HP 3437A System Voltmeter - first half of teardown, pending repair

[c4757p]

HP 3437A System Voltmeter teardown and repair, part one

Still quite a way to go, this is a complex piece of kit and not in the best shape.

Never mind the mostly uninitialized blog, it's brand new.

The analog people here might appreciate this preview:



Edit - added a couple photos I missed.

[echen1024]

Delightful. Some interesting construction quirks though. However, those white ceramics are absolutely beautiful.

[c4757p]

I tracked them down in the schematic (just updated the text to reflect it) - the two in the analog section are actually just precision matched resistor networks! The one in the digital section is an unobtainium microprocessor made by a company nobody's ever heard of, which specialized in MPUs for T&M applications. (I can't remember the name anymore) made by HP themselves, as free_electron pointed out - don't recall where I read that but I don't think it's true.

There are definitely some quirks. The digital board in particular looks like it was laid out rather carelessly - all sorts of pointless kinks and twists in traces, etc. The black jumper wire you can see in the picture of the MPU installed in its socket? There's a perfectly good space on the back of the board to run a trace for it. (Possibly it's there to allow disconnecting for service, but why a long wire with crimped connectors when a two-pin header-style jumper will do?) Parts of it are clearly designed with servicing in mind, but others are totally not - for instance, most of the connectors are directly soldered/crimped. Removing the GPIB port to get the digital board out was interesting.... the ribbon cable is mounted into the IDC connector and soldered directly to the board. I had to work the connector itself out through the back panel cutout (it was that or desolder twenty-something wires).

Of course, if you don't need to fully remove the boards, both of them are mounted in with flexible plastic hinges. Just unscrew on one side and fold out.

And I'm curious about that triax connector on the rear panel. It's directly paralled with a standard coax connector on the front panel, and the guard is undriven, so what's the point? Unless it was a junk box replacement by a clueless person. That is a possibility - there was one screw missing from the chassis, so I don't think I am the first one inside this.

I don't think I'm going to have to dig far into troubleshooting this one. I think it's just some stuck keys. But does anyone have access to a schematic that is actually legible? Agilent's copy is ridiculously smudged, as usual. (How the hell does the company that made it not have access to decent-looking schematics??)

[sync]

That turrets are teflon insulators. High impedance stuff.

[c4757p]

Ah yes, I see that now! Thank you. (That's pretty clever, actually.) I never noticed the ring around the metal, I saw the traces encircling them and thought they were connected to those traces. I guess they are actually guard traces.

By the way - am I right to think that the "crinkly" look to the top of this ROM is due to heat? The underside of the PCB around them is darkened a good bit, too. I'm really hoping there's no thermal damage to this thing, once I get the panel fixed - I don't remember hearing the fan spin up when I turned it on.

[kizzap]

Is the top of the chip feel greasy or sticky, or is it solid?

It looks like it might have some of the spray on it that is on the board.

-kizzap

[c4757p]

It is solid, and there is no spray on the board.

[BravoV]

Looks like the same conformal coating that is applied on the pcb, and those "crinkly" probably trapped bubbles when the chip was heated while the coating was not properly dried ? 

[free_electron]

you bet there is spray on the board. even the ceramic part is 'glazed' by some conformal coating ....
looks like this whole board was coated in something.

that cerqmic chip is an HP design processor. made by the Loveland division in colorado

[sync]

IFIAK that cap between the ROMs at the bottom of the picture is a wet tantalum. Don't replace these type of caps. They are high quality and expensive.

[c4757p]

If it looks that way, it's just the photo. I am 99.99% certain there is no coating on this board. You can clearly see exposed PCB material and metal in places where the soldermask has been pulled back (around vias and testpoints, for instance). There are even a couple places where uncovered copper (the traces on the digital board appear to be raw, unplated copper) has begun to corrode. I'd take a better picture but my camera is just a cell phone and it's really hard to photograph anything shiny...

Also note that a few of the pictures were taken shortly after washing the boards, there may have been water still in nooks and crannies and moisture on some parts that I didn't notice.

BravoV, the "crinkly" look is just 1970s soldermask.

Upon closer inspection, the ceramic chip appears to be permanently shaped that way. If you look very closely, even the screen-printed label follows the contours. (Again, can't manage a decent picture...) I think it rolled off the assembly line like that.

IFIAK that cap between the ROMs at the bottom of the picture is a wet tantalum. Don't replace these type of caps. They are high quality and expensive.

Yup, I don't replace wet tants. I haven't gone through the capacitors yet, so if there is one there I didn't notice it.

[sync]

It has lot of tantalums. Look in the service manual (Table 6-3 Replaceable Parts).
These are the metal ones with the positive lead soldered on them.

[c4757p]

that cerqmic chip is an HP design processor. made by the Loveland division in colorado

Is it? I remember reading that they were made by some defunct chip company. Don't remember where I saw that, though, probably was bullshit.

