TOFHIR2: The readout ASIC of the CMS Barrel MIP Timing Detector

TOFHIR2: The readout ASIC of the CMS Barrel MIP Timing Detector E. Albuquerquee, M. Araújod, A. Benagliac, A. Bolettid, R. Bugalhoe, T. Coutinhoe, F. De Guioc,f, P. Facciolid, L. Ferramachoe, M. Firleja, T. Fiutowskia, R. Franciscoe, M. Gallinarod, A. Ghezzic,f, J. Hollard, M. Idzika, H. Legoinhad, N. Leonardod, C. Leonge, M.T. Lucchinic, M. Malbertic, G. Marozzod, G. Da Molind, J. Morona, T. Niknejadd,e, L. Oliveirab, N. Oliveirae, S. Palluottoc,f, M. Pisanod, N. Redaellic, C. Silvad, J. C. Silvad,e, R. Silvad,e, M. Silveirae, K. Swienteka, T. Tabarelli de Fatisc,f, S. Taverniere, J. Varelad,e, V. Varelae, S. Whiteg, J. Wulffd a AGH University of Krakow, Faculty of Physics and Applied Computer Science, Krakow, Poland b DEE, CTS-UNINOVA, FCT-UNL, Caparica, Portugal c INFN Sez. Milano Bicocca, Italy d LIP, Lisbon, Portugal e PETsys Electronics, Oeiras, Portugal f University of Milano-Bicocca, Milano, Italy g University of Virginia, Charlottesville, Virginia, USA E-mail: [email protected] ABSTRACT: The CMS detector will be upgraded for the HL-LHC to include a MIP Timing Detector (MTD). The MTD will consist of barrel and endcap timing layers, BTL and ETL respectively, providing precision timing of charged particles. The BTL sensors are based on LYSO:Ce scintillation crystals coupled to SiPMs with TOFHIR2 ASICs for the front-end readout. A resolution of 30-60 ps for MIP signals at a rate of 2.5 Mhit/s per channel is expected along the HL-LHC lifetime. We present an overview of the TOFHIR2 requirements and design, simulation results and measurements with TOFHIR2 ASICs. The measurements of TOFHIR2 associated to sensor modules were performed in different test setups using internal test pulses or blue and UV laser pulses emulating the signals expected in the experiment. The measurements show a time resolution of 24 ps initially during Beginning of Operation (BoO) and 58 ps at End of Operation (EoO) conditions, matching well the BTL requirements. We also showed that the time resolution is stable up to the highest expected MIP rate. Extensive radiation tests were performed, both with x-rays and heavy ions, showing that TOFHIR2 is not affected by the radiation environment during the experiment lifetime. KEYWORDS: TOFHIR2 ASIC, Dark Noise suppression, Time of Flight, MTD timing detector, CMS – 1 – Contents 1. Introduction 1 2. State of the art 3 3. Architecture and implementation 5 3.1 Analog design 6 3.2 Digital design 10 4. Simulation results 12 5. Experimental setup 14 6. Characterization results 16 6.1 Power consumption 16 6.2 Pulse shape and noise 16 6.3 TDC performance 18 6.4 Time resolution of MIP equivalent pulses 19 6.5 Energy of MIP equivalent pulses 21 6.6 Rate performance 22 7. Radiation tests 24 7.1 TID radiation tests 24 7.2 SEE radiation tests 28 8. Conclusions 29 1. Introduction The Upgrade of the CMS experiment [1] for the future High-Luminosity phase of the Large Hadron Collider at CERN (HL-LHC) includes the construction of a new MIP 1 Timing Detector (MTD) to measure the time of charged particles with high precision [2]. The MTD barrel section (Barrel Timing Layer - BTL) is a thin cylindrical detector with a total active surface of about 38 m2 based on LYSO:Ce scintillating crystals [3] coupled to silicon photomultipliers (SiPM) [4]. The individual cell consists of a crystal bar with approximately 3.0×3.75×54.7 mm3 with two ~3.0×3.0 mm2 SiPMs glued at each end. The SiPM signals are conditioned and digitized by the TOFHIR2 integrated circuit described in this paper. The full BTL detector has 331’776 electronics channels. LYSO is a scintillating crystal with high density (7.1g/cm3), large light yield (~40 thousand photons/MeV) and scintillation decay time of about 40 ns. A charged particle crossing the detector 1 MIP - Minimum Ionizing Particle – 2 – deposits in the crystal on average an energy of 4.2 MeV 2 . A fraction of the scintillation light is detected at the two ends of the crystal bar by SiPMs. The individual SiPM photo-sensor has 14’400 micro-cells of 25x25 m2 each. The Photon Detection Efficiency (PDE) at 420 nm is 20-55% and the gain is 2.9-9.7105 for an overvoltage 3 in the range 0.9-3.1V. The SiPMs are operated with overvoltage in this range optimizing PDE and gain versus Dark Count Rate (DCR). The pulse amplitude is proportional to the number of photons emitted by the particle interacting with the crystal. The current pulse from the SiPM is a unipolar signal with peaking time of 30 ns and tail of approximately 200 ns. During its operation, the detector is exposed to a large flux of particles (reaching an integrated fluence of 2×1014 neq/cm2 by end of operation), which creates defects in the SiPM silicon crystal. These defects are responsible for an increase of DCR from 0.5 MHz before irradiation up to 10-20 GHz at the end of the detector operation. To reduce the DarkCount noise the SiPMs are operated at a temperature of -45oC and will be periodically annealed at 60oC. The main requirements to the TOFHIR2 ASIC are: (1) to measure MIP timing with a precision of the order of 30 (60) ps at the beginning (end) of HL-LHC operation; (2) to mitigate the effect of large DCR implementing noise suppression circuitry; and (3) to provide a measurement of the signal amplitude with <5% precision for time-walk corrections. The chip must cope with a MIP input