Performance evaluation of the FastIC readout ASIC with emphasis on Cherenkov emission in TOF-PET

Objective. The efficient usage of prompt photons like Cherenkov emission is of great interest for the design of the next generation, cost-effective, and ultra-high-sensitivity time-of-flight positron emission tomography (TOF-PET) scanners. With custom, high power consuming, readout electronics and fast digitization the prospect of sub-300 ps FWHM with PET-sized BGO crystals have been shown. However, these results are not scalable to a full system consisting of thousands of detector elements. Approach. To pave the way toward a full TOF-PET scanner, we examine the performance of the FastIC ASIC with Cherenkov-emitting scintillators (BGO), together with one of the most recent SiPM detector developments based on metal trenching from FBK. The FastIC is a highly configurable ASIC with 8 input channels, a power consumption of 12 mW ch−1 and excellent linearity on the energy measurement. To put the timing performance of the FastIC into perspective, comparison measurements with high-power consuming readout electronics are performed. Main results. We achieve a best CTR FWHM of 330 ps for 2 × 2 × 3 mm3 and 490 ps for 2 × 2 × 20 mm3 BGO crystals with the FastIC. In addition, using 20 mm long LSO:Ce:Ca crystals, CTR values of 129 ps FWHM have been measured with the FastIC, only slightly worse to the state-of-the-art of 95 ps obtained with discrete HF electronics. Significance. For the first time, the timing capability of BGO with a scalable ASIC has been evaluated. The findings underscore the potential of the FastIC ASIC in the development of cost-effective TOF-PET scanners with excellent timing characteristics.


Introduction
The detection chain for a light-based radiation detector system, used both in medical imaging and high energy physics experiments, comprises a scintillator to convert high energy particles into optical photons; a photodetector to convert the optical photons into an analog signal; readout electronics to amplify and/or discriminate the signal to provide information on the time-of-arrival and energy of the initial energetic particle; and lastly a data acquisition system to provide digital data for further physics analysis or image processing (Lecoq et al 2023).
This detector chain is complex, and to boost the performance, all components need to be optimized.For example, the state-of-the-art scintillator material for time-of-flight positron emission tomography (TOF-PET) is based on lutetium (LSO:Ce, LYSO:Ce, LGSO:Ce and Ca/Mg co-doped variants) providing a high number of photons per energy deposition, as well as a fast scintillation response with decay times of around 40 ns (Gundacker et al 2016b).For better sensitivity, but also to overcome limitations based on standard scintillation, new materials and processes are investigated, such as prompt Cherenkov photon emission in BGO (Kwon et al 2016).
Of similar importance is the choice of the photodetector.Nowadays, silicon photomulipliers (Gundacker and Heering 2020) (SiPMs) have mostly substituted photomultiplier tubes due to their compactness, better detection efficiency, and insensitivity to magnetic fields, which can be a requirement when PET is combined with magnetic resonance imaging.Important characteristics of SiPMs for TOF-PET imaging are the photon detection efficiency (PDE), single photon time resolution (SPTR), dark count rate (DCR) and crosstalk probability (Gola et al 2019).On the latter, significant crosstalk reduction has been reported due to the implementation of metal-filled trenches between individual single photon avalanche diodes (SPADs) (Merzi et al 2023).
Major time resolution breakthroughs have been achieved by the use of custom high frequency readout electronics to mitigate the contribution of electronic noise, provide a fast output signal, and apply leading edge discrimination at the software level using the digitized waveform (Cates et al 2018, Gundacker et al 2019).However, lowering the high power consumption when using off-the-shelf electronic components (Cates et al 2022, Krake et al 2022) is mandatory in terms of scalability (Cates et al 2022) for a scanner, consisting of thousands of individual channels, and thus imposing a challenge that various groups are actively working on (Pourashraf et al 2022, Gonzalez-Montoro et al 2023, Lee et al 2023, Nadig et al 2023b).
Application-specific integrated circuits (ASICs) are typically applied in systems because they can process many input channels with low power consumption in a compact area.A summary of available ASICs used for TOF applications is provided in table 1.The FastIC ASIC with 8 channels is of particular interest, due to a combination of relatively low power consumption of 6-12 mW ch −1 depending on the operation mode, excellent linearity of the input stage (about 3% linearity error for the full energy readout), and encouraging SPTR (Gomez et al 2021).The latter is of the most relevance to exploit Cherenkov emission, as it is the most determining parameter to obtain a good timing capability for low light intensities (Dolenec et al 2016, Cates and Craig 2019, Ota et al 2019, Gundacker et al 2020, Ariño-Estrada et al 2021, Kratochwil et al 2021a, 2021b, Terragni et al 2022, Gundacker et al 2023, Jens and Vandenberghe 2023, Kratochwil 2023).
In this manuscript, we evaluate the timing performance of the FastIC in combination with the latest generation of analog SiPMs from FBK (NUV-HD-MT technology) coupled to BGO and LSO:Ce:0.2%Cacrystals.Both the intrinsic timing capability using small crystals as well as the performance with long crystals, which are typically used in PET scanners, are investigated.To have a fair evaluation with the state-of-the-art time resolution with minimal impact of electronic noise and time walk effects, comparison measurements with custom high frequency readout electronics of the same detectors were performed.

