MATLAB: Estimate BER for LDPC 1/2 code

berldpc

I try to find the ber for LDPC code but I the result it seems wrong. pls help!! 
EbNoVec = (-1:0.5:11 )';      % Eb/No values (dB)
berEst = zeros(size(EbNoVec));
bpskModulator = comm.BPSKModulator;
bpskDEModulator = comm.BPSKDemodulator;
p = dvbs2ldpc(1/2);
ldpcEncoder = comm.LDPCEncoder(p);
ldpcDecoder = comm.LDPCDecoder(p);
msg = logical(randi([0 1],size(p,2)-size(p,1),1));
for n = 1:length(EbNoVec)
   
    % Reset the error and bit counters
    numErrs = 0;
    numBits = 0;
 
    while numErrs < 1000 && numBits < 1e6
        
         % Transmit and receive LDPC coded signal data
        encData = ldpcEncoder(msg);
     
         %channel
        dataMod_psk = bpskModulator(encData);
        
        
        % Pass through AWGN channel
        channel = comm.AWGNChannel('EbNo',EbNoVec(n),'BitsPerSymbol',1);
        dataMod_psk2=channel(dataMod_psk);
        
        dataDeMod_psk3=bpskDEModulator(dataMod_psk2);
        rxBits = ldpcDecoder(dataDeMod_psk3);
        
        MSG=rxBits;
        chk = isequal(msg,MSG);
            
       % Calculate the number of bit errors
        nErrors = biterr(msg,MSG);
        
        % Increment the error and bit counters
        numErrs = numErrs + nErrors;
        numBits = numBits + length(msg); 
    end
    
    % Estimate the BER
    berEst(n) = numErrs/numBits;
end
berTheory = berawgn(EbNoVec,'psk',2,'nondiff');
semilogy(EbNoVec,berEst,'-.')
grid
legend('Estimated BER ldpc')
xlabel('Eb/No (dB)')
ylabel('Bit Error Rate')

Best Answer

Hi Pertesis,
The issue you observe is due to the usage of comm.BPSKDemodulator with hard decision (which is the default) in conjunction with the LDPC decoder. Note that LDPC decoder expects the input to be the LLR's, but due to the hard decision from the BPSK demodulation, the output is bits from it. This is provided to LDPC decoder and it just wasn't able to decode any of it properly. I suggest you to make the following update to the BPSK demodulator and place the EbNoVec range from -11:0.1:2 to see the free falling curve.
bpskDEModulator = comm.BPSKDemodulator('DecisionMethod',"Log-likelihood ratio");
More details of the usage of LDPC decoder and BPSK demod can be accessed with the links.
Hope this helps.
Regards,
Sriram
Related Question