On Thu, 20 Apr 2006 12:37:04 +0200, Serge Bets wrote: > Hello Jonathan, > > On Monday, April 17, 2006 at 18:33:50 +0100, Jonathan Buzzard wrote: > >> http://www.buzzard.me.uk/jonathan/radioclock.html >> At the moment I am working on some new advanced statistical filtering >> of the signal that should provide *much* better decoding with weak >> signals. No promise on the timescale when it will be ready however. > > Have you seen the noisy signal decoding scheme of Poul-Henning Kamp? He > published code for the DCF77 signal on the ntp-hackers list last > February. It seems to me promising and very cheap. > http://lists.ntp.isc.org/pipermail/hackers/2006-February/001934.html >
Interesting but what I am aiming at is something much more sophisticated. The basics are I am using Bayesian statistics. That is I take a prior which is what I think the transitions from 0 to 1 and 1 to 0 should be occurring based on the current time on the computer. I then feed in the measured transitions, which I use to update my prior and out pops the result. I can do this every second, every two seconds, every three seconds etc. down to every minute depending on how much CPU time I wish to expend on the task. Oh and it's proper Bayesian statistics I am using, not some thing else dressed up as being Bayesian as seems to happen far to often. The only problem at the moment (apart from being written in Matlab) is that it does require your computers clock to be close to UTC to begin with. If you are more than a second or so out that it does not work very well. I also need to test it on something more than a particular 24hour period of the HBG signal that I captured last August. My normal code won't decode the HBG signal where I live, just too far from the transmitter. JAB. -- Jonathan A. Buzzard Email: jonathan (at) buzzard.me.uk Northumberland, United Kingdom. Tel: +44 1661-832195 _______________________________________________ questions mailing list [email protected] https://lists.ntp.isc.org/mailman/listinfo/questions
