HeathKit - Bell & Howell Schools Oscilloscope

[FlyingHacker]

Hi,

I am taking at look at an old Bell & Howell School oscilloscope marked as "Solid State Oscilloscope Triggered Sweep 5MHz Bandwidth" it is the blue one that is a little taller than it is wide.

Also known as:
Heath Model # 100-203-31
Bell & Howell # Model 9560-1
5 Inch Triggered Sweep Oscilloscope

Some docs here. I plan on assembling those into a PDF:

www.bunkerofdoom.com/lit/bellandhowell/index.html

Here is the PDF... The owner of Bunkerofdoom said to make it into a PDF was OK with him.

http://meanmutha.com/BellHowellScopeManual.pdf


Will post pics tomorrow.

I got it generally up and running. Was way out of cal and you couldn't find a trace at all. But I went through the cal procedures and it is generally OK now.

I am having an odd issue, though. When I set the horizontal division (uSec/cm) to 1 (the lowest setting) I am unable to trigger on a signal from about 80-120kHz. It will not get a stable trace even if you turn up the Stability control (need to research exactly what that is, as I am used to Tek scopes, which have no such control). I will not trigger with internal or external triggers in that frequency range. But only if the horizontal is set to 1uSec/cm.

I tried two function gens, and with and without termination. I tried sine, square and triangle waves. It just won't get a good trigger.

My wild guess is this is some ripple on the power supply. I am seeing a fair amount when measuring with an AC meter. I had around 1V on one rail, and half a volt on the other, etc.  I think I should replace the old electrolytic caps, and measure any resistors I can (lots of carbon composition).

Does anyone have a working one of these scopes? Would love to get some measurements. The docs have some voltage measurements, but no ripple tolerances like a Tek manual (professional tools vs. something you build in school).

Interesting little scope, really. It was a kit to be built by students and then used. They do this after they build their own Bell & Howell Nixie tube meter (got that in the sale too!). Both kits are made by Heath. I was planning on giving this to a friend who is new to scopes if I can get it going.

Thanks. Any info is appreciated.

[tautech]

Take a look at this part of the PSU:
http://www.bunkerofdoom.com/lit/bellandhowell/122.jpg
From the description there is only one voltage that you need to check for ripple: 31V.
All other 6 low voltages are supplied from this 31V via dropper resistors. 
You'd expect the ripple on the 31V to be less than 100mV IMO.

The other voltages supplied are listed as 180 & 150V, both unregulated.

[FlyingHacker]

Tautech,

Thanks for chiming in.

I shall check ripple on that 31V segment.

Do you think power supply issues could cause this weird trigger issue?

Note that I can see a trace when I turn up the "Stability," but it is overlapped. I find that frequencies near that bad range become much more touchy on the stability control. I have to have the stability control just barely above the point where the trace appears. If I go higher the trace begins to look all overlapped like it does in the bad range.

I tried contact cleaner on the stability pot (and all of them, worked wonders for intensity). I may try some more on stability.

Since the calibration procedure calls for a 100kHz wave with the sec/div set to 1uSec I am sure it it supposed to work in this setup.

[FlyingHacker]

I guess if the power supply tests out Ok (which I honestly doubt it will in terms of ripple anyway) then I should look at the stability pot itself, D305 and D304, R320 (the 220 ohm I may be misreading) and R330.

I am guessing either something is not switching fast enough, or is oscillating. Or, most likely, the power supply ripple is "rippling through" to the trigger circuit. Perhaps I will check the ripple around Q316 as well. I know I have TP2 at a steady DC 0.2V, but I didn't check for any ripple there. It could be through the +5V on the stability control as well, or right onto the collector of Q316.

[FlyingHacker]

Some more info:

Across emitter of Q201 to the anode of ZD201 I see 3.5 mV RMS of ripple. Doesn't seem like a ton? No?

If you look at the voltage measurement diagram for the power supply. I get <1mV from ground to AD, but ~310mV RMS from ground to the test point labelled AC.

~2mV RMS at the top of R214 as indicated as -18V

<1mV on K, R, and S to ground


The voltages are a little off, though (I doubt too far):

K = 4.596V

P = -8.37V

R = -4.539V

S = -8.88V

R214 = -18.05V

AC = 9.40V

AD = 8.94V

Q203e = 0V
Q203b = 0.206V
Q203c = 53.9V

V = 156.9V
X = 214.9V

[tautech]

If there is significant ripple on the PSU it will affect all operation of the scope.