Scintillating crystals
Measurements were performed with different crystals summarized in table 2, always wrapped in several layers of Teflon and glued with Cargille Meltmount (refractive index n c = 1.582) to the SiPMs.Two distinct crystal materials were evaluated: lutetium oxyorthosilicate doped with cerium and co-doped with 0.2% calcium (LSO: Ce:0.2%Ca) and bismuth germanate (BGO).

Silicon photomulipliers
Initial test measurements were performed with SiPMs from Hamamatsu photonics (S13360 3050PE), as they have been commercially available for a long time and thus the most common sensor used in literature, and with FBK NUV-HD-MT SiPMs.After confirming the superior timing performance (Gundacker et al 2020, Kratochwil et al 2021a) of the NUV-HD(-MT) technology, to be valid also for FastIC, in-depth measurements were performed using only the mentioned new MT-technology from FBK.The SiPMs have a pixel pitch of 40 μm with an effective photosensitive area of 3 × 3 mm 2 and no protective resin.The intrinsic SPTR of the SiPMs was measured to be about 50 ps FWHM after correcting for the electronic noise and contribution of the laser jitter/ reference detector (Gundacker et al 2023, Penna et al 2023).

Coincidence time resolution (CTR) setups
Two different CTR setups were used for the evaluation of timing performance.The first one is based on high frequency (HF) readout electronics (Cates et al 2018, Gundacker et al 2019) and a LeCroy DDA735Zi (3.5 GHz, 20 Gs s −1 ) oscilloscope (Gundacker et al 2016a).A balun transformer in combination with two radio frequency amplifiers was used to generate a fast analog time signal used for leading edge discrimination, while an analog operational amplifier (AD8000) was employed to monitor the number of detected emitted photons (energy signal).A comprehensive discussion of the setup and implementation of the custom readout electronics is given in Gundacker et al (2019).
For the second setup, two FastIC-evaluation demonstrator boards (ICC-UB 2023) were used, consisting of two stacked printed circuit boards, namely the FastIC board and a control board shown in figures 1 (a)-(c).
The FastIC board features sockets for testing individual SiPMs and SiPM arrays, along with a dedicated highvoltage power supply connection.Additionally, it offers various output access options for the ASIC (Gomez et al 2021).The control board facilitates the configuration of the FastIC ASIC via a computer, using a FPGA (Speed Grade 7 Intel MAX 10) for slow control and a USB driver for communication.The produced output signals were captured using a LeCroy WaveRunner 104MXi-A oscilloscope (1 GHz, 10 Gs s −1 ).The oscilloscope is set to trigger based on the simultaneous energy signals, and data regarding energy and time are directly measured in the oscilloscope from each acquired waveform.

