Investigation and Mitigation of Crosstalk in the Prototype ME0 GEM Detector for the Phase-2 Muon System Upgrade of the CMS Experiment
TThis work has been submitted to the conference proceedings of the 2020 IEEE NSS-MIC Conference forpublication. Copyright may be transferred without notice, after which this version may no longer be available.
Investigation and Mitigation of Crosstalk in thePrototype ME0 GEM Detector for the Phase-2Muon System Upgrade of the CMS Experiment
Stephen D. Butalla and Marcus Hohlmann,
On behalf of the CMS Muon Group
Abstract —The LHC is currently undergoing a high lumi-nosity upgrade, which is set to increase the instantaneousluminosity by at least a factor of five. This luminosityincrease will result in a higher muon flux rate in the forwardregion and overwhelm the current trigger system of the CMSexperiment. The ME0, a gas electron multiplier detector, isproposed for the Phase-2 Muon System Upgrade for the CMSexperiment to help increase the muon acceptance and to con-trol the Level 1 muon trigger rate. A recent design iterationof this detector features GEM foils that are segmented onboth sides, which helps to lower the probability of highvoltage discharges. However, during preliminary testing ofthe chamber, substantial crosstalk between readout sectorswas observed. Here, we investigate, characterize, and quan-tify the crosstalk present in the detector, and also estimate theperformance of the chamber as a result of this crosstalk viasimulation results of the detector dead time, efficiency loss,and frontend electronics response. The results of crosstalk viasignals produced by applying a square voltage pulse directlyon the readout strips of the detector with a signal generatorare summarized. We also present the efficacy of mitigationstrategies including bypass capacitors and increasing the areaof the HV segments on the third GEM foil in the detector.We find that the crosstalk is a result of capacitive couplingbetween the readout strips on the readout board and betweenthe readout strips and the bottom of the third GEM foil. Ourresults show that the crosstalk generally follows a patternwhere the largest magnitude of crosstalk is within the sameazimuthal readout segment in the detector, and in the next-nearest horizontal segments in eta. Generally, the bypass ca-pacitors and increased area of the HV segments successfullylower the crosstalk in the sectors where they are located;on average, we observe a maximum decrease of crosstalk insectors previously experiencing crosstalk from (1 . ± . to (1 . ± . with all HV segments connected in parallelon the bottom of the third GEM foil, with the addition ofan HV low-pass filter connected to this electrode, and anHV divider. However, with these mitigation strategies, wealso observe slightly increased crosstalk (cid:0) (cid:47) . (cid:1) in readoutsectors farther away. Manuscript received December 20, 2020. This work was supportedin part by the U.S. Department of Energy Office of Science (HEP) andthe National Science Foundation (Cornell University Subaward; P.I. Dr.Marcus Hohlmann).S. D. Butalla and M. Hohlmann are with the Aerospace, Physicsand Space Sciences Department at Florida Institute of Tech-nology, Melbourne FL, 32901 (email: sbutalla2012@my.fit.edu andhohlmann@fit.edu).
I. INTRODUCTION T HE high luminosity upgrade of the LHC at CERN isprojected to increase the instantaneous design lumi-nosity by at least a factor of five. In order to cope with theincreased muon flux rates from this higher luminosity,the CMS experiment is undergoing the Phase-2 MuonSystem upgrade [1], which will increase the redundancyof the muon system, as well as lower the Level 1 triggerrate. One of the detectors to be installed during theupgrade is the ME0 triple-Gas Electron Multiplier (GEM)detector, which will increase the muon acceptance from | . | to | . | in pseudorapidity (see Fig. 1), and also helpcontrol the Level 1 trigger rate. z (m) R ( m ) θ ° η θ ° η M E / M E / M E / M E / M E / M E / M E / M E / R E / R E / R E / MB1MB2MB3MB4Wheel 0 Wheel 1 RB1RB2RB3RB4Solenoid magnetSilicon tracker SteelWheel 2 R E / R E / M E / R E / R E / R E / CSCsRPCsDTs R E / G E / G E / GEMs R E / iRPCs R E / ME0 M E HGCALECALHCAL
Fig. 1:
Quadrant of the upgraded CMS experiment with theME0 in orange [1].
