Test-beam characterisation of the CLICTD technology demonstrator - a small collection electrode High-Resistivity CMOS pixel sensor with simultaneous time and energy measurement
R. Ballabriga, E. Buschmann, M. Campbell, D. Dannheim, K. Dort, N. Egidos, L. Huth, I. Kremastiotis, J. Kröger, L. Linssen, X. Llopart, M. Munker, A. Nürnberg, W. Snoeys, S. Spannagel, T. Vanat, M. Vicente, M. Williams
TTest-beam characterisation of the CLICTD technology demonstrator - a smallcollection electrode High-Resistivity CMOS pixel sensor with simultaneous time andenergy measurement
R. Ballabriga a , E. Buschmann a , M. Campbell a , D. Dannheim a , K. Dort a,1, ∗ , N. Egidos a,2 , L. Huth b , I. Kremastiotis a ,J. Kr¨oger a,3 , L. Linssen a , X. Llopart a , M. Munker a , A. N¨urnberg a,4 , W. Snoeys a , S. Spannagel b , T. Vanat a,5 ,M. Vicente a , M. Williams a,6 a CERN, Geneva, Switzerland b DESY, Hamburg, Germany
Abstract
The CLIC Tracker Detector (CLICTD) is a monolithic pixel sensor. It is fabricated in a 180 nm CMOS imagingprocess, modified with an additional deep low-dose n-type implant to obtain full lateral depletion. The sensor featuresa small collection diode, which is essential for achieving a low input capacitance. The CLICTD sensor was designedas a technology demonstrator in the context of the tracking detector studies for the Compact Linear Collider (CLIC).Its design characteristics are of broad interest beyond CLIC, for HL-LHC tracking detector upgrades. It is produced intwo different pixel flavours: one with a continuous deep n-type implant, and one with a segmented n-type implant toensure fast charge collection. The pixel matrix consists of 16 ×
128 detection channels measuring 300 µ m × µ m. Eachdetection channel is segmented into eight sub-pixels to reduce the amount of digital circuity while maintaining a smallcollection electrode pitch. This paper presents the characterisation results of the CLICTD sensor in a particle beam.The different pixel flavours are compared in detail by using the simultaneous time-over-threshold and time-of-arrivalmeasurement functionalities. Most notably, a time resolution down to (5 . ± .
1) ns and a spatial resolution down to(4 . ± . µ m are measured. The hit detection efficiency is found to be well above 99 . Keywords:
Silicon Detectors, Monolithic pixel sensors with a small collection diode, high-resistivity CMOS, PixelSensors
1. Introduction
The CLIC tracker detector (CLICTD) is a monolithichigh-resistivity (HR) CMOS sensor targeting the require-ments of the tracking detector for a future Higgs factorysuch as the Compact Linear Collider (CLIC) [1]. CLIC is aconcept for a linear electron position collider with centre-of-mass energies between 380 GeV and 3 TeV. The trackingdetector of CLIC is subject to stringent requirements [2].A single point resolution of < µ m in one spatial direc-tion needs to be combined with a hit time resolution of (cid:46) > . X per detector layer.Monolithic CMOS silicon sensors are attractive candi-dates for the large-area CLIC tracking detector due to ∗ Corresponding author
Email address: [email protected] (K. Dort) Also at University of Giessen, Germany Also at University of Barcelona, Spain Also at University of Heidelberg, Germany Now at KIT, Karlsruhe, Germany Now at DESY, Hamburg, Germany Also at University of Glasgow, U.K. their small material budget and relative ease of large-scaleproduction. It is advantageous to minimise the input ca-pacitance of these sensors to profit from a low noise level,a low detection threshold, a high signal, and a low powerconsumption [3]. A small capacitance can be achievedby minimising the size of the collection diode [4]. Vari-ous design features of the 180 nm CMOS imaging processhave been successfully tested within the framework of theALPIDE sensor development for the ALICE Inner Track-ing System upgrade [5, 6]. However, fast charge collectionis hampered by the limited depletion and the low elec-tric field in the small collection electrode design. To mit-igate this, modifications to the sensor design have beenintroduced in order to achieve full lateral depletion of theepitaxial layer [7] and to enhance the lateral field for ad-ditional acceleration of the charge collection [8]. The op-timised sensor designs have been shown to improve radia-tion hardness in the Mini-MALTA sensor developed in thecontext of the ATLAS upgrade Phase-II [9]. The CLICTDsensor is fabricated in two different pixel flavours affectingthe charge collection. In this document, the measurementsin charged particle beams are presented and the perfor-
Preprint submitted to Nucl. Instr. Meth. A February 9, 2021 a r X i v : . [ phy s i c s . i n s - d e t ] F e b ance of the two pixel flavours is compared.