rate of 2.5 Mhit/s per channel and a rate of low energy hits (<1 MeV) up to 5 Mhits/s per channel. Additionally, it should be able to accept the CMS Level-1 trigger and synchronization signals, with output bandwidth of 640 Mb/s. The chip should have power consumption lower than 20 mW per channel. Furthermore, the TOFHIR2 must be resistant to the expected total ionization dose TID (2.9 Mrad) and to the integrated particle fluence (2×1014 neq/cm2) in the BTL. A summary of the specification parameters is given in Table I. Table I -TOFHIR2 specification parameters. Number of channels 32 Voltage supply 1.2 V Reference voltages Internal Radiation tolerance Yes DarkCount noise filter Yes Number of TAC and QAC 8 TDC bin (ps) 10 10-bit SAR ADC (MHz) 40 Max MIP rate/ch (MHz) 2.5 Max low E rate/ch (MHz) 5 Output bandwidth (Mb/s) 640 Power consumption/channel (mW) <20 Tolerance to TID (Mrad) >3 Tolerance to particle fluence (neq/cm2) >2×1014 TOFHIR2 is a new chip designed by PETsys Electronics [5] 4 in CMOS 130 nm technology and fabricated in the same foundry where the HEP community is focusing the production of the 2 Most Probable Value of the energy deposition averaged over the entire BTL. 3 Overvoltage is the difference between the SiPM operation voltage and the breakdown voltage. 4 The design of the 10-bit SAR ADC was provided by AGH, Krakow. – 3 – ASICs for the HL-LHC upgrades. TOFHIR2 follows a first implementation with limited functionality (TOFHIR1) adapted from an existing chip developed by PETsys Electronics for PET applications in technology CMOS 110 nm technology of a different foundry [6]. The CMOS 130 nm technology was adopted for the final chip given its good behavior under irradiation (small shifts of the transistors threshold voltage and low leakage current). The first full version (TOFHIR2A) was developed and tested in 2020 [7], followed by TOFHIR2X in 2021 which implemented a full current-mode, noise suppression analog front-end [8]. The last prototype (TOFHIR2B), implementing improved protection against Single Event Effects (SEE) induced by radiation, was developed before the final production version (TOFHIR2C) in 2023. In this paper, we present a summary of the design principles of the TOFHIR2 ASIC and the results of the measurements performed 5 . Results obtained in test beam experiments with final sensor modules and TOFHIR2C shall be reported in another paper. 2. State of the art The sensor technology based on LYSO crystals and SiPMs adopted for BTL has been used during the last twenty years in medical imaging Positron Emission Tomography (PET). In PET, radioactive tracer molecules are injected into the patient’s body. The decay of tracer molecules generates a positron-electron annihilation event resulting in two back-to-back γ-photons 6 with an energy of 511 keV that are absorbed by a cylindrical array of LYSO crystals coupled to SiPMs surrounding the patient [9][10]. The LYSO crystal converts the incident particle into optical photons whose number is proportional to the energy deposited in the crystal. A fraction of the photons is detected at the SiPM. Detected photons are then converted to electrons 7 that are multiplied in avalanche by the SiPM, generating a current pulse that can be discriminated and digitized to obtain a measurement of the time of the particle detection, referred to as the “timestamp”. The SiPMs used in PET applications (typically with micro-cells of 50x50 m2) have a gain ten times larger than the SiPMs optimized for use under extreme radiation in BTL, where the maximum allowed SiPM gain is determined by the allowed SiPM power consumption due to DCR. The timestamp of the detected interaction is determined from the first optical photons detected by the SiPM 8 . PET systems can measure the difference in arrival times of two 511 keV gamma rays with a resolution the order of few hundred picoseconds (FWHM) [11]. This allows an improvement of the signal-to-noise ratio (SNR) of the image [12][13]. The precise digitization of event timestamps and γ-photon energies is typically achieved by employing applicationspecific integrated circuits (ASICs) [14]-[30]. Table II shows a comparison of the main characteristics of several ASICs developed for PET applications. However, none of these circuits is tolerant to radiation. 5 The results presented in this paper have been obtained mostly with the final version TOFHIR2C. When noted, we present results obtained with the previous versions of the ASIC. 6 γ-photons or gamma-rays have energy above the x-ray band 7 Usually referred to as photoelectrons (p.e.) 8 Due to the long decay time of the LYSO scintillating light only the first optical photons are relevant for a time measurement with resolution three orders of magnitude smaller than the decay time. – 4 – Table II – Comparison of ASICs for PET applications (adapted from [19]) ASIC Timestamp digitization Power (mW/ch) CTR FWHM (ps) Crystal height (mm) Ref. FlexToT external 11 123 5 [27][28] HRFlexToT external 3.5 180 20 [29] STiC3 TDC on ASIC 25 240 15 [21] PETA4 TDC on ASIC <40 460 25 [26] Petiroc external 3.5 n.a. n.a. [22][22] Triroc TDC on ASIC 10 432.7 10 [23][24] TOFPET1 TDC on ASIC 8-11 290.7 15 [15][16] TOFPET2 TDC on ASIC 3.6-7.2 119 3 [6][17][18][19] Along the detection chain several effects introduce stochastic fluctuations that determine the detector time resolution, namely the photo-statistics, the SiPM DCR, the analog front-end (AFE) noise and the time digitization. The contribution from photo-statistics is related to the stochastic fluctuation in the time-of-arrival of photons detected at the SiPM and scales with 1√𝑁𝑝𝑒 ⁄, where 𝑁𝑝𝑒 is the total number of detected photoelectrons. The contribution to the time resolution of DarkCount noise scales with 𝜎𝐷𝐶𝑅 𝑁𝑝𝑒 ⁄, where 𝜎𝐷𝐶𝑅 is the standard deviation of the DarkCount noise. 