Practical implementation of cross-layer design in broadband wireless access networks
International Journal of Mobile Network Design and Innovation
| Hi-index | 0.00 |
In the Frequency Division Duplex (FDD) mode of the Third Generation Partnership Project (3GPP) standard, implementation of the turbo decoder, especially for the mobile equipments, faces design decisions related to computational complexity, power efficiency, and memory requirements. In this paper we compare different approaches of low complexity implementation of the turbo decoder, with emphasis on the issues of signal scaling and quantization, the sliding window operation for memory size reduction and the iteration stopping algorithms. The demodulated signal at the output of the RAKE receiver may have a wide dynamic range and it may require many bits of precision. In order to overcome the numerical precision problem and to prevent Log Likelihood ratio (LLR) metric overflow, a scaling algorithm must be used. Our simulation results indicate that the Average Absolute (AA) algorithm using dynamic scaling outperforms other scaling schemes and it is less sensitive to the channel conditions. One of the major challenges in the implementation of a practical turbo decoder is optimization of memory requirements. In this paper we evaluate the performance of the sliding window algorithm using different main and guard window sizes. We show that the bit and block error rate performance of the sliding window scheme mainly depend on the guard window size rather than the main window size. The simulation results indicate that small guard window sizes can significantly decrease the iteration gain for large frames in fast fading channels. Iteration stopping algorithms reduce the power consumption and the latency of the decoder and help to dedicate more resources to other functions of the receiver. The block error distribution in the fading channels makes it even more essential to use an iteration stopping rule. Our simulations conclude that a rule called the minimum absolute value appears to be a very effective, low complexity and robust algorithm.