Resolution Studies of GEM/Timepix Detector with 5 GeV electrons
A. Bamberger, K. Desch, U. Renz, M. Titov, N. Vlasov, P. Wienemann, S. Zimmermann, A. Zwerger
aa r X i v : . [ phy s i c s . i n s - d e t ] S e p Resolution Studies of GEM/Timepix Detector with 5GeV electrons
A. Bamberger , K. Desch , U. Renz , M. Titov , , N. Vlasov , P. Wienemann ,S. Zimmermann and A. Zwerger
1- Albert-Ludwigs University of Freiburg, Physics Institute, Freiburg, Germany2- Rheinische Friedrichs-Wilhelms-Universit¨at, Physics Institute, Bonn, Germany3- CEA Saclay, DAPNIA/SPP, Gif sur Yvette, FranceThis contribution investigates a prototype of a TPC readout with a highly pixelatedCMOS ASIC, which is an option for charged particles tracking of the ILC. A triple GEMstack was joined with a TimePix and MediPix2 chip (pixel size of 55 × µm ) and itsreadout properties were investigated with 5 GeV electrons. The spatial resolution ofthe cluster center reconstruction was determined as a function of drift distance usingdifferent cluster alhoritms and compared with Monte Carlo predictions. The recent development of Micro-Pattern-Gas-Detectors (MPGD) allows an extended fieldof application for detectors with gas multiplication. For Time-Projection-Chambers (TPC)the readout with Gas-Electron-Multipliers (GEM) [1] has been demonstrated in conjunctionwith an unconventional readout option with high pixelation using the MediPix2 chip [2]. TheTimePix [3] and MediPix2 is investigated the first time with 5 GeV electrons from DESY II.They have the virtue of negligible multiple scattering. An external tracking with Si-telescopeprovides information on the track position which is used for resolution determination andfor the drift velocity measurements using a TimePix chip.
A cross sectional view of the triple GEM plus Medipix2 and Timepix detector with anelectron beam crossing the drift region is shown in Fig.1 (left part). The drift volume of6 mm thickness and 10 ×
10 cm in size serves for charged track detection. The drift fieldis about 1.1 kV/cm. Three CERN-produced GEMs of the same area as the drift volumewith 70 µ m holes of 140 µ m pitch are arranged in a stack above the readout plane with2 mm (2-2-1) or 1 mm distances (1-1-1) for the transfer gaps, the latter provides a collapsedsetup in order to study possible effects on the spatial resolution. With the triple GEMstack gas amplification of ≈ can be achieved with Ar/CO and He/CO - mixtures forthe high resolution detection of minimum ionizing particles. The MediPix2 and TimePixchip is positioned at a distance of 1 mm from the last GEM exposed to a field of E I = 4.0kV/cm. The advantage of a GEM setup is the robust operation, since the fields of the gasmultiplication region ( ≈
70 kV/cm) are well shielded insige GEM holes, see [2].
The MediPix2 and Timepix chip has a surface of 14 ×
14 mm and a pixel size of 55 × µm .It is positioned close to the border of the GEM stack, see Fig. 1 (right part). The remaining LCWS/ILC 2007 igure 1: Setup of drift volume: stack and MediPix/TimePix (left) and test beam setupwhere inserts show size of MediPix2/TimePix and the size of a pixel (right)surface of the GEMs is covered with 24 anode pads of 2 × size for monitoring purposes.The gas tight box contains the GEMs, the resistor chain, the TimePix and MediPix2 chipand the readout electronics of the pads.The readout of the MediPix2 and TimePix is done with MUROS2 using the fast shutteroption. The thresholds for the pixels used were 990 e − and 830 e − for the MediPix2 andthe TimePix, respectively. The electron beam is defined by trigger scintillating counters of1 × in size and a Si-telescope with 3 planes of strips (two planes allow measurementof the x-coordinate in front and behind the GEM plus Medipix2 and Timepix, and one planeis used for the y-coordinate measurement in front of the detector). The effective readoutpitch of Si-telescope is 50 µ m. The TimePix has a clock, which is distributed throughout the entire chip. A register oneach pixel counts the number of clock cycles in a way depending on the chosen mode ofoperation. For each pixel this mode can be set individually. In TIME-mode the cycles arecounted from the point when the signal crosses the threshold till a common stop by thegate signal (= ”Fast shutter”). The other mode ”Time-Over-Threshold” records the clockcycles as long as the pulse is above the threshold. The maximum number of counts in thismeasurement is limited by the chosen gate width of 12 . µ m, which is about 600 counts ata given clock frequency of 48 MHz.In a special configuration “Mixed Mode” every other pixel switches the TIME and TOTmode in checker board fashion. This results in 1/2 of all pixels are of TIME-type and another1/2 of the TOT-type. This allows a proximity information of both TIME and TOT. LCWS/ILC 2007
