Fluke 8100A Nixie Tube DMM

[WaveyDipole]

I have recently acquired a Fluke 8100A Nixie DMM. It seems a very nice unit but has some problems.

When I received it, there was a bit of a rattle inside the case. On opening it up, a number of things became evident:

- the original power socket had been removed and replaced with a connector block to which was connected a rather long flex
- the filament bulb had also been replaced by a much bigger 24V (automotive?) one which was attached by thick wires but floating loosely inside the case
- the meter came with the battery option (including all 18 dead NiMH cells). Two of the 6-cell holders had become detached and were floating loose inside the inner cover (screen)

Fortunately, there were no obvious signs of internal damage. The batteries and holders were removed. The unit was then powered up. All the Nixies lit up and displayed a minus with a row of zeroes. However, is no response to any DC volts input or touching the probes together on the resistance range. After a while when the DMM had warmed up, the three right-most Nixies started to flash random digits in quick very succession. There was still no response to any input.

The first step was to check the power supply rails. Obviously the HT supply to the Nixies is present otherwise they would not light up. The +5V supply to the logic circuits measures around 5.3V. Maybe a little high but so far so good.

I then checked the +7, +15 and -18 supplies and got a reading of around +3V, +9V and -25V when powered from mains, and +3V, +9V and -14.5V when powered from a 21.5V DC supply (based on 18 x 1.2V cells). I then spent some time probing these circuits but so far I have not found any faults. Eventually I lifted one leg of R1 and removed the fuse to disconnect the emitter of Q1 in order to test across E and C for leakage. The transistor tested as it was supposed to on the diode test, but I got a reading of around 6k in both directions on the resistance test. Thinking that I had found the problem, I removed the transistor and tested it again. Out of the circuit it tested perfectly OK with no less than several meg-ohms in one direction only across EC, OL the other way around. I then re-checked at the PCB and found that same resistance of 6k across EC, 3k between EB amnd CB. Since the fuse and R1 are disconnected and the transistor removed, how can there be a resistance between EC? I pressed the probes on a couple of random spots on the bare PCB surface and also got a resistance although a much higher one in the order of 10s to 100s of kohms.

Why is the PCB material resistive? Might this have a bearing on the problem?

What is the purpose of the filament bulb? Can it be replaced with something else? Its not particularly safe "flapping about in the breeze" and I could wrap some insulation tape around it at least, but it does look rather untidy.

The smoothing capacitors are always suspects, but I see no obvious signs of problems - no electrolyte leaks and no apparent shorts and although testing in circuit cannot be relied upon, capacitance approximately what it should be. C3 seemed more at variance than the others, but snipping one of its leads made no difference. Replacements for the caps and a spare RCA 40250 (just in case!) have been ordered.

I have attached the PSU circuit. It would be nice to get this early Fluke bench DMM going again if possible.

[Paceguy]

I own a 8100A line operated unit. I have never worked on a 8100A with the battery option. I do know that the 8000A's and 8600A's that have the battery option will not function properly with the batteries removed. The batteries take the place of the power supply's filter caps. The bulb is part of the charging circuit. You will have to make some modifications to the unit's power supply if you do not intend to keep it as a battery option unit. Compare the power supplies of the line and battery option units. You will have to add some filter cap(s)and a regulator, etc and maybe cut a few of the pcb traces and add a jumper or two. Hopefuuly Fluke used the same circuit board for the two types and you will just have to add the componants to make it a line operated unit. To test it further use a bench power supply if you have one to substitute for the batteries to see if it now works properly and if the voltage rails measure as they should. There are some Youtube videos on power supply modifications for the 8600A and there are also some here on EEVBlog. I wish that I could be of some more help than that.

[Paceguy]

Since you are in the UK, make sure two that the switch for the mains voltage is in the right position for 220VAC. Cheers!

[WaveyDipole]

Thank you for your posts. I had been thinking about the battery unit. The battery holders were glued to the inside of the top of the metal screen that covers the main board, so obth holders and cells and had hung there, upside down, over the PCB. Although it is less likely with NiMH cells, none of them seem to have leaked. Nevertheless, I don't care much for that arrangement. If it could be replaced with a LiON solution, that would be better, but it would have to be secured a bit better than at present. Looking at the diagram on page 6-6 of the manual where it describes the installation of the battery option, it shows a battery pack holder secured by 6 screws which is missing. For now I have just removed the holders and disconnected the wiring from the PCB.

