DVB-T2 LDPC Encoding Explained!

Well, I can't really explain LDPC encoding fully, but I have been looking at section  6.1.2 and the tables in Appendix A of the document ETSI EN 302 775 for a long time (maybe years),  without understanding anything.

Now I have written Matlab/GNU Octave code to demonstrate the algorithm.   I have tested it for the normal FEC block size (64800) and only for code rate = 2/3.   It is a lot slower than other encoders I found, but gives the same answers.  Whereas other encoders does one encoding in the blink of an eye, this demo code takes about 1 minute on my laptop.

% This is supposed to be ran interactively by copying and pasting. % NOTE! % You need the function t2_std_dvbt2_bl_ldpc from the % DVB T2 Common Simulation Platform.  Best is maybe to copy it % into your working directory.  In this function, find ct1 and ct2 % for code rate 2/3(SONY matrix).  Copy and paste them into Octave or Matlab. %  Then copy and paste he code below: % % Assemble ct1 and ct2 into one matrix ck, padding ct2 with zeros ck=zeros(size(ct1,1)+size(ct2,1), max([size(ct1,2) size(ct2,2)])); ck(1:size(ct1,1),1:size(ct1,2)) = ct1; ck(size(ct1,1)+1:size(ck,1), 1:size(ct2,2)) = ct2; % Set the LDPC parameters Nldpc=64800; CR=2/3; Kldpc = CR*Nldpc; np=Nldpc-Kldpc; % Create an information block %i=[0 1 0 1 0 1 0 1 0 1 ones(1,Kldpc-20) 0 1 0 1 0 1 0 1 0 1] ; i=randint(1,Kldpc,2); % Initialize the parity bits p=zeros(1,np); Qldpc = 60; tic; for ii=0:size(i,2)-1   ckr = idivide(ii, 360)+1;   m360 = mod(ii, 360);   % Some debug you can uncomment   %fprintf(1, 'ckr = %5d m360 = %d\n', ckr, m360);   % More debug you can enable/change to see some of the accumulation addresses   % if ii>355 && ii<365   %  mod(nonzeros(ck(ckr,:))'+m360*60,np)+1   % end   p(mod(nonzeros(ck(ckr,:))'+m360*60,np)+1) = mod(p(mod(nonzeros(ck(ckr,:))'+m360*60,np)+1) + i(ii+1), 2); end for pp=2:np   p(pp) = mod(p(pp)+p(pp-1), 2); end toc; % Verification agains Matlab code, if you have Matlab % I tested this with alternative code in Octave H = t2_std_dvbt2_bl_ldpc(CR, Nldpc, '1.3.1'); enc = fec.ldpcenc(H); cw = encode1(enc, i); ptest = cw(Kldpc+1:Nldpc); nnz(p) nnz(ptest) nnz(p-ptest)

I hope somebody finds this useful. I think it should be useful for at least two purposes:

• Educational.
• Verifying other LDPC encoders.