Connect with us

PCB high-voltage meltdown

Discussion in 'Electronic Design' started by Winfield Hill, Oct 18, 2006.

Scroll to continue with content
  1. After experiencing a PCB high-voltage meltdown, I am
    driven to ask for advice from experienced folks here
    on s.e.d. s/n 19 of my new +/-1100-volt amplifier
    suffered a severe insulation breakdown between two
    BNC output-connector-mounting holes and pads, spaced
    only 0.10-inch apart. A carbon path is now clearly
    visible on the bottom of the PCB. The HV breakdown
    is so severe that after being exposed to 500V, 3mA
    pumped into the output can't raise the voltage more
    than 100 volts (implying a 33k short).

    It's possible the carbon pathway first began forming
    in an uncleaned solder-flux region between the two
    pads. The carbonization eats well into the interior
    of the PCB, obscuring evidence of the initial path.

    The guilty pc-mounting BNC connector (not my choice,
    BTW), with its two holes and pads, is not being used.
    The high-voltage hole is needed for an output wire.
    The guilty ground hole 0.1" next to it was also used,
    but the ground connection can be made elsewhere.

    My solution for now is to completely drill out the
    dangerous ground hole, leaving only surface conduction
    to a ground plane 0.1" away. (The rest of the board
    has healthy clearances for the up to 2.5kV voltages
    seen, but sadly I didn't give the PCB design of the
    output-connector region, with its 1.1kV potential,
    the same attention.)

    I wonder if we can count on about 0.1" of PCB
    insulation to maintain a low-leakage (over 500M)
    over time. Perhaps we should spray or coat this
    region of the PCB with a conformal-coating sealant,
    after a careful cleaning.

    We have made 160 of these rather complex amplifiers,
    and now we're seeking advice for a robust, but we
    hope not too painful, solution to this problem.
     
  2. Phil Allison

    Phil Allison Guest

    "Winfield Hill"

    ** Suround the BNC with a ring of Ivory.

    I'm sure you have just loads of that stuff where you hang out.

    With all the other bats.





    ....... Phil
     

  3. Can you saw a slit in the board to remove the chance of carbon
    tracking? That is done a lot in the HV circuits on TVs and video
    monitors.


    --
    Service to my country? Been there, Done that, and I've got my DD214 to
    prove it.
    Member of DAV #85.

    Michael A. Terrell
    Central Florida
     
  4. J.Stockton

    J.Stockton Guest

    Win
    You didn't say what the pcb material was so I am going to assume FR4. If you
    clean the boards thoroughly and then bake them out to remove any water that
    was absorbed, they will be pretty high resistance. Then conformally coat
    with a silicon based coating and bake out. This should eliminate the
    problem. Cutting the slit in the board like Michael recommended is also a
    good idea. I have had problems with leakeage in FR4 boards and it gets much
    worse with increasing temperature. The conformal coating seems to help.
    Good luck
    Jim Stockton
     
  5. Eeyore

    Eeyore Guest

    IEC 60065 says 4.73mm for 2200V btw.

    Graham
     
  6. Eeyore

    Eeyore Guest

    Is it single ended or differential ?

    ~ 2.5mm is supposed to be OK for 1kV ( but it is the *minimum* ).

    Graham
     
  7. Without seeing a photo of the section it's a tad hard to make concrete
    suggestions, but I would cut away the bad bits and try to use ceramic
    insulators/standoffs for repair.
     
  8. Jon Elson

    Jon Elson Guest

    You're running 1100 V on a BNC? You can get away with this
    sometimes, but you are really pushing things. If the connector
    solder pins are .1" apart, my only surprise is that SN 1-18 didn't
    suffer the same fate.
    You didn't even CLEAN the flux off of an 1100 V small-clearance
    circuit? It is possible the PCB itself had enough conductivity due
    to contamination in it, to break down eventually. But, leave
    flux on the board and you really invite trouble. And, once it
    starts, there may be no way to repair it without a Dremel tool.
    Yes, I think that would be advisable. Look at the sweep transistors
    on computer monitors and such. They often make slots in the PCB
    between the drain or collector trace and the other transistor leads.
    The slot is a heck of a lot better insulator than even a clean PCB.
    If you still have 1 KV across .1" of board, then the coating sounds
    like a must. If you can remove the grounded area, then maybe just
    that and a thorough cleaning will be sufficient.

    I've seen PCB fires in the last vacuum tube gear made with PC
    boards with plate supplies in the 400 V range. These were in
    commercial audio gear. On the other hand, I guess PC board materials
    are probably a lot better today.
    I have made small production runs of several photomultiplier tube
    bases (voltage dividers) in the past, and some of these run 1000 to
    1800 V or so. I've never had a fire or breakdown in them. We use
    SHV connectors for the bias supply to avoid this sort of disaster
    when going over 1 KV. (We also used the .156" Molex connectors
    and skipped a pin between gnd and HV. This worked quite well,
    even in VACUUM!)

    Jon
     
  9. I suppose in a perfect world, with a clean, washed board, low humidity,
    no dust, low frequencies, and perfect DC balance you can run 10kv per
    inch. But if it gets muggy, and a little dust gets attracted, or
    there's some flux and a DC offset, things can get ugly. But you
    already know that.

    The grinding wheel sounds like a good solution. Air is a cheap and
    relatively good insulator.
     
  10. Jon Elson wrote...
    Jon, Jon, Jon, we're talking about students here!
    You don't think I'm going to build 160 amplifiers,
    with over 150 parts each, myself, do you?
     
  11. PeteS

    PeteS Guest

    I would be a little concerned at 0.1 inch at over 1kV. Is there any way
    of cutting back the ground plane at that zone, Win?