2. The CLICTD monolithic sensor
The CLICTD sensor features a matrix of 16 ×
128 detec-tion channels with a size of 300 µ m × µ m. In the 300 µ mcolumn dimension, the channels are segmented into eightsub-pixels, each with its own collection diode and analoguefront-end. This segmentation scheme reduces the digitalfootprint while maintaining fast charge collection and asmall detector capacitance. In the following, the sensorand readout design are outlined. A detailed description ispresented elsewhere [10]. The CLICTD sensor is fabricated in a 180 nm CMOSimaging process [7]. The sensor layout is shown schemat-ically in Fig. 1. The sensor features a small n-type col-lection electrode placed on a high resistivity p-type epi-taxial layer with a thickness of 30 µ m. The epitaxial layeris grown on top of a p-type bulk substrate, resulting ina total thickness of 300 µ m for the entire sensor. Sam-ples with a total thickness of 100 µ m and 50 µ m have beenproduced by backside grinding. The analogue and digitalon-channel electronics are located on deep p-wells, whichshield the CMOS transistors from the electric field in thesensor. Moreover, the sensor is shielded from the circuitry,which could act as a noise source. The shielding is also nec-essary to avoid charge collection by electrodes other thanthe collection electrode.From 3D TCAD simulations, it is expected that the de-pletion zone extends approximately 23 µ m in depth. Afull lateral depletion in the epitaxial layer can be achievedby including a low-dose n-type implant underneath the p-wells. In a second pixel flavour, a segmentation of theimplant is introduced in order to speed up charge col-lection [8], as illustrated in Fig. 1b. The segmented n-type implant generates a lateral electric field, resulting ina faster propagation of charge carriers to the collectiondiodes.The CLICTD sensor is fabricated in both pixel variants:the first flavour with a continuous n-implant and the sec-ond flavour with a segmented n-implant. For the secondflavour, the segmentation is only applied along the columndimension of the matrix, as indicated schematically on thebottom of Fig. 1b. In the row dimension, which wouldbe perpendicular to the magnetic field in the CLIC detec-tor, the implant is not segmented since charge sharing isdesired to improve the spatial resolution.Reverse biases are applied to the substrate and to the p-wells. The p-well reverse bias is limited to -6 V to avoid abreakdown of the on-channel NMOS transistors [11]. Thedifference between substrate and p-well bias is limited aswell to avoid punch-through between them [12]. The analogue front-end in each sub-pixel features a volt-age amplifier connected to a discriminator, where the volt-age pulse is compared to an adjustable detection threshold.Variations of the effective threshold in each sub-pixel canbe corrected with a a 3-bit threshold-tuning DAC. The dis-criminator outputs of the eight sub-pixels in each detectorchannel are combined in the on-channel digital front-endwith an OR gate. Each detector channel records the bi-nary hit pattern of its eight sub-pixels.For timing measurements, the time-of-arrival (ToA) isrecorded with a 100 MHz clock. A global shutter signalsets the time reference for the ToA, which is defined as thenumber of clock cycles from crossing the threshold untilthe shutter is closed. The granularity of the ToA setsa lower limit of 10 ns / √
12 = 2 .
3. Samples and readout
The results described in the following have been ob-tained with the Caribou versatile data acquisition sys-tem [13]. All the reported measurements were obtainedfrom a single 300 µ m thick assembly for each pixel flavour,except for those in Section 6.3 where samples of eachflavour with thicknesses of 50 µ m and 100 µ m were usedto compare the efficiency. The sensor bias voltage at thep-wells/substrate is fixed to -6 V/-6 V. The effect of a lowerabsolute bias voltage is presented elsewhere [14].
4. Laboratory characterisation
Initial laboratory measurements for the CLICTD sensorcan be found elsewhere [10]. In this document, measure-ments crucial for the calibration and reconstruction of thetest-beam data are summarised and partially extended.The results are summarised in Table 1.
The detection threshold is calibrated using the X-rayfluorescence spectra of up to four different materials. Alinear relationship between the threshold DAC values andenergy is found.2 ollection electroden-implant37.5 μ m μ m μ mcollection electrodelow dose n-type implantp- epitaxial layerp+ substrate n-well p-welldeep p-well V p-well V sub column r o w (a) Continuous n-implant column r o w collection electrodelow dose n-type implantp- epitaxial layerp+ substrate n-well p-welldeep p-well V p-well V sub collection electroden-implant37.5 μ m μ m μ m (b) Segmented n-implant Figure 1: Sensor designs
Parameter Continuous n-implant Segmented n-implant
Conversion factor [e − /DAC] 8 . ± .
02 (stat.) +0 . − . (syst.) 8 . ± .
02 (stat.) +0 . − . (syst.)Threshold dispersion [e − ] 24 ± ± − ] 11 ± ± − ] 170 +4 − (syst.) 178 +4 − (syst.)Front-end time resolution [ns] 5 . ± . ± . . ± . ± . Table 1: Results of the laboratory characterisation for the pixel flavour with continuous n-implant and segmented n-implant. The statisticaluncertainty is marked by (stat.), the systematic uncertainty by (syst.).
Two sources of systematic uncertainties are studied:First, the analysis is repeated with different numbers ofmaterials and varied fit ranges. A maximum deviation of ± . − /DAC is found. Second, a maximum charge collec-tion loss of 30 e − due to sub-threshold effects and chargecarrier recombination is assumed, which yields a one-sideduncertainty of − .
07 e − /DAC. The statistical uncertaintyamounts to ± .
02 e − /DAC.The resulting conversion factors are8 . ± .
02 (stat.) +0 . − . (syst.) e − /DAC , and 8 . ± .
