From: jfield on
There is no longer any MATLAB documentation on the "ddemodce" function, but my guess is that it includes an integrate and dump type demod. This means that if you give it 4 samples per symbol, it will take the average location of all the samples of that symbol and make a symbol decision.
Since the simulation has perfect TX to RX timing and does not add phase noise or a frequency offset, the only issue is AWGN. Averaging 4 samples per symbol will improve the symbol voltage decisions by 6 dB (i.e. 20*log10(sqrt(4)).
If you change bb_data to an all ones array, then all the QPSK symbols will be '3'. If you then set the data_len = 2, the Eb_No_dB = -5 dB and run the simulation many times, you will eventually see an bit error or 2 every time the average location of the 4 samples of the one symbol sent is outside the symbol '3' quadrant.
Hope this helps.
 | 
Pages: 1
Prev: a question about matrix
Next: wavelet