Comments to GEM GE1/1 TDR Version 5.03 from LHCC

Chapter 1 - Introduction

1. Physics reasons for low trigger thresholds for Run 4 (post-LS3) have to be explained.

: Overall, low thresholds will remain important in Run 4 as there is a number of very difficult scenarios that will remain viable in Run-4 (such as split or compressed SUSY scenarios where soft muons can be the only objects that can be reliably triggered on, some of these scenarios have been used to argue for higher energy colliders). Other examples are models with additional Higgses, which may have small x-sections. Examples could include scenarios where heavier Higgs decays to a pair of "conventional" Higgses, and channels with one of the "conventional" Higgses decaying to pairs of taus will remain important. This is discussed in the introduction (Chapter 1, Section 1.1, around lines 850-868) and further details can be found in the introduction to Chapter 6 (4th paragraph).

However, this TDR is not emphasizing very low thresholds for a conventional (prompt) muon trigger in Run-4 because, assuming successful deployment of the tracking trigger in LS3, triggering for prompt muons will be possible with fairly low thresholds due to much improved momentum resolution brought by the tracker. The CMS exercise done as part of the Technical Proposal shows that thresholds under 20 GeV are conceivable. While it is subject to optimization of the bandwidth allocations, even lower thresholds are possible given the large increase in the allowed Level-1 bandwidth. Studies show that the use of the GE1/1-ME1/1 bending angle can give additional suppression for the single-muon trigger by reducing punch-through in jets, when an energetic pion track is matched to a stub generated by muons produced in interactions of particles in the jet with the material of the calorimeter. But overall we assume that the L1 muon trigger system's dynamic range will be adequate to operate with sufficiently low thresholds. GE1/1 will provide support in ensuring good performance of the track+muon L1 trigger by improving the quality of matching using position and muon direction in station 1. GE1/1 can also become a fallback solution should the aging ME1/1 develop performance issues.

The area where the future track trigger will fall short is triggering on muons produced in decays of long-lived particles: for impact parameters beyond a few mm, the standalone muon trigger will be the only option to preserve sensitivity to such signatures. GE1/1 will be absolutely critical for the ability to maintain standalone muon triggering past LS3 to maintain CMS sensitivity to this type of signature. As many scenarios predicting these signatures imply fairly low momenta muons, thresholds for the standalone muon trigger of the order of 25 GeV for displaced muons will certainly be only possible if the forward trigger rate problem is addressed. GE1/1 addresses that problem and so that path remains open for CMS with the deployment of GE1/1. These points are mentioned in the introduction (Chapter 1) and further discussed in detail in Section 6.2.3, including performance illustrations.

2. Reasons why GEM Phase II upgrade has to be installed in LS2 have to be summarized.

The main reason is maintaining an efficient muon triggering with sufficiently low momenta in the forward region during the period between LS2 and LS3. That time period is particularly difficult for the CMS muon triggering. Here are the main points: 1. Even with the foreseen trigger upgrade (TDR), at Run2 L=2e34 maintaining the trigger threshold at pt=22 GeV (with L1 rate= 14 kHz) is not possible without substantial additional efficiency losses in the endcaps, which represent half of the CMS muon coverage. Those additional losses are at least 15% for a perfectly working system, which brings the overall trigger efficiency down close to 80%. Improvements to the performance of the trigger with the deployment of L1 Trigger Upgrade in 2016-2017 are very limited for the region of eta>1.6 as there is no second redundant system (for eta<1.6, where between CSC/DT and RPC there are at least 8 points for the track fit with the L1 Trigger Upgrade, which reduces the rate of momentum mismeasurements dominating the trigger rate; in eta>1.6 the number of available points is at most 4). See discussion in the introduction to Chapter 6 (third paragraph) and in Sec. 6.2. 2. Without the track trigger capabilities, which will only become available after LS3, and with the limited improvements from Phase-1 L1 Trigger Upgrade, forward muon rate problem cannot be addressed efficiently until LS3. GE1/1 provides an excellent solution that addresses this problem (see details in Sec. 6.2.2.) 3. There are also considerations related to planning of the work on the central tracker and the endcap calorimeter during LS3, which make installation of GE1/1 in LS2 a much preferred option.

3. Fig. 1.2:

a) Why stand-alone trigger rates matter? In Run 4 CMS will have central tracker trigger which might or might not benefit from extra GEM chambers.