02 (stat.) +0 . − . (syst.) e − /DACfor the pixel flavour with continuous and segmented n-implant, respectively. The sensor capacitance is expectedto be unaffected by the pixel flavours. Therefore, the con-version factors are the same within the uncertainties. The sub-pixel threshold variation is reduced using the3-bit threshold-tuning DAC. The systematic uncertainty isestimated by propagating the uncertainty on the thresh-old conversion factor to the threshold dispersion, whichyields +0 . − . e − for both pixel flavours. Additional sourcesof systematic uncertainties are studied by repeating theequalisation with different environmental conditions andreadout schemes. In total, a maximum deviation of ± − is found. The statistical uncertainty is in the sub-electronrange and therefore negligible. After equalisation, the RMS of the threshold dispersionamounts to (24 ±
3) e − for both pixel flavours. The thresh-old equalisation is repeated with additional CLICTD sam-ples and the results are found to be within the stated un-certainties. The single sub-pixel noise is estimated by varying thedetection threshold around the baseline and registering thenoise hits for every sub-pixel and every threshold value.The RMS of the noise hit distribution as a function of thethreshold is extracted on sub-pixel level.The systematic uncertainty on the average sub-pixelRMS is estimated with the same techniques used for thecalculation of the threshold dispersion uncertainty and avalue of ± − is found. The statistical uncertainty is be-low one electron.The mean of the single sub-pixel noise RMS is (11 ±
1) e − for both pixel flavours. The operation threshold is defined as the lowestmean threshold at which a noise-free operation ( < × − hits/sec) of the sensor is possible. Sub-pixels exhibit-ing non-Gaussian noise are excluded from this definitionby masking them online. The number of masked pixels isless than one per mille of the entire matrix.The systematic uncertainty is calculated by propagatingthe uncertainty in the threshold conversion factor to the3peration threshold. The statistical uncertainty with avalue below one electron is negligible.With this definition, the operation threshold in the lab-oratory is found to be 135 +4 − e − . In the test-beam, the op-eration threshold has been increased to 170 +4 − e − for thepixel flavour with continuous n-implant and 178 +4 − e − forthe flavour with segmented n-implant in order to gain mar-gin for stable operation. The ToT measurement is calibrated with test-pulses in-jected into the analogue front-end amplifier of individualsub-pixels. The relationship between the known amplitudeof the test-pulses and the measured ToT is parametrisedwith a non-linear function depending on four parame-ters [15].The ToT calibration has been shown to have limitedprecision for the following reasons: • The capacitance of the test-pulse injector is insuffi-cient to trigger the highest possible ToT in all sub-pixels. Some of the sub-pixels are therefore not cal-ibrated correctly in the high-ToT range. This con-straint affects single pixel signals with (cid:38) . − . • The on-channel NMOS transistors are affected by thenegative bias voltage applied to the p-wells. As aconsequence, the operation margin of the circuits isreduced leading to a strong non-uniformity and non-linearity in the ToT response [10]. • In cases where several sub-pixels are hit in the samereadout frame, the combined ToT response is assignedto all hit sub-pixels. This can lead to a wrong conver-sion depending on the observed hit multiplicity.The ToT calibration is applied to evaluate the signalin Section 6. For all other analyses, the ToT values arenot converted to electrons to prevent the introduction ofsystematic errors due to the above mentioned limitations.The insight acquired from the characterisation of theCLICTD front-end will have important implications forthe front-end development of the next generation of sen-sors.
The time resolution of the front-end is estimated withtest-pulses injected in the analogue and digital front-endof each sub-pixel. The test-pulse injection is triggered ap-proximately 1 µ s before the end of a CLICTD frame andoccurs asynchronously to the ToA clock in order to en-sure a random phase between the clock and the injectiontime. For each sub-pixel, the time residuals between theToA values recorded for the analogue and digital test-pulseare evaluated. To estimate a lower limit on the front-endtime resolution, only pulse heights that induce a ToT of 11are considered, which is equivalent to the most probableenergy deposition of a minimum ionising particle.
3x Mimosa26 (upstream)
3x Mimosa26 (downstream)
CLICTDTimepix3 Particle beam
Scintillators + PMTs
67 cm
Figure 2: Telescope setup at the DESY II test-beam facility. TheDUT is placed between three MIMOSA-26 planes in the upstreamand downstream arm, respectively. An additional Timepix3 plane isplaced behind the last MIMOSA-26 plane in the downstream arm.
The systematic uncertainty introduced by the uncer-tainty on the threshold conversion factor is negligible. Theanalysis is repeated by lowering and raising the consideredsignal range by one ToT bin and a deviation of ± . ± . . ± . ± . . The front-end time resolution is therefore larger than es-timated from the 10 ns ToA binning.
5. Test-beam and analysis setup
In the following, the test-beam setup and the offline re-construction are outlined.
Data was recorded at the DESY II Test Beam Facil-ity [16] using a 5.4 GeV electron beam.For the measurements presented in this document, thedevice under test (DUT) is placed in a EUDET telescopewith six MIMOSA-26 monolithic active pixel sensors [17]and a Timepix3 time-reference plane [18], as illustrated inFig. 2.The AIDA Trigger Logic Unit (TLU) [19] provides atrigger signal when recording a coincidence between twoscintillators placed in front of the first telescope plane.The telescope and DUT are controlled and read out usingthe EUDAQ2 data acquisition framework [20].The CLICTD sensor is operated with shutters closing200 ns after a trigger signal is received from the TLU. Theshutter is opened again after the readout is completed.A compression algorithms ensures that only data com-ing from those detector channels that registered a hit are4hifted out. The resulting readout frequency is approxi-mately 500 Hz.
