A 30fJ/b Current-Biased Inverter Based RO TRNG with High Temperature and Supply Voltage Stabilities

In this paper, we present an ultra-low power true random number generator (TRNG) based on ring oscillator (RO) with current-biased inverters. Random numbers are extracted by symmetrically designed arbiter depending on the jitter-noise-caused phase difference of a pair of ROs. Using a bias transistor shared by 16 inverters (8 for each RO), we can obtain a virtual power supply lower than V DD (denoted as V V DD ) and a subthreshold bias current to increase the oscillation frequency while ensuring high randomness. In addition, power consumption is significantly lowered by using V V DD to reduce the amplitude of oscillation. Moreover, V V DD can be designed with high temperature and supply voltage stability by optimizing the transistor sizes, which allows the proposed implementation to generate random numbers and pass both NIST and auto-correlation test suites over wide ranges of supply voltage (1.0V∼1.4V) and temperature (0°C∼120°C). The proposed TRNG is implemented using standard 65nm CMOS process, with the bit generation rate and energy efficiency reported to be 169Mbps and 30fJ/bit, respectively (1.2V and 27°C).