high voltage equipment repairs special considerations/stories?

[coppercone2]

So while I don't want to work on the miller dynasty welding machine any more, it got me thinking about high voltage/high current systems (like IGBT drivers). I got the feeling that these sort of repairs are generally different.

My experience is mostly limited to things that have a normal fuse and transformer, meaning the fault damage/currents are not especially high.

I am wondering if anyone has experience working on repairing damaged equipment that is not really limited well and has a mixture of microelectronics and power electronics.

I am wondering about stuff like components that measure properly with DMM but fail under high voltage or current (isolation related faults). Does anyone have any stories about stuff that would just not get fixed with conventional troubleshooting techniques? I see a little bit of this on mr.carlsons lab. Miller welding equipment seems particularly hard to fix (at least forum comments generally give these machines a very low repair ability score).


Do people tend to use specialized test equipment for these sort of tests, i.e. high voltage/current ohmmeter (the desire for a bridge comes to mind)?

I know its alot different because I felt unsafe powering my machine up without a leather apron, gloves, arc glasses and ear muffs lol. I kept thinking that some capacitor is gonna blow up on me. There is alot of information on high power industrial, but the area where PCB electronics mix with bricks seems different, none of the tools I typically use seem appropriate/convenient. I kinda feel like that everything is treated like a fuse, i.e. if it can be mechanically removed and replaced it is, and stuff that is soldered is treated like a fried ASIC to us. I am guessing no one takes 'component level' repair seriously with this kind of equipment as of now, and that all falls under 'not economically viable'?

[abquke]

I used to work on repairing/refurbishing equipment that was reasonably high voltage, though not high power. 5-15kV or so.There isn't fundamentally a whole lot different from working on any other electronics except that one can't 'just probe' voltages.

[strawberry]

screwdriver is nice HV test instrument for up to 40kV (low energy up to 100W source)
voltage estimated by distance ark jumps over
vacuum tube sets are more robust and can easily be tested HV against chassis return
(usually 8..18kV was for B&W and for color more)

about right to repair
old Grundig or Telefunken sets were equipped with built in special HV tube discharge lead for repairs
and many easy accessible and marked adjustment pots (usually color/pictogram marked and in rows on back of the chassis so you dont have to search for some codes in some random locations on PCB). Marked test voltages on PCB.

Apple and other modern devices cant be opened not breaking them. They say it is done for our safety... It was back then 30kV baby 

[Kleinstein]

I had an example that worked OK with the high voltge, but tested bad with a more normal meter:  it was a contact to aluminum with a bit too much oxide to give a good contact. A the high voltage it did not really mater, though it created some "noise". So it was good to fix anyway.

Most failures are still the same. With mechanical larger parts bad solder joints may be a bit more common. Also just collected dust can be a problem with a relatively easy fix.

[Smith]

I occasionally work on power supplies up to 30kV. One consideration is make sure your measurement device is up to the max output voltage + ~50%. I had some systems with a bad feedback circuit, that cost me some really expensive resistor networks for expensive multi meters. Also If there is any one else present I make sure they know what I am working on. Older HV cabling tends to dry out, making it prone to bend back out of inputs.

My power supplies usually go up to a few mill amps. One does eventually shuts down >32mA with some shady circuit at >2kV. I do NOT let anyone use it but myself. I've been zapped so many times by voltages over 10kV, never had more issues that a sore arm and a bruised self esteem.

I don't use much expensive high voltage cabling and connectors. I've used standard BNC cables up to 8kV, and default speaker wire >10kV. If the distance to other conductors and your fingers is big enough, it is NOT a problem.

Where I live it isn't a problem. I've been to France, and even the safety instances believe any voltage above 1kV will kill you, regardless of how low the current is. People freak out if you come near anything with high voltages. I knew this, so one time I measured a few kV from a rather large power supply with a paperclip as they did not have the right connector, those people looked horrified.

[coppercone2]

I had an example that worked OK with the high voltge, but tested bad with a more normal meter:  it was a contact to aluminum with a bit too much oxide to give a good contact. A the high voltage it did not really mater, though it created some "noise". So it was good to fix anyway.

Most failures are still the same. With mechanical larger parts bad solder joints may be a bit more common. Also just collected dust can be a problem with a relatively easy fix.

that sounds like it could easily drive me crazy. Works ONLY at high voltages. But I do think I heard of this one before or at least might anticipate it, since its similar to relay wetting and dry joint. Have you ever seen this in reverse that it works only in low voltages? This one fortunately follows the adage "put some more juice through the circuit'.

