Anritsu MG3700A Vector Signal generator - Unstable rubidium reference

[cf4044]

I have recently acquired an Anritsu MG3700 having the rubidium clock option but unfortunately there is something wrong with it.  The frequency is not steady.  I don't mean just not as steady as one would expect a Rubidium oscillator to be but unsteady like a simple LC oscillator would be better.
It goes in cycles, the frequency slowly rising from about 20kHz below its intended frequency of 10MHz and as soon as it gets to within about a kHz from where it should be it suddenly drops back to -20kHz and keeps repeating.  Each cycle takes several seconds.

I am checking the frequency on the 10MHz reference output at the back of the instrument.  I don't know if the actual rubidium oscillator itself is faulty or the PLL and whatever other circuitry it uses to bring the frequency down to 10MHz. I'll be trying to find out but would appreciate any advice beforehand if anyone can help. , perhaps having seen this problem before. I think checking the module itself should be quite easy, I just need to look for a pin with the ~6.8GHz on it and see if first of all if there is any signal at all and secondly whether it is stable or not.

If I apply a good 10MHz to the reference input, bypassing the internal reference, the the instrument works perfectly, so the automatic reference clock selection and subsequent use of the reference works fine.  I am in fact using the instrument only with external clock.

I took apart the reference clock module and found it has quite a lot of bodges.  I don't know if these are an attempted failed repair or an official Anritsu bodge to fix some design issue. The bodges are on the support board, not the Rubidium module itself - I didn't open that and don't intend to unless I am sure it is faulty. I would really like to fix it as part of the reason I bought this particular instrument was the rubidium clock. 

I have attached pictures of the reference clock module, with the RF covers removed.

[paul@yahrprobert.com]

I only have experience with a Stanford PRS10 rubidium module, but it has an ouput lock indicator signal.  Perhaps you can find out if yours has one.  Those rubidium things have a tube that has a finite lifetime.  If the unit sat in a lab where people left it on all the time then it could very well have hit its limit.

[edpalmer42]

I have recently acquired an Anritsu MG3700 having the rubidium clock option but unfortunately there is something wrong with it.  The frequency is not steady.  I don't mean just not as steady as one would expect a Rubidium oscillator to be but unsteady like a simple LC oscillator would be better.
It goes in cycles, the frequency slowly rising from about 20kHz below its intended frequency of 10MHz and as soon as it gets to within about a kHz from where it should be it suddenly drops back to -20kHz and keeps repeating.  Each cycle takes several seconds.

I am checking the frequency on the 10MHz reference output at the back of the instrument.  I don't know if the actual rubidium oscillator itself is faulty or the PLL and whatever other circuitry it uses to bring the frequency down to 10MHz. I'll be trying to find out but would appreciate any advice beforehand if anyone can help. , perhaps having seen this problem before. I think checking the module itself should be quite easy, I just need to look for a pin with the ~6.8GHz on it and see if first of all if there is any signal at all and secondly whether it is stable or not.

I'm not familiar with that Fujitsu Rb module, but I've never seen one that brought out the 6.8 GHz signal.  They always take a crystal oscillator, sometimes but not always an OCXO, and multiply it up to 6.8 GHz where the magic happens.  The output of the module is derived from that crystal oscillator, not the 6.8 GHz.  Through the magic of synthesizers, the output frequency from the module can be whatever the designers want.

I only have experience with a Stanford PRS10 rubidium module, but it has an ouput lock indicator signal.  Perhaps you can find out if yours has one.  Those rubidium things have a tube that has a finite lifetime.  If the unit sat in a lab where people left it on all the time then it could very well have hit its limit.

No Paul, a Rb standard doesn't work like that.  You're probably thinking of a Cesium standard.  A Rb standard doesn't have any defined wear-out modes other than typical electronic faults and aging.  They can run for decades.  But your comment about the lock indicator is totally correct.  Many (most?, all?) pieces of test equipment with Rb standards have either an LED or a message available on the screen to tell you if the module is locked.

