Test Equipment Anonymous

Several years ago I succumbed to the Easter egg demo on this function generator at the Milton Keynes radio rally, when it was still at the Irish centre. I don't recall how much I paid but it might've been £60. I've never used it as much as I wanted to because it had a very bad habit of rebooting itself after an hour or so, which isn't good if it's doing something important where the loss of signal could cause expensive and annoying problems.
These unexpected restarts happened without warning and set all the parameters to their default, a constant 1kHz 10mV sine after the normal countdown startup.
Yesterday I decided it was time to find out why it was doing this and if I could stop it.
Checking online I found various people with similar problems and a few solutions. There was a theory that an LM339 comparator suffered from drift and caused the power supply to protect itself either from overcurrent or overvoltage.
I wasn't sure if that was happening but LM339s are cheap so I changed it just in case. An hour or so into testing and it was clear that the LM339 wasn't the root cause but it did give me a big clue.
Monitoring the output I noticed some strange pulsing, almost like the output of a digital noise generator, amplitude of the pulsing was several hundred mV and it was really quite strange as the pulses were quite well defined shifts in amplitude but with seemingly random duration and period. There was no obvious source of these pulses, the supply rail had some digital noise but nothing like this and very periodic. The pulses must've been coming from downstream. I slept, worrying that it might be "the MOSTEK ROMS" or the "1k x 4 bit static RAMs" that I'd read about, then decided to look at the rest of the schematic.

The next part of the circuit cleans up the LM339 output, mixes it with a power-on pulse and runs on the 3314A's battery when the power is off, Schottkying its way to the 5V rail when the power is on. The job of this part of the circuit is to cleanly reset the computer on startup. Hmm.

It's a CMOS hex Schmitt trigger similar to a 40106BE but presumably a bit older. I decided to monitor this 14584B's output, which is formed by 3 paralleled inverters. I set the DSO to trigger when it noticed something happen, in single trigger mode. After an hour or so it had caught the output of these inverters being very analogue and delivering a range of voltages that shouldn't be coming out of a Schmitt trigger. The inverters must've been oscillating and steadily building up electrons in all the wrong places until such time as the outputs flipped out. Old CMOS does the silliest things.

I changed the hex inverter (I had one in stock, but only one, though I'm sure a 40106 would've worked too) and just in case the flipping out was triggered by a leaky 2N3906 on its input, I changed that too. This has fixed the problem, it has been playing random arpeggios to me for several hours now, under HP-IB control from Python + linux-gpib.

A magnificent effort! CMOS pathology has driven good men to drink. Circuits that ought to work but are stubbornly dysfunctional until an oscilloscope probe is put somewhere that ought to be irrelevent, and then they spring to life.

I can't believe HP would have released it in that state, but there may be ageing effects at work. They did make a few unfortunate decisions, such as the crappy plastic gears in the 8640A/B.

I have two bugbears:

A Schlumberger/CRC sweeper with Nixie tubes, which I had working on the low range at one point, but whenever I worked on it things went wrong with the TE I was using with it. Eventually a varicap diode died when I was trying to get the high range to work. I tracked a couple down in the end, but I found I couldn't raise the steam to face it. I've since acquired better sweepers.

A Racal 1991 frequency counter which is barely responsive. I have a 1992 which is a joy to use and doesn't have any naff switches. It has the double overnised OCXO. The 1991 has the dubious TCXO. I've tried to track down the problem methodically and have changed a few chips I suspected. I copied the EPROMs from the 1992 and 1991 and they were the same, then I re-flashed the 1991 EPROM thinking that might reinforce any dodgy bits. No difference. I'd really like to get hold of the source for the firmware then it would be sensible to put a logic state analyser on the buses. The documentation is reasonably good.

I have a go at them from time to time.

Congratulations on the repair. Have you considered sending it to one of the groups.io mailing lists?

I regard such repairs as being better than crosswords/soduku at staging off dementia. But, unless the equipment has unique qualities, not necessarily entirely rational

No, I'm not entirely rational.

tggzzz wrote: ↑Fri Jan 09, 2026 10:04 pm No, I'm not entirely rational.

Are any of us?

