34401a bumbling repair attempt

[Swainster]

I recently commented on another thread about a blown up 34401a, and was inspired to relate my own (ongoing) experience. I originally didn't plan on documenting this so the photos start half way through.

A couple of weeks ago I succumbed to temptation and starting cruising evilBay again. I had my eye on a refurbished Keysight MSO, but the multi thousand dollar price kept me from pulling the trigger. However the 'bay doesn't let you off so easy and over the next few days started parading all kinds of old electronics kit before my eyes. Well, I took the bait and decided to buy a "project" 34401a to play with. As it turned out, two of the potential targets on my list turned out to be in my city so a few days later I met the lister outside a locked industrial unit and waited patiently as he revealed an Aladdin's cave of old electronics test and measurement gear, packed in so tight that he had to start dragging stuff out into the corridor to gain entry. A short while later he had half a dozen 34401a DMMs of varying ages and reduced functionality piled up for me to choose from. Suffice it to say, by this point I wasn't thinking straight. All of them had faults of some kind, so lacking better knowledge I went for one of the newer Agilent branded models, which actually booted up for about 1 second then the display went blank (except if you pressed the shift key). But I couldn't leave there without a spares unit right? So some time later I was on my way home with an empty wallet and 2 broken Agilent 34401a units plus a basic bench power supply.

As demonstrated by the seller, the primary unit was would power up like normal, but usually before it finished starting up the display would go blank and everything would stop working. My guess was that this could be a power supply issue and possibly an easy fix. The spares unit was actually slightly newer but couldn't even get past all display segments on, and had a suspicious single followed by double beep. No idea what its condition was but the assumption was that the parts inside would be compatible with the primary unit.

Google brought up a few threads reporting the same or similar symptoms as the first unit, typically after the DMM had been brought out of extended storage. Started to think about depolarised e-caps, however one thread suggested that the problem was related to the ground referenced logic. I'm afraid that I can't shed light on this issue as after poking around inside and measuring expected power supply voltages, the unit just started to boot up normally, and has been working normally every since, without even reporting any errors during the extended self test. Score! (Well, actually I was a bit disappointed as I bought it for the fun of repairing it). This thread will be about the other unit, however attached is a pic of the working one being pressed into service testing IEC60990 fig 4 touch current (just to get an idea if the DUT will pass the test at the test house). For safety eagles, there's a mains isolation transformer just out of shot feeding the test network. It turns out that the VFD display is ideal for photos.

To be continued...

[wn1fju]

I have experienced this kind of thing several times with vintage pieces.  At first, I thought that maybe while I was poking around with a scope probe, I could have pressed the probe point down on a marginal solder joint which then fixed the problem.  But now I think a more likely explanation is that the electrolytic capacitors simply reformed to the point that things were now working.

After these pieces were "fixed," I never had any problems with them again.  But do keep an eye on the caps.

[Swainster]

So on to the main subject, the "spares" model. As a reminder, on power up all the segments light up on the display and a single beep is heard, followed shortly by 2 close beeps. Is there any information on what these beeps mean? I guess I should RTFM

Opening the unit up and damage is revealed - sooty smudges on the PCB around the front terminal. Flipping the board over and it's clear that the Lo/common trace from the front panel to the front/rear selector switch has gone missing, presumably deposited over the PCB and metal cover. A little further in, a small portion of the analogue ground under the custom resistor PGA had also vaporized. However, the fuses were good and there was no sign of heating on other traces, yet current requires a loop - a bit of detective work was called for. A look at the inside of the cover revealed the answer. The maximum common mode voltage must have been significantly exceeded, causing the floating 'ground' to arc over to the cover, perhaps starting from the sharp pins of the PGA. This in turn probably blew the exposed trace on the 'Lo' net (presumably this exposed trace is intended as a kind of fuse to prevent fires?). This moved the arc closer to the front terminal where it burnt it's way all the way back to terminal. It must have been quite exciting for the user (abuser?). The only good thing about this hypothesis is that it doesn't involve arcing through the transformer, the other likely culprit.

I've attached some pics - the flying wires were added by me just to see if I could get anything to work.

edited for clarity

[Swainster]

@wn1fju Definitely will be keeping on eye on it - if only because I want to see if it is stable enough to make it worth sending for calibration.

[Swainster]

Well the burnt traces were discouraging but a big part of the reason I bought the DMMs was for the challenge of making a repair. One of them spontaneously fixed itself so...

