Hi,

On 2021-03-25 17:29, Attila Kinali wrote:

Look at the BIPM data, as each clock is identified and you have
continuous measurement of a whole bunch of 5071s there so you can even
line them up and make statistics of the spread that way, just a bit of
number crunching away. There may be articles available also that gives
relevant hints.

Cheers,
Magnus

Attila Kinali writes:

As I understand it, cesium beams are considered "primary" standards
because once the ADEV hits the floor, it stays down there ?

--
Poul-Henning Kamp      | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG        | TCP/IP since RFC 956
FreeBSD committer      | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.

The ADEV of the 5071 follows the typical curve of
improving at a constant rate vs increasing Tau
until it levels out due to limiting by flicker.  AFAIK,
this flicker noise never turns back up.  There is
no observable aging either.  Nor are there any
observable environmental effects.

The 5071's traditionally counted for 85% of the
weight in the TAI weighted average.  You might as
well start by asking what is the long term stability
of the TAI.

Just based on what you are saying, your client reminds
me of many consulting clients I have had who started
out asking the wrong question.  I usually responded,
not by attempting to answer the question, but rather
explain why it was the wrong question and then proceed
to figuring out the right question.  Just a comment;
maybe that is not what is going on in your case.

Rick N6RK

On 3/25/2021 9:29 AM, Attila Kinali wrote:

Dear Attila

We have clock comparison data going back 25 years so you could DIY.
But I thought I sent that to you a couple of years ago?

Cheers
Michael

On Fri, 26 Mar 2021 at 3:34 am, Attila Kinali attila@kinali.ch wrote:

Hello Attila

I realised that I had recently made a plot of the kind you were after
recently for our quality system documentation.

The plot is for a standard tube. The UTC plots are from CircularT. The
local Cs-Cs plots are for a local comparison of the UTC clock with another
Cs; I have divided by sqrt(2). I have used about 4 years of data. The Cs is
free running during this time.

I don’t know why the UTC data seems to go lower - perhaps there’s some
averaging in Circular T. Nor do I know why it starts to walk off. The utc
curve starts to walk off at one year which may mean something.

Cheers
Michael

On Fri, 26 Mar 2021 at 3:34 am, Attila Kinali attila@kinali.ch wrote:

Hi,

On 2021-03-25 19:21, Poul-Henning Kamp wrote:

"primary" standard means different things in different context.

In telecom, it means it adheres to ITU-T G.811 specifications, which
effectively puts within 1E-11 in maximum frequency error, which is what
analog cesiums can deliver. Most of the cesiums we attain as hobbyists
was designed to meet this spec. The underlying specification driving it
was to keep data-slip rate between two operators down to once in 70
days. It was reasonably achieveable with the technology at hand and for
the total cost so I think it was fair.

In metrology "primary standard" has a complete different meaning, and in
practice all clocks we hobbyists gets to have would not fit, they would
all be more or less good "secondary standards".

The sales people for vendors will be happy to underblow the
understanding of you being able to buy and have your own "primary
standard". If it where, you would not be needing traceability to
anything else, but you end up needing to have that anyway, and in
reality the "primary" reference is actually one of the few that
contributes to the international realization. I've seen a few of those,
but not in my basement.

The "primary reference" is not about ADEV hitting the floor, all devices
do that (for a strict definition of what should be measured in ADEV).
It's about the context one consider it "primary".

It would be cool to say one has a "primary standard", and depending on
context I have several working or none. Coolness aside, when talking to
metrology folks and national metrology labs, I might have some clocks,
but I do not call them "primary reference", because most of them does
not have that either.

So, I think we should be careful with the term, it's get thrown around
too lightly.

Cheers,
Magnus

Maybe HP5071s will get their ADEV when optical clocks will officially rule...

On Fri, Mar 26, 2021 at 9:40 AM Magnus Danielson magnus@rubidium.se wrote:

You can do the ADEV just fine using Circular T. Long-term 5071 does not
dominate the TAI phase/frequency, but they do dominate for stability.

This is the difference between EAL and TAI, the 5071's go into EAL, but
EAL is then frequency corrected with a handful of primary references
(cesium fointains) into TAI. EAL just try to be as stable as possible,
without careing too much about phase and frequency.

Cheers,
Magnus

On 2021-03-26 11:00, Azelio Boriani wrote:

I’ve attached an old paper of mine, from PTTI-2000.

I’ve also attached the latest BIPM file of TT.  If you want long-term stability it is best to use that file to compute TT-EAL and TT-TAI and then re-reference the clock data you get from the BIPM.  The BIPM algorithm for EAL/TAI/UTC changed in 2014 so if you stick with TT you re fine.  Since it is computed every 10 days, you might want to spline-interpolate to get the in-between points.  Or just interpolate.

