An Analog-Digital Neural Processor for Integrated Sensor Conditioning

In this work a CMOS neural processor is presented. Based on a mixed analog-digital architecture, the system processes data in analog current mode, using digital registers to store weights safely. The proposed processor consists of two main blocks: A mixed-signal four-quadrant multiplier and a class...

Full description

Saved in:
Bibliographic Details
Main Authors: Medrano, N., Sanz, M.T., Martinez, P.A., Celma, S., Zatorre, G.
Format: Conference Proceeding
Language:English
Subjects:
Analog-digital conversion
Artificial neural networks
Circuit simulation
CMOS process
Electronic circuits
Energy consumption
Giant magnetoresistance
Inverters
Magnetic sensors
Neurons
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In this work a CMOS neural processor is presented. Based on a mixed analog-digital architecture, the system processes data in analog current mode, using digital registers to store weights safely. The proposed processor consists of two main blocks: A mixed-signal four-quadrant multiplier and a class AB current conveyor that implements the non-linear output function. Achieved from the circuit-level simulation, the neuron system-level model is used to study the application of this architecture to tackle linearization of a magnetoresistive sensor. Simulation results show the efficiency of the new implementation.
DOI:10.1109/ICECS.2006.379941