-Pat

Cubdriver wrote: ↑Fri Jan 09, 2026 10:46 pm

tggzzz wrote: ↑Fri Jan 09, 2026 10:04 pm No, I'm not entirely rational.

Are any of us?

Pleasingly... no.

Zenith wrote: ↑Fri Jan 09, 2026 9:40 pm A magnificent effort! CMOS pathology has driven good men to drink. Circuits that ought to work but are stubbornly dysfunctional until an oscilloscope probe is put somewhere that ought to be irrelevent, and then they spring to life.

The probing of the defective CMOS hex inverter was like this, I used a DIP clip to extend the pins and clipped the oscilloscope's 10:1 probe to the top of the output, then waited. And waited.

The chaotic state machine that was happening inside the CMOS had been affected by the extra load of the clip and probe and instead of the usual hour before the reset, it was closer to three times that. I don't usually leave the LeCroy 9400 running for that long and was becoming concerned that maybe it would fail before it had detected the anomaly. That was when I decided that it was a watched pot and the only way to make it boil was to ignore it, so I had a long soak in the bath and when I returned the 9400 was displaying a mess of pulses with varying amplitudes in the region of 200 to 300mV, not the normal 0V or 5V Schmitt trigger outputs.

synx508 wrote: ↑Sat Jan 10, 2026 7:00 am I don't usually leave the LeCroy 9400 running for that long and was becoming concerned that maybe it would fail before it had detected the anomaly. That was when I decided that it was a watched pot and the only way to make it boil was to ignore it, so I had a long soak in the bath and when I returned the 9400 was displaying a mess of pulses with varying amplitudes in the region of 200 to 300mV, not the normal 0V or 5V Schmitt trigger outputs.

Is the LeCroy 9400 a dodgy number, likely to overheat or something? I wouldn't hesitate to leave either Siglent set up to catch something like that.

Normally I'd leave a logic analyser running to capture such rare events. But LAs presume good signal integrity, which isn't the case here.

I once had a pig of a problem with my 485: it would take a long time to start up from cold, but would restart quickly. I could do a partial test once every three hours and a full test every 12 hours. Turned out to be a dicky electrolytic only used to kickstart the SMPS.

At least your problem didn't fall into the boring "it's always the electrolytics" category

Zenith wrote: ↑Sat Jan 10, 2026 12:05 pm Is the LeCroy 9400 a dodgy number, likely to overheat or something? I wouldn't hesitate to leave either Siglent set up to catch something like that.

Not really, it's an early high performance DSO so it turns a fair amount of power into heat, it's also about 40 years old. It has never given the slightest hint that it won't be completely dependable and this just makes me more suspicious of it.

Haha! Or should I say something much worse?

I have not fixed it.

Yes, it has rebooted itself. I guess I'm going to be changing the non-volatile RAM. All the nvram presets that I'd stored had been replaced by defaults.

Definitely a gremlin. I've heard you can remove them by placing the infested equipment in a pentagram with a goats fat candle at each point and making certain incantations. I haven't tried it myself.

Zenith wrote: ↑Sat Jan 10, 2026 3:13 pm Definitely a gremlin. I've heard you can remove them by placing the infested equipment in a pentagram with a goats fat candle at each point and making certain incantations. I haven't tried it myself.

Accompanied by a flame thrower.

I occasionally have another go at the Racal 1991, but it's wearing thin and I'm getting fed up with it. The day might arrive when I forget the pentagram and just use the flamethrower on it.

The problem with a lot of TE with microprocessors, is that the documentation very often gives no details what the microprocessor is doing, not even a flowchart. There are EPROM dumps available online, which can be useful, but the source code for the microprocessor never seems to be released. I suppose it should be possible to feed the firmware through a disassembler and reverse engineer it. It wouldn't have the original variable names and comments, and it would be a lot of work.

I also have a RACAL 1992, the red power button fell off a few years ago and I never found it - perhaps the vacuum cleaner did. I like the 1992 but I found that, a bit like the HP 3314A, the computer tended to do strange things if left powered on for too long. It's probably a thermal issue in the RACAL.

I have a Racal 1992 with the double oven frequency standard, which works perfectly. The OCXO seems good to a few parts in 10^9 and there are no problems with the buttons.