Back to basics, get the PSU right first.
Check all main 'lytics even those used for local bulk capacitance.
http://www.bunkerofdoom.com/lit/bellandhowell/4bellandhowellscope.jpg
Eg. C307 on the 9V rail

[FlyingHacker]

The base and emitter voltages of Q316 read like the schematic reference voltages (with Stability off of course).  I have the base tuned to 0.2V per the cal procedure. The emitter is 3.685V (note that the student assembling this is using a 2.5 digit multimeter which they assembled).

I think it may just be that I cannot get a fine enough adjustment on the Stability pot. Even putting my light load of my Fluke 87 on the circuit can occasionally get it stable.

I hooked a Simpson 260 up to the wiper of the Stability pot to I could turn ithe knob and look for jumps or dead spots. It didn't have any dead spots. It seemed to double back on itself a TINY bit right around the area of interest, but this could have just been a change in the loading on the rest of the circuits caused by the trace popping up. I was measuring the voltage coming off the wiper, rather than the resistance on an open circuit.

I am tempted to add a trim pot between pin 3 of the Stabilty pot and the five volt rail in order to give me finer control. It looks like the stability pot must be logarithmic or something because it is at the five volt supply voltage when full off, but by the point of interest, which is maybe a quarter to a third the way on the knob, the wiper is at 1.8V or so. If I could bias the whole thing down by lowering the voltage at pin 3 (the five volt rail input to the Stability pot, which used as a voltage divider) then I should have more of the Stability knob in the more useful range.

But... It really should work as is, or they would not have had students build it like this. So I do think something is amiss someplace, but I am not sure where.

Thoughts? Thanks.

[FlyingHacker]

If there is significant ripple on the PSU it will affect all operation of the scope.

Back to basics, get the PSU right first.
Check all main 'lytics even those used for local bulk capacitance.
http://www.bunkerofdoom.com/lit/bellandhowell/4bellandhowellscope.jpg
Eg. C307 on the 9V rail

Are you saying 3.5mV RMS is significant ripple? I don't know what is SHOULD be...

[tautech]

If there is significant ripple on the PSU it will affect all operation of the scope.

Back to basics, get the PSU right first.
Check all main 'lytics even those used for local bulk capacitance.
http://www.bunkerofdoom.com/lit/bellandhowell/4bellandhowellscope.jpg
Eg. C307 on the 9V rail

Sorry missed that. 
Too eager to help.
Carry on. 

Are you saying 3.5mV RMS is significant ripple? I don't know what is SHOULD be...

[FlyingHacker]

Sorry missed that. 
Too eager to help.
Carry on. 

Are you saying 3.5mV RMS is significant ripple? I don't know what is SHOULD be...

So you think 3.5mV RMS sounds acceptable then? It didn't seem inordinately high, but I have seen some Tek scope rails require less than 1mV. A Tek scope this is not... 

I will check the caps with my ESR ("Transistor Tester"works in circuit for larger caps).  I bet they are not great.

[tautech]

Yep. I'd work with 3.5mV. 

[FlyingHacker]

I replaced C307 and C308. These were the only two I had on hand. I don't have the 350V caps on hand.

The ESR of the 350V caps was generally in spec. Though C205 was a hint high.

The ESR of the 2000uF 15V caps I replaced was ~.46ohms. Higher than the spec of around .12ohms, but not like a short or open or anything. Of course the new Wun Hung Lo replacements were around .32ohms, but were 50V.

Anyway, this had no effect on the issue, though was probably a good plan.

What do you think of my trim pot idea?

I am not sure where else to look. I will check some resistors.

[FlyingHacker]

Checked R328 and R330 as well as lifted legs and tested D304 and D305. All fine.

[FlyingHacker]

I put this on an isolation transformer and hooked up my four channel Tek 2465. Here is a photo of the waveforms.

From bottom to top the waveforms are (vertical scales not consistent):

Q316e
Q310c
Q308b
Q203b (over on power supply)

Photo one is with it working correctly, with a signal of 150kHz. If I then adjust the frequency generator to 146kHz the waveforms stay the same, even though the waveform on the scope under test goes away.

So it looks like the trigger is still working correctly, but perhaps the blanking circuit is in error for some reason.

The second photo shows the waveforms if I adjust the Stability control upward more until I get an image on the scope under test (but it is not properly triggered).

