in Paris, France, September 2125 (2009), pp. 347–
349. DOI: 10.5170/CERN-2009-006.
[5] ALICE Collaboration. Upgrade of the ALICE Read-Out and Trigger System. Technical Design Re-
port, CERN-LHCC-2013-019, ALICE-TDR-015. 2013. URL: http://cds.cern.ch/record/
1603472.
[6] ALICE Collaboration. “Centrality dependence of the charged-particle multiplicity density at mid-
p
rapidity in Pb-Pb collisions at sNN = 2.76 TeV”. In: Phys. Rev. Lett. 106 (2011), p. 032301. DOI:
10.1103/PhysRevLett.106.032301.
[7] ALICE Collaboration. “Charged-particle multiplicity density at mid-rapidity in central PbPb colli-
p
sions at sNN = 2.76 TeV”. In: Phys. Rev. Lett. 105 (2010), p. 252301. DOI: 10.1103/PhysRevLett.
105.252301.
[8] J. Wiechula. “Commissioning and Calibration of the ALICE TPC”. PhD thesis. Goethe-Universität
Frankfurt am Main, 2009.
[9] H. K. Soltveit et al. “The preamplifier-shaper for the ALICE TPC-Detector”. In: Nucl. Instr.
Meth. A 676.0 (2012), pp. 106–119. DOI: 10.1016/j.nima.2012.02.012.
[10] H. Stelzer et al. The ALICE TPC Readout Chamber: From Prototypes to Series Production.
ALICE-INT-2003-017. 2003. URL: https://edms.cern.ch/document/384259.
[11] M. De Gaspari. “Systems-on-Chip (SoC) for applications in High-Energy Physics”. PhD thesis.
Ruprecht-Karls-Universität Heidelberg, 2012.
[12] G. Trampitsch. “Design and Characterization of an Analogue Amplifier for the Readout of Micro-
Pattern Gaseous Detectors”. PhD thesis. Graz University of Technology, 2007.
[13] L. Jönsson. Front end electronics for a TPC at future linear colliders. EUDET-Memo-2010-30.
2010. URL: http : / / www . eudet . org / e26 / e28 / e86887 / e105928 / EUDET - Memo - 2010 -
30.pdf.
[14] D. A. Huffman. “A Method for the Construction of Minimum-Redundancy Codes”. In: Proc. IRE
40.9 (1952), pp. 1098–1101. DOI: 10.1109/JRPROC.1952.273898.
TPC Upgrade TDR 165

[15] Mass production Tests of the ALICE TPC Front End Electronics. 2007. URL: http://ep- ed-
alice-tpc.web.cern.ch/ep-ed-alice-tpc/testing.htm.

