Resources for Project

Vidya Kumar's readme on the programs

These are the codes required in addition to the ldpc_decode from Igors page (click on software link)) Just write a small code to generate H matrix. As the dimensions of H matrix might be really big its preferable to use sparse format. Once you have your H matrix as a H.mat file, you can run the binmain program. Do download both the programs. The sgm variable in binmain is the sigma of the AWGN channel and the SNR computed is rate compensated. Right now, the program computes for all the values in sgm variable and hence would take around one night to give you the output. Instead I would suggest you run it for just three values i.e SNR=2dB to SNR=3 dB i.e outer for loop in binmain(with j variable) for j=5:7, first just to have a feel of the performance(a BER of 10^-4 at 3 dB is decent for length of code around 4000. The output is in array BER(bit error rate) and is displayed from 0 to 6 db in 0.5 db intervals. You could change this range by changing the range of s in the sigma program attached here. Thats it. Also the number of iterations in ldpc_decode program is 100. You can change this to 15 or 30 to get quicker results.

More Updates

About the SNR: The j value in binmain.m is the indicator of SNR. Rate compensated SNR for guassian channel is E/(2(sigma)^2*r) where sigma is the noise variance ,E is the symbol(bit) energy and r is the rate of code. the sigma.m computes the sigma values for SNR corresponding to 0 to 6. So j=1 SNR=0 dB, j=2 SNR=0.5 dB j=3 SNR=1 dB and so on...