FastIC signal processing
The input signal generated by the SiPM following a gamma-ray interaction in the crystal is processed by the FastIC front-end electronics.It triplicates the current pulse to three signal processing paths with different internal gains with the task of individually providing information on the timing, energy and triggering as illustrated in figure 2.
The timing path always creates an output pulse after the signal crosses a certain time threshold set by a comparator.The timing information is encoded on the rising edge.The energy processing path produces a linear energy output pulse with the energy information encoded in the pulse width.The FastIC offers two configurable options for energy output signal generation.The first option involves generating the signal when it exceeds the set timing threshold.The second option generates the signal only when it reaches a predetermined threshold in the triggering path.As illustrated in figure 2, this feature provides the flexibility to choose the most appropriate method for a certain application.In this work, the FastIC was configured with the second option.The gain on the energy path can be adjusted to count either individual photons or measure thousands of scintillating photons.In  a similar manner, the timing and trigger threshold can be adjusted with the aim of providing a low leading edge threshold (typically at the level of 10% to 200% of a single SPAD signal amplitude) and a flexible trigger threshold aiming to discard dark count events and events with low energy deposition.Energy (based on the linear measurement of the signal amplitude) and timing signal are combined into one signal, with the first pulse providing the timing information and the second pulse containing the energy information.However, in this study, we disabled the timing pulse and routed it via fast-OR signal to a different output channel.
Figure 3 illustrates the two potential setups.In option 1, depicted in (a), the energy and timing signals are merged into a single signal, where the initial pulse conveys timing details through the rising edge, and the subsequent pulse encodes energy information within the pulse duration.For this study, option 2 was employed, as shown in (b), where the timing and energy pulses are distinct and captured using separate channels on an oscilloscope (c).
Although option 2 doubles the number of readout channels, it simplifies the oscilloscope's data acquisition process, allowing individual signal processing.In total 4 channels (2 for timing, 2 for energy) were used.
Since the FastIC output signals are binary pulses, the performance of the digitizer is less demanding with respect to fully analog signals as long as the sampling rate is sufficient to capture points on the rising edge.

Data analysis
At least 10 000 coincidence events were recorded for both setups and each configuration.The CTR was determined by calculating the time difference between the arrival times of signals from opposite-facing sensors, as shown in figure 4(a).The signal integral (HF readout) or signal amplitude (encoded in the width of the pulse) was used for energy discrimination selecting at 511 keV photopeak events as illustrated in figure 4 (b).In particular, a Gaussian fit of the photopeak region is performed, and the center of the Gaussian at 511 keV plus or minus 1.5 times the standard deviation is used to perform the energy cut.With a measured energy resolution of BGO (without SiPM saturation correction) of 20%, this translates to an energy window between 440 keV and 570 keV.For the selected events, the time difference is histogrammed as shown in figure 4 (c) and the CTR value is obtained as the full width at half maximum (FWHM) of the coincidence time delay histogram.For measurements with LSO, the time difference was modeled with a simple Gaussian distribution, while for BGO measurements, a fit consisting of two Gaussian distributions was used (Kratochwil et al 2021a) and the FWHM calculated as the difference between the crossing points of half the maximum.

Photon counting for FastIC threshold configuration
To calibrate the threshold settings for the FastIC, staircase plots were generated by quantifying the number of cells triggered at different threshold levels (in units of LSB) with the FPGA.The photon count data obtained at different SiPM overvoltages and different gains (I LSB dictates the increment between two consecutive thresholds) are presented in figure 5.The I LSB therefore controls the size and number of steps in the threshold sweep.It can be seen that for high overvoltage only a few stairs are visible and each step ranges over multiple threshold values, while for lower overvoltage the effective DCR is drastically reduced due to a combination of low gain (thus the threshold is at the level of several triggered cells) and low SiPM crosstalk probability (it becomes less likely that a single dark count in a cell triggers multiple cells).
To have sufficient dynamic range, but also fine granularity around the level of a single triggered SPAD, a global I LSB setting of 15 was used for the following measurements.