I left those welding machine PCB in the ultrasonic for 30 min last night too, I will try one more run with fresh diodes before I mothball it. I thought maybe there was something arcing in the connectors or something, like in the housings, etc. Peeled off some discolored varnish too.


Did anyone here ever see conformal coating go bad and conduct electricity? The welder I failed to repair had some conformal coating that was turning mustard color. I just want to rule that out as messing with the low voltage 30V/20/20/20 flyback supply before it goes into indiana jones warehouse.. the only theory I had is that maybe there is a duty cycle control for the flyback thats being driven to a higher duty cycle by leakage or something

[coppercone2]

Well, not to disapoint anyone but I managed to get my welding machine from smoking diodes. I think it was a bad capacitor or inrush. When I hooked it up through a 20A variac @110V (which I guess is not that much of a inrush limiter), and replaced a very mechanically weak capacitor that read 60Kohms in a damaged state (lead came off while removing from PCB), the welder no longer smokes the diodes. I raised the voltage to 118V and nothing happened. I also bathed the boards for a good 25 minutes in the ultrasonic.

Pending thermal camera study and further analysis to determine the loss of function problem.. but at least its stable enough that I am not scared to measure it.

[jonpaul]

A few HV mémoires...

1950s....ozone poisoning from Tesla coil....

1965....College rotating machines laboratoire... 208 V delta three phase 5..10 hp synchronous motor-generator...

A student closed a synchronize circuit by mistake with  180 degree out of phase...short circuit...WOW, horrible  NOISE and .fire and sparks in the brushes and slip rings...

I jumped to the main lab breaker, (480V 200A(?) hit the RED EMERGENCY disconnect.
 BAM...breaker opens, saved the day.  Lab prof gave ma an A.

1982...100 W medical gas laser, 20 kv arcs and corona...

1983...55 kV 1 kw arc lamp cinema ignitor... start-up transients blow up DVM and start fire in MOV  line protection, that caught carpet on fire

1990s  100 nS Xénon lamp pulse 30 kV ignition crashes microcontroller in medical illuminators

1990s 30 kV lab supply, 10 ma...shorted output...blew up regulator...

Bon courage,


Jon

[abquke]

A tool I used to use was a "bat detector" kit. It basically mixes ultrasonic down to audio levels so kiddos can listen to bat squeaks, but it was great for listening for hisses and pops characteristic of coronas.

[coppercone2]

and I found a shorted TVS diode taking out a rail in the display section now that its stable.

I measured the voltage and its 400VDC on the DC bus running on 120V.

Nothing seems open at these voltages, its all shorted.

Needle insulated probe tips from fluke have been essential for this work

Speaking of hiss, when it turns on now, despite not frying any parts, it makes a modulated hiss noise

[coppercone2]

And I will write down this repair tip/guide for this kind of problem with switchmode weld/plasma systems that need to have their control electronics stabilized.

before hand:
1) Use rubberized abrasives to clean arc damage till light brown PCB color. If stuff is charred badly use a dremel with a ball nose burr to route away sections of PCB. Stick tape on it and refill with DP280 epoxy and fiberglass powder, I cut trenches near sooty areas also and filled with glue. Repair any damage you see as you go along fully using PACE techniques, don't leave it for later because I would have thought that circuit is irrelevant.
2) heavy ultrasonic cleaning if there was soot damage on the board. Unless there is something unrepeatable there like a MCU, its likely the ultrasonic will help you find a problem or get rid of a intermittent. Bake in regulated oven after rinse and air-dry (I use 65C for 1 hour). Probobly putting it under a piece of black material outside in the hot sun will work also. 
3) use THF to get rid of varnish in obvious repair areas
4) after a few attempts with pure mains to maybe clear faults or fully damage parts, stick it on a large variac to maybe reduce inrush a little and operate the unit at different voltage levels within the manual. If you are not welding I am not sure how much difference the current increase from low voltage makes compared to dealing with standard line voltage, so I chose 110V as its the lowest operating voltage. This is just for equipment safety not your own, so its still just as hazardous
5) probe slowly and carefully with needle probe shrouds to avoid damage to circuits. Take your time because one mistake can cost you an entire circuit path. I brought a chair into the garage and got extra lighting to be certain.
6) since the system is misbehaving, use safety precautions. This means brand new probes, a fluke meter, and make sure all the stuff is clean, inserted properly, and you have a good work area.