[lowimpedance]

Nice fingerprint on the ROM next to the crinkly 'finished' one !, testing for heat??.
I have a vague recollection that I have seen this surface finish before, I would be surprised if it was heat damage.
 It was one grubby 'bugga' that's for sure. How people let stuff end up like this browns me off!.
On a somewhat different note I see the Micro is in a 'RED' socket. I recall seeing a service note for the 3456a regarding units which contained these sockets that would cause lock up problems, and should be removed. See the third note in the thread below.
Of course I would just put in a high quality replacement socket.

https://www.eevblog.com/forum/testgear/hp3456a-service-notes-for-owners-info!/msg144820/#msg144820

[c4757p]

Thanks for the note about the socket. Can't read it on my phone though, I'll check back tomorrow.

f_e, you were right (as usual   ) about the coating. It's very, very thin, but definitely there, and definitely a spray: there is a ceramic capacitor touching a crystal, and the interface between them remains uncoated. If it weren't for that, I swear to god the components look perfectly natural and uncoated! They are just slightly duller without their coating...

I was also wrong about the unfinished copper corroding - it's not unfinished, it is gold, and still had some gunk on it.

[Vgkid]

I can't wait to see more of the teardown, along with it working properly.

[c4757p]

It works!

It's got a few segments out on the LED display, and a few discrete LEDs out, so I'll replace them*, along with any aluminum caps. Needs a cal, too. (What? You're trying to tell me that giving a precision sample-and-hold voltmeter a literal bath screws up its calibration?)

Looks like one of the LED drivers might be out as well, one of the annunciators seems to be stuck half-on.

I was right, the fan is completely stopped. It has its own discrete three-phase driver circuit, so probably just a transistor blown in there. They completely omitted any back-EMF diodes... I'm sure in a three-phase configuration they aren't technically required, but when the fan jams because it's caked with filth, those poor driver transistors are just sitting there plucking the motor windings like a harp.

*I'm hoping I can find new 7-seg displays that fit in the front panel cutouts. I'll just stick them on a PCB to match the one that it has. The old one with tiny LED modules glued and bond-wired to a separate PCB is cool for old-shit nostalgia points, but if it doesn't work...

[sync]

It works!

It's got a few segments out on the LED display, and a few discrete LEDs out, so I'll replace them*, along with any aluminum caps. Needs a cal, too. (What? You're trying to tell me that giving a precision sample-and-hold voltmeter a literal bath screws up its calibration?)

Congratulation!

How do you did the bath? These high impedance circuits must be very clean.

[c4757p]

Hot water, detergent and a soft scrub brush. Followed by an alcohol rinse and a second bath.

Hmm... the fan drive circuit is completely dead. Powered, but no sign of oscillation. (And none of the motor windings are open or short.) Ah, screw it, I'll just replace all six transistors. Can't be bothered to probe around for an hour trying to find which 2N2222 or 2N4403 has kicked the bucket...

[sync]

The water and detergent will leave a film which can affect the high impedance circuits. As final step I usually gently scrub and rinse them 2-3 times with pure IPA and then 2-3 rinse cycles only. It's important that the alcohol is pure and leave no residues.

[c4757p]

Yep, of course. As of yesterday, I was only concerned with getting this thing rapidly to the point where I can tell whether or not it's beyond economic repair. Now that I can see that everything works, I have components to replace, so my grubby hands will be all over the boards and I'll be leaving plenty of flux and crap behind. They'll get another, much more proper cleaning after that.

The motor driver works too, now. Just had to change out the 2222s. Even after the bath, that motor was frozen solid... finally poured alcohol directly in it and started twisting it back and forth, and out pours a huge slurry of grime and dust. I want to know where this meter was before I came to have it...

And yes, those ROMs get very hot. No wonder they (and the PCB under them) look scorched. The microprocessor also gets very hot. I would not want to run this for long fanless.

[sync]

Maybe overheating affected the calibration.

[c4757p]

Doubtful. The analog section and ADC are completely encased in heavy metal shielding with little ventilation, so I doubt they respond much to the fan being disabled. I think it's more a combination of "probably last calibrated in the 1980s", "spent a decade in a mud pit", and "scrubbed with a brush that may have pushed trimmers". (I know for a fact that cleaning upset at least one trimmer - I intentionally rotated an open-frame capacitor to clear out any dirt between the rotor and stator.)

[sync]

Yes, of course. No software calibration but trimmers. And I'm currently working on a 3450A which also adjusted by trimmers.

[macboy]

*I'm hoping I can find new 7-seg displays that fit in the front panel cutouts. I'll just stick them on a PCB to match the one that it has. The old one with tiny LED modules glued and bond-wired to a separate PCB is cool for old-shit nostalgia points, but if it doesn't work...

I have several very similar multi-digit 7-segment LED displays at home. Post more details (pictures, dimensions) of the faulty one(s) and I'll see if I have a potential match.