SiPM overvoltage scan and saturation of the energy signal
One advantage of the new MT-technology is the reduction of internal crosstalk, which allows to bias the SiPMs at higher overvoltage to increase gain, PDE and SPTR (Merzi et al 2023).To make use of this improvement, CTR measurements were performed at different SiPM overvoltages with 20 mm long LSO crystals.Figure 6 shows the obtained energy histograms at different SiPM overvoltages (a)-(c) as well as the CTR as function of the overvoltage (d).It can be seen that for high overvoltage the dynamic range of the ASIC reaches its limitation, as the Compton edge and photopeak are no longer well resolvable.Thus, at higher overvoltage also events depositing less energy in the crystal were considered which degrades the measured CTR.The best CTR value of 133 ± 5 ps was achieved at 10 V overvoltage.For these measurements, the timing discriminator threshold was not optimized.

CTR results for BGO and LSO with the FastIC
For the optimum overvoltage of 10 V, a threshold scan was performed for all tested crystals.Following the performed calibration in figure 5, the amplitude of a single SPAD cell corresponds to about 8 LSB, thus the scan ranges from [0-4] triggered SPADs for BGO and [0-2.5]triggered SPADs for LSO.The top of figure 7 (a)-(c) shows the selected energy selection for 20 mm long BGO crystals and the resulting coincidence time delay distribution, while in the bottom the CTR as function of the discriminator threshold is plotted for (d) BGO and (e) LSO.
The reference measurement of the small BGO crystals coupled to HPK SiPMs is also included, showing that the FBK sensors achieve about 30% better time resolution with respect to the HPK technology.This trend is consistent with published data in Kratochwil et al (2021a) using HF readout electronics.For BGO, operating at very low threshold settings is vital to preserve the good timing response arising from the detection of prompt Cherenkov emission with measured CTR values of 330 ± 8 ps FWHM for small pixels and 490 ± 12 ps FWHM for 20 mm long crystals.Given the bright and fast scintillation response of LSO, there is only a small threshold dependency for LSO (Seifert et al 2012), resulting in measured CTR values of 95 ± 3 ps and 129 ± 4 ps FWHM for 3 mm and 20 mm long crystals respectively.
The energy resolution obtained by FastIC at 511 keV is approximately 22% for the 20 mm long crystals and 20% for the 3 mm long crystals without applying any correction for SiPM saturation.The non-linearities of SiPM at large signals should not be significant for BGO due to its low light yield compared to LSO crystals that requires to compensate this non-linearity to compute the energy resolution (David Sánchez et al 2022).

Timing limits with HF readout electronics
The timing performance of the same detectors (crystals glued to SiPMs) was also investigated with high power consuming readout electronics to asses the intrinsic limits.In this case, there was no limitation on maximum overvoltage, meaning that the SiPMs were biased up to 16 V above breakdown voltage.In figure 8 the measured time resolution as function of the leading edge threshold is shown for (a) BGO and (b) LSO.
Particularly for the long BGO crystal, one can observe a strong threshold dependency and the optimum settings are reached for values just above the electronic noise floor.
Although not scalable to a full system, the obtained CTR values for LSO and BGO (upon time walk correction) are consistent with measurements reported in Gundacker et al (2020), ( 2023

