A coupled oscillator based ising chip as a physical entropy source for hybrid true random number generation

This work presents a dual-purpose Coupled Oscillator Based Ising (COBI) chip that serves as a combinatorial optimization problem (COP) solver and as a physical entropy source for a hybrid True Random Number Generator (TRNG). By disabling coupling weights, oscillator phases drift randomly, creating a physical entropy source for various probabilistic compute applications. To meet the statistical randomness and throughput requirements, we adopted a hybrid approach. Post-layout simulation and static timing analysis project an expanded-output throughput of 351 Mb/s at 0.9 V, corresponding to an estimated energy of 71.2 pJ/bit when using the measured COBI-chip power (Geffe-block power excluded). We report NIST SP 800-22 results for the output obtained by applying von Neumann correction in software after data collection, while the unconditioned raw output was evaluated using NIST SP 800-90B. The proposed TRNG approach was demonstrated in graph decomposers, enabling a real-world probabilistic computing application without the need for a separate physical entropy-source macro or a software algorithm generating pseudo random numbers.