I also see a description of the "ballast lamp that provides constant current charge even though line and battery levels may vary". I should have spotted that earlier.

The installation involves installing the meter, the battery holder, a couple of resistors and three diodes as well as the lamp. The battery holder wires are then plugged into the PCB using the two provided pins. I did wonder about the absence of any filtering caps with the exception of the smoothing cap C1. There seems to be only one version of the circuit diagram provided, with the battery option identified by a box with a dashed line. Removing the battery option would mean simply lifting on end of the 3 diodes and maybe removing the lamp.

I had wondered what purpose connecting the two transformer taps to CR32 and CR33 serves? Does this have something to do with regulation?

I temporarily re-instated Q1, reconnected the 470Ω resistor R1 and the fuse and tried connecting a 21.6 supply to the battery terminals to see whether it would make a difference to the voltage rails but it didn't. Unfortunately I am now seeing another problem. Only 1 Nixie is lit and the 5V supply is now at 3V.... It seems like something is slowly failing - I just hope it isn't the transformer. The 470Ω resistor gets quite hot. Oh, and BTW, it measures 560Ω so is our of spec anyway and needs to be replaced.

Incidentally, what is the transistor+zenner part marked DH8041, Q11 on the schematic? I am unable to find any information about this part.

[TimFox]

When dealing with other equipment that contained NiCd batteries charged from a line-operated supply, where taps on the string of NiCd cells provided working voltage to the circuits, I made some appropriate TL431 shunt regulator circuits to produce the voltage differences between the taps, supplied from the charging circuit, to allow operation from the AC line without batteries installed.

[factory]

Incidentally, what is the transistor+zenner part marked DH8041, Q11 on the schematic? I am unable to find any information about this part.

Given it's part of the rail marked +7VR, with a cal adjustment, it could be the voltage reference.

The failing display could be a result of a failing capacitor on the 200V rail, loading the rail down, the capacitor will be warm if it's failed to reform, had this with one or two items of vintage TE.

David

[tggzzz]

I have recently acquired a Fluke 8100A Nixie DMM. It seems a very nice unit but has some problems.

When I received it, there was a bit of a rattle inside the case. On opening it up, a number of things became evident:

- the original power socket had been removed and replaced with a connector block to which was connected a rather long flex
- the filament bulb had also been replaced by a much bigger 24V (automotive?) one which was attached by thick wires but floating loosely inside the case
- the meter came with the battery option (including all 18 dead NiMH cells). Two of the 6-cell holders had become detached and were floating loose inside the inner cover (screen)

They would be NiCds, not NiMH. The different chemistries have different charging characteristics.

What is the purpose of the filament bulb? Can it be replaced with something else? Its not particularly safe "flapping about in the breeze" and I could wrap some insulation tape around it at least, but it does look rather untidy.

They protect the other components when the battery has failed and become short circuit. In normal operation the lamp is on very dimly, but is a lighthouse with a shorted battery.

See https://www.eevblog.com/forum/testgear/fluke-8125a-military-dmm-teardown/msg2540730/#msg2540730 for a teardown of a similar military variant, including pictures.

[WaveyDipole]

When dealing with other equipment that contained NiCd batteries charged from a line-operated supply, where taps on the string of NiCd cells provided working voltage to the circuits, I made some appropriate TL431 shunt regulator circuits to produce the voltage differences between the taps, supplied from the charging circuit, to allow operation from the AC line without batteries installed.

Tim, thank you for that idea.

Incidentally, what is the transistor+zenner part marked DH8041, Q11 on the schematic? I am unable to find any information about this part.

Given it's part of the rail marked +7VR, with a cal adjustment, it could be the voltage reference.

The failing display could be a result of a failing capacitor on the 200V rail, loading the rail down, the capacitor will be warm if it's failed to reform, had this with one or two items of vintage TE.

David

David, that makes sense and I should have realised that myself. I have seen voltage references in a ceramic package (LM299, LM399) but didn't immediately recognize this plain metal TO5/TO39? can component as one. The 7.0V should have given it away.

