Code concatenation is used to construct code families with good

  theoretical and practical properties. The underlying idea has been

  first proposed by Forney in 1966 where simple and powerful code

  constructions have been derived. In this talk we address concatenated

  code constructions based on simple subcodes, where we are interested

  in asymptotic properties of the underlying random coding

  ensembles. For example, a linear growth of the minimum distance of the

  overall code with block length guarantees that for good communication

  channels the residual error probability after decoding gets

  arbitrarily close to zero if the block length tends to infinity. On

  the other hand, since concatenated codes are generally decoded

  iteratively, iterative decoder convergence is another issue

  which needs to be addressed besides distance growth.

  

  In particular, we consider the ensemble of codes formed by a serial

  concatenation of an outer code (for example a repetition code) with

  multiple accumulators through uniform random interleavers. Based on

  finite length weight enumerators for these codes, asymptotic

  expressions for the minimum distance and an arbitrary number of

  accumulators larger than one are derived. In accordance with earlier

  results in the literature, we first show that the minimum distance of

  repeat-accumulate codes can grow, at best, sublinearly with the block

  length. Then, for repeat-accumulate-accumulate codes and code

  rates of 1/3 or smaller, it is proved that these codes exhibit

  linear distance growth with block length, where the gap to the random

  coding bound can be made arbitrarily small by increasing the number of

  accumulators beyond two.

  

  Finally, we briefly present hybrid concatenated coding schemes which

  represent a mixture between serial and parallel concatenation.  We

  show that such structures also exhibit linear distance growth with

  block length (albeit with a smaller growth rate coefficient), but as

  an advantage have better convergence properties than multiple serially

  concatenated codes.