This is briefly discussed in the introduction paragraph on the bottom of page 3 beginning "After the new silicon tracker..." As far as triggering on prompt muons is concerned, that will rely on L1TrkMu trigger utilizing matching of the inner tracks with standalone muons. The quality (and purity) of the matching depends significantly on the availability of muon segments in the first station, as that is where matching utilizes both the position of the segment as well as the direction of the segment. In other stations, radial magnetic field "unbends" muons making direction measurement either less meaningful (in station 2) or nearly completely useless (stations 3 and 4). Position measurement is also most helpful in station 1, as in other stations both multiple scattering and the "unbending" (affects deflections of the positions of the segments) make matching not as efficient. GE1/1 enables good measurement of the direction of segments and, should ME1/1 develop performance issues, the redundancy provided by GE1/1 will provide position measurements in station 1.

While GE1/1 improves performance of muon+track trigger and serves as an important fallback solution should aging become a problem, the main motivation for GE1/1 after LS3 stems from the necessity to preserve CMS sensitivity to physics with displaced muons arising from decays of long-lived new particles in Phase-2. Track trigger is inefficient to tracks with impact parameters beyond a few mm, and CMS sensitivity to such signatures will uniquely rely on standalone muon trigger. Physics scenarios predicting such signatures vary substantially, typically these are low cross-section signatures requiring large luminosity, and in many cases typical momenta of muons require trigger thresholds of the order of ~20-25 GeV. These points are mentioned in the introduction (Chapter 1) and further discussed in detail in Section 6.2.3, including performance illustrations.

b) Add rate of the real muons (mainly from light quark decays, b/c quarks as well as W/Z) to this plot to indicate what are physics driven limits for the trigger rates.

The candidates reconstructed by muon triggers are dominated by true muons arising from various sources, including heavy flavor, decays in flight, electroweak processes, some come from punch through due to interactions in the material yielding true secondary muons. The problem that the CMS muon trigger has is that a fraction of these muons get reconstructed with a much higher momentum than they really are due to tails in momentum resolution. We are attaching a plot showing the trigger rates are much higher than the rate of muons arising from Ws and Zs. We also added several sentences at the end of the first paragraph in Sec, 6.2.2 to relay this information to the readers.

  • Trigger rate and W muonic rate versus pt overlaid.:

c) Why extra chambers provide only ~30% improvements in the trigger rates at ~7 GeV while ~10 times at 25 GeV?

One should first note that trigger rate is dominated by real muons with mismeasured momentum. The purpose of GE1/1 upgrade is to "unsmear" the trigger curve, so to first order it has to "push" mismeasured muons to the left where they belong while preserving the integral of the full distribution from pt=0 upwards. Therefore, such reduction in rejection power is natural. Second, at fixed signal efficiency, the improvement with the use of the bending angle is always stronger at higher pt because the bending angle is more smeared for low pt muons by multiple scattering, which further reduces discrimination. That said, one can increase the suppression at lower pt at some cost to efficiency. The exercise shown in the plot implies maintaining at least 98% efficiency near the target momentum value (efficiency continues growing reaching nearly 100% at high pt) as the goal was to demonstrate that this large reduction in trigger rate can be done without losing efficiency. We won't be able to get an order of magnitude reduction at the low pt range, but rejection can no doubt be enhanced as we optimize the final system. We added a comment on this topic at the end of the first paragraph in Sec, 6.2.2 to relay this information to the readers.

d) What is planned L1 rate in Run 3? Total and for the muon triggers.

During Run3 (starts after LS2), as in Runs 1 and 2, the L1 total rate is 100 kHz and the target muon allocation is not to exceed 20 kHz. In Run4 (post-LS2 when tracking trigger will become available), the L1 maximum rate is 750 kHz, but the single mu allocation is expected to remain within a few tens of kHz.

4. Try to escape CMS jargon in presentations of the TDR, especially abbreviations, as it makes hard for the referees to understand the proposal.


Chapter 2 - "GEM Chambers"

1. How optimization of the gas mixture is done? Only two options are presented and it looks as one of them is not viable due to greenhouse gas concerns.

Answer 1: The optimisation has been done by previous experiment running with similar GEM detectors, i.e. same gas volume geometry with 3-1-2-1mm gaps and same GEM foil design (70 micron holes diameter and 140 micron hole pitch). These experiments, LHCb and COMPASS, already performed such gas studies and we benefit from their results and experience. There is a section in the TDR describing those experiences from other experiments in some detail (Sections 2.1.2 and 2.1.3). The gas performances clearly do not depend on the overall size of the detector.

