Comparative Study of ASE-ASE and ASE-Laser Based Quantum Random Number Generators

Quantum random number generators (QRNG) based on amplified spontaneous emission (ASE) are known to achieve multi-gigabit-per-second rates through high-speed sampling of intensity and phase fluctuations. Two popular implementations of these QRNGs are based on balanced detection of ASE-ASE and ASE-Laser beating. The high QRNG rates reported so far however come at the price of increased system complexity, making it a clear deterrent for widespread use. In this paper, we perform a direct one-to-one comparison study between ASE-ASE and ASE-Laser-based QRNGs using detailed theoretical modeling and supporting experiments, with the goal of achieving 1 Gbps QRNG using simple, low-bandwidth photodetectors and low sampling rate digitizers. We develop, for the first time to the best of out knowledge, closed-form analytical expressions to describe the signal variance as a function of realistic, non-ideal fiber-optic optical components, the signal correlations as a function of digitizer sampling rate, and minimum entropy as a function of signal variance. We observe a trade-off between the correlations in the detected signals and the sampling rate used, thereby obtaining an optimum sampling rate and hence the digitizer specification for extracting the best randomness. We also observe the standard deviation in the signal distribution is higher for ASE-Laser QRNG in comparison to the ASE-ASE implementation. Using off-the-shelf telecom fiber components, balanced detectors with bandwidth as low as 100 MHz, and an 8-bit digitizer operating at a sampling rate of 100 Megasamples-per-second (MSPS), we achieve QRNG rates of 546.03 Mbps and 507.94 Mbps for the ASE-ASE and ASE-Laser beating implementations respectively. The ASE-ASE QRNG rates are subsequently doubled to more than 1 Gbps using polarization multiplexing and bit-interleaving.