I wanted to add Additive White Gaussian Noise (AWGN) and involve effect of slow flat fading on my transmitted signal in Matlab.

I created noise with zero mean and variance of one by

awgn_noise = random('norm',0,1,1,NoOfSamples);

Then I created Rayleigh fading by

rayleigh = sqrt( (random('norm',0,1))^2 + (random('norm',0,1))^2 );

As far as I understand, the rayleigh values will attenuate the whole signal while gaussian noise values will be added to each sample of transmitted signal as:

received_signal = (transmitted_signal*rayleigh) + awgn_noise;

Am I doing it in a right way?


1 Answer 1


As I described in this answer, it is generally assumed in the Rayleigh fading model that the energy is conserved on average (i.e. that the fading coefficient $R$ is such that $E\left\{R^2\right\} = 1$). As a result you would need to generate the Rayleigh distribution from two normal distribution with standard deviation of $1/\sqrt{2}$ (instead of with unit standard deviation). This can be achieved with:

rayleigh = sqrt( (random('norm',0,1/sqrt(2)))^2 + (random('norm',0,1/sqrt(2)))^2 );

or alternatively, since you are already using random from the statistics toolbox, with:

rayleigh = random('Rayleigh',1/sqrt(2));

Other than that, the expressions you have properly represent a slow flat Rayleigh fading channel for the transmission of a one-dimensional signal (such as BPSK, M-PAM, ...)

  • $\begingroup$ Thanks a lot for the answer. The answer you referred really helped me already while writing these equations. Yes, I am using BPSK signal but I didn't get why the energy is conserved on average i.e. 1 for Rayleigh fading? $\endgroup$
    – Rok
    Dec 22, 2015 at 13:02
  • $\begingroup$ In your second last line "(e.g. lossy channel)", were you supposed to write "(e.g. lossless channel)"? It is making the definition vague. $\endgroup$
    – Rok
    Dec 23, 2015 at 14:06
  • $\begingroup$ This fading model comes about to account for situations where there is a large number of scatterers, but ultimately the energy that is scattered has to come from somewhere. E{R^2}=1 means that all the transmitted power eventually reaches the receiver (lossless channel). Also, you can see this as the model normalization condition: once you have the performance of the system for E{R^2}=1, you can get it's performance for other values of E{R^2} (e.g. lossy channel) by correspondingly adjusting the SNR. $\endgroup$
    – SleuthEye
    Dec 23, 2015 at 14:10
  • $\begingroup$ @Rok I meant lossless channel, but I didn't type it at the right place (now it should be). Thanks! $\endgroup$
    – SleuthEye
    Dec 23, 2015 at 14:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.