𝜎𝐷𝐶𝑅 scales with √𝐷𝐶𝑅. The contribution of the front-end noise is given by 𝜎𝐴𝐹𝐸/𝑆𝑅, where 𝜎𝐴𝐹𝐸 is the standard deviation of the front-end noise and SR is the slew rate of the pulse rising edge at the level of the discrimination threshold. These contributions are not correlated and therefore the total time resolution 𝜎𝑡 is obtained by the quadratic sum of them: 𝜎𝑡=𝜎𝑡𝑝ℎ𝑜𝑡⊕ 𝜎𝑡𝐷𝐶𝑅 ⊕ 𝜎𝑡𝐴𝐹𝐸 The contribution arising from the time-to-digital conversion (including the clock jitter) is normally of the same order of the front-end noise or smaller. The jitter of the SiPM response is small when compared to the other sources of time uncertainty. In bench-top experiments using optimized discrete circuits associated to small LYSO crystals and SiPMs with improved single-photon time resolution (SPTR) it was possible to measure a Coincidence Time Resolution (CTR) of the two PET γ-photons of the order of 60 ps FWHM, which corresponds to a time resolution per channel of about 20 ps r.m.s (standard deviation) 9 largely dominated by photo-statistics [31]. In measurements performed with PET ASICs, typically the best CTR is of the order of 120 ps FWHM (see for example [6]), corresponding to single channel resolution of 36 ps r.m.s, which results from the contributions of photo-statistics, front-end noise and digitization, estimated at around 20 ps each, added quadratically. In BTL before irradiation, the target contributions to the time resolution are 20 ps from photostatistics, 10 ps from front-end noise and 10 ps from the digitization, for a total of 25 ps [2]. After accounting for the various factors contributing to the pulse amplitude (energy deposit, light collection efficiency, SiPM PDE and Gain), pulses in BTL before irradiation are about five times smaller than in PET. Therefore, achieving a noise contribution 𝜎𝐴𝐹𝐸/𝑆𝑅 two times smaller than 9 In PET applications, the time resolution is usually given by the full width at half maximum (FWHM) of the distribution of the time difference of the two γ-photons (CTR). The conversion factor between FWHM and root mean square (RMS) for a Gaussian distribution is 2.35. The single channel resolution and the CTR differ by a factor √2 . – 5 – in PET applications requires one order of magnitude improvement of the AFE by reducing the solid-state noise and increasing the bandwidth. During BTL operation the DarkCount noise increases by several orders of magnitude due to radiation, as already mentioned. The time resolution is progressively dominated by the term 𝜎𝑡𝐷𝐶𝑅. To achieve the goals of the BTL detector, it is necessary to implement noise suppression in the AFE to mitigate the negative impact of the DCR, by a factor larger than 2. The AFE of TOFHIR2 implements DarkCount noise suppression, for the first time in radiation detectors. On the other hand, in order to limit the increase of SiPM current due to DCR, staying within the power limits of the SiPM bias voltage supply system, the SiPM overvoltage is decreased from about 3V to 1V during the detector operation. The consequent drop of SiPM PDE and gain translates into a decrease of the number of photoelectrons by a factor larger than two and of pulse amplitude by a factor larger than five, which is responsible for the deterioration of 𝜎𝑡𝑝ℎ𝑜𝑡 and 𝜎𝑡𝐴𝐹𝐸. Accumulated exposure of silicon chips to ionizing radiation creates trapped charges at oxide/silicon interfaces causing quasi-permanent device shifts that may lead to chip malfunction [32]. The degradation depends on the device technology, process, and bias conditions. The effects due to TID in the technology used in TOFHIR2 have been characterized in [33][34]. The PMOS devices lose their current drive capability by up to 5% at 1 Mrad and 10% at 10 Mrad. NMOS devices have increased channel leakage of up to four orders of magnitude in the 1~10 Mrad range. NMOS and PMOS also experience threshold voltage shifts. These effects are higher in the 1~10 Mrad range. They depend strongly on the device size but also on the site where the chips are fabricated. Versions of TOFHIR2 have been fabricated in two different foundries of the same company, named in the following as A and B. Foundry A has considerably lower TID sensitivity than foundry B [33]. Single event upsets and transients, usually referred to as single event effects (SEE), caused by local ionization created by recoiling ions hit by heavy particles (mainly protons and neutrons) can cause a chip either to “lock-up”, to lose synchronization or to modify the configuration, thus requiring a time-consuming reset process. These effects are strongly dependent on the ion charge and therefore, induced ionization charge. The techniques used in TOFHIR2 to mitigate the radiation effects are described in the next section. 