Coincidence time resolution of radiation detector based on 6×6 mm² ToF SiPM detectors with different readout schemes

1 Introduction

In recent years, Silicon Photomultiplier (SiPM) had been increasingly used in many research fields such as nuclear medicine [1–5]. Time-of-Flight (ToF) performance is a very important factor to improve the image quality in Positron Emission Tomography (PET) reconstruction. Excellent timing resolution enables direct annihilation photons localization using ToF information without tomographic reconstruction [6]. Recent studies in timing resolution are mainly about fast scintillators and Cerenkov emitting materials by using low-noise readout electronics [7, 8]. In the development of ToF-PET scanners, timing resolution of the PET detector is still being improved [9–11]. At present, the best Coincidence Timing Resolution (CTR) is about 30 ps Full Width at Half Maximum (FWHM) when utilizing a Cerenkov detector coupled with a Microchannel Plate Photomultiplier (MCP-PMT) [12]. In high energy physics, the ToF performance also dominates particle identifications in the spectrometer. In the BESIII, the plastic scintillators serve as the ToF system [13]. In the AMS-100 proposal, a ToF system is constructed with small scintillator rods read out by SiPM arrays at both ends. The proposed ToF system have achieved a time resolution of 50 ps for particle mass reconstructions [14].

For practical application, it is very challenging to enable the ToF capability of SiPM arrays or strip SiPMs that comprise so many detection channels. For signal processing of the SiPM array, one approach is to use multiplexing network circuit, which consists of resistor-based [15] and capacitor-based [16] schemes. The number of readout electronics channels can be reduced in these schemes. However, contributions of extra RC constant in these schemes will lower rise time of the SiPM signals, thus degrading the ToF performance. For readout electronics of the SiPM detectors, Time-Over-Threshold (TOT) is a commonly used method [17]. In the TOT method, the timing information can be uniquely determined by measuring the leading edge of the SiPM signals. Due to non-linearity between the deposited energy and the digital pulse, it needs additional corrections in this method. Another approach is to apply compact Application-Specific Integrated Circuit (ASIC) design. Compared with regular electronics readout schemes based on discrete components, the ASIC designs are able to maximize ToF capability such as NINO ASIC [18], HRFlexToT [19].

In nuclear medicine, the small-size SiPMs, such as those with 3 × 3 mm² and 2 × 2 mm² active area, have been well studied [20–22]. In high energy physics, however, the large-size SiPMs, such as 6 × 6 mm², are more desirable, such as Super Tau-Charm Facility (STCF) [23] to be constructed in China. In this paper, the timing performance of radiation detector based on large size SiPM is evaluated. The evaluation is mainly focused on SiPMs with fast output terminal. For the purpose of practical application, two kinds of amplification circuits were designed to amplify the fast timing signal.

2 Materials

In the evaluation of timing performance, two identical one-to-one coupling detectors were mounted. Each one had a LYSO crystal bar with a size of 3 × 3 × 10 mm³ and a SensL J-series SiPM [24] evaluation board (MicroFJ−SMTPA−60,035) [25]. The evaluation board provided a standard output for energy and a fast output for timing pickoff. The crystal was wrapped with Teflon (PTEE) tapes except for the exit face and coupled to the SiPM window by use of silicon grease. For the signal observation, an adapter board was designed and assembled under the evaluation board. The detectors were mechanically fixed using a designed 3-D printed holder. The SiPM has an active area of 6.07 × 6.07 mm² with 35 $\mu$ m microcell. The breakdown voltage of the SiPM is 24 V. To obtain best timing performance, the CTR values of both detectors at different bias voltages were measured.

3 Methods

This work mainly focuses on readout method for best timing performance. The fast output of the SensL J-series SiPM has a sharp leading edge, giving rise to a better timing performance compared with the standard output. The evaluation setups described in the following sections have been applied for timing pickoff.

3.1 RF transformer based output

The CTR measurement setup is illustrated in Figure 1. In this system, the timing pickoff was based on the reference readout circuit. The RF-transformer was applied to decrease the impact of the parasitic and passive capacitances of the SiPM detector, which facilitated the extraction of the fast signal components [26, 27]. The adapter board was designed to directly connect the fast output signal to a 50 Ω terminated oscilloscope for signal observation. The fast output from the SiPM was used for timing pickoff. The standard output for the SiPM was also sent to the oscilloscope with a 50 Ω termination, which was for energy calculation. The oscilloscope features a bandwidth of 1 GHz and a sampling rate of 5 GS/s. In the experiment, a Na-22 point source was applied to irradiate the two detectors. In the oscilloscope, a logic trigger mode was applied. In the logic trigger, four waveforms including two energy signals and two timing signals were simultaneously recoded once when both of the energy signals passed the thresholds. Note that the timing signals were not used for event trigger because of their small amplitudes. The trigger thresholds for both energy signals were identical. The bias voltages for both SiPMs ranged from 27 V to 32 V, and the trigger thresholds were 50 mV, 100 mV, 150 mV, 200 mV, 250 mV and 300 mV respectively. For each bias voltage, we collected 20,000 events.

