Replacing IGBTs in a 2.8KVA UPS

[Medtech1]

I wish to replace some bad IGBTs in a 3000VA APC UPS that have blown.

Pictures of the IGBTs as I was removing them (it blew the emitter lead right off on three, and caused a short between the emitter and collector on another):

http://redhat.systems/IGBT_bad1.JPG

http://redhat.systems/IGBT_bad2.JPG

They are a GP4063D.. this is an image of one of them: 


 
There are a few part numbers for the GP6063D when I searched for replacement ((ie: TO-247AC AUIRGP4063D (TO-247AC)  /AUIRGP4063D-E (TO-247AD), IRGP4063DPbF, IRGP4063D-EPbF,), that vary in max ratings ie: continuous collector current  96A vs 100A etc 

As others have mentioned, I'm concerned about purchasing from Ebay or Aliexpress given the fakes out there.

Given these are from 2011 or so, can I be relatively sure this replacement IGBT is a matched replacement?

https://www.infineon.com/dgdl/auirgp4063d.pdf?fileId=5546d462533600a4015355ba62491525


Even though four of them blew (I'm thinking a high overload current (ie: short) from the output of the UPS did this), I was going to replace all six of them, but since two are still ok, is it ok to mix two old working ones with 4 new?


Also, many of the "GP4063D" IGBTs on eBay show them as being 96A (Continuous Collector Current at 25 C), which doesn't match the datasheet for a GP4063D I posted above (which shows it as 100A.. maybe this is negligible?)

https://www.ebay.ca/itm/10pcs-IRGP4063DPBF-GP4063D-TO-247-600V-96A-transistor-Original-IR/291637989330?



Thanks in advance.

[TheMG]

Buy from a reputable source, such as: https://www.digikey.ca/en/products/detail/infineon-technologies/IRGP4063DPBF/1300610

I would replace the whole set. Why take the chance? The other IGBTs that appear to be ok might have been strained and could potentially fail in short order (seconds, minutes, hours, days....), taking all of your brand new replacement IGBTs with them.

Also ALWAYS verify the integrity of the gate drive circuitry, which could have been damaged by excessive voltage backfeeding on the gate pin. This can be done by removing all the transistors out of the inverter circuit, and using an oscilloscope to see if there is a gate drive waveform on the gate pin at each of the transistor locations. This will likely be very brief, as the UPS control circuitry will probably very likely fault off rather quickly, so a DSO in single-shot capture mode makes this easier.

[james_s]

I would second this, there's no reason to buy from unknown sellers when you can get a suitable part from a reputable supplier.

I would also replace the full set and then keep the pulls that test good and use them for personal projects. Used parts like that are good for tinkering around and prototyping or when something blows up and you need a quick fix.

[exmadscientist]

+1 on replacing the whole set of six (who knows what happened to the other two?), and +1 to checking out the gate drivers (they have a tendency to fail when transistors fail). It might be wise, if you are placing an order anyway, to grab a set of gate driver ICs even if you decide not to install them.

The reason you're seeing discrepancies in the 25C current figure between 96A and 100A is that there are two different products: the commercial 96A IRGP4063D and the automotive 100A AUIRGP4063D. Don't ask me why; vendors are awful. Your photos show the commercial parts in TO-247AC, but the automotive ought to be fine as well.

[Medtech1]

Buy from a reputable source, such as: https://www.digikey.ca/en/products/detail/infineon-technologies/IRGP4063DPBF/1300610

 

I was curious what is the difference between these IGBTs and your url (other than the $4/per difference)?  Automotive a higher grade?

https://www.digikey.ca/en/products/detail/AUIRGP4063D/AUIRGP4063D-ND/3455192?itemSeq=345693792

[exmadscientist]

Automotive a higher grade?

Yes, exactly. Usually automotive parts have tighter screening for higher reliability... and much higher cost.

In this case the automotive part has slightly better specifications, which is unusual but does happen now and then.

Either is perfectly suitable for this use, get whichever you can get your hands on. Sometimes automotive parts are easier to buy or cheaper, as they may be specified for "long production life" and so you can end up with the commercial grade part discontinued and only available from vultures, while the automotive remains in normal production. Like I said, vendors suck.

