[time-nuts] Zeeman frequency oddness

[Magnus Danielson]

Hi John,

Don't know if I address your real issues, but I want to add some more 
pieces to the puzzle.

On 10/27/2012 09:16 PM, John Ackermann N8UR wrote:
> Since I recently got an HP Cs degausser (thanks, Stijn!), I though I'd
> go through the whole setup routine for my 5061B/004 and see how close
> the C-field-via-Zeeman setting would bring me to GPS-derived frequency.
> It turned out to be an interesting and puzzling exercise.
>
> This 5061 seems to be in perfect working order -- quick lock, good meter
> readings, and measurements indicate frequency within parts in e12. I
> don't have any reason to believe that it's not tuned or working
> properly, except for the Zeeman-setting results.
>
> The problem is that when I tune the audio source around 53.53 kHz, per
> both the manual and the sticker on the door, I don't see any change in
> Beam I at all. Nor do I see anything at the alternate frequency of 42.82
> kHz.
>
> Instead, I see the expected three peaks -- primary with a smaller
> secondary on either side -- at about 48.21 kHz, which doesn't show up
> anywhere in the literature I've found.

There are in total 7 peaks, you want the center peak of those.

Tom has made measurements:
http://leapsecond.com/images/cfield.gif

I have done the same to one of my tubes, but I don't recall where I have 
that flimsy picture, so the above is a good start.

> Below, I've cut and pasted a years-old message from TVB and Corby that
> explains the Zeeman frequencies. I've measured the synthesizer output
> and it's nominally 12.6317725 MHz, which per that message should
> correspond to a 53.53 Zeeman. Where 48.21 kHz comes from, I have no idea.

The separation of the peaks depends on the C-field value you have.

For low C-field strength, the side-peaks separate my the square of the 
C-field (B), and the cluster shifts gently linear with the C-field.

You want to spread the side-peaks out, such that they do not confuse 
your measures. Also, looking at the above you have the 7 Rabi 
distributions, and on top of those the Ramsay fringes, at which you want 
to lock onto the center one. It may be interesting to learn that certain 
systematics skews the shape of these, and thus causes a systematic 
miss-tuning, so great care is taken to reduce that effect when 
manufacturing the tube.

Look at the above C-field plot again and you see how the side features 
move with different C-field settings.

Modern "digital" caesium clocks measure the side-features in order to 
Servo the C-field into a stable value, and hence also stabilize C-field 
drift out of the equation first degree.

> I'm using a Rigol arbitrary function generator locked to an external
> reference as the audio source, in sine wave mode. I know that's not the
> cleanest device in the world, but the wave doesn't look too bad on my
> scope and a counter indicates the frequency is what the dial says. As I
> adjust the audio amplitude, the beam current responds, and I see a peak
> at around 500mV, which the 5061B manual says is correct.
>
> Any ideas why I might be seeing this very off-the-wall result? Could
> distortion in the audio source cause something like this? I'm more
> inclined to blame technique or gremlins than the 5061B -- again,
> external measurements indicate that the thing is tuned correctly and
> operating properly, just having this goofy Zeeman response.

You want a clean source, as spurs or distortion would cause you to look 
at multiple points in the spectrum at the same time and you would get 
the combined result of those features. Still, the distortion values 
doesn't have to be stellar to get decent readings.

Cheers,
Magnus