Help needed for repair of Efratom FRS-C rubidium oscillator

[ejd.pol]

Hello,

Recently I have acquired faulty Efratom FRS-C 2A8A4C Rubidium oscillator. It was the internal reference of a PM6685 frequency counter.
The supply of the frequency counter is toast, and maybe the main board is too. So I am in the process of seeing what can be saved.
The first and foremost thing to try to save is the reference. The serial number of the oscillator is 20702, date code 50/94.
 
Unfortunately, the coax output signal is not the expected 10 MHz sine wave. Instead, the output is a sine wave that hovers somewhere around 60 Hz.
Fortunately though, the device still seems to be able to lock: after a few minutes warming up (it does get warm), the "unlock" led goes off.
So this gives me some hope that not all is lost, and a repair is possible. (The unlock led is connected to pin 1 of the external connector: lock monitor)

So, I took off the top cover, to see what was underneath. A nice construction of four boards making up the sides, interconnected by a flex board on top.
The flex board seems to have a number of numbered test points, so I measured the DC values on these points, after warming up and unlock off.

2: 0.65 V
3: 6.38 V
4: 2.55 V
5: 23.22 V
6: 17.09 V
7: 0.00 V (I used this point as measurement reference for the other points)
8: 8.86 V
10: 5.92 V
11: 5.75 V
12: 0.00 V
13: 0.00 V
14: 2.51 V

While doing these measurements, i noticed that point 11 is very sensitive: by just touching it, I can unlock the device (led on).
And in that state, I can see on my scope that the output sine frequency changes, and even seems to cross 0 Hz.
The highest frequency that it produces during the sweep is about 160 Hz.

Perhaps the symptoms as described are known for the FRS-C type, and hopefully the repair is known as well.
Of course, more measurements can be taken (DMM or scope) if useful.
I have found some documentation on the FRS-C on the web, but the complexity involved is daunting.
I hope someone can give me some guidance on how best to proceed.

I have taken some pictures of the disassembly, so if there is any interest, I can post these as well.

Many thanks in advance, Evert-Jan

[sanleontexas]

May I suggest that you relist this in the metrology section? There's a lot of talent there,
and likely that someone is familiar with this.

[ejd.pol]

Sure, thanks for the good advice, much appreciated. I will open a new thread in Metrology.

[Halcyon]

Don't "relist" posts in multiple parts of the forum.

If something needs to be moved, we can move it.

[ejd.pol]

Hello, I am sorry to have overstepped the rules! Can you help in moving this thread to "Metrology"?

Many thanks in advance, Evert-Jan

[edpalmer42]

I was hoping that someone with direct experience with the FRS-C would pop up, but I have repaired a few Efratom FRK and M100 units that use similar technology, so maybe I can help.  I have a version of the FRS-C that has an odd part number of 100334-004.  I think it's a private label version for some OEM.  It works fine so I can follow your tests for comparison.

First, a bunch of questions ....

1.  Are you working in or out of circuit?  Out of circuit is preferred so that you have full control without any interference from the PM6685 (whether it's alive or dead).  At the very least, you need to tap into the power and measure the start-up (~1.8A) and steady-state (~0.5A) current at 24V.  This will confirm that all the ovens are alive.

2.  Depending on internal options, there are 1 or 2 fuses on the power supply board.  Are they both good?

3.  What type of scope do you have?  You mentioned that the noise on the output is about 60 Hz.  How sure are you about the frequency?  Could it be 50 Hz?  The mains frequency in the Netherlands is 50 Hz isn't it?

3.  Does pin 11 on the oscillator board show a 10 MHz TTL square wave?  If that's not present, nothing else can work.  I know that pin 11 is also test point 4, but you want to check pin 11 in case there's a problem with the flex cable.

4.  Does pin 9 on the oscillator board show a 20 MHz 3Vpp sine wave?

That's enough for now.

Ed

[ejd.pol]

Hi Ed, Many thanks for your response!  I'll try to answer your questions accurately.

1) I have taken the unit out of the broken frequency counter. That counter also contained a 24 V supply private for the oscillator,
    so (after verifying by DMM and scope that it functions correctly), I have also removed the supply, and use it to power the oscillator.
    It is amazing: one third of the counter box is reserved for the oscillator and its power supply!
    Actually, I also took out the base plate on which the oscillator was mounted, and use it as the heat sink.

2) So there are two completely separate supplies, the only have the 220 V mains in common. They are both switching supplies.
    The supply for the counter electronics has one fuse on the primary side, which is blown. A replacement fuse blew immediately.
    The supply for the oscillator works, and is completely functional, as far as I can ascertain.

3) I have a PM3375 (100 MHz, 2 channels). The output is indeed a suspicious 60 Hz, but not precisely so.
    Moreover, It changes frequency in unlocked state in a sweeping manner, reaching DC on the low side (It seems to "cross" DC),
    and reaches about 160 Hz maximum before it goes down again. Indeed in NL we have 50 Hz, so it is certainly not some kind of stray mains voltage.

4) Indeed, after digesting some Efratom manuals (older ones with decision charts, newer ones with schematics), the xtal oscillator was my first stop as well.
    I measured a nice 20 MHz signal on pin 9, and 10 MHz on pin 11 of the oscillator board. That was reassuring.
   
I followed the signal into the power supply board, where the output buffer is located; no problems on the flex board.
Q3 is the output transistor (2N2222A), but it has no heat sink mounted on it, as shown in the assembly diagram in the manual.
I felt it, and it was only a bit warm to the touch.

I probed the base, emitter, and collector, and all three have reasonable DC and 10 MHz AC signals.
Q3 is the last transistor in the 10 MHz chain, so it seems that the active electronics are fine! :-)

This is where I stopped for today, as it seems impossible to trace the output signal from Q3 to the connector, at least, not without severe disassembly.
It might be as simple a problem with the coax cable, or some short on the power supply board (although I find that unlikely).
For now, the only way I can see to probe a point between the connector and Q3 would be to supply the box through crocodile clamps or such,
so that I do not need to plug in the external connector. And I could actually start by doing a continuity/short check on the coax...

