I am starting with what should be a frequency measurement at the sample rate. Timelab can import that and should be able to determine ADEV from that. I agree that something is not right though. I unfortunately am trying to measure with what I already have. I don't have a TIC.
Regards, Mark On Mon, Dec 3, 2018 at 5:10 PM Bob kb8tq <[email protected]> wrote: > Hi > > Let’s back up a bit: > > The input to an ADEV calculation is a record of phase (time between wide > spaced > edges). Often this is from a PPS output on a device. It could equally well > be from > 10 MHz edges spaced a count of 10,000,000 edges apart. > > The first part of the formula takes the difference between adjacent > samples. > > The next step puts that difference into a standard deviation calculation. > > If you are starting with something else, then you likely are not > calculating ADEV. > > Bob > > > On Dec 3, 2018, at 11:21 AM, Mark Goldberg <[email protected]> > wrote: > > > > I have looked at it multiple times and so far cannot see a discrepancy. > The > > ration of the ADEVs for the two sample rates is very close to 2 for a > wide > > range of Tau in the 10-200 sec range. That can't be a coincidence. It's > > between 2 and 3 for higher Tau but that may be due to temperature > changes. > > > > Basically, the Perseus SDR, clocked by my Bodnar GPSDO is sampling the 10 > > MHz from my TruePosition and using the DDC to down convert to 48 kHz. I > > then do either a 32768 or 262144 point FFT, resulting in window times of > > 0.6816 or 5.453 sec and use a built in function in Spectrum Lab that as I > > understand it interpolates the peak frequency from the highest two bins. > > That calculated peak frequency is saved every 0.6816 or 5.453 seconds. > That > > data is fed to Timelab. > > > > I could do the same size FFT for both and decimate the input by 1 and 8 > > respectively, which should result in the same FFT window time and check > > that. > > > > I'm scratching my head here. > > > > Regards, > > > > Mark > > > > > > On Mon, Dec 3, 2018 at 6:28 AM Bob kb8tq <[email protected]> wrote: > > > >> Hi > >> > >> ADEV is pretty well documented and the results should *not* be dependent > >> on the > >> sample rate. The proper approach to dropping the sample rate is > decimation > >> of the > >> data. I think you may want to look at the way you are doing ADEV. > >> > >> Bob > >> > >>> On Dec 3, 2018, at 12:30 AM, Mark Goldberg <[email protected]> > >> wrote: > >>> > >>> I ran them again today and the results are almost identical. The slower > >>> sample rate still shows a lower ADEV. I am not sure what the algorithm > to > >>> determine the peak frequency is but it is described as a linear > >>> interpolation between the two highest peaks on the FFT. I could see > how > >>> that could be subject to math errors. For the slower sample rate, the > >>> frequency bins are smaller. I have not researched how all this could > >> affect > >>> the peak frequency and possibly introduce noise (or by truncation > remove > >>> small amounts of noise). > >>> > >>> I will run it a few times in the next few days and will also try to > move > >>> the antennas to a location with lower multipath. > >>> > >>> Thanks for the suggestions, > >>> > >>> Mark > >>> > >>> > >>> On Sun, Dec 2, 2018 at 2:16 PM Bob kb8tq <[email protected]> wrote: > >>> > >>>> Hi > >>>> > >>>> Looking at the data, I’m guessing the slower sample rate was done > last. > >> If > >>>> so, > >>>> you may not be letting the GPSDO’s “warm up” long enough. Most designs > >>>> take days (if not weeks) to get to their ultimate stability. > >>>> > >>>> Bob > >>>> > >>>>> On Dec 2, 2018, at 11:16 AM, Mark Goldberg <[email protected]> > >>>> wrote: > >>>>> > >>>>> I broke down and drilled a couple holes in the wall and moved the > >>>> antennas > >>>>> for both GPSDOs outside with a better view of the sky, but still > >> somewhat > >>>>> shielded by the house and a metal RV cover. The number of sats is way > >> up > >>>>> and the TruePosition is in normal mode most of the time. The ADEV is > >> not > >>>>> improved and actually measured worse. I took data for twice as long > at > >>>> two > >>>>> sample rates, 0.6816 sec and 5.453 sec. The ADEV from the slower > sample > >>>>> rates is lower, possibly due to the measurement method, but they all > >> are > >>>> in > >>>>> the same order of magnitude. I've named the Heather screen captures > for > >>>> the > >>>>> inside and outside locations and shown one combined Timelab capture. > >>>>> > >>>>> Regards, > >>>>> > >>>>> Mark > >>>>> > >>>> > >> > <TruePos_Outside_Ant_Heather.png><TruePos_Inside_Outside_Compare_Ant_Allan.png><TruePos_Inside_Ant_Heather.png>_______________________________________________ > >>>>> time-nuts mailing list -- [email protected] > >>>>> To unsubscribe, go to > >>>> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > >>>>> and follow the instructions there. > >>>> > >>>> > >>>> _______________________________________________ > >>>> time-nuts mailing list -- [email protected] > >>>> To unsubscribe, go to > >>>> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > >>>> and follow the instructions there. > >>>> > >>> _______________________________________________ > >>> time-nuts mailing list -- [email protected] > >>> To unsubscribe, go to > >> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > >>> and follow the instructions there. > >> > >> > >> _______________________________________________ > >> time-nuts mailing list -- [email protected] > >> To unsubscribe, go to > >> http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > >> and follow the instructions there. > >> > > _______________________________________________ > > time-nuts mailing list -- [email protected] > > To unsubscribe, go to > http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > > and follow the instructions there. > > > _______________________________________________ > time-nuts mailing list -- [email protected] > To unsubscribe, go to > http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com > and follow the instructions there. > _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com and follow the instructions there.
