Construction of partial unit memory encoders for application in capacity-approaching concatenated codes

In order to achieve capacity approaching performance, the use of partial unit memory (PUM) component codes with good distance properties in turbo codes and woven turbo codes have been suggested. PUM codes are no more complex, but perhaps less well known than traditionally used convolutional component codes. Turbo codes and woven turbo codes based on PUM codes have been shown to outperform those based on convolutional codes with no extra decoding complexity. In this paper, the convolutional properties of PUM codes are exploited and how these can be used to construct encoder structures and their trellises is described so that they are suitable for use in the aforementioned parallel concatenated structures. The encoder and trellis construction techniques are illustrated by two example PUM codes. The performance of these codes as component codes in a turbo structure is presented to show the influence of the parameters of the PUM code and how this impacts on the design of suitable component codes in a concatenated structure with capacity approaching performance.