5

A small ESD spark is able to find its way across my PCB at a relatively low voltage (low kV). It is not obvious which path the spark is taking and the sparks only jump when the board is sealed up inside an assembly so it is not possible to visually see it happen.

Are there any tricks for getting the spark path to leave some sort of trail so after the event we can look at the board and figure out which gaps it is jumping?

I'm thinking there might be some kind of powder or film that discolors or fuses when a spark passes though it, leaving a visual witness mark of the sparks passage.

Any ideas?

winny
  • 13,064
  • 6
  • 46
  • 63
bigjosh
  • 9,888
  • 29
  • 48
  • 2
    In some sense, it doesn't really matter where the arc is coming from. If this is happening on your board then you've probably violated one or more design rules for HV applications. Personally, I'd review those design rules and compare them with what you've done. Fix everything you've done wrong and the problem should just go away. – J... Dec 24 '18 at 12:31
  • @J... "Fix everything you've done wrong and the problem should just go away." It is not always practical to throw away all your inventory and stop everything for a few months while you produce a new design every time you find a new problem! And being the flawed human that I am, it is likely that the new design will have even more things wrong with it than the existing one that has already been few a few refinement cycles! – bigjosh Dec 24 '18 at 19:23
  • The point is that if you've gotten yourself into this situation then there are probably other things you've overlooked as well. Presumably you're not discovering this *after* you've completed prototyping? Now is the best time for a design review - before this thing gets ramped into production, the tooling is finalized, and you're really in trouble. – J... Dec 24 '18 at 20:47
  • And sometimes all you can do is cut your losses... Poor design, poor review will always result in poor product. I had to push for a design to be completely scrapped due to poor layout. Yes managers didn't like it,they got vitriolic BUT a fast turnaround is sometimes all you can do. Flash test a bare board. Review your grounding scheme, identify all HV separation, re-assess the needed gap. Measure at Gerber level the separation. Consider conformal coating *if* external flashover. –  Dec 25 '18 at 10:59
  • The problem was related to changes in the assembly process. A mounting screw was omitted and another had a plastic washer added. The combination allowed a large polarization voltage to build up on a big metal part, and the (now) empty mounting hole provided a path for the spark to jump directly to a trace. Switching the plastic washer to a metal one cured the issue. Because of the way the product is assembled, this was VERY hard to track down. I am leaving this question open because having a way to find that spark would have saved hours. I'm going to try something with 35mm film next time! – bigjosh Feb 10 '19 at 21:07

2 Answers2

5

Do you have a blank board, no components? That's how we usually do it.

Also, what is the source of the kilovolts causing the sparking? Is this your calibrated pulser during an EMC test? Or HV crap coming into the board from long comm lines? Or a 50KV environment such as industrial conveyor belts and rollers?

With a DC HV supply, apply slowly-increasing kilovolts (with current limiter) to the wires soldered to the suspected traces, and watch the obvious visible sparks. Less ideal than a regulated DCKV supply is one of those BD-10 hand-held Tesla coils.

Even try this: figure out how to detect any sparking at all (such as a de-tuned AM radio nearby.) Apply your kilovolts to the pcb to create sparking at a mysterious location. Now, for each suspected sparking location on the board, cover it with a blob of silicone caulk. Assemble everything and apply the kilovolts again. If the same sparking is still there, then do it again, another blob of caulk at another spot. Eventually you'll hit the offending location.

I've seen pcbs with proper HV-holdoff spacing around certain traces, but where nobody checked for HV standoff going vertically through the board. At one spot the 2KV PMT supply trace went right over a piece of internal groundplane. Multilayer board, so very thin epoxy-glass between those layers. After a few years of operation, the HV chewed right through the pcb, creating an unwanted carbon resistor. See Cary Spectrometer PMT shorting.

wbeaty
  • 10,761
  • 23
  • 39
  • 1
    Testing an empty board should always be the first step, of course, but it's also quite likely that the arc is starting from one of the components, either a sharp edge of a pin or leg, a solder spike, etc. A high sensitivity camera works well testing in complete darkness - a long exposure during operation can produce an image of any arc activity. – J... Dec 24 '18 at 12:28
  • Also, +1 for the Cary link... our 6000 feels like it needs service about every six weeks. I'm going to have a look at that board next time I see it. – J... Dec 24 '18 at 12:42
  • wires soldered to the suspected traces The problem is even knowing where to start since I don't even know where the sparks are making it from the housing to the PCB. The jumps only happen when the unit is assembled and sealed up, and no way to see inside when it is happening. Was hoping for a trick to speed up the search! There must be something that will record some kind of change when a spark passes though it? – bigjosh Dec 24 '18 at 18:50
  • @bigjosh white paint turns black for power-arcs (using your 30watt KV supply to a dead pcb.) Or this: On a dead board on the bench, first apply KV and detect sparking (with flashing neon bulb to ground, etc.) Then, seal possible spots one at a time with silicone caulk, apply KV each time, until it stops, or until it takes much higher KV. Also, why was KV present? EMC test? Some kind of mysterious failure in the field? (Perhaps add to main description.) I've seen lots of 20KV stuff from conveyor systems. – wbeaty Dec 25 '18 at 09:40
  • @J... we service only inhouse, draw schematics etc., rebuild charred NMR probes and kilowatt pulse amps. I see design errors all the time. If we sent units out, the same design flaw would just kill it again and again. Heh, a nasty trick is to use an electrolytic cap for supply timing, then put it against a boiling hot to220 regulator package. Your product will mysteriously die in 5-10 years, guaranteed. See the BLAX300 elba-artesyn story, the Eppendorf centrifuge story: http://staff.washington.edu/wbeaty/chemtrik.html http://staff.washington.edu/wbeaty/chem-eppndrf.html – wbeaty Dec 25 '18 at 09:48
2

Air gaps to sharp points are 1kv/mm and flat surfaces are 3kV/mm, assuming clean air low humidity.

Add more gaps for margin. Exposed 5mm LEDs were once notorious for failing ESD tests in keyboards due to the small gap from case to leads with plastic coupling capacitance.

Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182
  • All true - but currently looking for some and diagnosing and debugging advice rather than design advice! Sometimes better to figure out why your car won't start and fix it rather than to buy (or design!) a new car! :) – bigjosh Dec 24 '18 at 19:27
  • Can't you measure the gaps? you can create some magnet wire twisted pair current loops and bring out to a scope to find the ESD path. Is it destructive or not? You can current limit the probe with 1M if it is. Otherwise show photos – Tony Stewart EE75 Dec 24 '18 at 20:10
  • "Can't you measure the gaps?" Which gaps? So far we can not definitively figure out where the spark is jumping from the enclosure to the PCB. – bigjosh Dec 25 '18 at 02:45
  • YOu haven't given much details on the test or the geometry of dielectrics and conductors so if you can't follow my instructions, you need someone who can. – Tony Stewart EE75 Dec 25 '18 at 03:08
  • "You haven't given much details on the test or the geometry" I am not looking for free ESD consulting here! :) Instead I am looking for general techniques to save time by recording the path of a spark when it is not possible to directly see it. If this exists, it would be a handy tool for debugging problems like the one I am facing now. – bigjosh Dec 26 '18 at 03:00
  • The debug method depends entirely on your missing details, There are many methods then I can choose which method would be best suited as there are too many to get into details. I have dont this for large enclosed power transformers from invisible Partial Discharge ( a tick tick effect much less than corona, which is is not ESD as you describe. SO please elaborate if you want – Tony Stewart EE75 Dec 26 '18 at 03:22