An FPGA-actuated MTJ True Random Number Generator with Probability Modulation

Random number generation is important for a variety of applications, such as Monte Carlo simulations that require large quantities of random bits with precise probability. Recently, we have shown using perpendicular nanopillar magnetic tunnel junctions (pMTJs) with Field Programmable Gate Arrays (FPGAs), true random number generation (TRNG) can be achieved at 5MHz using one XOR operation done in postprocessing of the data. One challenge of high-volume TRNG with pMTJs is an observed drift in probability bias across several hours of data collection. To compensate for this and improve the precision of our bitstream, we introduce XOR and a feedback loop on the FPGA board to modulate the write voltage over the course of data collection. With feedback and XOR, we generated 10^12 random bits. Our data collected with feedback passes the NIST statistical tests for randomness, and we perform further FFT and Chi-squared testing to check for correlations between bits. FPGA pulse modulation also allows us to quickly tune the switching probability of the junction with precision within ±0.5 percent before XOR. We further demonstrate that pulse modulation can consistently create bitstreams with an extremely low switching probability of 1 in 10 million, useful for simulations of rare events.