A subthreshold mode programmable implementation of the Gaussian function for RBF neural networks applications

A new CMOS analog programmable Gaussian function for RBF neural networks, based on current mode operations and using devices operating in the subthreshold region is proposed. In order to maintain low voltage operation, the proposed circuit uses the gate and backgate terminals of the devices. The power supply is as low as 1.2 V. Analysis and simulations have shown that the new structure has a power consumption of less than 200 μW. The proposed scheme produces an output which is related with a Gaussian input current without any nonlinear approximations. Although operating at subthreshold current levels, reasonable speed can be obtained since voltage swings are in the range of a few V_T.