In case you have some more suggestions on how to continue, please let me know!   

[edpalmer42]

Hi Ed, Many thanks for your response!  I'll try to answer your questions accurately.

1) I have taken the unit out of the broken frequency counter. That counter also contained a 24 V supply private for the oscillator,
    so (after verifying by DMM and scope that it functions correctly), I have also removed the supply, and use it to power the oscillator.
    It is amazing: one third of the counter box is reserved for the oscillator and its power supply!
    Actually, I also took out the base plate on which the oscillator was mounted, and use it as the heat sink.

Yes, I have the PM6681 which is basically a PM6685 without the Rb standard.  It also uses a second supply if you have that option.

2) So there are two completely separate supplies, the only have the 220 V mains in common. They are both switching supplies.
    The supply for the counter electronics has one fuse on the primary side, which is blown. A replacement fuse blew immediately.
    The supply for the oscillator works, and is completely functional, as far as I can ascertain.

No, I was referring to the power supply board that's inside the FRS-C.  There are one or two fuses on that board.

3) I have a PM3375 (100 MHz, 2 channels). The output is indeed a suspicious 60 Hz, but not precisely so.
    Moreover, It changes frequency in unlocked state in a sweeping manner, reaching DC on the low side (It seems to "cross" DC),
    and reaches about 160 Hz maximum before it goes down again. Indeed in NL we have 50 Hz, so it is certainly not some kind of stray mains voltage.

The sweeping is normal and good to see.  The system is tuning the oscillator up and down, searching for the Rb signal.  If your scope is good enough or you have a frequency counter, you should see the 10 MHz from the oscillator sweep up and down by something in the range of the 160 Hz that you mentioned.  I don't understand why that 160 Hz would be appearing on the output.

One problem that old Rb standards can have is that the quartz oscillator drifts beyond the point that it can be pulled back to 10 MHz.  The sweeping continues forever.  I'm not aware that the FRS-C has this feature, but some Rb standards might not output a signal until the unit locks.  But your unit is reporting a lock, so that likely isn't your issue.  If you have access to a well-calibrated frequency source like a GPSDO, another known-good Rb, or similar you could check the square wave out of the oscillator board to confirm that it sweeps above and below 10 MHz.

4) Indeed, after digesting some Efratom manuals (older ones with decision charts, newer ones with schematics), the xtal oscillator was my first stop as well.
    I measured a nice 20 MHz signal on pin 9, and 10 MHz on pin 11 of the oscillator board. That was reassuring.
   
I followed the signal into the power supply board, where the output buffer is located; no problems on the flex board.
Q3 is the output transistor (2N2222A), but it has no heat sink mounted on it, as shown in the assembly diagram in the manual.
I felt it, and it was only a bit warm to the touch.

No heat sink on Q3?  Mine has the heat sink.  Maybe try an in-circuit test of Q3 to confirm that it looks like 2 diodes and doesn't have a C-E short.

I probed the base, emitter, and collector, and all three have reasonable DC and 10 MHz AC signals.
Q3 is the last transistor in the 10 MHz chain, so it seems that the active electronics are fine! :-)

This is where I stopped for today, as it seems impossible to trace the output signal from Q3 to the connector, at least, not without severe disassembly.
It might be as simple a problem with the coax cable, or some short on the power supply board (although I find that unlikely).
For now, the only way I can see to probe a point between the connector and Q3 would be to supply the box through crocodile clamps or such,
so that I do not need to plug in the external connector. And I could actually start by doing a continuity/short check on the coax...

In case you have some more suggestions on how to continue, please let me know!   

If you have to do any disassembly, move very carefully!  That flex cable has been cooking for 25 years.  I would expect it to be as brittle as glass.  If you can work without moving it at all, that would be best.  I would expect the traces to fall off the board at the gentlest touch of a soldering iron.  Don't even consider working on it unless you have both a temperature-controlled iron and temperature-controlled vacuum desoldering unit.  Don't use desoldering wick - it requires applying heat for too long.  Basically, be paranoid!!

Ed

[edpalmer42]

The heatsink might not be required on Q3.  I just found a note that says that the heatsink is only required if your unit has the TTL output option.

I'm quite confused about which output option my unit has.  It has the jumper from A-B and the heatsink, but it also has L1 and C18 which are only used with a sine wave output!  It also has a TTL sticker on the flex cable!  What a mess!

[ejd.pol]

Hi Ed,

Quick reply (before rushing off to work):

-Have not checked yet the fuses on the power board. Will do so!
  (given the high functionality of the unit, I would expect they are intact, but it is good to be certain.)

-The 160 Hz occurs just at the top of the frequency sweep, and only lasts a short time before it sweeps to lower frequencies.

-Even if the crystal has drifted so much that it can not be pulled into the proper range, the coax output should have the crystal oscillator signal present.
 My unit does not. That is the first thing to repair. After that has been fixed, the locking can be examined. Mind though that my unit does indicate lock after the warm-up!

-I did test Q3 with power off, and saw the diodes between B-E and B-C. Did not check for a C-E short though, will do that too!

-Your last remark about the flex board is an important one, much appreciated! Together with the complexity, the inexperience in dismantling these units (if unavoidable)
 were the two reasons for me to ask for help. On second sight, the complexity is not as overwhelming as I first thought ,
 but the handling of the unit (flex board, and especially the physics package with the rubidium lamp and heater and such) is certainly beyond my experience.
 I will increase my paranoia level! 