What is interesting is that I can turn the Stabilty downward and see it lose trigger completely.

To phrase another way, with 150kHz (working frequency) as I increase Stability from nothing  upwards I see first no trigger and no waveform, then trigger on the Tek and the proper waveform on the scope under test, and then the improperly triggered waveform on the Tek (photo two) and an improperly triggered waveform on the scope under test.

With 146kHz (failing frequency) as I increase Stability from nothing upwards I see first no trigger and no waveform on the scope under test, then the triggered waveforms on the Tek (first photo) but *still no wave* on the scope under test, finally I see the improperly triggered waveform on the Tek (photo two) and the improperly triggered waveform on the scope under test.

So it almost seems like the blanking circuit. I need to examine it further. Note that the collector of Q203 looks fine, I was just probing it on channel 4 of the Tek, which has limited vertical range, so I went with the base of Q203 instead.

[FlyingHacker]

I am going to study the circuit further and figure out where I need to probe. Have to be careful with the HV caps right there on the power supply.

[FlyingHacker]

As per above I *think* I have established that the issue is not in the trigger. The trigger waveforms on the Tek remain the same when I adjust the frequency (they change in frequency as well of course). However the wave disappears from the scope under test. When adjusting the frequency through the cutoff region I hear a slight buzzing from the neighborhood of IC201 or Q202. It is hard to pinpoint, and both of those are at very high voltage so I can't get too close.

I wish I had a HV scope probe  luckily I have been able to check them for a moment wit the Fluke 87, which won't read over 1000V, but at least doesn't blow up. I do this before I touch the scope probe to anything 

So, here are some other measurements...

The waveform on Q309e, as well as IC201 pin 1 do not change when the wave goes off the scope under test.

Collector of Q308 looks good, amplified version of the base. Also no change when the scope under test stops showing the wave.

[FlyingHacker]

I think I have exhausted my ideas on the horizontal and sweep trigger board. Unless something is wrong with the high voltage optocoupler IC201 or the HV transistor Q202 it seem like everything getting there is in order to sweep and blank the CRT.

Perhaps I should take a look at the vertical board. Maybe something is amiss there causing the beam to go offscreen or something. The scope has no Beam Find function, which would be useful to look for that. I may start by checking C105, an electrolytic I hadn't noticed because i was looking at the other side of the scope.

[Tomorokoshi]

Looks like a fun project.

I have one of these:
http://www.digibarn.com/collections/systems/appleII-bell-and-howell/index.html

[FlyingHacker]

Well, the vertical deflection waveforms look fine. They change based on the input waveform and frequency.

I am starting to think it's in the HV section. Maybe Q202.

I am definitely seeing the issue at 10 uSec/cm as well as at 1 uSec/cm.

I noticed that what is sometimes happening as I adjust the frequency is that the waveform on the scope under test gradually fades in and out over some different ranges. Adjusting the intensity control sometimes causes the waveform to come back, but then a change in frequency causes it to fade out again. This seems to be the most prevalent around the 10-30 kHz range.

I actually see that the fading has a sawtooth pattern as I increase the frequency. I ramp the frequency up and the waveform gets brighter and brighter, and then vanishes. Keep going up in frequency and it gets brighter and then vanishes. And so on... Definitely sawtooth.

It's like something is resonating in the Q202 circuit?

Any thoughts, anyone?

I really need a high voltage probe, but I doubt I can afford one to fix this old scope. Anyone in or near St. Louis have a high voltage probe they could lend? I have an airplane, and can fly a few hundred miles to pick it up... Thanks.

Update: the Chinese 2kV probes aren't that pricey, and might suffice for this.

Thanks.

[FlyingHacker]

Looks like a fun project.

I have one of these:
http://www.digibarn.com/collections/systems/appleII-bell-and-howell/index.html

Ah yes, the badass Black Apple ][. A friend of mine had one as a kid. I really wanted it. Were those a kit, or just branding for classrooms?

[Tomorokoshi]

Ah yes, the badass Black Apple ][. A friend of mine had one as a kid. I really wanted it. Were those a kit, or just branding for classrooms?

As I understand it, Apple was looking for a way to get into the educational market, and Bell & Howell had a good distribution channel into schools.

Bell & Howell was looking to add computers to their offerings, so it was a convenient match for both at an important time in the early computer industry.

