Author Topic: Homemade IGBT MIG welder with ATX PSU parts. [blown yet again][blown count: 5]  (Read 28554 times)

0 Members and 1 Guest are viewing this topic.

Offline magic

  • Super Contributor
  • ***
  • Posts: 7453
  • Country: pl
I think the part around the power transformer, primary switches and secondary rectifiers would be more relevant ;)
In particular, secondary diode voltage is something you should know, not guess. What's the turn ratio? How is it rectified?
« Last Edit: July 24, 2019, 10:03:46 pm by magic »
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
I think the part around the power transformer, primary switches and secondary rectifiers would be more relevant ;)
In particular, secondary diode voltage is something you should know, not guess. What's the turn ratio? How is it rectified?

I don't quite remember the transformer parameters but they're here in this thread somewhere along with all the other details.
Secondary is configured in a half bridge (center tap ground), primary in a push pull.
Mains is rectified into a capacitor bank with two series capacitors, where the transformer goes in the center, everything is very similar to an ATX PSU.
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
Whew, spent some time surfing through aliexpress, reading all the reviews and found a few gems.


The image is from the reviews and it's quite hilarious, as there's an SMD device stuck into a TO-247 package.
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 
The following users thanked this post: Electro Detective

Offline magic

  • Super Contributor
  • ***
  • Posts: 7453
  • Country: pl
Re: Homemade IGBT MIG welder with ATX PSU parts.
« Reply #128 on: July 25, 2019, 07:26:31 am »
Open circuit voltage doesn't matter because PWM.
:scared:

Also because of the way SMPS PSU works you have to divide your 230VAC peak voltage by 2.
A 10:8 ratio would mean about 17% duty cycle at idle to get 22V.
230*1.41 = 324V(peak)
324*8/10 = 259V
259/2 = 130V
22/130 = 17% duty cycle.
Wait, so is the primary 2×10T and the secondary 8T in total, i.e. 2×4T?

Because if not and it actually is 2×10T and 2×8T (or center-tapped 10T and 8T in total) then you got the math wrong. You apply primary voltage to only one 10T winding at a time, but both 8T secondaries react, producing a combined 1.6×Vin across the secondary rectifier.

Is this the circuit you use?
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
It is the typical half bridge PC PSU topology,

The transformer turns ratio is best calculated from the primary turns divided by the turns of one of the secondaries.
« Last Edit: July 25, 2019, 07:43:33 am by xavier60 »
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 
The following users thanked this post: Refrigerator

Offline magic

  • Super Contributor
  • ***
  • Posts: 7453
  • Country: pl
OK, thanks for the correction. I got to "primary is a push-pull" and stopped reading there :palm:

If the primary is 10T and the secondaries are 8T each than it's 330/2·0.8=130V indeed. So the diodes are killed by transients or their dubious provenance :P
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
OK, thanks for the correction. I got to "primary is a push-pull" and stopped reading there :palm:

If the primary is 10T and the secondaries are 8T each than it's 330/2·0.8=130V indeed. So the diodes are killed by transients or their dubious provenance :P
And undamped ringing would come close to doubling that.
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
OK, thanks for the correction. I got to "primary is a push-pull" and stopped reading there :palm:

If the primary is 10T and the secondaries are 8T each than it's 330/2·0.8=130V indeed. So the diodes are killed by transients or their dubious provenance :P
And undamped ringing would come close to doubling that.

The snubbers on the diodes are another thing i forgot  :palm:
I had them on the crap 150EBU04 diodes.  :-BROKE
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
Took the MOSFETs out and they're both blown.
Cracked one open to check the die size and it's 5.5x5.5 = 30.25mm2, 0.5A/mm2.
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7453
  • Country: pl
So what actually is the point of having ±65V open circuit secondary voltage for 22V output? Do you expect 33% efficiency under actual load? :-//

I don't know what can be realistically expected from such converter, but 33% seems rather extreme. SMPS work better if transformer ratio is close to ideal and duty cycle close to 50%. Isn't it be better to cut the secondary in half and parallel the two parts for 32V output at lower impedance, still giving margin for 66% efficiency?