The offline analysis of the test-beam data is per-formed with the test-beam reconstruction framework Cor-ryvreckan [21, 22].The offline event building is based on the readout framesprovided by the CLICTD sensor. For the MIMOSA planes,only pixel hits that are associated to a TLU trigger sig-nal with a timestamp within the CLICTD frame are con-sidered. Likewise, Timepix3 pixel hits are required tohave a timestamp that lies within a CLICTD frame. Thisevent building scheme ensures that the telescope data wasrecorded when the DUT was active.For each telescope plane, adjacent pixels hits are com-bined into clusters . The cluster position is calculated by acentre-of-gravity algorithm.Track candidates are selected by requiring a cluster oneach of the seven telescope planes. The tracks are fit-ted with the General Broken Lines (GBL) formalism [23],which takes into account multiple scattering in the ma-terial traversed by the beam particles. In the alignmentprocedure of the telescope planes, the track χ is min-imised. For the subsequent analysis, only tracks with a χ per degree of freedom less than or equal to five areconsidered.For the measurements in this document, the resolutionof the track impact position on the DUT is determined tobe between 2 . µ m and 2 . µ m, depending on the telescopeplane spacing [17, 24]. The timestamp of each track isgiven by the ToA measurement in the Timepix3 plane.The track time resolution is determined with the methodsdescribed in [25] and a value of 1.1 ns is found.A reconstructed track is associated to a CLICTD clusterif the spatial distance between the track intercept positionon the DUT and the nearest pixel in a cluster is less than1.5 pixel pitches in both directions and the track time iswithin the same CLICTD frame as the cluster.The hit detection efficiency of the DUT is defined as thenumber of tracks associated to a CLICTD cluster over thetotal number of tracks. The tracks used for the hit de-tection efficiency calculation are required to pass throughthe acceptance region of the DUT. The acceptance regioncomprises the physical pixel matrix of the sensor excludingone column/row at the matrix edge. It has been verifiedthat the exclusion of additional columns/rows close to thematrix edge does not alter the results. If the track inter-cept position on the DUT lies within a masked pixel or itsdirect neighbours, the track is rejected. The efficiency un-certainty is calculated using a Clopper-Pearson confidenceinterval of one sigma [26].For the CLICTD sensor, the cluster position is cor-rected with the η -formalism in order to account for non-linear charge sharing [27]. The η -function is constructedby plotting the in-pixel position of the telescope tracks as m] m In-pixel row (cluster) [ - m ] m I n - p i x e l r o w ( t r a ck ) [ - Data function h CLICdp - Threshold = 178 eBias =-6V / -6V
Figure 3: In-pixel telescope track position as a function of recon-structed in-pixel cluster position for the pixel flavour with segmentedn-implant at a threshold of 178 e. The 5th order polynomial fittedto the distribution (shown as red line) is defined as η -function. a function of the reconstructed in-pixel cluster position,as shown in Fig. 3 for a sample with segmented n-implantat a threshold of 180 e. A 5th order polynomial fitted tothe distribution is used to correct the reconstructed clus-ter position. The η -correction is only applied to clustersthat have an extent of two pixels in row direction.
6. Performance in test-beam measurements
In this section, the characterisation of the CLICTD sen-sor in test-beam measurements is presented. First, theperformance is evaluated at fixed operating conditions,also referred to as nominal conditions . For the compar-ison of the two pixel flavours, the detection threshold isfixed to 178 e − and the bias voltage to -6 V at the p-wellsand the substrate. In a second step, the applied detectionthreshold is varied for both flavours in order to study theimpact on the performance parameters. The results aresummarised in Table 2. The pixel flavour with the segmentedn-implant was designed to speed up charge collection. Asa consequence, charge sharing between sub-pixels in a sin-gle channel is reduced. This is reflected in the cluster sizein the column and row direction in Fig. 4 and Fig. 5, re-spectively.The statistical uncertainty on the cluster size is of theorder of 1 × − . For the systematic uncertainty, the un-certainty in the threshold conversion factor is propagatedto the cluster size using the threshold scans in Fig. 10 andFig. 11. A systematic uncertainty of ± .
01 is found.5 arameter Continuous n-implant Segmented n-implant
Cluster size [pixels] 1 . ± .
01 (syst.) 1 . ± .
01 (syst.)Cluster column size [pixels] 1 . ± .
01 (syst.) 1 . ± .
01 (syst.)Cluster row size [pixels] 1 . ± .
01 (syst.) 1 . ± .
01 (syst.)Max. threshold with efficiency > .
7% [e − ] 387 ±
12 (stat.) +9 − (syst.) 537 ±
20 (stat.) +12 − (syst.)Efficient operation window [e − ] 207 ±
12 (stat.) +5 − (syst.) 357 ±
20 (stat.) +8 − (syst.)Spatial resolution (column) σ ( s )col [ µ m] 6.7 ± ± σ ( s )row [ µ m] 4.6 ± ± σ ( t ) [ns] 6.5 ± ± Table 2: Results of test-beam characterisation for the two pixel flavours. The results are obtained at the operation threshold of 178 e − exceptfor the efficiency measurements, where a threshold window is listed. Cluster column size F r a c t i on Continuous n-implantSegmented n-implantCLICdp - Threshold = 178 eBias = -6V / -6V
Figure 4: Cluster size in column direction for the pixel flavour withand without segmentation of the n-implant at nominal conditions.The error bars reflecting the statistical uncertainty are not visible.
The cluster column size is 1 . ± .
01 and 1 . ± .
01 forthe pixel flavour with continuous n-implant and segmentedn-implant, respectively. The reduction by approximately5 % for the flavour with the segmented n-implant confirmsthat charge sharing is suppressed due to the electric fielddistribution that enforces fast charge collection. In therow direction, the size is 1 . ± .
01 and 1 . ± .
