Author Topic: How an Atomic Clock Really Works, Zeeman Alignment  (Read 2257 times)

0 Members and 1 Guest are viewing this topic.

Online BrianHGTopic starter

  • Super Contributor
  • ***
  • Posts: 8275
  • Country: ca
    • LinkedIn
How an Atomic Clock Really Works, Zeeman Alignment
« on: October 31, 2022, 09:03:52 am »
How an Atomic Clock Really Works, Round 2: Zeeman Alignment....

Holly Sh!t, even after over 20 minutes of 2 separate free running cesium clock's 5MHz outputs keep dead perfect relative phase.  Are even 2 separate rubidium 10Mhz references that good?  (Answered below thanks to EEVB member 'ejeffrey')


« Last Edit: December 03, 2022, 04:46:03 am by BrianHG »
 
The following users thanked this post: RoGeorge

Offline MathWizard

  • Super Contributor
  • ***
  • Posts: 1772
  • Country: ca
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #1 on: November 19, 2022, 10:43:30 am »
And I complain about the cost of a good DMM or scope. Lab equipment like AC's cost a fortune.

So how good are the electronics in it ? What sort of tolerances do the component's have ? I wonder about the calibration and maintenance. Do they replace stuff every so many months or years like airplane parts ?
 

Online BrianHGTopic starter

  • Super Contributor
  • ***
  • Posts: 8275
  • Country: ca
    • LinkedIn
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #2 on: November 19, 2022, 04:44:12 pm »
Do any members here at EEVBlog own 2 rubidium 10Mhz references.
I would like to see the 2 10MHz clocks superimposed on a 2 channel scope to see if they keep perfect phase and for how long.
(Maybe I should have originally posted this thread in the TEA sub-forum.)
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 15156
  • Country: de
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #3 on: November 19, 2022, 05:10:07 pm »
I think they do show the superposition of the 2 atomic clocks, giving a rather boring static picture. Even with one of them not locked (so just the OCXO running) the picture is still relatively stable.
A Rubidium clock should be more stable than even a good OCXO and the superposition should thus be also stable to the normal observer (no time laps).
One can see the slight shift in frequency with the C-field (during demagnetization), but still not that much.

Besides the cesium tube and a good OCXO there should be not be many components that are that special. As cost was not a real concern there may be a few hermetic wet tantalums that now cost a small fortune, but they should not be essential, just more common some 50 years ago.  Normally there should be no need to replace things regularly - maybe a few more normal electrolytic capacitors if they are inside an ovenized part. Much of the cost is in the developement and adjustment / operation instructions divided by only a small number of systems build. The Cs tube has a limited, but still quite long lifetime.
 

Online BrianHGTopic starter

  • Super Contributor
  • ***
  • Posts: 8275
  • Country: ca
    • LinkedIn
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #4 on: November 19, 2022, 09:33:42 pm »
I think they do show the superposition of the 2 atomic clocks, giving a rather boring static picture. Even with one of them not locked (so just the OCXO running) the picture is still relatively stable.

Once both clocks were tuned and aligned, at 36:10 in the video, the 2 clocks were dead stable.
All the earlier scope shots were either GPS or comparing with the second not-yet-tuned clock.

I just wanted to see if comparing those 2 of those ebay Rubidium clock modules would keep their relative phase after an hour on the 10MHz out.  We have enough TEA members here that someone may be able to make the measurement.

The Cesium clocks, if kept at the same height above the ground, should keep relative phase for how long?
We got to be talking something like months to years if they were accurate enough to measure Einstein's when one was placed on a 747 flight back and forth to measure if the time dilation effect was real.
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 15156
  • Country: de
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #5 on: November 19, 2022, 09:46:58 pm »
If the clocks are accurate to some 1E-14 this would 1E7 second to get 1 period of the 10 MHz clock. So depnding on how good you are measuring the phase shift it may be hours before seeing a small difference of 1/1000 a period. So the test schown looking at the crossing point moving on the scope can be quite sensitive.

Operating 2 clocks near by may lead to syncronization: it only takes a rather weak interaction to make them lock to each other.

 

Online BrianHGTopic starter

  • Super Contributor
  • ***
  • Posts: 8275
  • Country: ca
    • LinkedIn
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #6 on: November 19, 2022, 10:27:18 pm »
Operating 2 clocks near by may lead to syncronization: it only takes a rather weak interaction to make them lock to each other.
I was well aware of this for simple analog free running oscillators, however, I did not think that such a shielded and well isolated Cesium tube within another shielded case separated by a reasonable few feet would succumb to the same problem.

What would be the mixing path?

My only guess would be that the 5MHz output is also somehow directly fed into the excitation of the Cesium as a necessary type of mixing of signals.  Then, an accurate enough externally sensed 5MHz signal may additive mix just enough to perform such a lock pulling the Cesium core along.  One would think such an expensive reference clock may have been designed to counteract this problem.
 

