Repairing a scrapped Fluke/Philips PM3082 Oscilloscope

[dml]

Hello.

I'm new here. 50% n00b and 50% hobbyist since I was about 9. My day job is not and has never been EE. This is not a guide to repairing anything. This is a long post. End of disclaimers!


So a while back I picked up a couple of Philips scopes - a PM3392 and a PM3082. The 3392 was to replace the USB DSO I had been using for some hobby stuff, and don't really enjoy using all that much. I wanted a standalone analog unit. The 3082 was practically scrap - a donor unit for knobs and encoders / front panel stuff for the main scope.

The PM3392 had some front panel issues but generally worked ok. I lucked out on a second source for the broken panel elements and had it fully working before the 3082 arrived. So... I decided to try and *fix* the 3082, either to have a second scope or maybe hack it into something fun like a curve tracer or some other project.

The PM3082 'journey' is the the subject of this post And it was a journey. I'm writing this up in the hope it saves someone else from making a few mistakes. I'll point out quickly that the unit was sold for parts and claimed 'may have parts missing'. I agree that could mean anything - and indeed it did.

I link some images here crammed with annotations. This is mainly to save linking extra pics of the same stuff in some cases.

So after a quick inspection and a quick belching-flames-power-up test, it seemed like the tube would not light. No whine from the flyback but some activity in low voltage rails. The 5v rail had some life at least and I think the front panel had power (maybe a LED).

So I removed the PSU board. Not easy - there are nasty plastic catches behind the board which need carefully prized with a long thin screwdriver. Better still, a long thin insulated stick as strong as a screwdriver (good luck with that) because discharging all the caps in-situ first is a bit sketchy. I discharged the main offenders at the front first, lift it out then do the rest.

At this point I'll point out *one* of the blue HV caps near the front of the board always retains a nice 2KV charge. Discharge that before handling. The rest are either low voltage or always tested discharged. Still - discharge all of the HV & mains caps and the CRT with a suitable probe, every time.



Also that funny looking diode will test open with a normal DMM. it's a 14kv diode with a forward of 50V+. Don't replace it with a normal diode

I examined the unit - top side is all large thru-hole components. No SMT at all. This is misleading because the back of the board has a few small islands of SMT stuff. On a more careful inspection, I see two 'parts missing'. These turned out to be opamps.

Checking the schematic, these are LM358s (N1101, N1102) which provide drive levels for the fan, graticule bulb, trace rotation and ... the EHT convertor circuit - that's N1101. See pic of crusty gaps where chips should be:



I don't remember the order of the next two steps as it was a couple of weeks ago - but at some point I tried to power up the PSU outside the chassis, in an effort to check voltages. I think I had figured since there was no EHT to worry about, I could get away with it. But Philips was one step ahead of that - board does not like running without a load. The unit screeches with no output.

Checking the schematic, there are some confusing over/under voltage protection circuits including a crowbar across the +/-58V rail, on top of the voltage regulation feedback. It takes some time and headscratching to absorb whats going on. This PSU needs to be considered as an integral part of the scope and not some standalone module you can play with on its own.

Not all rails are sensed for protection - so I tried to fool it by placing dummy loads across a few filter capacitors on +5v, -12v, +12v which I figured were the ones needing the most care for the sake of ICs everywhere. I used some 12V halogen bulbs in series with some high wattage, low-ohm resistors. i.e. a precarious mess. This did bring the board to life. Checking the rails showed good voltages for most of them ...except the +/- 58V rails which drive the EHT convertor.

These rails were reading +/- 135v. This was my first really big mistake - I later noted that the filter caps on these rails are rated for 100v. I was intending on recapping anyway - but beware when trying to hack this unit to work out of the chassis.

So rather than try to find more dummy loads to make a bigger, more precarious mess, I abandoned that and went back to the faults I already knew about.

So I replaced the missing ICs. With chips I had handy (that might have been another mistake). This was made more tricky by the fact one of the pads was lifted and turned upside down. I remove the pad and made a new track from kynar wire, and raised the existing SMT resistor over it to avoid contact. I actually repeated this whole exercise twice, but I'll come to that folly later (once would have been enough).



I hadn't cleaned up the board at this point but it was dabbed with alcohol well enough to test (the 'bodge resistor' was not yet installed - that's later). So I restored the board, attached the EHT sucker to the CRT (which was a trial because the clip pin had fallen out of the tube).