References for Chapter 7

[1] ALICE Collaboration. Technical Design Report of the Time Projection Chamber. CERN/LHCC
2000-001. 2000. URL: https://edms.cern.ch/document/398930/1.
[2] ALICE Collaboration. Performance of the ALICE Experiment at the CERN LHC. 2014. arXiv:1402.
4476 [nucl-ex].
[3] ALICE Collaboration. “Centrality Dependence of Charged Particle Production at Large Trans-
p
verse Momentum in Pb–Pb Collisions at sNN = 2.76 TeV”. In: Phys. Lett. B 720.1-3 (2013),
pp. 52–62. DOI: 10.1016/j.physletb.2013.01.051.
[4] ALICE TPC Collaboration. “The ALICE TPC, a Large 3-Dimensional Tracking Device with Fast
Read-out for Ultra-high Multiplicity Events”. In: Nucl. Instr. Meth. A 622.1 (2010), pp. 316–367.
DOI : 10.1016/j.nima.2010.04.042.
p
[5] A. Ortiz Velasquez. “Production of pions, kaons and protons at high pT in sNN = 2.76 TeV Pb-Pb
collisions”. In: Nucl. Phys. A 904-905.0 (2013), pp. 763c–766c. DOI: 10.1016/j.nuclphysa.
2013.02.129.
[6] ALICE Collaboration. Upgrade of the ALICE Experiment: Letter Of Intent. CERN-LHCC-2012-
012 / LHCC-I-022. 2012. URL: http://cds.cern.ch/record/1475243/.
[7] ALICE Collaboration. Upgrade of the ALICE Inner Tracking System. Technical Design Report,
CERN-LHCC-2013-024, ALICE-TDR-017. 2013. URL: http : / / cds . cern . ch / record /
1625842.
[8] P. Christiansen et al. “The influence of detector effects on TPC performance”. In: Nucl. Instr.
Meth. A 609.1 (2009), pp. 149–155. DOI: 10.1016/j.nima.2009.08.052.
[9] G. Van Buren et al. “Correcting for distortions due to ionization in the STAR TPC”. In: Nucl. Instr.
Meth. A 566.1 (2006), pp. 22–25. DOI: 10.1016/j.nima.2006.05.131.
[10] G. Van Buren. private communication.
[11] F. V. Böhmer et al. “Simulation of Space-Charge Effects in an Ungated GEM-based TPC”. In:
Nucl. Instr. Meth. A 719.0 (2013), pp. 101–108. DOI: 10.1016/j.nima.2013.04.020.
[12] M. Mager, S. Rossegger, and J. Thomas. Composed correction framework for modeling the TPC
field distortions in AliRoot. ALICE-INT-2010-018, 2010. URL: https : / / edms . cern . ch /
document/1113105/1/.
[13] S. Rossegger, B. Schnizer, and W. Riegler. “Analytical solutions for space charge fields in TPC
drift volumes”. In: Nucl. Instr. Meth. A 632.1 (2011), pp. 52–58. DOI: 10.1016/j.nima.2010.
12.213.
[14] M. Mager, S. Rossegger, and J. Thomas. The Langevin equation expanded to 2nd order and com-
ments on using the equation to correct for space point distortions in a TPC. ALICE-INT-2010-
016, 2010. URL: https://edms.cern.ch/document/1108138/1.
[15] M. Mager, S. Rossegger, and J. Thomas. Space-charge effects in the ALICE TPC: a comparison
between expected ALICE performance and current results from the STAR TPC. ALICE-INT-2010-
017, 2010. URL: https://edms.cern.ch/document/1113087/1.

References for Chapter 8

[1] LHC Programme Coordination. LHC Luminosity Plots for the 2010 Heavy-Ion Run. Website.
2010. URL: http://lpc.web.cern.ch/lpc/lumiplots_ions_2010.htm.
166 The ALICE Collaboration

