A 256-point discrete Fourier transform processor fabricated in a 2 /spl mu/m NMOS technology

Abstract

The authors describe a 2-/spl mu/m NMOS single-chip processor which computes a 256-point Fourier transform in 6.5 ms by means of the DFT algorithm. This chip is primarily intended for speech processing applications such as voice recognizers, pitch extractors, and DFT vocoders. The DFT algorithm has been implemented wih two parallel multipliers, an accumulator/shifter and two ROMs with separate address computation units. The processor contains over 20,000 transistors and dissipates ~400 mW when operating at a clock frequency of 10 MHz. The die size is only 12.5 mm/SUP 2/ as a result of the use of a full-custom design method and the incorporation of low-resistance implanted As/SUP +/ undercrossings within the logic circuitry.

Keywords

This publication has 7 references indexed in Scilit:

A novel method for pitch extraction from speech and a hardware model applicable to vocoder systems
Published by Institute of Electrical and Electronics Engineers (IEEE) ,2005
Custom LST chip-set for speech analysis
Published by Institute of Electrical and Electronics Engineers (IEEE) ,2005
Improvements of the harmonic-sieve pitch extraction scheme and an appropriate method for voiced-unvoiced detection
Published by Institute of Electrical and Electronics Engineers (IEEE) ,2005
DFT-Vocoder using harmonic-sieve pitch extraction
Published by Institute of Electrical and Electronics Engineers (IEEE) ,2005
A 256-point discrete Fourier transform processor fabricated in a 2 /spl mu/m NMOS technology
IEEE Journal of Solid-State Circuits, 1983
A Compact, Flexible LPC Vocoder Based on a Commercial Signal Processing Microcomputer
IEEE Journal of Solid-State Circuits, 1983
Signal Processing Using MOS-VLSI Technology
Published by Elsevier ,1981