HP-3440 - Nixie multimeter - Repair or not repair?

[Paulo Peres]

Hi All!
Merry Christmas!

Today I was checking if my old HP-3440 was working and discovered that an old issue with one of the displays is back.
I rescued this equipment from a scrap dump almost 20 years ago with same problem.

This equipment is notorious because it is a Digital Multimeter that does not use integrated circuits.

Internally, it is a frequency counter that uses the read voltage values to change an oscillator frequency. Basically, it's a voltage-frequency converter.
At the time of its conception, the suitable display for the application was the nixie tube, so all circuitry gravitates around it.

Turn out that the counter are made with two transistor flip-flop cells, that produces a BCD code as result of the counting. Each BCD digit is produced thru its own board with a nixie display. There is 4 identical boards here. And each digit must be converted to the decimal code used in the nixies.

As I said, it doesn't use ICs. So its creators conceived a solution that had the touch of a genius of the HP designers of the time: a baquelite box with 8 neon bulbs at one side and a conveniently designed PCB on opposite side. And under the bulbs light beam, a small layer of photo-conductive material is painted on PCB (see photos). The BCD digit shown by the neon bulbs is decoded by the electrical path formed when the photoresistive paint is illuminated.

Wonderful, isn't?

But let's back to the unit on my bench.

Its a equipment manufactured in 1971, full of carbon mass resistors (yes, that cursed ones). The repair I did years ago was done with silver ink, because there was interrupted tracks in the PCB with the photoresists and was effective.

Today I tested and found no broken tracks, so it could be something else. I haven't yet devised a proper test for the entire circuit, but I'm wondering: should I really try to fix this?
I mean, I tested some of that resistors and found up to 7% of resistance drift on some 5% units. If the photoresistive PCB is defective, I have no replacement and would have to redesign it with some kind of modern equivalent component.

So, I'm asking to the colleagues that uses to repair this things: Do you think it worth the effort? The general accuracy of the equipment is fine (the measurement circuits have more stable components), but I'm not sure if I shoud change something in the original design just to make it work again. It is a wonder of its time just because that circuitry.

What do you think?

[factory]

Repair it, early DVMs are getting much harder to find, far too many people destroy our history, many more frequency counters were made than this DVM.

A few of the boards are shared with HP counters of the era, they did replace the optical decoder section with a CMOS BCD IC & 10 transistors, in the 1980s for the 5245L, not found any of these later boards, but I believe they won't fade from one digit to another in the same way.
The decoder neons are a known problem as they age, they can fail to light in darkness. The decoder plate corrosion is a less common problem, I've had this too, I bought a selection of used boards for parts.

The carbon resistors on the counter boards should be a problem, other brands used 10% resistors and 20% ceramic caps instead of the silver micas, for the same type of circuit, more critical areas will use film resistors.

David

[doktor pyta]

++ for the repair and please publish a short summary under this thread.
The multi function unit makes this device useful even in today's workbench or just as a kind of 'living fossil'.

[Andy Chee]

If you have the time, resources, and skill.... then repair it.

If not, then give it to someone who does.

Most definitely do NOT throw it out in e-waste!!!

[Paulo Peres]

If you have the time, resources, and skill.... then repair it.

If not, then give it to someone who does.

Most definitely do NOT throw it out in e-waste!!!

Well... I didn't mean to scrap it.
My main question is how far should I go if I can't find original replacements. The Russian version of the 7442 with some extra circuitry could be the basis for a functional equivalence of the board, but this is an instrument designed *without* ICs.
I don't know. It's like swapping a Spitfire's Merlin for a PT6 turbine.
It could work, but the thing loses its soul.

[doktor pyta]

Don't be so pessimistic.
You have four identical counter/nixie driver modules, so You can swap them, compare the waveforms etc. to narrow the culprit.
From my experience the most common issue with that modules are the germanium transistors, not the custom LDR arrays.

[Stray Electron]

  If you do decide to not repair it, I could use three of the Nixies to replace the three that I'm missing in my HP  5321B Freq Counters.

   IIRC my counter (built in 1969) uses 7442s and one of them is also missing. It looks like someone tried to work on it and made a pretty good mess of it and also left some of the parts out!

[Arts]

Repair it; defects in these units are rarely terminal and are generally easily fixed. They can take some time to track down though.
And even if you do have to make some modifications, meh. Whatever you need to do to keep her running today can always be reversed tomorrow when a better solution becomes available

[factory]

Don't be so pessimistic.
You have four identical counter/nixie driver modules, so You can swap them, compare the waveforms etc. to narrow the culprit.
From my experience the most common issue with that modules are the germanium transistors, not the custom LDR arrays.

This might vary depending where in the world you are & how badly TE is treated, here in the UK they get abandoned in cold damp storage conditions (lofts, sheds, garages), corrosion of those traces on those ceramic phototransistor plates is definitely a problem here. Along with rusty iron wires to the various components (neon lamps included), they break off.

Not seen much problems with the Ge transistors on the low speed boards, from the HP TE in my collection, but the higher speed boards with early Si transistors, I've had a few failures on those.

  If you do decide to not repair it, I could use three of the Nixies to replace the three that I'm missing in my HP  5321B Freq Counters.

   IIRC my counter (built in 1969) uses 7442s and one of them is also missing. It looks like someone tried to work on it and made a pretty good mess of it and also left some of the parts out!

This has almost nothing in common with this DVM, it's a whole generation newer, also the counter ICs are all custom parts designed by HP, there is a HP Journal covering those. Only a few HP counters of this age used standard 74xx series logic in the display counter and the 5321 isn't one of them.