Comparison of readout electronics for LSO
Comparing the obtained CTR results for the FastIC with HF readout electronics, we notice an offset of about 30 ps for LSO with the ASIC for both crystal geometries.Such an offset is expected due to the following reasons: first, the FastIC input stage might be limited in the bandwidth meaning in combination with the capacitance of the SiPM and additional impedance from the connectors, it reduces the effective bandwidth (Fernández-Tenllado et al 2019).This means that the SiPM signal slew rate (dV/dt) is lower, resulting in an additional contribution of electronic noise (Cates et al 2018).Due to this electronic noise contribution, the optimal threshold is reached at higher leading edge threshold settings (eg.for 20 mm long LSO at 5 LSB (≈60% of a single SPAD signal amplitude), compared to ≈20 mV (25% of a single SPAD signal amplitude) for the FastIC and HF readout, respectively).Although LSO is a bright scintillator with many photons produced in the first few  nanoseconds, having the timestamp at a later time can worsen the statistical contribution of the scintillation response (Seifert et al 2012).Due to the interplay between electronic noise contribution (scaling inverse, CTR 1 threshold µ ) and photo statistics (scaling linear, CTR threshold µ with the leading edge threshold (Gundacker et al 2020)), the optimum settings are reached at higher values for the FastIC.
Next, the timestamp for the FastIC measurement is obtained by the in-build timing comparator, which has limited performance.On the contrary with the HF readout measurement, the timing is obtained via leading edge discrimination directly on the high-performance oscilloscope with negligible timing jitter.
Lastly, the FastIC was limited in terms of signal saturation at high overvoltage, thus lower bias voltage had to be applied.Both PDE and measured SPTR improve further by a few percentage when increasing the excess bias voltage above 10 V (Merzi et al 2023).Adjusting the overvoltage to 10 V for the HF readout measurement results in a mild CTR degradation from 57 ± 5 to 64 ± 4 ps for the small and from 95 ± 3 to 104 ± 3 ps FWHM for the long LSO crystals.
As mentioned before, the performance of the digitizer/oscilloscope is negligible due to the digital-like output signal of the FastIC.With a measured signal slope (dV/dt) of 1.2 V ns −1 and a RMS of the baseline of 3 mV, this translates into a jitter (

Comparison of readout electronics for BGO
All the mentioned considerations are also valid for BGO, however the impact is more severe in the case of prompt photons.Given the low amount of detected light, the analog signal slew rate of the SiPM is smaller, hence higher electronic jitter.Placing the leading edge threshold at higher values when the signal is steeper does not help either, because by then we face strong statistical fluctuations based on the number of detected prompt photons (Kratochwil et al 2021a).Given those considerations, having a low electronic noise contribution already at very low leading edge thresholds is crucial for low light intensities (Dolenec et al 2016, Kratochwil et al 2021b).Another aspect, more dominant for Cherenkov photons in BGO, are time-walk effects.For a higher leading edge threshold, time-walk effects have more time to propagate and contribute.This behavior can be well observed for the threshold scan with the HF readout measurement displayed on the left of figure 8, with measured CTR values of 157 ± 4 ps at 10 mV (12% of the single SPAD signal amplitude), 176 ± 4 ps at 20 mV (25% of the single SPAD signal amplitude), and 199 ± 4 ps at 40 mV (50% of the single SPAD signal amplitude).Also, the limitation in overvoltage contributes, for instance at 12 V overvoltage instead of 16 V, the measured CTR values with HF readout electronics are 172 ± 5 ps FWHM and 296 ± 6 ps for the small and long crystals, respectively, while for 16 V overvoltage the CTR values are 157 ± 4 ps and 273 ± 5 ps.