[krivx]

It works!
*I'm hoping I can find new 7-seg displays that fit in the front panel cutouts. I'll just stick them on a PCB to match the one that it has. The old one with tiny LED modules glued and bond-wired to a separate PCB is cool for old-shit nostalgia points, but if it doesn't work...

I think I have some old 7-segs that are marked HP. If you can find a P/N on them I'll check and you can have a few.

edit: haha, I didn't see you already had an offer! Still stands anyway.

[c4757p]

Thanks, guys!

PN is 1990-0589 according to the manual. The only marking on it is 5677708J. Dimensions are attached.

[c4757p]

More pictures and information in a couple days - just got my capacitors and fan

[lowimpedance]

And yes, those ROMs get very hot. No wonder they (and the PCB under them) look scorched. The microprocessor also gets very hot. I would not want to run this for long fanless.

The data sheet for the 'A' revision of the 3624 gives a max. temp of 135 deg 'under bias' , 90nS speed, o/p currents 100mA.
I wonder how fast they are working?. Guess they will get a little toasty. Having another look at the picture of them the one with the odd surface texture still looks like it has had a coating applied rather than any heat damage. do all the ROM's appear to be original?, none replaced?.
Anyway its nice to see its not ready for landfill yet.

[c4757p]

The data sheet for the 'A' revision of the 3624 gives a max. temp of 135 deg 'under bias' , 90nS speed, o/p currents 100mA.
I wonder how fast they are working?

The system clock is 3.333 MHz. Not sure what the memory clock is. Not incredibly fast.

Yeah, like I said I did eventually recognize the coating... It is very thin, but definitely there. Still, the PCB under the ROMs is darkened - pretty sure that's from heat! Anyway, I have the new fan installed now, and everything runs just a bit warm to the touch with the fan blowing over it.

[macboy]

I checked my stock of displays and I don't have anything to match.

[c4757p]

Thanks for looking! After carefully cleaning all the contacts, I've discovered that most of the "bad segments" were actually poor contacts, and there appears to be only one segment dead. It's the "A" segment on a digit which only displays 0 and 1, so now it displays U and 1. I can live with that.

[free_electron]

i think that final coating is some moisture protectant . applied like you would apply hairspray. it's not dipped or painted with a brush but vaporised and then probably dried in a pass-thru oven at 60 degree centigrade to get the solvents out.

[c4757p]

Well, I haven't touched this in quite a while - I needed LEDs to replace the ones it has, and that package got tied up in the mail for a while.

This, my friends, is what happens when you put 2013 LEDs in a device from the 70s. Guess I'll be changing a couple resistors too

Every single thing appears to work. I think the only thing I will have to adjust is that trim cap I messed with while cleaning - compensation for the 10V range attenuator, I believe. Everything else is within specifications still (absolutely astounding considering what it's been through).

By the way - does anybody know where I can source a couple replacement buttons? It doesn't seem to be nearly as easy to find replacement parts for HP equipment as for Tek...

One thing that just occurred to me - if the front panel LEDs were failing, what about the optocouplers between the digital and analog boards? Are the LEDs used in optocouplers are unreliable as the ones used on panels?

[c4757p]

Er... am I reading this right? Are they really giving the LEDs (VCC-VF-VCESAT)/R = (5-2-1)/33 = 60mA? I guess that works out to a lower average, once you consider the multiplexing duty cycle...

[sync]

The LED current will be lower. The annunicator data lines are switched with a transistor to VCC.

[c4757p]

Nope, this is the voltage across one of the 33 ohm resistors. (Looks like it ended up offset, must be I swapped the probes after configuring the scope.) Peak is about 2.2V, so 66 mA, and average, compensating for the accidental probe swap, is about 0.3V, so 9 mA. I guess the average is about what I would have expected. I think I will replace the 33 ohm resistors with 91 or 100 ohm resistors, for an average of about 3mA.

(Yes, I notice my calculation is flawed because that resistor serves multiple LEDs. Oh well, I still want about 1/3 current.)

[c4757p]

Update: I will follow up in more more detail soon. Finally got myself a GPIB interface! So I will try out and demonstrate some of the more useful applications of this thing. Just have to wait for shipping.

Hmm... I was planning on selling it when I finish with it, but now I'm just thinking how useful it would be paired with MATLAB...

[echen1024]

Update: I will follow up in more more detail soon. Finally got myself a GPIB interface! So I will try out and demonstrate some of the more useful applications of this thing. Just have to wait for shipping.

Hmm... I was planning on selling it when I finish with it, but now I'm just thinking how useful it would be paired with MATLAB...

Us on the forum buy test equipment to fix and sell. After we fix it, we fall in love with it and keep it.

[c4757p]

That's how I got almost all of my equipment! I sell most of it, but I do fall in love with the occasional item. That's OK, most of them do eventually end up sold for an upgrade..