I have also been so focussed on the LT supply, I forgot about the HT one! The meter had a chance to cool down while I was having dinner. When powered it on again afterwards, all 4 Nixies lit up again, but started fading after two or three minutes up time. Obviously it is a power issue although I didn't notice any caps going warm/hot. This is not to say, however, that one of them does not have some kind of problem that manifests itself after a bit of warm-up time. Since the HT supply is derived from  the LT supply, it makes sense that something dragging down the HT supply might also drag down the LT supply. Hopefully I will have time to have a further look over the weekend.

I have recently acquired a Fluke 8100A Nixie DMM. It seems a very nice unit but has some problems.

When I received it, there was a bit of a rattle inside the case. On opening it up, a number of things became evident:

- the original power socket had been removed and replaced with a connector block to which was connected a rather long flex
- the filament bulb had also been replaced by a much bigger 24V (automotive?) one which was attached by thick wires but floating loosely inside the case
- the meter came with the battery option (including all 18 dead NiMH cells). Two of the 6-cell holders had become detached and were floating loose inside the inner cover (screen)

They would be NiCds, not NiMH. The different chemistries have different charging characteristics.

Agreed that the original batteries in the day would have been NiCD. The ones currently installed are marked NiMH however. They are industrial plain clad cells undoubtedly will have had been installed as replacements at some point. The original Fluke battery holder is missing and the 3 x glued in 6-way holders are evidently a DIY effort. However, your point is valid. NiMH and NiCD have different chemistry and characteristics.

What is the purpose of the filament bulb? Can it be replaced with something else? Its not particularly safe "flapping about in the breeze" and I could wrap some insulation tape around it at least, but it does look rather untidy.

They protect the other components when the battery has failed and become short circuit. In normal operation the lamp is on very dimly, but is a lighthouse with a shorted battery.

See https://www.eevblog.com/forum/testgear/fluke-8125a-military-dmm-teardown/msg2540730/#msg2540730 for a teardown of a similar military variant, including pictures.

Thank you for that explanation and link. I had a quick look and intend to have a more detailed read later. Looking at the pictures in that thread, the internals of that 8125A are very similar to my Fluke 8100A. The layout of the controls and the PCB are near identical, but the military 8125A has some additions including the daughterboard with the three pots and the addition of some fuses, power connectors and lights at the bottom of the front panel together with the WW resistors strapped on to the internal underside PCB cover screen and, of course, the rugged field case. I have attached some photos of my 8100A for comparison. You can see the replaced bulb and the cobbled mains lead wiring.

EDIT: Having fully read that thread this morning, it does contain mentions comparing to the 8100A and the 8120A.

[Paceguy]

Don't waste your time looking further than the power supply first. You must resolve the power supply rails before you go any further. Only once the power supply is working properly will you be able to properly troubleshoot the rest of the circuitry.

[WaveyDipole]

I tried several times to post a further update but kept getting:

Database Error
Please try again. If you come back to this error screen, report the error to an administrator.

It seems that removing Paceguys quote solved that...
I added it in again afterwards and it then posted OK, so not sure what gives with that?

Anyway, here is what I was trying to post:

Don't waste your time looking further than the power supply first. You must resolve the power supply rails before you go any further. Only once the power supply is working properly will you be able to properly troubleshoot the rest of the circuitry.

Agreed. Good advice. I am going to have to wait a few days for some of the caps to arrive, but in the meantime C6 (2µF)  at 150V has been replaced with a 4.7µF  at 250V that I had to hand. I also found a 3µF at 250V cap next to it which I don't see on the diagram and replaced it with another 4.7µF but at 450V. I also took further voltage readings and all the rails were still and sinking over a few seconds after power up. While mains powered, I checked the voltage at the fuse, F2, and found it to be -39V which is way higher than the expected -18V output from Q1, but the output at C3 was actually at just -15V. The 17V supply at the emitter of Q6 is at about 20V. I started to wonder whether there is a problem with Q1, or else the DH80417 voltage reference.

It is possible to decouple the 200V/5V supply (Q4/Q6 etc) from the remaining supply rails by means of a wire link. Doing so allowed independent testing of the 200V/5V supplies. With the link cut and and 18V supply applied from the bench PSU (taking care of polarity), the current limiter was slowly turned up and the current settled at around 250mA. At this point both the 200V and 5V lines sagged to around 140V and 2.9V respectively. Since the 7441 driver chip felt quite warm, it was removed and the circuit re-tested. Tis time, the current draw was around 140mA on power up and settled at just over 100mA after a few seconds - a difference of 150mA. The 200V line was now at around 195V and the 5V line at 5.03V. According to the datasheet, the typical supply current for the 7441 should be around 21mA, topping out at 36mA. It seemed that the driver chip is faulty and drawing excessive current, which was dragging down the other supply rails.