PWM surely is magic, but you pay for it right here.
« Last Edit: July 25, 2019, 03:00:12 pm by magic »
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
So what actually is the point of having ±65V open circuit secondary voltage for 22V output? Do you expect 33% efficiency under actual load? :-//

I don't know what can be realistically expected from such converter, but 33% seems rather extreme. SMPS work better if transformer ratio is close to ideal and duty cycle close to 50%. Isn't it be better to cut the secondary in half and parallel the two parts for 32V output at lower impedance, still giving margin for 66% efficiency?

PWM surely is magic, but you pay for it right here.

I can remove a few turns in the transformer but not now, i still want to test the current setup out.
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7453
  • Country: pl
Not my circus, not my monkeys :P
Higher turn ratio means lower peak secondary voltage, less primary current and even less resistive losses for the same amount of trafo wire.

https://www.tme.eu/en/details/idw100e60fksa1/tht-universal-diodes/infineon-technologies/

Is this good enough? 600V, 100A and 200W continuous at Tc=100°C. Should be hard to destroy unless you forget the snubber again :P
And you are guaranteed it's not a 1N4148 inside and you get a proper datasheet corresponding to the actual silicon in your hands, no guessing, no smoke, just calculate your power lossess and heatskinking and it works.
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
A little update on the whole project.
I've since blown the welder up two more times.  >:D
Each time getting closer to proper operation.
With the new diodes i have (IDW100E60) and new MOSFETS (IRF460) the welder does power up and survives a trigger pull just fine.
No load with the trigger pulled (operational) the inverter draws 50mA from the mains, which i think is not too bad.
But the MOSFETs popped as soon as the welding wire touched the metal scrap i tried to weld.
I then doubled up on the MOSFETs, previously there were one per side, now there are two per side (a total of four) and also removed two turns off the transformer (10 primary, 5.5 secondary).
Did some tests, everything seemed ok-ish, i then separated a few strands of my negative welding wire and shorted them out to the positive to see how well it melts wire.
It lasted about 30 seconds before blowing up yet again, although a lot more spectacularly than before.

So for once i decided to do some actual research and calculations (groundbreaking stuff, i know).
Couldn't find much info on FBT cores, but Bmax on them appears to be in the range of 2000 to 3400 gauss.
I took 1500 as my Bmax to be extra safe and calcuated a primary of 10.7 turns (rounded off to 10, because generous Bmax estimate).
From that i calculated my secondary, which turned out to be around 3 turns.
I'll rewind my transformer in the same way i did before with the primary between two secondaries, but i'll split the secondary wires to make them bifilar.
So my transformer will be wound with the primary between two bifilar secondaries.
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7453
  • Country: pl
That's progress :-+

What's the switching frequency, duty cycle, dead time, average input current and primary inductance? Here some math needs to be done: checking if idle magnetizing current of the primary is sane, calculating FET Rds(on) losses.

It seems like your topology shouldn't subject FETs to significant overvoltages because of the clamping action of the opposite FET's body diode, but that's only if everything is laid out tightly and the current loop through the FETs and the capacitor bank is short.

BTW, I have no idea what will happen but you could wrap one turn of cable around the transformer, bring the voltage further down with resistor dividers and hook it up to the scope. It might give at least some approximate picture of voltage waveforms at the working windings.
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
Did some estimates and decided to go with 10:2.5 setup on the transformer.
Also the bifilar thing didn't work out because the wire was too tangled to separate.
With only 2.5 turns on the secondaries it was a bit of a challenge to wind, in the end i decided to loosen the twisted strands of my wire to spread them out, this way it would cover most of the surface. Copper sheet would be best for situations like these.
Anyways, i'll assemble the welder tomorrow and will see how well it works this time.
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 

Offline SiliconWizard

  • Super Contributor
  • ***
  • Posts: 15797
  • Country: fr
I've since blown the welder up two more times.  >:D

Congratulations! :P
 

Offline strawberry

  • Super Contributor
  • ***
  • !
  • Posts: 1199
  • Country: lv
there is no 2.5
 