01 forcontinuous and segmented n-implant, respectively. Withinthe uncertainties, the cluster row size is unaffected by thepixel flavour, confirming that the n-implant segmentationonly affects the column direction. The results have beencross-checked with several CLICTD samples in order toexclude systematic uncertainties such as a rotational mis-alignment of the DUT in the test-beam setup. The clustersize values agree within the measurement uncertainties.Figures 6– 9 show the cluster size and cluster columnsize as a function of the in-pixel track intercept position.The reduction of the cluster size for the pixel flavour withsegmented n-implant is indeed only visible in the in-pixel
Cluster row size F r a c t i on Continuous n-implantSegmented n-implantCLICdp - Threshold = 178 eBias = -6V / -6V
Figure 5: Cluster size in row direction for the pixel flavour with andwithout segmentation of the n-implant at nominal conditions. Theerror bars reflecting the statistical uncertainty are not visible. column dimension, where the segmentation is introduced.
Threshold scan.
With increasing detection threshold, thecluster size decreases, as illustrated in Fig. 10 and Fig. 11for the cluster size in column and row direction, respec-tively. In column direction, the impact of reduced chargesharing for the flavour with the segmented n-implant isparticularly pronounced for low detection thresholds. Itdecreases for high thresholds due to inefficiencies formingat the pixel edges, which are especially sensitive to the dif-ferent deep n-implant structures. In the row direction, thesize for both pixel flavours is identical over the scannedthreshold range.
The seed signal is defined as the highest single pixel sig-nal in a cluster. In Fig. 12, the seed signal distributionfor both pixel flavours is depicted. The distributions arenot expected to follow Landau-Gauss functions owing to6 - m] m Column coordinate [ - - - m ] m R o w c oo r d i na t e [ C l u s t e r s i z e - Threshold = 178 e
CLICdp
Bias = -6V / -6V
Figure 6: In-pixel cluster size for the flavour with continuous n-implant at nominal conditions. - m] m Column coordinate [ - - - m ] m R o w c oo r d i na t e [ C l u s t e r s i z e - Threshold = 178 e
CLICdp
Bias = -6V / -6V
Figure 7: In-pixel cluster size for the flavour with segmented n-implantat nominal conditions. - m] m Column coordinate [ - - - m ] m R o w c oo r d i na t e [ C l u s t e r c o l u m n s i z e - Threshold = 178 e
CLICdp
Bias = -6V / -6V
Figure 8: In-pixel cluster column size for the flavour with continuousn-implant at nominal conditions. - m] m Column coordinate [ - - - m ] m R o w c oo r d i na t e [ C l u s t e r c o l u m n s i z e - Threshold = 178 e
CLICdp
Bias = -6V / -6V
Figure 9: In-pixel cluster column size for the flavour with segmentedn-implant at nominal conditions.
500 1000 1500 2000 2500
Threshold [e] M ean c l u s t e r c o l u m n s i z e Continuous n-implantSegmented n-implantCLICdp
Bias = -6V / -6V
Figure 10: Mean cluster size in column direction as a function ofdetection threshold. The hatched band represents the statistical andsystematic uncertainties.
Threshold [e] M ean c l u s t e r r o w s i z e Continuous n-implantSegmented n-implantCLICdp
Bias = -6V / -6V
Figure 11: Mean cluster size in row direction as a function of detectionthreshold. The hatched band represents the statistical and systematicuncertainties.
Cluster seed charge [e] F r a c t i on Continuous n-implantSegmented n-implantCLICdp - Threshold = 178 eBias = -6V / -6V
Figure 12: Cluster seed signal distribution for both pixel flavours atnominal conditions. The error bars reflecting the statistical uncer-tainty are not visible. charge sharing and the limitations of the charge measure-ment and calibration. The lower seed signal for the pixelflavour with continuous n-implant is a consequence of thehigher charge sharing.The cluster charge is not evaluated quantitatively in thisdocument owing to the limited precision in the conversionof ToT values to physical units as discussed in Section 4.
The hit detection efficiency as a function of the detec-tion threshold is depicted in Fig 13. A close-up of the high efficiency range is shown in Fig. 14. The maximumthreshold with an efficiency of > .
7% and the range be-tween this value and the operation threshold (defined as efficient operation window ) are listed in Table 2. For thetwo thinned assemblies, the operation threshold is 180 e.The statistical uncertainties arise from the uncertaintieson the efficiency values. The systematic uncertainty isgiven by the uncertainty on the threshold value.The efficient operation window evaluates to207 ±
12 (stat.) +5 − (syst.) e , for the continuous n-implant and357 ±
20 (stat.) +8 − (syst.) efor the flavour with the segmented n-implant. The efficientoperation window for the second flavour is more than 1.5times larger owing to the reduced charge sharing, whichgives rise to a higher seed signal.For high detection thresholds, inefficient regions startto form at the pixel edges, as illustrated in Fig. 15, wherethe in-pixel hit detection efficiency is shown at a thresholdof 1950 e − for the flavour with continuous n-implant. Thepixel corners are especially affected as a result of enhancedcharge sharing in these regions. The spatial residuals, defined as thedifference between the reconstructed cluster position andthe track intercept on the DUT, are shown in Fig. 16 forthe column direction. The RMS of the central 99.7% ofthe distribution amounts to 7 . µ m for the pixel flavourwith continuous n-implant and 8 . µ m for the one withsegmented n-implant.8
00 400 600 800 1000
Threshold [e] E ff i c i en cy Continuous n-implantSegmented n-implantCLICdp
Bias = -6V / -6V
Figure 13: Hit detection efficiency as a function of threshold for bothpixel flavours. The hatched band represents the statistical and sys-tematic uncertainties.