The 1991 I bought for £40, which is cheap for one of those. It has the TCXO. Sometimes when it's powered up a few digits appear. The buttons seem to do nothing. The standby works until it stops working. I've checked the EPROM and its contents are the same as the 1992. I've swapped the microprocessor for one out of the HPIB board - no difference. I've thought I've traced the problem to defective ICs and have changed a few. No difference. There's a clock going into the microprocessor and some of the pins show activity I'd expect on the ports. The manual is good, but not brilliant. I can't see much point in putting an LSA on the µprocessor because I don't really know what to look for. I've no clear idea what the µprocessor is doing. There's a fault finding flow chart, which runs out of road before pinning down the problem. I can't see anything like damaged components or dry joints. The 1992 has known good replacement units, which is an obvious way to isolate problems, but I'm reluctant to mess about with something that works so well. A really good technical description would be useful.

There doesn't seem to be anything wrong with the power supply lines, but sometimes when it's powered down then powered up quicky, a few digits appear. I wonder if it's something to do with the way in which the power supply lines are brought up

I occasionally have another go at it. I'd buy another spares/repair 1991/2 in the hope that out of the two one could be made to work, however, I wouldn't be prepared to pay much for another non-working one, and they are never particularly cheap.

It's been a source of great frustration.

Some PCB assemblies refuse to be repaired, sometimes this is because they've absorbed something that makes them conductive or that has dissolved vias. I had the former with a Marconi 2019A that I'd bought for £100, it mostly seemed fine except it couldn't manage to tune over its full range.
The oscillator is split into chunks to achieve a linear voltage to frequency characteristic - keeping away from the low voltage end of each varicap by switching to a different oscillator, each with its own PCB line inductor.
More than one of these switched oscillators would only oscillate at one end of its range, which was rather strange, the tuning voltages were also well off as if there was some extra capacitance in parallel with each tuning diode.
I started by changing a few BFR91s or whatever it uses, but it didn't help. On one of the oscillators, I tried removing all the components and replacing them with fresh new ones but it was the same - which pointed to the PCB striplines being the problem.
In the world of music synths there are several models from the 1980s that have issues relating to the degradation of epoxy resin, either in hybrid modules or on capacitors. The Marconi's oscillator did have some sort of sealant or adhesive around some parts, possibly to help with microphonics or make it more resistant to handling shock, so perhaps that was the root cause.
In the end I paid about £50 to have one of the Israel based scrap test equipment companies send me a new oscillator block, which solved the problem. It was also an opportunity to learn about calibrating the 2019A's frequency modulation system, which was quite interesting, as I recall you have to adjust the modulation depth at a whole series of frequencies and the 2019A stores these values in EAROM. In use it uses interpolated values to adjust the modulation depth correction, which keeps it quite close to the calibrated values.

While I'm thinking of Marconi, I also have a 2305 modulation meter which works except for the optional distortion/SINAD board, which is not recognised at all by the firmware. One day I'll figure out why.

synx508 wrote: ↑Sun Jan 11, 2026 4:46 pm In the world of music synths there are several models from the 1980s that have issues relating to the degradation of epoxy resin, either in hybrid modules or on capacitors. The Marconi's oscillator did have some sort of sealant or adhesive around some parts, possibly to help with microphonics or make it more resistant to handling shock, so perhaps that was the root cause.

Interestingly, one of the Hi-Fi comics (Hi-Fi News January 2026, pp133) mentions a similar issue with gloop attacking PCBs. It's well-known that you shouldn't allow PVC cable to touch expanded polystyrene (gives off vapours that dissolve the polystyrene into a sticky mess), and I've learned to be careful with plastics. Then there's that rubber(?) coating on cameras etc that gives a nice grippy feel but after twenty years goes sticky and yucky. My Nikon D100 went that way, Oh, and unstable binders in 1970s magnetic tape lost a lot of master tapes from that era until the technique of baking and quickly recording onto another medium was developed.

