[time-nuts] Droitwich Frequency Error

[Magnus Danielson]

Hi Andy,

On 2020-10-18 16:07, Andy Talbot wrote:
> Hi Folks
> I've just joined this list, but have had an interest in accurate frequency
> measurement for decades, ever since finding a fully operational HP5061A in
> skip with what appears to be plenty of Cs life left in the tube.
First of all, welcome!
>    It only
> gets turned on when needed here.    Anyway :
>
> This morning, mainly to prove something it to myself I made a plot of the
> off-air UK Droitwich LF transmitter whose carrier is supposed to be a
> national standard - although I believe it uses a Rb source that is
> periodically updated with a Cs one - manually.

As I recall it, it runs of an GPSDO with a cesium as hold-over.

I do not consider it to be the national standard, but it is one of
several handy sources of off-the-air frequency sources in UK and
valuable as such. For national standard, I would think NPL's UTC(NPL) be
relevant and well worth a visit.

> I have a Leo-Bodnar GPSDO had been turned on only some 20 minutes before
> the plot was started, as teh reference into a custom own-design LF receiver
> that gives a digitised baseband I/Q output to a PC.  Its LO is tuned using
> an  AD9852  48 bit DDS clocked at 10MHz and set by an algorithm that allows
> a micro-Hz but completely deterministic tuning error - it was this error I
> was trying to show by doing a plot of Droitwich off air using the GPSDO as
> a reference.
As been already pointed out by others, GPSDOs take some time to settle,
and 20 min is far less than what I would be satisfied with. I recommend
you to keep it on.
> The plot at http://www.g4jnt.com/DropF/clipboard_202010181333.png shows the
> result - the phase is the red line.    My calculated tuning error at this
> frequency is 3uHz, but there is a slope to the line showing around 20
> degrees of phase shift in a little under 2 hours.   This corresponds to
> about 7.8uHz error. If my tuning error contributes 3uHz of this, that's
> still 4.8uHz on 198kHz , or 0.024 PPB.   It was well after sun rise, and
> the transmitter is only 100km away from me, so oughtn't to expect
> propagation anomalies
Since you have variations due to 24 h period of sun, you for sure want
to measure multiple periods to separate out those variations.
> My question,  does anyone know if a GPSDO, has any inherent drift over its
> first hour or so?   I wouldn't have thought there was any mechanism for
> that, but who knows.
It does. If it bothers you is a mater of scale. The oscillator heating
up and drift, that takes time for the lock loop to fully track in, and
it is only as the oscillator settles that the PLL loop can settle. The
PLL loop also start of with arbitrary phase and frequency errors, and it
takes time before those errors die away, and this depends on the
time-constant, and well those is kept long for good performance
typically. This is why patience is needed. Fire up and let run. Another
aspect is that you want the position average to work out fine-grained
position, as this reduces the wobble in phase that incorrect position
produces.
>
> Finally, I guess 2.4E-11 isn't really outrageous for a rubidium source,
> even one that is 'supposed' to be corrected from 'time-to-'time' Or  Is it?

Not completely, if that is the real offset. This is the real
time-nuttery, trying to factor out things in once own setup to be able
to measure deeply into the abyss. You learn so much just trying, and it
is loads of fun. With that, again welcome! :)

Cheers,
Magnus