[singapol]

Check the crt circuit, that's where INTENSITY and FOCUS are. Check for cracked solder joints.

[FlyingHacker]

Check the crt circuit, that's where INTENSITY and FOCUS are. Check for cracked solder joints.

Thanks for the input. I think the problem almost has to be in the CRT circuit.

But what would cause the intensity to vary based on the input frequency? Doesn't sound like a solder issue, or am I missing something? Maybe the audible whine I hear could wiggle a solder joint or something?

I will take a look at the board. If I don't see anything I will pull and test Q202 with a meter as a start.

Thanks.

[FlyingHacker]

I'll post some lore info about why I am asking this soon, but can somebody else please compare D306 andD307 on the schematic (transistors used has diodes for a limiter, I think) with the solder mask on the circuit board shown here http://www.bunkerofdoom.com/lit/bellandhowell/371b.jpg

Those transistors look like they do not match the schematic. Thoughts? Are they supposed to be a limiter?

UPDATE: Ack! My stupidity... Although D305 and D306 are transistors being used as diodes I was looking at Q306 and Q307!

So ignore that.

[FlyingHacker]

So the CRT circuit and HV side turned out to be a red herring (I did find Q202 was poorly soldered, and cleaned that up)...

When I previously checked the horizontal side I had only checked the trigger, not the horizontal deflection. Turns out the horizontal deflection (collectors of Q323 and Q324) also change waveforms when the trace goes away! They get a flat line at the head of each sawtooth (or could be tail, not sure what I was triggering off of). The point is that I guess the beam is offscreen long enough that you can't see the trace (or the timing gets out of phase with the blanking).

Anyway, I started tracing the changing waveform backwards through the circuit (looking for anyplace where the change of the input frequency causes the waveform to jump). Since it was the same on both Q323 and Q324 I figured it had to be someplace that affected both circuits.

I have attached some more scope screenshots from another scope.

For all of these photos the top trace is the left hand input (schematic-wise) of R336. I am triggered off of this with a hold off since the other waveform is longer.

The bottom trace in these two images is measured at the emitter of Q306 (labeled D306) UPDATE: NO, this is actually at the emitter of Q306, NOT D306... Why they chose to use transistors and name them D using the same number as an actual transistor is beyond me.

In photo one the input frequency is tuned such that waveform is displaying correctly on the scope under test.

In photo two I have slightly adjusted the input frequency such that the scope under test is no longer displaying the waveform (i.e. I caused the problem I am trying to solve). You can see the resulting waveform on the transistor change.

That is a weird pattern. Is that some sort of oscillation?

[FlyingHacker]

In these two photos the top trace is again the left hand (input) side of R336.

The bottom trace is the Gate of Q317.

In photo three (the first photo in this post) the input frequency is tuned such that the scope under test is working correctly.

In photo four (the second photo in this post) the input frequency is again slightly adjusted such that the scope under test no longer displays a waveform (no trace).

You can see how the waveform changes.

[FlyingHacker]

So you notice in my UPDATEs above I was confused, and was scoping Q306 and Q307 when I wanted D306 and D306 (which are actually transistors used as diode in the schematic. )

So I am going to do some more probing around.

At least I am getting changing waveforms when the scope under test misbehaves. Now I should be able to track this down, I hope.

[FlyingHacker]

Well, I have been looking at this guy periodically. Still jot having any luck.

I replaced all electrolytics. This got all the LV power supplies to good values and ripples.

I was hunting a red herring on the horizontal, trigger, and vertical boards. I misunderstood why a waveform looked the way it did. I learned a lot about scopes exploring this, and am now pretty sure the problem is in the HV side, as I originally thought...

I replaced Q202, which I really thought was going to fix it, but the problem persists.

I checked the diodes, including the Zener ZD202, and they test OK. ZD202 seems to break down right at the 9.1V. However, when I put a meter on ZD202 in action I read close to 2kV on either side of it. I do not notice a difference. Should these two side of ZD202 read the same?

Of course I am reading this with an old Simpson 260 on the 5000V scale, which puts the reading way over to the left in the low accuracy area. I do not have a HV probe for any of my DMMs. So until I get one the Simpson is all I have. I also do not know how accurate it is on that scale. So it may not actually be right around 2kV. By the circuit design it should be about 1.6kV or less.

The only other thing I can think would be the three HV capacitors. These are all rated for just 1.6kV (apparently a known poor choice on Heath's part, and an issue if the circuit really is at 2kV). I get reasonable readings at the two diodes in the voltage multiplier circuit. And the display only fails at certain frequencies of input. At these frequency ranges the trigger is working, the blanking circuit on the LV side of the optocoupler IC is working, the horizontal and vertical deflection are all working.

Anyone know of a cheap source for these HV .1uF caps? Ideally 2kV. These guys seem pricey, and I hate to waste any more money of this thing, but it has become a bit of a Moby Dick obsession to get going.

I don't have any way to measure the HV side of the optocoupler. I need to get a HV probe for the scope.

I took C201 out and it measures OK.
All the resistors on the HV side read right near their ratings.

I am not sure what ZD202 is supposed to do, if someone could explain that this might help.

Any help would be very much appreciated.

Thanks.

[FlyingHacker]

I found either these:

http://www.mouser.com/ProductDetail/Cornell-Dubilier-CDE/942C20P1K-F/?qs=sGAEpiMZZMukHu%252bjC5l7YerOgAzP4CWkBluaUkSdZA8%3d

Or these:

http://www.ebay.com/itm/New-7x-0-1uF-2000V-20-Metallized-Polyester-Film-Capacitors-1uF-2KV-DC-/140590893448?hash=item20bbdec988:m:mowZQvyrA0zNS2WSDEQudJA

Open to opinions on those, and any other cheaper options.

[tautech]

ZD202 is to set a basic voltage differential between the cathode and the sucessive grids that is then further adjusted by Focus and Intensity and their various resistive dividers. Grids will be more positive than the cathode.

Many other scopes also have a PDA (Post Deflection Acceleration) source often +10kV+ to further accelerate the electron beam. Yep there are 2 +'s in that, little test for you. 

AS for the HV caps, Aliexpress or eBay will be a cheaper source, even 10kV caps are well priced.
104, 0.1 uF is a high value for an HV cap and you may find 473, 0.047 uF easier to get and parallel a pair.

[FlyingHacker]

Thanks.

Do you think the caps are the more likely culprit to fail at specific frequencies? Or should I be looking at the optocoupler? thanks.

[tautech]

Thanks.

Do you think the caps are the more likely culprit to fail at specific frequencies? Or should I be looking at the optocoupler? thanks.

Honestly, who knows  it might be a design fault. 
Or originating from somewhere else within the scope.

Have you taced the signal when it fails to be displayed throughout its path to the CRT deflection plates?
It's certainly a brain teaser type of fault. 

[FlyingHacker]

Thanks.

Do you think the caps are the more likely culprit to fail at specific frequencies? Or should I be looking at the optocoupler? thanks.

Honestly, who knows  it might be a design fault. 
Or originating from somewhere else within the scope.

Have you taced the signal when it fails to be displayed throughout its path to the CRT deflection plates?
It's certainly a brain teaser type of fault. 

I traced it to the output from the various board. In other words it is OK at the output transistors for the horizontal and vertical deflection. I traced the blanking waveform to the LV side of the optocoupler.

I guess I should follow the wires to the actual attachment point on the CRT. It would least be worth a look. Do you think a physical issue, like a bad solder joint or loose connector could fail at various frequency ranges?

I just got a Fluke 80k-6 probe, like new for $29 off eBay. For this the HV scope probe would be more useful, but that was such a good deal I couldn't pass it up.

[tautech]

Thanks.

Do you think the caps are the more likely culprit to fail at specific frequencies? Or should I be looking at the optocoupler? thanks.

Honestly, who knows  it might be a design fault. 
Or originating from somewhere else within the scope.

Have you taced the signal when it fails to be displayed throughout its path to the CRT deflection plates?
It's certainly a brain teaser type of fault. 

I traced it to the output from the various board. In other words it is OK at the output transistors for the horizontal and vertical deflection. I traced the blanking waveform to the LV side of the optocoupler.

I guess I should follow the wires to the actual attachment point on the CRT. It would least be worth a look. Do you think a physical issue, like a bad solder joint or loose connector could fail at various frequency ranges?

I just got a Fluke 80k-6 probe, like new for $29 off eBay. For this the HV scope probe would be more useful, but that was such a good deal I couldn't pass it up.

Very unlikely IMHO.

[FlyingHacker]

That was my thought.

The horizontal and vertical seem to work fine at the connection to the tube.

I think I will pull the optocoupler and try to test it at different frequencies (at a lower voltage of course). It is the cheaper part...

[FlyingHacker]

I replaced Q203 just because I had more spares. The frequency at which the issue occurred changed slightly, but that is all.

I ordered another optocoupler, just because it was only $3. 

Here is a linear frequency response (image) of the current optocoupler with a range from 20Hz to 200 kHz. Note that the problem occurs in the neighborhood of 70-140Hz depending possibly upon temperature.

Nothing is too remarkable here. This is feeding 15V into the output side of the optocoupler through a 1k resistor. The input is a sweep generator from 20Hz to 200kHz. I set the input to roughly match the voltage swing on the scope at the emitter of Q309. The input side of the optocoupler is almost linear at this frequency range through a 220ohm resistor. I suppose it is normal to expect some roll off. I have no idea what this looks like at 1.5-2kV.

I did notice some variance in the phase shift in the output wave vs. the input. Possibly this is an issue.

I guess I try the new optocoupler when it shows up, and if that doesn't work get some HV caps. If that still doesn't work then maybe defenestrate this thing.

In all honestly I will likely end up using this tube and basic control boards to create a scope clock or some other hacked display, like a game with an Arduino. A 5MHz scope is decent for simple logic stuff, but not all that useful.

I probably should have given up on this long ago. Some guys on the TekScopes mailing list said throw it away! I just couldn't bring myself to do it. I have learned a lot about how scopes work, since this circuit is so simple compared to a Tek scope.

[GreyWoolfe]

A high voltage probe really isn't that expensive- http://www.ebay.com/itm/High-Voltage-Probe-80K-6-6kV-PK-Max-Fluke-983T-bb4-/371460379690?hash=item567cc3942a:g:9G0AAOSwhcJWGIqk
For a little more, you can get a Fluke 27/FM meter with it- http://www.ebay.com/itm/TEST-EQUIPMENT-FLUKE-27-FM-MULTIMETER-WITH-HIGH-VOLTAGE-PROBE-AND-CASE-/191727208894?hash=item2ca3d509be:g:79oAAOSw9N1VpXay
You can never have enough DMMs.

[FlyingHacker]

A high voltage probe really isn't that expensive- http://www.ebay.com/itm/High-Voltage-Probe-80K-6-6kV-PK-Max-Fluke-983T-bb4-/371460379690?hash=item567cc3942a:g:9G0AAOSwhcJWGIqk
For a little more, you can get a Fluke 27/FM meter with it- http://www.ebay.com/itm/TEST-EQUIPMENT-FLUKE-27-FM-MULTIMETER-WITH-HIGH-VOLTAGE-PROBE-AND-CASE-/191727208894?hash=item2ca3d509be:g:79oAAOSw9N1VpXay
You can never have enough DMMs.

Yeah, in a posting above I mentioned I just ordered the Fluke 80k-6. Got a good deal at $29 for like new.

I have way too many meters, and keep buying more. Triplett 631 combination VOM/VTVM is my latest acquisition. None of the digital ones had a HV probe because I didn't really need one until now. But I am sure it will come in handy.

I do need to grab one of the cheaper HV 100x scope probes, though. Methinks that would be the most useful thing for debugging this issue.

[FlyingHacker]

FYI... Here is a PDF of the docs, which makes them much easier to read:

http://meanmutha.com/BellHowellScopeManual.pdf

[FlyingHacker]

Grrrr....

Well, I replaced the optocoupler... No change.

I did get the Fluke HV probe. It seems the HV is actually sitting at about -1950V. So the Simpson was correct.

It seems this is because the HV tap from the transformer (before the doubler circuit) is about 940V.

Of course the HV caps are clearly marked 1600VDC!

I know I have spent way to much time and money on this silly old student scope, but it really has become a challenge at this point,

I am trying to source some appropriate caps. I low balled some 3kV caps on eBay... Will see if that goes through.

Any other thoughts.

Thanks.

[FlyingHacker]

I have really turned into Captain Ahab here...

I ordered three new caps for the HV section. All are rated 3kV.

My only other thought is that this had a MCT2E optocoupler in it, rather than the TIL115 listed in the docs. Of course I replaced it with another MCT2E, because it was cheaper. There is also a 1K resistor bodged from Pin 1 of that optocoupler to ground (presumably a voltage divider with R202). I tried disconnecting it, but that did not help.