[2] LHC Programme Coordination. LHC Luminosity Plots for the 2011 Heavy-Ion Run. Website.
2011. URL: http://lpc.web.cern.ch/lpc/lumiplots_ion.htm.
[3] The ALICE DAQ group. ALICE Electronic Logbook. Website. 2012. URL: https : / / alice -
logbook.cern.ch/logbook.
[4] G. Bregliozzi. Private Communication. 2013.
[5] ALICE Collaboration. Upgrade of the ALICE Experiment: Letter Of Intent. CERN-LHCC-2012-
012 / LHCC-I-022. 2012. URL: http://cds.cern.ch/record/1475243/.
[6] D. A. Huffman. “A Method for the Construction of Minimum-Redundancy Codes”. In: Proc. IRE
40.9 (1952), pp. 1098–1101. DOI: 10.1109/JRPROC.1952.273898.
[7] T. Kollegger. “The ALICE high level trigger: The 2011 run experience”. In: Real Time Conference
(RT), 2012 18th IEEE-NPSS. 2012, pp. 1–4. DOI: 10.1109/RTC.2012.6418366.
[8] ALICE Collaboration. Technical Design Report of the Time Projection Chamber. CERN/LHCC
2000-001. 2000. URL: https://edms.cern.ch/document/398930/1.
[9] D. Decamp et al. “ALEPH: A detector for electron-positron annihilations at LEP”. In: Nucl. Instr.
Meth. A 294.1-2 (1990), pp. 121–178. DOI: 10.1016/0168-9002(90)91831-U.
[10] Walter Blum. The ALEPH handbook. CERN-ALEPH-89-077. Geneva: CERN, 1989. URL: http:
//cdsweb.cern.ch/record/227125.
[11] A. De Min et al. “Performance of the HPC calorimeter in DELPHI”. In: IEEE Trans. Nucl. Sci. 42
(1995), pp. 491–498. DOI: 10.1109/23.467923.
[12] M. Ball et al. “Technical Design Study for the PANDA Time Projection Chamber”. In: (2012).
arXiv:1207.0013 [physics.ins-det].
[13] B. Ketzer. “A Time Projection Chamber for High-Rate Experiments: Towards an Upgrade of the
ALICE TPC”. In: (2013). arXiv:1303.6694 [physics.ins-det].
[14] J. Wiechula. “Commissioning and Calibration of the ALICE TPC”. PhD thesis. Goethe-Universität
Frankfurt am Main, 2009.
[15] ALICE TPC Collaboration. “The ALICE TPC, a Large 3-Dimensional Tracking Device with Fast
Read-out for Ultra-high Multiplicity Events”. In: Nucl. Instr. Meth. A 622.1 (2010), pp. 316–367.
DOI : 10.1016/j.nima.2010.04.042.

References for Chapter 9

[1] J. F. C. A. Veloso et al. “THCOBRA: Ion back flow reduction in patterned THGEM cascades”. In:
Nucl. Instr. Meth. A 639.1 (2011), pp. 134–136. DOI: 10.1016/j.nima.2010.10.083.
[2] F. D. Amaro et al. “The Thick-COBRA: a new gaseous electron multiplier for radiation detectors”.
In: JINST 5.10 (2010), P10002. DOI: 10.1088/1748-0221/5/10/P10002.
[3] SciEnergy Co. Ltd. URL: http://www.scienergy.jp/.
[4] B. Yu et al. “Investigation of Chevron Cathode Pads for Position Encoding in Very High Rate, Gas
Proportional Chambers”. In: IEEE Trans. Nucl. Sci. 38.2 (1991), pp. 454–460. DOI: 10.1109/
23.289339.
[5] Takayuki Sumiyoshi et al. “Development of a gaseous PMT with micro-pattern gas detectors”. In:
Nucl.Instrum.Meth. A639 (2011), pp. 121–125. DOI: 10.1016/j.nima.2010.10.032.
[6] P. Colas, I. Giomataris, and V. Lepeltier. “Ion backflow in the Micromegas TPC for the future
linear collider”. In: Nucl. Instr. Meth. A 535.1-2 (2004), pp. 226–230. DOI: 10.1016/j.nima.
2004.07.274.
TPC Upgrade TDR 167

[7] S. Kane et al. A STUDY OF MICROMEGAS WITH PREAMPLIFICATION WITH A SINGLE

GEM. World Scientfic. 2004.
[8] M. Vandenbroucke et al. Discharge reduction technologies for Micromegas detectors in high
hadron flux environments. 2nd International Conference on Micro Pattern Gaseous Detectors.
2011.
[9] G. Charles et al. “Discharge studies in Micromegas detectors in low energy hadron beams”. In:
Nucl.Instrum.Meth. A648 (2011), pp. 174–179. DOI: 10.1016/j.nima.2011.05.056.
[10] M. Vandenbroucke. Development and Characterization of Micro-Pattern Gas Detectors for In-
tense Beams of Hadrons. Ph.D. Thesis, Universite Pierre et Marie Curie and Technische Universi-
tat Munchen. 2012.
[11] G. Sekhniadze. Construction and experience with a 2.4 x 1. m2 MMG chamber. RD51 meeting,
Zaragoza, 2013. 2013.

