Fluke 732A with popcorn noise SZA?

[TiN]

I know, we already have many 732 threads. Yet, I don't remember anyone talking about popcorn noise issue in them as of today.
Well, there is place for everything, so couldn't resist recent lures, and bought one of those 699$ sick 732A's.

Box arrived yesterday, and quick check revealed poor darlington on AC mains board cooking to +130 °C.



Quick swap with TIP122 (original is lower voltage TIP120) fixed that issue, so I hooked the output to some random meters:



After a day of runtime, I observe "popcorn spikes" on the 10V output , and buried in the noise 1.018V output. 
Yes, I am sure it is not the 3458A (meter used for 10V 732A log.



Jumps are random, at -0.4 ppm level. 

Guess I am looking at replacing SZA263 reference chip with a replacement? Anybody else here have this failure mode in their 732 unit, to share the experience?
Reminds my jumpy LTZ1000A chip, which I still have somewhere.

I have 8842A for donor parts, so probably least resistance path is to steal SZA chip from it and put into this 732A. Of course I'll have to test SZA in 8842A first, to make sure that one is quiet.
Also will replace all electrolytic capacitors in 732A, but I doubt those can cause the noise like this.

[Andreas]

did you check the zener itself?

might be a mis-handled output stage or capacitor.

with best regards

Andreas

[Kleinstein]

The noise does not look 100% like the popcorn noise seen with the jumpy LTZ. This one is more like random drops - the more usual type of popcorn noise is with same size steps. It is still possible to get peak like behavior if the lower state is only rarely populated and short lived.

[martinr33]

732as can pick up noise from transformers through induction, switch  mode power supplies - and UPS systems. I'd get it away from your meters, and maybe even try to run it off of a battery, or a clean 30V supply.

I often wonder if Fluke used the SZA in so many meters because they needed a few really good ones for the 732a, so I wouldn't replace it just yet (they may also have gained some economies of scale from doing that, so it is just a hypothesis).

[TiN]

Sure, more testing is required before doing any ref lobotomy. I also have spare 5700A refdac , but rather not take that one apart just yet.

Graph chart have averaging math applied, so what matters are raw sample (circles in blue).

[Pipelie]

the curve doesn't look likes popcorn at all.
it probably has a fault in power supply.
and due to the averaging math applied on Keithley 2002, I can't find those pulse on curve of 1.018V. 

[Echo88]

Dont sacrifice the 5700 RefDAC, it hasnt done anything wrong. 

[Kleinstein]

Checking the supply (e.g. measure in parallel to output) could be a good idea. The faulty TIP120 might not have been the only problem there.

[floobydust]

Checking the supply (e.g. measure in parallel to output) could be a good idea. The faulty TIP120 might not have been the only problem there.

+1  I could not find a TIP120 Darlington on the 732a schematic, and
never met a transistor that runs hot because it is defective. It's usually an overload causing it to heat up. A shorted cell roasts many gel-cell chargers.

My second guess is RF interference from cell phones etc. as the 732a does not incorporate any ferrite beads or common-mode chokes that I can see. It's from an era when the airwaves were quieter. I would try some clamp-on ferrites to the outside test leads and see if the (~4uV) jumps clear up.

edit: it's Q1 on the A3 pre-regulator board.

[ManateeMafia]

The TIP120 is listed as Q1 on the A3 pcb in the newer 732A/AN manual.

When he gets the parts to change on the power supplies and new batteries, the problem just might clear itself. The 18.6v supply to the reference would also be a good place to monitor without going too deep.

[dl1640]

A local second-hand offered me about 1300USD for one 732A without batt. I am struggling.

通过我的 PRA-AL00 上的 Tapatalk发言

[TiN]

I decided that January 1, 2019 is a good day to get back on this project. After bit of struggle I got into oven box and took the reference board out to the day light.





Nasty corrosion on the tracks and even component legs was immediately detected.



Maybe it cause excessive leakage and unwanted resistance paths? Maybe not... But this is not the way precision circuit should look like, so off we go, to remove all parts, inspect them, clean them and refub the PCBA.



Mark all the connections and remove the wiring harness from the board. Done..



Now we can go ahead with removing each resistor, carefully marking where it was originally located. One could argue that soldering stress cause resistor to drift/shift...



You right, but because I got this unit with zero history, and it must be stabilized afterwards for weeks/months anyway, I don't see much harm already.



We can however see lot of area to clean up, and it would be nearly impossible with all those tall Fluke's wirewound sticking around.



To be continued.... 

[chuckb]

TiN
Your bravery really has no bounds!
Have a great new year.

[branadic]

Good luck on that road and happy new year.

