TIG Welder Failure Analysis and Help for Repair

[olNick]

Hello all,

New member here and joined after reading some of your welder repair posts.
In order not to pollute this space too much, I'm adding a link to my post in the diysmps forum. The admin there was helpful with some comments, but I'm asking for more help here. I you deem it better I can transfer it all to here.
Basically I blew up the welder using an under powered, cheesy generator that alsmost stalled. The saga is here...

http://www.diysmps.com/forums/showthread.php?697-SMPS-Failure-Analysis-and-Repair&p=11808

The schematics for the welder are referenced there, but I add the links here also...

Schems are here,

My welder 160 amp https://dl.dropboxusercontent.com/u/85561940/th160.pdf
Component placing FWIW, https://dl.dropboxusercontent.com/u/85561940/thcomp.pdf
Same welder more less but in 200A version https://dl.dropboxusercontent.com/u/85561940/th200.pdf

My first request here is feedback on whether my thinking as to why it blew up is on track. After alot of reading, and getting a handle on how smps work, this is what I think happened..

My thinking is the extreme under voltage/brownout when the generator almost stalled dropped the +-15 ac voltage coming out of the transformer which then gets regulated to +- 15 volts for gate drive, and then regulated again to +-12 vdc for for the control circuitry. With this low voltage the control/pwm gdt drive etc stopped working and the igbts stayed on and fried.

Or the +- 15v gate drive was too low to switch the igbts off.... or too high whhen generator was overvoltage and exceeded the 20v gate-emmiter limit. and cooked the device.

Once this happened the short sent 300vdc across the gdt secondaries.
After this replacing parts (IGBTs, bridge, fuse etc.) w/o replacing the open GDT it would still blow up (which it did) due to open g-e circuit, since the GDT was now open.

my thoughts, comments welcome,
regards,
nick

[olNick]

I have ordered an new GDT, and wonder if simply scoping the gate waveform without the high voltage connected (B1 disconnected) would be sufficient, as I have no real easy way to check with 300V between e-c.

This was my last post over there...

There are 2 rectifiers for the incoming ac.

One is a big bridge rectifier (B1) on schem that feeds the main filter caps and generates the ~300 VDC for the IGBT's. This is turned on by the relay K1, I'm guessing with a delay giving the logic circuits a chance to start.

There is a +-15V transformer on board (TV1) on schem that feeds another smaller rectifier (B2) that gives +-15 Vdc (for gate drive) and also gets regulated 10 +-12V for the low voltage circuit. This gets 220VAC as soon as the power switch is turned on.

What I am saying is I can disconnect the B1 so no 300VDC goes to the IGBTs but I have my low voltage circuit working and can check for pulses waveform quality across g-e on device. Granted this is without load...

I cannot use a variac (at least simpley) to bring up the voltage slowly because I will not have enough ac voltage going into B2 -- the control voltage. I believe this under voltage is what cooked things in the first place....

I can use a dim bulb tester instead and limit current....
Also, my scope is an old analog 2 channel scope and I dont have a differential probe...

regards,
nick

[Richard Head]

OlNick

I would disconnect the HVDC supply to the IGBTs as you suggest and power only the small internal DC PSU.
Obviously replace all suspect components beforehand. Gate drive resistors can also blow open circuit with very little visible damage so check them. You will probably have to lift one end due to the GDT resistance.
If you have a dual channel scope then check the gate drive of both low side MOSFETs at the same time. (use an isolation transformer for scope) The duty cycle should be maximum and there should be no cross-conduction.
Repeat the procedure for the high side MOSFETs remembering that this particular measurement can only be performed with no power applied to the IGBTs.
The gate drive waveforms should be clean with no significant ringing/overshoot and the transition times should be less than about 200ns (preferably faster).
Confirm the operating frequency is correct (if you know what it is). Probably about 20-30kHz.
The next step is to connect the supply to the IGBTs collectors again and using an externl isolated PSU to supply the control circuitry gradually turn up the variac while monitoring the gate drive and DC output voltage. Also, turn the output current pot all the way down. The output voltage should climb steadily as you increase the variac until the DC voltage is about 70VDC and then hold steady. As this happens the PWM of the gate drives will pull back and the waveforms will become jittery. This indicates that the outer voltage loop is regulating the output voltage.
Remember that a DC welder is designed as a constant current source so you need to load the output to confirm current regulation. You will need a resistive load that can draw about 30-40A at 12V. Using a DC current clamp (or shunt) on the output leads confirm that the current is being controlled by adjusting the front panel current pot. This test should ensure that the unit is ok the switch on directly, after disconnecting the external PSU etc.

[olNick]

Thank you for the response Richard,

I am waiting on the gate drive transformer which has the secondaries open.
Yes, the gate resistors have blown open and the transils (tvs) look suspect so I will replace them also.

Currently I only have access to a very old single channel Trio 1303D 1 meg scope as my Tek 7603 is in storage so I guess I can only check for gate drive one by one, igbts installed but no power...

I dont know if I mentioned it, but this welder only had a pair if HGTG40N60B3's one  on the low and one on the high side even though the board is drilled for 3 igbts per side.

The schem shows a pair of HGTG30N60B3 per side.
Do you foresee any problem by using a pair per side as shown or should I get the thing working with the single, more powerfull device per side as it was? I feel this would make for a more reliable welder.
The same  board (S300 pcb) is used to drive all the models so I'm guessing that there is enough drive signal there.
I do not know if they use a different GDT in the other models, but I would not think so.
The heatsink is also drilled for 2 devices per side.
OTOH, I have read that paralleling devices opens up another can of worms...


Anyway, would appreciate your thoughts..

regards,
nick

[Richard Head]

OlNick

I had a look at the circuit and it's a double forward topology, not a bridge. Crazy at this power level these days.
Anyway, I would discourage you from paralleling IGBTs as they don't share current well. Better to use a larger device.
I would be inclined to first repair it to the state it was when it failed. Once it's up and going again then you can try to improve it.
Don't forget about heatsink compount between the power devices and the heatsink, it's vital at higher powers.

Good luck with the repair.