References for Chapter 10

[1] S. Schmeling. “Common tools for large experiment controls: A common approach for deployment,
maintenance, and support”. In: IEEE Trans. Nucl. Sci. 53.3 (2006), pp. 970–973. DOI: 10.1109/
TNS.2006.873706.
[2] B. Franek and C. Gaspar. “SMI++: An Object Oriented Framework for Designing Distributed
Control Systems”. In: IEEE Trans. Nucl. Sci. 45.4 (1998), pp. 1946–1950. DOI: 10.1109/23.
710969.
[3] ETM professional control GmbH, A Siemens Company, Marktstrae 3, A-7000 Eisenstadt, Austria.
URL : http://www.etm.at.

[4] C. Gaspar, M. Dönszelmann, and Ph. Charpentier. “DIM, a portable, light-weight package for in-
formation publishing, data transfer and inter-process communication”. In: Computer Phys. Com-
muni. 140 (2001), pp. 102–109. URL: http://dim.web.cern.ch/dim/papers/CHEP/DIM.
PDF.

References for Chapter 11

[1] ALICE TPC Collaboration. “The ALICE TPC, a Large 3-Dimensional Tracking Device with Fast
Read-out for Ultra-high Multiplicity Events”. In: Nucl. Instr. Meth. A 622.1 (2010), pp. 316–367.
DOI : 10.1016/j.nima.2010.04.042.

[2] H. Müller. SRS commercial status 2013. Talk at the RD51 Collaboration Meeting, Zaragosa,
Spain, July 1 – 6. 2013. URL: http : / / indico . cern . ch / contributionDisplay . py ?
contribId=129&sessionId=6&confId=258852.

References for Chapter A

[1] L. Betev and P. Chochula. Definition of the ALICE Coordinate System and basic rules for Sub-
Detector Components numbering. ALICE-INT-2003-038. 2003. URL: https : / / edms . cern .
ch/document/406391/2.
[2] ALICE Collaboration. “ALICE: Physics Performance Report, Volume II”. In: J. Phys. G32 (2006),
pp. 1295–2040. DOI: 10.1088/0954-3899/32/10/001.
168 The ALICE Collaboration
List of Figures

1.1 Inclusive e+ e invariant mass spectrum for central Pb–Pb collisions . . . . . . . . . . . 3

2.1 Schematic view of the TPC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2 View of TPC endplates and rods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3 Clearance during IROC insertion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.4 Detail of OROC insertion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.5 Side view showing clearance during OROC insertion . . . . . . . . . . . . . . . . . . . 10
2.6 Detailed view of skirt electrode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.7 Overall view of TPC with Service Support Wheel . . . . . . . . . . . . . . . . . . . . . 11
2.8 Schematic diagram of the TPC gas system . . . . . . . . . . . . . . . . . . . . . . . . . 12

3.1 Drift velocity and diffusion gas mixtures based on Ne and CO2 . . . . . . . . . . . . . . 13
3.2 Distortions at the central electrode for all gases . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Townsend and Attachment coefficients for three gas mixtures . . . . . . . . . . . . . . . 14

4.1 Photograph of standard GEM foil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

4.2 Simulated avalanche in GEM hole . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.3 Dimensions of the ALICE TPC readout chambers . . . . . . . . . . . . . . . . . . . . . 18
4.4 Exploded view of a GEM IROC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.5 Exploded view of a GEM OROC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.6 Schematic exploded cross section of the GEM stack . . . . . . . . . . . . . . . . . . . . 21
4.7 Photograph of an IROC GEM foil in the stretching frame . . . . . . . . . . . . . . . . . 22
4.8 Dimensions of an IROC GEM foil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.9 Photograph of an IROC GEM foil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.10 Detailed view of the HV distribution on an IROC GEM foil . . . . . . . . . . . . . . . . 25
4.11 Bias resistors and HV supply of an IROC GEM foil . . . . . . . . . . . . . . . . . . . . 25
4.12 Dimensions of OROC GEM foils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

