Error Event Statistics for Convolutional Codes

Viterbi (1971) introduced a structure generating function for convolutional codes and used it to bound the probability of a decoding error . Viterbi\´s result is used to approximate the probability function on error events of length . Using , approximate values are found for the error event statistics , the expected number of symbol errors in an error event; , the expected number of branch errors in an error event; and , the expected length (in branches) of an error event. The statistics are technically approximate, but are practically upper bounds, loose at high channel error rates, and tightening as the channel error rate drops. The per-unit-length statistics and appear to be of use in finding good codes.