Agilent 34401A repair log

[golf32]

My work was getting rid of a beautiful Agilent 34401a. It was in a pile with craploads of other test equipment (most not worth taking, but if I get around to it I will post the repair log of other stuff I got). It had a tag saying that the ohms ranges were way off. I didn't write them down, but it was throwing basically every ohms range error code on self test. I started my troubleshooting by measuring the current coming out of the front panel when measuring resistance. Every range was basically zero.



Some quick theory on the ohms range current source from the manual:

U201-A, Q201, R201, and R202 work together to turn the reference voltage into a current which produces a precise voltage drop across the 28.57k resistor in U102D. U201-B and Q202 work together to replicate the voltage drop across the other resistors in U102-D to create a constant current. Q203-Q210, CR202, and Q211 are a protection circuit to prevent overvoltage at the input.


My first port of call was the reference voltage drop from the 28.57k resistor. When I measured it, it was way out of spec. I suspected the JFET Q201, so I removed it. My transistor tester measured it as two resistors, meaning my suspicion was correct and the part was bad.


Next, I bypassed the protection circuit by lifting one leg of Q202 and CR202 and jumpered them. I was still not getting the output current required and suspected Q202. I measured it out of circuit and again, it was dead. Finally after that, I was getting the correct current readings at the front panel. When I reconnected Q202 and CR202 however, the current was off. This made me suspect the protection circuit. I didn't want to troubleshoot each individual transistor, so I replaced them all. I tested the jfet Q211, and lo and behold, it was dead. I did not test any of the BJTs.


Finally, after all the replaced parts, it did not error out! upon testing with my fluke resistance standard, the readings were nearly spot on. Pretty great considering the non-exact replacements and poor job at cleaning flux.




I hope this helps someone in the future with this same issue. Don't trust the jfets!

[floobydust]

I believe the 34401a Ohms source does not have protection for -ve going transients, it only has protection from +ve overloads by CR202. So the JFET Q211 gets hit and does fail, or worse.
Also note there is an ECN for early S/N's below 3146A30500 that made changes to the Ohms source, was R207 5k62 and CR203 4.7V zener.

[Kleinstein]

The block of BJTs provides the protection for negative going voltage. This should reasonably work up to some -1000 V, though maybe only for a limited time with the 1 mA test current. The voltage rating of the risistors is also a bit on the low side.
Very fast transients (e.g. ESD) could be a problem, as the transistors and CR203  have a limited speed.

It is still odd to also have a bad Q201, as this one is quite a bit away from the input.

[Dr. Frank]

Finally, after all the replaced parts, it did not error out! upon testing with my fluke resistance standard, the readings were nearly spot on. Pretty great considering the non-exact replacements and poor job at cleaning flux.


I hope this helps someone in the future with this same issue. Don't trust the jfets!

Hello, I completely disagree with your conclusion, and I think that you did not repair the DMM properly at all.

The readings are not spot on, but completely out of specification.
100Ω range is -227ppm, 10kΩ is +400ppm, 10MΩ is -2800ppm.
These extreme deviations I call "still defective".

Btw, you should set the 34401A to 6 digits, and always use full scale reference values, if you want to make a statement about its accuracy, which might be down in single ppm figures.

I "dislike" your analysis and repair method:
You use too much trial and error, instead of extensively testing all voltages in the circuit and from that find the root cause, and in turn ALL defective components.
I'm saying that in a harsh manner, for several reasons.

First, do not try to adjust the DMM in this condition, that would hide the faults into the calibration.
Second, you still do not know, which kind of event caused the damage, and vv. you can't derive which components as well had been damaged by this event. I'm wondering, if you always have used the equivalent/correct replacement types, especially  for the FETs.
Third, your trial-and-error method should not be an example for others.

If even Q201 REALLY was defective, how should this have happened?
All the components towards the CC output must have been damaged as well, especially U201.

I don't remember precisely at the moment, but the value of V_GS of Q202 was crucial for correct operation, I think. Maybe it was even a selected part? See BOM.
Yep, here you go:
Q202 1855-0863 1 Transistor J-FET P-Chan D-Mode 27014 MMBF5461SEL

Please search for similar 34401A repair threads, where we had documented the correct voltages of the circuit, and retrace that for your DMM.
In most cases, only U201 was defective, causing the same failures which you described initially.

Frank

[strawberry]

ohms range protection work up to +/-300V
Q211 high leakage  ~500Mohm or test method
My 34401A got on repair bench for ~20ppm but 2800ppm is crime to universe

[wje]

"My 34401A got on repair bench for ~20ppm but 2800ppm is crime to universe". Love it, hilarious.

I really like my 34401A as a regular bench meter, but I'm currently having a crisis because my 2 Datron 4910's disagree by 2ppm, and my Solartron 7081 doesn't agree with either of them, although it's between the two. Life for us electronics guys is so hard.

