Analog-to-Digital Conversion using ANNs with Non-Linear Feedback

Analog-to-Digital conversion is a basic signal processing task that is needed at various places in the context of modern day mixed-signal systems like instrumentation & control systems, system-on-chip, etc. It is because of the fact that most real-world signals are analog in nature whereas most on-chip computation is digital. The technical literature is replete with electronic implementations of analog to digital converters including, but not limited to, Flash ADC, Successive Approximation ADC, and Sigma-Delta ADC. Given their promise of parallel processing and fast convergence, artificial neural networks have also been employed for analog-to-digital conversion. The first such attempt employed the Hopfield Neural Network and later several variants were introduced. However, most of the existing neural circuits for analog-to-digital conversion have an underlying similarity in the sense that they are derived from the Hopfield Network Architecture. A new scheme for analog-to-digital conversion utilizing a neural circuit for solving systems of linear equations is presented. The circuit employs (2n) opamps and (n+3) resistances for an n bit ADC.

Autorentext

Mohd. Samar Ansari received B. Tech., M. Tech. and Ph. D. degrees in Electronics Engineering from AMU, Aligarh, India, in 2001, 2007 and 2012 respectively. He is an Assistant Professor in Dept. of Electronics Engineering, AMU. He has research interests in Analog Signal Processing and Neural Networks, and has 54 Journal and Conference publications.