169
170 The ALICE Collaboration

4.13 Schematics of the HV distribution scheme . . . . . . . . . . . . . . . . . . . . . . . . . 26

4.14 Potentials on all GEM electrodes of a quadruple GEM stack . . . . . . . . . . . . . . . 27
4.15 Point resolution in the magnetic bending plane for MWPC and GEM . . . . . . . . . . . 29
4.16 Fraction of one-, two-, three-, and four-pad clusters for MWPC and GEM . . . . . . . . 30
4.17 Cluster sizes for MWPC and GEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4.18 Pad layout of an ALICE OROC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.19 Leakage current measurement setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.20 Leakage current results from a good GEM foil . . . . . . . . . . . . . . . . . . . . . . . 35
4.21 Leakage current results from a recovered GEM foil . . . . . . . . . . . . . . . . . . . . 35
4.22 Setup of the high resolution scanning system . . . . . . . . . . . . . . . . . . . . . . . . 36
4.23 Distributions of geometrical parameters of GEM holes . . . . . . . . . . . . . . . . . . 36
4.24 Example map of hole diameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.25 Setup for gain mapping measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.26 Setup for gain mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.27 Pulse height distribution for a single pad . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.28 Relative gain map of a single GEM foil . . . . . . . . . . . . . . . . . . . . . . . . . . 39

5.1 GEM setup at TUM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

5.2 Signal and current vs. time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.3 Gain fluctuations in Ne-CO2 (90-10) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
5.4 IB and sigma for S-LP-LP-S for various DUGEM2 . . . . . . . . . . . . . . . . . . . . . 44
5.5 Measured IB vs ET1 for different settings of ET2 . . . . . . . . . . . . . . . . . . . . . . 45
5.6 2D IB scan in a quadrupole standard GEM . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.7 2D IB scan for S-S-LP-S quadruple GEM setup . . . . . . . . . . . . . . . . . . . . . . 47
5.8 IB and energy resolution for S-LP-LP-S configuration . . . . . . . . . . . . . . . . . . . 47
5.9 Energy resolution and ion backflow in a quadruple S-LP-LP-S GEM . . . . . . . . . . . 48
5.10 Probability distribution of random alignment between two GEM layers . . . . . . . . . . 50
5.11 IB vs ET1 and ET2 from measurements and simulations (Ne-CO2 -N2 (90-10-5)) . . . . . 50
5.12 IB as a function of ET1 from measurements and simulations for quadruple GEM setup . . 51
5.13 Layout of the GEM foil . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.14 Support frame details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.15 Foil gluing procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.16 Assembly of the prototype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.17 Test box with field cage and IROC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
TPC Upgrade TDR 171

5.18 HV distribution of the prototype . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

5.19 Effective gain of the chamber as a function of HV . . . . . . . . . . . . . . . . . . . . . 58
5.20 55 Fe spectra obtained in Ar-CO2 (90-10) and Ne-CO2 (90-10) . . . . . . . . . . . . . . 58
5.21 Detector equipped with front-end electronics . . . . . . . . . . . . . . . . . . . . . . . . 59
5.22 Gain map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.23 dE/dx spectrum for electrons and pions . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.24 dE/dx resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5.25 Separation power between pions end electrons. . . . . . . . . . . . . . . . . . . . . . . 62
5.26 Simulated dE/dx spectrum for 1 GeV/c pions . . . . . . . . . . . . . . . . . . . . . . . 62

6.1 Schematics of readout electronics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

6.2 Equivalent dNch /dh fluctuations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.3 Expected signal occupancy (worst case) . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.4 Schematic of SAMPA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.5 Time properties of clusters for two gases . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.6 Noise distribution on current system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
6.7 SAMPA testing robot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
6.8 Schematic of FEC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
6.9 Schematic of readout system with FEC, CRU, trigger, DCS and online farm . . . . . . . 80

