Hi David, I have actually implemented the second option. Detecting the the sequences of "calms" where the modulation is off and gating the ADC sampling at those moments. There are plenty of calm zones between the sequence of modulated bits, in addition the modulation is off for the whole 59th second before each minute. I considered picking the signal only during that 59th second which could indeed lead to an improvement. Will try that. Thx, Gilles. > If the modulation is phased locked to the carrier which is common, > then that suggests two other ways to extract the timing from the > carrier without interference from the modulation. > 1. Integrate the phase comparison over a whole number of modulation > cycles; the modulation will then cancel out. This could be seen as > using a sin(x)/x frequency response with the nulls aligned to remove > just the modulation components of the signal. > > 2. Gate the phase comparison for periods where the modulation is not > occurring. > > On Sun, 5 Mar 2017 11:54:56 +0100, you wrote: > > >I have been working for a "while"…> on a reference oscillator disciplined by an > >RF signal (162khz), Cs stabilized, but phase modulated with time coding > >information. The phase modulation is quite significant : +/-1rad 10Hz > >triangular shape, leading to a frequency shift of +/- 20PI (+/-6,3Hz). > >Extracting the base signal was quite an interesting challenge…> I tried many > >options (and even a bit more…> ) but it was - and is still - fun... > > > >... > > > >My questions to the respected audience: > >- Can I get any better? > >- Have I reached the limit of what the RF signal can provide? > >(with all the perturbations encountered on the path from the transmitter to my > >antenna...) > >- Am I limited by the local oscillator? > >- Or the measurement gear? > >- Or anything else…> ? > > > >Thanks very much for your feedback, > > > >Gilles.