-My unit has indeed only the analog sine output, not the TTL output. At the factory, L1 and C18 are always installed, as the default is the sine wave output.
 I have read in the manual somewhere that conversion to TTL is done after factory sale. That may explain the presence of L1 and C18.

-Final thought: this morning it dawned on me that the fail may have to do with the main power supply fail. Reasoning is as follows.
 It seems that the active components in the 10 MHz chain are fine, up to and including Q3. The reason there is no 10 MHz to be seen on the coax
 might be something mechanical (a break in the coax at or near the connector, for example), but there is _some_ signal, so that's unlikely.
 Alternatively then, it could be, say, a failing capacitor. And for the AC-coupling capacitor, that might have been triggered by the fail of the main power supply.
 If so, then I have lost all hope for the main board, but it looks like the failing capacitor has protected the oscillator very well. Makes sense?
 
 

[edpalmer42]

Hi Ed,

Quick reply (before rushing off to work):

What is this 'work' thing you speak of?  I don't remember ever hearing that term. 

-Have not checked yet the fuses on the power board. Will do so!
  (given the high functionality of the unit, I would expect they are intact, but it is good to be certain.)

True, but missing a blown fuse is high on the list in the Face Palm Hall of Infamy!

-The 160 Hz occurs just at the top of the frequency sweep, and only lasts a short time before it sweeps to lower frequencies.

-Even if the crystal has drifted so much that it can not be pulled into the proper range, the coax output should have the crystal oscillator signal present.
 My unit does not. That is the first thing to repair. After that has been fixed, the locking can be examined. Mind though that my unit does indicate lock after the warm-up!

Agreed.  However, we're starting to run out of obvious things to check so I'm starting to think about less obvious possibilities.

-I did test Q3 with power off, and saw the diodes between B-E and B-C. Did not check for a C-E short though, will do that too!

-Your last remark about the flex board is an important one, much appreciated! Together with the complexity, the inexperience in dismantling these units (if unavoidable)
 were the two reasons for me to ask for help. On second sight, the complexity is not as overwhelming as I first thought ,
 but the handling of the unit (flex board, and especially the physics package with the rubidium lamp and heater and such) is certainly beyond my experience.
 I will increase my paranoia level! 

-My unit has indeed only the analog sine output, not the TTL output. At the factory, L1 and C18 are always installed, as the default is the sine wave output.
 I have read in the manual somewhere that conversion to TTL is done after factory sale. That may explain the presence of L1 and C18.

-Final thought: this morning it dawned on me that the fail may have to do with the main power supply fail. Reasoning is as follows.
 It seems that the active components in the 10 MHz chain are fine, up to and including Q3. The reason there is no 10 MHz to be seen on the coax
 might be something mechanical (a break in the coax at or near the connector, for example), but there is _some_ signal, so that's unlikely.
 Alternatively then, it could be, say, a failing capacitor. And for the AC-coupling capacitor, that might have been triggered by the fail of the main power supply.
 If so, then I have lost all hope for the main board, but it looks like the failing capacitor has protected the oscillator very well. Makes sense?

One other point that you can check for signal is 'pad G' which is shown on the parts drawing and is silkscreened on the power supply board in my unit.  It's between the main connector and the 10x1 connector labelled as J1.  If the signal is present there it proves that L1 and C9 are good.

It seems unlikely that the failure of the main power supply would affect the output of the FRS-C.  Perhaps a pre-existing fault that either wasn't noticed or wasn't important.  Just use an external reference and carry on.

Have you confirmed that the counter's main board is dead?  Many (most?) failures in SMPSs just make the output disappear rather than go high.

Ed

[ejd.pol]

Hey Ed,

The work thing is something I inherited from my dad, he was inflicted too! No need to worry though, it is not contagious or anything like that. 

After some puzzling, I checked both fuses. It seems my power supply board is a different revision from what is depicted in the manual I am using.
(Yet another thing to be paranoia about: incorrect or outdated information in the official documentation!)
Both fuses seem to be intact: they both measure almost as a short, but not quite: they measure 0.05 ohms, while shorted leads measure 0.02 ohms.
So that seems to be OK. No face palm award yet! (DMM = Fluke 8842)

Good advice about testing pad G, I had missed that! I found a nice steady 10 MHz sine wave on that point. That means that Q3 is also in the clear,
no need to test for shorted C-E anymore. So tomorrow, I will have a good look at the connector and the coax cable!

No, I don't know the status of the main board yet. I do have another 10 MHz generator that would come in handy there, a home-brew 10 Mhz GPSDO.
(Designing and building that one was a nice adventure. Based on an Arduino Uno, but the micro does not do any counting. I use a 74LV8154 for that.)
I would have to find a replacement power supply first, though. Do you know if there is a PM6685 service manual floating around on the web?

But before putting significant effort and time into that part, I would like to know if the rubidium oscillator is still ok! 

Cheers, and may thanks for all your help and guidance!

P.S. in case you're interested, I can post the design-and-build doc of the GPSDO. The costs run only in the $50 to $60 range,
      about half of which is the OCXO. I definitely learned a lot from it.

[edpalmer42]

Hey Ed,

The work thing is something I inherited from my dad, he was inflicted too! No need to worry though, it is not contagious or anything like that. 

After some puzzling, I checked both fuses. It seems my power supply board is a different revision from what is depicted in the manual I am using.
(Yet another thing to be paranoia about: incorrect or outdated information in the official documentation!)
Both fuses seem to be intact: they both measure almost as a short, but not quite: they measure 0.05 ohms, while shorted leads measure 0.02 ohms.
So that seems to be OK. No face palm award yet! (DMM = Fluke 8842)

Yes, another thing we can check off the list.

Good advice about testing pad G, I had missed that! I found a nice steady 10 MHz sine wave on that point. That means that Q3 is also in the clear,
no need to test for shorted C-E anymore. So tomorrow, I will have a good look at the connector and the coax cable!

