How does ofdmChannelResponse function really work ?

Question

Seyyed Erfan 2025-7-18

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2178568-how-does-ofdmchannelresponse-function-really-work

评论： Seyyed Erfan 2025-7-23

https://uk.mathworks.com/help/comm/ref/ofdmchannelresponse.html#mw_47820a6f-11d0-4482-a5db-e137a9df42db_head

I have written this code and I want to know what exactly happens under the hood of the ofdmChannelResponse block and how can I replicate it manually in my code ? In other words, what kind of mathematics happens to obtain hest given channel pathgains, path filters, nfft and CP size ?

%% Parameters
nFFT = 128;            % Number of FFT bins
cpSize = 32;           % CP length
M_mod = 4;             % Size of QAM constellation
M_bits = log2(M_mod);  % Bits per QAM symbol
N_bits_perfram = nFFT * M_bits;
SNR_dB = 0:3:30;
Nf = 1e4;              % Number of Monte‐Carlo trials per SNR
delta_f = 15e3;        % 15 kHz subcarrier spacing
fs = nFFT * delta_f;   % Sampling Frequency (1.92 MHz)
pathDelays = [0 50 120 200 230 500 1600 2300 5000]*1e-9;
delaySamples  = round(pathDelays * fs);  
P = length(pathDelays);
avgPathGains_dB = [-1 -1 -1 0 0 0 -3 -5 -7];  % (dB)
%% Doppler shifts to sweep
doppler = [0, 100];
numD    = length(doppler);
fd1 = doppler(1);
fd2 = doppler(2);
%% Initialize BER accumulators for this Doppler
BER   = zeros(1, length(SNR_dB));
%% Create Rayleigh channel object (with current fd)
rayleighchan = comm.RayleighChannel( ...
    'SampleRate',           fs, ...
    'PathDelays',           pathDelays, ...
    'AveragePathGains',     avgPathGains_dB, ...
    'MaximumDopplerShift',  fd1, ...
    'PathGainsOutputPort',  true);
chaninfo = info(rayleighchan);
coeff    = chaninfo.ChannelFilterCoefficients;
%% Loop over SNR points
for q = 1:length(SNR_dB)
    snr = SNR_dB(q);
    for i = 1:Nf
        %-----------------------------------------
        % 1) Data Generation
        %-----------------------------------------
        data_info_bit = randi([0,1], nFFT*M_bits, 1);
        data_c = bi2de(reshape(data_info_bit, nFFT, M_bits));
        qamTx = qammod(data_c, M_mod, 'gray','UnitAveragePower', true);       
        %-----------------------------------------
        % 2) OFDM Transmitter 
        %-----------------------------------------
        txout = ofdmmod(qamTx, nFFT, cpSize);
        %-----------------------------------------
        % 4) Fading Channel + AWGN for OFDM path
        %-----------------------------------------
        [chanOut, pathGains] = rayleighchan(txout);
        [RXIN_noisy, nvar] = awgn(chanOut, snr,"measured");
        %-----------------------------------------
        % 6) Channel Estimation (OFDM)
        %-----------------------------------------
        hest = ofdmChannelResponse(pathGains, coeff, nFFT, cpSize);
        % Demodulate OFDM
        rxsym = ofdmdemod(RXIN_noisy, nFFT, cpSize);
        %-----------------------------------------
        % 8) MMSE Equalization (OFDM)
        %-----------------------------------------
        G_mmse   = conj(hest) ./ (abs(hest).^2 + nvar);
        X_hat    = G_mmse .* rxsym;
        %-----------------------------------------
        % 11) QAM Demapping + BER Accumulation (OFDM)
        %-----------------------------------------
        data_demapping = qamdemod(X_hat, M_mod, 'gray','UnitAveragePower', true);
        data_info_est1 = reshape(de2bi(data_demapping, M_bits), nFFT*M_bits, 1);
        BER(q) = BER(q) + sum(xor(data_info_est1, data_info_bit));
        
    end  % end Monte Carlo loop Nf
end  % end SNR loop
%% Normalize errors to obtain BER for this Doppler
ber    = BER   ./ (N_bits_perfram * Nf);

2 个评论
显示无隐藏无

dpb 2025-7-18

在 MATLAB Online 中打开

which -all ofdmChannelResponse
/MATLAB/toolbox/comm/comm/ofdmChannelResponse.m

is an m-file, you can go see what is under the hood, at least until it drops into either builtin or p-code functions, if it isn't all in the m-file.

Normally, TMW does provide links to supporting papers in the Algorithms section but it seems they didn't here. But, that's what comes with the proprietary language; they can document only what they feel like doing keeping internals hidden either for what may be considered competitive advantages or simply lack of resources to put more into the documenattion or just an oversight.

Seyyed Erfan 2025-7-23