High-Speed Discrete Gaussian Sampler With Heterodyne Chaotic Laser Inputs

Considering the limited throughput of a true random number generator (TRNG), the time independence of speed in a real discrete Gaussian sampler design was usually sacrificed. By utilizing the broadband chaos from an optically injected semiconductor laser with heterodyning, we propose a laser-field-programmable gate array system which generates discrete Gaussian numbers with a time-invariant throughput of 35.7 and 7.8 MSPS on Zedboard for two ring learning with errors schemes. Applying the proposed compressed cumulative distribution table method, the throughput of the proposed architectures reaches up to 40 and 8.3 MSPS, which is the fastest amongst the existing architectures based on TRNG inputs.