In this contribution we derive the super-trellis structure of turbo codes.
We show that this structure and its associated decoding complexity depend s
trongly on the interleaver applied in the turbo encoder. We provide upper b
ounds for the super-trellis complexity. Turbo codes are usually decoded by
an iterative decoding algorithm, which is suboptimum. Applying the super-tr
ellis structure, we can optimally decode simple turbo codes and compare the
associated bit-error rate results to those of iterative algorithms.