HP 3490A rescue (FIXED)

[GregDunn]

I was at our local electronics store picking up some miscellaneous parts - it's about the only one left which doesn't specialize in TVs, computers and knick-knacks.  I was surprised to see that they had a shelf piled with old test gear; several expensive HP analog meters and some stuff which I couldn't identify.  But there in front was a fairly clean looking HP 3490A.  This is the unit which has a built-in logic test suite and is supposed to be fairly accurate within its 5.5-digit range.  There's a nice article about it in the HP Journal for August 1972.

There was no way to plug it in and test it, so I hazarded a price and the owner seemed almost glad to get rid of it.  Though when we looked inside (to make sure it hadn't been stripped) he said it was lucky he didn't realize there was so much gold in it or he probably would have salvaged it a couple of years ago.   

Anyway, home it went.  Powered up OK, and the PSU seems to be providing the correct voltages.  The logic test runs to completion with all the correct displays, except for an added decimal point which appears 2 digits over from the "right" one during parts of the test.  However, every other display in the test suite as well as all the measurement ranges yields the "-12000.2" overload display (max range is 120000 counts).  I still haven't gotten a really good sense of how this beast is put together, but I suspect that since the flow is Signal conditioner -> DC Amp -> ADC -> Logic -> Display that the fault probably lies in the DC Amp or the ADC.  I thought it might be a quick fix when I found a toasted resistor (A2R8) in the DC amp but it turned out to only be the guard ground connection.  That, unfortunately, hints that the DC amp may have taken a jolt.

That's a problem, because the big hybrid, A2U2, is complete unobtainium - in fact, the service manual pretty much says "if it's bad, exchange the board for a new one."     Nevertheless, I'm hoping I can learn a few things about this unit even if it's not repairable without an expensive donor.  There are plenty of other things it could be, but apparently either the DC amp is pinned to one rail or the ADC front end is likewise malfunctioning.  The logic test, if I decipher the block diagrams correctly, does not test the signal path directly, only the digital portion; however, the other tests do attempt to verify the DC amp, ohms converter, and reference voltages.  So I'm leaning toward that deceptively sparse DC amp board which sits between the signal converter and everything else.

Oh well - now that I have all my Fluke DMMs purring along, time for something a little more challenging, right?

[duak]

I'd always wanted an hp 35 calculator, an hp 970A handheld DMM and an hp 3490A.  I remember that the first place I worked professionally had all of these things.  The 3490A was bought specifically for its AC performance to test some custom acquisition systems that needed to have 20 filter channels matched to some insane tolerance.  Back in the 70's the analog way with tweaking was the only way.

Here's some advice on repairing one: https://www.eevblog.com/forum/repair/hp-3490a-fixed/

Kudos to hp providing a manual with enough info to repair this thing.  Looking at the schematic, it should be possible to replicate the hybrid with modern parts.

Best o' luck,

[GregDunn]

I did in fact review that thread before diving in.     Several of the chips made little crunching noises, so they were indeed a bit loose.  But sadly, it made no difference.

No doubt that hybrid can be recreated with new tech, and I'll do that if needed.  I'd just like to keep it as original as possible - not always an option when the goal is to get it running again.

I'll update this thread when I get a chance to look more closely at it.  The queue is getting shorter now, so hopefully it will be soon!

[Stray Electron]

  I've repaired a lot of that old HP gear and I almost always found that the problem was in one of the mechanical connections and not in the electronics.  Take a second, closer look before you condemn that 3490 and don't ASSUME that just because two parts are touching that they're making contact.  Measure the input signals and voltages on the IC itself and measure the output signals on the device that they are going to.

[GregDunn]

Yes, as soon as it gets on the bench that's the plan.  The only things I've done so far have been replacing the toasted resistor, adjusting the PSU and removing/reinserting all the cards; I have no idea about signal continuity or TP voltages on the main circuitry.  I know HP's troubleshooting flowcharts are pretty thorough; hopefully they'll get me to a point where I can see the real issue.

That said, the other three HP devices I've repaired all had power supply issues rather than mechanical ones, so I think it was reasonable to start there on this one.   

[perdrix]

Any time you see 120000.2 displayed on a 3490A you should check *all* the power rails and adjust them into specification according to the manual, and ISTR that you may need a quick "flip through" the calibration procedure.

HtH
Dave

[GregDunn]

All the power supply voltages are within spec, even the ripple values.  I have removed and reseated all cards and socketed ICs.

The display (except when running the #1 logic test) always reads some variant of 120002 (OL), with the decimal point(s) appearing in different places on different ranges or test modes.  Sometimes I get 1 dp in the right place, sometimes 2 or 3 of them.  The logic test gives exactly correct values except for the spurious decimal locations.  That seems to be a problem with the display logic, so I took a look at the display troubleshooting path first.  It's supposed to have a pulse with period 1.6 mS and duration 0.4 mS but the period is 1.2 mS.  This does not change with settings.  The manual suggests replacing the '161 counter but it seems more likely that the counter is being programmed incorrectly since the waveform is consistent and clean.

I guess I'm going to have to start with the basic clock generating section of the logic circuit and work forward from that next.

[GregDunn]

Talked to someone on the groups.io HP forum and got some good tips.  After a bit of poking around with the scope, it was pretty clear that the input hybrid A2U2 was not toasted, but that the buffer amplifier section was hard against the -30VDC rail.  I started measuring the bits around A2U2 and found that A1Q505 measured about 22V between base and emitter.  Fail!  It's a bog standard 2N3904 so I stuffed one in from my parts box and it came alive.     DCV and ACV ranges are close enough to convince me that the input amp, ADC and AC Converter all are functional.

Still have the DP problem, but it doesn't keep me from reading the display, so it can be sorted later.  Same with a bit of ripple on the +30V supply, though that seems to be a bad transistor and not just the PSU capacitors.  Biggest problem remaining is that Ohms mode won't converge on a reading - it autoranges like crazy no matter whether it's shorted or open.  I'll probably look at that next.

So, one step closer.

Edit: how many people read this post and shouted "Did you use 4-wire mode for Ohms?"  Let's see hands.  You are correct; this unit does not have 2-wire mode.  Hooking my kelvin leads up to the meter gave me working ohms measurements.       So, one less thing to fix.

[GregDunn]

OK, let's wrap this up.  The power supply issue seems to have been either a poor test ground or something related to the input buffer being driven to the rail by that bad transistor.  In either case, it looks fine now, so I don't need to play with it any more.  The only remaining issue was the spurious DPs on the display.

Turns out this is the older rev of the display board, so it uses plain old 7404 and 7400 chips instead of the 74LS14 and 74LS00 listed for the newer board.  Still, close enough.  From all indications, it looked like U3 (a 7404) which buffers the range select lines was faulty, and spurious signals were getting to the DP drivers.  Just to be thorough, I put in a socket so I could try something else if a modern LS chip had problems in the circuit.

Nope. It worked first time.  I declare this one fixed, and after a warm-up and thorough re-check of the ACV, DCV and Ω modes I'll move to the next unit in the queue.  Incidentally, it now passes the built-in test 100%.

[highdesert56]

That is so cool.

I wish I had places with stuff like that laying around in my neck of the woods.

[GregDunn]

This guy had a number of ex-pen-sive HP analog voltmeters on the shelf for $200-300, plus some weird custom telecom stuff.  The 3490A was about the only thing he had which wasn't really high or off-the-wall.  He doesn't get much test gear, mostly surplus parts.  I'm glad I went in there that day.

Oh, and after a couple of hours warmup, it measures smack on with my first DC standard.  I'm curious to try the others now...