[time-nuts] Z3801 lockin behaviour

[Magnus Danielson]

From: "Tom Van Baak" <tvb at leapsecond.com>
Subject: Re: [time-nuts] Z3801 lockin behaviour
Date: Sun, 23 Jul 2006 17:22:25 -0700
Message-ID: <001001c6aeb7$3dfd5aa0$dc0ff204 at computer>

> > Hi!
> >
> > I have powered up my Z3801A and monitored the 10 MHz on a counter (with
> fancy
> > histogram and statistics) with the Cs as reference. The Z3801 is moving
> around
> > quite a bit and the SatStat shows that it hasn't really converged towards
> > precission (the 24h time error prediction). Infact, it drops out of lock
> into
> > holdover every now and then. When it jumps in or out of hold over, a
> "large"
> > (50 mHz) frequency jump is clearly vissible. I have about 1 mHz of Allan
> > Deviation and about 4 in Standard Deviation.
> >
> > I have made comperative measurements of my counter setup by hooking it up
> to my
> > Rb, and that is noticeably lower in noise.
> >
> > Is this to be expected or not?
> >
> > Any thoughts & recommendations?
> >
> > Cheers,
> > Magnus
> 
> While watching the syst:stat screen did your Z3801A
> already complete its multi-hour site survey and the
> coarse/fine frequency adjustments?

Yes. I explicity had it redo the survey to make sure it was running of the
current antenna position.

> If it's still in a
> start-up acquisition mode ignore the 10 MHz output.

It was ramping like hell until the oven stabilized.

> If it doesn't lock after a few hours, I'd suggest a full
> reset. Also check your signal levels, and elevation
> mask. Did it get your lat/lon/alt correct?

Seems alright. Get SS values which seems good, occassionally above 200 but
rarely below 100. Usually track 5-6 satelites but if it had been a 12 channel
receiver I would have had 12 satellites tracked over the 10 degree elevation
mask.

> The predicted time error won't settle down for a few
> days (it's a long time constant), so don't worry about
> that quite yet.

OK. I'll let it simmer for a few days then.

> Not sure what you mean by "The Z3801 is moving
> around quite a bit". You should get 10 solid digits
> of accuracy out of a Z3801A at short sample times.
> That grows to 11 or 12 over several hours; 12 to 13
> over several days (at which point you won't quite be
> sure if the errors are your counter, your Z3801A, or
> your cesium).

The end goal is to compare the Cs, Rb and GPS at various time constants, in
which case I will be able to resolve the phase-noise of each of them.

Right now I'd say I only have a solid 9 digits.

Cheers,
Magnus