Do you actually have a coax cable on the connector?  Mine, and every picture I can find online, shows the power supply board as a single board with the connector soldered directly into the board.

No, I don't know the status of the main board yet. I do have another 10 MHz generator that would come in handy there, a home-brew 10 Mhz GPSDO.
(Designing and building that one was a nice adventure. Based on an Arduino Uno, but the micro does not do any counting. I use a 74LV8154 for that.)
I would have to find a replacement power supply first, though. Do you know if there is a PM6685 service manual floating around on the web?

Google is your friend.  Also, fixing the original power supply is a reasonable thing to consider.  But don't go probing it with your scope unless you're REALLY sure what you're doing.  FYI

But before putting significant effort and time into that part, I would like to know if the rubidium oscillator is still ok! 

Why?  I mean, yes, it's fun to fix a Rb standard, but it's basically a seperate project from fixing the counter.  AFAIK, your PM6685R is also known as a PM6685 (without the R) which has a built-in oscillator.  If you want better accuracy, you can use an external reference or purchase a plug-in OCXO internal reference from the web site we love to hate!  Actually, there's only a couple available right now.  Search for pm6685 and pm66xx.  There are add-in oscillators and RF prescalers that just drop into place.

By the way, I sorted out the confusion in my unit regarding TTL vs. Sine output.  I think this unit left the factory with the TTL option.  Somewhere along the line someone converted it to Sine output - all the component values match the requirements for a 10 MHz Sine output.  But they forgot to remove the factory-installed jumper between pads A-B and didn't bother to remove the TTL sticker that's on the flex board.  When I clipped the jumper the waveform purity and amplitude improved dramatically.  Nice!  I didn't notice before because since the unit worked, I had no reason to pull it apart.

Cheers, and may thanks for all your help and guidance!

P.S. in case you're interested, I can post the design-and-build doc of the GPSDO. The costs run only in the $50 to $60 range,
      about half of which is the OCXO. I definitely learned a lot from it.

Thanks, but I have too many GPSDOs .... and too many Rb standards .... and too many fancy OCXOs ..... I think I'm seeing a pattern here ......

Ed

[ejd.pol]

Hi Ed,

>> Do you actually have a coax cable on the connector?  Mine, and every picture I can find online, shows the power supply board as a single board
>> with the connector soldered directly into the board.

Yes, there is a connector that is indeed soldered onto the power board in my box as well. There are two other connectors though. At least, I would call them connectors too.
There is some wiring between the power supply and the oscillator, and these wires are soldered to a second connector, which mates with the connector on the power board.
And the coax I was referring to runs from that second connector to a third one (single line with regular 0.1" pitch) that plugs into the main board.

So, this morning I sat down to see if I could find some short or discontinuity in that part of the signal path. To my surprise, I found that there is a 50 ohm resistor
somewhere in the coax/connector assembly. I can not locate it exactly, but I would expect it to be at the end of the third connector.
After measuring, and checking again, I could find nothing wrong with the cable. So, in desperation, I hooked everything up, to see if I had perhaps missed something
in my earlier measurements. And to my astonishment, I now have an accurate 10 MHz signal at the third connector!! 
Of course I am very happy about it, but I really have no clue what has changed to make that happen... 
And now that it seems to work, it is impossible to find the fault...

I have hooked up my lowly Tek CDC250, which I have repaired some years ago by replacing the dead reference by a TCXO.
I have calibrated that reference with my GPSDO. It has been a while since I last used it (more than a year...), but when I hooked it up
to the Rubidium reference, and allowed for some warming up, it is giving a very steady 10,000,000 reading, with every now and then 10,000,001 .
Not bad at all! 

A few remarks/questions in hindsight:
-Given your warning, I have taken a good look at the flex board, and it seems to be in excellent condition. No sign whatsoever in the direction of it being brittle.
-I found that the screws to mount the box on the base plate are just a bit too long, and they nudged a capacitor and bent it out of ts original orientation. i did not bend it back.
-On lifting the outer shell (after removing the four bolts on top), I found only two plastic washers. Are there supposed to be two, or four? Or perhaps none?

So, it seems that we can now shift attention to the "rest" of the PM6685.    And I agree completely, that is a entirely independent project in and of itself.
It is clear that the supply needs to be repaired or replaced, so I want to disassemble it from the counter, at least to protect whatever is still alive on the main board.

Heh, I was laughing when I looked at the video you posted: at 7:45, the guy simply lifts the supply board from the main board, effortlessly.
That will not be possible at all in my case! In fact, due to the construction, I think it is impossible to remove it without actually destroying one thing or the other.
During assembly, the board was mounted on the main board with nylon pylons, and then soldered in place on both ends at multiple pins...

So for disassembly, I see no other option but to destructively cut the pylons, and to make a jig from copper wire that allows me to heat all connector pins at the same time,
allowing me to desolder the board one end at a time. The heat sink close by does not make it any easier, though.
I truly wonder how they serviced the power supply in the factory....  And please do let me know if you can think of any alternatives!
If and when I can assemble a repaired or a replacement power supply back onto the main board, I will use nylon standoffs with ditto nuts and bolts!

>>Google is your friend. 

Indeed it is, and I found the service manual without much difficulty. I am reading the relevant parts on the power supply.
And actually, the first thing I would like to do, is to power up the main board with a stand-in supply to see if there is still some life in it..
That replacement power supply would need to generate +5, +15, and -7 V, and the 5V is used as-is on the main board. Hmmmm.....
These supplies are not offered on FlukeBay. Fortunately, my unit already has the 2.7 GHz option and the GPIB option.
Thanks for the warning about doing measurements on the supply. I will be careful with that. But first it need to get out!



 

[edpalmer42]

Hi Ed,

>> Do you actually have a coax cable on the connector?  Mine, and every picture I can find online, shows the power supply board as a single board
>> with the connector soldered directly into the board.