FIGURE 1. Experimental setup (A) and photograph (B) for timing evaluation.

3.2 Readout circuit based on common-emitter amplifier

Considering that the fast output has a small output capacitance of 160 pF [24], it will result in a very fast leading edge. However, the amplitude of the timing signal from a gamma event is less than 15 mV. In practical application, it is not suitable for the timing pickoff. In second readout scheme, the timing signal was amplified by a two-stage amplification circuit. The circuit consisted of two common-emitter amplifiers generated by discrete components, as illustrated in Figure 2. A low noise silicon bipolar RF transistor (BFR106, Infineon) with a maximum 5 GHz frequency was applied to create the CEA circuit. R1 and R2 were used to provide a bias for the transistor, establishing a proper operating Q-point. A bypass capacitor C2 was to enhance the AC gain at high frequency. R3 was for thermal stability by negative feedback. The gain of the CEA was determined by R4. In between, the two CEAs were coupled with a 100 nF capacitor. The fast output of the SiPM was directly fed to input of the CEA circuit with a 50 Ω termination. The output of two-stage CEA circuit was directly sent to the oscilloscope with a 50 Ω termination. Overall, the gain of the two-stage CEA circuit was about 70.

FIGURE 2. Schematic (A) and photograph (B) of two-stage common-emitter amplification circuit.

FIGURE 3. Schematic (A) and photograph (B) of two-stage amplification circuit based on minicircuit RF amplifier.

3.3 Readout circuit based on RF amplifier

The last readout scheme is similar with the CEA. The timing signal was amplified by two high speed Minicircuits MAR-6+ RF amplifiers, as illustrated in Figure 3. Both amplifiers were cascaded for signal amplification before timing pickoff. The commercial amplifier was a wideband amplifier offering high dynamic range from DC to 2 GHz。The cascaded structure provided a very high gain for small signals. The RF amplifier needed an external bias resistor. In our experimental setup, the resistor was 150 Ω. After measurements, the overall gain of the two-stage RF amplification circuit was about 60.

3.4 Signal processing

Figure 4 shows a typical example of the acquired waveforms by the oscilloscope, including the energy and timing signals of the two detectors (Det1 and Det2). Here we show the timing signal output based on the RF amplifier. The waveforms on the oscilloscope were sent to a PC via USB interface and recorded via MATLAB software with TekVISA driver. Totally, three kinds of experiments based on RF-transformer, CEA circuit and RF amplifier were carried out. For each experiment, totally 20,000 events were recorded.

FIGURE 4. Typical waveforms of a coincidence event captured by the oscilloscope.

With the recorded waveforms, we firstly made the baseline correction, that was, subtracting the mean baseline value at first 200 sample point. The energy was the integration of all data point from energy signals. In the meantime, we used the timing signal to determine the CTR. Picking off the leading edge of the timing waveform, we performed a cubic spline interpolation at the rising edge of the timing signals in 10 ps step. In order to obtain a better timing performance, different trigger thresholds from 1 mV to 10 mV in 1 mV step were used. For each threshold, we acquired a histogram showing the distribution of time difference between Det1 and Det2 signals for all back-to-back photon events. Additionally, a 400–600 keV energy window was applied to each detector to select 511 keV gamma photons. In the timing distribution, we directly applied the FWHM of the histogram to represent the CTR.

4 Results

The timing signal waveforms of the three readout schemes are shown in Figure 5. The recommended readout based on RF transformer resulted in very small signals. The signals can be amplified by the commercial RF amplifiers. For the CEA-based schemes, the waveform was a little distorted. However, it ensured the high gain in high frequency. Therefore, the optimum timing can be picked off.

FIGURE 5. Typical timing waveforms in three experiments based on transformer-based readout (A), CEA-based readout and RF amplifier-based readout (B) schemes.

Figure 6 shows energy spectra for Det1 and Det2 respectively. For the energy resolution calculations, a referenced Cs-137 point source was applied to irradiate the two detectors. Compared with the energy spectrum irradiated by the Cs-137 source, the evaluated energy resolutions were 9.41% and 9.38% for the two detectors respectively.

FIGURE 6. Typical energy spectra for Det1 and Det2 respectively.

