|
چکیده
|
The neuron is the most important part of a biological neural network, which is known as the core of neural networks. In most neural circuits, supply voltages are fixed due to the adjustment of parameter values of circuit elements during the design process. In this paper, we propose a new neural circuit that mimics the biological neuron behavior, including sodium and potassium channels and pumps, while operating independently of the supply voltage value. Due to their abundance in neural networks, power consumption and die area are the most important factors in the analog design of these circuits. The proposed circuit has low power consumption compared to similar CMOS circuits and has small capacitors to reduce die area. The proposed circuit operates reliably across a supply voltage range of 0.3–1.2 V without any change in firing frequency. To our knowledge, this is one of the first circuits capable of maintaining stable frequency over a wide range of supply voltage variation. Power consumption is 560 pJ spike−1 at a firing rate of 30 kHz with a supply voltage of 0.3 V. The circuit is simulated in Hspice software using 90 nm CMOS technology.
|