The greenhouse effect of CF4 is known and we are seriously considering any possible alternative candidates, as stated in the TDR (Section 2.2.3). As stated in the TDR, in case CF4 is completely banned and if no alternative candidate is found, the required performances in terms of triggering will be achievable using the Ar/CO2 70:30 gas mixture if one keeps in mind that each super chamber has two GE1/1 chambers working in OR mode. Based on the performance data with Ar/CO2 we obtained from a GE1/1-V during our most recent 2014 H2 test beam campaign we have illustrated that the probability for a super chamber to trigger in one 25ns clock cycle is approximately 99%.

2. What is GEM chamber discharge rate?

Answer 2: We have measured the discharge probability using an Am-241 source. To observe discharges with this source, for our gap configuration, it was necessary to operate the detector at gains in the range of 4e5 to 6e5. Here the discharge probability was measured to be 1e-5 to 1e-3. However, these operating conditions are well beyond those that will be used in CMS as can be seen from Figure 2.8 in the TDR. Extrapolating the experimental data to a gain of 5e3 (reasonable CMS operating condition) the discharge probability is found to be approximately 9.2e-10. The Am-241 source is an alpha emitter and causes a factor of ~10^2 more primaries to be created than the MIP case. Thus for MIPs the discharge probability should be divided by this factor to be approximately 1e-12 to 1e-11. Similar behavior to this has been reported in the literature by the COMPASS experiment, see for example Figure 22 of S. Bachmann et al. Nuclear Instruments and Methods in Physics Research A 470 (2001) 548–561.

Added Section on discharge probability.

3. Are you planning to repeat long-term stability and aging tests with final version of the chamber assembly?

Answer 3: The final version of our chamber will not have new material or any different mechanical solution with respect the final prototypes assembled. Our aging studies at GIF++ will continue with final prototypes until the projected 100mC/cm^2 for the HL-LHC is reached. Outgassing test on chamber materials will be continued to accumulate statistics and to confirm the present positive results.

4. Selected pitch of the sensitive electrodes provides resolution more than twice better of what is required. Why not to increase pitch (and size of the pick-up electrodes)? Doubling pitch will simplify production and reduce cost of on-chamber electronics by about a factor of two (as well as reduce costs of the power supplies, etc.).

Answer 4: The requested angular resolution of 300 microrad is an upper limit, lower is better. In addition to resolution, chamber alignment also needs to be considered, so there is a safety margin built into the actual resolution. Our collaboration has profited by the already existing VFAT2 chip (of which an upgraded design is being developed). To maintain the usability of this chip for the slice test we need to ensure the collected charge per strip does not exceed the dynamic range of the chip (i.e. ensures undesirable behavior due to saturation does not occur). This, in addition to the angular resolution and alignment concerns mentioned above, drives the choice of the strip pitch.

Added a paragraph in Section in the "Angular resolution" part.

5. Did you consider replacement of CF4 with small amount of CH4 which is a good quencher, fast gas, and in small % amounts not flammable? [Kadyk, Wise, V'avra 1979-88; 2001 + many others]

Answer 5: We know CH4 is a good quencher and quite fast but we would like to avoid flammable gas for obvious safety reasons. If CERN will ban CF4, GE1/1 can run safely with AR/CO2 70:30

The most compelling reason for CF4 over CH4 is shown in the literature: With CF4 higher drift velocities can be obtained by a factor of ~2 over the considered range of electric field when using CF4 in place of CH4. We see that the total number of electron-ion pairs per cm is 100 (53) for CF4 (CH4). So we the spatial distribution of primary ionization clusters for mixtures w/ CF4 is more uniform than for mixtures w/ CH4. Consequently, eplacing CF4 with a small amount CH4 leads to a poorer detector time resolution which would be undesired. Tests are ongoing on different molecules, similar to banned ones but with lower Global Warming / Ozone Depleting Power (3,3,3-tetrafluoropropene HFO-1234yf , 1,3,3,3-tetrafluoropropene HFO-1234ze, 3,3,3-trifluoropropene HFO-1233zd , trifluoroiodo- CF3i)

Regarding CH4 vs CF4: In the literature it is well illustrated that CH4 will contribute to aging of wire chambers and SWPC. In Kadyk’s paper they illustrate for wire chambers that Ar/CH4 80/20 gives an approximate gain loss of less than 10% per C/cm (loss defined as (-1/G)*dG/dQ) while Ar/CO2 80/20 gives an approximate gain loss of 0% per C/cm.

In Va’vra’s paper the bond strength of C-H (F-C) is shown to be 4.3 (5.4) eV and they state the energy to destroy CH4 (CF4) is 4.5-4.6 (5.2-7.8) eV depending on the reaction process. As a result CH4 will polymerize more easily than CF4.

Added paragraph in Section 2.2.3

6) Did you consider option of using CSC vs. GEM? Fluxes after Phase II upgrade in the region of these chambers are reasonable, so even typical wire based detector should work.

