Low-power turbo decoding based on reverse calculation

Abstract

Turbo code is one of the most powerful forward error correcting codes. But, as turbo decoder is included in a class of highly memory-intensive systems, turbo code suffers from large power consumption. Although several low-power techniques have been proposed, power consumption is still a major issue to be solved in practical implementation. This thesis presents a reverse calculation method for forward metrics in order to reduce memory accesses in turbo decoding. First, the reverse calculation method will be presented for MAX-log-MAP algorithm. Then, we will show that the same method can be applied to log-MAP algorithm after some modification in the log-MAP algorithm. Experimental result shows that in W-CDMA standard, 80% of memory accesses can be reduced through the reverse calculation method. By employing the method in the modified-log-MAP turbo decoder, power consumption is reduced by 34.4% and memory size is reduced by 39.2%.