The
William E. Brice
Colloquium Series

Dr. Andrew J. Viterbi

Vice Chairman and Chief Technical Officer
QUALCOMM Incorporated
San Diego, CA

Approaching the Shannon Limit:
Theorist's Dream and Practitioner's Challenge

Nearly half a century after Shannon established the channel capacity of a noisy channel as the insurmountable limit on the rate of reliable communication, a combination of long known but dormant concepts and a fresh approach has brought us to within a remarkably small distance of that limit. For an additive white Gaussian (AWGN) channel, low error probabilities have been achieved at rates greater than eighty percent of capacity, or in terms of bit energy-to-noise density ratio, Eb/N0, within less than 1dB of the minimum value which corresponds to operation at channel capacity.

The underlying concept which led to these results, first established through simulation by Berrou et al [1993], is that of iterative codes with soft decision output decoding. Four principal characteristics distinguish this new class of coders and decoders, which were originally denoted as "turbo" codes.

  1. the code is systematic with two or more sets of parities, all but one generated from a different (pseudorandom) permutation of the information bits;
  2. each set of parities is generated recursively by a linear shift register with feedback;
  3. the decoder successively decodes the information based on each set of parities individually, generating for each information bit a soft decision output to be used to improve the decoding accuracy for the next set of parities. Iterations continue as long as the decisions improve;
  4. performance depends strongly on the choice of output soft decision metrics. The preferred alternative is the maximum a posteriori probability (MAP) algorithm devised by Bahl et al [1974].
In this lecture we shall present a simplified approach to the MAP decoder and relate it to the conventional convolutional decoder in common usage in most digital wireless communications systems.
Thursday, September 19, 1996
4:00p.m. - 301 Sewall Hall
Rice University
Electrical and Computer Engineering Department

URL for this page: http://www-ece.rice.edu/Brice/Sep19-96.html

Brice Colloquium Series

Comments and questions to: mandynev@ece.rice.edu

Last modified: August 1, 1996