3. Architecture and implementation The functional block diagram of the TOFHIR2 ASIC is shown in Figure 1. The ASIC has 32 independent channels, a service block, and global control and trigger logic. Each channel integrates the analog front-end, time and charge digitizers and the channel digital control. The service block provides bias currents, several reference voltages and currents, and monitoring features. The global control and trigger logic handles the chip configuration, the digital data flow and transmission, and the trigger filtering (CMS Level 1 trigger). – 6 – Figure 1. Functional block diagram of the TOFHIR2 ASIC 3.1 Analog design A block diagram of one of the 32 TOFHIR2 channels is shown in Figure 2. Each ASIC channel contains one pre-amplifier, two post-amplifiers (T and E), three leading edge discriminators (T1, T2 and E), two Time-to-Amplitude Converters (TAC), one Charge-toAmplitude Converter (QAC), one 40 MHz 10-bit SAR ADC and local control logic. TACs and QACs are replicated eight-fold to handle Poisson fluctuations of the event 10 rate. The preamplifier provides a low impedance input to the sensor’s current signal. The input current is replicated into three branches for timing, energy discrimination and charge integration. Pulse filtering is included in the post-amplifiers to mitigate the deterioration of time resolution due to the large DCR induced by radiation and due to pile-up of LYSO pulse tails, as described below. For testing purposes, the chip implements an analog test pulse generator. 10 Unless otherwise specified, in this paper the term event refers to a signal at the channel input. – 7 – Figure 2. Block diagram of the TOFHIR2 channel. To obtain the required time resolution, since the SiPMs output is a current, the developed circuit is designed to process the input current without the usual conversion to voltage. Full current-mode increases the bandwidth, since the signal flows through low impedance nodes, making the circuits less sensitive to parameter variations and parasitic elements, thus reducing the mismatch between channels. The preamplifier is a current buffer to isolate the large SiPM load from the core of the AFE. The output current is mirrored to different timing and energy branches. Considering the effect of SiPM DCR, time-tagging relevant events directly at the preamplifier output with a time resolution better that 60 ps at EoO is impossible without proper signal processing. The challenge arises from the fact that signal events and DarkCount pulses have the same shape and frequency content, although they differ in amplitude. In addition, the high event rate hitting the preamplifier leads to baseline drifts due to severe piling-up of residual pulse tails. Since event flagging is done with comparators that have their threshold set to a fixed value referenced to the preamplifier baseline, baseline drifts will lead to wrong comparator firing time. Therefore, the baseline must be very stable for accurate timing. Figure 3. DLED block diagram. Motivated by the above considerations, we used the signal processing technique DLED (Differential Leading Edge Discriminator) [35], implemented here for the first time in a CMOS integrated circuit (Figure 3 and Figure 4). With appropriate delay settings, DLED preserves the – 8 – rising edge of the pulse and cancels the pulse tail (Figure 5). This approach reduces DarkCount noise and avoids baseline fluctuations and pile-up, since the input unipolar signal is converted to a bipolar signal. As shown in Figure 4, at the DLED output node Io is the current difference (I1I2) between the direct and delayed paths, hence the DLED signal processing in full current mode. Figure 4. Full current mode AFE. An ideal delay line cannot be implemented in a CMOS process; hence we have used a series of RC taps behaving as a low pass filter. In TOFHIR2 a delay line with eight taps can be programmed to have an equivalent delay in the interval between 200 and 1400 ps. The ability to vary the delay value also provides an additional degree of freedom to fine tune the pulse shape, since increasing the delay increases the pulse amplitude. The pulse shape also depends on the fabrication process mismatch between PMOS device MP1 and NMOS device MN1 (Figure 4). A pulse calibration feature has been implemented via an array of PMOS devices in parallel with MP1 forming a 5-bit DAC, which can be programmed to trim the pulse shape. Since the BTL detector will have more than 300k channels, it is important to have a quick and expeditious trimming process to set the pulse shape on demand. This can be done by simply measuring the width of the pulse lobe above the baseline using the T1 comparator and the two TDCs. At the DLED output, the baseline is stabilized with a baseline holding feedback loop. The baseline holder amplifier has a single low frequency pole in the mHz region, therefore the dynamics of the baseline holder does not modify the pulse shape. Figure 5. DLED input (left) and output (right) waveform. The example is for EoO conditions. Table III shows the results of a simulation in one possible scenario for the end of operation (EoO) conditions (DCR 55 GHz, signal yield of MIP pulse 6000 p.e., SiPM gain 1.5x105), Pulse width 200 ns Pulse width 25 ns 1.62mA 23µA – 15 – Figure 10. The BTL front-end board prototype with two TOFHIR2 and two ALDO2 ASICs. Figure 11. The BTL sensor module composed of 16 LYSO crystal bars glued at both ends to linear arrays of 16 SiPMs (left). The slit on the external wrapping allows shining the UV laser directly on the crystal bars to induce photo-excitation in the LYSO (right). Measurements were performed using laser pulses generated by the HPK PLP-10 Picosecond Light Pulser (wavelength 405 nm) or by the NKT Photonics PIL1-037-40 UV laser (wavelength 375 nm). The laser power is adjusted to obtain a given number of photoelectrons per pulse. The calibration of the pulse amplitude is obtained by measuring the SiPM output current. The tests of EoO performance are made with un-irradiated SiPMs. DarkCount noise is emulated by using a blue LED. Measurements with blue laser beam impinging directly on the SiPMs were also performed. Figure 12 shows a picture of a typical test setup inside a thermally controlled black box used in the measurements. Access to data via the ASIC I/O digital links is done using the PETsys Readout System [44]. Flexible cables connect the test board to the FPGA in the readout motherboard. The readout system allows the readout of two TOFHIR2 ASICs. The system provides the front-end board with all the necessary power, SiPM bias voltages, configuration, and readout. The TOFHIR2 supply voltage and the SiPM bias voltage are regulated by the ALDO2 ASIC [45]. – 16 – Figure 12. Example of the test setup used in the TOFHIR2 characterization measurements. In this case, we have laser (signal) and blue LED (DarkCount noise) light impinging directly on SiPMs mounted on the TOFHIR2A test board. A beam splitter is used to produce synchronous signals. 6. Characterization results 6.1 Power consumption The measured consumption of the 32-channel TOFHIR2C ASIC is 455 mA when the chip is powered, the clock is active, the 32 channels are firing at 2.5 MHz with the digital test pulse and 3% of the events are accepted by the L1 trigger and transmitted to the readout system. In this measurement there is no dynamic consumption on the analog frontend and there are no low energy hits activating the TDC/QDC and respective digital logic. From simulation, the corresponding consumption in BoO conditions is about 35 mA. The analog dynamic contribution is not expected to change significantly along the detector lifetime since the decrease in signal yield is compensated by the increase of DCR. Therefore, taking this into account, we estimate that the TOFHIR2C consumption in operating conditions is 490 mA. In TID radiation tests we have observed an increase of current consumption of the order of 5-10%, but we have shown that the natural annealing due to pauses of the accelerator bring the radiation effects to below 1% (see section 7.1). Therefore, we neglect the radiation effect in the power consumption. In conclusion, as the supply voltage is 1.2 V and the chip has 32 channels, the estimated power consumption of TOFHIR2C per channel (static + dynamic) is 18.4 mW. 6.2 Pulse shape and noise We measure the pulse shape tuning the UV laser intensity to yield a number of photoelectrons representative of BoO conditions. The shape of the pulses is obtained by scanning the discriminator threshold (LSB of 1.25 A). The time of the leading and trailing edges of – 17 – discriminator output pulse are measured by the TDC1 and TDC2, respectively, allowing the reconstruction of the pulse shape. Using this method, the peak of the pulse is not observed due to the configurable range of the discriminator threshold. Figure 13 shows a good agreement between simulation and data for the rising edge of the pulse. The slew rate in the rising edge at the level of the optimum threshold for timing measurement is 28.6 A/ns (to be compared to 28.0 A/ns in simulation). The observed small discrepancy in the pulse trailing edge is ascribed to a different DLED delay and trim (see section 3.1) in the chip used in the measurement and in the simulation. Figure 13. Pulse shape at the input of the discriminator reconstructed with a threshold scan as obtained with simulation data and experimental data using TOFHIR2X. The combined contributions of the electronics noise and TDC to the time resolution are estimated with blue laser light shining directly on two naked SiPMs using a beam splitter. The coincidence time resolution (CTR) is measured between the two channels and the channel time resolution is obtained as CTR/√2. Figure 14 shows the measured and the simulated channel time resolution as a function of the pulse slew rate (𝑑𝐼 𝑑𝑡 ⁄). Measurements and simulation are in good agreement. A fit of the data points with the function: 𝜎𝑡=𝜎𝑛𝑜𝑖𝑠𝑒 (𝑑𝐼 𝑑𝑡 ⁄ )⨁𝜎𝑇𝐷𝐶 ⁄ yields 𝜎𝑛𝑜𝑖𝑠𝑒= 0.360.03 A and 𝜎𝑇𝐷𝐶= 121 ps. The noise measurement agrees with the simulation estimation of the electronics noise at the input of the discriminator (0.42 A). The TDC resolution agrees with the direct measurement presented in the next section. The results reported in this section were obtained with TOFHIR2X and SiPMs of cell size 15×15 m2. – 18 – Figure 14. Time resolution of laser pulses directly detected by a SiPM as a function of the slew rate of the pulse rising edge in data and in simulation. The blue line represents a fit of the data points (see text). Results obtained with TOFHIR2X. 6.3 TDC performance The TDC calibration uses an external digital pulse synchronous to the system clock and distributed internally to all TDC inputs in the chip. The time of the test pulse relative to the clock edge, defined in the FPGA of the readout board, is scanned in steps of ten picoseconds. The distribution of the TDC binning derived from the calibration data is shown for 2032 TACs in 4 different ASICs