Potential of Ultralow-Power Cellular Neural Image Processing With Si/Ge Tunnel FET

This letter studies the application of tunnel FET (TFET) for ultralow power image processing through cellular neural network (CNN). Through steeper switching slope, and thereby higher g_m/I_DS, a TFET-based CNN synapse can deliver the same performance as MOSFET even with a lower power. A TFET-based synapse is also scalable to the ultralow power regime; hence, by comprising more cells than MOSFET at the same power, TFET can reduce the multiplexing overheads in image processing with CNN. Utilizing unique properties of TFET, we show an improved performance for low power image processing using TFET.