IGBT Brick based inverter design question

[applicanon]

Hello!!

I salvaged an SCR full bridge rectifier, DC link, and IGBT H-Bridge complete with snubber network from a rather large Medium frequency DC Inverter that was used as the power source for a rather large spot welder.

I am intending to use these components to build myself a TIG welder.

The TIG output frequency I would like to be able to adjust between 50hz - 800hz.

Initially I was thinking that the amperage would be controlled using PWM on the IGBT H-Bridge using a carrier frequency of something like 10khz.  The duty cycle would have been adjusted using a PWM control IC which would take an amperage reading of the output and adjust accordingly.

This seemed to bring up a large number of potential issues.  Mainly having to finess the system to be able to run consistantly at such a high switching speed for the 10khz PWM rate of the IGBT's.

I decided I had a much simpler and more elegant solution.

Would it be possible to instead simply forget all about controlling the IGBT h-bridge with PWM for amperage control?
Instead, I think it would make much more sense to just use phase control of the SCR rectifier to control the amperage.

If I wanted full amperage, then the SCR's would be full on full time, and the IGBT h-bridge would also have the IGBT's on for the entire cycle length for each cycle of the h-bridge switching sequence.  For less amps, the IGBT h-bridge switching scheme remains the same, but the gates of the SCR bridge rectifier will only be turned on for a portion of the sinewave, depending on the amount of amperage reduced.  Want 50% amps, then only half sine wave gets thru the scr portion.  25%?  Then only 1/4 of the sine wave is rectified.

Is this a viable solution?  I think it should save me tons of hastle with trying to get my IGBT's to switch at the insane frequency needed if I were to use PWM with them instead.

[mag_therm]

Do you have the gate driver boards?
To check the IGBT, measure all the gate capacitances. They should all be within +/- 5% of datasheet
To check the scr, measure gate resistances then check thay are in off state with meter in low Ohms or diode check.
To check the snubbers, you might have to unbolt a lead to get the capacitors out of circuit
Are you intending to run it on 575V 3 phase?

It sounds like a major project,
 I would suggest first simulate, then make a small benchtop model on 24 or 50 V ac  to prove your control ideas and get the controls working.

[T3sl4co1l]

Semiconductors are small potatoes.  What do you have for transformers, inductors, capacitors, heatsinks, etc.?

What overall architecture did you have in mind?

Note that 120V input is basically useless, and you'll still need reasonable capacity at 240V to do much of anything.  (15A 240V is enough to do, probably not quite a full 100A, unless you have a power factor correction input stage.  30A will be enough for most purposes, and 50A+ for heavy duty.)

You need an isolation and conversion stage of some sort, basically any old power supply, using current mode control to get the constant current output.  Which can always be made resistive or CV afterwards with an error amp, plus the current mode control protects the semiconductors far better.  So, you need a big transformer for that, and output filter choke.  Which will be pretty big indeed if your semiconductors are quite slow?

Output chop is whatever.  It's wasteful of voltage drop to use IGBTs there, but they can handle the current as long as you've got the heatsinking.

SCRs, won't respond very fast, and need a very large inductor.  Plus you still need the isolation stage.  Sure, it can be a simple chopper, though it still needs voltage and current protection because of the terribly slow response.

Tim

[applicanon]

Both the SCR rectifier module and the IGBT H-bridge module are mounted on separate aluminum base plates which have liquid cooling channels machined within.  These are beastly heatsinks and can dump a ton of heat.

I've tested everything out already it's all functional.  SCR gate trigger board is already completed and works with microcontroller level voltage input.  Same with the IGBT drivers.  Already completed and ready to rock.

The output transformer from the H-bridge is made from a substantially large toroidal core made from a material which when in operation within the frequency and the power range expected, it shouldn't go into saturation at all and keeps losses low.  Intrinsic air gaps built into the core material also ensure saturation does not occur.

Same goes for output choke, it's done.  no worries about that.

The system starts at 240vac mains on a 50 amp breaker.  it goes through a 1:1 isolation transformer which is pretty beefy and will do the job fine.  next it hits the SCR module where I would like to have amperage measurement from the output of the welder fed back and used to set the Phase angle that the SCR's turn on at.  The 240AC mains is 60hz so it's not a big deal how slow they are.  they can handle 60hz easy, right?

After the SCR module, it goes into a DC Link which utilizes some rather chunky capacitors which are very good for the job.  430V rated, 3300uF, 8 of them.  Should smooth out the scr output pretty fine i think.

next it hits the igbt h-bridge which converts it to a square wave at something between 40hz - 1600hz, whatever the person welding chooses.  Thats put through the toroidal core transformer which steps down the voltage considerably.  secondary of the transformr is centertapped and has diodes to make the output back into DC.

filter choke (along with a secondary winding which will provide feedback measurement), then the tig stinger.  That's the project.

So yea.. does that make more sense to control the current via the SCR phase angle chop instead of using the current feedback to pulse the IGBT's at the H-bridge with PWM?

[T3sl4co1l]

Damn, so it really was the old school real deal huh?

Ah, mains isolation transformer?  Crude, but I guess that'll work.

Geez, that thing must be, well, everything must be massive, really.  What did this all fit in originally, one or two racks worth?

And 50A circuit?  Nice.


So it will be DC output?  I was confused because you said pulsed output earlier.  But then why would the user control the inverter frequency, that's internal, who cares..?

So, if I understand this right, your problem is the transformer, and IGBT control.  It can't be a fixed chop, it has to adapt to load.

The SCR stage won't respond nearly fast enough: 26.4mF is a fuckload of energy (2.7kJ at 320V!) and, just do a ballpark smell test: if your TIG torch is doing 20V at 100A that's 2kW continuous, or 2kJ/s.  The SCRs could fuck off for the better part of a second and you wouldn't even know anything's different at the other end.  Yup, smells alright!

Sure you don't have to use all the caps, but the inductor's going to be a respective value, and you'd prefer to shrink both at the same time.  Changing the inductor is a much harder affair.  Removing capacitance raises the supply impedance and output ripple, which is probably suitable here, but only to a point.  (That said, if the choke is of suitable construction, maybe you can just reduce the airgap: this increases inductance and reduces saturation current ~proportionally.  Likely you don't need but a fraction of its original current rating, either.)

Yeah, SCRs in and of themselves are capable at mains frequency -- maybe even up to a few kHz depending.  That means they're good for regulating output against variation in mains voltage, even fairly sudden changes.  But that's not the problem, it's the filter after them.  And you can't make it too small because, well, it's gotta ride through the 120Hz rectified waveform at least, but will need a cutoff frequency about a tenth of that to get reasonable output ripple.  And a 12Hz cutoff frequency is pretty slow -- you will literally feel it in the pedal response, or how the arc stretches and breaks.

So to maintain arc stability, regulate current versus arc length, control arc current at all -- you need PWM into the transformer, and this will be limited by the time constant of the output filter choke, which should be only some milliseconds, fast enough to account for shaky hands for example.

Tim