-branadic-

[alm]

Let's hope that the soldering heat does not affect the annealing of (manganin?) wire on those mica card resistors, and affect the temperature compensation. Although I would expect the resistors that are critical for stability to be the hermetically sealed ones.

[chuckb]

FYI
I will be trying some low temperature solder to help minimize thermal shock on sensitive components. Mouser has -
https://www.mouser.com/ProductDetail/910-SMDIN52SN48

Which melts at 118 deg C verses 182 deg C (?) for normal 63/37 solder. This low melting temperature along with a normal heat sink clip should minimized thermal shock to resistors and zener diodes.

I don't know the thermal EMF generated by this solder but it can't be worse than the 34uV / deg C for a normal Kovar / copper connection.

I will also use a dedicated soldering iron tip for this solder to minimize contamination with normal solder.

[lowimpedance]

Happy New Year indeed !.
 Just a thought  , since you have infinite patience for restoring gear that most would have junked after one look at it  , how about re creating the PCB (I assume its bog standard FR4 by the look) and transfer the now removed precious parts to the new PCB !?! .
 As I said just a thought   .

[TiN]

how about re creating the PCB (I assume its bog standard FR4 by the look) and transfer the now removed precious parts to the new PCB !?! .

That idea crossed my mind, but decided not to do it as LTFLU/SZAs are not trivial to source and trim procedure for them is tedious. Making board for just one unit repair rather pointless, as there is long list of more interesting projects already. So for this 732A I'm going for least resistance path to get it going and be done with it. Otherwise we will end up with putting LTZ1000 chipsy inside , you know how that goes...

I have better reference box to work with than this old 732A, and need attention there before the exciting calibration this spring.. 

[Kleinstein]

It looks like some of the flat wire would resistors are not actually used, but just there to be eventually used if jumpers for trimming are set different.

The corrosion does not look that bad to do real damage to the traces. So a good cleaning should be enough to restore it.

Any idea what caused that type of damage ?
Kind of looks a little like a leaking cap, but I don't see a candidate for this.

The corrosion looks bad, but it is still possible the popcorn noise is still from the reference chip. So one may want to do a check before full assembly back to the oven.

[TiN]

Alright, slowly but surely, we have some progress:

Condition before:



Board stripped of precision wirewounds and ready for cleaning (SZA and film caps also removed temporarily in process)



Condition now:



I messed up few parts on PCB, but should be alright still. Also replaced plastic LM358 with hermetic fancy-pancy one. 

As by product of resistors removal I did actually run temperature coefficient plots on most of them. I can tell so far, that these "gold sealed" Fluke resistors are one of the best wirewounds I ever tested, typical WW's hysteresis is suprisingly very small.

Plan to hook up harness and put it all together into oven for a smoke test tomorrow.

[ManateeMafia]

Looks better.  I was hoping to see a new xDevs red 732Z pcb. 

My last 732A purchased at the same time as yours is a high plains drifter. I think it drifted down 60+ ppm before I unplugged it. Maybe mine had the same corrosive damage?
Have you considered 1v and 100mV outputs like the 732C?

[Inverted18650]

thumbs up brotha, hope it functions better and settles in.

[TiN]

My 732A is not a drifter (afaik), but a jumper .

Well, and all that teardown and cleaning was all for nothing.



Same ~0.5ppm jumps as before. To prove that it's not fluke in the measuring 3458, second meter connected to 1.018V output shows correlated jumps on it's data too.
Data off the chart in first hour or so is just warming up. Oven thermistor measurement provided by third meter (OHM 10K range, dark-red line).

Now I swapped original SZA263 reference chip to SZA263A from Fluke 8842A donor board. Yes, I know the resistors are not matching this other chip, but I don't care about tempco much at this point, just want to see if jumps are indeed sourced from refamp or not.

[Inverted18650]

 

[TiN]

 

Initial data from 8842A SZA263A chip looking promising.
Obviously tempco/stability is thru the roof, as resistors on 732A board are not matching the refamp? It's going down at rate of about -2 ppm/hour.



But no spikes yet! 

Setup rig capture from livecam. Excuse the poor IQ.



Of course there could be big possibility that SZA263A sorting for 8842A is way less strict than sorting/conditioning process for 732A-grade SZA, so I could end up trimming resistors till the cows come home to no avail.

Which make me thinking to ditch all this SZA/LTFLU non-sense and make quick and dirty PCB around good proven LTZ1000A ref design.
Just reuse external 732A oven regulation circuit and output stage resistors to form 10V/1.018V/1.0V.

I know at least one more person that might be remotely insterested in such a zombie board?   

What say you? Yay for LTZ-based 732Az or leave the thing rest in peace?