Stabilizing:
1) start at the big high power/current parts with thermal optics. If you do not find a problem there
2) you can assume the control regulation loop is being fucked with some how by something, like a leakage.
3) test the small stuff near the big components, no matter how innocuous it is. In this case it was a presumably a 100nF capacitor making a diode on the other side of a transformer light up that was physically 8 inches away. However, you could assume the 100nF capacitor was doing something related to the IGBT/MOSFET/Flyback circuit startup or regulation. I.e anything within range of a power electronics part could be causing all sorts of mayhem
4) use DMM min/max peak mode to get sanity check if you don't have a scope meter
5) you can try over rating components to get a feel of what is going on, like I substited the broke 1W 22V diodes with 5W 20V diodes. Sure the IGBT is not working properly at 18V but you are changing power levels to more manageable levels, maybe. You have to switch it out later but it might give you some opportunity to see something. Since you don't plan on welding, it can't hurt.
5) still not sure what happens if you ground isolated IGBT rails.

I am surprised I managed to get this far without any tools more then a DMM and thermal cam. I have a feeling if I was trying to figure out the switching pattern with a scope meter it might be counter productive actually, so use common sense.

Hopefully people will resurrect some millers

[fzabkar]

https://www.acs.org.au/content/dam/acs/acs-publications/ACS-ebook-2017_A-Vision-Splendid_The-History-of-Australian-Computing.pdf

The next step was finding someone to build SILLIAC. The Australian subsidiary of British company Standard Telephones and Cables (STC) was contracted to do the chassis and wiring, but would not be responsible for the project as a whole. The senior engineer was Brian Swire, hired out of the Government Aeronautical Research Labs in Melbourne.

Swire was tragically killed in a high-voltage accident in the University of Sydney’s School of Physics Plasma Lab in 1964, one week before Adolph Basser died.

[Neomys Sapiens]

You could look at Mil-HDBK-454B: Guideline 45 Corona and Electrical Breakdown Prevention
for some hints. It is surprising how much of it does have repair relevance. For example, don't bend wires to sharper angles then in the original wiring, avoid lead ends sticking out of solder joints, etc.

[coppercone2]

Well Ok, I am getting no where. I found that a LF347 is running alot hotter then any of the other LF347. Since there was a shorted protection diode somewhere, I wonder if that rail was being pulled low by the shorted ESD diode could have damaged the LF347. I think I replaced it, but I also think I got that LF347 on eBay, so I will replace all 3 from proper supply place. a CD4013 was also running pretty hot (140F), not sure if this is normal. Anyway I am gonna replace all those hot chips, and also replace some resistors that were getting really hot (the 150K orange ones). I think maybe there is something going on with their temperature coefficient but its a shot in the dark. I am also waiting for 1 level shifter to replace that was on the PCB with the damaged 15V rail diode.

I am just going for these guys because they all run on 15V and since I found a short on the 15V rail, maybe its a good measure to maintain sanity.

I did notice when I replaced the shorted diode (and all TVS diodes) and the OpAMPS on that PCB, the power supply noise changed. Before it sounded like it was an musical device now it sounds like its infrequently changing pitch that sounds like its more related to some kind of hiccup or whatever. The 7815 is running hot too, not that I expect it not to (the voltage is correct), but maybe its an indicator of overloaded parts.

Since the 150K resistors were screaming hot at 400V (single phase), I think the split phase (800V operation) would potentially damage them.. maybe its a sign of damage. The god damn conformal coating is so annoying I just snip em off and replace em rather then screw around trying to measure at high voltage since they get damaged during removal unless  I go through too much bull shit.

I can't trust any of the hot parts based on what I saw on the HV pcb with that stupid capacitor causing all that ruckus

peek a boo, you transient fuck you, i see you! To anyone who actually cares, the problem is it shows the 5-over temp error and shows current and voltage as kind of high. The Vref for the control card is spot on. Hopefully its just dickered cheap parts on the 15V rail.

And I noticed when I cleaned up sharpie marks off one of the LF347's that I marked as being hot, the text rubbed off a little. To me thats a sign of poor parts (maybe counterfeit), so when I redid that PCB I might have installed a shoddy ebay part.

