[time-nuts] Primary Time Standards

[Frank Stellmach]

Hi,

I think, the correct definition for a "Primary (Time) Standard" is not 
yet given here, yet.
And I wonder, if there really exists an official definition, in form of 
a norm / a standard by BIPM.
There should be one, does anybody know?


I try to give some characteristics which have to be fulfilled all 
together, in the order of importance (think so):

1) A "Primary Standard" has to be accepted and defined by the 
Metrological Communitiy, i.e. by all the national standards institutes 
joining the BIPM. For time and frequency, its the CCTF which decides 
over the definition.

A certain value is assigned to the Primary Standard, and an official 
definition is given.

2) A primary Standard is state of the art, it is currently the best one 
in terms of reproducibility.

It must not be the most stable one (like Cs-beam/fountain compared to 
H-maser)

What is accepted as a Primary Standard today, will be a Secondary one 
only by tomorrow,
like the Rubidium, or even Quartz oscillators, which once were 
state-of-the-art  "Primaries".

Which standard is better, can be measured and judged by comparing the 
"Primary" against the "Secondaries", and by comparing an ensemble of 
Primaries against each other.

3) The known side effects or influence parameters (like magnetic field, 
pressure, temperature, particle motion, gravity, quantum effects, etc.) 
have to to be identified theoretically and experimentally, which means 
they are well known by magnitude and understood by a physical model. 
Experiment and theory have to fit.

These side effects then can be effectively removed by either cancelling 
the influence parameter (e.g. in a Cs standard to zero the external 
magnetic field), or by evaluating the disturbance to calculate an upper 
boundary of the total deviation, or by making precise corrections to the 
output.

4) The Primary should not be an artefact (as the "kilogram", currently), 
but should be defined and reproduced by basic physical effects, 
preferrably quantum standards, which can be reproduced everywhere in the 
universe, simply by a construction recipe, instead of sending an 
artifact around.



So to answer Jims questions:

A Rubidium standard is not a primary any more, because

- it is less reproducible than Cs, due to the ~100 times higher 
susceptibility of the inner energy levels to magnetic fields.
That means:  External magnetic fields can be cancelled in both systems 
in the same way.
But this residual field,  produces a much smaller output deviation in 
the Cs clock.
- The metrological community has redefined Cesium as the new primary 
many years ago, instead of the former Rubidium.


The TAI is theoretically defined by the SI definition, whereby the UTC 
also accounts for the earth rotation by leap seconds.
In theory the TAI / SI second is ideal (zero uncertainty)

The TAI /  SI second is practically realized by the complete ensemble of 
Cs Primaries in the National Standards Institutes worldwide.
This "real" time scale is not perfect, as "side effects" on  Cs are well 
known and cannot be eliminated completely.
The reproducibility of the ensemble is around 2e-15 only..

But being the Primary Standard for the second, I think, this ensemble is 
simply defined to have zero practical uncertainty also, at least after 
the correction table is published.

The trimming of the Cs clocks is intended for a controlled cancellation 
of those side effects, i.e. magnetic fields, but not for a sort of 
calibration!
The different clocks principally cannot be adjusted against anything 
else, because nothing better exists at the moment.

If the members of the ensemble all deviate from the total average in the 
boundaries of error calculations only, then they are "Known Good". The 
statistical model of the complete ensemble gives an idea about the 
overall stability and reproducibility of the second, and also detects 
"Bad" clocks with systematic failures.

On the other hand, it is not a good and accepted metrological practise 
at all , to tune one or more of those "good" Cs clock artifically close 
to the average. This would simply spoil the independance of the 
ensembles members.


Today, the Cs is still the Primary Standard, although there might 
already exist more reproducible physical systems (e.g. Hg ion clock,???)


The H maser is much more stable (short term), but not as reproducible 
(long term) as Cs clocks.
That means,  it is obviously not possible to get an ensemble of H-masers 
that close together "by design only" as it is the case with Cs.

Therefore, H-masers cannot be Primary Standards.


Frank