Firing up the scope showed the CRT was back to life.

The display was very wobbly and super, super bright. Scorching, unhappy levels of bright. Something still wrong. And sometimes the vertical axis would just collapse to a line, at random.

Also the PSU was making strange intermittent whiney, clicky, chirpy sounds from one of the transformers. Not good.

I could control the trace and text brightness from the front panel, but the lowest setting was 'super bright' and one notch lower would turn it off completely. The highest setting was 'sunburn'.

I'll cut this part short(-er) because it took days of mucking about to make progress. I had got concerned that the PSU or the tube were swapped in from another machine as part of a previous donor/repair exercise and were 'unmatched'. However this theory didn't pan out. The model numbers tied up ok and checking components all over the board didn't find anything bad.

I took a thermal camera to the board (SeekThermal for mobile) and it showed several areas getting really hot. I wasn't too happy about that either.




Note: The main things which get warm/hot are the NTC anti-surge resistor, the whole snubber circuit on the 58v rails to the EHT convertor (which uses a tiny 100R resistor & ceramic cap), several other small resistors around the board, the switching transistors, a small transformer and most of the glass power diodes. Comparing temps with the working scope showed similar values in most areas so this seems 'normal'.

I wasn't about to repeat the dummy load thing, so I soldered wires to test points all over the board and reinstalled it. All the rails checked out ok including the +/-58v supplying the EHT convertor for the tube. Hmmm. Not finding anything obviously wrong.

Studying the schematic again, one of the opamps I had replaced (N1101) controls the EHT convertor's duty cycle by feedback from -2200v output to the tube, via a resistor divider. This had me wondering if I had upset that feedback circuit, overdriving the tube. Some difference in opamp specs, offset or whatever.

I decided to swap N1101 for a second IC in case the first was damaged or bad, or just different tolerances. No change. This was annoying because I had to redo the track & bridge trick and meant more heat on the board etc.

I soldered wires to test points around N1101 and noted the compared input voltages looked ok(-ish) @ 17v each but the output was very high, IIRC 27v or -27v. Seemed like the on-time of the EHT convertor was excessive, which might explain the tube being so bright and putting stress on the PSU via the 58v rails.

So, noting the resistor divider was a 31.6Meg, high-power resistor on the top of the board, and two SMT resistors on the bottom (121k+19k), I removed the 121k from the bottom and installed 151k (see 'bodge' resistor above). This did indeed reduce the CRT brightness to something sensible - similar to the PM3392 unit.

This felt like progress. But the vertical axis was still going, the PSU was still making unstable sounds and the text was wobbling and jumping randomly.

Notes: the schematic shows 149k for PM3082 and 121k for PM3392. I found 121k, which drove the CRT extra bright. 149k would probably have been about correct and would have matched the schematic. Maybe this PSU was swapped in after all, but not from a PM3392 because it is missing the 5v postregulator found in those units. However, this story is still open - 121k may have been correct after all, as I'll explain later.... BTW There are several other inconsistencies/revisions in the schematics, and the back of my PSU doesn't even match the schematic or layout for the EHT convertor in any of the service manuals i found!

At this point I realised 'probably out of my depth' but I kept going out persistence/stupidity etc.

I recapped the whole PSU board in the hope the chirping, ticking sounds would go away. I tested all the caps which came out. I'd say the exercise was not exactly necessary but I still feel better having done it. At least, it kept me busy. Some of the caps which came out read equal or a touch better in some readings than the ones going in (and they were carefully spec'd and sourced, not ebay!). Others were reading a bit less happy, but probably still not worth the cost & effort spent.

For reference, this is what came out. These are original caps from the factory btw - a match for the ones I found in the other scope.


The 400V/150uF mains caps gave such good readings I just put them back in, even though I had new ones!

And no, the sounds from the PSU didn't go away. Grrrrrr.

I swapped the NTC resistor (70'c temp!) for a new one @ 15R/4A, but same result. Same noises from PSU. Taking thermal camera to other scope showed similar result although rechecking the repaired PSU showed temps of some areas had lowered, now matching the PM3392.

I had already checked all of the semiconductors in or out of circuit and nothing bad. So probably an issue somewhere else in the scope...