FWIW here are plots of TAI-weight from the 'w' file for 2021.01. The
5071A's are at most 8%. Masers are >87%.
These sum up to ca 95% - not sure where the missing 5% is..
Going back to much before 2014 might show larger weight for the 5071As - I
can post python code for the figures if someone is eager to try...

AW

[image: TAI_weight_clockype.png][image: TAI_weight_lab.png]

On Sat, Mar 27, 2021 at 4:03 PM Demetrios Matsakis via time-nuts <
time-nuts@lists.febo.com> wrote:

Hi

I would guess that fountains would be in there somewhere ( at least in the
2021 version ).

Bob

Anders,

I am surprised to find that hydrogen masers was so dominant these days.
It only shows that the ALGOS improvement that handled the linear drift
that allowed their inclusion have been very effective and should have
improved the stability.

I am also surprised to find my local NMI SP clocks in as fourth weighted
laboratory. I should rig up so they have another maser to measure.

Cheers,
Magnus

On 2021-03-28 15:03, Anders Wallin wrote:

Bob,

Check the "Ground state" section of the graph.

Cheers,
Magnus

On 2021-03-28 15:15, Bob kb8tq wrote:

The pie charts that Anders created show clocks in the UTC "flywheel".
That excludes most research fountains and optical clocks because they
can't or aren't run continuously.

Of the 427 clocks [1] in that data set there are 247 5071A (58%) and 167
H-masers (39%) and 6 fountains (<1.4%). So the short answer to your
question is that there are only 6 fountains in the chart: 4 Rb fountains
at USNO and 2 Cs fountains at PTB.

When playing with charts, it's important not to confuse the number of
clocks with the weighting of clocks. And even the weighting is
misleading because for practical reasons any clock's weight in UTC is
capped at about 1%. All 4 of USNO's Rb fountains are at this level, for
example. About 60 of the H-masers are also at this level. And none of
the 5071A; not even close.

If you are a 5071A sales person you are likely to emphasize that 60% of
the clocks in UTC are 5071A. You are less likely to mention they
contribute only 8% of the weight these days.

If you are in H-maser sales you can claim 40% of the clocks in UTC are
masers and also they contribute 88% of the weight.

It's a good time to be in Microchip clock sales: they now own both the
hp 5071A and the Sigma Tau H-maser manufacturing plants. Yes, they still
make PIC chips too ;-)

/tvb

[1] https://webtai.bipm.org/database/clock.html

• clock type 35/36 are 5071A
• clock type 40/41 are H-masers
• clock type 92/93 are Cs/Rb fountain

On 3/28/2021 6:15 AM, Bob kb8tq wrote:

The BIPM’s new algorithm, implemented around 2014, weights clock by their predictability, and also estimates the frequency drift of the masers.  That’s why masers have risen to their level of prominence.  But masers do have all kinds of variations at some level, which I easily saw when I worked at the USNO and had decades of data uncorrupted by time transfer noise.  I was going to publish a paper in 2019, but I didn’t because getting approval was impossible.  It will be interesting to see what you guys can come up with using BIPM data to create long time series of frequency variations for cesiums and masers too.  Be careful because the BIPM data may not tell you when a cesium beam tube was changed, or a maser underwent repair.  Or a clock was moved.  So I would treat any data gap as a new clock.

Maybe one year ago Microsemi started advertising lower drift in their masers, and I suspect that is because they found a way to compensate for the initial drift so the user doesn’t see it.  (I have no idea if this is really true.)  Years ago in a paper I published with Mike Garvey and Paul Koppang I found the frequency drift noticeably decreased after three years.  Extending my unsupported theory, that might mean new masers pick up a drift over time.  But components change, and whatever they do can be modeled.

Another thing about the meaning of weights is that clocks used to be characterized through comparison with the EAL timescale, which does not involve primaries.  But the new algorithm characterizes them against a TT-guess which carries the frequency of the primaries.  So that in effect degrades masers, cesiums, and Rb fountains to the status of frequency interpolators between the primary standards.  (This is an observation I have made many times and no one has contradicted me, so I guess it is obvious.)

Hoi Demetrios,

Sorry for the late answer, last week was a bit crazy.

On Sat, 27 Mar 2021 09:58:30 -0400
Demetrios Matsakis via time-nuts time-nuts@lists.febo.com wrote:

Thanks! I hadn't seen this paper yet.

I actually wanted to avoid extracting the data on my own,
as there are a lot of pittfalls in this. And I'm sure I am
not even aware of half of them.

Your paper contains part of the information I was seeking.
And quite a bit that I wasn't but is very intersting. I need
to spend some quality time with it :-)

Thanks a lot and have a nice week!

		Attila Kinali

--
The driving force behind research is the question: "Why?"
There are things we don't understand and things we always
wonder about. And that's why we do research.
-- Kobayashi Makoto