200 300 400 500 600
Threshold [e] E ff i c i en cy Continuous n-implantSegmented n-implantCLICdp
Bias = -6V / -6V
Figure 14: Hit detection efficiency as a function of threshold for lowthresholds. The hatched band represents the statistical and system-atic uncertainties. - m] m Column coordinate [ - - - m ] m R o w c oo r d i na t e [ E ff i c i en cy CLICdp - Threshold = 1950 eBias = -6V / -6V
Figure 15: In-pixel hit detection efficiency at a threshold of 1950 e − for the pixel flavour with continuous n-implant. The spatial telescope track resolution at the DUT isquadratically subtracted from the measured RMS to ob-tain the spatial resolution of the DUT.The statistical uncertainty on the spatial resolution isof the order of 1 × − µ m. To quantify the systematicuncertainties, the telescope single plane resolution is var-ied within its uncertainties given in [17], which yields anuncertainty of ± . µ m. The propagated threshold uncer-tainty is ± . µ m as well. The total systematic uncertaintyis given by the quadratic sum of the individual values.The spatial resolution in row direction evaluates to 4 . ± . µ m for both pixel flavours. Observing identical valuesis in agreement with the similar cluster row size presentedin Section 6.1. The value is well below the requirement of7 µ m for the CLIC tracking detector.In column direction, the spatial resolution for the pixelflavour with segmented n-implant is 7 . ± . µ m, whichis approximately 12% larger compared to the 6 . ± . µ mthat is measured for the pixel flavour with continuous n-implant.In both dimensions, the spatial resolution is superiorto the binary resolution that would be expected with-out charge sharing. The binary resolution is given bypitch / √
12 and evaluates to 8 . µ m in row direction and10 . µ m in column direction. Threshold scan.
In Fig. 17, the spatial resolution is shownin row direction as a function of the detection threshold.With increasing threshold, the spatial resolution degradesowing to the decrease in cluster size. The binary resolutionof 8 . µ m is never exceeded.For threshold values greater than 1000 e − , the efficientpixel area starts to shrink from the pixel edges leading toan improvement in the spatial resolution.9 - - m] m Residuals (col.) [ F r a c t i on Continuous n-implantSegmented in n-implantCLICdp m m = 7.2 continuous(s) RMS m m = 8.1 segmented(s) RMS - Threshold = 178 eBias = -6V / -6V
Figure 16: Residuals in column direction between track interceptposition and reconstructed cluster position on the CLICTD. Theerror bars reflecting the statistical uncertainty are not visible.
The spatial resolution in column direction as a functionof threshold is depicted in Fig. 18. It illustrates that thereduced charge sharing for the flavour with the segmentedn-implant causes a degrading spatial resolution regardlessof the detection threshold.
The time residuals are defined as thedifference between the track timestamp and the ToA ofthe DUT. In Fig. 19, the time residuals are depicted as afunction of the seed pixel ToT for the pixel flavour withcontinuous n-implant. A slower response is observed forlow signal heights ( time-walk ). The effect is particularlystrong in the pixel corners, where a lower seed signal isexpected, as can be seen in Fig. 21 for the continuous n-implant and in Fig. 22 for the segmented n-implant. Asa consequence, the time-walk is more pronounced for theflavour with continuous n-implant. This result is partic-ularly interesting for applications where precise timing isrequired before an offline time-walk correction can be ap-plied.For the CLICTD sensor, a time-walk correction is per-formed for each ToT bin separately by subtracting themean time difference between the track and the measuredToA. The time residuals after correction as a functionof the seed pixel ToT are shown in Fig. 20. The widthof the time residuals is larger for small seed signals dueto the stronger impact of amplitude jitter. The in-pixeltime residuals after time-walk correction are depicted inFigs. 23 and 24 for continuous n-implant and segmentedn-implant, respectively. After time-walk correction, thetiming is more homogenous across the pixel cell for bothpixel flavours. The remaining in-pixel pattern suggests that slow signals arise predominantly from incident posi-tions at the pixel corners, which is in agreement with 3DTCAD simulations indicating that the time resolution de-grades in the pixel edge regions [8].The one dimensional residual distributions are depictedin Fig. 25 for both pixel flavours. The time resolutionis calculated using the RMS of the central 99.7% of thedistribution, which amounts to 6.6 ns and 5.9 ns for thepixel flavour with continuous n-implant and segmented n-implant, respectively.The time resolution associated with the track timestamp(1 . .
01 ns. The sys-tematic uncertainty is estimated by repeating the time-walk correction for every sub-pixel position in a channelindividually. The spread of the sub-pixel specific time res-olution is ± . . ± .
1) ns, which is about10% better than the (6 . ± .
1) ns for the flavour withcontinuous n-implant. The values are within the require-ments for the CLIC tracking detector. The similar timingperformance for both pixel flavours, despite the acceler-ated charge collection for the flavour with segmented n-implant, can be attributed to front-end timing limitationsas explained in Section 4.
Threshold scan.
In Fig. 26, the time resolution is de-picted as a function of the detection threshold. For bothpixel flavours, the time resolution increases with increasingthreshold owing to the stronger contribution of amplitudejitter.