PS - I sent one 4910 to Fluke for calibration against their JJ array some years ago, that was way too expensive to repeat. Unless I can't stand the uncertainty anymore. And yes, I'm also a volt-nuts member.  At least modern GPS disciplined frequency refs meant I sold my cesium-beam std.

[golf32]

I don't know how much I agree with your idea that too much trial and error is bad, but different strokes for different folks I guess. If I had the time I'd measure the voltage of every node in the circuit and do the math to figure out what's wrong, but I don't.

I never did the math to check if it was in spec, I was just happy to see it close. I'm going to calibrate my resistance standard before I do anything else. It probably hasn't been calibrated in decades.

I never expected the high resistance range to be entirely accurate. I did a crappy job of cleaning flux, so leakage is my first thought.

I thought Q201 being dead was strange too, but it's kind of a moot point if the custom ICs are dead. Maybe I killed it with my fiddling. Haven't looked into replacements but I'd imagine it's hard.

Thanks for the feedback all.

[strawberry]

https://www.eevblog.com/forum/testgear/34401a-ohms-drift/

My 34401A was calibrated being faulty and drifted again. possible reason it was dumped on ebay

left std Cynel 60/40 solder hard flux on PCB and couldnt register difference (dont have reference at some ppm range to compare with)
cleaned with IPA ~99

[Kleinstein]

There is a chane that Q201 was not bad: with Q202 not working  there can be quite some current between the inputs of the U201B  OP-amp and this may look like a broken Q201.
Transistor testers are often not that great recognizing JFETs, expecially those with a higher threshold.

Usually the 34401 does not drift very much and often is in spec even after decades. Some of the difference seen could also be from the reference resistors used.  Flux residue and related leakage could explain the higher resistance ranges beiing bad.  It is not a good idea to just get a new calibration, because the leakage current from flux residue can be quite variable, and may change with time and humidity. The correct way is to get it cleaned well. Some flux types may be OK, but it is hard to know upfront and can depend on the temperature profile.

If one has a suitable meter for the small current, one could check if the current of the ohms current source in the 1 M and 10 M range depends on the DUT voltage. Ideally there should be not much change as this would be something like extra resistance in parallel to the DUT causing a linearity problem.

Another possible problem is U201 being damaged in a way to have higher than normal input bias or offset. Changing U201 could effect the calibration noticible.

[Dr. Frank]

Hello golf32,
obviously you still didn't get the point, that your DMM is still defective.
the readings are not even close, they are extremely way off. For the 10MΩ  range, instead of speculating that it wouldn't be that precise, better read both specifications of the 34401 and the 5450A, and you'll see, that 2000ppm is an order of magnitude off, at least, I'd guess.
if you you'd caught all the failures, and not implemented new ones by wrong Q202, your DMM would deviate not more than maybe 20ppm. That's the only math you have to do.
Even the 5450A is very stable, and probably wouldn't deviate more  than 20ppm in ranges up to 10MΩ. Especially 10k is ultra stable.. so definitely there's still a big fault in your 34401A.
You also don't have to measure every node, and you don't have to calculate anything inside the circuit.
please search for these diagrams, where me and others documented the reasonable voltages. From that you can easily derive, where your errors are residing. Especially Q202 is crucial.
I expect, that the OpAmp is defective as well. That can easily be seen, if + and - are not exactly on the same potential. That's really no big effort, and requires no math at all.
Another volt-nuts had a defect Q202, but the circuit dud not work properly before a correctly selected V_GS was assembled.
About the method:
You suspected all three JFETs to be defect, plus the HV transistor chain.
as others said already, transistor tester sometimes/often, I don't know, don't handle these correctly. So I really doubt, that all are defective.
A better way would have been, to properly identify really defect transistors, so you could trace the assumed over voltage event, how it digged backwards into the circuit. Just blindly replacing parts does not allow to trace that any more.
Frank

[golf32]

Well I'll do some more poking. I believe I already measured the +- differential on the opamp, but I will see if any of my "fixes" have changed that.