Time walk correction
For leading edge discrimination, it is normal to experience some kind of time-walk as signals with a high number of detected photons pass the threshold earlier compared to signals with a lower number.Hereby, the selected energy window is a determining factor: for a narrow energy selection, we can assume that the signal variation is small, and hence little jitter comes from these variations.In addition, the magnitude of time-walk effects depends on the value of the leading edge threshold-as mentioned before placing it at high values means variations of the signal have more time to propagate, enhancing the effect.
To assess the potential contribution with the FastIC measurements coming from time walk, the coincidence time delay is plotted against the signal amplitude (expressed as time due to the nature of the energy measurement) for 20 mm long LSO (left) and 20 mm long BGO (right) crystals in figure 9.
In both cases we observe time walk effects when selecting all the events (top), while after photopeak selection (bottom) there is no clear time-walk visible and therefore no additional room for improvement.
For the special case of Cherenkov radiation in BGO, there is another fluctuation due to the number of detected prompt photons.It was shown in Kratochwil et al (2020) that by measuring the signal rise time of the analog amplified signal, these fluctuations can be monitored on an event-to-event basis and corrected with respective time resolution improvements.
Although for the measurement with HF electronics we were able to access the signal rise time and perform such a correction (resulting in a CTR improvement from 157 ± 4 to 123 ± 4 ps) for 3 mm and 273 ± 5 to 228 ± 6 ps) for 20 mm long BGO crystals, in depth discussed in Kratochwil (2023)), for the FastIC measurements we only have one timestamp available, and thus no access to monitor signal variations coming from the detection of prompt photons.One solution-on a single pixel level and solely as proof of conceptmight be to utilize the trigger signal to get a second timestamp and from this calculate the signal rise time as the difference between the timestamps generated by the FastIC timing output and the trigger output.However, this requires substantial modifications on the data acquisition and yields a lot of unwanted signals (as the trigger threshold should be set low, in the level of a few triggered SPADs, where we capture many dark count events).
Moreover, the trigger signal has a larger jitter, adding another source of variation that needs to be taken into account in accurately measuring the rise time with two thresholds.Verifying this workaround to validate the applicability of time-walk correction for prompt photons with an ASIC is subject to dedicated future work.
4.4.Impact of the energy window on the timing performance with BGO and the FastIC Similar to Kratochwil et al (2021b), we tested the impact of the energy window on the achievable timing resolution with the 20 mm long BGO crystals.Adjusting the upper energy windows (above the 1.5 standard deviations, about 570 keV) has a negligible effect on the CTR values, whereas modifying the lower energy threshold can impact the CTR.Changing the lower energy threshold from 440 to ≈400 keV, ≈300 keV and ≈150 keV yields CTR values of 490 ps, 500 ps, 540 ps and 600 ps respectively.The timing degradation is mild, mostly because the majority of the events (given the high photofraction in BGO) deposit the full energy in the crystal.
This means, that accepting low energy events on a BGO TOF-PET scanner based on the FastIC would further enhance the good sensitivity with only a mild timing degradation.However, this sensitivity gain comes with a drop in spatial resolution and selecting the optimum configuration will depend on the specific imaging task.

FastIC+ and considerations toward a system
In this study, an evaluation of the FastIC was performed based on single channels.The aim was to expand the range of performance already studied (Gomez et al 2021, Mariscal-Castilla et al 2023) and to gain a better understanding of how this ASIC works with the most recent SiPM technology in combination of prompt Cherenkov emission.Although the CTR values are better with HF readout electronics, the obtained results with the FastIC are scalable.Each of the FastIC boards is equipped with two ASICs, thus in total 16 channels (e.g. 4 x 4 SiPM crystal array) can be read out with low power consumption of only 12 mW/channel.The current version does not include a time-to-digital converter (TDC), thus the requirements on the signal processing are high.This major limitation is addressed in the successor chip FastIC+, which will include a TDC with a simulated jitter contribution of only 22 ps FWHM.
We can expect mild timing degradation (Nadig et al 2023b) when using 1:1 crystal-SiPM coupling and due to light sharing / enhanced SiPM crosstalk for the full SiPM array.However, already a CTR of 500 ps on a system level constitutes a more than factor 5 gain of effective sensitivity with respect to non-TOF capability assuming a diameter object of about 40 cm (Maurizio 2009).