[coppercone2]

well Ill be, shorted 3906. 0.05V diode test 

[coppercone2]

Ok, that did nothing. WHen I replaced the transistor a logic gate got real hot and I found what appeared to be a overheating MOSFET switch, which I also replaced, but the thing is still stuck making that techno beat, displaying 5 and weird current/voltage readings.

gonna have to go through the signal paths when its turned on since nothing is getting excessively hot anymore 

holy crap I never had anything give me so much trouble or have spare parts that are so over priced! gonna start running boards off power supplies now because I don't know what the hell to go look for anymore.. nothing here looks too special, standard industrial control shit in a weird form factor

[jonpaul]

bonjour coppercone2,

1/ Fixing a blown SMPS is not easy as many sections interact and a fault in the switch transistor easily damages the driver, LV PS, feedback, etc.

2/ Even veterans can have such problems.

3/ We used to isolate the LV PS, driver and feedback and check each section with dummy loads, to isolate the faults.

Bon chance

Jon

PS: For real HV stuff (megavolts)

my old friend Greg LEYH = Lightning on Demand....
https://lod.org/

Detusches Museum in Munich Electric Hall
https://www.deutsches-museum.de/en/museum-island/exhibitions/electric-power/high-voltage-installation

Photo is 1200V transistor blown in Greg's 13 m Tesla Coil....

[coppercone2]

lol. the input rectifier was slightly cracked but that looks exploded.

Well I noticed that the display should say HEL P-5, but it says just -5. However, those segments work when displaying the stuck voltage and current.

I wonder if the MCU is corrupted or if there is a power draw issue with saying HELP for the logic assosiated with the display, maybe a sag.. I wonder if that power supply oscillation syncs up to the display . I am thinking the display board not showing the HELP message is maybe not a symptom of the inverter but the processor board. I gotta run it cold with fake inputs to see what it does, I think a blown GPIO pin might make it get stuck in 'safety hault' mode but I am not sure..

[coppercone2]

the noise change of the power supply coincides with the display changing states from -5 to voltage/current readings.

I wonder how that is possible, it should not draw a significant amount of power to switch seven segment displays. Any theories on how a atmega8 can be doing this?

I wonder if there is a shorted part somewhere that is loading the 5V rail hard and the little bit of extra juice used to light up the digits is making it go near the limit or something.

[jonpaul]

Copper, just realized you have a welder,

I am not xperienced in welders but your MCU question reminded me of a xenon lamps 750 W 60 kV ignition medical illuminators we designed in 1990, with industrial CPU control

Ignition transients coupled conducted and radiated into the CPU clocks and control lines.


Very careful shields, grounding and EMI filters plus transients absorbing Zeners achieved reliable operation

Just a thought

Jon

[coppercone2]

Hmm well, it looks like the intermittent fault with the part getting hot was smoke from the a 3 ohm resistor connected to a HV PMOS dip-4 package.

I let it run while it was making a high pitched noise and I watched the resistor turn colors and the gate die melted the plastic and it was starting to turn orange. I wonder if I could have zapped that with ESD or got a counterfeit ebay part or if there is a problem down the line, or if operating the with the shorted undiscovered 3906 transistor or the shorted some how damaged the PMOS device. I will replace it and see what happens. Maybe it was just a bad ebay part.

Replacing that PMOS was one of the first things I did , but there were many repairs after that, so something may have happened.


what does medicine need xenon light for?

[coppercone2]

got a new PCB. I  forgot to mention that before it broke I heard periodic clicking.. i.e. whine increases in volume then a click. I think that maybe the PCB in question that I repaired some arc damage on has deeper arc damage internal to the PCB. I see miller has a contingency for this, they refer to board layer seperation.. so I assume its possible that there was internal arcing. I thought maybe I was hearing a voltage level rising and then a spark gap effect or something. When I did my repair I took a little out with a ball mill and refilled with epoxy but I am imagining even if it did work, do I trust that kind of incomplete repair at 800V when I run it on split phase? The other board had isolated damage that I completely milled out because it happened in a mostly depopulated corner, so I trust that, but the inverter control had alot of char in a area that was densely populated... I started to get the feeling that trusting my repair there is risky. Maybe with X-rays or something you can be sure... but its just beyond economical repair for me (and the parts on the densely populated board are all HV mosfets)

I am imagining that board is a damaged car bumper and the other board is some damage to a mirror housing. Replace the bumper (mains connected 800V mosfet board) and repair the mirror (isolated 15V diodes). Repairing a totaled bumper is too hard for something that is going 'off road' IMO. Too much frustration from one board. Playing whack a mole with dip4 mosfets no thanks