[time-nuts] FMT on October 13
magnus at rubidium.dyndns.org
Wed Sep 26 05:11:07 EDT 2007
From: "Tom Van Baak" <tvb at LeapSecond.com>
Subject: Re: [time-nuts] FMT on October 13
Date: Tue, 25 Sep 2007 23:45:12 -0700
Message-ID: <000b01c80008$ca040aa0$0800a8c0 at pc52>
> ); SAEximRunCond expanded to false
> Errors-To: time-nuts-bounces+magnus=rubidium.dyndns.org at febo.com RETRY
> > I guess it depends on signal to noise ratio. With reciprocal counters, you
> > only need one period to measure as acurately as you need, but to have good
> > acuracy, you need very good S/N, as there is no filtering possible.
> > For example, the HP 5370 can measure a single period of a signal with a
> > resolution of 20pS (excluding noise and trigger imperfections), so excluding
> > these errors, the HP 5370 could measure a single period of a ~3.5 MHz signal
> > with 7 x10-5 precision (if I have not goofed the calculations....) More
> > periods improve the resolution proportionately to the quare root. Accuracy
> > is another matter.
> > Didier KO4BB
> The jitter on a single period is likely very, very high, especially
> if it comes over the air. That's why one usually measures over
> a duration of thousands or even millions of periods (effectively
> called the gate time).
> The HP 53132A makes something like 200,000 measurements
> per second. As a result, for a certain range of frequencies, it
> claims 12 digits/sec of resolution (vs. HP 5370 ~11 digits/sec).
As was discussed recently, didn't they do averaging such that they updated
value every second but the raw singel-shot resolution doess not give you the
12 digits/sec. There was a nice explanation in an article on how this was
not improving say ADEV measurements in the end.
More information about the time-nuts