However, this was not the full picture. After reconnecting the link with the chip removed and powering the circuit via the battery pins from the bench PSU set at 21.60V, the -18V and +5V rails were now showing the correct voltage. The 200V rail was still slightly under. However, the 7V and 15V rails were still both low at 3.8V and 5.6V respectively, so some there is still some other problem to be discovered in the voltage reference part of the circuit.

Powering from mains sends the -18V rail to something like -25V and produces significant over-voltage on all rails, so modification to add regulation (and filtering) will be required as previously mentioned by Paceguy and TimFox. For test purposes, I will have to use the bench PSU for a supply for now.
 
Incidentally, I have seen on another thread a discussion regarding the 7441 vs the 74141 and that the latter is preferred. Would the circuit need to be modded for the 74141 or is it a drop in replacement?

[factory]

Incidentally, I have seen on another thread a discussion regarding the 7441 vs the 74141 and that the latter is preferred. Would the circuit need to be modded for the 74141 or is it a drop in replacement?

The 74141 has output blanking for BCD 10 to 15, the 7441 doesn't, that is the difference.
Does the Fluke have blanking for unused digits, instead of displaying all the zeros with no input?

I find old 7441 & 74141 ICs quite unreliable, until recently new production 155ID1 (Soviet 74141 equivalent) were available in bulk directly from the country they are made in, some went in my 5340A counter (which does blanking using separate transistors) to replace defective 7441 ICs.

I read here that older Soviet era 155ID1 can be either 7441 or 74141. http://www.tube-tester.com/sites/nixie/74141-NDT/74141-NDT.htm

David

[WaveyDipole]

Incidentally, I have seen on another thread a discussion regarding the 7441 vs the 74141 and that the latter is preferred. Would the circuit need to be modded for the 74141 or is it a drop in replacement?

The 74141 has output blanking for BCD 10 to 15, the 7441 doesn't, that is the difference.
Does the Fluke have blanking for unused digits, instead of displaying all the zeros with no input?

All Nixies show zeros when there is no input.

I find old 7441 & 74141 ICs quite unreliable, until recently new production 155ID1 (Soviet 74141 equivalent) were available in bulk directly from the country they are made in, some went in my 5340A counter (which does blanking using separate transistors) to replace defective 7441 ICs.

I read here that older Soviet era 155ID1 can be either 7441 or 74141. http://www.tube-tester.com/sites/nixie/74141-NDT/74141-NDT.htm

I have seen those on eBay. Thank you for the link to the info. That is most helpful. Might actually order one of those.

In the meantime, some further progress has been made on the PSU. I powered up this morning to take some readings and the voltages were all over the place again so I broke the link and determined that the Nixie HV and logic 5V rails were still OK. The rails were being pulled down by something in the remainder of the circuit. That something turned out to be Q6 which had a faulty BE junction. It read OL one way, and around 1.5V the other way, which of course should be about 0.7V for both junction. I also snipped C5 and checked the capacitance, which read way above 50μF, so not shorted, but definitely leaky. I replaced the 2N3906 and found a capacitor to use temporarily until my replacements arrive. The good news is that the voltage ref settled at exactly 7.000V and the 15V rail read around 14.8V. The -18V rail read exactly -18V even when I increased "battery" voltage from 18V to 21.60V. I resoldered the link to reconnect the remaining rails and they are still correct as well. I left it on for a few minutes and it does seem stable.

BTW, I noticed that the voltage reference takes maybe 2-3 minutes to settle from around 9.5V to 7.000, but once it reaches that, it stays stable. I presume this is normal?

The next step, I think will be to order a replacement driver IC and wait for the caps to arrive.

[Kleinstein]

BTW, I noticed that the voltage reference takes maybe 2-3 minutes to settle from around 9.5V to 7.000, but one it reaches that, it stays stable. I presume this is normal?

The next step, I think will be to order a replacement driver IC and wait for the caps to arrive.

