True random number generation based on arrival time and position of dark counts in a multichannel silicon photomultiplier

Abstract

Silicon photomultiplier (SiPM) arrays normally contain tens of photon detection channels, with each channel consisting of several thousand microcells and each microcell being a normal single-photon avalanche diode with a quenching resistor. The dark counts of SiPM arrays have independent spatial and temporal randomness, which will be used for true random number generation in this paper. Since the arrival times of the high dark count rate in each channel are measured with a fast, high precision time-to-digital converter, along with the number of channels, the entropy can be extracted with very high efficiency. The bias in the raw data due to the exponential distribution of the arrival time differences between successive dark counts is removed using the transformation of the cumulative distribution function. Except for the preamplifiers for a signal readout from a 4-channel SiPM, all electronics components in our prototype are implemented inside of one chip of a field programmable gate array. The prototype has a 63.54 Mbps generation throughput, and the statistical quality of the generated random numbers is evaluated. Since the property of the dark count is compatible with single-photons from laser beams, one can easily increase the generation bitrate by either adding more SiPM channels or irradiating the SiPM with extra laser beams.