I'm glad you mentioned the Marconi 2019A. I have a 2019 I bought from Stewart of Reading about five years back, when my HP8640A began to play up because of the geriatric gear problem. He said he'd sorted out a good one, had changed one of the LCD displays and burned it in. When it arrived, it didn't work. I had a chat with him and agreed it had probably been shaken up in transit. I reseated all boards and connectors and it was fine. The frequency reference is a bit off, but that can be expected with 30 to 40 year old OCXOs and I have plans to replace it with one of the excellent Aliexpress ones. I use the 2019 when I have a need for a serious RF signal generator.

I bought a second 2019 for £70 at MK last year, assured that it was fine, apart from a LED not working. It works after a fashion but the frequency is nothing like expected and a LCD digit segment is dark when it shouldn't be. I expect that's a bad connection or driver IC. This is a case of the seller not being outright deceitful, but he probably switched it on and found it did things he expected and didn't look any further. I really bought it as a parts mule, but as is often the way with parts mules, it seems so close to working properly that it seems a shame to cannibalise it. It's in the queue to be looked at, with a view to making it fully functional. I wouldn't pay £50 for a sub-assembly for it. I might pay that much for another 2018 or 2019 as a parts mule. This is the slippery slope.

As regards degradation of PCBs, that's worth bearing in mind, especially with twitchy RF circuitry relying on the properties of the board. Ageing can be variable and storage conditions can involve heat cycling and damp. A lot of items I come across have obviously been kept in a damp garage for ages. Things can happen, such as contamination with tobacco smoke, or being given a good squirt of WD40 cure all. Although I don't often do it, storing recently acquired treasures in the airing cupboard for a week to remove damp is a wise move.

Zenith wrote: ↑Mon Jan 12, 2026 11:34 am ... and a LCD digit segment is dark when it shouldn't be. I expect that's a bad connection or driver IC.

I've frequently "rejuvenated" LCDs by using IPA to clean the zebrazone connector between the LCD glass and PCB. I'm not sure that would help with a dark segment.

As regards degradation of PCBs, that's worth bearing in mind, especially with twitchy RF circuitry relying on the properties of the board.

Tek 4x5 scopes have quite a few phantom capacitors in the schematics, annotated as being part of the PCB.

Ageing can be variable and storage conditions can involve heat cycling and damp. A lot of items I come across have obviously been kept in a damp garage for ages. Things can happen, such as contamination with tobacco smoke, or being given a good squirt of WD40 cure all. Although I don't often do it, storing recently acquired treasures in the airing cupboard for a week to remove damp is a wise move.

I'm always doubtful about spray-on gunk. I do have two DeOxit sprays, one for carbon wipers which I will spray into pots. The other, D100, I spray into the cap and then use a jewellers screwdriver to deposit a small drop exactly on the contact.

The Solartron 7081 manual has a relevant "warm-up" procedure.

You'd imagine that proper contact cleaner sprays would be very pure and leave no residue, and would have been thoroughly tested with PCBs and electronic components. I don't like the spray and pray approach, especially with RF circuitry. IMHO it's much better and less wasteful to spray a bit into a plastic cap and apply it where it's needed with a soft bristled paintbrush reserved for the purpose.

I've found zebra strips fairly easy to deal with. Clean them up with IPA and put them back together, as you say. In the factory, they may have given them a wipe with IPA, but they would have assembled them without special jigs etc. Very often the LCD cell has clearly been damaged, and since they were mainly custom made, they are not easy to replace unless you have an undamaged pull. The LCD in the 2019 does not appear to be damaged and a problem with the zebrastrip would be more likely to cause a segment not to activate, rather than cause unwanted activation. That's why I assume the driver IC is playing up. I'll have a clearer idea when I've investigated it properly. If it can't be fixed, it's a slightly irritating fault, but not a showstopper.

Zenith wrote: ↑Mon Jan 12, 2026 12:36 pm You'd imagine that proper contact cleaner sprays would be very pure and leave no residue, and would have been thoroughly tested with PCBs and electronic components. I don't like the spray and pray approach, especially with RF circuitry. IMHO it's much better and less wasteful to spray a bit into a plastic cap and apply it where it's needed with a soft bristled paintbrush reserved for the purpose.

When cleaning sprays are used there has to be a residue of some sort or other. That could either be the cleaning compound, or the muck moved (cf removed) by the spray and left behind when the spray evaporated.