It is highly unusual for the reference / supply to have that much initial drift and especially a significant higher voltage to start with. This is more indicating a problem, like a bad capacitor so that the circuit oscillates or many be bad resistor / cold solder joint.  Normal settling for the reference would be more about a few 10 mV or less.

[WaveyDipole]

Thanks Kleinstein. I did think this rather strange as the LM299 does not do this, but I am not familiar with this particular voltage ref. However, what you say makes sense. I have finally got around to having a look and cannot find anything obviously wrong. Snipping C8 (50µF) to take it out of circuit has no effect. All of the resistors seem OK. R22 (1.242k) may be a little high, but I need to lift one end to be sure. I figured that since one bypass (Q6) developed a problem with one of its junctions, that the other bypass (Q9) might also have an intermittent problem that is not showing in circuit so I swapped it out, but the problem is still there.

I found that the actual circuit differs from the original drawing. There is a 1k resistor and 22nF capacitor inserted in series with Q9c and Q7e/Q10e. I presume this has been added as a filter? I added these to the simulation I had already drawn up, but they make no real difference to the output voltage which remains stable at the target voltage (+7V nominal, +6.995 in my simulation). I did for a while think that the light bulb filament warming up might be having an effect, but then wouldn't that affect the other rails also? Yet they are not affected. Nevertheless, to eliminate this possibility, I removed the bulb but it made no difference.

That still leaves Q10 and the reference itself, or a dry joint/broken track which is being very difficult to spot.

[Kleinstein]

I don'e see a way for the reference to give a much higher voltage initially. A cold solder joint or similar may be possible. However the parallel resistors are on the low side of the divider - so a cold jolder joint there would lead to a lower voltage.  Some old resistors may also fail, though I would not expect carbon resistors in the critical part.

[WaveyDipole]

Agreed. R13, R14, R15, R20, R21 and R22 are all grey bodied and have their values printed on them so are probably specialised high stability parts. The remainder,  including the two 3.3K and the 100k are standard carbon type with 5% tolerance, but neither seem to be in a critical position.

I have also now replaced Q7 (NPN, 2N3904) as well but once again, no difference.

This is beginning to look more and more like a dodgy DH80417A reference, although I am not necessarily ruling out a cold joint just yet. I decided to re-flow every joint in that part of the circuit and for a while after doing that and cleaning up the Flux I got a reading of 7.01V on that rail on power up. However, coming back to it 15 minutes later, it was back at 8.92V again. Possibly heating the PCB while re-flowing various joints did something thermally temporarily and it has reverted to the fault condition on cooling off? Its a weird one.

BTW, I measured the voltage across the Zenner, which I should have done earlier I suppose, and its at 6.061V at startup, rising to around 6.067 after a couple of minutes, but it does seem to stay at around that value.

If I do consider replacing the reference, even temporarily, then the first question is, is it safe to de-solder, or with that harm it?

Secondly what would it be replaced with? Neither this part, nor the similar SZA263 or LTFLU are available to purchase anywhere. I couldn't even find a datasheet. I figured that I could temporarily use an NPN transistor and 6.2V (according to data on Radiomuseum) zener diode as I did in the SPICE model, but I expect that it will need to be replaced with a proper reference oftherwise precision will be lost. Curiously, I had to use a 5.6V zenner in the SPICE model to get close to 7.0V.

I have seen mentions of the LTZ1000 as well and circuit diagrams. I would have to devise a suitable PCB so this would be a more complex undertaking.

I would also like to ask whether the reference is also responsible for regulating the voltage of the +15V rail, or is the connection via R15 to supply power? I have noticed that when the +7 rail goes up to almost 9V, the voltage on the 15V rail also rises, but only by a few 10s of milliamps, from around 14.80ish to 15.15ish, but certainly not by almost 2V.

[Kleinstein]

The 3.3 K resistors should be non critical they are more for protection on transisents and I don't see even a 50% change making a difference. The 100 K (R12) may be a little more relevant, mainly for the +15- One still exchange the 3.3 K resistors - just in case, especially R19, as a way to high resistance there could cause a higher voltage for the 7 Ref.

The 15 V also uses the same zener, but not the transistor instide the ref. chip. So the 2 are somewhat linked.

A possible way to get the higher voltage would be part of the regulator oscillating, e.g. depending the ESR of one of the electrolytic capacitors.