One common fault for old Rb standards is that the internal crystal oscillator has aged so far that the module can't lock to it anymore.  In that case, the unit continually hunts up and down in frequency looking for the Rb signal.  Sound familiar?  If this is the case, there might be an internal pot or trimmer cap that can be adjusted.  Smart modules have serial ports that you can access and change a data register to shift the oscillator.  Many will require component changes to shift the crystal's frequency to allow it to lock.

Ed

[cf4044]

I'm not familiar with that Fujitsu Rb module, but I've never seen one that brought out the 6.8 GHz signal.  They always take a crystal oscillator, sometimes but not always an OCXO, and multiply it up to 6.8 GHz where the magic happens.  The output of the module is derived from that crystal oscillator, not the 6.8 GHz.

Yes, now looking more closely I see the chips are all simple 74 series logic, nothing that could prescale a microwave frequency. Not sure what all that stuff does though, I was expecting it would be the PLL and all that magic but it seems that is all inside the sealed module.  AFAIK all the external circuitry needs to do is to automatically switch to the external sync input when there is a 5 or 10MHz signal present and indicate to the system which clock it has selected.

It seems though that anything that is wrong must be inside the sealed Fujitsu module.  If it does indeed have a 10MHz output I'll probably just replace it with an ordinary 10MHz OCXO and maybe one day I open up the Rb module and take a shot at fixing it, if it is even possible.

[paul@yahrprobert.com]

No Paul, a Rb standard doesn't work like that.  You're probably thinking of a Cesium standard.  A Rb standard doesn't have any defined wear-out modes other than typical electronic faults and aging.  They can run for decades.

Well, there's this old paper, https://ieeexplore.ieee.org/abstract/document/1537723 , which says that in the Rubidium discharge cell the Rubidium diffuses into the glass envelope and can result in a lifetime of about 10 years. 

[edpalmer42]

No Paul, a Rb standard doesn't work like that.  You're probably thinking of a Cesium standard.  A Rb standard doesn't have any defined wear-out modes other than typical electronic faults and aging.  They can run for decades.

Well, there's this old paper, https://ieeexplore.ieee.org/abstract/document/1537723 , which says that in the Rubidium discharge cell the Rubidium diffuses into the glass envelope and can result in a lifetime of about 10 years.

The Rb lamp is indeed the closest thing to a 'weak link' in any Rb standard.  But, in spite of the paper you referenced, the lamps in many different types of standards have turned out to be surprisingly robust.  HP ceased production of the 5065A in the late 1980s but they are still greatly prized and still sell for somewhat eye-watering prices.  That wouldn't be the case if the lamps were starting to fail.  I'm sure it does happen and may become more common, but apparently not yet.

It has also been found that if the lamp does start to fade due to Rb deposition or absorption on the inner surface, it can usually be rejuvenated by heating it with a heat gun.  Apparently, this forces the Rb out of the glass and results in a lamp that performs almost like new.  AFAIK, no one has formalized this procedure as far as time, temperature, bulbs that can or cannot be rejuvenated, etc.

Ed

[wkb]

Sounds like what I had on an Efratom FRK Rb. The unit hunts for atomic lock, never manages to lock and so forth. In my case it proved to be the crystal heater not working. The cold crystal oscillated too far from the nominal 10 MHz to allow the regulation loop to lock.

Once I repaired the heating circuit (had multiple failures no less) things worked OK.

The scope pics show the hunting, the 1st one shows acquiring atomic lock. On the Efratom this is an easy accesible signal.

[cf4044]

Sounds like what I had on an Efratom FRK Rb. The unit hunts for atomic lock, never manages to lock and so forth. In my case it proved to be the crystal heater not working. The cold crystal oscillated too far from the nominal 10 MHz to allow the regulation loop to lock.

Once I repaired the heating circuit (had multiple failures no less) things worked OK.

The scope pics show the hunting, the 1st one shows acquiring atomic lock. On the Efratom this is an easy accesible signal.

Thanks, that trace looks very similar to what I see, except that I see a much faster drop compared to the increase.  As you describe it does seem very likely what I have. I'll only know once I open it up.

What I'm concerned is that the whole thing seems to be potted. For sure it is in a sealed can that I'll have to open it with a dremmel, only then will I know if it is also potted.