[time-nuts] Setting Rubidium to match GPS source
magnus at rubidium.dyndns.org
Fri Sep 28 12:10:35 EDT 2007
From: "Pete" <peterawson at earthlink.net>
Subject: Re: [time-nuts] Setting Rubidium to match GPS source
Date: Fri, 28 Sep 2007 09:07:29 -0600
Message-ID: <000e01c801e1$4da7efa0$0200a8c0 at BASE1>
> Using oscilloscopes to measure relative phase has been widely
> used for a long time. The setup appears in many H-P Op/Srv
> manuals in the adjustment procedure section to calibrate the
> instrument timebase to a shop standard.
> But, it's not a useful technique for settings more precise than
> 1 or 2E-10 in the best situation.
> The process degrades badly as the 'scope sweep speed
> increases past 10ns/div since a 10MHz sinewave slope is
> approaching a rather flat line. Watching 2 rather flat lines
> for precise zero crossings isn't as simple as it seems. In
> addition, any noise/drift or outright shifts in vertical position
> obscure the result & aren't unusual during a 16 minute window.
This is why you gain up both channels more and more. This converts them more
into vertical lines again and you can zoom in more. However, my experience is
that evenetually the 1/f phase-noise will become so apparent that the average
motion due to frequency error becomes harder to detect by eye.
The method is thus still limited. We have better methods available (i.e.
various TIC systems).
> Delayed sweep operation just adds more uncertainty, so
> you end up waiting for the phase difference to drift to the
> point that both traces are displayed in the same window.
> But now, they won't stay there for 16 minutes, unless they're
> already VERY close to matching. The whole process demands
> that you walk your way down to locking up the 2 sources.
> It's more painful than it sounds!
Can be. The delayed trigger window as such only require a short-term stability
and for 10 MHz we are talking about up to 100 ns. Just as with a time-error
estimator for a TIC.
More information about the time-nuts