I don't know a direct replacement for th ref. chip.
Many of the more normal zener diodes have quite some noise and they would not be a real option. A suiteable single low noise zener to work with a transistor may be obtained from China (2DW232 using 1 diode). Some 5.6 to 5.8 V for the zener may be the correct value.
The zener part of the reference at least seems to be OK - so one could worst case still used the zener of the old ref chip and only use a replacement for the transistor part.

[WaveyDipole]

In the attached photo, what would the expected value of the resistor on the RHA marked '1242' be? I measured it at 12.52k when disconnected from the circuit, which would mean its value is 100Ω higher than it should be, but I see no decimal point? This is R22, one of the resistors matched to the voltage reference, the other being R15 which is marked 73.2 and measures 73.3, which is very close to what it should be.

Otherwise I checked every resistor in the circuit. The two 3.3k resistors (R17, R19), which were both over 3.5k, were replaced. The 100k was also replaced as it was high at 117k so also out of spec (5%). In addition, Q9 and Q10 have been replaced along with C8. Apart from R22 in the above question, all 1% seem to be in spec.

At this point I thought that pretty much leaves just the reference. I checked for oscillation, but there was no evidence of any on the scope.

Finally, I realised that I needed to check R181, the 3.3k across the output of the +7V rail and that appears to have nailed it! I de-soldered it at the joint to the +7V rail and checked its value - around 3.6k. Its not critical, so I soldered it back. The rail now read +7.01V - which was not unusual as this happened several times previously after I had just been soldering on the PCB. I powered off, went away and came back a few minutes later, powered on expecting the usual rise to around 8.92V but this time it was still at 7.01V! I repeated the test to make sure and so far so good. Conclusion: a cold joint at the R181 junction with the +7VR rail. It seems that your hunch, Kleinstein, was correct. Since this is a reference rail and it feeds several high impedance points in the 8100A circuit, I wondered whether perhaps the 3.3k across the output provides the minimum loading needed for proper regulation and stabilising of the rail, in which case, a cold joint here would have interrupted the load causing the rail to float high? I might just replace that 3.3k as I still have one left. Next time I should maybe start troubleshooting from the opposite end of the circuit....

The +15V rail is slightly low at +14.88V, which has been consistent throughout and is probably not too critical, but will be re-checked once the Electrolytic cap has been replaced.

In the meantime I have also received the replacement Nixie driver chips. I ordered both the DM7441 and the SN74141 so that I could compare them. I decided to try both of them to see the difference, but because all I am getting is just random flashing digits, no difference was evident. Current draw with the chip inserted was between 155mA and 160mA, which is less than 20mA increase when compared with the chip removed, which is in line with what should be expected according to the datasheet.

So then next step, while I wait for the replacement Electrolytic caps, will be to determine why I am getting just random digits on the display and no response to any input.

UPDATE: I repeated the test again some 30 minutes later. This time the rail started at over 8V again, but rapidly fell to 7.02V, taking around 5 seconds to do so rather than the previous 2-3 minutes. Its definitely better but still appears to need a little settling time. I have now replaced that 3.3k and will re-test again in 30 mins or so.

UPDATE2: Have repeated the test an hour later and this time it took around 10 seconds to drop from around 8.5V to 7.02V, but again, not two or three minutes. Replacing the resistor seems to have made little difference. Not sure whether this can be called a fix yet, but it seems to be an improvement.

UPDATE3: another test this morning confirms that the problem is NOT solved. We are back to the same 2-3 minute settling time so it seems I may have spoken too soon!

[WaveyDipole]

Having seemingly ruled everything out in the +7V circuit, I figured it was time to investigate the DH80417B. Since the +15V rail appears relatively stable at around +14.8V,, I had the idea to  measure the voltage difference across R14, across R15 and across BC of the reference. Theoretically I was expecting these to be pretty much static, but they showed the same drift. Eventually I de-soldered the reference from the board and tested both the transistor and the zener. The transistor was tested on the diode setting with a DMM and tested as one might expect for a good silicon transistor. The zener was tested on a transistor/diode/zener tester which showed the zener breakdown voltage to be around 6.2V which is consistent with the information available online.