Cluster reconstruction and point resolution
Typically, 8 - 9 non-separated clusters per track were observed, see Fig. 2 (left part).Images with double tracks were not considered in the current analysis. The following threeclustering methods were used, where last two can separate the overlapping clusters using thecharge deposition information (TOT or MIXED modes) resulting in everage 10 - 11 clustersper event: • “Contiguous areas” method leaves overlapping clusters not separated. That method isapplicable for any TimePix mode. The result is a lower number of clusters dependingon the effective threshold settings and therefore not covered here • “Saddle Point” method separates the contiguous clusters making a line for secondarymaximum in a projection transverse to track. The line divides the merged clusters atthe saddle point • “Island” method joins the adjacent pixels with a nonzero TOT values into clusters inthe way that a pixel is joined into the cluster of its neighbor with a highest TOT value.The procedure is repeated for each pixel to produce a unique assignment of pixels toclusters. It results in a separation in two spacial coordinates available compared withthe above case Drift distance y [mm]0 1 2 3 4 5 6 m ] m [ m ea n s / ndf c s – clel /n Dt 8.806 – c s – clel /n Dt 8.806 – > = 12 CLUSTERS Figure 2: Image of 5 GeV electron track (left part) and σ and D t /n elcl for Ar/CO (rightpart)The evaluation follows basically [2]: A cluster is rejected if it contains less than 9 hits.The noise is removed beyond the cluster region. Two different methods of the straight linefits to the centers of the clusters result in an inbiased estimation of the standard deviation.Residuals of the fit to all cluster centroids (N) and a fit to (N-1) centroids resulting in aresidual of the exempt cluster which is permutated over all clusters. These residuals enterinto two Gaussian fits which produce σ unbiased = √ σ N × σ N − . LCWS/ILC 2007 he left part of Fig. 2 shows a measured dependance of the resolutions versus the driftspace for Ar/CO . In the drift region to separate the lateral diffusion within the drift spacefrom the intrinsic GEM plus TimePix and MediPix2 resolution the y-coordinate informationof the external telescope is used. The following parametrisation is applied for the fit: σ = σ + D t × yn elcl . The σ equal to 24.9 ± . The results of spatial resolution σ and D t /n elcl corresponding to different clustering meth-ods for Ar/CO and He/CO are summarised in Tab. 2. The comparison with HEEDsimulations [4] is also given in the table.DATA SimulationsGas σ D t /n elcl σ D t /n elcl “Island” Ar/CO ± ± ± ± 14 – –“Saddle Ar/CO ± ± 36 15.2 ± ± ± ± 78 19.4 ± ± σ and D t /n elcl for the data and Monte Carlo simulations. Thecomparison with Monte Carlo simulation is present at the moment only for the “SaddlePoint” clustering method.The agreement between two clustering methods is found to be good for both σ and D t /n elcl values. It is found that simulations considerably overestimate the slope D t /n elcl comparing to the measurements for both gases.The slope depends on the size of merged clusters because a higher number of electrons perdetected cluster than for an ideally resolved cluster is present. This leads to a reduction ofthe effective transverse diffusion through the active volume and reduces the slope parameter.Based on this observation it is possible that the discrepancy of the Monte Carlo with respectto the slope is present, since the average number of detected clusters lower by a factor of 2than expected from simulations [4]. We would like to thank the EUDET project for its financial support and the MediPixcollaboration for supporting us work with readout software and hardware, especially withrespect to the TimePix chip used for the very first time in an experiment. We would alsolike to thank Michael Campbell, Erik Heijne, Xavier Llopart and Fabio Sauli for stimulatingdiscussions and a lot of valuable advices. References [1] Nucl.Instr.Meth.A386(1997)531[2] A. Bamberger et al.,“Readout of GEM Detectors Using the MediPix2 Chip”, NIMA573, 361-370(2007)[3] X. Llopart, see this conference[4] M. Hauschild, http://hausch.cern.ch/hausch/MediPix.html[1] Nucl.Instr.Meth.A386(1997)531[2] A. Bamberger et al.,“Readout of GEM Detectors Using the MediPix2 Chip”, NIMA573, 361-370(2007)[3] X. Llopart, see this conference[4] M. Hauschild, http://hausch.cern.ch/hausch/MediPix.html