Stochastic Quantization Using Magnetic Tunnel Junction Devices: A Simulation Study

In this paper, we investigate a magnetic tunneling junction (MTJ) device as a stochastic quantizer. Thermal noise in an MTJ is one of the parameters in determining its output, 0 or 1. This thermal noise, which causes random switching in a magnet, can be successfully exploited to implement high-performance stochastic quantization on an analog input (the function relating MTJ's output state to its analog input is defined as an MTJ quantization function). Subsequently, an optimized signal-to-noise ratio (SNR) is found to maximize the information throughput. The best output to input correlation is achieved when the input signal's amplitude is 1.5 times the standard deviation of the MTJ's quantization function. We also demonstrate that an MTJ-based quantizer offers the highest performance when the information to be quantized is normally distributed in the input signal. Simulations to quantize images containing information, such as light intensity and pattern, also confirmed the highest information throughput when this optimum SNR is utilized.