[Dr. Frank]

https://www.eevblog.com/forum/repair/hp-34401a-error-612-613-615-617-618-619-621/

In this thread, there are many failure modes described, and you'll find this diagram with the faulty or correct voltages.
The case with wrong Q202 V_GS is not contained, so please search for yourself, using a similar headline
Frank

[oz2cpu]

i do agree with Frank, sorry..
you meter is still too much off,

but come on, be happy :-) and be friendly frank :-)
i know you, and i know this hp meter, it is WOW nice and good and stable,
and deserves respect and to be handled with white gloves on..
But :
it is alive, clearly found the spot and the problem, keep digging, replace with the correct parts,
be sure to clean the area very good, perform all the calibrations, and then see if it can perform to spec again,
a super score, i take it any time of the day

[Dr. Frank]

i do agree with Frank, sorry..
you meter is still too much off,

but come on, be happy :-) and be friendly frank :-)
i know you, and i know this hp meter, it is WOW nice and good and stable,
and deserves respect and to be handled with white gloves on..
But :
it is alive, clearly found the spot and the problem, keep digging, replace with the correct parts,
be sure to clean the area very good, perform all the calibrations, and then see if it can perform to spec again,
a super score, i take it any time of the day

Yes I know that I'm sometimes too direct.. that's my way anyhow, and nothing personal, of course.
42 years ago, @ Airforce Lab, when each instrument had to be field repairable, we had to order expensive HP spare parts, 10times the regular price.
Therefore, I've learnt to analyse first, and then only to replace parts on root cause find.
I've been doing failur analysis this way for the last 40 years, so I will always un-like trial-and-error methods.
In this case, I'm really interested, which parts really have been blown (check the replaced parts by proper methods), which further components are damaged and what was the initial event, probably.

[strawberry]

now we know why repairs are dumped
parts cost 10x and must pay someone to do repairs
new instrument would be cheaper they say
only f..er are those who cant afford new instruments
less engineering "better" for industry
more cr..p brakes more money is made
more pollution on earth

[oz2cpu]

>Yes I know that I'm sometimes too direct.. that's my way anyhow

no problem, we love you anyways :-)

>more cr..p brakes more money is made
>more pollution on earth

Exactly, I love to score old defective equipment, play with it, repair it, use it, or sell it or trade it, when i am done playing..
win win for all parts

[Kleinstein]

The repair was not that nonsystematic: The initial guess that Q201 was bad was not out of thin air, but based on measurements - though possibly with the error in assuming that the OP-amp input current is small. That is a bit tricky. Q201 is also a relatively easy to replace part.

The next step of suspecting Q202 is also not that bad - it is known to be a part that can fail, and testing outside may be easier than analysing the circuit. Even after neasuring all the votlage it is not so easy to come up with which part is broken - it may work with 1 broken part but 2 broken parts could already have many odd combinations. Especially if one happens to have a somewhat suitable P JFET at hand, swapping Q202 is not such a bad idea.

replacing all the transistors of the protection is a bit odd - most broken BJTs show up as a dead short and are thus easy to find also in circuit. A test to narrow down the fault would be relatively easy. I would prefer it over unsodering all the parts.

The result after the repair is not so clear. The deviations are at least suspect and deserve more tests. Even of the 3 values would be spot on, a little more testing would be a good idea.
The observed difference can have severa reasons, e.g. flux, leaky relays, a leaky Q202 or Q201, a partially bad U201 (chances are it got stressed beyond specs), a damages switch chip, drift of the 34401 internal resistors or just some drift in the reference resistors. Diviations in both directions are a bit strange.

[TaylorD93]

well done, i scored a 34401A which was being scrapped at work with AC and Ohms failures at Calibration, and during self test.

Iirc U201 was the fault about £12 for the IC and it was back up and running, for memory it buffers the 7VDC reference. I maybe mistaken. Such a great feeling of achievement when you can repair such a capable DMM for so little, when the Calibration house said it was not financially viable to repair it, so pushed us to replace it with a 34461A (imo a less reliable DMM)

[floobydust]

To finish the repair, OP would have to dig in a bit. This is supposed to be fun, enjoy the puzzle lol. I'd be so happy to pull a 34401A out of the scrap heap.

I repaired a (different) dumpster dive multimeter that ended up literally 1" away from my own garbage can, my patience running out. It ended up being one FET op-amp input bias current that measured 70nA instead of spec 1pA typ./60pA max., which caused loading and offset. When a semiconductor gets damaged and not totally wiped out due to an overload, it can be very difficult to troubleshoot. One clue I found was the leakage current varied greatly with small temperature changes of a few degrees.

I would say either the new Q202 JFET is far off from the selection criteria, or U201 AD706 is not well. Warm it up a little (non-ESD finger touch) and see if readings move too much.
JFET's will naturally read as ohmic resistance D-S and some transistor testers cannot supply enough -ve voltage to turn off the JFET, and then wrongly report the part as something else.
There must be a sweet spot for the AD706 output voltage with the zener CR201A and Q202 MMBF5461 spec is V_GS(off) +1 to +7.5V, I_DSS -2.0 to -9.0mA pretty wide range off the reel.

[golf32]

Just commenting to say that I am not abandoning this repair. I do agree that I didn't do the best job, but hey, that's how you learn. I haven't had much time to work on it, but I will over the long weekend.

