Eevblog #235 - Rubidium Frequency Standard

[metalphreak]

I didn't notice any sync issues until I read the comments, then I noticed a very slight lag between the audio and Dave's lips moving. I would say its maybe 50ms or so, nothing really serious.

[king.oslo]

I am trying to work out how much the thing drifts per year measured in time. Datasheet says it has long term stability of 2 x 10^-11. But 2 x 10^-11 of what? 1 hz? 10mHz or whatever the physics pacakage is running at? Or something else?

Thanks.M

[ejeffrey]

It is a fractional change.  Since it has a 10 MHz output (10^7 Hz), a drift of 10^-11 is a change of 10^-4 Hz.  The 6 GHz internal microwave frequency will change by 0.06 Hz.  If you divide it down to a PPS (1 Hz) signal, that will change by 10^-11 Hz.  Use it to stabilize a frequency synthesizer and the output of the synthesizer will drift by 10^-11 of whatever the operating frequency is.

[tnt]

It's always a deviation from nominal value.

dev = abs( (f_real - f_nom) / f_nom )

and the absolute frequency doesn't matter ...

Imagine you have a 10 MHz output and a 1 Hz error, then if you derive a 1 GHz frequency from that source then you'll get a 100 Hz error.

[rf-loop]

I am trying to work out how much the thing drifts per year measured in time. Datasheet says it has long term stability of 2 x 10^-11. But 2 x 10^-11 of what? 1 hz? 10mHz or whatever the physics pacakage is running at? Or something else?

Thanks.M

If 10MHz (10 Mega Hz) have 2x10Exp-11 error, say example it is this amount too fast.
Its frequency is 10 000 000.000 2 Hz. (10.0000000002 MHz)

In rough this stability means error. But "long term stability"  - and I ask "how long"
How about short term..

Then, in what situation. In stabilized outside temperature, pressure, magnetic field etc.

If compare example cheap Efratom LPRO-101. (DATUM LPRO-101) these are factory adjusted inside +-5x10exp-11  error, same accuracy what example I have done as I have sometimes calibrated these to someone.) It means that frequency is (with given enviroment variables) between 9.9999999995 - 10.0000000005 MHz.
Of course if same error  in one Hz. It means  0.99999999995 - 1.00000000005Hz.
Now it can ask how many seconds need that it have (accumulated) one second error.

But in practice year level drift is more than +-2x10exp-11.

I have checked tens of LPRO-101 and some these FEI models also after they have take out from active original use.

After different time of use and diffrent sometimes also very hard enviroments used. Then I have checked (before my adjustments) they have all been typically well  inside +-10x10exp-10 error window. (10MHz under 10mHz error) Mostly drifted to up some mHz. (why?) No one can exactly tell how long they have used and how long after any calibration so this do not give much of information.
Also most of Rb need be 48-72 hour or even more on before check. (but well.. <30 minutes they are typically very near real working freq)


C-field setting may also drift.  (It is sensitive to magnetic field strenght and direction and it is also method for adjust it) Even earth magnetic field affect it.

Rb physic package have many mechanism for drift.   It is NOT primary standard as example Cesium or Hydrogen maser.

[don.r]

What is the lifetime of the lamp and is there any way to measure the life left after purchase? This would be my biggest concern. Considering you can get an Thunderbolt or equivalent for about $150, I'm not completely sold on these.

[wkb]

If compare example cheap Efratom LPRO-101. (DATUM LPRO-101) these are factory adjusted inside +-5x10exp-11  error, same accuracy what example I have done as I have sometimes calibrated these to someone.) It means that frequency is (with given enviroment variables) between 9.9999999995 - 10.0000000005 MHz.

What do you use to calibrate the Rb units?  Curious here..

[king.oslo]

So, how many years can i expect before it is off by 1 sec? It's off by 2x10^-11 every year.  I it correct that it will be off by 1 sec every (2x10^11)/2 yrs? Circa 100 000 000 000yrs?

Is my math wrong?M

[ejeffrey]

What do you use to calibrate the Rb units?  Curious here..

You can reference them to a GPS pulse-per-second source which has better stability over time periods greater than a day or so.  I am not sure what the mechanism for aging is, certainly the rubidium atoms themselves don't change.  Perhaps it is a drift in the offset voltage of the lock-in amplifiers phase detector, or some non-linearity in the dither signal applied to the microwave frequency.

SRS has some more information about how these things work: http://www.thinksrs.com/products/PRS10.htm.  They sell both the modules like Dave shows here, and also complete instruments with power supplies and multiple outputs for synchronizing a whole rack full of equipment.

[Blue]

The Stanford Research Systems make nice products. I've got a lock-in amplifier from them.
But I do not understand their price politics and I hate them for that: International buyers have to pay 10% more that US customers  .

Why? Do they hate foreigners so much that they have to screw them?

[rf-loop]

If compare example cheap Efratom LPRO-101. (DATUM LPRO-101) these are factory adjusted inside +-5x10exp-11  error, same accuracy what example I have done as I have sometimes calibrated these to someone.) It means that frequency is (with given enviroment variables) between 9.9999999995 - 10.0000000005 MHz.

What do you use to calibrate the Rb units?  Curious here..

Long time averagin method with two  HP Z3801  one HP/Symmetricom 58503A (it is nearly same as Z3801)
and  HP53131A  for statistical analyze so that I can think if can trust or not. Also I can follow with some poor Trimble Thunderbolt and 
With careful long time method this can easy go to this accuracy. If I see some disturbance in stability I just put this follow to garbage and take agen.  (of course also Rb under test/adjust need control with temperature. Normally I watch this plate temp where Rb under test is sitting.)
Of course Rb's under test/adjust need also "burn in" after maybe long time not powered. There are some small drifting maybe after long time not used. Minimum is 3-4 days. Rb have not so big this effect but example OCXO adjustment is more difficult. Just becouse ageing curve and maybe high retract if long time not used. Also it is littlebit unknown how it go agen to follow its nominal ageing curve. Every OCXO is just different.