Online coppercone2

  • Super Contributor
  • ***
  • Posts: 11336
  • Country: us
  • $
been waiting like 3 years to power up my IGBTs, looks like I made the right call doing nothing
 

Offline SiliconWizard

  • Super Contributor
  • ***
  • Posts: 15797
  • Country: fr
been waiting like 3 years to power up my IGBTs, looks like I made the right call doing nothing

As an old colleague of mine used to say: "it's urgent not to do anything". ;D
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
Tested the welder but not fully.
Seems either i made a mistake when rebuilding or the previous failure did more damage than expected, because at a certain input voltage the MOSFET half bridge starts oscillating.
Above 130VAC if i pull and release the trigger the MOSFET half bridge starts oscillating at around 90kHz, while my switching frequency is set to 50kHz.
I think i'll remake the MOSFET board and if that doesn't help i'll remake the TL494 board along with the GDT.
Otherwise everything seems to be looking pretty well so far. Under no load output reaches 19V at about 130VAC input and with a 1.3-ish Ω load it maintains 14V, which might be limited by the input voltage.
With the 1.3Ω load at 14V the ripple on the output is 1.4VRMS
I haven't yet tested the welder on anything above 140VAC because i fear it might blow up.
Also my variac is only rated up to around 300W and starts buzzing pretty loud on anything above 1A.
After operating with the said load for a few minutes all of the heatsinks are cold.


« Last Edit: August 15, 2019, 06:45:46 pm by Refrigerator »
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
Hooked up my scope to the gate of one of the low side MOSFETs to capture the waveform during oscillation.
I also made a quick and dirty current clamp with an inductor core and a clothespin, with 100 turns and a 100Ω resistor it should give me 1mV/1mA.
Yellow - gate voltage, Red - transformer current.
Any idea where this could be coming from? Connecting a 23.5Ω load on the output stops the oscillation.
« Last Edit: August 16, 2019, 09:08:33 am by Refrigerator »
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7453
  • Country: pl
I would guess the PWM controller just drives it that way. :-//
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
I decided that the 23Ω load resistor is an ok fix and ran some tests at full voltage.
For these tests i made a 2Ω load resistor out of some old welding wire, but 2Ω was a bit much so i stepped it down to 1Ω and later ~0.75Ω for my tests.
Checked the voltage range and it goes from 18V to 25V, testing was done at 25V with the aforementioned resistances.
I let the welder do its thing for about 15 minutes and checked the temps. Both heatsinks were warm, could say almost hot but not bad, considering there was no fan plugged in, that's with the 1Ω load.
Too bad i don't have a temp probe so i can only go by feel. Of course i unplug and discharge everything before i go poking my fingers inside.
Before i do any more testing i'll go make a temperature controlled PWM speed controller for the fan i'll be using because it uses a PWM input and is very loud at full power (afc1212de).


Edit: had to downsize the image.
« Last Edit: August 16, 2019, 02:29:50 pm by Refrigerator »
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
I would guess the PWM controller just drives it that way. :-//
When the trigger is released the GDT drive transistors are turned off along with the controller output (DTC pulled high) so it's not the controller.
I guess it was because there was no current going through the GDT that allowed the oscillation, anyways it's fixed now.
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 

Offline RefrigeratorTopic starter

  • Super Contributor
  • ***
  • Posts: 1593
  • Country: lt
I think it's ready for a true weld test, but i'll do that tomorrow, because it's dark already and all the mosquitoes have come out to bite me in the arse.
It's quite amazing what a little airflow can do, with the fan barely idling the MOSFET heatsink is only lukewarm and the diode heatsink is ice cold when running just shy of 700W.
I also "accidentally" melted the load resistor i made.  >:D When i shorted the melted ends out i got some hissing arcs.
One thing i'm not feeling too confident about was the squeal coming from the inverter during moderate loads (above 1kW).
Last time right before it blew up it started squealing, which is why it worries me so much.
I have a blog at http://brimmingideas.blogspot.com/ . Now less empty than ever before !
An expert of making MOSFETs explode.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf