Hewlett-Packard 3314A function generator

[tggzzz]

Hmmm.

Dropout for half a mains cycle.

Could there be a (safety?) capacitor on the mains side that is occasionally shorting out and dragging down the input voltage, followed by the short being destroyed and the input voltage recovering?

Do you have the "high voltage output" option, since that appears to have a separate PSU connected across the transformer output.

[synx508]

It's as if you're tempting me into building a pulse swallower that works on the mains! That would be quite a neat gadget and cheaper than buying one of those huge AC power sources that can be programmed to misbehave.

I don't think it can be a missing cycle, though, there's a reasonably large capacitor in there and I don't think it'd empty that quickly, also i'd see a ramp across the whole period of the glitch on the unregulated side and that's not what's happening.

There have been no glitches since yesterday evening, the LeCroy 9400 is sitting here waiting for one, producing its serious kit AC fan growl.

[tggzzz]

I'm far from convinced by the concept of a problem with the rectified output. Nonetheless, I'd like to understand what produced the linear rise/fall ramps shown on the scope CH1 output.

I don't think it can be a missing cycle, though, there's a reasonably large capacitor in there and I don't think it'd empty that quickly, also i'd see a ramp across the whole period of the glitch on the unregulated side and that's not what's happening.

Knowing the current drawn by the PSU, you could calculate dV/dt=I/C and see if they are in the same ballpark.

The duration of the regulated droop will partially depend on how long it took before the unregulated input drooped so far that the regulator ran out of headroom.

If the problem is after the bridge rectifier, it might only be a half cycle.

How about disconnecting the mains completely and connecting the bench PSU across the bridge rectifier output? Different current limits could be used to test different hypotheses.

[Zenith]

I'm not convinced by the dropped cycle theory, but it can't be ruled out. The mains transformer could be disconnected and a bench PSU used in its place. The rectifiers are cheap so replace them if there's any doubt. Is the High Voltage Output Option fitted? That takes AC power from the transformer secondary and may be a problem. You'd have to work out whether it could be removed or run from DC. Is it fitted? Could it be causing the glitch? Then there's C108 and C109. Probably plastic film and generally unlikely to cause problems at all - unless they have cracked cases. I can't really imagine one breaking down for a cycle then reforming. There's the fan to consider.

The transformer could be tested separately. Say with a rectifier, capacitor and load, and watch for a glitch with a DSO.

The difficulty with all this is that it takes an age for the fault to appear.

To cast the net wider, could this be a fairly regular mains glitch caused by something like a fridge? It shouldn't cause a problem, but the 3314A may be particularly sensitive. As I think you pointed out, there's a lot of capacitance to cope with a dropped cycle. I think you'd notice it with other things.

[Zenith]

Hmmm.

Dropout for half a mains cycle.

Could there be a (safety?) capacitor on the mains side that is occasionally shorting out and dragging down the input voltage, followed by the short being destroyed and the input voltage recovering?

No lurking Rifas, but there's a mains input filter with three safety capacitors. After all this, it couldn't be a bloody standard input filter playing up.

Do you have the "high voltage output" option, since that appears to have a separate PSU connected across the transformer output.

I should have read your posts before submitting mine.

[tggzzz]

Do you have the "high voltage output" option, since that appears to have a separate PSU connected across the transformer output.

I should have read your posts before submitting mine.

Life's too short; been there done that, many times!

[tggzzz]

I'm far from convinced by the concept of a problem with the rectified output. Nonetheless, I'd like to understand what produced the linear rise/fall ramps shown on the scope CH1.

The transformer could be tested separately. Say with a rectifier, capacitor and load, and watch for a glitch with a DSO.

It might be useful to observe the transformer output directly. A DSO might offer some form of missing pulse trigger, e.g. trigger if input has not exceeded the trigger level for 10ms.

To cast the net wider, could this be a fairly regular mains glitch caused by something like a fridge? It shouldn't cause a problem, but the 3314A may be particularly sensitive. As I think you pointed out, there's a lot of capacitance to cope with a dropped cycle. I think you'd notice it with other things.

45 years ago I had an old fridge that produced a spike on my audio amplifier output. Took a very long time to notice the correlation! Cured by wrapping the fridge's lead through a ferrite toroid several times.

[synx508]

I've already gone some way to exhausting the mains-borne trigger theory by producing many, many glitches by turning on some of my more unruly appliances. I managed to trigger the DSO a few times with very short duration spikes but the 3314A didn't respond to them at all.

Still no glitches today and I'm starting to wonder if this is a rectifier thing, I've not had a glitch since I very roughly wiggled the rectifier diodes. It has a pair of those very HP metal clad diodes that look a bit like TO5 transistors with one lead coming from the can and the other in the middle underneath doing the positive side and a TO220 style dual diode doing the negative side. However, I did that while it was powered on and there was no glitch. If it doesn't glitch soon I'm going to give it a few days rest before resuming, a kind of TE cliffhanger.

[Zenith]

To cast the net wider, could this be a fairly regular mains glitch caused by something like a fridge? It shouldn't cause a problem, but the 3314A may be particularly sensitive. As I think you pointed out, there's a lot of capacitance to cope with a dropped cycle. I think you'd notice it with other things.

45 years ago I had an old fridge that produced a spike on my audio amplifier output. Took a very long time to notice the correlation! Cured by wrapping the fridge's lead through a ferrite toroid several times.

I had the same problem, which is what brought it to mind. It was a well known phenomenon. I solved it by using a large mains filter which came out of a tape drive for a mini computer.

[Zenith]

I've already gone some way to exhausting the mains-borne trigger theory by producing many, many glitches by turning on some of my more unruly appliances. I managed to trigger the DSO a few times with very short duration spikes but the 3314A didn't respond to them at all.

Still no glitches today and I'm starting to wonder if this is a rectifier thing, I've not had a glitch since I very roughly wiggled the rectifier diodes. It has a pair of those very HP metal clad diodes that look a bit like TO5 transistors with one lead coming from the can and the other in the middle underneath doing the positive side and a TO220 style dual diode doing the negative side. However, I did that while it was powered on and there was no glitch. If it doesn't glitch soon I'm going to give it a few days rest before resuming, a kind of TE cliffhanger.

Have you remade the soldered joints to the rectifiers and checked with a magnifying glass for cracked circuit lands? I've found that the joints to rectifiers are prone to going bad, probably because they can get hot. Dry joints can cause obscure problems. Cracked circuit traces are not always easy to spot, and are another source of PITA problems.

[tggzzz]

I've already gone some way to exhausting the mains-borne trigger theory by producing many, many glitches by turning on some of my more unruly appliances. I managed to trigger the DSO a few times with very short duration spikes but the 3314A didn't respond to them at all.

Still no glitches today and I'm starting to wonder if this is a rectifier thing, I've not had a glitch since I very roughly wiggled the rectifier diodes. It has a pair of those very HP metal clad diodes that look a bit like TO5 transistors with one lead coming from the can and the other in the middle underneath doing the positive side and a TO220 style dual diode doing the negative side. However, I did that while it was powered on and there was no glitch. If it doesn't glitch soon I'm going to give it a few days rest before resuming, a kind of TE cliffhanger.

Have you remade the soldered joints to the rectifiers and checked with a magnifying glass for cracked circuit lands? I've found that the joints to rectifiers are prone to going bad, probably because they can get hot. Dry joints can cause obscure problems. Cracked circuit traces are not always easy to spot, and are another source of PITA problems.

Pull and push and twist the buggers?

[synx508]

I didn't resolder the rectifiers when I last had the board out, because they looked fine under magnification and at that point they didn't seem like a likely cause. It might not even be the solder end that is bad, perhaps the rectifiers themselves are mechanically knackered and achieving near magical levels of thermally triggered break faults that cure themselves.
This would fit some more of the fault profile, though, with the -15V being entirely unaffected. The 5.1V gets reset and is held up by being a SMPS, where the reference is also moving in sympathy.

[synx508]

Pull and push and twist the buggers?

They're quite hot at the moment, but wiggling them does nothing, no restarts, no DSO triggering.

[tggzzz]

Do you have a thermal camera, or even one of the point-and-shoot non-contact thermometers? They have a surprisingly and usefully narrow field of view.

[synx508]

No, but I'm not worried about the heat there, they were toasty but not danger toasty, more hot radiator that you wouldn't want to keep a hand on too long. I've decided the interconnect between A2 and A3 needs to be replaced. It's just too sensitive to position and movement to make faultfinding productive. Had a quick look on RS, almost everything that might be close to useful is out of stock. Had a look on Farnell, the filtering was impossibly broken. I've got a lot of crimp terminals, 3 and 4 way housings for both real and compatible 0.1" molex connectors, I could use a load of them, it'll take forever to build the cables. I might do the A1 interconnect too if it works.
Might have a look on digikey and mouser later because I don't really like the idea of doing that much crimping.