7. Conclusions
The CLICTD monolithic pixel sensor has been charac-terised in a charged particle beam for two different pixelflavours. In one flavour, a deep continuous low-dose n-implant ensures full lateral depletion of the epitaxial layer.In a second flavour, the n-implant is segmented to enhancethe lateral electric field for accelerated charge collectionand reduced charge sharing.The requirements for the CLIC tracker in terms of spa-tial and timing resolution as well as hit detection effi-ciency are fulfilled. In addition, the sub-pixel segmen-tation scheme of the front-end has shown to deliver therequired accuracy while minimizing the digital footprint.Previous results have shown that the estimated power-consumption [10] as well as the material budget[14] alsocomply with the requirements.The measurements confirm that charge sharing is af-fected by the pixel flavour, but only, as intended, in thecolumn dimension where the segmentation was introduced.The position resolution of 4 . µ m in the other directionremains unaffected. The reduced charge sharing for thepixel flavour with the segmented n-implant improves the10
00 1000 1500 2000 2500
Threshold [e] m ] m S pa t i a l r e s o l u t i on (r o w ) [ Continuous n-implantSegmented n-implantCLICdp
Bias = -6V / -6V
Figure 17: Spatial resolution in row direction as a function of thresh-old for both pixel flavours. The hatched band represents the statisticaland systematic uncertainties.
Threshold [e] m ] m S pa t i a l r e s o l u t i on ( c o l . ) [ Continuous n-implantSegmented n-implantCLICdp
Bias = -6V / -6V
Figure 18: Spatial resolution in column direction as a function ofthreshold for both pixel flavours. The hatched band represents thestatistical and systematic uncertainties.
Seed pixel ToT - - - ) [ n s ] h i t - t t r a ck ( t - - - - F r a c t i on CLICdp - Threshold = 178 e
Bias = -6V / -6V
Figure 19: Residual between track timestamp and reconstructedcluster timestamp as a function of seed pixel charge before time-walk correction for the pixel flavour with continuous n-implant. Theblack crosses denote the mean of each ToT bin.
Seed pixel ToT - - ) [ n s ] h i t - t t r a ck ( t - - - - F r a c t i on CLICdp - Threshold = 178 eBias = -6V / -6V
Figure 20: Residual between track timestamp and reconstructedcluster timestamp as a function of seed pixel charge after time-walkcorrection for the pixel flavour with continuous n-implant. The blackcrosses denote the mean of each ToT bin. - m] m Column coordinate [ - - - m ] m R o w c oo r d i na t e [ - - - ) [ n s ] h i t - t t r a ck ( t CLICdp - Threshold = 178 eBias = -6V / -6V
Figure 21: In-pixel time residuals for the pixel flavour with continuousn-implant before time-walk correction. - m] m Column coordinate [ - - - m ] m R o w c oo r d i na t e [ - - - ) [ n s ] h i t - t t r a ck ( t CLICdp - Threshold = 178 eBias = -6V / -6V
Figure 22: In-pixel time residuals for the pixel flavour with segmentedn-implant before time-walk correction. - m] m Column coordinate [ - - - m ] m R o w c oo r d i na t e [ - - - ) [ n s ] h i t - t t r a ck ( t CLICdp - Threshold = 178 eBias = -6V / -6V
Figure 23: In-pixel time residuals for the pixel flavour with continuousn-implant after time-walk correction. - m] m Column coordinate [ - - - m ] m R o w c oo r d i na t e [ - - - ) [ n s ] h i t - t t r a ck ( t CLICdp - Threshold = 178 eBias = -6V / -6V
Figure 24: In-pixel time residuals for the pixel flavour with segmentedn-implant after time-walk correction. - - - ) [ns] hit - t track (t F r a c t i on Continuous n-implantSegmented n-implantCLICdp = 6.6 ns continuous(t)
RMS = 5.9 ns segmented(t)
RMS - Threshold = 178 eBias = -6V / -6V
Figure 25: Time residuals between track timestamp and CLICTDtimestamp after time-walk correction. The error bars reflecting thestatistical uncertainty are not visible.
500 1000 1500 2000 2500
Threshold [e] T i m e r e s o l u t i on [ n s ] Continuous n-implantSegmented n-implantCLICdp
Bias = -6V / -6V
Figure 26: Time resolution as a function of the detection threshold.The hatched band represents the statistical and systematic uncer-tainties. measured time resolution by 10 % to 5 . CRediT authorship statementR. Ballabriga
Methodology, Supervision
E.Buschmann
Investigation, Software
M. Campbell
Methodology
D. Dannheim
Investigation, Methodology,Supervision, Writing - Review & Editing
K. Dort
Formal analysis, Investigation, Software, Visualization,Writing - Original Draft
N. Egidos
Resources
L.Huth
Investigation
I. Kremastiotis
Investigation,Resources
J. Kr¨oger
Investigation, Software
L. Linssen
Project administration, Funding acquisition
X. Llopart
Resources
M. Munker
Investigation, Methodology,Supervision, Writing - Review & Editing
A. N¨urnberg
Resources
W. Snoeys
Conceptualization, Resources
S. Spannagel
Investigation, Methodology, Software
T. Vanat
Resources, Software
M. Vicente
Investigation
M. Williams
Investigation, Software
Acknowledgments
This work has been sponsored by the Wolfgang GentnerProgramme of the German Federal Ministry of Educationand Research (grant no. 05E15CHA). The measurementsleading to these results have been performed at the TestBeam Facility at DESY Hamburg (Germany), a memberof the Helmholtz Association (HGF). This project has re-ceived funding from the European Union’s Horizon 2020research and innovation programme under grant agree-ment No 654168. This work was carried out in the frame-work of the CLICdp Collaboration.