in Figure 15. The average bin size is 11.3 ps which matches well the value of 10 ps expected from simulation. It is worth noting the low dispersion of TDC binning (r.m.s. = 0.4 ps). Figure 15. Normalized distribution of the TDC bin measured for 2032 TACs in 4 different ASICs. The linearity of the TDC is derived from the code density distribution measured with random digital pulses following a uniform distribution in time. Figure 16 shows the Differential NonLinearity (DNL) and the Integral Non-Linearity (INL) as a function of the TDC code obtained for two different TACs. Before linearity corrections, DNL is less than ± 0.5 LSB and the INL is less than ±2 LSB. – 19 – Figure 16. DNL and INL as a function of the TDC code obtained for two different TACs. The TDC resolution is derived from coincidences between two TDCs in the chip receiving a common test pulse. The average coincidence time resolution (CTR) estimated with 28 TDC pairs is 18.8 ps, from which we derive a TDC resolution of 13.3 ps as shown in Figure 17. In this measurement, the dispersion of the resolution is 5.3% r.m.s. Figure 17. Normalized distribution of TDC resolution. 6.4 Time resolution of MIP equivalent pulses In BTL, the timing of a MIP particle is obtained from the average of the two measurements in a single LYSO bar. The bar time resolution may be derived from the CTR of the two channels in the crystal bar (bar = CTR/2) 13 . We performed measurements with the UV laser tuned to generate a LYSO pulse with 9800 photoelectrons when the SiPMs are operated at overvoltage of 3.1 V (SiPM gain 9.7105), characteristic of BoO conditions. Figure 18 shows the bar time resolution as a function of the discriminator threshold for the optimal settings of the delay line in the noise suppression circuit. At the optimal threshold, we obtain a resolution of 24 ps, compatible with the simulation results. 13 Equivalent to the average of two measurements for a fixed impact point position along the bar if there are no correlated uncertainties between the two ends. – 20 – Figure 18. Time resolution of LYSO pulses characteristic of BoO as a function of the discriminator threshold. Figure 19. Time resolution of LYSO pulses characteristic of EoO as a function of the delay line in the DarkCount noise suppression circuit (right). To reproduce the large DCR at EoO, we use background LED blue light emulating the SiPM DarkCount noise. The calibration of the LED light is established by measuring the SiPM current as a function of the LED voltage. The SiPM current is converted to equivalent DCR by considering the SiPM gain at the operating overvoltage. We measure the channel time resolution – 21 – of laser pulses characteristic of EoO conditions, while illuminating the SiPMs with blue light emitted by the LED. Figure 19 shows the time resolution as a function of the delay in the DarkCount noise suppression circuit for pulses with 4400 p.e. and SiPM gain of 3.6105 and equivalent DCR of 18 GHz. In this measurement, the SiPM is operated at an overvoltage of 1.0 V as foreseen at EoO to limit the DarkCount noise. For the optimal setting of the delay line (590 ps), we obtain a resolution of 58 ps. The signal yield and SiPM parameters used in the measurements reported above are representative of measurements performed with the revised sensor module, assembled both with non-irradiated and irradiated SiPMs, in a test beam at CERN 14 . In order to assess the impact of the channel rate on the time resolution, we performed measurements of the time resolution in a single channel (one side of the LYSO bar) as a function of the average rate of UV laser pulses. The measurement was done in a single channel to reach 2.5 MHz without saturating the chip output bandwidth. The laser is triggered with a pseudorandom sequence of pulses. We chose an amplitude of the laser pulses such that the bar resolution is typical of EoO. We have observed that the laser is not able to generate pulses of constant amplitude when separated by less than 250 ns, therefore we removed from the analysis these pulses. Figure 20 shows that the time resolution remains stable as a function of the effective average rate of pulses used in the study. Figure 20. Time resolution as a function of the laser pulse rate (see text). 6.5 Energy of MIP equivalent pulses We have performed measurements of the event energy as given by the TOFHIR2 pulse charge integration using UV laser pulses impinging in the sensor module. The laser power was tuned to deliver a given number of photoelectrons per pulse as derived from the measurement of the SiPM current. We performed measurements at two values of the SiPM overvoltage, namely 3.1 V and 1.0 V, typical of BoO and EoO operations. In each of the two cases, we configured the QDC attenuator to match the expected range of pulse amplitude. 14 Test beam results will be described in a future publication. – 22 – Figure 21. Average energy (left) and relative energy resolution (right) as a function of the number of photoelectrons in the signal for two values of the SiPM overvoltage, typical of BoO (3.1V) and EoO (1.0V) conditions, respectively. In these measurements, as shown in Figure 21 (left), we observe a linear response in the amplitude range typical of BoO (overvoltage 3.1 V) and EoO (overvoltage 1.0 V). We also observe in Figure 21 (right) that in all cases the relative energy resolution is below the specified 5%. 