7.1 Current performance: 1/pT resolution in p–Pb collisions . . . . . . . . . . . . . . . . . 82

7.2 1/pT resolution for central barrel with current and GEM TPC . . . . . . . . . . . . . . . 84
7.3 dE/dx resolution and energy resolution as function of electron transparency . . . . . . . 85
7.4 Tracking efficiency at different interaction rates . . . . . . . . . . . . . . . . . . . . . . 86
7.5 Momentum resolution at different interaction rates . . . . . . . . . . . . . . . . . . . . 87
7.6 dE/dx resolution at different interaction rates . . . . . . . . . . . . . . . . . . . . . . . 87
7.7 Average space-charge density map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
7.8 Space-charge distortions for Ne-CO2 -N2 (90-10-5) . . . . . . . . . . . . . . . . . . . . 89
7.9 Space-charge distortions at inner field cage . . . . . . . . . . . . . . . . . . . . . . . . 90
7.10 Space-charge distortions vs. epsilon . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
7.11 Ion pileup simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
7.12 Influence of space-charge fluctuations on distortions . . . . . . . . . . . . . . . . . . . . 93
7.13 Relative fluctuations of the space charge . . . . . . . . . . . . . . . . . . . . . . . . . . 94
7.14 Residual space-charge distortions after mean correction . . . . . . . . . . . . . . . . . . 95
172 The ALICE Collaboration

7.15 Fluctuation of drift distortions; z scan . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

7.16 TPC Performance after distortion corrections. . . . . . . . . . . . . . . . . . . . . . . . 96

8.1 Schematic LHC filling scheme and bunch train structure . . . . . . . . . . . . . . . . . 98

8.2 Distribution of time differences between two collisions . . . . . . . . . . . . . . . . . . 99
8.3 Schematic outline of the calibration flow . . . . . . . . . . . . . . . . . . . . . . . . . . 100
8.4 Data compression factor via cluster finding in RUN 1 . . . . . . . . . . . . . . . . . . . 101
8.5 Comparison of the tracking performance . . . . . . . . . . . . . . . . . . . . . . . . . . 105
8.6 Cluster-to-track association efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
8.7 ITS and ITS-TRD track precision inside the TPC. . . . . . . . . . . . . . . . . . . . . . 107
8.8 Space-charge calibration from track interpolation. . . . . . . . . . . . . . . . . . . . . . 108
8.9 Results from the ITS-TRD interpolation method. . . . . . . . . . . . . . . . . . . . . . 108
8.10 Resolution of the ITS-TRD interpolation method. . . . . . . . . . . . . . . . . . . . . . 109
8.11 Comparison of the momentum resolution with and without distortions . . . . . . . . . . 109
8.12 Schematic of the seeding procedure and t0 estimation. . . . . . . . . . . . . . . . . . . . 111
8.13 Deviation of t0seed from real t0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
8.14 Matching of the local-y coordinate of ITS and TPC tracks. . . . . . . . . . . . . . . . . 112
8.15 Simulated laser event . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