References [1] The Compact Linear Collider (CLIC) - 2018 Summary Report2/2018. doi:10.23731/CYRM-2018-002 .[2] D. Dannheim, K. Kr¨uger, A. Levy, A. N¨urnberg, E. Sicking.(Eds.), Detector technologies for CLIC, CERN Yellow Reports:Monographs, 2019. doi:10.23731/CYRM-2019-001 .[3] W. Snoeys, Monolithic pixel detectors for high energy physics,Nucl. Instrum. Meth. A 731 (2013) 125–130. doi:10.1016/j.nima.2013.05.073 .[4] S. Senyukov, et al., Charged particle detection performances ofCMOS pixel sensors produced in a 0.18 µ m process with a highresistivity epitaxial layer, Nucl. Instrum. Meth. A 730 (2013)115–118. doi:10.1016/j.nima.2013.03.017 .[5] G. Aglieri Rinella, et al., The ALPIDE pixel sensor chip for theupgrade of the ALICE Inner Tracking System, Nucl. Instrum.Meth. A 845 (2017) 583–587. doi:10.1016/j.nima.2016.05.016 .[6] D. Dannheim, et al., Comparison of small collection electrodeCMOS pixel sensors with partial and full lateral depletion ofthe high-resistivity epitaxial layer, Nucl. Instrum. Meth. A 927(2019) 187–193. doi:10.1016/j.nima.2019.02.049 .
7] W. Snoeys, et al., A process modification for CMOS monolithicactive pixel sensors for enhanced depletion, timing performanceand radiation tolerance, Nucl. Instr. Meth. A 871 (2017) 90 –96. doi:https://doi.org/10.1016/j.nima.2017.07.046 .[8] M. Munker, et al., Simulations of CMOS pixel sensors with asmall collection electrode, improved for a faster charge collec-tion and increased radiation tolerance, JINST 14 (05) (2019)C05013–C05013. doi:10.1088/1748-0221/14/05/c05013 .[9] M. Dyndal, et al., Mini-MALTA: Radiation hard pixel designsfor small-electrode monolithic CMOS sensors for the High Lu-minosity LHC, JINST 15 (02) (2020) P02005. doi:10.1088/1748-0221/15/02/P02005 .[10] I. Kremastiotis, et al., Design and characterisation of theCLICTD pixelated monolithic sensor chip, IEEE Trans NuclSci PP (2020) 1–1. doi:10.1109/TNS.2020.3019887 .[11] J. W. van Hoorn, Study and Development of a novel SiliconPixel Detector for the Upgrade of the ALICE Inner TrackingSystem, PhD thesis (2015).URL https://cds.cern.ch/record/2119197 [12] M. Munker, et al., Simulations of CMOS pixel sensors with asmall collection electrode, improved for a faster charge collec-tion and increased radiation tolerance, JINST 14 (05) (2019)C05013. doi:10.1088/1748-0221/14/05/C05013 .[13] T. Vanat, Caribou – A versatile data acquisition system, PoSTWEPP2019 (2020) 100. doi:10.22323/1.370.0100 .[14] K. Dort, The CLICTD Monolithic CMOS Sensor (2020). arXiv:2011.09389 .[15] J. Jakubek, Precise energy calibration of pixel detector workingin time-over-threshold mode, Nucl. Instrum. Meth. A 633 (2011)S262–S266. doi:10.1016/j.nima.2010.06.183 .[16] R. Diener, et al., The DESY II Test Beam Facility, Nucl. Instr.Meth. A 922 (2019) 265–286. arXiv:1807.09328 , doi:10.1016/j.nima.2018.11.133 .[17] H. Jansen, et al., Performance of the EUDET-type beam tele-scopes, EPJ Tech. Instrum. 3 (1) (2016) 7. doi:10.1140/epjti/s40485-016-0033-2 .[18] T. Poikela, et al., Timepix3: a 65k channel hybrid pixel readoutchip with simultaneous ToA/ToT and sparse readout, Journal ofInstrumentation 9 (05) (2014) C05013–C05013. doi:10.1088/1748-0221/9/05/c05013 .[19] P. Baesso, D. Cussans, J. Goldstein, The AIDA-2020 TLU: aflexible trigger logic unit for test beam facilities, JINST 14 (09)(2019) P09019. doi:10.1088/1748-0221/14/09/P09019 .[20] Y. Liu, et al., EUDAQ2—A flexible data acquisition soft-ware framework for common test beams, JINST 14 (10) (2019)P10033. doi:10.1088/1748-0221/14/10/P10033 .[21] D. Dannheim, et al., Corryvreckan: A Modular 4D Track Re-construction and Analysis Software for Test Beam Data (2020) arXiv:2011.12730 .[22] J. Kr¨oger, S. Spannagel, M. Williams, Corryvreckan User Man-ual, latest version.URL http://cern.ch/go/db9Z [23] V. Blobel, A new fast track-fit algorithm based on broken lines,Nucl. Instrum. Meth. A 566 (2006) 14–17. doi:10.1016/j.nima.2006.05.156 .[24] S. Spannagel, H. Jansen, Gbl track resolution calculator v2.0,2016. doi:doi:10.5281/zenodo.48795 .URL https://github.com/simonspa/resolution-simulator [25] F. Pitters, et al., Time resolution studies of Timepix3 assemblieswith thin silicon pixel sensors, JINST 14 (05) (2019) P05022. doi:10.1088/1748-0221/14/05/P05022 .[26] C. J. Clopper, E. S. Pearson, The use of confidence of fiduciallimits illustrated in the case of the binomial, Biometrika 26 (4)(1934) 404–413. doi:10.1093/biomet/26.4.404 .[27] K. Akiba, et al., Charged Particle Tracking with the TimepixASIC, Nucl. Instrum. Meth. A 661 (2012) 31–49. doi:10.1016/j.nima.2011.09.021 ..