[time-nuts] GPS noise reduction

Bruce Griffiths bruce.griffiths at xtra.co.nz
Sat Apr 5 18:00:15 EDT 2008


Richard H McCorkle wrote:
> Hello,
> I looked at the dual standard data from yesterdays run, and
> once again the predominance of the short-term variations in
> phase occur simultaneously in both systems. As pointed out
> earlier a 3-corner hat is the wrong methodology as what I
> am doing is closer to a common view comparison. If I relax
> my goal to just attempting to reduce the GPS variations that
> are common to both data sets is there any way this can be
> accomplished without adverse effects on the long-term
> stability?
> When I flip between the charted data between the standards
> there are small variations unique to each standard but the
> GPS variations common to both data sets are 10 to 100 times
> larger than the unique variations. It would seem logical
> that some approach could be used with multiple oscillators
> disciplined to a common receiver to reduce the common data
> variations so the unique data would become more predominant.
> Would going to a triple oscillator design allow prediction
> of the common variations?
>
> Any thoughts?
>
> Richard
>
>   
Richard

Unless your oscillators are exceptionally noisy little will be gained.

A significant proportion of the observed variation between oscillators 
may be due to TIC quantisation noise and not oscillator instability.
Depending on your GPS timing receiver a higher TIC resolution may be 
necessary to reduce this effect.
If you are using a GPS timing receiver with a performance equivalent to 
that of an M12+T then a TIC resolution of 1ns or better may be required 
to reduce the observed variations between the 2 GPSDOCXOs.

Using 3 or more oscillators wont help as you can only determine the 
phase variance (using an N cornered hat technique) of each oscillator 
wrt the GPS signal and not the actual phase errors.
However since the GPS instability dominates for short averaging times 
the error in the phase variances of the individual oscillators will be 
very large.

Bruce



More information about the time-nuts mailing list