Citation: Wang, Q.; Pan, G.; Jiang, Y.
An Ultra-Low Power Threshold
Voltage Variable Artificial Retina
Neuron. Electronics 2022, 11, 365.
https://doi.org/10.3390/
electronics11030365
Academic Editors: Luis
Hernández-Callejo,
Sergio Nesmachnow and
Sara Gallardo Saavedra
Received: 3 December 2021
Accepted: 17 January 2022
Published: 25 January 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
An Ultra-Low Power Threshold Voltage Variable Artificial
Retina Neuron
Qiguang Wang , Guangchen Pan and Yanfeng Jiang *
Department of Microelectronics, School of Internet of Things Engineering, Jiangnan University,
Wuxi 214122, China; wqg1997@outlook.com (Q.W.); 6171916008@stu.jiangnan.edu.cn (G.P.)
* Correspondence: jiangyf@jiangnan.edu.cn
Abstract:
An artificial retina neuron is proposed and implemented by CMOS technology. It can be
used as an image sensor in the Artificial Intelligence (AI) field with the benefit of ultra-low power
consumption. The artificial neuron can generate signals in spike shape with pre-designed frequencies
under different light intensities. The power consumption is reduced by removing the film capacitor.
The comparator is adopted to improve the stability of the circuit, and the power consumption of
the comparator is optimized. The power consumption of the proposed CMOS neuron circuit is
suppressed. The ultra-low-power artificial neuron with variable threshold shows a frequency range
of 0.8–80 kHz when the input current is varied from 1 pA to 150 pA. The minimum DC power is
35 pW when the input current is 5 pA. The minimum energy of the neuron is 3 fJ. The proposed
ultra-low-power artificial retina neuron has wide potential applications in the field of AI.
Keywords: artificial retina neuron; spike; CMOS; Axon-Hillock circuit; ultra-low power
1. Introduction
Compared with the traditional Von Neumann architecture computer, the human
brain shows stronger associative memory and thinking in images. It also has a greater
potential ability than existing computers in solving complex problems such as function
approximation, complex classification and clustering [
1
]. Moreover, compared with current
existing computers, the human brain is not only more powerful, but it is also smaller and
consumes less power. Therefore, the realization of the artificial neural network (ANN) to
mimic the human brain intelligence has become a hot subject for research recently [
2
]. The
human brain is composed of many complex interconnected neurons, and the information
interaction between neurons is what forms the thinking ability. Designing a reasonable and
efficient neuron unit is an important point for imitating the thinking ability of the human
brain [3,4].
The first-generation ANN consists of threshold gates [
5
]. Its principle is using the
threshold gate to judge the output result by counting the binary sum of the inputs. If the
inputs’ summation is larger than the threshold value, it is considered to be high level (1);
otherwise it is low level (0). It can be seen that the function of the first-generation ANN is
very limited and that it can only process binary data. This is still far removed from the real
biological neuron. The second-generation ANN is based on the encoding of the frequencies
of the neuron pulses [
6
]. By stacking multiple layers of the neurons and applying a back
propagation algorithm, a neural network can be constructed, which is known as deep
learning neural network. This network is widely used in machine learning, brain-machine
interfaces, image sensors, etc. [
7
]. Although the second-generation ANN is powerful, its
energy consumption and efficiency are still not good enough compared with the biological
network. Moreover, there is a big difference in the process of communicating with the
spikes of neurons in the human brain in the underlying logic. Faced with these problems,
the third-generation ANN has been proposed recently. Its neuron units are much closer
to biological neurons, in that they can communicate with each other using spike signals.
Electronics 2022, 11, 365. https://doi.org/10.3390/electronics11030365 https://www.mdpi.com/journal/electronics
