Why BER simulated result and theoretical one in Simulink does not match over AWGN channel?

This is my simulink block diagram:

The parameters set in this plant:

Random Integer Generator:

Stop time = 1000000s (Get sufficient data)

AWGN Channel:

After getting the result under Eb/No from 0 - 20 dB and plotting them against theoretical one:

The result does not match the theoretical one.

Note that after 10dB, the data in array (log10) is -inf, and that in Simulink the display shows the BER is 0.

In the same model, I still simulated the 16QAM, 64QAM, 256QAM:

This is even more weird that, with increasing QAM bits per symbol (M=16 to 256 (k=4 to 8)), the simulated BER is decreasing..., which is definitely not the right answer.

So could anyone help me check if there is anything wrong with my model configuration. Or I should do some modification of my model? Thanks..

Answer 1

[Aastha]

Hi @稼润,

I tried to reproduce the Simulink model according to the parameters you provided.

On reproducing the model, I found that you are possibly computing the Symbol Error Rate (SER) instead of the Bit Error Rate (BER). To compute the BER from the SER, scale the error rate obtained from the "Error Rate Calculation" block by log₂(M). This is because the input to the "QPSK Modulator" block is an integer, not bits.

From your attached plot, it can be observed that there is a 3dB shift between the simulated and theoretical results. I have attached a comparison plot of the BER vs. Eb/No before and after scaling. From the plot it can be seen that the theoretical and analytical BER now match.

I hope this helps!