Effects of Quantization in Adaptive Processes. A Hybrid Adaptive Processor.
Abstract
An analysis of two adaptation algorithms for finite-dimensional, discrete-time, linear estimation filters has been completed. The analysis allows the prediction of rate of convergence and of performance level for these algorithms which use information about the polarity of the received data to achieve automatic adaptation to the optimum least-mean-square error filter. An adaptive signal processor which uses digitally controlled analog weights is described. The signal paths through the system are purely analog, thus eliminating the need for output reconstruction from digital samples. Consequently, the adapt-rate does not restrict the frequency response of the system. The computation of new weight values and the control of the system of weights is carried out by a small general purpose digital computer which is interfaced with the adaptive processor. A novel feature of the adaptive signal processor is the preprocessor which consists of a set of simple RC-filter stages instead of the customary delay lines. A simple theoretical analysis of the preprocessor and qualitative comparison with systems utilizing delay lines is given. The results of a set of experiments involving both single and multi-channel filtering of period signals are presented. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 30, 1971
- Accession Number
- AD0725818
Entities
People
- B. Widrow
- John Kaunitz
- John L. Moschner
Organizations
- Stanford University