    As I recall, the breakdown of air at 20% RH is ~10kV/inch. ( I fully
    expect contrary responses ;)

    At that, it makes 0.1" good up to 1kV in free air. So a coating might
    well help. FR4 is much better - about 50kV per inch or more, so a
    conformally coated board (with suitable breakdown characteristics of the
    conformal coat) might very well solve the issue.

    Apart from that, good advice from many - I had 'holes in boards' from
    insufficiently cleaned units that were then conformally coated (sealing
    in the contaminants). As others, I would suggest thorough cleaning and a
    silicone based conformal coat to prevent other issues (apart from the
    distance issue above).


    Cheers

    PeteS
     

  12. I thought BNC was only good for ~600v. I've seen PL259's used beyond that.

    Cheers
     
  13. Martine Riddle wrote...
    Yes. Although I can think of cases where 1, 1.5
    or even 2kV uses have worked for years on end.
    But then there's the spectacular failure at 1kV,
    or less. So, yes, 600 volts sounds about right.
    Indeed, with substantial modifications, all the
    way to 15kV. I've found there are two issues,
    1) basic coax-cable dielectric HV withstanding
    capabilities, which are dependent on pinholes
    and the like, and 2) issues having to do with
    the outer woven-shield connector termination of
    the coax, creating small, high, local electric
    fields that can disastrously break down at 1/3
    to 1/10 of the coax dielectric-breakdown limit.

    Well-designed high-voltage connectors deal with
    the shield termination in an innocuous fashion.
     
  14. Eeyore

    Eeyore Guest

    Sounds far more suitable to me.

    Graham
     
  15. On 18 Oct 2006 14:28:53 -0700,
    What do you need 160 for these beasts for?

    Just curious.
    robert
     
  16. Robert Baer

    Robert Baer Guest

    *If* the board was cleaned, 25 mils is sufficent to hold off 1250V up
    to 210C; soldermask is the coating.
    At 2600V and 20C, surface arcing over the sodermask will occur;
    cannot say where the voltage "breakdown" is.
    Helps to severely limit the current (20uA seems to not create
    carbonized paths in time periods less than 10 seconds).
    Granted, there are no intermediate layers of copper "in the way" and
    that the board material we use is not FRxx.
    Perhaps so-called "no-clean" solder was used, which obviously leaves
    crap behind.
    TINSTAAFL.
     

  17. I thought the HN connector was designed for high voltage coaxial
    connections? I remember using them in a early '50s RCA TV transmitter.


    --
    Service to my country? Been there, Done that, and I've got my DD214 to
    prove it.
    Member of DAV #85.

    Michael A. Terrell
    Central Florida
     
  18. Robert Latest wrote...
    Anti-proton trapping experiments at CERN; the
    HV amplifiers drive the trapping electrodes,
    and apparently there are a lot of them!
     
  19. Guest

    Manufacturers and suppliers readily publish data on FR4 dielectric
    strength, usually around 400-500V per mil. But this is only valid for
    80 or 100 seconds. I haven't seen a datasheet contain any practical
    information on operational breakdown performance.

    Does anyone have more accurate guidelines for the operational voltage
    range of FR4 if you want to, for example, avoid any possibility of a
    failure within 10 years of continuous operation? In electrical printed
    circuit boards, the breakdown voltage means something only to those few
    that are worried about transient spikes. I've always wondered why they
    dont publish specs that would be useful to the other 95% It would be
    nice to stick to the UL guidelines, but 500V sot-23 transistor pad
    spacings make a mockery of those. There must be a study out there some
    where.

    There is a related thread at:
    http://groups.google.com/group/sci....nk=st&q=breakdown+fr4&rnum=1#a7d1c2bacacbb04b


    A useful trick to decrease the likelyhood of breakdown and arcing is to
    bond a layer of copper clad Kapton to the bottom side of the PCB,
    ofcourse this only works for SMD only boards. The Kapton film is copper
    on one side and 0.1mil teflon on the other (Kapton FN). The teflon
    makes an excellent thermoplastic adhesive above 280degC and sticks to a
    PCB nicely. The copper side is grounded which puts a perfect ground
    plane 2mil below the PCB. If the PCB is thin, this causes the electric
    field to be directed downward, and due to the dielectric properties,
    concentrate in the Kapton, thus avoiding breakdown of the FR4. With the
    addition of some thermal vias, and an aluminium plate, you can also
    make an excellent heat-spreader or heat-sink attachment, in this case
    forget the copper layer, use two sided kapton FN and bond direct to the
    metal backing.

    You can also lay the components directly on kapton, but you only get
    traces on one side. A better option is microwave type, copper clad
    teflon, which is relatively low-cost and can be processed like 2 layer
    FR4, except you wont get plated through vias unless you deal with a
    specialty manufacturer.

    /Andrew
     
  20. joseph2k

    joseph2k Guest

    Nothing of the sort. Kapton(tm) has no chemical relation to Teflon(tm).

    Kapton® Polyimide Film



    DuPont™ Kapton® is a leader in the high performance films industry, offering
    over 40 years of diverse products, global technical support and customer
    service. DuPont has set a high standard in the polyimide film markets with
    its durability and performance in extreme temperature environments.
    Kapton® has a unique combination of electrical, thermal, chemical and
    mechanical properties and retains these properties over a wide range of
    industrial environments and applications.


    From miniaturized electronic components to Mars rover heaters, from high
    speed locomotive motors to airbag seat sensors, DuPont™ Kapton® polyimide
    films make innovative design solutions possible.

    Other than that your post ain't tooo bad.
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-