[time-nuts] M12 100 Hz vs. 1PPS measurements

Bruce Griffiths bruce.griffiths at xtra.co.nz
Sat Dec 6 20:18:53 UTC 2008


WarrenS wrote:
> Tom, 
> This is great work and Information to see.
> I do think that this will inspire, and make if not a giant leap, at least a small step to better GPSDO performance. 
>
> 1st question:
> Does anyone have or can they tell me where to get a copy of the actual raw data Tom used for these plots?
> http://www.leapsecond.com/pages/m12-adev/
>
> The Link on his leapsecond site to the raw data seems to be missing.
> www.LeapSecond.com/logs/ - Logbook and raw data (ask) 
> A request to Tom, so far has gone unanswered.
>
> 2nd question 
> Can anyone provide me with the filter formula or algorithm or any information 
> regarding that is used by any of the better GPSDO units to calculate the ECV 
> Dac output voltage.
>
> I know about the simple basic 1st order averaging  filter used in some of the units, 
> and the even more basic but effective over under check to increment / decrement 
> an accumulator.  BUT it would seem to me there has got to be filters that give much 
> better result than either of these available and being used out there.
>
> Allan deviation (ADEV) and Modified Allan deviation (MDEV) 
> http://www.mail-archive.com/time-nuts@febo.com/msg07211.html
>
>
>
>
> I'm looking for just the basic formula used to process the phase error data before 
> being applied to the DAC. I am not looking for anything secrete or with all the extra 
> things some use to verify that the data is valid or appropriate before being applied 
> or thrown away.
>
> Thanks, for any information you can provide me.
> WarrenS
>
>   
Warren

The better GPSDOs also model OCXO, aging, tempco, and other effects such
as the effect of oven current (can be significant when Kelvin
connections to the EFC aren't available and the OCXO heater and EFC
share a common ground pin). Correction for these effects is important
when the GPSDO enters hold mode for any length of time.

A Kalman filter is probably more effective for such applications, given
sufficient training time, than a standard digital phase lock loop with
or without additional filtering.

Bruce



More information about the time-nuts mailing list