First things first, I checked the power supply - immediate score! The -18V rail was out of spec - unfortunately it was running at over -36V i.e. unregulated transformer voltage. This is potentially a very bad thing, as anything on the -18V and -15V rails will have taken some serious abuse. This includes the display, which I decided to tackle first so I could at least confirm if the main controller was running and read out the error codes.

The display logic runs between -18V and analog ground, so the nominal 18V input had been doubled by the fault. On the display board this supply is regulated down to 5V for the logic, but this local 5V rail was down at around 0V, showing a near short to ground with signs of heat damage to an inductor. Desoldering the VFD to allow for further investigation revealed the ugly truth - the 5V regulator (78L05, or something similar), much like the LM337T on the -18V mainboard rail, had failed by shorting input to output, letting through the unregulated supply voltage. This had cooked the inductor, leaving it open circuit, but not before destroying, amongst other things, the display processor. The only thing which may be salvageable is the VFD itself, which at least showed all the segments lit up, though its brightness is suspect given that it was being supplied by about 150% of its nominal voltage. I pretty much decided to write-off the display board at this point, however the main board could be tested by swapping in the display from the working DMM.

Suprisingly the burnt unit booted up with swapped display, and the 3 start-up beeps were reduced to 2 (i.e. the second double beep had become a single beep). Less surprisingly, the DMM reported error 604, which is ADC related, which also meant that it had to skip the rest of the self test, at least for the tests which rely on the ADC value. I decided to get the ADC working so that the rest of the self test could be completed. This is where I made my first big mistake - not properly reading, and more importantly, understanding the service manual and schematic. I should have first checked that the ADC input was sensible  - ideally near zero, as that is what error 604 actually means - it was expecting zero and was getting something else. As it turned out, the ADC integrator op amps actually were cooked, but even after I replaced them (twice) it was still flagging a 604 error. 'scoping things out showed that the integrator output was not ramping as expected, being stuck at the negative rail. I had pretty much pulled out all my hair before noticing that the service manual states that the integrator can get stuck at one rail or the other if the input is out of range. Sure enough, the input to the ADC was at +18V. Incidentally, trying to troubleshoot the newly replaced op-amps introduced me to something I've never seen before - op-amp overload recovery time. Annoyingly this can make a perfectly good op-amp look like it has failed. Anyway, once I'd forced the ADC input to 0 then the integrator recovered and the expected output ramp output around +2.5V was seen. This cleared error 604... and of course led to a long list of other self test errors due to me forcing the DC amplifier output to zero.

I believe this main board is pretty much the latest layout, as unlike the older, working meter, which has a USA made mainboard with no silkscreen and dual JFET DC amplifier input, this board is made is Malaysia, has almost full component labelling in the silkscreen, and has replaced the dual jfet with a bootstrapped OPA130. This is pretty annoying as I cant find the full schematics online, but it looks like someone in China has posted a reverse engineered schematic. As far as I can tell, this reversed schematic is pretty accurate, although doesn't include the correct designators/labels, and the resolution is so low that is it almost illegible. I've added designators to part of it and attached it below in case anyone finds it useful. Can't guarantee that it's correct but it's probably a good place to start.

One of the main functions of the DC amplifier is the decade range selection, achieved by switching the negative feedback resistors. The reason for the saturated input to the ADC turned out to be lack of any feedback at all. The resistors themselves were fine but feedback switching is done by the custom agilent "switch chip", one of the chips which reside across the +18V to -18V rails - it looks like this hadn't survived the +18V to -36V excursion. Sadly these are no longer available from Keysight, at least according to the questionably named official Keysight FAP (find a part) website, so this is where I am currently stuck. I confess to having splurged on an Aliexpress "genuine" replacement, but it will be at least a few weeks before it comes in and I figure that there is only about a 50% chance it will actually work, but I've spent enough energy on this project that I figure it is worth a gamble. Hopefully I will be able to add more to this thread when it arrives...

[razvan784]

I am very sorry for your trouble, I hope you didn't overpay for your DMM.
That switch IC and the custom resistor network are central to the 34401A performance. Let's hope your replacement is genuine. Otherwise, you might be able to get some functionality back by replacing the MC, MZ and PRE switches with a standard analog switch IC.
Please don't throw away the damaged switch IC, there are people who'd like to see what's inside.

[Swainster]

@razvan784 I guess I did kind of overpay a little given the condition of the meters (both were non-working at time of purchase), but as I mainly bought them to play around with, and maybe learn something from, in that respect they have been very worthwhile. I started out with limited experience of analogue electronics as my background is in firmware, but now I almost feel like could design my own multimeter 
In fact, now I'm starting to think that I need to get a 7.5 digit multimeter just to confirm the performance of this multimeter once it is rebuilt (assuming I manage to get it working again). It's a slippery slope...

