On Wed, Jul 29, 2009 at 2:46 PM, George Nychis
<address@hidden> wrote:
On Wed, Jul 29, 2009 at 1:43 PM, Shizheng Li
<address@hidden> wrote:
Dear all,
I am reading the GMSK demodulator code (gmsk.py) and I have a question. In the modulator, Gaussian filter is used before the FM modulator. In the demodulator, no matched filter is used. The signal go through FM demodulator, then clock recovery, then make the decision. Why not use a matched filter after the demodulator, as in standard diagrams of the receivers? Thank you very much!
I think that one reason may be is that the output from the USRP to the host is not normalized. With a matched filter, you're going to convolve the input signal with your coefficients, and if your input signal is not normalized then a higher input signal to the USRP will generate stronger output from the matched filter (whether it's a match or not). Therefore you're going to have trouble thresholding to determine the peak in synchronization.
actually, I'm thinking an acquisition matched filter, so you might want to ignore that :)
- George