Entropy extraction in metastability-based TRNG

Abstract

True Random Number Generators (TRNG) implemented in deep sub micron (DSM) technologies become biased in bit generation due to process variations and fluctuations in operating conditions. A variety of mechanisms ranging from analog and digital circuit techniques to algorithmic post-processing can be employed to remove bias. In this work we compare the effectiveness of digital post-processing using the XOR function and Von Neumann Corrector with circuit calibration technique for a meta-stability based reference TRNG design. The energy consumption per bit is used as the metric for comparison of the different techniques. The results indicate that the calibration technique is effective for 12% larger process variation than the XOR function and extracts entropy comparable to the Von Neumann Corrector at 56% lesser energy/bit. The analysis thereby demonstrates that circuit calibration provides an efficient tradeoff between entropy and energy/bit for removing bias in lightweight TRNG.