David

[factory]

Well... I didn't mean to scrap it.
My main question is how far should I go if I can't find original replacements. The Russian version of the 7442 with some extra circuitry could be the basis for a functional equivalence of the board, but this is an instrument designed *without* ICs.
I don't know. It's like swapping a Spitfire's Merlin for a PT6 turbine.
It could work, but the thing loses its soul.

This is the evolution of the 5212L-4A latching display counter board (+1224 logic variant);

The original was 5212L-4A, Ge diodes & transistors for counter part, neon optical decoder for the display & storage.



This became 05212-6016 after HP renumbered the boards, older boards used Ge parts, later ones (around mid 1970s) replaced the Ge parts in the counter part with Si parts, Ge 2N404 got replaced by Si MPS404A (still made in surface mount form), a few resistors & capacitor also changed with the Ge replacement.



The last versions 05212-6018 (1979) & 05212-06027 (1980) replaced the optical decoder part with a CMOS 4514 or 14514 decoder/latch IC, 10 transistors, the IC power supply was a Zener based circuit.





David

[jonpaul]

MUSEUM PIECE! Valuable to save.

easy to check the Nixies with a 300V PSU and resistor.

Most switches and PCB edge connectors may have corrosion/oxidation, through cleaning first.

j

[Paulo Peres]

Thank you guys!

I'm convinced. I will repair the DVM.

I'll post the pics here...

[factory]

This was hard to find a picture of, it's not suitable for the 3440A, as this one is for an option the 3440A never had, but to give an idea of the short lived 1979 design, with a extra board added to the older design board, before they redesigned it to get everything onto one board.



David

[Paulo Peres]

Wow...

Very interesting. Is not a direct replacement, but is something to think about.

I found the schematics. Thank you!

[factory]

That diagram is wrong for your boards, there are no pictures of the +1224 BCD logic version on the web.

The diagram in reply #9 is the correct +1224 BCD logic version, see the different outputs used from the decoder IC. The photoresistor plate was made in two versions for the same reasons.

Of course you should figure what parts are the problem first.

David

[Paulo Peres]

That diagram is wrong for your boards, there are no pictures of the +1224 BCD logic version on the web.

The diagram in reply #9 is the correct +1224 BCD logic version, see the different outputs used from the decoder IC. The photoresistor plate was made in two versions for the same reasons.

You are right.

I liked the idea and didn't pay attention to the details.

***

Later today I will do some tests in the instrument. I'll share the results.

[Paulo Peres]

And here we go...
I did some tests and found some things with wrong values (voltages, mainly).
These are some things I saw:

This is how my Fluke 8050A reads the test voltage (11.83V)

And this the HP-3840/3844 version

This is how my Fluke 8050A reads the test voltage (38V)

And this the HP-3840/3844 version

The semiconductors are apparently ok, and when I change the module position the defect follows it. I did some comparative tests too, setting the two more significative displays to 3 (33V) and checking the flip-flops Q outputs, them comparing the results. No significant differences observed.

D flip flip in both boards. The result for C is the same. These waveforms was acquired in cathodes of CR5&CR6 and CR7&CR8. All in 20V/div scale (2V/div signaling is because the last 2 probes don't have the attenuator signaling).

I'll check the rest of circuitry from begin before continue.

[factory]

That haze with no single digit lit, is how mine with faulty photoresistor plates were, I proved this by swapping the plates from one board to another. Be careful if doing this. I avoid removing the tubes unless absolutely necessary as they can crack, they are often in poor condition from pin corrosion (look up oxide jacking) in the UK.

The picture in the first post shows possible corrosion in the circled areas, also the repair could be bridging traces (silver migration?).



David

[coromonadalix]

it is damaged on many places ... not just the red circles

[Paulo Peres]

Well...
Actually I tested the red circles and these points are ok. The blurred track at lower right is the old silver ink. The other track is just scratched, not cut.

I also swaped the photoresist boards between two modules and the problem really is in the pictured board.

Tomorrow I'll check it again with a microscope.

I was playing to the schematics and the logic to narrow the search. I believe I can fix a broken track, even if tiny. But if is the photoresist, I'll need another board.

[Paulo Peres]

Yep!
I checked the board with the microscope (unfortunately its a old model and I have no reliable means to take good pics).

The tracks are a bit ugly, with several scratches but without interruptions.

I have a problem with the photoresist. 

Happy new year, btw.

[Paulo Peres]

While I don't have a solution for the photoresist, I still troubleshooting the DVM.

I found a bad Electrolytic Cap and was able to get a fair calibration, but now I found this:

" alt="" class="bbc_img" />

My plugin is a HP-3444A and I just discovered that part of Ohms scales are dead.

I have the PDF manual that Keysight made available in the site, but its schematicas are low resolution.

Does anybody have one with beter resolution?

[factory]

Try a wanted post for the photoresistor, or a another board.

The thumbnail picture showing the broken wire, I can't seem to make it bigger, it worked fine this morning.
I've got a 3444A here somewhere, I could have a look to see where it connects to.

David

[The Doktor]

I do have one possible suggestion for replacing the photoresist section. I am not at all familiar with this meter, maybe somebody who is can say whether my idea is workable or just complete nonsense, but either way here goes. Is there any chance of using a bunch of CDS cells to replace the photoresist strips? You can buy a bag of 20 of them for about three dollars off of aliexpress. It would probably not be a very pretty repair, but as I doubt the resistance values are supercritical, it may very well work.

[Kleinstein]

 There are some 17 LDRs in the chip and they see quite some voltage (150 V). The available space for the LDRs circuit is limited and especially the smaller LDRs may not like the high voltage.
 Also the neons have a limited size and thus a size limite from that too.
 The easier replcement would be with a decoder chip like in later versions.