BPSK BER simulation

I need to solve the problem with BER calculation. I made a coherent correlationtype detector in MATLAB for a BPSK signal, but when I demodulate the noisedsignal BER is much worse than it sould be in theory for coherent detection. Iadd noise using awgn() function setting the desired SNR[dB]. I assumeEb/No[dB]=SNR[dB]+10log(Tbit/Tsample) So why is it? My M-Files are attached. Iused 1000 bits to simulate and SNR=-12dB which equals according to the EquationEb/No=8dB. Theoretically the result shoul have been around 1e-4 and I getBER=4e-2. The BER curves of the desired and of the Reality I have are also enclosed.Please help!