I have met the same problem in simulating an OFDM system in baseband. When I finished a single-user BPSK modulated simulation in flat Rayleigh fading channel. The BER performance is very strange and cannot match the ideal performance for BPSK signals in flat Rayleigh fading channel. When I check the Matlab program, I can find that if I use yt(x), xing*sig.+AWGN,the results are strange. If I change the model to be y=ifft(fading*Sig.)+AWGN, the results is ok (but I did not compare it with the ideal performance. It look like ok. So I investaged the IFFT of AWGN, I found the mean and the vance changed greatly, even though I am not sure weather or not the pdf after FFT the noise is still Gaussian distributed. I also examined another OFDM program for OFDM channel estimation in frequency. I did that according to an IEEE paper. I found the results can match the paper when I use the Y=HX+N after FFT with N to be AWGN. Any help will be appreciated Thanks in advance Li Pingan