Yes, there is a connector that is indeed soldered onto the power board in my box as well. There are two other connectors though. At least, I would call them connectors too.
There is some wiring between the power supply and the oscillator, and these wires are soldered to a second connector, which mates with the connector on the power board.
And the coax I was referring to runs from that second connector to a third one (single line with regular 0.1" pitch) that plugs into the main board.

Okay, I was getting confused between J1 (the single line connector) and P1 (the main I/O connector for the unit).

So, this morning I sat down to see if I could find some short or discontinuity in that part of the signal path. To my surprise, I found that there is a 50 ohm resistor
somewhere in the coax/connector assembly. I can not locate it exactly, but I would expect it to be at the end of the third connector.
After measuring, and checking again, I could find nothing wrong with the cable. So, in desperation, I hooked everything up, to see if I had perhaps missed something
in my earlier measurements. And to my astonishment, I now have an accurate 10 MHz signal at the third connector!! 
Of course I am very happy about it, but I really have no clue what has changed to make that happen... 
And now that it seems to work, it is impossible to find the fault...

The fact that you touched it and it started working is very suspicious.  Check the solder joints on J1 and P1 *VERY* carefully.  I use a 20X magnifying glass.  Connectors, switches, even some component leads tend to unsolder themselves over time.  This get particularly bad if there's power and/or heat involved.  Even at 20X magnification it's almost a guess whether you're looking at a good solder joint or one where the component lead has completely broken free of the solder and is just sitting in the hole.  Is that a crack that I'm seeing or just a trick of the light?  It's often easier to just add some liquid flux ( that's usually necessary ) and resolder all the pins. 

I have hooked up my lowly Tek CDC250, which I have repaired some years ago by replacing the dead reference by a TCXO.
I have calibrated that reference with my GPSDO. It has been a while since I last used it (more than a year...), but when I hooked it up
to the Rubidium reference, and allowed for some warming up, it is giving a very steady 10,000,000 reading, with every now and then 10,000,001 .
Not bad at all! 

A few remarks/questions in hindsight:
-Given your warning, I have taken a good look at the flex board, and it seems to be in excellent condition. No sign whatsoever in the direction of it being brittle.
-I found that the screws to mount the box on the base plate are just a bit too long, and they nudged a capacitor and bent it out of ts original orientation. i did not bend it back.
-On lifting the outer shell (after removing the four bolts on top), I found only two plastic washers. Are there supposed to be two, or four? Or perhaps none?

I didn't find any washers in mine, but it was messed with before I got it.  Mine has a plastic sheet that's folded into the inside of the cover.  Does yours use different screws to attach the base plate and the cover?  Mine has 5 screws for the baseplate and 4 for the cover.  They look identical but I noticed that 5 of the screws are non-magnetic and 4 are magnetic.  Coincidence?  I think not!

So, it seems that we can now shift attention to the "rest" of the PM6685.    And I agree completely, that is a entirely independent project in and of itself.
It is clear that the supply needs to be repaired or replaced, so I want to disassemble it from the counter, at least to protect whatever is still alive on the main board.

Heh, I was laughing when I looked at the video you posted: at 7:45, the guy simply lifts the supply board from the main board, effortlessly.
That will not be possible at all in my case! In fact, due to the construction, I think it is impossible to remove it without actually destroying one thing or the other.
During assembly, the board was mounted on the main board with nylon pylons, and then soldered in place on both ends at multiple pins...

No, he did the disassembly off-camera.  He referred to an earlier video where he repaired a similar unit so he didn't want to waste time doing the same thing again.  Be sure to check out that earlier video.

So for disassembly, I see no other option but to destructively cut the pylons, and to make a jig from copper wire that allows me to heat all connector pins at the same time,
allowing me to desolder the board one end at a time. The heat sink close by does not make it any easier, though.
I truly wonder how they serviced the power supply in the factory....  And please do let me know if you can think of any alternatives!
If and when I can assemble a repaired or a replacement power supply back onto the main board, I will use nylon standoffs with ditto nuts and bolts!

>>Google is your friend.

Indeed it is, and I found the service manual without much difficulty. I am reading the relevant parts on the power supply.
And actually, the first thing I would like to do, is to power up the main board with a stand-in supply to see if there is still some life in it..
That replacement power supply would need to generate +5, +15, and -7 V, and the 5V is used as-is on the main board. Hmmmm.....
These supplies are not offered on FlukeBay. Fortunately, my unit already has the 2.7 GHz option and the GPIB option.
Thanks for the warning about doing measurements on the supply. I will be careful with that. But first it need to get out!

Working on a mains-connected power supply, particularly a switching one, can go really bad, really fast!  You should do as much as you can with the supply unplugged and only after you made damn sure that the capacitors are discharged!!  Then check the rectifiers and all the electrolytic capacitors on both the primary and secondary sides of the transformer.  They are easy to check and are often the culprits.  The fact that the fuse blows instantly suggests a shorted capacitor.  It's possible that such a short could then take out the rectifier associated with that capacitor.

Ed

[ejd.pol]

Hi Ed,

Yeah, a bad solder joint, or a crack in the flex board could be the culprit here.
The solder joints are not easy to check though, as they are all hidden inside the box.
So on one hand, it would be nice to check everything, on the other hand, that check could easily cause a new problem!
So at the moment I am considering to err on the safe side: not mess with it now that it seems to work.
Of course, I would like to have proof that the system achieves a good lock. What is a sure sign of proper lock for you?

Yes, my unit has the plastic sleeve also on the inside. There are holes in the sleeve that are much larger than needed to pass the bolts through.
So, given the flexibility of the outer shell, when no washers are used, the bolt will bend the shell locally.
I can imagine that washers are used to prent this bending, but that couls also have been achieved by making the holes in the inner sleeve small.
So I do not understand the function of the plastic washers. (It is not isolation, as the bolt will provide electrical contact.
Thanks for the tip about magnetic and non-magnetic bolts, I will check mine for that!