Answer 6: Yes, but the available space and envelope for installation spatial resolution considerations, total charge integrated on wires in a CSC and rate capabilities indicated that the choice of GEM technology to be used in that region is the most reliable one, as demonstrated by similar upgrades in LHCb and ATLAS based on micropattern detectors.

Added a sentence in Section 2.1.1

Chapter 3 and 4

1. What is the advantage of running the VFAT3 in "lossless" mode compared to running in SPZS mode?

A.C PA GDL & AS: The advantage of the lossless mode is that the 128 bits are not compressed, so that the receiver should not have a decompression algorithm. The disadvantage is that transmits the 128 bits even if only one of them is at "1", with enormous waste of bandwidth.The lossless is simple, and compatible with the VFAT2 Instead the SPZS mode implements a kind of zero suppression, so it will uses the band in a more efficient way but it requires additional implementation on the receiver. We are going to make the system configurable in order to be able to switch between the two options. With 1 time slot per event both SPZS and lossless satisfy the bandwidth requirements for the high rate environment. If 3 time slots per trigger are preferred then only SPZS satisfies the bandwidth requirements. Lossless will be very important during the debugging phase.

A paragraph has been added at the end of section

2. Timing resolution will be critical. What is the expecting timing dispersion per VFAT3 chip and per superchamber?

Timing resolution is dominated by the detector performance. The spread will also be dominated by detector performance. The spread from chip to chip will be very small, expected around the ~% level.

A sentence has been added in the next to last paragraph of section 3.2.1.

3. For a prototype chamber, 97% of the hits are associated with the correct bunch. What does this do to the fake trigger rate? That is, how sensitive is the trigger rate to this fraction? Suppose it were 95%, what would happen?

The results from the test beam show that even without a precise synchronization and phasing relative to the LHC bunch structure ( which we can reach using the LHC Timing, Trigger and Control (TTC) system) we were able to have a 95% of hits in 25 ns time window, which simulate approximately the LHC bunch crossing. Simulation used to design the matching GEM-CSC algorithm (to be implemented in the CSC ME1/1 OTMB FPGA) has been tuned to the measurements obtained in the test beam. The current algorithm has been designed to have modest sensitivity to timing mismeasurements (most of the time, it requires only one of the two GE1/1 hits to be reconstructed in the right BX), and the next version of the algorithm will use +/1 BX matching window in comparing CSC and GEM data. That feature will allow mitigating effects of even a very significant reduction in the GE1/1 timing resolution (e.g. taking ArCO2 case as the worst case scenario). We added a new paragraph at the end of Section 6.2.1. providing these details to the readers.

4. What inspection will be done on the GEM foils itself, beside tensile strength?


1) An optical inspection is first performed to identify defects, scratches, irregular hole sizes, and contact between top and bottom metalized surfaces. A leakage current test is part of the quality control of the GEM foils. A microscope is also used when necessary to further investigate defects.

2) The quality of the foil (leakage current and impedance) is checked using a picoammeter. With an applied potential difference of 500 V between the GEM metal sides, the GEM foil should draw a current of no more than 30 nA.

Discussed in chapter 5, section 5.2

5. What is the critical path for the production?

AS: The VFAT3 chip; however we have allowed for sufficient time, modularity, quality controls and even two foundry runs to ensure that the risk is mitigated. VFAT3 design and submission. Two submissions have been foreseen in the schedule and budget. Each VFAT3 module is designed to be testable independently to the other modules. This facilitates any debugging. In addition we intend to start the test board and routines before the submission of VFAT3 via a separate designer to the VFAT3 designers. This will allow the designer to concentrate on design and any bug fixes necessary whilst the test and characterisation is performed in parallel.

This answer will not be added in the TDR.

6. Regarding the slice test (appendix A): given that the final VFAT and GBT will not be available by summer 2016, what running information will not be provided by the slice test and how will that information be obtained?

AS, AC: The Slice Test uses VFAT2 which has a trigger resolution of 16 channels and readout at 40 MHz. The maximum trigger rate is limited to 200kHz. Also the LV1A latency is limited to up to 6us. Nevertheless the system will emulate the VFAT3 and GBT protocols hence it will be transparent to their non availability:

1. DAQ system will be integrated in CMS DAQ;

2. combined CSC+GEM trigger

3. Operation procedure implemented

4. Reconstruction included in official CMSSW

5. Validation done with standard tools

6. Background and noise rate included in simulation.

Once installed and commissioned we will:

1. Gain integration experience with the final electronic system;

2. Reduce the GEM commissioning period

3. Back-end electronics installed and commissioned in advance of the installation of the FE electronics.

4. All components (Incl. detectors) will have been qualified beforehand at the TIF

5. Trigger commissioning and performance check;

6. Background measurement

7. Cross-check with data what expected by simulation.

A paragraph has been added at the end of section A3.