The recent design of this detector differs from a previ-ous generation of CMS GEM detectors, the GE1/1, in thatit features GEM foils with high voltage (HV) segmentswith protection resistors on both sides of the foils, whichfunction to protect the chamber from HV discharges.By contrast, only the side of the GEM foils in GE1/1detectors facing the drift electrode is segmented. Thefoils are divided into 37 HV segments with a range ofareas from 98.7 cm to 103.4 cm . During the quality con-trol testing of the first prototype, we observed crosstalkin neighboring readout sectors. This paper provides asummary of the investigations of the crosstalk in the a r X i v : . [ phy s i c s . i n s - d e t ] F e b E0 detector. We characterize and quantify the crosstalkpulses in this detector by injecting signal pulses into thereadout sectors with a signal generator (see Fig. 2 fora picture of the ME0 GEM detector and the associatedreadout sector partitioning). We also discuss simulationresults of expected detector efficiency loss, which arebased on results from crosstalk rate measurements usingpulses created by alpha and beta sources, backgroundrate simulations, and dead time simulations from theresponse of the frontend application specific integratedcircuit (ASIC) mounted on hybrid cards due to thecrosstalk. Finally, we discuss mitigation techniques in-cluding the use of bypass capacitors on the GEM foiland increasing the area of the HV segments on the thirdGEM foil to reduce the magnitude of the crosstalk inother readout segments.
Fig. 2:
The CMS ME0 triple-GEM detector and its readoutsector labeling. Sector numbering ranges azimuthally from 1to 3 and vertically from 1 to 8.
II. C
HARACTERIZATION OF THE C ROSSTALK
During the preliminary alpha-irradiation testing of theprototype ME0 detector, we observed bipolar pulses insectors that were not being irradiated. To investigatefurther, we injected a square pulse of 1 microsecondwidth into a readout sector (128 readout strips), andthen read out the crosstalk signal on 128 ganged stripsin all of the other readout sectors with an oscilloscope.To measure and quantify the crosstalk, these oscilloscopetraces were recorded and their amplitudes manuallymeasured. Figure 3 displays such a measurement: onchannel 1 of the trace (top), we see the injected square A typical pulse width in a GEM detector is on the order of 10ns. However, with the capabilities of the signal generator used in thisstudy, the amplitude of the pulse was preserved, but became distorteddue to an impedance mismatch below one microsecond. Thus, we useda one microsecond pulse width for these studies. pulse, and on channel 2 (bottom), we see the resultingcrosstalk pulse in an adjacent readout sector.
Fig. 3:
An example oscilloscope trace of the input square pulse(channel 1, top) and the crosstalk signal (channel 2, bottom).
The crosstalk percentage is quantified as the ratio ofoutput pulse amplitude to the input amplitude, mul-tiplied by 100%, with the error given by the standarderror propagation formula below. These tests showed amaximum crosstalk amplitude of (6 . ± . of theinjected amplitude. XT % = V out V in · (1) δ ( XT %) = | XT | (cid:115)(cid:18) δV in V in (cid:19) + (cid:18) δV out V out (cid:19) · (2)Comprehensive crosstalk “maps” (see an example inFig. 4) were made by reading out the signal in all ofthe other readout (RO) sectors in the chamber. Thesemaps identify the largest observed crosstalk and theextent to which neighboring sectors in the chamber areaffected. Sectors with 0.00% crosstalk were those thatdid not display a crosstalk signal differentiable frombaseline. Consequently, the error listed for these sectorsis undefined as prescribed by (2). The range of theaverage observed crosstalk across the detector for pulseinputs into each φ partition of three η segments are listedin Table I. TABLE I:
Range of Observed Crosstalk for Pulse Inputs IntoEach of the Three Phi Partitions of the Listed Eta Segments. η Sector Minimum Crosstalk (%) Maximum Crosstalk (%)1 0.24 ± ± ± ± ± ± The crosstalk signal is a result of CR differentiation:the capacitive coupling between RO sectors (and thecoupling due to the bottom electrode of GEM3 that facesthe strips) results in an average measured intersector ig. 4:
An example crosstalk map with pulse input in sector ( η = 5 , φ = 2) . Note the nearly symmetric behavior in adjacent φ partitions (and η partitions). Sectors with XT = 0 . showed no discernible crosstalk. capacitance C = 702 ± pF), and the resistance is thatof the 50 Ω characteristic impedance of the cable. Thetime constant is then τ ≈
35 ns. This hypothesis wasverified by examining the time constant of the observedcrosstalk pulses and also by varying the input squarepulse widths T , which shows the characteristic behaviorof a CR differentiator for pulse widths T (cid:29) τ and T ≈ τ (see Fig. 5). These results are also verified via circuitsimulations reported in a complementary paper by M.Hohlmann submitted to these proceedings, see [2].III. E STIMATING THE I MPACT OF THE C ROSSTALK ON D ETECTOR P ERFORMANCE
