[time-nuts] Common sky pps errors for any GPSDOs?

Matt Ettus boyscout at gmail.com
Tue Jan 6 07:40:10 UTC 2009


On Mon, Jan 5, 2009 at 10:59 PM,  <SAIDJACK at aol.com> wrote:
> Hi Matt,
>
> I must admit I don't fully understand your requirements. Are you looking  for
> correlation between errors, or absolute UTC accuracy, or short term
> jitter/wander?
>
> If you have two systems with self-surveyed antenna positions, you will
> likely have 1 - 10 feet of antenna height error in the self survey on the  Motorola
> timing receivers (typically).
>
> This position-hold error in itself will give you much more than 140ps  error
> (offset, drift, wander) between the units as satellites fade in and  out of
> solution, even if the units are sitting right next to each other and are  seeing
> the same systemic GPS errors.
>
> For example, let's say both units share the same antenna, and after
> auto-survey one reports it's height as 10 feet MSL, the other unit as 15  feet MSL
> (M12M's have easily more than 5 feet height error after  self-survey).
>
> So if you now compare the outputs of the units, satellites  directly overhead
> could cause a 5 feet, or ~5ns error, while sats at  the horizon will not be
> affected by the height error, but rather the long/lat  errors (which are much
> smaller).
>
> So unless you have a perfectly surveyed antenna position stored in the two
> receivers (to within < 1 foot) you will get GPS systemic errors as well  as
> timing errors due to position error - especially due to antenna height  errors.
>
> When we say units typically have 25ns unit-to-unit variation on the 1PPS on
> un-calibrated units, then I believe most of this is caused by the auto-survey
> position errors of the GPS receiver. One could get much better performance by
>  manually entering the exact position-hold position of the antenna, and then
> calibrating for antenna cable delay (in 1ns steps).
>
> This seems to yield down to 2ns performance as reported by
> Motorola/Synergy/NIST with careful calibration, and using a "proper" GPS timing  antenna with
> multipath choke-ring etc.
>
> But again, this requires a perfectly surveyed antenna position, as well as
> offset correction due to antenna cable length delay.
>
> There are also antenna cable length variations due to ambient temperature
> changes :)
>
> Bruce and others had discussed these errors not too long ago. 140ps error
> (or 70ps per GPSDO unit) may be possible on a long antenna cable just  due to
> temperature changes on the cable..
>
> Lastly, our units seem to have a residual PLL tracking noise floor  of down
> to 1.9ns rms when using a good double oven OCXO as can be seen  on the unit
> running in Mexico using a properly surveyed antenna position:
>
>   _http://resco.ucol.mx/Fury/gpsstat.htm_
> (http://resco.ucol.mx/Fury/gpsstat.htm)
>
> Getting 140ps matching offset error between two different units'  1PPS
> outputs may be tough to achieve.
>

I see what you are saying here.  I guess that it will be necessary to
have some way to calibrate out the long term phase variations from the
received signals just like in VLBI.

I think it would be interesting to see the raw data from a time
interval counter on the PPS of two identical GPSDOs sharing an
antenna.  If anyone has that sort of data, maybe over a day or two,
I'd love to get a copy.

Thanks,
Matt



More information about the time-nuts mailing list