6.6 Rate performance The rate performance of the TOFHIR2 chip was assessed with simulations and with test pulse measurements. The eight-fold analog buffering in TACs and QAC before digitization is potentially a source of inefficiency due to buffering overflow. We simulated pseudo-random pulses at the channel input of two types: MIP pulses at an average rate of 2.5 MHz, which are digitized by the ADC, and low energy pulses that are buffered in the TAC but rejected for digitization. The rate of the low energy pulses was varied between 5 MHz, the estimated rate of energy deposits above 100 keV in BTL, and 32 MHz, the average bunch crossing rate at LHC. In the simulation, we conservatively assumed 100 ns per measurement digitization, and considered the cases of two measurements (one timestamp and one charge) and three measurements (two timestamps and one charge). As shown in Figure 22, the channel inefficiency is 1% at the nominal rate of low energy events of 5 MHz raising to 4% at the rate of 32 MHz. Configuring the chip to perform only two measurements, the inefficiency at 5 MHz becomes negligeable and the maximum inefficiency is below 3%. – 23 – Figure 22. Channel inefficiency for a MIP rate of 2.5 MHz as a function of the rate of low energy events. Using the TOFHIR2 test system with adequate firmware, we could inject pseudo-random test pulse sequences in the ASIC that emulate BTL operation parameters, namely average 2.5 MHz digitized events per channel and 3% L1 trigger acceptance (2.67 MHz events transmitted per ASIC). Every 25 ns a pseudo random number, x, is generated. If x < tp a test pulse is generated triggering the TDC, and if x < tgr a L1 trigger acceptance is generated, where tp and tgr are tunable parameters. The L1 trigger acceptance is transmitted with a 12.125 µs delay, the expected latency in the experiment. L0 is not used in these measurements. Correct ASIC L1 latency setting can be identified by sweeping the setting and observing event counts (Figure 23). Figure 23. Event output rate as a function of the L1 latency setting. When generating test events in a single channel, we could observe that the rate of transmitted events follows the trigger acceptance rate up to the maximum chip output rate, as shown in Figure 24. Generating test events simultaneously in various channels we have observed that the output event rate grows linearly with the number of activated channels up to the maximum output rate of 2.67 M events/s (Figure 25). These results prove that TOFHIR2 is capable to operate in the high-rate conditions of BTL at the HL-LHC. Indeed, a GEANT4 detector simulation with 200 pile-up events per crossing has shown that the maximum channel occupancy for energy deposits above 1 MeV is 8% [2], corresponding to an average of 2.56 activated channels per crossing. Therefore, a L1 trigger rate of 750 kHz as planned in CMS leads to an output rate of 1.92 M events/s, which is 72% of the chip output bandwidth (one Tx link). – 24 – Figure 24. Channel event rate as a function of the L1 trigger rate. Figure 25. Output event rate as a function of the number of activated channels for an input rate of 2.5 MHz per channel and for two values of the L1 trigger acceptance. 7. Radiation tests 7.1 TID radiation tests The resistance to TID radiation of the three TOFHIR2 prototypes was tested at the X-ray irradiation facility at CERN (Figure 26-left). The maximum expected dose in the barrel MTD is 2.9 Mrad. At each campaign, we have irradiated two ASICs, the first one up to 7 Mrad in steps of 0.5 Mrad and another up to 3 Mrad in steps of 1 Mrad. Several measurements were performed between steps, as well as after the irradiation. The irradiations were performed at -25 oC. In the first campaign, we irradiated TOFHIR2A fabricated in foundry B. Using the test pads, we measured the current consumption and the bandgap voltage, and we performed several DAC voltage scans (amplifier baseline, discriminator threshold, QAC/TAC baseline, ALDO DAC). Using the TOFHIR2 data acquisition, we monitored the front-end noise from a scan of the T1 discriminator threshold, measured the shape of the internal analog test pulse using a scan of the T2 discriminator threshold, and performed the TDC and QDC calibrations in all TACs and QACs in the two ASICs. The time interval between steps was of the order of 10 minutes to prevent annealing effects. The same measurements were performed 12 hours after the irradiation in order to assess the effect of annealing at room temperature. Additionally, the time resolution with laser pulses was measured using the ASIC irradiated at 3 Mrad. – 31 – 10.1088/1748-0221/8/ 02/C02050. [17] A. Di Francesco et al., TOFPET2: a high-performance ASIC for time and amplitude measurements of SiPM signals in Time-of-Flight applications, Journal of Instrumentation, vol. 11, no. 03, p. C03042, 2016. DOI: 10.1088/17480221/11/03/C03042. [18] D. Schug et al., Initial Measurements with the PETsys TOFPET2 ASIC Evaluation Kit and a Characterization of the ASIC TDC, IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 3, no. 4, pp. 444-453, Jul. 2019. DOI:10.1109/TRPMS.2018.2884564. [19] V. Nadig et al.,, Investigation of the Power Consumption of the PETsys TOFPET2 ASIC, in IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 4, no. 3, pp. 378-388, May 2020, DOI:10.1109/TRPMS.2019.2955032. [20] H. Chen et al., A dedicated readout ASIC for