A New Look at the Error-Correcting Capabilities of Convolutional Codes ∗

Communication systems that employ a single convolutional code for error correction often operate at rates far below capacity. In that region, where channel errors are sparse, the performance of a code is well described by the free distance. The discovery of Turbo codes, however, made investigations of the channel region close to, and even above, capacity necessary. Recently, Jordan et al. proposed to use a lower bound on the active distances to describe a code’s performance close to capacity and defined in this context a family of maximum slope (MS) codes. Here, we point out some issues that show that maximum slope cannot be the whole story. We look at indicators of error-correcting capability of convolutional codes from a new perspective. Based on the average length of the error bursts at the output of a Viterbi decoder as well as on the first event error probability, we present a new method to gain insight into a code’s performance. Simulation results support our findings that are applicable to all channel regions, no matter whether operating far below or close to capacity. Keywords—Error-correcting capability, maximum slope, average burst length, first event error probability.