LDPC Code simulation with soft decoding

Hi. I'm trying to create an LDPC code simulation for learning.

The code I made earlier is as follows.

% Define DVB-S.2 LDPC code parameters
dvbS2Rate = 1/2; % Code rate for demonstration
dvbS2ldpc = dvbs2ldpc(dvbS2Rate); % Get the 
LDPC object for DVB-S.2
numInfoBits = size(dvbS2ldpc, 2) - size(dvbS2ldpc, 1); % Number of information bits
% Set up simulation parameters
SNR_dB = 1:10; % SNR range from 1 to 10 dB
Nstop = 1000; % Number of frames to simulate
WER = zeros(1, length(SNR_dB)); % Pre-allocate WER for each SNR value
% Create LDPC encoder and decoder objects
ldpcEncoder = comm.LDPCEncoder('ParityCheckMatrix', dvbS2ldpc);
ldpcDecoder = comm.LDPCDecoder('ParityCheckMatrix', dvbS2ldpc, 'DecisionMethod', 'Soft decision');
% Simulation loop over SNR values
for snr_idx = 1:length(SNR_dB)
    % Convert SNR from dB to linear scale for noise variance
    SNR_linear = 10^(SNR_dB(snr_idx) / 10); 
    noiseVar = 1 / (2 * SNR_linear * (1/dvbS2Rate));
    % Create AWGN channel object
    awgnChannel = comm.AWGNChannel('NoiseMethod', 'Variance', 'Variance', noiseVar);
    % Initialize error counters
    numberOfBitErrors = 0;
    numberOfWordErrors = 0;
    for frame = 1:Nstop
        % Generate random information bits
        infoBits = logical(randi([0 1], numInfoBits, 1));
        % LDPC encoding
        encodedBits = ldpcEncoder(infoBits);
        % BPSK Modulation
        modulatedSignal = 2 * double(encodedBits) - 1;
        % Pass through AWGN Channel
        receivedSignal = awgnChannel(modulatedSignal);
        % BPSK Demodulation
        receivedBits = receivedSignal < 0;
        % LDPC Decoding
        decodedBits = ldpcDecoder(double(receivedBits));
        % Calculate the number of bit errors
        bitErrors = sum(infoBits ~= logical(decodedBits));
        numberOfBitErrors = numberOfBitErrors + bitErrors;
        numberOfWordErrors = numberOfWordErrors + (bitErrors > 0);
    end
    % Calculate WER for the current SNR
    WER(snr_idx) = numberOfWordErrors / Nstop;
end
% Plot the WER vs SNR graph
figure;
semilogy(SNR_dB, WER, 'b-o');
xlabel('SNR in E_b/N_0 in dB');
ylabel('WER');
title('Word Error Rate vs. SNR for DVB-S.2 LDPC Code');
grid on;

But Result is..

Like that.. So I wondered

[Jongmin]

The Ideal result is this(It is hard decision result when I run that code)

Answer 1

[Jaswanth]

Hi Jongmin,

In your LDPC code simulation for a DVB-S.2 setup, the key issue lies in the input format expected by the decoder. Your current approach uses binary bits, whereas the decoder requires Log-Likelihood Ratios (LLRs). To resolve this, you need to modify the demodulation phase to calculate LLRs from the received signals, considering the noise variance.

Please refer to the following response given by a MathWorks Staff member for potentially helpful information.

https://www.mathworks.com/matlabcentral/answers/1409347-how-to-use-the-ldpc-encoder-decoder-in-matlab#comment_2969247

Also, it is advised in MathWorks documentation to use ldpcEncode over comm.LDPCEncoder, which will be discontinued in future updates. Please refer to following MathWorks documentation to learn more about ldpcEncode: https://www.mathworks.com/help/comm/ref/ldpcencode.html

I hope the information provided above is helpful in accomplishing your task.