Conclusion
In this study, we have presented the results of our examination of the FastIC readout ASIC with Cherenkov emitting scintillators and one of the most recent SiPM detector developments.The best measured coincidence resolution at 511 keV the FastIC were 330 ps and 490 ps FWHM for small and long BGO crystals.Although the CTR values are substantially better using power-hungry high frequency readout electronics, already these figures show the potential of the FastIC as a readout ASIC for TOF-PET applications utilizing Cherenkov-emitting scintillators.For state-of-the-art LSO scintillators coupled to FBK NUV-HD MT SiPMs and read out with the FastIC, we demonstrate sub-100 ps for 3 mm long crystals, and even for 20 mm long crystals the timing deterioration is only about 30 ps with respect to the limits with HF-readout electronics.
Further research is necessary to determine whether and to which extent the addressed limitations in terms of electronic noise and time-walk contribution can be overcome.Additionally, a study is required to determine whether an increase in the dynamic range of the ASIC would improve the CTR performance for SiPMs with MTtechnology.Although this study focused on the performance evaluation on a single channel level, the reported encouraging results are scalable, and no substantial timing degradation is expected.The findings underscore the potential of the FastIC ASIC in the development of cost-effective TOF-PET scanners with excellent timing characteristics.

Figure 1 .
Figure 1.(a) Schematic depiction of the measurement setup with FastIC and the data acquisition with an oscilloscope.(b) CTR measurement setup picture and (c) assembled control-and FastIC-board picture.FPGA and USB driver are placed under the FastIC board.

Figure 2 .
Figure 2. Schematic depiction of the signal processing path through the FastIC ASIC and its produced outputs for data processing.

Figure 3 .
Figure 3. (a) ASIC output configuration showing timing and energy information combined in one signal.(b) ASIC configuration for separated time and energy signal.(c) Simplified measurement setup showing the difference between option 1 and option 2 on the oscilloscope.

Figure 4 .
Figure 4. Schematic depiction for measuring and obtaining the CTR value.(a) Time difference signal produced by two opposite detectors based on the time-of-arrival.(b) Energy spectrum for photopeak event selection.The time differences are further analyzed on the basis of the energy selection.(c) CTR values is obtained by the FWHM of the time delay histogram based on the energy selection.

Figure 5 .
Figure 5. Recorded photon counts for (a) different bias voltages and (b) different global current threshold I LSB configurations for the FastIC ASIC obtained with the FPGA.The I LSB parameter represents the resolution of each LSB, smaller I LSB , smaller size of each LSB.The measurement time for each threshold was about 250 ms.

Figure 6 .
Figure 6.(a)-(c) Measured 22 Na energy spectrum for different overvoltages with 20 mm long LSO scintillator, measured at a timing threshold of 5 LSB (with I LSB = 15).(d) CTR values obtained for different overvoltages.

Figure 7 .
Figure 7. Top: energy spectrum for the two 20 mm long BGO crystals in coincidence with selected events (a)-(b) and resulting time delay histogram after energy selection with a double Gaussian fit function (c).Bottom: measured CTR values with the FastIC for (d) BGO and (e) LSO.

Figure 8 .
Figure 8. HF electronic, CTR measurement values obtained for (a) BGO and (b) LSO:Ce:0.2%Cascintillator crystals.The best HF CTR values were 157 ± 4 ps (time walk corrected, 123 ± 4 ps) for 3 mm and 273 ± 5 ps (time walk corrected, 228 ± 6 ps) for 20 mm long BGO crystals and 57 ± 5 ps and 95 ± 4 ps for 3 mm and 20 mm long LSO:Ce:0.2%Cacrystals.With this electronics, the amplitude of a single fired cell at 16 V overvoltage corresponds to about 80 mV.Details on the rise time correction method for BGO can be found in Kratochwil et al (2020) and Kratochwil (2023).
per channel.Correcting for does only change the decimal place, eg.from 95 to 94.6 ps FWHM.

Figure 9 .
Figure 9. Measured time differences in coincidence at time-of-arrival for 20 mm long LSO (a) and BGO (b) crystals, as well as a detailed view of the time corresponding energy selection for LSO (c) and BGO (d).

Table 1 .
Summary of ASICs used for TOF measurements.Due to the use of different SiPMs, scintillators and measurement setups, it is not always possible to directly compare the values, thus the table is intended to provide an overview and references to literature.

Table 2 .
Overview of scintillator crystalls used in this work.