To estimate the impact that this crosstalk has on de-tector performance, experimental results determining theprobability of observing a crosstalk pulse were used asinput parameters to a simulation of the ensuing back-ground rate in the detector from crosstalk.. To determinethe crosstalk probability, a GE1/1 GEM detector withdouble-segmented foils was irradiated with alpha andbeta sources through a small hole in the GEM drift cath-ode PCB, and the hit rate of the pulses above thresholdwere recorded. Dead time and timing error simulationsof the frontend ASIC hybrid cards were performed byinjecting a signal pulse at a fixed time into the simulatedshaping circuit, and then varying the injection time of acrosstalk pulse into the same, simulated circuit of theASIC. It was found that a maximum timing error ofabout 550 ns results from the interference of the largecrosstalk signal with actual signal pulses on neighboringstrips. Results of heavily-ionizing background particlerate simulation in CMS were used in tandem with the
Width = 20 ns
Width = 100 ns
Width = 250 ns
Width = 1000 ns
Fig. 5:
Crosstalk pulse shapes for different input square pulsewidths. Note that below 1 µ s, the input pulses were distorteddue to impedance mismatch between the cable and the stripsector. imulations to determine the loss of efficiency of thedetector. Figure 6 displays the results of these studies,which shows the nominal detector efficiency and theefficiency loss from the dead time due to crosstalk, foreach readout partition ( η number) in the detector. Fig. 6:
Plot of the simulated reconstruction efficiencies and thelosses due to crosstalk with increasing pseudorapidity (for animpact time equivalent to 50 bunch crossings).
IV. MITIGATION STRATEGIES FOR REDUCINGCROSSTALKSeveral mitigation strategies were tested to amelioratethe crosstalk: increasing the area of the HV segments onthe bottom of the third GEM foil (GEM3B), both with andwithout a low-pass filter, and installing bypass capacitorsin one η segment. The idea is to reduce the impedanceto ground for AC signals by increasing the capacitancebetween strips and the bottom of the GEM3 foil or bycreating a direct AC path with a capacitor [2]. For thefirst study, we soldered five, 330 pF bypass capacitors inparallel with the protection resistors to the HV segmentsin the η = 8 sector on the bottom of the third GEM foil,and removed the protection resistors in η = 5 , connectingthese HV segments in parallel to increase the capacitanceof the third GEM foil (see Fig. 7).Crosstalk maps were then taken for pulse inputs intoeach RO connector in η = 5 , . We then repeated crosstalkmeasurements with all of the 37 HV segments on GEM3Bconnected in parallel with solder (in effect recreatingan unsegmented GEM3B electrode). This configurationwas measured both with a low-pass circuit and bothwith and without the HV divider. An example map forthe configuration of all HV segments connected, a lowpass circuit, and the HV divider, along with the originalmap in the baseline configuration, as well as a map thatdisplays the total change in percentage, is presented inFig. 8. We see that although new sectors farther awayfrom the pulse injection are experiencing a small amountof crosstalk, there is an overall decrease in the sectorspreviously suffering from crosstalk. The results, quoted Fig. 7:
The bypass capacitors (covered with a layer of Kaptontape) installed in η = 8 on GEM3B (top) and the HV segmentsconnected in parallel with solder in η = 5 on GEM3B (bottom). as a change in percentage of crosstalk amplitudes, arelisted in Table II.For a summary of all mitigation strategies for pulseinput into readout sector (5,2), see Fig. 9. It should benoted that a small value of crosstalk (cid:0) (cid:47) . (cid:1) wasobserved in all φ partitions in all η sectors after thesemodifications were made. This is expected because thecontiguous GEM3 bottom can couple a small amount ofcrosstalk into all RO sectors. Overall, the average ob-served crosstalk is reduced by each mitigation strategy,with the largest decrease in crosstalk occurring when thethird GEM foil is contiguous (i.e., protection resistors onthe top-side of the foil, only), with the HV divider andlow-pass filter connected.V. S UMMARY AND C ONCLUSION
During the initial performance testing of a prototypeME0, a triple-GEM detector proposed for thePhase-2 Muon Upgrade of the CMS experiment, sub-stantial crosstalk was observed in neighboring readout ig. 8:
Crosstalk maps of the observed crosstalk percentage for the unmodified, segmented GEM3B (left), the modified version ofGEM3B where all HV segments are connected in parallel with solder, with HV filter and HV divider connected (center), and thechange in the crosstalk percentage between the unmodified and the modified chamber (right), for pulse input into sector (5,2).