So my next excellent mistake was to sacrifice the good scope for the bad one. Yes I took the PSU out of the fully working 3392 and stuck it in the 3082. Fired it up with a dummy load on the EHT (don't try that) and was met with fizzing but no other bad sounds or effects. I chickened out of this exercise very quickly and put the PSUs back in their respective homes. Firing up the PM3392 resulted in a small emergency, with beeps and magic smoke from the PSU. :-(  I powered it off very fast, reseated the board, tried again and this time it fired up ok. However the inductor has cooked off its insulation. I'll have to replace or rewind it. Lesson here - don't do that.



I still don't know what happened there but I figure probably not well seated in its socket. Or something.

Observation: These scopes were not made for DIY repairs.

Now the fun part! The magic sparkles.

I spent ages prodding, poking, shaking, twisting the unit while powered up to see if i could get the trace vertical to stop jumping and collapsing at random. This sometimes felt successful, like a mechanical or contact fault. But half the time it felt like teasing. I couldn't find a single spot on the board which would reliably fix or fail that issue. Hair dryer, hot air station, blowing on it - all failed.

In frustration, I re-flowed the back of the PSU, the output stage board and neck board. At first, this seemed to fix the issue. But it came back. I started prodding the neck board with a wooden stick and got the idea this area was involved. It's the one area which made sense and I hadn't inspected as much as the rest. I found a heatsink spring-soldered against 4 ICs on the back of this board. One of the ICs was missing thermal compound almost completely and there was a gap between it and the heatsink. So I added thermal compound and tried to bend the heatsink to be sprung against the IC again. In the process, I managed to get thermal compound on the SMT components nearby. With alcohol to the rescue, I managed to get this compound equally spread over ALL of the components, including under some of them. And removed compound from the other 3 ICs in the process. It just got better and better.

Eventually I washed the board, redid the compound and put it back. There's still some compound under a couple of ICs but the stuff is not conductive and I'll try again if/when i get the other remaining problems resolved.

Still the trace was jumping. At this point I was 99.9% towards giving up on this unit. I probably should have long before now. It had got dark outside and I was seriously hungry, ready to make a late dinner. But I just sat and watched the glow from the back of the tube through a hole in the neckboard, remembering my parents first TV with valves glowing through the back. Yes that's what kind of a sad state I was in from this  repair exercise.

Then I notice something really weird. I saw a tiny yellow light come on. And go off, and come on again. It looked a lot like a SMT led, but dim. This was news because this scope seemed a bit old to have tiny diagnostic LEDs, especially near things that get hot. So I got a magnifying lens and sat watching it.

I grabbed my mobile and recorded what I could through that lens:

https://www.dropbox.com/s/1tqdj9jd763xgo8/pm3082-neckboard-fault.mp4?dl=1

3 frames cut from that video:


This capture was the last time I saw the 'light'. The first time I noticed this it was brighter and frequent. By the time I recorded it, getting dim and not happening often.


Confusing. Magic lights on a surface mount board... ?!?

Removing the board and inspecting closely, I find what looks like a burned via - converted into a resistor. I couldn't tell at first if it was shorting inside the board, or just disconnecting. The via itself was not shiny but dark. Holding the board against a bright light I could see it could be shortcut with a bodge wire, so that's what I did:



And guess what? The angry PSU chirping went away. The vertical axis isn't collapsing to a fuzzy line. The text stopped jumping. CH1 shows a trace!

Lesson here - if you're going to give up on a repair, don't give up until its dark.



However it's not all good, yet.

On power up, the PSU does still emit some ticks and clicky noises at first. It goes away after 30 seconds or so. Something still up there and I think it has to do with incorrect current draw by other parts of the scope. Maybe still concerning the EHT/intensity problems which still don't have a good explanation. Perhaps I installed two equally defective LM358s picked from ebay long ago, even if the trace rotation & graticule bulb circuits seem to work ok via the same ICs. Or maybe a winding short a the transformer causes multiple problems - but no signs of that from the outside and rail voltages all check out now.

The trigger doesn't seem quite right. I can get it to work by fiddling with the menus and adjusting the trigger level - but it's supposed to show a triggered signal using just the calibration square wave hookup and the AUTOSET button. But it does not. Flipping it from EDGE mode to TV  mode makes it lock also - but by default, EDGE mode doesn't trigger without messing with menus & controls. It also seems like the trigger level isn't aligned with the signal position/amplitude on the display. Maybe due to the calibration aborting. Or user error. Another journey there probably.