The standard analog switch idea is good - replacing the functionality of the custom switch chip would be a worthy project in its own right!

[Swainster]

Ok, time for an update!
While I was waiting for the "new" switch chip to come in from Aliexpress I continued to troubleshoot the remaining circuitry. By jumpering over the pads of the defunct switch chip, I figured that I could feed in a signal from the input jack and scope out the output where it feeds into the ADC. Well, that didn't go well and another burnt out trace was identified, this time on top of the board (but hidden under a MOV). It also confirmed that fact that most of the Op-amps had fried during the voltage excursion, however by this stage I had got fed up of waiting for specific chips to come in and I had basically splurged on replacements for every op amp in the meter, plus almost any other part that was connected to the -18V rail. This included the RMS to DC converter, which brought particularly sharp tears to my wallet. I had to make a few subsitutions, mainly for the AD706 which is out of stock at the usual places. However, in the end I didn't even use these latter 2 parts as the 2 of the 3 AD706s run off the 0 ->15V or 0->18V rails, so escaped the -18V disaster, and the 3rd one (-10V ref) seemed to still be working.

Anyway, after repairing the trace and swapping out the various ICs, the DC amp now seemed to be working, though running surprisingly hot. Just in case, I changed the jelly bean transistors which make up part of the input, which seemed to cool things down, but I may take a look at this again at some point.

[Swainster]

And so on to the AC circuit. I didn't really spend much time investigating this as it's a function that I rarely use. I just swapped out the dead amps and the suspect analog multiplexers. What could go wrong?...

Well, it turned out that the 2 separate -15V supplies for this section had failed (they were derived from the -18V by a couple of 3.3V zeners, which had died). And without these -15V sub-rails then the newly replaced analog multiplexers tried to self destruct. This was detected by the time honoured method of burning a probing finger, so these 2 new (and somewhat pricey) parts are now back on the susect list. I didn't have replacements so they were left alone for the timebeing (and seem to behave themselves now that the -15V supplies had been restored. Note to self... always always check the power supply first.

Substitutions (that I can recall):
Replaced the DG411DY with a DG411HSDY as it was the only version I could get at short notice which could handle the supply voltage
Replaced Analog Devices AD825 with TI OP828 for a similar reason
Replaced Analog devices DAC with the Maxim equivalent as it was a couple of dollars cheaper from the particular supplier that I happened to be using.

Didn't really take many photos of this stage as not much to show
Edited for part number typo

[Swainster]

Eventually the switch chip came in from China, and initial inspection looked good. The new part looked unused and had a 2010 date code, vs the original 2004 part. The label was laser etched rather than silkscreened, but this was a plausible change. All in all, I was feeling quite hopeful. I considered fitting a PLCC socket but there wasn't quite enough space, so the new part was soldered in place. I have to admit I find PLCCs one of the worst packages to hand solder - give me a TSOP any day of the week.

By this stage I had reassembled my working 34401a so didn't have a working display handy, however a quick power up resulted in only 2 beeps, instead of 3, indicating that the ADC power on self test had passed (the other beep was for the missing display).

[Swainster]

And here I have to admit I took a shortcut. While I have ordered parts to repair the original LCD (just waiting on an Atmel flash microcontroller to replace the original Intel masked part), I ran out of patience and ordered a brand new LCD module from Keysight. They still have a few left but if you want one you should probably hurry up as they will end support for this model early next year. This arrived this morning (along with a new rear panel as neither of my 34401As had one). It was final assembly time!

And the result is...

Pass!

All self test passed and a quick check of the ranges give comparable values to the working 34401a. It will need adjusting to take into account the new voltage reference (as the original had walked off before I got the meter), but overall, the repair looks to be successful. DC voltages were tested up to +-100V, AC was tested up to 20V and 1MHz (though definitely some difference between the 2 meters at higher frequencies). Will test some resistances and currents tomorrow but so far so good.

Remaining tasks:
More testing!
Let the new reference burn in for a few weeks or more, then adjust.
Make a 3D model of the rear panel so I can print a new panel for the other meter.

Note:
It doesn't really show up in the photos but the new display has a more blueish tinge compared to the older display. You wouldn't notice unless they are right next to each other.

[m3vuv]

Well done that man,nice write up Bravo!!!