Abstract
A true random number generator (TRNG) is a fundamental component of hardware security. Latch-based TRNGs (LTRNGs), with their simple circuitry, are well-suited for energy-constrained IoT devices. However, a mismatch between the two inverters in the latch can result in little or no entropy output. This work presents chip measurement results demonstrating the use of negative bias temperature instability (NBTI) to mitigate mismatch in LTRNG. We propose a sequence to quantify the initial mismatch and selectively applying NBTI stress to PMOS device with higher conductivity for the first time. Measurement results in a 130nm CMOS process show that after 1000 minutes of NBTI stress at 3.0 V/120°C, the mismatches are reduced to within the overcompensated target of ±20 mV, and the output entropy improves from zero to over 0.85. Mismatch recovery at 20°C is negligible. Recovery at 120°C is accelerated, with mismatches still remaining below 50% of their initial values, and the output entropy remaining at 0.47 and 0.80 for the measured two chips. The randomness of post-processed data is validated using NIST SP 800-22 and SP 800-90B.
Funding Information
  • Grant (JSPS 21K04201)
  • Grant (JSPS 24K07596)
  • Grant (JPMJMS2214-5)
  • Grant (JPMJCR24U1)
  • Grant (JPMJCS24K9)
  • Grant (JPNP14004)
  • Grant (22H03557)
  • Grant (JPJ012368C06201)
  • Activities of VLSI Design and Education Center (VDEC), The University of Tokyo, in Collaboration with Cadence Design Systems and NIHON SYNOPSYS G.K