Third... the unit fails auto-calibration. Something to do with display position something-or-other. While I don't know what that is, I notice the text boxes don't quite fit on the display horizontally and I'm wondering if that has to do with the 151k resistor bodge I applied to get the brightness in a sensible range - i.e. correcting the problem at the wrong location, in the EHT drive. Perhaps the 121k resistor was good all along, and the fault is elsewhere. The filament current maybe (although its fed *direct* from a EHT secondary), or some Z-output driving area. I just don't know. Suggestions welcome.


So that's the 'summary' of my journey with this, so far. I wrote this up not because I'm desperate to get help fixing it, but because it records a few fun mistakes/hazards and some weird problems that might help someone else trying to investigate a similar unit. And maybe someone out there *does* know what's going on with the CRT intensity going 10x overdrive without bodging the sense divider.

I'll come come back to this and continue the 'repair' when I get my patience back, but probably after a break

[dml]

Quick update on this.

I re-tested the bodge resistor using a resistor+trimpot around the same range 120k-160k.

Adjusting this affects intensity, X+Y display scaling and focus all at once. So I now have the menus fitting inside the display. Although the scaling is still guaranteed to be off spec now after messing with that.

Still the value which happens to get the scaling approximately correct and still allows the trace intensity to be reduced properly lies somewhere between 121k - 151k. Definitely above the fixed 121k which was installed. Interesting, but still not explained.

Since this divider is only a voltage comparator preset for N1101 and the output from this device just sets a duty threshold for the flyback circuit which follows, I doubt replacing N1101 with some other (assured genuine) part will help at all. If it compares at all and is already adjustable, it probably compares 'enough'.

So I still don't know whats going on here. It's closer, but still not right. PSU still ticks sometimes.

Something that does come to mind is the response time of the N1101. If this has somehow changed, it might cause some weird feedback problems and add to the confusion. I could investigate that.

Also, the auto-calibrate feature won't activate unless this EHT voltage preset is in a fairly narrow range - I just figure that out. It complains about warming up (and ignores the 2-second button press option) if this is too far out. It seems to care more about the output of this N1101 than temperature or time. That's a bit strange.

However despite all this the calibration still fails with error #31 and I'm still not sure exactly what that error represents. It appears auto-calibrate always takes the same amount of time but will 'sleep' (blank screen, or dot) on all sub-tests which follow the first failed test. Depending on the preset voltage above, it gets a varying distance into these tests but never beyond about 70% progress without hitting the 'sleep' state and ending with error #31.

I'll study that part of the service manual again later and see if anything makes more sense than it did last time.

[EDIT]

The 'ideal' value for this resistor, for display size/intensity seems to be 139k, just over half way between the two values in the schematics. Not sure what that means - but perhaps the 31Meg resistor has drifted. It does seem to run a bit warm. I don't have a reliable way to test that just now without adding more uncertainty but i might try to check it another time.

[dml]

I figured out why the PM3392 power supply ended up with a cooked inductor after restoring it to the chassis after the swap test. This became more clear from the schematic.

I accidentally plugged the fan & battery-backup headers in each others' sockets.   

They both have identical sockets side by side, except the polarity is swapped on each. So the fan circuit was shorted and this took out the drive transistor and damaged the inductor. The battery circuit wasn't affected, protected by a pair of diodes.

Fortunately the fan circuit is pretty isolated so nothing else was damaged. A capacitor saw reverse polarity for a few seconds but that's about it. Might have to swap that now just in case.

So the PM3392 is again repaired and back on the mini bench. 

[tautech]

Thanks for the wonderful write up, an excellent read and for sure it will help somebody. 

I'll drop a couple of links.
Operators maunal
http://assets.fluke.com/manuals/PM3082__umeng0000.pdf

Service manual
Trusted file repository, login required.
http://www.ebaman.com/index.php/remository/ELECTRONICS/Test-Equipment/Fluke-Philips/Scopes/Fluke-PM3082-PM3084-PM3092-PM3094-Service-Manual/

[dml]

Thanks for the wonderful write up, an excellent read and for sure it will help somebody. 

Thanks! It was fun. I'll certainly come back to the 3082 after a bit more digging. And there might be some other repairs on the way in the not too distant future.