About Rb lamp. Mostly it is not real problem if so not use these oldest Ball Efratom models.
I do not know who and why have started this Rb lamp paranoic.
I have not seen real lamp fail in example LPRO-101. But many fails what "looks like" lamp fail.
I littlebit think this lamp paranoic come from history when Rb lamps was littlebit different and also units price was high.

Today there is there mass produced Rb's and...
... just if cheap shirt come dirty buy new.

[JBeale]

I started a FAQ about this particular model, the FE-5860A (to be more precise, the cheaper of the versions currently available on the auction sites, since FEI made so many different variants with the same model number.)

http://www.ko4bb.com/dokuwiki/doku.php?id=precision_timing:fe5860a_faq

List of FAQ topics:

    FE-5680A Rubidium Frequency Standard FAQ
        What is a FE-5680A and why would I want one?
        Useful Links
        Parameter Specifications
        Connector Pinout
        Why doesn't the C-field adjustment trimpot through the hole in the side do anything?
        What is the range of frequency adjustment via RS-232 commands?
        What if my 5680A output does not lock up after several minutes?
        Can I get a square wave 10 MHz output instead of a sine wave?
        How can I get a 1 PPS output?
        What is the typical drift with temperature (tempco)?
        Long Term Frequency stability
        Input Voltage Requirements (official spec is 15 - 18 volts DC)
        Is a Rb frequency standard sensitive to external magnetic fields?
        Mechanical Dimensions
        Electronic Parts
        Undocumented RS-232 commands

[JBeale]

You can reference them to a GPS pulse-per-second source which has better stability over time periods greater than a day or so. [...]

In the long term, certainly GPS has better accuracy, but not necessarily in one day. Just for reference: at my location (poor sky visibility, and restrictions on antenna mounting) the only GPS I have that works, has as much as 400 ns of drift relative to a stable reference, at any given time (due to multipath issues I assume). Anyway, 400 ns in 24 hours is a relative stability of 5E-12, and at least one report from a surplus FE-5860A I've seen is showing less than 1E-12 drift after 8 weeks (though some units are worse).  If I want to be conservative and have my reference stability 10x better than the drift I'm measuring, I'd need to measure against GPS for over a month to be confident of a drift measurement at the 1E-12 level.

[wkb]

I started a FAQ about this particular model, the FE-5860A (to be more precise, the cheaper of the versions currently available on the auction sites, since FEI made so many different variants with the same model number.)

http://www.ko4bb.com/dokuwiki/doku.php?id=precision_timing:fe5860a_faq

Thanks!  I have a unit on its way from HKO so this might come in handy.  Been collecting URLs myself too but the centralised location of the FAQ is sure nice to have.

[mikeselectricstuff]

Extreme teardown of the physics package:

[king.oslo]

mikeselectricstuff, you creative genius!

Keep them coming. Next do a atomic battery teardown!  M

[Blue]

  Cool

can you repeat it with the lamp underwater instead of above the waterline? 

[JBeale]

Why does this come with an OCXO?

The "free gift" OCXO has nothing to do with the FE-5680A, and the odd frequency is not particularly useful for anything (unless you run your own cell base station, it's related to a CDMA sampling frequency I think). The guess is he just had some extra parts no one wanted, and wanted to sweeten the deal and distinguish from the several other sellers offering the same 5680 unit.

[EEVblog]

Extreme teardown of the physics package:

Excellent update Mike.
How did you get the close up shot of the bulb?
What camera are you using?, and did you shot the video under a magnifying lamp or something?
One of things I'm disappointed at with my Canon HF G10 camera is it's lack of ability to do good macro closeups.

Dave.

[mikeselectricstuff]

Extreme teardown of the physics package:

Excellent update Mike.
How did you get the close up shot of the bulb?
What camera are you using?, and did you shot the video under a magnifying lamp or something?
One of things I'm disappointed at with my Canon HF G10 camera is it's lack of ability to do good macro closeups.

Dave.

It's a Panasonic HDC-SD10 camcorder (£70 on ebay bought as faulty & fixed!).
It will focus close enough to get a 20mm wide object full-screen.
I think you can get macro add-on lenses for some camcorders

[EEVblog]

It's a Panasonic HDC-SD10 camcorder (£70 on ebay bought as faulty & fixed!).
It will focus close enough to get a 20mm wide object full-screen.
I think you can get macro add-on lenses for some camcorders

Thanks.
Yeah, have to look into a good macro lens for the Canon.
Bound to be one given the standard lens thread size.

Dave.

[BravoV]

Thanks a lot Mike, really an amazing extreme teardown, especially the exploding rubidium bulb !

Just make sure you don't drink that RbOH solution by accident ! 

[jahonen]

Measured warm-up transient from my FE-5680A.

I'm wondering if the initial sweep should be symmetrical around 10 MHz or not.

Regards,
Janne

[KD0CAC John]

Mike ,
You mentioned getting some more rubidium , when do we get to see the next water test ?
I would say that there was a much larger reaction to the metal than breaking the bulb , because of the sloshing of the water .
Next time I would suggest holding the bulb under water and then using a punch / etc. to just tap on glass while being held steady underwater , just a slight tap on glass .

If is doesn't explode , then its only half as fun

[firewalker]

Don't expect to see something at all. The amount of Rubidium inside the lamp is minimal.

Alexander.