I've now completed the simulations of the 512 state, rate 1/4 MAP decoder. This time I fixed the signal amplitude A to 7 so as to avoid limiting the state metrics (SM) too much. This is less than the "optimum" values for 6 bit quantisation, but with only 8 bit state metrics, SM limiting becomes more of a factor.
You can obtain a plot a these results from
http://www.itr.unisa.edu.au/~steven/turbo/codec/map49t.ps
You can see that at low BER, the implementation loss is only 0.05 dB. Note that at low BER the performance of the Viterbi and MAP algorithms are almost identical. At high BER, the MAP decoder gains about 0.5 dB over ideal Viterbi.
Also plotted on the graph is the sub-MAP decoder performance. This is where the E operand is simplified to the min function. This was very easily done with the MAP decoder by setting the lookup tables to zero. The signal amplitude was also set to A = 7. As can be seen, at low BER the sub-MAP performance is almost identical to the MAP decoder. However, our sub-MAP always performs worse than ideal Viterbi, loosing about 0.3 dB at high BER.
My next step is to modify the MAP decoder to the Berrou et al. rate 1/2 16 state code.
Steven S. Pietrobon, Satellite Communications Research Centre
University of South Australia, The Levels SA 5095, Australia.
steven@spri.levels.unisa.edu.au
http://www.itr.unisa.edu.au/~steven/