Connect with us

Power inverter design for a welding machine

Discussion in 'Electronic Design' started by orvillefpike, Mar 29, 2007.

Scroll to continue with content
  1. orvillefpike

    orvillefpike Guest

    I had this topic on another forum but we got the boot because we where
    told that we where off topic. We had a good thread going on and
    somebody told us that this forum might be more appropriate for this
    kind of discussion.
    I am trying to design a welding machine based on the inverter
    principle. So far I have the AC line being fed through a diode bridge,
    filtered through a capacitor, modulated through a IGBT "H" bridge
    feeding current to a 10 Ohms load resistor. Eventually I want to feed
    a step-down transformer and then turn up the power in order to have
    sufficient current to weld but I am afraid of the transformer's
    voltage kick-back destroying the IGBTs in the "H" bridge. I have made
    a lot of tests with LT Spice's circuit simulator and it looks good but
    I am afraid of the pitfalls, in real life, that I don't know about. If
    anybody is familiar with power inverter might be able to tell me what
  2. Amps? Volts?
  3. Vector MFG. make real nice basic low cost inverters.

    A 4000 watt version could run a commercial version. And i do
    encourage using commercial methods when welding. Homemade welders are
    bad because they are required to have small importent things obvious
    to the expert but not to the untrained.

    And on/off switch must do two thing I believe, Cut power and interupt
    the main welders line. BUt maybe I am wrong and my advice is

    Bottom line get a certified design from a reliable source. NO

    Even running a power inverter driven welder requires advice, can it be
    done. I do not know, but it sounds simple.
  4. orvillefpike

    orvillefpike Guest

    I would it go be able to deliver around 100 to 120 Amps at 25 Volts.
  5. John Barrett

    John Barrett Guest

    current mode buck regulator, anyone ?? International Rectifier has some 70A
    IGBTs that derate to 50A at 100C that should handle the job if you use 2 or
    3 in parallel on a hefty heat sink with cooling fans... dont forget a
    thermal cutout :)

    might be a little easier to work out compared to a transformer design at
    those currents.. ground reference would be the only real issue to cope

    or pick one up on ebay for less than $100
  6. orvillefpike

    orvillefpike Guest

    M. Williams

    I am afraid that the amount of energy through the freewheeling diode
    might destroy it, hence the need for a snubber circuit.
    Have you got a schematic of your induction heater?

  7. Tim Williams

    Tim Williams Guest

    Check my website, I'm building an induction heater. Same thing, just a
    different output.

    If the IGBTs are co-pack with diodes and the H-bridge is constant voltage
    (stiff voltage supply, bypass capacitors aplenty), the flyback energy will
    be shunted back into the power rail by the diodes.

    When driving a pure inductive load, the voltage waveform is a squarewave, so
    the current waveform is a triangle wave. When the current passes through
    zero, the diode turns off and the IGBT part starts carrying the current,
    producing a visible step of a few volts on the flat part of the square wave.
    When real power is being drawn, the position of the transition changes.

  8. Tim Williams

    Tim Williams Guest

    Actually, that's my brother (or mother).
    The diode IS the snubber. The energy doesn't go into the diode, it is
    directed by the diode. The diode sees amperage, but not energy.
    Umm..? It's in every post of even posted it in quoted text.

  9. orvillefpike

    orvillefpike Guest

    OK, found it.

  10. Werty

    Werty Guest

    Modern is single , low cost NPN forward

    converter . Use a low cost Thyrister

    to soft start , if the choke is not enough

    to control inrush .

    I have all the parts for a 10KW welder .

    MJE13007 will drive 1KW into a microscopic

    ferrite core . IGBT are CRAP , MOSFETs

    a bit better , 20 year old bipolars work best .

    13007 will have a winding to drive the

    base , and a pull down circuit to shut down

    the oscillator .

    This is the latest technology in switchers .

    You can rectify output for D.C. welding ,

    This oscillator starts in less than 10 microseconds

    so you can "control" it to produce any waveform .

    by merely turning it on for 100 to 1000 cycles

    then controlling another , opposite circuit .

    Oscillating is safer , cause its the "failure"

    of the core , that shuts off the drive to the

    13007 . Self limiting ...

    Buy a $17 450 Watt , IBM PC power
    supply from MCM , open it up .
    The core is TINY !
    Its 1/3 the size used , just 10 years ago
    in all PC's . 450 watts , 100% duty cycle .
    But many still use push pull circuits .
    A few use the newer , better single
    drive .

    Its really mind bending watching a welder
    do 1/2" steel plate , with a tiny box , with
    no heat , 100% duty cycle ..

    And if ya build/design it yourself ,
    its better .

    about $300 will do for 10KW .
  11. Werty

    Werty Guest

    You cant buck regulator , its too much

    voltage drop .

    You will use 310 vdc up to 600vdc

    "forward converter" .

    IGBT are Vsat = CRAP .

    old MJE13007 can beat the IGBT .

    buy the new round center ferritte E cores .

    and use copper sheet for secondary .

    30 amp Shottky diodes ...

    I got parts , just need time to build it .

