Ravil makes some good points, but I will also point out an error in your project. A common mistake for calculating the BER is using the BER Analyzer for a QAM signal. Since in your design you are detecting the QAM seqeuence and then generating a new one, the eye diagram will be near perfect with no noise.
Search the forum for BER Test Set examples or you could also use the error vector magnitude estimation from the constellation diagram.
Regards