[Medtech1]

Thank you for all the input/comments.. I'll order some additional IRS21844S IC's as well which appear to be the gate drivers for the IGBTs.

[Medtech1]

Hey Guys..  sorry to ask if this is a dumb question..   but is the PBF in this part number just mean Lead free?   IRS21844SPBF

https://www.digikey.ca/en/products/detail/infineon-technologies/IRS21844SPBF/1189741


Sorry.. I'm not an engineer..

[james_s]

Yes, at least that's what it means everywhere I've seen that suffix.

[Medtech1]

I've replaced the IGBT driver IC's (all three).

As I replaced the IC's under the microscope, I found some other damaged components on the board and decided to replace  (in addition to the six IGBTs) because once the IGBTs are soldered back with the huge heatsinks, I can't get all these things again (this is the UPS view when it is assembled - right now I have the main board out -  http://redhat.systems/P1004673.JPG ):

1) the schottky diodes (one was physically damaged)
2) at least one resistor (physically damaged)

I may also replace the 47uf capacitors.

This shows the components I'm focusing on:

http://redhat.systems/SMX3000_UPS_cctboard1.JPG

http://redhat.systems/SMX3000_UPS_cctboard2.JPG

I am starting to wonder what happened to this UPS.

btw, if anyone wondered, I'm only an electronic hobbyiest (long time).. and took this thing on as a challenge, as it was an eBay purchase. I received my money back due to it being defective. It powers up, and just shows some faults (obviously). It's a  $2800 UPS (here in Canada anyways), so it would be nice to get it running if it isn't a crazy amount of work).

[james_s]

Unfortunately with power electronics, a fault can be triggered by something as simple as a cracked solder joint or some small component failing and instantly cascade into a catastrophic failure involving many often expensive components. When repairing something like that you have to be extra careful to locate all of the faults otherwise when you power it up the thing may go bang again. It can be tricky even for an experienced tech but with a little luck this thing may just work once you replace the bad parts you've found.

[TheMG]

That is why when repairing such things, usually soldering in the new IGBTs would be one of the last things I do. I always verify for a switching waveform at each of the gate pads with all IGBTs removed.

Test every diode, resistor, etc in the gate drives even if it look physically fine.

All it takes is one faulty component in the gate drive circuitry of one IGBT and the whole thing goes bang again. If you test without the IGBTs installed, if there is still a fault, no magic smoke will escape because there is no current path.

Do also test to ensure no short circuit on the inverter output within the UPS itself (filter caps for instance).

[Medtech1]

That is why when repairing such things, usually soldering in the new IGBTs would be one of the last things I do. I always verify for a switching waveform at each of the gate pads with all IGBTs removed.

Test every diode, resistor, etc in the gate drives even if it look physically fine.

All it takes is one faulty component in the gate drive circuitry of one IGBT and the whole thing goes bang again. If you test without the IGBTs installed, if there is still a fault, no magic smoke will escape because there is no current path.

Do also test to ensure no short circuit on the inverter output within the UPS itself (filter caps for instance).

I did go through all the resistors and caps, mostly looking for shorts since they are in the circuit... (including between all the IGBT connections on the pcb).

I'd like to try testing the gate pins with a scope.. (the IGBTs aren't install yet) I just need to research how to do it.     

From the spec, the gate threshold voltage is 4 to 6.5v.. so I would expect to see pulse between the gate and the emitter?

I'll admit I've put most my focus on the area around the IGBTs (shown in the pic below - bottom right of the image).
 
https://redhat.systems/P1004680.JPG

I can power it on with a 15A/120v source... even though it is a 120/30A power in, without any load or batteries connected, I doubt it would need much for amperage. One interesting thing about this UPS is the battery power supply is 10 x12VDC batteries in series, so at full charge, the total battery voltage in is around 136V. Therefore, not much of a requirement for a big heavy transformer to raise the battery voltage up to the 120AC line voltage for loads, like most other UPSs.
 

[james_s]

Yes, typically I'd expect to see 12-15V or so.

[Jetes]

Hi Medtech1.  I have the exact same UPS with the same issue.  Before I go replacing all these IGBTs, I wanted to see if you had success getting yours running again.  Thanks to everyone else that replied to this post, the info is very helpful.