Fig. 9:
Summary plot of the observed crosstalk percentage in all sectors for pulse input into (5,2) for different mitigation measures.
ABLE II:
Average Change in Crosstalk for All Sectors Previously Experiencing Crosstalk With Bypass Capacitors ( η = 8 ) andHV Segments Connected ( η = 5 ), and With All HV Segments Connected on GEM3B with HV Filter, Both With and Without anHV Divider Pulsing into Bypass Cap. & HV segments GEM3B Continuous, HV Filter GEM3B Continuous, HV Filterconnected in η = 5 (w/o HV Divider) (w/ HV Divider) η = 8 (-0.47 ± ± ± η = 5 (+0.03 ± ± ± η = 1 N/A (-0.17 ± ± ± ± ± sectors. This motivated our investigations to better un-derstand and mitigate this crosstalk, as outlined in thispaper. By applying square voltage pulses to all 128 read-out strips in a RO sector, and reading the signal out ofeach remaining RO sector, we observe that, in the versionof the ME0 triple-GEM detector where GEM foils haveHV protection resistors on both sides, a range of crosstalkbetween 0.24%–6.40% is seen across all φ partitions ofa readout sector that is being pulsed, with crosstalkextending to the nearest neighboring η segments. Thiscrosstalk is due to the capacitive coupling between ROsectors and the coupling between RO sectors and theelectrode at the bottom of GEM3. We see that whilea small value of crosstalk is introduced into other ROsectors by making the bottom of the third GEM foilcontiguous again, the crosstalk is successfully reduced,with a maximum average reduction from (1 . ± . to (1 . ± . with a low-pass filterand HV divider. Simulations of the frontend hybrid cardASIC circuit response indicate that without these modi-fications, we could expect a maximum efficiency loss of ∼
6% when operating the detector in CMS. Consequently,a design with a contiguous bottom electrode on GEM3but double-segmented GEM1 and GEM2 foils has beenadopted by the CMS muon group as the final designfor mass production of GE2/1 foils (see Fig. 1 for thelocation of the GE2/1 chambers in the endcap of theCMS experiment), which face the same crosstalk issuesas the ME0 module discussed here.We note that the purpose of the initial double-segmentation in all three GEM foils was to limit dis-charge propagation and rate in the GE2/1 detector. Testsshow that the discharge probability is reduced by threeorders of magnitude [3]. The effect on the discharge prob-ability with the adopted configuration (with the thirdGEM foil segmented only on one side and the other twofoils segmented on both sides) is currently under study.Preliminary results indicate that this “mixed design”can simultaneously reduce discharge propagation andcrosstalk with final results to be published soon.A
CKNOWLEDGMENTS
We gratefully acknowledge support from FRS-FNRS(Belgium), FWO-Flanders (Belgium), BSF- MES (Bul-garia), MOST and NSFC (China), BMBF (Germany), CSIR(India), DAE (India), DST (India), UGC (India), INFN(Italy), NRF (Korea), QNRF (Qatar), DOE (U.S.A.), andNSF (U.S.A.). We would also like to thank all members of the CMS GEM group for their contributions to thisproject. R
EFERENCES[1] CMS Collaboration, “The Phase-2 Upgrade of the CMS MuonDetectors Technical Design Report”, CERN-LHCC-2017-012, CMS-TDR-016, Sep. 12, 2017. Available: https://cds.cern.ch/record/2283189?ln=en.[2] M. Hohlmann, “A Model for Crosstalk in Micro-Pattern GasDetectors,” in
Proc. Virtual 2020 IEEE Nuclear Science Symp. andMedical Imaging Conf. (NSS/MIC) , Boston, MA, USA, 2020, (theseproceedings).[3] J. Merlin, “GEM stack optimization for GE21 and ME0 (cross-talkstudies, discharge mitigation),” presented at the