GYS Inverter 3200 MMA welder cooking freewheel diodes.

[shakalnokturn]

I'd like a little advice here from anyone who has a little experience on arc welder repair.

This welder GYS Inverter 3200 has a few clones with different names. PCB model is 64184IND7-P, that being the best way to find any schematics for GYS equipment.
The welder has a rated no load output of 80V and a maximum current of 130A@25.2V

I bought this broken looking for something different to repair and then learn to use. The previous owner said that it got hot (thermal protection LED on) but no longer worked once cooled.
When I received it the SMPS was in hiccup mode, this was narrowed down to 1 of the 3 STTH2003CG double diodes on the heatsinked IMS. (Total of 6 parallel diode junctions for freewheeling along with 2 parallel R-C series snubbers.)
One diode of the 2 in the middle STTH2003CG was shorted, the surviving one had an abnormal curve.

I ordered a replacement from Littlediode on eBay then got hot and bothered replacing it on the IMS and once replaced and reassembled took it for a quick test. Quick because it failed after 5min...

The diode I had replaced has failed again, I do believe the layout is not perfect and that the centre one gets hit by more current than the others but this time it got hit so hard that it partly desoldered itself from the IMS !

Seeing this I have the feeling that I was maybe sent a fake diode, I don't really trust any eBay semiconductor sellers.

What else should I be checking before attempting to replace this diode once more?

Schematic can be found here: http://valvol.qrz.ru/schems/Gysmi-131.pdf
GYS is a French brand and the French really are NO good at schematics... sorry for the headache.
The only difference is my PCB does not have the small section with Q33P01, Q33P02... of that part it only has D33P06.

Also what are good practices for rework on IMS?

[shakalnokturn]

After removing the IMS board and heatsink, on closer inspection the heat has unstuck the insulating layer from the back plate so now I'm in trouble.
Will have to remove the remaining freewheel diodes and dremel part of the IMS board away to fit TO220 diodes in place.

The STTH2003CG I received from LittleDiode looked just like the ones here:
https://fr.aliexpress.com/item/32827360393.html

Only mine didn't have the ST logo sanded away (is that confession to forgery?), funnily not many photos these Morocco produced diodes out there... And all other photos have the logo bottom-left for whatever that's worth.

[TERRA Operative]

I'd start by purchasing components from reputable sellers.

[shakalnokturn]

I have to agree although I still get tempted once in a while to cut shipping costs,but then the extra damage, wasted time, doubt... It isn't really worth it.

[shakalnokturn]

I'm going to dremel away the portion of the IMS where the 3 freewheel diodes were and fit three IXYS DSEP30-06A in place, another option would be some Fairchild ISL9R3060G2.
TO-247 package rather than TO-220 because the thermal resistance must have been pretty low with the original IMS design (I don't know how much power they actually have to dissipate) and I have room to fit them although it's going to be an ugly fix.

On the diode side I'm going from three dual 10A diodes (total 60A continuous & 110A IFSM per device) to three single 30A diodes (total 90A continuous & 250A IFSM per device)

Original R-C snubbers for these diodes were 2 networks of three parallel 22ohm 1206's in series with 3.3nF ceramics each squeezed in between two STTH2003's.

Is there any benefit to spread this out as 3 snubbers of 10ohm in series with 2.2nF soldered right at the diode legs?
Any suggestions?

[shakalnokturn]

I did say it would be an ugly fix...!

I went for the cheaper ISL9R3060G2 diodes having a mounting hole whereas the IXYS don't, bought from ASWO (not always a reliable source, the diodes came in individual sealed antistatic bags with a Radiospares number on them so I felt confident) unfortunately, as often with this supplier they were low on stock, I only received 2 diodes.
Being impatient to repair I decided to go along with "only" 60A worth of diode.

For the time being it is working well, I gave Little(turd)Diode (eBay seller) all the feedback and respect they deserved. If in need for not so cheap fakes to set alight they're a decent source.

[langwadt]

I don't know much difference it makes but you should probably try to make the wire lengths equal, or possibly a bit longer and separate to help current sharing 

[shakalnokturn]

I have tried to do that on the present layout.

AC-in are 3 pins bottom right, +DC-out are four next pins, four top pins are ground.

On the transformer side: The right STTH2003 is closest to the current source so DC output is soldered on left STTH2003's tab.
On the arc side: Right ISL9R3060 is closest to current source (ground) so left ISL9R3060's cathode is first linked to DC-out.

Typical central-heating plumbing...

[TERRA Operative]

Yeah, I've had dubious experience with Little Diode before too, I refuse to buy from them anymore...

Also, I like the PCB surgery.