A High-Throughput and Configurable TRNG Based on Dual-Mode Memristor for Stochastic Computing.

In this work, a novel configurable true random number generator (TRNG) based on VO 2 /HfO 2 stacked dual-mode memristor is demonstrated. The $\mathbf{V}_{\mathbf{th}}$ fluctuation in the volatile IMT mode of the device is exploited as the entropy source for the TRNG, ensuring the generation of high-quality random numbers. Additionally, the non-volatile RRAM mode allows for in situ control of the TRNG's on/off state. The proposed TRNG achieves a high throughput of 5 Mb/s, enabling seamless integration into a stochastic computing (SC) system for image edge detection. The experimental result showcases superior performance with an average error rate as low as 0.73% with a bitstream length of 256.