For RF circuits I imagine the residue could change the capacitance/inductance/losses, thereby "detuning" the circuit.

For precision voltage and low current circuits, even a slightly conductive residue could be a problem.

Washing assemblies (cf spraying) in solvents would allow the residue contaminants to be diluted/removed - but you've got to have the right solvent in "large" quantities and use it safely.

tggzzz wrote: ↑Mon Jan 12, 2026 1:00 pm

Zenith wrote: ↑Mon Jan 12, 2026 12:36 pm You'd imagine that proper contact cleaner sprays would be very pure and leave no residue, and would have been thoroughly tested with PCBs and electronic components. I don't like the spray and pray approach, especially with RF circuitry. IMHO it's much better and less wasteful to spray a bit into a plastic cap and apply it where it's needed with a soft bristled paintbrush reserved for the purpose.

When cleaning sprays are used there has to be a residue of some sort or other. That could either be the cleaning compound, or the muck moved (cf removed) by the spray and left behind when the spray evaporated.

I suppose so. I've never been able to find what's in contact cleaners such as Servisol, Maplin's juice, etc. It says on the can that the contents are extremely flammable, but it doesn't list the ingredients. I'd have thought they were fairly common organic liquids. I suppose there might be a trace of lubricant and a reducing agent to deal with oxides. I normally clean rotary switches with IPA on cotton buds, which removes a lot of the grime, then apply the contact cleaner with a brush. Sometimes I've flooded sealed switches with IPA as there really is no other way. If there's no other option I point and squirt. I always lose the extension tube that comes with a can of contact cleaner.

Is there a material difference in effect between pure IPA and purpose made contact cleaner?

Zenith wrote: ↑Mon Jan 12, 2026 1:57 pm

tggzzz wrote: ↑Mon Jan 12, 2026 1:00 pm

Zenith wrote: ↑Mon Jan 12, 2026 12:36 pm You'd imagine that proper contact cleaner sprays would be very pure and leave no residue, and would have been thoroughly tested with PCBs and electronic components. I don't like the spray and pray approach, especially with RF circuitry. IMHO it's much better and less wasteful to spray a bit into a plastic cap and apply it where it's needed with a soft bristled paintbrush reserved for the purpose.

When cleaning sprays are used there has to be a residue of some sort or other. That could either be the cleaning compound, or the muck moved (cf removed) by the spray and left behind when the spray evaporated.

I suppose so. I've never been able to find what's in contact cleaners such as Servisol, Maplin's juice, etc. It says on the can that the contents are extremely flammable, but it doesn't list the ingredients. I'd have thought they were fairly common organic liquids. I suppose there might be a trace of lubricant and a reducing agent to deal with oxides. I normally clean rotary switches with IPA on cotton buds, which removes a lot of the grime, then apply the contact cleaner with a brush. Sometimes I've flooded sealed switches with IPA as there really is no other way. If there's no other option I point and squirt. I always lose the extension tube that comes with a can of contact cleaner.

Is there a material difference in effect between pure IPA and purpose made contact cleaner?

Yup. I use an elastic band to hold the extension tube to the can.

Even MSDSs aren't much help with the interesting ingredients.

The difference between IPA and a contact cleaner is almost certainly the trade secrets that end up as the residue.

If you are really serious about this sort of thing, there's the thread that EC8010 started "Defluxing such that single-digit fA electrometers work". He does seem to have exacting requirements. It's a related problem, but not quite the same as a PCB which plays up because its properties at hundreds of MHz have altered.

Zenith wrote: ↑Mon Jan 12, 2026 11:34 am I
I bought a second 2019 for £70 at MK last year, assured that it was fine, apart from a LED not working. It works after a fashion but the frequency is nothing like expected and a LCD digit segment is dark when it shouldn't be.

I had this with my 2019, it was the level shifting CMOS chips used to drive the display, a unit in one chip had gone bad and an output was stuck.
I bought a few tubes of replacement chips and I note that they're not discontinued and quite hard to find so if you remind me I'll try to keep a tube with me at the next rally we both attend.

I did the repair from the top side of the PCB because you're supposed to need a special tool to dismantle the LCD assembly from its PCB.