On the ensemble performance of turbo codes

We derive the ensemble weight enumeration function of turbo-codes, where the ensemble is generated by a uniform choice over the interleaver and a uniform choice over the component codes from the set of time-varying recursive convolutional codes. This reduces the study of the performance of turbo-codes to a two parameter family, namely the interleaving length N (which determines the delay and the memory requirements) and the memory length m (which determines the complexity).