[synx508]

Instead of that horror I've just reversed the cable end to end which seems to have fixed it. This shouldn't work but it did, for now at least. Let's see if it glitches…

[Zenith]

I hope it fixes it. Bad connections and bad cables have caused me plenty of problems in the past.

There have been a few false dawns so far, but fingers crossed.

[EC8010]

Instead of that horror I've just reversed the cable end to end which seems to have fixed it. This shouldn't work but it did, for now at least. Let's see if it glitches…

Ah yes, over the years, all the electrons had piled up at one end.

[synx508]

I hope it fixes it. Bad connections and bad cables have caused me plenty of problems in the past.

There have been a few false dawns so far, but fingers crossed.

No, no, the false dawns will continue.

Swapping the cable has fixed the occasional E51, E30, E31 and E34 but very late last night, after I'd removed the +15V probe and after a whole day with no glitches it glitched and corrupted the NVRAM to the extent that it triggered the E09 which causes it to restore everything to defaults on restart.

I'm not the only person to have been tormented by this 3314A, I'm sure that the previous owner only replaced the CPU out of desperation, before selling it cheap at a rally but judging by the wear on the connector on the A2 board it has been on and off many, many times.

[Zenith]

Oh well, back to the rectifiers.

I'm not the only person to have been tormented by this 3314A, I'm sure that the previous owner only replaced the CPU out of desperation, before selling it cheap at a rally but judging by the wear on the connector on the A2 board it has been on and off many, many times.

Without detailed information on the unit, which the designers would have had, such as source code for the µprocessor, systematic fault finding will only narrow things down so far, so it becomes educated guessing and swapping out parts, especially with a fault that isn't hard and fast.

They ask what seems to me a lot of money for these on thieves bay.

[synx508]

I captured a nicer picture of the ramp down from 15 to 11.08 volts. It's not quite the smooth slope that I thought previously, but it's still ramp-like.

[Zenith]

It looks to be as if the slope is modulated with a 100Hz FW rectified waveform. It's not very clear. That may shed some light on the problem.

[tggzzz]

It looks to be as if the slope is modulated with a 100Hz FW rectified waveform. It's not very clear. That may shed some light on the problem.

It does look a little like that.

Sometimes tack soldering a capacitor in parallel with the existing capacitor can help indicate if there's a fault with the original capacitor.

[synx508]

It looks to be as if the slope is modulated with a 100Hz FW rectified waveform. It's not very clear. That may shed some light on the problem.

Yes, I agree, it looks like the ripple on the unregulated side of the pass transistor. The ripple stops when it reaches 11.1V, which has to be about what you'd expect if the regulator circuit had fallen back to the 12V zener powered bootstrap circuit based on Q106 (which is drawn with the zener facing the wrong way and an NPN where there should be a PNP, I think edit: in the -15V side, it's drawn correctly on the positive side). Am I misunderstanding this part of the circuit? I don't think so, but perhaps I am. Once the output of the supply forward biases Q106 the control loop is closed again, this is what happens if the op-amp isn't doing its thing and given that the op-amp is powered from the regulated ±15V rails, it would seem to be prudent to have another option for getting those rails to do something sensible.

I could replace the two 1N4148 or similar diodes in the emitter of Q104 or Q102, which is the current limiter, there to reduce the B-E voltage on Q100 when there's a voltage across the current shunt or if there's a short on the output (via R104).

That's it, I'm going to replace CR118, CR20, R108 and C102.

And, on the off-chance that it's the input side of the op-amp that's glitching (remember I've not managed to record the output of the op-amp in the feedback path, as I waited all day and it didn't glitch), CR108, R151, R112, R114 (trimmer for +15V), R116.

Tomorrow. Unless I decide not to.

[synx508]

I've replaced almost all the things on the list, I left the original 20kΩ reference/feedback divider resistors in place, though. Loads of new diodes, including CR125, which feeds +15V to the op-amp's power pin, in case that was intermittent. It's back on test.

Edit: I have higher hopes this time because I've been reading about 1N4148 failures and one of them is very similar to what I've been experiencing. I reviewed the settings that I used for the -15V test and it's possible that I'd set the trigger such that -11V would not have triggered it, so it's possible that CR125 was opening and turning off one rail of the 1458 op-amp. With that in mind it takes the number of 1N4148s that could cause the problem up to 3. I've replaced 4, just to be sure.