[time-nuts] Primary Time Standards
magnus at rubidium.dyndns.org
Thu Jul 14 13:00:27 UTC 2011
On 14/07/11 04:26, Jim Palfreyman wrote:
> Hi All,
> I've just realised I don't understand something. Something quite basic.
> Primary Standards are ones which don't have to be calibrated against others.
> My understanding is that Caesium and Hydrogen masers are Primary Standards
> (in our field).
In the ideal world...
> Secondary Standards are calibrated against the Primary Standards. My
> understanding is that Rubidium is an example of a Secondary Standard.
> So I can calibrate my Rubidium clocks by adjusting the C Field. All good.
Concept wise yes...
> But why is it that Caesium Clocks and Hydrogen Masers have an adjustment
OK, let's look at the factors which play in here...
We have two issues here... stability and frequency correctness.
The rubidium gas cell standards (where the gas cell and not the rubidium
vapor is the relevant thing) has several frequency pulling mechanisms
such as cavity pull, wall-shift, ligth-shift, buffert-gas-shift etc.
etc. This makes the gas cell standard (rubidium or whatever) unsuitable
for realization of the SI second, but it is pretty stable for the size
and power and has found good use in many applications due to this fact,
not the least in telecom situations.
The active hydrogen standard (there are passives as well) has wall-shift
and cavity pulling among is systematic errors. It doesn't have a buffert
gas so that pulling effect isn't there. Automatic cavity tuning has
improved cancellation of the cavity effect and the active hydrogen
standard has a very high mid-term stability of the classical standards.
The cesium beam standard also has systematic shift components, but there
is no wall-shift and no real cavity pulling. However, phase errors in
the Ramsay resonators can introduce shift in frequency.
All these standards also has shifts due to things like black body
temperature, doppler effects etc.
Another common shift is due to magnetic fields. The ideal frequency of
these atoms is with no magnetic field (aka C-field). However, in reality
this they always have a C-field. However, modern commerical cesium beams
have no manual C-field adjustment as the early beams had, but rather a
control-loop which looks at the side-band responses and servo the
Rabi-splittings to a stable C-field. For a certain C-field the shift of
the Ramsey response for the fine-grained middle is predictable and the
shifted frequency is used for reference.
Together with a number of other control-loops to help stabilize the
frequency shifts and maintain noise sufficiently low the modern cesium
is much better as a primary standard.
So older commercial cesium beams where not really that good for your
ideal primary reference... but good enought to be very useful for many
> And what about the clocks used to determine UTC around the world? Do they
> have an adjustment facility? What are they adjusted to? Wouldn't that make
> them Secondary Standards?
You are confusing the time of TAI and realization of SI seconds with the
adjusted time-scale of UTC, which when needed jumps a second to maintain
the UTC-UT1 difference within bounds. The frequency of the clocks does
not change by this, it's the phase for UTC.
> Now I'm aware that the "average" of those clocks is UTC, so are those clocks
> adjusted regularly to get closer to that average?
Your 10 MHz can be adjusted to be closer to TAI/GPS if you like. It is
the digitis of the display which leap-seconds changes.
You want your frequency to follow TAI and your time follow UTC. If you
let your frequency adjust to UTC then it will be following coarsely the
earth rotation as represented by UT1 rather than atomic time of TAI. I
know it sounds confusing, but there it is.
> I'm sure someone can clear this up for me.
... or confuse further. Let's see.
More information about the time-nuts