Abstract

A turbo decoder uses an iterative procedure and its performance is dependent on the number of iterations it performs. In a turbo decoder, the number of iterations required to achieve a certain FER is a function of the channel SNR. An example of this is a base station in a mobile communication system, where turbo decoder receives data frames from different users each with a given level of fading and as a result, with different SNR. If a fixed number of iterations is allocated to each frame, it is possible that certain frames remain erroneous while decoder is idle part of the time for other frames. In this thesis the idea of sharing the processing power among different frames, possibly belonging to different channels, is used. This time-sharing approach can increase the processing throughput and/or the performance of the turbo decoding scheme. We present two schemes and give their performance for the 3GPP and DVB-RCS codes. The first one is serial processing algorithm, which decodes the input frames one by one and the other one is parallel processing scheme that tries to decode all of those received channels simultaneously. We investigated the implementation of these schemes using a Digital Signal Processor (DSP) chip (TI TMS320C6416 DSP CPU). We have developed an efficient implementation algorithm by combining those two multi-channel processing schemes. All the presented schemes can also be used for a single-channel communication system