After some searching, I found the video that the repair guy was referring to in the video that you posted.
It is about a repair of a PM6681, and indeed it shows the repair of the supply, which seems pretty much identical to the PM6685 supply.
In his case, the TL431 needed to be replaced. But... the video does not show what technique he uses to take the supply out!   
That second video does show how he examines his supply, rigs it such that he can probe and such, that is definitely useful.

So, I did what I have described earlier: destroy the pylons, and then desolder the connecting wires on both ends of the PCB.
Removing the pylons was easy: the nylon has become brittle, so only using a needle-nose pair of pliers, and a bit of twisting did the job.
For the desoldering, I first went in with some solder-wick, and that worked just fine on the primary side.
The secondary side has some more pins, and it took a bit more effort to get it loose, but in the end I got that too. Phew! 

What was clear immediately was that the two 470uF/35V secondary end caps definitely need to be replaced.
The green plastic sleeve is discolored to black at the base, and there is some signs of leakage in the from of gunk between the caps and pcb.
And while I am at it, the 5V end cap of 10,000uF/6V3 will be replaced as well. I plan to use a 10V type for that, if I can find one that fits in the available space.

But before actually doing anything with the caps, I have two more checks to do: the rectifiers (as you rightfully mentioned), and the transformer.
I already did some measurements on the transformer, and what worries me is that on the primary side, I do see the coil on the 1 - 6 pins (0.5 ohm),
but the coil on the 3 - 4 pins measures in the Mohms. I guess that means conclusively that my transformer is broken, doesn't it?
If so, is it worth-while to try to fix the transformer, or can they be found somewhere? Should I start looking out for a replacement supply?

Cheers, and many thanks for your help in this adventure!



 

[edpalmer42]

I already did some measurements on the transformer, and what worries me is that on the primary side, I do see the coil on the 1 - 6 pins (0.5 ohm),
but the coil on the 3 - 4 pins measures in the Mohms. I guess that means conclusively that my transformer is broken, doesn't it?
If so, is it worth-while to try to fix the transformer, or can they be found somewhere? Should I start looking out for a replacement supply?

It seems unlikely that the transformer is blown.  Note that the service manual says that the only time the fuse should blow is if there's a catastrophic failure on the primary side of the transformer.  The winding from 3-4 is a low voltage, low power winding that only provides power for one small IC.  I don't see anything there that would be capable of frying that winding.  R57 is shown on the schematic as 100R while the parts list says 47R, but only 1/8 W.  That would blow open long before the winding would fail.  R33 is 10R 1/8W.  It would blow as well.  Also, V08 is set up as a current limiter that won't allow more than about 130ma to flow.  Recheck your measurement.

If the transformer is blown, it's not practical to fix or replace it.  It's a custom part that only Fluke has and I doubt they would sell it to you.  Your only option at that point would be to do a complete replacement.  That would be a significant, but perhaps straight forward job.

Ed

[ejd.pol]

Hello Ed, I did do some more measurements to get a better view on what may need repair, and doing so, I can follow a path of destruction.
All measurements are done in circuit. What follows is a list of broken primary components. Primary components not mentioned are ok.
Secondary components have not been measured.

-The recheck of the winding on pins 3 and 4 of the transformer gave the same result: 17 Mohm, and rising during the measurement.
-Both diodes in D09 measure open.
-R57 measures 4 Mohm
-R29 measures 2 kohm
-Transistor V08 shows a short between C and E, and that brings us to ground again.

I can not tell easily if IC U03 (UC3842A) is still intact.

I was thinking about this a bit, and the only way that I can think of for these failures to happen, is if the two primary windings shorted.
In turn, that may have been caused by the secondary output caps failing? Large secondary currents heating up the transformer perhaps,
combined with a weak spot somewhere in the insulation of the primary windings?

In any case, I think it is very likely that I do have a broken transformer. So, before sinking anymore time or money into the supply,
I would want to see the main board functioning! If the main board is toast, then the PM6685 can be scrapped, except for the reference, of course!
(Unless the physical parts may be of use to someone.)

So, some thinking is needed on how to power up the main board with some kind of emergency supply.
I'll keep you posted on progress, and as always, your feedback is highly appreciated.

[edpalmer42]

Wow!  Some serious carnage there.  Yes, something obviously went horribly wrong!

I'm not hopeful, but since the blown circuits are nothing more than power for U03, you could try, with the unit unplugged, injecting power from an external supply to see if U03 wakes up.  I'd recommend a variable lab supply.  Set it for maybe 200ma current limit and slowly increase the voltage to a maximum of 10 - 12 volts.  Be sure to check the resistor chain (R24 - R26, R27, R30 - R32) and D12, D13 to make sure there are no blown components there.  If U03 draws some current, but doesn't cause the supply to current limit, you can consider whether to try a power test with the unit plugged in.  I'd STRONGLY suggest that if you do a power test, you use a dim-bulb tester.  If you're not familiar with that, search the forum.  There are a few good threads that describe the tester and its use.  It's trivial to throw one together if you don't have one.

I've had a few cases where I needed to simultaneously power up multiple supplies.  I just used a switched power bar with multiple supplies.  That would be appropriate for the above power test and for powering up the main board.

If the main board survived, you could replace the three supplies.  +5 and +15 modules are easy to find.  -7 would be a variable unit.  You might be able to find a Fluke parts unit for sale that uses the same power module.  There seems to be lots of commonality between the various PM66XX models, so that might be a practical option.  Determine which units use the same power module and then start looking around.

Ed

[ejd.pol]

Hello Ed,

After giving the situation some thought, I think I will construct a replacement three-output supply first, delivering +5, +15, -7.
The +5 V can be fixed, that is needed more often. The other two voltages variable, so I can use it in other situations too.

