Design of an efficient VCMA controlled spintronic random number generator

Abstract

Random number generators play an essential role in stochastic circuits, neuromorphic computing and cryptographic systems. Existing CMOS based true random number generators (TRNGs) incur a significant performance, power and area overheads. The systems based on magnetic tunnel junctions (MTJs) offer favorable features such as high resilience, low energy consumption, good performance, and less complexity. Compared with the conventional MTJ switching approaches, voltage controlled magnetic anisotropy switching (VCMA) features low switching current, minimal control circuitry and calibration process, and insensitivity to the write pulse duration. In this paper an efficient design for TRNG based on VCMA using MTJ is presented. The proposed magnetic TRNG consists of a standard two-terminal MTJ connected to the drain terminal of a PMOS transistor which controls the pulse width of the applied bias voltage. MTJ is used as a noise source device and its precessional switching characteristics are exploited to design the TRNG. The resistance of the device fluctuates while operating in precessional switching regime and the magnetization state is selected randomly via thermal fluctuations. The voltage developed across the MTJ corresponding to its random state results into a random bit. A 64-bit parallel random number generator has been implemented and evaluated using Spice simulating tool (with 45 nm CMOS technology node). The proposed circuit has an advantage of less circuit complexity and high energy efficiency. It achieves a high throughput of 64Gbps with an energy consumption of 5.6 fJ/bit.