The Digital-Analog SiPM Approach: a Story of Electronic and Excess Noise

Modern PET detectors rely on fast and accurate timing performances in order to provide improved lesion detectability and reduced patient scan times. Optimizing the coincidence timing resolution (CTR) is an important facet that helps achieve high PET detector performance. On the way of improving the CTR to sub-100 ps, and ultimately to 10 ps, research on all detector components is necessary, including the scintillator, photodetector and electronic readout. The digital SiPM, where each single photon avalanche diode (SPAD) is connected to its own readout, has gained a lot of attention as the ultimate photodetector. However, extended power consumption and the enormous number of channels makes the realization challenging. Hence, in this contribution we investigate a semidigital approach, realized with the pixelation of an analog SiPM. We tested a single SiPM from HPK (S13360-6075V) compared to a 2x2 SiPM array (S13361-3075N-02) and a 3x3 SiPM array (S13361-2075N-03) with a 6x6x3 mm
³
crystal and a 6x6x20 mm
³
crystal, readout by our in-house developed acquisition system employing NINO as front-end ASIC. Preliminary results show that the single SiPM array routinely performed better than the pixelated arrays while reading each of the individual SiPMs independently, correcting for time walk and properly combining the time-stamps. A minimum CTR value of 117 ± 53ps wascrystal achieved using the single SiPM array with the 6x6x3 mm for the NINO electronics readout. Using a HF bipolar transistor readout, a similar trend was observed, with a minimum CTR of 116 ± 5 ps for the single array and the 6x6x3 mm3crystal. A-priori this result is surprising because, due to the lower capacitance of the smaller SiPMs, electronic noise should be reduced leading to better timing. The most probable explanation for this counter-intuitive result is the 4 or 9 times lower number of scintillation photons generating each SiPM signal, which together with correlated noise sources (optical cross-talk) leads to excess noise, especially in the applied leading edge time discrimination.