(Side note: I happen to have a Fluke 7261A that needs the same treatment. Same transformer problem: short in the primary windings.
Completely different transformer, though. Also three output voltages to replace. With a bit of work, I should be able to rewind the primary.
Transformer disassembly started, E-I core removed. But before going any further, I want to know if it is worth while!)

So, now I am thinking about constructing a fixed+variable power supply. That would come in handy more often, I'm sure.
I have some surplus computer supplies, that should take care of the +5, and gives a decent +12 for free.
It is the variable outputs that require some thought.

----------------------------------------------------------

Coming back to the Rb oscillator.

Magnetics:

I have the same situation: the bolts for the outer shell are non-magnetic, the bolts for mounting on the base plate are magnetic.
And additionally, and this is where I had not taken care enough to keep things separate: the same is true for the washers!
I have four magnetic ones, and four non-magnetic.  I have sorted them using a weakly magnetized screwdriver.

Root cause of the failure:

The signal that I saw on the output before the unit repaired itself, was a completely different and well-defined signal.
That tells me that the connector itself is good. I remember reading that there are some analog switches that the 10 MHz signal must go through
before it reaches the output buffer on the supply board. That means that other signals can be routed to there as well.
If that is correct, the failure was in that switching part! Would you agree? Was it truly a failure to begin with?
What is the purpose of the analog switches? (I do not recall reading anything about that.)

The signal observed earlier is some kind of modulation signal. From the frequencies observed, most likely it was the 127 Hz signal.
If I am not mistaken, the 127 Hz signal is used to modulate the Ghz resonator frequency. But is that 127 Hz signal itself modulated as well?
I think I need to go back and read more carefully about how the modulation and locking mechanism works.

Calibration:

In order to get the oscillator out for inspection, I also removed the 20(?)-turn Bourns pot from the front panel.
(So now I know how to circumvent the calibration sticker, which is still very much in tact!) After removal from the front panel,
I carefully measured the position of the pot, so that I could always put it back in position in case it would be accidentally turned.
The resistance measurements have in 5 digits, thanks to the 8842. The unit was originally calibrated to 0.2 mHz.
I have not touched the Bourns pot, nor have I touched the setting on the aux power supply of the unit (the one that delivers 24 V).
Would that suffice to get me somewhere close to the originally calibrated value, or is that just wishful thinking?

Regards, EJ

[edpalmer42]

Hello Ed,

After giving the situation some thought, I think I will construct a replacement three-output supply first, delivering +5, +15, -7.
The +5 V can be fixed, that is needed more often. The other two voltages variable, so I can use it in other situations too.

(Side note: I happen to have a Fluke 7261A that needs the same treatment. Same transformer problem: short in the primary windings.
Completely different transformer, though. Also three output voltages to replace. With a bit of work, I should be able to rewind the primary.
Transformer disassembly started, E-I core removed. But before going any further, I want to know if it is worth while!)

So, now I am thinking about constructing a fixed+variable power supply. That would come in handy more often, I'm sure.
I have some surplus computer supplies, that should take care of the +5, and gives a decent +12 for free.
It is the variable outputs that require some thought.

There are two ways to look at it.  You can build or buy 3 power supplies with the hope that the board powers up and works properly and then those power supplies become the permanent replacement for the entire blown U39 module.  If the counter is dead, you repurpose those supplies as lab supplies.

You could also build or buy three lab supplies that cover the voltages necessary.  In that case, you end up with three nice lab supplies and then, if the board works, build or buy 3 dedicated supplies to repair the counter.  You end up with three lab supplies.  You can *NEVER* have too many power supplies in your lab! 

One complication to keep in mind is that the input to the U39 module is high voltage DC rather than AC.  I don't know if typical AC switching power supplies will tolerate high voltage DC input.  You might have to rearrange the input power to feed them with AC.

Coming back to the Rb oscillator.

Magnetics:

I have the same situation: the bolts for the outer shell are non-magnetic, the bolts for mounting on the base plate are magnetic.
And additionally, and this is where I had not taken care enough to keep things separate: the same is true for the washers!
I have four magnetic ones, and four non-magnetic.  I have sorted them using a weakly magnetized screwdriver.

The mix of magnetic and non-magnetic screws didn't make much sense to me.  I dug through my stash of non-magnetic screws and found four more non-magnetic ones so now I've got a full set.

Root cause of the failure:

The signal that I saw on the output before the unit repaired itself, was a completely different and well-defined signal.
That tells me that the connector itself is good. I remember reading that there are some analog switches that the 10 MHz signal must go through
before it reaches the output buffer on the supply board. That means that other signals can be routed to there as well.
If that is correct, the failure was in that switching part! Would you agree? Was it truly a failure to begin with?
What is the purpose of the analog switches? (I do not recall reading anything about that.)

I'm not seeing any analog switches.  The oscillator itself is U1, 54LS320 which is a dedicated quartz oscillator chip, followed by a 54HCT393 which is a ripple counter, i.e. a divider.  In any case, you had a solid 10 MHz signal coming out of the oscillator board and I don't see anything between the oscillator board and the power supply board.

The signal observed earlier is some kind of modulation signal. From the frequencies observed, most likely it was the 127 Hz signal.
If I am not mistaken, the 127 Hz signal is used to modulate the Ghz resonator frequency. But is that 127 Hz signal itself modulated as well?
I think I need to go back and read more carefully about how the modulation and locking mechanism works.

I don't think you stated what the amplitude of the bad signal was.  Did it look like a solid signal or something that was basically noise being coupled in to the output line and then amplified by Q3?  Bad solder joints or other circuit faults anywhere in the system are a possibility, but the clean 10 MHz coming out of the oscillator board but not on the output connector points toward the power supply board rather than the oscillator board.