Time-of-Flight Positron Emission Tomography using silicon photomultiplier (SiPM), in Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC), 2014, pp. 1–5. DOI:10.1109/NSSMIC.2014.7431045. [21] V. Stankova et al., STIC3 - silicon photomultiplier timing chip with picosecond resolution, Nuclear Instruments and Methods in Physics Research A, vol. 787, pp. 284– 287, 2015. DOI: 10.1016/j.nima.2014.12.073. [22] J. Fleury et al., PETIROC, a new front-end ASIC for time of flight application, IEEE Nuclear Science Symposium and Medical Imaging Conference, 2013, DOI: 10.1109/NSSMIC.2013.6829018. [23] S. Ahmad et al., TRIROC: a multi-channel SiPM readout ASIC for PET/PET-TOF application, IEEE Transactions on Nuclear Science, vol. 62, no. 3, pp. 664–668, 2015. DOI: 10.1109/ TNS.2015.2397973. [24] S. Ahmad et al., TRIROC, a versatile 64-channel SiPM readout ASIC for Time-of-Flight PET, IEEE Nuclear Science Symposium, Medical Imaging Conference and RoomTemperature Semiconductor Detector Workshop (NSS/MIC/RTSD), 2016. DOI:10.1109/NSSMIC.2016.8069882 [25] P. Fischer et al., Multi-channel readout ASIC for TOF-PET, IEEE Nuclear Science Symposium Conference Record, vol. M11-140, pp. 2523–2527, 2006. DOI:10.1109/NSSMIC.2006.354423 [26] I. Sacco et al., PETA4: a multi-channel TDC/ADC ASIC for SiPM readout, Journal of Instrumentation, vol. 8, 2013. DOI:10.1088/1748-0221/8/12/C12013. [27] A. Comerma et al., FLEXTOT - current mode ASIC for readout of common cathode SiPM arrays, 2013 IEEE Nuclear Science Symposium and Medical Imaging Conference (2013 NSS/MIC), 2013. DOI: 10.1109/nssmic.2013.6829761. [28] J. M. Cela et al., A Compact Detector Module Design Based on FlexToT ASICs for Time-of-Flight PET-MR, IEEE Transactions on Radiation and Plasma Medical Sciences, vol. 2, no. 6, pp. 549–553, Nov. 2018. DOI: 10.1109/TRPMS.2018.2870927. [29] S. G. Fernandez et al., A High Dynamic Range ASIC for Time of Flight PET with monolithic crystals, 2019. DOI: 10.22323/1.343. 0085. [30] X. Zhu et al., Development of a 64-channel readout ASIC for an 8x8 SiPM array for PET and TOFPET applications, IEEE Transactions On Nuclear Science, 2016, DOI: 10.1109/TNS.2016.2518808. [31] S. Gundacker et al., High-frequency SIPM readout advances measured coincidence time resolution limits in TOF-PET, Physics in Medicine & Biology, vol. 64, no. 5, p. 055 012, Feb. 2019. DOI: 10.1088/1361-6560/aafd52. [32] Radiation Handbook for Electronics, Texas Instruments. [33] L. Gonella et al., Total Ionizing Dose effects in 130-nm comercial CMOS technologies for HEP experiments, Nuclear Instruments and Methods in Physics Research Section A, Volume 582, Issue 3, 1 December 2007, Pages 750-754, DOI 10.1016/j.nima.2007.07.068. [34] F. Faccio et al., Radiation-Induced Short Channel (RISCE) and Narrow Channel (RINCE) Effects in 65 and 130 nm MOSFETs, IEEE Transactions on Nuclear Science, Volume: 62, Issue: 6, December 2015, DOI 10.1109/TNS.2015.2492778 [35] A. Gola, C. Piemonte and A. Tarolli, The DLED Algorithm for Timing Measurements on Large Area SiPMs Coupled to Scintillators, IEEE Transactions on Nuclear Science, Vol. 59, No. 2, April 2012, DOI: 10.1109/TNS.2012.2187927. [36] H. Traff, Novel Approach to High Speed CMOS Current Comparators, IEEE Proceedings - Electronic Letters, vol.28, No.3, 1992. DOI: 10.1049/el:19920192. – 32 – [37] E. Albuquerque et al., A Full Current-Mode Timing Circuit with Dark Noise Suppression for the CERN CMS Experiment, ESSCIRC 2022IEEE 48th European Solid State Circuits Conference (ESSCIRC), Milan, Italy, 2022, pp. 177-180, DOI: 10.1109/ESSCIRC55480.2022.9911529. [38] M. Firlej, T. Fiutowski, M. Idzik, S. Kulis, J. Moron, K. Swientek, An Ultra-Low Power 10-bit, 50 MSps SAR ADC for Multi-Channel Readout ASICs, Journal of Instrumentation, JINST 18 (2023) P11013. DOI: 10.1088/1748-0221/18/11/P11013 [39] G. Anelli, M. Campbell, M. Delmastro, F. Faccio, S. Floria, A. Giraldo, E. Heijne, P. Jarron, K. Kloukinas, A. Marchioro, P. Moreira, W. Snoeys, Radiation tolerant VLSI circuits in standard deep sub-micron CMOS technologies for the lhc experiments: practical design aspects, IEEE Transactions on Nuclear Science 46 (6) (1999) 1690–1696. DOI:10.1109/23.819140. [40] W. Snoeys et al., Layout techniques to enhance the radiation tolerance of standard CMOS technologies demonstrated on a pixel detector readout chip, Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment 439 (2) (2000) 349–360. https://doi.org/10.1016/S0168-9002(99)00899-2. [41] F. Faccio, Radiation issues in the new generation of high energy physics experiments, International Journal of High Speed Electronics and Systems 14 (02) (2004) 379–399. DOI:10.1142/S0129156404002429. [42] Techniques for radiation effects mitigation in ASICs and FPGAs handbook, ESA-ESTEC, ECSS-QHB-60-02A. [43] S. Kulis, Single Event Effects mitigation with TMRG tool, 2017 JINST 12 C01082, DOI:10.1088/17480221/12/01/C01082. [44] https://www.petsyselectronics.com/web/public/products/6 [45] P. Carniti et al., ALDO2 a multi-function rad-hard linear regulator for SiPM-based HEP detectors, Nuclear Instruments and Methods in Physics Research Section A, Volume 1039, 11 September 2022, 167028, DOI:10.1016/j.nima.2022.167028 [46] A. Bornheim et al., Integration of thermo-electric coolers into the CMS MTD SiPM arrays for operation under high neutron fluence, 2023 JINST 18 P08020, https://iopscience.iop.org/article/10.1088/1748-0221/18/08/P08020