On decoding bit error probability for binary convolutional codes

An explanation is given for the paradoxical fact that, at low signal-to-noise ratios, the systematic feedback encoder results in fewer decoding bit errors than does a nonsystematic feedforward encoder for the same code. The analysis identifies a new code property, the d-distance weight density of the code. For a given d-distance weight density, the decoding bit error probability depends on the number of taps in the realization of the encoder inverse. Among all encoders for a given convolutional code, the systematic one has the simplest encoder inverse and, hence, gives the smallest bit error probability