Rationale for replacing all the transistors in the protection circuit was that when bypassed it "worked" (which I now know is not actually in spec), so I thought the problem was somewhere in there. I also saw where I could still buy the parts so I thought, hey, I'll just replace them all instead of tracking down an individual transistor.

I didn't know about transistor testers not being able to test certain JFETS. Definitely good to know.

[Kleinstein]

If the threshold of Q202 is wrong one would expect the circuit to be not possible to turn the fet all the way off or on. So the problem would be with the lowest or higherst current settings to be likely way off. It would rather very unlikely to be just at the edge to be only a little off.
As far as I see the circuit the main point is to have a threshold of less than some 5.5 V to be able to turn off and an IDss of more than some 1-2 mA. So there is quite some tolerance window.
It still looks like they need some selection for the MMBF5461.


The current problem with some 200 ppm off is more like a thing of leakage or an offset.  A test with an increased temperature of some critical parts is a good idea. Leaking semicinductors usually get way worse when warmer. This is not guaranteed, but the normal case.

Another short at cleaning is another idea - when clean an additional cleaning step should not change much. So it would help to check the reading in the 10 M range before and after to get the difference with good resolution.
If possible it would also help to measure all the ohms ranges: so far it looks a bit odd with some values high and some low. Leakage current would mainly effect the 10 M range and an offset should efect all ranges in the same direction.
Another point to check is the input current in the voltage mode and also for the 4 wire ohms sense Hi. Higher than nomal input bias is possible after things like ESD damage or with contamination / dirt.

[golf32]

I've taken some measurements and now I suspect U201. I'm seeing a differential of ~150uV on U201A and ~500uv on U201B. The first is technically within spec if this was the worst AD706 to leave the factory plus long term drift, but the second is obviously not. Neither of the outputs are pinned to a rail so this leads my to believe that the chip has excessive offset.

[Dr. Frank]

I've taken some measurements and now I suspect U201. I'm seeing a differential of ~150uV on U201A and ~500uv on U201B. The first is technically within spec if this was the worst AD706 to leave the factory plus long term drift, but the second is obviously not. Neither of the outputs are pinned to a rail so this leads my to believe that the chip has excessive offset.

I still can only recommend a systematic approach, i.e. measuring these about 10 test points in the schematic, and writing that into the picture which I have linked, or see the unit that I have repaired: https://www.eevblog.com/forum/repair/hp-34401a-error-612-613-615-617-618-619-621/msg3571485/#msg3571485

This way only we together could identify the still existing (U201?) or new (Q202?) failures.
Btw: This AD706 always fails with excessive input currents (it's a low I_bias type), so that afterwards causes high offsets on its inputs.
Your description does not necessarily imply a defective U201.

Frank

[golf32]

It seems the Q201 transistor I used is an adequate replacement. I get a ~5.05 and ~0.505 drop across the 28.57k resistor. I don't think there's an actual specification for that (datasheet says "About").

I assume for that post you linked the first value is pre-repair, and the value after the error is post repair?

[Dr. Frank]

(Attachment Link)

It seems the Q201 transistor I used is an adequate replacement. I get a ~5.05 and ~0.505 drop across the 28.57k resistor. I don't think there's an actual specification for that (datasheet says "About").

I assume for that post you linked the first value is pre-repair, and the value after the error is post repair?

Yes, I think I described that in my text, that the first value is faulty, the second behind the => is correct/repaired.
The voltages are about ok.
Which instrument did you use to measure these? One with 10M or 1GOhm input impedance?
The difference between +/-, pins 5, 6 is 20mV, that seems to be a bit too high, although I measured the same, or a 10MOhm DMM could create an error.

The Gate voltage of Q201 is too high, 6.4V, all others measured around 4.6V or lower.

That might be an indicator, that there is a slightly excessive bias current out of pin5 of U201, which drives an additional current into node pin 1 of Q201 / 28k57.
Is there a big difference of this Gate voltage when you switch through the different, upper ranges?
There's a table in the description of Theory of Operation, p. 98, which of both current sources R201 or R202 is used.
If you apply a short in 2W mode, and switch all ranges from 100Ohm through 1MOhm, there should be no difference in Gate voltage, as R202 only is used.
In 10M and 100M range, this could change, but these both ranges should have the same Gate voltage.
That could indicate an excessive current out of pin 6.

Q202 seems to operate correctly.

How old is your instrument? The AD706 often is going to leak after about 20years.

I would replace U201, if this Gate voltage varies strongly, as described.
Such an effect in this place could explain the varying big errors, including the sign, from 100Ohm over 10k to 10MOhm.
You could measure and calculate the errors between your 34401A and the 5450A for each of the resistance ranges, maybe a systematic trend can be seen, like growing error from 100Ohm to 1M, and negative error for both 10M and 100M ranges.

Frank