Oh... thanks for that! I wasn't aware of it. I will try to modify it later, but after a first inspection, I see I will need to fill the preamble not only modulated by square-gaussian pulses, but also modulated by frequency, and use this block with base band signal instead of Quadrature Demod output (as I'm doing right now). This puzzles me because I will not be able to put the following PFB Clock Sync in baseband since GFSK isn't a linear modulation (can't be stated as a sum of shaped pulses) and there is no proper matched filter.
Should I use the PFB Clock Sync AFTER Quadrature Demod? Will the "time_est" tag make sense? Many thanks! 2016-01-28 20:03 GMT-03:00 Tom Rondeau <t...@trondeau.com>: > On Thu, Jan 28, 2016 at 11:04 PM, Francisco Albani < > francisco.alb...@gmail.com> wrote: > >> Hi to all. >> >> I want to test if the block Correlate and Sync improves my GFSK receiver, >> so I started modifying ./gr-digital/examples/demod/test_corr_and_sync.grc . >> >> 1) I replaced the BPSK modulator (with rrc filter) by a Chunks to Symbols >> [-1,+1] followed by a Interpolating FIR with gaussian taps. >> 2) I replaced the matched filter taps for Correlate and Sync and >> Polyphase Clock Sync with the ones I think are correct BUT OBVIOUSLY ARE >> NOT [<-- help needed here]. >> 3) I eliminated frequency offset and Costas Loop because I believe they >> are not applicable after a FM demodulation (which I did not included in >> this first exploration). >> >> (Flowgraph attached) >> >> Things I observe: >> >> * Correlate and Sync detects preamble start, but the clock recovery looks >> wrong (this can be seen zooming in or in the constellation plot). >> * QT GUI Time sink showing correlation does not triggers with the >> hardcoded value (200). I had to change it to 45. >> >> I'm having a hard time figuring out from the c++ code and documentation >> the right way to design the taps to put into Correlate and Sync and PFB >> Clock Sync blocks. >> >> I always find myself feeling I get contradictory clues from code, >> examples and documentation when choosing the taps for blocks that depend on >> PFB. >> >> Can you help me? >> >> Thanks! >> > > Check out the test_corr_est.grc example, instead. The corr_est block was > written to replace the correlate_and_sync to improve the internal > performance issues as well as make it more applicable to other modulation > types like GMSK and GFSK. See if that example doesn't make more sense for > your situation. > > Tom > >
_______________________________________________ Discuss-gnuradio mailing list Discuss-gnuradio@gnu.org https://lists.gnu.org/mailman/listinfo/discuss-gnuradio