Online BrianHGTopic starter

  • Super Contributor
  • ***
  • Posts: 8275
  • Country: ca
    • LinkedIn
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #7 on: November 20, 2022, 02:19:58 am »
If the clocks are accurate to some 1E-14 this would 1E7 second to get 1 period of the 10 MHz clock.
Wow, at least 230 days for a possible single 360 degree phase shift at 5MHz.  :scared:
I see why everyone would just prefer using GPS receivers for their 10MHz reference.
(Ok I also realize GPS is cheaper and easier to get than Rubidium references.)
« Last Edit: November 20, 2022, 03:59:40 am by BrianHG »
 

Online ejeffrey

  • Super Contributor
  • ***
  • Posts: 4034
  • Country: us
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #8 on: November 20, 2022, 11:36:15 pm »
Do any members here at EEVBlog own 2 rubidium 10Mhz references.
I would like to see the 2 10MHz clocks superimposed on a 2 channel scope to see if they keep perfect phase and for how long.
(Maybe I should have originally posted this thread in the TEA sub-forum.)

I have several at work I can test.  Probably not until after turkey day.  My understanding is that basic cesium clocks are 1-2 orders of magnitude more accurate than rubidium oscillators.
 
The following users thanked this post: BrianHG

Online ejeffrey

  • Super Contributor
  • ***
  • Posts: 4034
  • Country: us
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #9 on: December 02, 2022, 07:35:39 pm »
I was able to measure this.

Yellow: This is from a SRS FS725 that warmed up for ~10 minutes before starting the test.  The locked light was on, but it might improve stability if it is on for a few hours/days.  It is also coming from a distribution amplifier that doesn't have the cleanest waveform

Green: this is from a separate SRS FS725 that has been powered up for probably a few weeks.  The output is taken directly from one of the instruments 10 MHz output.

There is no perceptible drift or jitter, the waveforms look completely stable on the scope.  I took two photos 8 minutes apart to show the slow drift: In that time, the green trace has slewed approximately 30 ns, or about 62 ppt.  That is well within spec, the accuracy at shipment is supposed to be +/- 50 ppt, and aging increases that further.

Since there is some definite offset, I don't think there can be injection locking, this is the real stability.  I will try again after the first source warms up for a few hours.
 
The following users thanked this post: BrianHG

Online BrianHGTopic starter

  • Super Contributor
  • ***
  • Posts: 8275
  • Country: ca
    • LinkedIn
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #10 on: December 02, 2022, 10:23:04 pm »
I was able to measure this.

Yellow: This is from a SRS FS725 that warmed up for ~10 minutes before starting the test.  The locked light was on, but it might improve stability if it is on for a few hours/days.  It is also coming from a distribution amplifier that doesn't have the cleanest waveform

Green: this is from a separate SRS FS725 that has been powered up for probably a few weeks.  The output is taken directly from one of the instruments 10 MHz output.

There is no perceptible drift or jitter, the waveforms look completely stable on the scope.  I took two photos 8 minutes apart to show the slow drift: In that time, the green trace has slewed approximately 30 ns, or about 62 ppt.  That is well within spec, the accuracy at shipment is supposed to be +/- 50 ppt, and aging increases that further.

Since there is some definite offset, I don't think there can be injection locking, this is the real stability.  I will try again after the first source warms up for a few hours.

Thank you ejeffrey.  At least now I know my claim of the Rubidium VS the properly aligned Cesium don't even come close to comparison.  They're truly in 2 completely different worlds.  (@36 minutes in the video.)

The video I pasted by CuriousMarc shows us there is a lot more involved in attaining the true Cesium precision via the 'Zeeman Alignment' and how to extrapolate a 5MHz reference from it.  Those 20 minutes of a dead perfect cross alignment between the Cesium clocks, the second one turned on and aligned separately, all together shows how difficult and still amazingly unbelievable how 2 free running oscillator references can achieve such a perfect reference of time.
« Last Edit: December 02, 2022, 10:30:16 pm by BrianHG »
 

Online ejeffrey

  • Super Contributor
  • ***
  • Posts: 4034
  • Country: us
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #11 on: December 03, 2022, 04:13:39 am »
I came back a few hours later and the frequency difference  between the two had stabilized at very close to 5 ns/ minute.  The stability is considerably better than that and they could be recallibrated for better accuracy but it's nothing like a cesium clock.
 
The following users thanked this post: BrianHG

Online BrianHGTopic starter

  • Super Contributor
  • ***
  • Posts: 8275
  • Country: ca
    • LinkedIn
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #12 on: December 03, 2022, 05:10:26 am »
Here is the Cesium difference after 20 minutes: (*seen at 36min in the video)



It looks to be 1/5 of a horizontal time base.
Now, if I could only make out the H-time base dial setting on the scope and H-zoom settings shown a minute later in the video.  The we could calculate the ppt drift.
« Last Edit: December 03, 2022, 10:11:44 pm by BrianHG »
 

Online BrianHGTopic starter

  • Super Contributor
  • ***
  • Posts: 8275
  • Country: ca
    • LinkedIn
Re: How an Atomic Clock Really Works, Zeeman Alignment
« Reply #13 on: December 03, 2022, 05:53:33 am »
Here are the 2 images separate...
The second trace slanted like this: ' \ ' has moved ever so slightly to the right after 20 minutes.


« Last Edit: January 22, 2023, 04:09:58 pm by BrianHG »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf