Fluke 732A - ripple on output

[Rax]

Hi all,
Just managed to snatch my first primary-grade voltage standard, and I am seeing some ACV on the output.

I was aware of the potential of finding ripple on the output, so the first thing I did - before powering it up - was to replace the three axial electrolytics. One had clear signs of trouble inside - rubber cap being crumbly dry and bulging - and I replaced all three of those with long life axials (27.5k hrs PEG124RL3150QE4 for C1 on the Pre-Reg and 8k hrs MAL211819471E3 for C3 on the same board and also on the Reg board).

I enclose a scope reading of the output. There's indeed about 70mV RMS at about 293kHz.

I am continuing looking at others caps, and I also plan to grab some batteries today and see if it's happening in floating mode too.

Appreciate input and maybe similar experience and hopefully solutions.

[Rax]

Some pics and screenshots.

[Rax]

Obviously, trying to cure a 293kHz ripple with repairing the main PS caps was quite a load of wishful thinking - which I was conscious of, but I planned replacing those anyway. I was hoping there would be more of an effect on the issue at hand, but I'm seeing no improvement.

Being AC out, I am still suspicious of the caps in the PS, but I'm running a thin and diminishing hope I'll avoid having to mess with the reference inside the oven - which I was hoping would not be the case.

[bdunham7]

Is there any possibility of this being due to external interference?  Can you try shutting anything and everything off other than the 732B and your scope?  I had an issue with the LED lights in my workbench where they had a very large conducted interference at 238kHz.  One of the instruments that was badly affected by this was my 731B.  Are the batteries at all OK in your 732A?  i don't know what is in the unit itself that could possibly cause this.

[Rax]

Is there any possibility of this being due to external interference?  Can you try shutting anything and everything off other than the 732B and your scope?  I had an issue with the LED lights in my workbench where they had a very large conducted interference at 238kHz.  One of the instruments that was badly affected by this was my 731B.  Are the batteries at all OK in your 732A?  i don't know what is in the unit itself that could possibly cause this.

I think you're right on the money, as this ripple (exact frequency and level) was disclosed prior to the sale. So my assumption prior to committing to the purchase was a "frugal" tech in the seller's shop didn't consider the possibility of EMI from LED circuitry and such, exactly as you posit. Additionally, I couldn't put my finger in anything inside the 732A that could generate this odd garbage on the output. I was willing to bet a decent chunk of money there's nothing wrong with the unit and the seller got the short end of the stick.

But I am seeing all of that at my bench too. Admittedly, the two benches - mine and theirs - may as well be inflicted by having essentially the same LED drivers... I'll look into whether turning off all lights here would suppress this garbage.

[bdunham7]

I think you're right on the money, as this ripple (exact frequency and level) was disclosed prior to the sale.

That would seem to be too much of a coincidence.  But there aren't too many things in there that would oscillate.  Perhaps somewhere in the vicinity of U1 on the A5 board?

[Rax]

Been working on this:

• Got a set of fresh batteries, and the ripple is still there (!)
• Replaced the line filter, suspecting it's shot and lets through all the switching crap pestering my household
• Replaced all tantalum caps on the pre/regulator(s), in the off chance that, for instance, the battery charge Schmitt trigger is becoming a relaxation oscillator (still holding a magnifier at this hypothesis, and trying to figure out if I should suspect some other weak or failed part turning the ST into a RO)
• Replaced the rectifying bridge, in case there's some odd ringing I end up seeing at the output
• Replaced R12 (.39ohm) and waiting for R13 to be delivered, in case they became noisy (although I really doubt that's what I'm seeing; that said, those two resistors definitely have to be replaced)
• Between TP2 and TP1 in the regulator board (the main 18.6V rail into the reference), I see a couple of mV of crap (can't lock on a specific frequency)

Needless to say, none of the above mitigated anything.

I'm afraid this seems to point to issues with the reference board inside the oven, unfortunately. Not sure what inside the oven can cause this, other than the op amps (which, if I gather my spirit and open the oven, I may try to replace).

[Rax]

Also, although sometimes the output stabilizes at what my environment thinks is about 6ppm high of 10V (sometimes for good stretches of very good stability), I'm often getting variations which can be as high as 50ppm or more (I have both my Prema and the Datron 1281 looking at this and they agree with these variations). Both meters are running filtered, but I am still thinking the oddness with the output in DC is just registering what is going on with this AC garbage out.

The oven thermistor is at 4.52kohm and change, but very stable.

[Bill158]

Radu:

You are just postponing the inevitable, taking out the oven/reference assy and finding what is causing the oscillation.  I took one out a long time ago.  I followed the guide in the manual.  My biggest fear was the ribbon cable breaking at the reference module end.  I found that if I turned the 732A on it's right side, as stated in the instructions, it seemed to be easier to remove the module.  I agree with bdunham7 that the problem is probably around U1.  As I vaguely remember once you take the top off the module, then the PC Board comes out when you take the retaining screws, or whatever, out.  You are NOT seeing anything like what you are seeing on the output, at the +18.6 volt test point.  So the regulated supply is working fine.  The insulating foam can be very brittle after all these years of being heated.
Like Dr. Frankenstein once observed "There are some things that man was not meant to tamper with" but this isn't one of them.  It is hard but not impossible.  My experience with the 732A is that once you are done fixing it the output voltage will recover very nicely to it's previous value.
Good luck, I know you can do it.
Bill

[Rax]

I took one out a long time ago. 

Bill: some of the information I'm finding on this is that the reference PCB is conformally coated. Is that what you've seen? If yes, difficult to remove, and, I guess, reapply? Any tips?
Obviously, not to put you on the spot, any others having info on this are welcome to pitch in.

[bdunham7]

the reference PCB is conformally coated.

Oh, yuck!

https://paceworldwide.com/video/pace-rework-and-repair-conformal-coating-removal

[floobydust]

When a few electrolytics go bad, I figure it's time to check all and replace them all. It's just less hassle in the end.
I don't know how the tantalums age in this product, but I would test them too.

I believe you've replaced C1, C3 on the pre-regulator board. C5 I would test.
That leaves:
Four on the A4 Regulator board:
C1 330uF 80V
C2 82uF 20V tant.
C3, C4 - 10uF 35V tant.
C9 22uF 25V tant. (across heater)

One at the front panel 10V output jack terminal strip:
C1 82uF 20V tant.

There are no electro's/tants on the Ref board that I can see.

You can check if the heater regulator is oscillating at TP1 (+HEATER).

[Bill158]

I took one out a long time ago. 

Bill: some of the information I'm finding on this is that the reference PCB is conformally coated. Is that what you've seen? If yes, difficult to remove, and, I guess, reapply? Any tips?
Obviously, not to put you on the spot, any others having info on this are welcome to pitch in.

I did this in May 2009.  I can't remember if the PCB was coated or not.  I went in to change RT-1.  As I remember I simply heated up the two solder joints on RT-1, pulled it out, cleaned out the holes and put a new one in.  This was all from the component side, upper, of the PCB.  I did all of this without removing the PCB.  I didn't worry about the coating around the two solder holes.
Bill

[Rax]

I think I can report progress... After a bunch of hours in a joint venture work session with neighbor bdunham7, I think we figured it out. I'll post a more detailed report later, but it seems the compensation scheme of U1 has become compromised over time due to baking and drifting of parts and IC vs. a likely marginally stable design. We've adjusted the value of R25 for a more aggressive compensation. It seems the internal oscillation has cancelled due to this adjustment.

Most of the hours to get this done was spent on mechanics of taking this apart and putting it back together. Overall, a very puzzling industrial engineering design product. We agreed a weekend project by either one of us may be more easily serviceable than this turbo-diy-ed product (as much as this delivers stellar performance, but the criticism doesn't intersect with that point). And a project nearly impossible to get done without two pairs of hands... By oneself, this wouldn't be fun at all.

I'll monitor how this behaves in the next days, weeks, and maybe months, and report back here.

[Rax]

BTW - no conformal coating at all we could detect. As easy desolder/solder as one can get.

[bdunham7]

I'll monitor how this behaves in the next days, weeks, and maybe months, and report back here.

Did that front panel voltage shift issue go away?  I'm wondering if it wasn't just my test leads--perhaps we should have been more careful and used a low-EMF z-plug to bare wire setup.

I've rethought adjusting the unit to nominal and perhaps others can chime in.  My first thought was that since we've upset it with a repair, it will need time to drift to a new steady state anyway so why not go ahead and trim it as close to 10.000000V as possible.  But perhaps just changing the compensation of U1 won't actually destabilize or shift it at all and it may quickly settle to its old value--which, b/t/w, we don't know. 

[Rax]

Did that front panel voltage shift issue go away?  I'm wondering if it wasn't just my test leads--perhaps we should have been more careful and used a low-EMF z-plug to bare wire setup.

I've rethought adjusting the unit to nominal and perhaps others can chime in.  My first thought was that since we've upset it with a repair, it will need time to drift to a new steady state anyway so why not go ahead and trim it as close to 10.000000V as possible.  But perhaps just changing the compensation of U1 won't actually destabilize or shift it at all and it may quickly settle to its old value--which, b/t/w, we don't know.

I'm not seeing much crankiness at the front connections (and haven't before either) but I'd need to further investigate to tell for sure. I've always used "AB" low-emf cables with this, which I feel are very solid, so maybe that helped. It may be worth opening the front panel if only to see if the parts there are nominal, particularly the 82uF tant. And maybe the diode for leakage.

Currently, running this seems to maybe make the fluctuations I'm seeing - within about 1ppm of 10V - diminish ever so slightly with passing of time. But it has only been a couple of days, so a little premature to say. But before the repair I was seeing pretty wild flares of tens on ppm, probably triggered by the instability of the op amp.

Right now, per my Prema, this is mostly on the "underside" of 4ppm high of 10V, and as I said, seems to flutter inside a ~1ppm window (which is a little wider than I'd like to see it). Of course, my meter and any number of things may also factor into this "noise."

[Rax]

Well, further investigation shows there's indeed a problem there. I'm not sure if the hours of work during open heart surgery has maybe worsened something that was there before, or I just didn't notice, or whatever it was.

But the symptoms are this:

• The 10V HI terminal is sensitive to mechanical action. Fiddling with it a bit can restore it to its about 4ppm high which is what I regard is its "stable" or "true" value.
• Sometimes it'd just drift away from there. Touching a bit the HI terminal restores its level.
• Touching it can also make it jump but a few tens of ppm.
• It can also drop to "zero" - in fact, always about 14mV. I have no good explanation for this (other than that the cap and others parts on these terminals - 10V HI and LO - have some sort of leakage that elevates HI from closer to zero. Then, why 14mV?...

I've cracked open the front panel - note this "PCB configuration" doesn't seem documented in either of the manuals I am aware for the 732A - and the best I can come up with is the #23 wire being connected to the HI binding post via a solder joint that looks pretty untidy, and then also a trace. Also, if the LO terminal has everything soldered together, including the nut, the HI doesn't, so it's possible, because it relies on the mechanical contact between the nut, washer and the board pad, it may not connect perfectly. That #23 has to do with supplying U1 internally, so maybe that node being mechanically sensitive to action on the front terminals can explain the full drop of the output.

I dread working on the front terminals - almost as much as inside the oven - but I could:

• Infiltrate a micro-drop of deoxit at the nut-to-pad interface
• Rework that solder joint, and clean well with IPA.
• Gently tighten the nut to enhance the galvanic contact with the board PCB.

[dietert1]

I'd guess the bad contact is between the body of the binding post and the PCB trace and turning the nut and the washer can solve the problem. The problem may be linked to the instability observed before, if the red capacitor on the board is affected as well.
Before soldering a "pure copper" binding post body, you need to loosen/separate the plastic parts as they will crumble.

Regards, Dieter

[Bill158]

Well, further investigation shows there's indeed a problem there. I'm not sure if the hours of work during open heart surgery has maybe worsened something that was there before, or I just didn't notice, or whatever it was.

But the symptoms are this:

• The 10V HI terminal is sensitive to mechanical action. Fiddling with it a bit can restore it to its about 4ppm high which is what I regard it's "stable" or "true" value.

Radu:

The 732A that I had the oven out of to replace the thermistor also had another problem which is exactly like what you are seeing.  Actually from the beginning with me, this 732A had two problems.   First was the internal temp kept dropping, very slowly, according to the resistance measurement of the thermistor from the front panel terminals.  That is why I tried a fix, replacing that indicator thermistor.  That solved nothing!  So someday I will have to work on that issue.
But the second issue was that the +10 volt output was unstable, slowly rising but not out of spec.  Then one day a few years ago the +10 volts started rising at a fast rate, to over 3 ppm in a month, every time I measured it.  So I did the same as you and wiggled the terminals and sure enough it suddenly dropped back to where it had been several weeks before.  I took off the front panel and I didn't like the looks of the solder joints either.  So I reflowed all of those 23 ga (?) solid wires from the oven to the banana jack terminals and the unit has been absolutely stable as can be, and it was back to it's original value a month before.  That was done on 12/28/21.  Since then the output has only risen about 0.4 ppm or less.  At this level it is hard to be certain without a JJA.  So I have had very good results in reflowing the connections! 
Bill
[/list]

[Rax]

I reflowed all of those 23 ga (?) solid wires from the oven to the banana jack terminals
Bill

Bill, thank you for this, it's pretty reassuring and seems like good confirmation I'm on a good track with this (another long saga!...). You also seem to confirm your sample has this PCB implementation, which I don't think is very typical of the 732A. Maybe a later revision?...

If you recall - did you also have to tighten the nut(s) on the post(s)? As an over-planner, I am trying to figure if the risk of harming in some way the numerous manganin wires soldered on the tip of that post by manhandling that binding post is worth it. A very careful path would be to reflow and clean those joints, and then put it back and watching it for a while. And only trying the tightening if it seems necessary.

[Rax]

The problem may be linked to the instability observed before, if the red capacitor on the board is affected as well.
Regards, Dieter

We've considered that as well, but wagered the parts at the output address transients from the load rather than instability of the reference circuitry. The only reason I'd see the former as likely is out of consideration for the length of those wires from the reference module to the front panel. But once the oscillation was canceled, that seemed like a pretty neat and deterministic resolution. I feel some oddness of those parts may have more to do with the 14mV leakage, but I guess I'll have to continue working on curing the symptoms and see where it takes me.

[Rax]

I reflowed all of those 23 ga (?) solid wires from the oven to the banana jack terminals
Bill

Bill - I think I misread you somewhat and I now understand you reflowed all of the wires - thick and thin, as there's a large variety - to the actual rear post of the binding post. I was thinking of just reflowing #23, and possibly #28, as they don't connect to the binding post itself, but the PCB pad and connect to the HI node via a trace.

As I was heading to doing this, I realized these may be done using Cadmium-based solder (which I'm almost positive not to be the case). Is that known regarding these? Are they using regular solder? I haven't seen that mentioned anywhere (including Todd M's article: https://xdevs.com/fix/f732/), but I'm very squirrely about chemicals and heavy metals. 

[Rax]

The other question I can't find an answer to is why are #23 and #28 soldered to the PCB pad, instead of the binding post directly, as all others?

If I'd trust the other wires would stay attached while reflowing the blob of solder on the post (which is the Pandora's box I'm dreading), I'd remove those two above from the PCB pad and solder directly to their respective posts.

[Rax]

I've reflowed specifically the #23 solder pad. The next step is tightening the nut, but I need a correct size wrench - I'm not willing to risk #33, a nearly hairline gauge size wire that solders to the same post with trying to tighten with pliers or other inappropriate tools - and will get it tomorrow.

After just reflowing #23 I feel it's much easier to restore to stability the 10V output by touching the posts a bit. It feels like the "dysfunctional state" is the exception, statistically (easier to restore than mess it up). I feel I'm on the right track with this. Now apparently, the output is a little more than 2ppm of 10V in agreement with my Prema, which I take as another good sign. This is from about 6ppm at relative stability before.

I really can't figure the industrial design employed by Fluke in this - the way they arranged the front connectors is one further puzzling set of choices - and it'd be interesting to see how long this "industrial DIY" instrument was built according to this design. I think the "A" was relatively short, and I think I understand why.