9.1 Photograph of COBRA 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

9.2 Working principle of COBRA GEM for the suppression of the ion backflow . . . . . . . 120
9.3 Setup for the measurement of gas gain and ion backflow . . . . . . . . . . . . . . . . . 121
9.4 Effective gain and IB in Ne-CO2 (90-10) as a function of the voltage across the GEM . . 122
up
9.5 Effective gain and IB as a function of DUAC . . . . . . . . . . . . . . . . . . . . . . . . 122
9.6 Effective gain and IB scan vs. X-ray tube current for GEM 50 – COBRA 2 – GEM 50 . . 123
up
9.7 Effective gain and IB vs. DUAC and ET2 for GEM 50 – COBRA 2 – GEM 50 . . . . . . . 124
9.8 Effective gain and IB vs. X-ray tube current for GEM 50 – COBRA 2 – COBRA 2 . . . . 124
9.9 Effective gain and IB voltage scan for GEM 50 – COBRA 2 – COBRA 2 . . . . . . . . . 125
9.10 55 Fe energy spectra with GEM 50 – COBRA 2 – COBRA 2 . . . . . . . . . . . . . . . . 125
9.11 One-, two-, and three-pad clusters for rectangular and chevron-shaped readout pads . . . 126
9.12 Space-point resolution in rj obtained with different pad geometries . . . . . . . . . . . 127
9.13 ITS-TPC momentum resolution with different pad geometries and gas mixtures . . . . . 128
9.14 Setup to measure hybrid 2 GEM + MMG system . . . . . . . . . . . . . . . . . . . . . . 129
9.15 IB, energy resolution and energy spectrum for hybrid 2 GEM + MMG system . . . . . . 129
TPC Upgrade TDR 173

10.1 DCS subsystems overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132

10.2 Schematic of the front-end electronics DCS . . . . . . . . . . . . . . . . . . . . . . . . 133
10.3 Current readout for online calibration and DCS . . . . . . . . . . . . . . . . . . . . . . 134

11.1 Yellow platform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

11.2 Schematic of current monitoring system . . . . . . . . . . . . . . . . . . . . . . . . . . 137
11.3 Schematic of HV mapping scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
11.4 Schematic of RR system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
11.5 Schematic of LV setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
11.6 Working principle of the calibration pulser system. . . . . . . . . . . . . . . . . . . . . 140
11.7 Schematic setup of the calibration pulser system. . . . . . . . . . . . . . . . . . . . . . 140

12.1 TPC upgrade project structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

12.2 Timeline of TPC upgrade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

A.1 Global coordinate system used in this document . . . . . . . . . . . . . . . . . . . . . . 147

A.2 Local coordinate system used in this document . . . . . . . . . . . . . . . . . . . . . . 148
174 The ALICE Collaboration
List of Tables

1.1 Synopsis of parameters of the upgraded TPC. . . . . . . . . . . . . . . . . . . . . . . . 6

3.1 Properties of gas mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

4.1 Parameters of GEM foils for the ALICE TPC . . . . . . . . . . . . . . . . . . . . . . . 23

4.2 Typical high voltage settings for minimal ion backflow . . . . . . . . . . . . . . . . . . 28
4.3 Dimensions and parameters of readout planes and pads . . . . . . . . . . . . . . . . . . 31

5.1 Geometry test detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

5.2 Standard and ion backflow high voltage settings . . . . . . . . . . . . . . . . . . . . . . 57

6.1 Data rates and partitioning for 5 readout partitions . . . . . . . . . . . . . . . . . . . . . 66

6.2 SAMPA parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
6.3 System parameters and ENC values for the current TPC . . . . . . . . . . . . . . . . . . 71
6.4 Contributions to the capacitances for the current TPC readout . . . . . . . . . . . . . . . 71
6.5 Simulated and measured ENC for PASA, PCA16, S-ALTRO and SAMPA . . . . . . . . 72
6.6 Occupancies and data sizes in RUN 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

7.1 Overview of simulation parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

8.1 Event size and data compression factors . . . . . . . . . . . . . . . . . . . . . . . . . . 101

9.1 Geometries of COBRA GEMs and standard GEMs used for the measurements. . . . . . 119

11.1 Time required for the most relevant activities . . . . . . . . . . . . . . . . . . . . . . . 136

11.2 Specifications of the existing LV power supply system. . . . . . . . . . . . . . . . . . . 139

12.1 List of institutions participating in the TPC upgrade. . . . . . . . . . . . . . . . . . . . . 142

12.2 Sharing of responsibilities for construction and installation of the TPC upgrade . . . . . 143
12.3 CORE cost estimate for the TPC upgrade. . . . . . . . . . . . . . . . . . . . . . . . . . 144

175
176 The ALICE Collaboration