I then temporarily replaced the reference with a 2N3904 and a 5.6V zener as I had done in the SPICE model, but on power-up, the output started at around 13.5V and then started sinking slowly as it had done previously with the original voltage reference in place. The model predicted an output of around 6.59V. Unfortunately there does not appear to be a link provided to disconnect the reference output from the rest of the circuit, so the only way I can test whether the problem is coming a voltage leak somewhere further along the line is to cut the track, which I am not too keen on doing. The other option, I suppose, it to experiment with the SPICE model and see whether the problem can be reproduced through breaking links in the circuit.

If anything, this exercise appears to indicate that the reference is not the problem, but I am puzzled by why the voltage drop across R14, R15 and REFbc were showing the same downward drift over time as the +7V rail does.

[Vgkid]

Since the 15v is stable , is it possible to isolate the 7v reference circuit from the rest of the board?

[Kleinstein]

The transistors Q9+Q10 form a Sizlaky pair and this circuit part is possible to oscillate sometimes at rather high frequencies.
This can especially be an issue with more modern, fastger  transistors and the relatively high impedance at the base of Q10. Chances are the added R25 + C10 may be because of this, but not sure if they are enough.  Is R15 really only 4.5 K  - in most applications the collector current is much smaller and thus an expeced R15 of more like 100 K.

[WaveyDipole]

Since the 15v is stable , is it possible to isolate the 7v reference circuit from the rest of the board?

Had the same thought. I have been hesitant until now because because it required cutting tracks, however, it does seem to be a necessary next step. Since the reference tests perfectly fine out of circuit, I first put this back on to the PCB and then found a place to discretely cut the track, which I was able to do on the underside. The initial result at power up indicated 7.001V on the +7VR rail and +8.49V on the severed upstream end. I will recheck in a few minutes to be sure, but initially it would appear to indicate a leak somewhere further along in the circuit.

The transistors Q9+Q10 form a Sizlaky pair and this circuit part is possible to oscillate sometimes at rather high frequencies.
This can especially be an issue with more modern, fastger  transistors and the relatively high impedance at the base of Q10. Chances are the added R25 + C10 may be because of this, but not sure if they are enough.  Is R15 really only 4.5 K  - in most applications the collector current is much smaller and thus an expeced R15 of more like 100 K.

R15 is actually 73.3k (nominal value 73.2k). I have since updated my SPICE model to reflect this. R22 is 12.5k.

UPDATE: Just gone back and powered up. +7VR started at 7.54V. Took around 10 secs to drop to +7.01V.

[Kleinstein]

The 15 V regulator has C59+R196 as a damping element, likely to help with the compensation and avoid oscillation. Maybe a similar combination could also help the 7 V ref.

The simulation may tell if the 7 V part is well stable or tends to oscillation: e.g. add a AC current load to the output and in an AC simulation look at the 7 V voltage. Ideally there should be no resonances an no phase shift of more than some 100 deg, ideally less than 85 deg.

[WaveyDipole]

The simulation may tell if the 7 V part is well stable or tends to oscillation: e.g. add a AC current load to the output and in an AC simulation look at the 7 V voltage. Ideally there should be no resonances an no phase shift of more than some 100 deg, ideally less than 85 deg.

Thank you for the suggestion but AC analysis is an area that I haven't really experimented with yet so I am unsure how to proceed. I am using LT Spice and have added a current source to the output and presume that I need to set it for small signal AC analysis? What amplitude should I use, and do I need to set an AC phase value? (assuming 0). When configuring the simulation command for AC analysis, what starting and ending frequency do I use? In the meantime I am doing a bit of Googling in an attempt to answer these questions.

BTW, I did one other quick test by reconnecting the +7VR rail to the rest of the circuit while monitoring its voltage while the meter was powered up. It didn't move from 7.01V, so maybe the voltage on the rail is parasitic rather than as a result of a leak?

[Kleinstein]

The amplitude and phase can be set to the default 1 and 0. A phase is only relevant with more than 1 AC source at a time.
The small signal AC analysis could run over something like 1 Hz to maybe some 100 MHz, some 20-50 points per decade - it is very fast anyway.
Looking at the AC voltage at the output would than directly give the output impedance (response scalled to 1 A of excitation). So 0 dB would be 1 Ohm.

This type of analysis is quite handy for voltage regulators / lab supplies. A phase shift of more than +-90 deg. indicates that with a "suitable" load impedance it will oscillate. With less than 90 deg. it should be stable with any passive (that is no power source) load.