A Lightweight and High Yield Complementary Metal-Oxide Semiconductor True Random Number Generator with Lightweight Photon Post-Processing

Abstract

This paper introduces a novel TRNG architecture that employs a wave converter to generate random outputs from the jitter noise in a customized ring oscillator (RO). Using a current-starved inverter, the proposed RO offers the option of operating three different oscillation frequencies from a single oscillator. To assess its performance, the core TRNG proposed in this work was designed with multiple samples, employing various transistor sizes for 28 nm CMOS processes. The measurements show that only a small number of measured TRNG samples passed the randomness NIST SP 800-22 tests, which is a common problem, not only with the proposed TRNG but also with other TRNG structures. To solve this issue, a lightweight post-processing algorithm using the Photon hash function was newly applied to the proposed TRNGs topology. The lightweight Photon hash function-based post-processing was implemented with the proposed TRNG topology in a 28 nm CMOS process. The design occupies 16,498 µm2, with a throughput of 0.0142 Mbps and power consumption of 31.12 mW. Measurements showed significant improvement, with a 50% increase in chips passing the NIST SP 800-22 tests. Compared with the conventional DRBG post-processing method, the proposed lightweight Photon post-processing reduces area occupation by five times and power consumption by 65%.