    Circuitry is much simpler than the old
    Push-Pull , 2 transistor PC power supplies
    10KW , 100% duty cycle ..
  12. Werty

    Werty Guest

    Non sense , there is no flyback energy .

    No Catch diodes , because there

    is no energy !

    Open up a NoteBook computer power

    supply and draw the schematic !

    Single bipolar , forward converter ,

    but it is oscillating , control is to simply short

    the base drive to ground .

    It is the most elegant , low cost and hi

    power switcher ever .

    This "controlled" oscillator starts in less than

    10 microseconds , so you can get any

    waveform you need .

    Want a perfect sine wave for your R.V. ?

    2 mirror circuits , one starts the positive

    sine , the other rests .

    No H bridge , no hi-cost components ..

    Simple , fault tolerant , for the ferritte

    cores are saturating 200 times a millisecond.

    100% duty cycle . No heat .
  13. Terry Given

    Terry Given Guest

    yeah right. although at these puny power levels a BJT wouldnt be too
    awful, go read an MJE13007 data sheet.

    for a start, Vceo = 400V, Vces = 700V. So Vdc = 350V (unless you fancy
    emitter switching which is a lot messier). So for 2500W out at say 85%
    efficiency, thats about 9A. it fits in the SOA, just (best not to think
    about lifetime).

    then the gain is ~ 6. For Ib = 1.5A, Ic = 8A, Vcesat = 2V. Hmm, thats
    about what the IGBT does.

    now make it switch fast (and dont forget your floating gate drive has to
    supply 1.5A continuous, along with hefty turn-on and turn-off spikes).
    at those sorts of current levels, you will need to do something pretty
    special. and dont get it wrong, lest your SMPS explode.

    the humble IGBT is looking pretty good.

    I think its pretty clear from the data why you dont see 2.5kW SMPS using

    1mOhm = 10W. better be a good transformer. Still with MJE13007 it will
    barely be above 20kHz, so that helps a bit. except when you count how
    many layers in the transformer, and try to keep Fr = 1.5. Then there is
    the lead-out.
    at 100A? thats a good trick.
    yeah right. big bipolar SMPS did exist, but they were complex, and very
    hard to make work correctly. way, way beyond a do-it-at-home project,
    unless you happen to be an expert in that particular field.

    this is why IGBTs are the clear winner above a few kW, and FETs below
    that - the removal of Rdson = k*Vdsmax^2.5 pretty much killed the need
    for BJTs in smps.

    I caught the tail-end of big bipolar drives, using tralingtons. even
    then the 300A parts had a gain of about 30, so the base drivers were

    for several kW, FETs are the best bet here, you can get nice 600V FETs
    fairly cheaply (although you need Rdson < 0.3R to beat an IGBT in terms
    of conduction losses), and some awesome LV FETs for synchronous
    rectification. even then its not trivial, but a shitload easier when you
    can ignore SOA and the gatedrive is comparitively easy

    by the time you get above a couple of kW, the required Rdson makes the
    FET very expensive (assuming you want high efficiency), and a
    non-starter above about 10kW.

  14. Terry Given

    Terry Given Guest

    what, no leakage inductance in the transformer, and no stray inductance
    in the layout. thats a good trick.
    betcha the output doesnt ramp up that fast.
    not in a forward converter they dont.

  15. Terry Given

    Terry Given Guest

    OK, build one then post an MPEG of it welding. Tim has plenty of pics of
    his induction heater working.....

    thats a bit different from 10kW.
    well it was in 1963.
    this is a blocking oscillator. again, all the rage in 1963.
    thats because using FETs allows them to switch at 100kHz - 200kHz, as
    opposed to 25kHz - 40kHz with bipolars.

    and, of course, its 450W. now incre3ase the power 22x. Wow, I^2R losses
    just got 500x higher. Hmmm.
    perhaps you should look a bit harder at a modern PC SMPS. no blocking
    oscillator circuits there, nor will you find a saturating core (except
    perhaps as a mag-amp auxiliary regulator, but even then not likely)

    next you will say it runs first time. and never fails. LOL

  16. orvillefpike

    orvillefpike Guest

    I am using some International Rectifier IGBTs in the "H" bridge that
    have, the specs are similar to the one you describe. I am using a "H"
    bridge do drive a transformer because it seems simpler to me and it is
    also safer, because of the isolation factor, and by driving a
    transformer, the current going through the IGBT should be smaller and
    It might work with 4 IGBTs in the bridge instead of 8 or 12.
  17. orvillefpike

    orvillefpike Guest


    How much current where you able to pass through the IGBTs. How much
    current did it draw from the utility line.

  18. Tim Williams

    Tim Williams Guest

    Well, I've seen peaks up to 200A (maybe even 300), under very unusual
    circumstances. I think I fixed the suspect that was causing that. Don't
    want too many like that through the 4PC50's, though!

    I haven't had power draw more than about 1kW, which is around 10A from the
    120V line. The transistors appear to be good for at least 50A and 300V,
    which corresponds roughly to one fat ass weld bead in your case.

  19. orvillefpike

    orvillefpike Guest


    Isn't your circuit connected to 240 Vac?

  20. Tim Williams

    Tim Williams Guest

    It will be... still in the testing stages (regardless of what any of my
    drawings may indicate).

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