DocumentCode :
2030088
Title :
A mixed signal multiplier using A2 binary representation dedicated to neural networks applications
Author :
Boukadida, Hatem ; Hassen, Néjib ; Gafsi, Zied ; Besbes, Kamel
Author_Institution :
Lab. de Microelectron. et Instrum. (μEi), Univ. Pierre et Marie Curie, Paris, France
fYear :
2011
fDate :
6-8 April 2011
Firstpage :
1
Lastpage :
4
Abstract :
The main goal of our work is to investigate the delay performances of redundant binary arithmetic operators using different CMOS logic and mixed signals styles. This paper presents a novel technique for A2 redundant binary multiplication in mixed-signal circuits especially for neural networks applications. The proposed multiplier consists of three cascaded blocks: Two 3-bit Arithmetic Multiplier Digital to Analog Converter (AMDAC) cells, and one 6-bit Flash ADC. Our technique indicates that this approach significantly reduces the silicon area occupied by such multipliers compared to the classical scheme using combinational multipliers. The circuit being studied is optimized for speed efficiency at 0.35 μm CMOS process.
Keywords :
CMOS logic circuits; mixed analogue-digital integrated circuits; neural nets; voltage multipliers; CMOS logic; arithmetic multiplier digital to analog converter cells; binary representation; mixed signal circuits; mixed signal multiplier; neural networks applications; redundant binary arithmetic operators; size 0.35 mum; Artificial neural networks; Biological neural networks; CMOS integrated circuits; Computers; Digital arithmetic; Neurons; System-on-a-chip; Computer arithmetic; Redundant binary representation; VLSI integration;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Design & Technology of Integrated Systems in Nanoscale Era (DTIS), 2011 6th International Conference on
Conference_Location :
Athens
Print_ISBN :
978-1-61284-899-0
Type :
conf
DOI :
10.1109/DTIS.2011.5941407
Filename :
5941407
Link To Document :
بازگشت