Figure 7 shows the CTR distributions when scanning the trigger threshold at different bias voltages. Under each bias voltage, it should be noted that the CTR degrades with the increased thresholds. From these timing distributions, one biased at 32 V showed best timing performance. These experiments are repeated three times. The CTR value of both coincidence detectors was given by the mean ± standard deviation of the three measurements. The best CTR value at 32 V based on the RF transformer-based scheme was 228.3 ± 1.4 ps. Figure 8 shows a typical CTR spectrum at 2 mV trigger threshold when biased at 32 V.

FIGURE 7. FWHM CTR results by scanning trigger thresholds when the SiPMs are biased at different voltages.

FIGURE 8. The CTR spectrum at 2 mV threshold based on the RF transformer-based scheme.

Since the best timing result was obtained when biased at 32 V, the following experiments based on the CEA-based and RF amplifier-based readout were also carried out at 32 V. Figure 9 was a plot of the CTR results when scanning different trigger thresholds. The CTR distribution for the CEA-based readout scheme was relatively uniform. The best FWHM CTR values were 235.4 ± 1.1 ps at 24 mV threshold for the CEA-based readout and 231.1 ± 1.5 ps at 14 mV threshold for the RF amplifier-based readout, respectively. Figure 10 shows typical CTR spectra for both readout schemes.

FIGURE 9. CTR distributions at different trigger thresholds based on CEA-based and RF amplifier-based readout schemes.

FIGURE 10. Best CTR spectra based on CEA-based (A) and RF amplifier-based (B) readout schemes.

4 Discussions

In this work, we have evaluated and compared the CTR values of three different timing readout schemes for two 3 × 3 × 10 mm³ LYSO crystals coupled with a 6 × 6 mm² SiPM detector respectively. The timing readout configurations were based on RF transformer-based, CEA-based and RF amplifier-based circuits respectively. From our measurements, the optimum timing can be obtained at 32 V bias voltage. Scanning the trigger thresholds, the best FWHM CTR values were 228.3 ± 1.4 ps, 235.4 ± 1.1 ps and 231.1 ± 1.5 ps for the RF transformer-based, the CEA-based and RF amplifier-based readout respectively. Although we can obtain a good CTR performance under the RF transformer-based readout, it is not practical for multi-channel, high-density readout electronics system where one can not set 1 mV trigger threshold because it is very sensitive to the noise level. The best way is to amplify the timing signals. From our studies, the amplified timing signal retained the fast leading edge in the CEA-based and RF amplifier-based schemes. In comparison, the CTR distribution in the CEA-based scheme is more uniform than that in the RF amplifier-based scheme. In practical multi-channel readout electronics system, it is better to set a unified trigger threshold under the condition of the optimum timing performance.

In the three readout schemes, the RF transformer is a passive component. The power consumption for the RF transformer-based circuit can be negligible. In normal operation, a 6-V power supply (223A-30-3, Tektronix) was applied on the CEA-based and RF amplifier-based circuits. 18 mA and 16 mA current were observed for the CEA-based and RF amplifier-based circuits respectively. Although the power consumptions were a little large, the number of timing channels is generally far less than that of the signal readout electronics. It is acceptable for practical SiPM-based detector system. One needs to make a trade-off when choosing the number of the timing signals in whole readout system.

There have four factors that affect the total timing performance in the experiment in terms of intrinsic timing performance of the instrumentation, scintillation statistics of the scintillator, SiPM technology itself and signal readout method. Firstly, from the perspective of the instrumentation, the oscilloscope we used had a bandwidth of 1 GHz and a sampling rate of 5 GS/s. Splitting the fast output of the SiPM into two identical signals, the intrinsic timing performance of the oscilloscope was measured, as shown in Figure 11. The intrinsic CTR value was 116.0 ps. Deconvolving the contribution of intrinsic timing performance of the instrumentation to the CTR, the best FWHM CTR value for the RF transformer-based readout method became 196.6 ps, which was a little better than the results in [28] where it showed a 203 ps CTR value when two 3 × 3 × 20 mm³ LYSO crystals were coupled with 3 × 3 mm² SiPMs. Although many ToF measurement experiments with multiplexed SiPM [29–32] had proven the importance of reducing capacitance in the readout electronics path, and the output capacitance (160 pF) of the 6 × 6 mm² SiPM at fast output end is four times larger than that (40 pF) of the 3 × 3 mm² SiPM, the contribution to rise time of SiPM output signal should be negligible because the sense resistor R_S of the 6 × 6 mm² SiPM is five times smaller than that of the 3 × 3 mm² SiPM.

FIGURE 11. Intrinsic CTR spectrum of the oscilloscope.

