Scribd
Upload
84 views

MRC Matlab

% Maximal Ratio Combining Example % luminouslogic.com % Goal: prove that by weighting multiple copies of a signal by each's SNR, the result is % a single signal whose SNR is the SUM of the individual copies % Initialize workspace clear all; close all; % Simulation parameters snr_db_copies = [20 15 12]; % The SNR in dB of each of the copies num_iter = 10000; % # of combines to perform % Create BPSK signals tx_symbols = round(rand(num_iter,1))*2 - 1; % Define how much to scale the mu=0 s=1 noise to get desired SNRs % Now, I know that SNR_dB = 10*log10( signal_var / noise_var) % And by experimentation, I see that if noise created by randn is scaled by K, % the resulting variance of that noise is K^2. So, % SNR_dB = 10*log10( signal_var / K^2) % SNR_dB / 10 = log10( signal_var / K^2) % 10^(SNR_dB / 10) = signal_var / K^2 % 10^(-SNR_dB/10) = K^2 / signal_var % K^2 = signal_var * 10^(-SNR_dB/10) % so use K = sqrt(signal_var * 10^(-SNR_dB/10)) to scale noise for desired SNR

Uploaded by

Yoftahi
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
84 views

MRC Matlab

% Maximal Ratio Combining Example % luminouslogic.com % Goal: prove that by weighting multiple copies of a signal by each's SNR, the result is % a single signal whose SNR is the SUM of the individual copies % Initialize workspace clear all; close all; % Simulation parameters snr_db_copies = [20 15 12]; % The SNR in dB of each of the copies num_iter = 10000; % # of combines to perform % Create BPSK signals tx_symbols = round(rand(num_iter,1))*2 - 1; % Define how much to scale the mu=0 s=1 noise to get desired SNRs % Now, I know that SNR_dB = 10*log10( signal_var / noise_var) % And by experimentation, I see that if noise created by randn is scaled by K, % the resulting variance of that noise is K^2. So, % SNR_dB = 10*log10( signal_var / K^2) % SNR_dB / 10 = log10( signal_var / K^2) % 10^(SNR_dB / 10) = signal_var / K^2 % 10^(-SNR_dB/10) = K^2 / signal_var % K^2 = signal_var * 10^(-SNR_dB/10) % so use K = sqrt(signal_var * 10^(-SNR_dB/10)) to scale noise for desired SNR

Uploaded by

Yoftahi
Copyright
© © All Rights Reserved
Available Formats
Download as TXT, PDF, TXT or read online on Scribd
You are on page 1/ 2

% Maximal Ratio Combining Example

% luminouslogic.com
% Goal: prove that by weighting multiple copies of a signal by each's SNR, the r
esult is
% a single signal whose SNR is the SUM of the individual copies

% Initialize workspace
clear all;
close all;

% Simulation parameters
snr_db_copies = [20 15 12]; % The SNR in dB of each of the copies
num_iter = 10000; % # of combines to perform

% Create BPSK signals


tx_symbols = round(rand(num_iter,1))*2 - 1;

% Define how much to scale the mu=0 s=1 noise to get desired SNRs
% Now, I know that SNR_dB = 10*log10( signal_var / noise_var)
% And by experimentation, I see that if noise created by randn is scaled by K,
% the resulting variance of that noise is K^2. So,
% SNR_dB = 10*log10( signal_var / K^2)
% SNR_dB / 10 = log10( signal_var / K^2)
% 10^(SNR_dB / 10) = signal_var / K^2
% 10^(-SNR_dB/10) = K^2 / signal_var
% K^2 = signal_var * 10^(-SNR_dB/10)
% so use K = sqrt(signal_var * 10^(-SNR_dB/10)) to scale noise for desired SNR

% Create channel noise


K = sqrt( var(tx_symbols) * 10.^(-snr_db_copies/10));
noise = randn(num_iter, length(K)) .* K(ones(1, num_iter),:);

% Create received copies of noisy signals


rx_symbols = tx_symbols(:,ones(1,length(snr_db_copies))) + noise;

% Estimate SNR on each channel to verify coded up correctly


S = var(tx_symbols);
for i=1:length(snr_db_copies)
N = var(rx_symbols(:,i) - tx_symbols);
fprintf(1,'SNR on channel %d is supposed to be %2.0f dB. Actually is %2
.2f dB\n',i,snr_db_copies(i),10*log10(S/N))
end

% Create Maximal Ratio Combining (MRC) weights


mrc_weights = 10.^(snr_db_copies/10); % Weight by SNR (but not in dB!)

% Perform MRC
mrc_rx_symbols = rx_symbols * mrc_weights.' / sum(mrc_weights);
% Did it work? Well, isn't that something!
N = var(mrc_rx_symbols - tx_symbols);
fprintf(1,'SNR after MRC is supposed to be %2.1f dB. Actually is %2.2f dB\n'
,10*log10(sum(mrc_weights)), 10*log10(S/N))

You might also like