[ntp:questions] Re: Best way to harness multiple stratum 1 servers

John Ackermann N8UR jra at febo.com
Fri Apr 7 01:17:34 UTC 2006


Darryl Otzen said the following on 04/06/2006 08:08 PM:

> For comparing Cs oscillators I use an old fashioned analogue phase meter
> coupled with a chart recorder.
> The meter is centre zero with +/- 5 degree FSD.
> Once I'm happy with the performance of the analogue outputs of the
> oscillator I'd start looking at the PPS.
> How accurate is your TIC?
> Feed the same signal to both inputs, check if there is any unstable
> measurement error.

Hi Darryl --

I have also used an HP 8505 (I think that's the number) vector voltmeter
which has a phase meter, feeding the chart recorder output into an ADC
and then logging the data.  That does work well -- as long as you don't
slip a cycle!

The main issues with the TIC are resolution and jitter; when you're
measuring intervals of ~10us (which is what I use) the stability of the
timebase aren't really important because any drift over such small
intervals is way below the resolution of the counter.  Be that as it
may, I normally use an external timebase, typically an Rb standard which
has better short term stability than the Cs.

Theinexpensive HP 5334A counter I dedicate to long-term logging has a
2ns resolution, but I'm using it in 100 sample average mode, which
yields effectively 200ps.  Against a tau of 100 seconds, that's about
2x10e-12.  There is probably still some counter jitter in that, so the
actual usable resolution is maybe 5x10e-12 in 100 seconds.  For long
term measurements, with averaging times of an hour to several days, the
resolution becomes even less of an issue -- 10ns/day is about 1.15x10e-13.

I haven't measured the single-shot jitter of this particular counter,
but based on other HP counters I have looked at, the jitter is typically
a couple of times the resolution (e.g., for 2ns single shot, jitter
might be in the 5ns range).

John




More information about the questions mailing list