From the perspective of scintillation statistics in the crystal, the CTR is related to the scintillation decay time (τ_d), the scintillation rise time (τ_r) and the number of detected photons n), i.e. $CTR\propto \sqrt{{\tau }_{\mathrm{d}}\times {\tau }_{\mathrm{r}}/n}$ [8, 33–35]. Lowering τ_d and τ_r can extremely increase the CTR performance. In practical applications, imperfect light collection efficiency of initially emitted scintillation lights (n) limits the achievable timing performance. However, Enhanced Specular Reflectors (ESRs) can increase light collection efficiency in the end-coupled configuration, both energy resolution and timing resolution can be further improved in comparison with the Teflon-based wrapper [36]. Additionally, the side-coupled configuration [37] between the scintillator and the SiPM provides not only excellent light collection efficiency but also low photon transit time jitter. In this kind of coupling configuration, the lateral surface of the crystal is directly coupled to the multiple SiPMs. About 100 ps CTR can be obtained when using an array of 4 × 4 mm² SiPMs [38]. With the ESR wrapper, the timing performance can be further enhanced for our experiment.

From the perspective of SiPM technology itself, the novel Near Ultraviolet (NUV) SiPM from FBK provides a lower-afterpulse (Low-AP) substrate and a lower electric field (Low-F) feature in comparison with the SiPMs from Hamamatsu and SensL. The low-AP substrate can achieve an improvement of the CTR of about 30 ps [39]. Coupling a 3 × 3 × 5 m³ LYSO crystal to a 4 × 4 m² NUV SiPM, an excellent CTR value of 130 ps can be reached.

Last but not least, from the perspective of the readout method, the time resolution in the electronics path can be given by ${\sigma }_{\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{i}\mathrm{n}\mathrm{g}}=\frac{{\sigma }_n}{\left(dV/dt\right)}$ where ${\sigma }_{\mathrm{n}}$ and dV/dt are the electronics noise and slew rate of the signals respectively. Electronics noise mainly originates from the thermal noise of the discrete components and 1/f noise from the parasitic capacitance in the electronics circuit [40], which has clear physics limits and is hard to be considerably improved. By increasing the signal amplitude V and decreasing the rise time (τ_r) of the signal, the slew rate can be enhanced. Hence, the electronics timing jitter can be further improved. In this work, we utilized the very fast signal of the SiPM, i.e., the fast output. In order to achieve a high slew rate, two different high-frequency amplification circuits were designed and evaluated. The CEA-based amplification circuit showed a better uniformity with increasing the trigger thresholds. However, the electronics noise (${\sigma }_{\mathrm{n}}$) was also amplified in the single-ended readout method, which had a limit for the best time resolution. More efficient way should be the differential signal readout prior to the high-frequency amplification, which doubled amplitude of the signal and suppressed common noise [8]. With the differential readout method, a 98 ± 2 FWHM CTR can be obtained when a long LSO:Ce:0.2%Ca 2 × 2 × 20 mm³ crystal was coupled to a 4 × 4 mm² FBK NUV-HD SiPMs.

For future work, we plan to glue ESRs outside the LYSO crystal except for the exit face. It can raise the number of scintillation lights that the SiPMs receive. It should contribute to the leading edge of the timing signals. In addition, we will construct a ToF PET detector including a SiPM array and a LYSO array. Multiple 6 × 6 mm² SiPMs will be designed to form the SiPM array, which is used to read out the 2 × 2 mm² LYSO array. The LYSO array consists of many 2 × 2 × 10 mm³ LYSO bars. The LYSO array is directly 9-to-1 coupled with the SiPM array. Consequently, we will further improve the CTR performance of the ToF PET detectors constructed with 6 × 6 mm² SiPM. The other experiment is for detector verification of the STCF experiment [41]. In the STCF, plastic scintillators work as ToF detectors. We plan to apply 6 × 6 mm² SiPM to study the timing performance of the scintillation detector with a dual-end readout.

5 Conclusion

In summary, we evaluated the timing performance of both detectors based on large-size SiPMs using three different readout circuit. The best CTR value can be obtained at 32 V bias voltage when using RF transformer-based readout scheme for timing signal processing. The CTR distribution for the CEA-based readout scheme was relatively uniform. For practical application, the amplified timing signal based on the CEA circuit is more desirable because it is more feasible for trigger threshold selection in multi-channel readout electronics system.

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.

Author contributions

KH: Conceptualization, methodology, writing, supervision. CL: Software, visualization, validation. YL: Data curation, investigation. JX: Investigation, resources, visualization. SX: Conceptualization, methodology, writing, supervision.

Funding

This work was supported by the Natural Science Foundation of Shandong Province (No. ZR2022QA039) and the Program of Qilu Young Scholars of Shandong University.

Conflict of interest

JX was employed by Anhui Specreation Instrument Technology Co., Ltd.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References