Calibration:

In order to get the oscillator out for inspection, I also removed the 20(?)-turn Bourns pot from the front panel.
(So now I know how to circumvent the calibration sticker, which is still very much in tact!) After removal from the front panel,
I carefully measured the position of the pot, so that I could always put it back in position in case it would be accidentally turned.
The resistance measurements have in 5 digits, thanks to the 8842. The unit was originally calibrated to 0.2 mHz.
I have not touched the Bourns pot, nor have I touched the setting on the aux power supply of the unit (the one that delivers 24 V).
Would that suffice to get me somewhere close to the originally calibrated value, or is that just wishful thinking?

Regards, EJ

Rubidium standards do drift over time.  Considering the age of the unit, recalibrating it by comparing it to your GPSDO isn't a bad idea.  It will be challenging to do it with a scope, you really need a time interval counter.  The frequency will also shift slightly with temperature and input voltage.  But since you've got that pot, that's a good starting point.

FYI, the PM6681 power supply is very close, if not identical to the PM6685.  The service manual warns that if you touch the +5V trimmer, you have to adjust 'the complete instrument', whatever that means.  The PM6685 service manual doesn't say that, but since it's so similar, keep that warning in mind.

Ed

[ejd.pol]

Hello Ed,

There are two ways to look at it.  You can build or buy 3 power supplies with the hope that the board powers up and works properly and then those power supplies become the permanent replacement for the entire blown U39 module.  If the counter is dead, you repurpose those supplies as lab supplies.

You could also build or buy three lab supplies that cover the voltages necessary.  In that case, you end up with three nice lab supplies and then, if the board works, build or buy 3 dedicated supplies to repair the counter.  You end up with three lab supplies.  You can *NEVER* have too many power supplies in your lab! 

Yeah, agreed. What I would like best as end result is to have a working counter, with no visible modifications on the outside. Inside is a different story.
And I am willing to put some effort into achieving that result. That is why I would like to know up front whether or not it is worth while going down that path.
So my preference is the second approach, with as extra goal to build the 3 dedicated supplies into the counter.

Same thing applies to the Fluke 7261A. I'd like to know if it still works, before I decide to rebuild the primary side of the transformer.

One complication to keep in mind is that the input to the U39 module is high voltage DC rather than AC.  I don't know if typical AC switching power supplies will tolerate high voltage DC input.  You might have to rearrange the input power to feed them with AC.

The conversion from AC to DC is done with a straightforward rectifier. That rectifier should be able to deal with momentary mains peak voltages, so I would expect it to be ok, no?
And if the replacement power module has AC input, it can also be hooked up tot he AC line coming into the box.

I'm not seeing any analog switches.  The oscillator itself is U1, 54LS320 which is a dedicated quartz oscillator chip, followed by a 54HCT393 which is a ripple counter, i.e. a divider.  In any case, you had a solid 10 MHz signal coming out of the oscillator board and I don't see anything between the oscillator board and the power supply board.

I went back to the manual, and saw that indeed I misremembered: there are analog switches, but they are in the modulation circuitry, not in the 10 MHz path.

I don't think you stated what the amplitude of the bad signal was.  Did it look like a solid signal or something that was basically noise being coupled in to the output line and then amplified by Q3?  Bad solder joints or other circuit faults anywhere in the system are a possibility, but the clean 10 MHz coming out of the oscillator board but not on the output connector points toward the power supply board rather than the oscillator board.

Indeed I didn't, and indeed it may be that it was just a stray signal weakly coupled into the 10 MHz line. I wish I had made some scope screen shots... 
In analyzing the fault location, I started at the connector, and worked my way back to the oscillator, making big steps. Once I found the signal,
I followed it forward towards the connector, making small steps. And to my surprise, I could then follow it all the way to the connector. (Simplified account, but basically correct.)
I expect that at some point in the future, the 10 MHz signal disappears again. Then I will do the same thing, but with improved recording of my findings!

Rubidium standards do drift over time.  Considering the age of the unit, recalibrating it by comparing it to your GPSDO isn't a bad idea.  It will be challenging to do it with a scope, you really need a time interval counter.  The frequency will also shift slightly with temperature and input voltage.  But since you've got that pot, that's a good starting point.

FYI, the PM6681 power supply is very close, if not identical to the PM6685.  The service manual warns that if you touch the +5V trimmer, you have to adjust 'the complete instrument', whatever that means.  The PM6685 service manual doesn't say that, but since it's so similar, keep that warning in mind.

Ed

Yeah, if I get the entire box working, then I can use that to measure my GPSDO, and adjust the pot for an exact 10,000,000.000 reading.
That would be the best I can do, as my other counters have less digits! I have a few more 74LV8154 lying around,
so I am tempted to use those to make a dual 12 (or even higher) digit counter, with the help of a micro dealing with overflows. Hmmm.... 

Cheers for now, variable supplies are ordered, I will report back once I have rigged up the whole shebang.   

[edpalmer42]

One complication to keep in mind is that the input to the U39 module is high voltage DC rather than AC.  I don't know if typical AC switching power supplies will tolerate high voltage DC input.  You might have to rearrange the input power to feed them with AC.

The conversion from AC to DC is done with a straightforward rectifier. That rectifier should be able to deal with momentary mains peak voltages, so I would expect it to be ok, no?
And if the replacement power module has AC input, it can also be hooked up tot he AC line coming into the box.

It isn't clear to me why a switching supply would have trouble with high-voltage DC input.  The only thing I can think of is that with DC, the input voltage won't ramp up at mains frequency, it will just hit hard at maximum voltage.  Maybe the internal surge limiter wouldn't be able to cope with that.   

Cheers for now, variable supplies are ordered, I will report back once I have rigged up the whole shebang.

Good luck.  Let us know how it goes.

Ed