For a lot of people the FFT seems to be the "one size fits all" solution to any frequency and phase related problem in DSP. It is NOT! For frequency/phase detection & comparisons from sets of sampled data the methods explained in
http://ipnpr.jpl.nasa.gov/progress_report/42-121/121G.pdf are MUCH more appropiate. Best regards Ulrich Bangert > -----Ursprungliche Nachricht----- > Von: [email protected] > [mailto:[email protected]] Im Auftrag von Hal Murray > Gesendet: Mittwoch, 3. Juni 2009 20:00 > An: Discussion of precise time and frequency measurement > Betreff: Re: [time-nuts] Sound Cards for locking to GPSDO 10 > MHz references > > > > > One of the main problems is that in working at milli-Hz > binwidths the > > FTT word length needs to be very long to cover even a few > tens of Hz > > range and we run into memory problems. > > I'm missing something. How much memory do you have on your laptop? > > I'm not a DSP wizard. If you have 10 Hz bandwidth and you > want milli-Hz > bins, that takes 2x10x1000 samples. Right? I'd expect that > to fit easily. > > That's 20K samples, at 8 bytes each, round up to 10, call it > 200K bytes. > > Jumping to micro-Hz might get interesting. That would be 200 > megabytes. > Lots of laptops have room for that. Maybe not an old one. > > Even with an old laptop without much memory, I'd expect you > could do several > factors of 2 better than milli-Hz bins. > > On the other hand, how much bandwidth do you really need? > Junk crystals are > 50-100 ppm. 100 ppm at 1 KHz is 1/10 Hz. So why do you need > more than 1 Hz > input bandwidth? You can probably get closer than that by > calibrating the > crystals in your particular gear. > > Connie's numbers were 250 micro-Hz drift with a 500 micro-Hz > offset. (That > was with reasonably stable temperature.) So a few milli-Hz > bandwidth looks > like enough. > > > > > > -- > These are my opinions, not necessarily my employer's. I hate spam. > > > > > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to > https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts > and follow the instructions there. _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts and follow the instructions there.
