Threshold levels on Correlation Receivers

[ptmartins]

Hi all,

I'm doing reception on BPSK modulated symbols, using correlation receivers. I am testing several templates to correlate with the received signal.

To make the bit decision I compare the value of that correlation with a threshold. But i'm having trouble in computing that treshold. If is too low all bits are noise... If is too high all bits are right (even in 0 dB of SNR...). I can't find references that talk about this problem. Is there a rule for the treshold ?
What does say your expirience?

thanks

[tmcdavid]

PT:
If you are unfotunate enough to suffer from a DC component in your correlation output, you should not use a threshold. If no DC is involved, you would use zero threshold. That is to say, if your correlation coefficients match symbol A and your contrasting phase is symbol B and the cross-correlation of A with B is -1, then the correlation of A with the received mix of symbols with noise will have a positive correlation when the signal contains sufficient A energy to overcome the noise and a negative correlation when the B energy dominates.

If you cannot easily set up for zero threshold, all is not lost!! You need to correlate the input against both A and B. Twice as much compute work, but then the most likely symbol mixed in the noise is the one with a higher correllation value than the other.

Let us know how you solve the challenge.

Regards,

Terry