Chapter 9 - Organization, responsibilities, planning and costs

1. Fig. 9.1. How "GEM collaboration" is interacting with CMS collaboration, especially with Phase 2 CMS wide upgrade management, resource management, Technical Coordinator, etc.?

MT,PK,AS: These answers will not be added to the TDR In 2014, the GEM project was officially recognized by the CMS Muon-IB as a new subsystem/subgroup, next to the existing muon subsystems (CSC, DT and RPC groups). When CMS officially endorses the GE1/1 construction project, the GEM Collaboration will disappear as a separate entity and will instead be fully absorbed as a regular CMS muon subgroup with a standard CMS management structure. As shown in Fig.9.1, already now, the GEM Collaboration is organized in subgroups, with a management structure of a typical CMS subsystem.

Today, the GEM Project Manager along with corresponding coordinators/managers interacts with the central CMS management, including Upgrade and Resource Management and Technical Coordination. As shown in Fig.9.1 and elaborated on in the text, the GEM Collaboration management already includes a dedicated coordinator/manager for each of these items. Once fully absorbed into the CMS Collaboration, the Muon-GEM subgroup will continue to have a dedicated Resource Manager, Technical Coordinator, Phase-2 Upgrade Coordinator, Detector Physics Group Coordinator and so on, that will interact with the rest of CMS in the same way as present subsystems do.

2. It is hard to clearly identify roles of the Chairperson and Project Manager. Even charge for them is "mixed" in the paragraph 9.3. Could you split and define responsibilities for each of these two management positions separately?

MH: While the PM will be coordinating all technical tasks, the IB Chair will focus more on the institutional side ensuring that all collaborating institutions can engage as planned, e.g. in terms of manpower and collaborating on tasks. MT: text will be modified, clearly identifying the role of the IB Chairperson and Project Manager

3. There are a few items in the task matrix (Fig. 9.2) where few groups are involved, like single group in the hardware alignment. What are risks associated with such assignments?

MT: Today, the interest in the GEM project is still increasing. Many groups have expressed interest to join the effort. When new groups join the collaboration, they will be asked to cover mainly the underpopulated areas of the project.

AS: There is no risk. In particular the alignment group is a consortium of several different Universities and Research Institutions in Hungary. Traditionally, they have looked after the hardware alignment of the in the muon system, hence expertise lies there.

Explanation: Hungarian institutes (ATOMKI and University of Debrecen , Wigner from Budapest) are traditional hosts of CMS muon alignment activity since the beginnings (~20 years). The financial background is also common, either joint or consorcial fund in which all the collaborating partners are involved. The same policy is applied in the muon institutional board where the Hungarian institutes have a common representative who is leading the muon alignment activity in CMS.

In addition new groups, as mentioned earlier, have expressed interest.

4. What is happening during a year from 6/2017 to 6/2018 after all chambers are ready, but not yet installed (Table 9.1)? AS: Improved text and an improved schedule with dependencies has been added to the TDR.

AS: QC7 until QC10 is foreseen after inserting final electronics and the superchambers . PK: Text will be added to describe the storage facility and tracking procedure.

5. Provide detailed cost estimates for a subset of main items to understand how major costs are actually determined. This should include cost estimates from companies, etc. Concentrate on the main items, such as chambers components, electronics, power supplies (including HV), etc.

AS: A detailed cost spreadsheet was prepared with all major components, including quality flags. That cost estimate was made following the LHC CORE rules and was examined internally by the CMS cost committee; it is used for an improved summary table and pie charts in the TDR are included.

6. Provide expected needs in FTEs from the collaborating institutions (not accounted in the core cost estimate) vs year required to accomplish the project by main categories: scientist (including students), technician, engineer.

AS: The project needs about 70 FTE in total for construction and installation with a distribution as follows: 72% Physicists, 12 % Technicians and 16% Engineers, this phrase is added in the TDR A.C : Here is the FTE table as of today March 20th 2015 - . PK: The collaborating institutes have pledged support for installation, commissioning, and operation. Based on extensive experience of the collaborators with the CSC and RPC systems, we will include a table summarizing the FTE needs.

-- ArchanaSharma - 2015-02-27

Edit | Attach | Watch | Print version | History: r39 < r38 < r37 < r36 < r35 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r39 - 2016-01-03 - MichaelTytgat
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    MPGD All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright &© 2008-2023 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback