Real or fake IGBT FGL40N120?

[masterx81]

Hi!
I'm repairing an UPS (bought used quite cheap as functioning, but with all igbt shorted and some other mess around, i doubt that it was working), and i'm having fun finding all the components needed, trying to avoid counterfait parts...
How i can test an IGBT for know if it's genuine?


Left is original, right the new one. The package seem consistent, but the markings are a bit different.
The dca75 tells me:
Vge(on)=4.569v at Ic=5.04mA and Ig=3uA
Vge(off)=3.628v at Ic=5.1uA
gfe=29.9mA/V at Ic=3.0mA to 5.0mA
Vcesat=0.526V at Ic=5.0mA and Vge=8.0V with 0.52Vf body diode
From what i can tell, the few parameters that i can match on the pdf seem spot on, but the DCA75 test at way too little current...
I not have a good old IGBT for measure the parameters, all were shorted or exploded.
There are some tests that i can try to see if it's genuine? I've a 30A electric load, a signal generator, oscilloscope, etc.....
Thanks!

[ArdWar]

Looking at visuals alone both aren't convincing tbf. Both logos are in the wrong proportions compared to official Fairchild marking, unless there were batches where laser bloom was normal.

[BrokenYugo]

Who/where did they come from?

All power transistors from questionable sources can be assumed fake, buy from trusted suppliers, Digikey, Mouser, etc. if you want a trustable product. Not eBay, Amazon, Chinese marketplace sites, etc. half those sellers don't even know what they're selling.

Try scratching the gloss black in the middle, if it's painted it's definitely fake.

If you insist on testing I guess mapping out the whole safe operating area per datasheet and see if one blows where it shouldn't would be fairly definitive, or decap one and compare to real, fakes will have smaller/different die.

[T3sl4co1l]

Right one is most likely bogus: notice the checkerboard finish on most of the surface (milled or lasered away), lack of mold mark (it's been ground down), and take a close look at the relief around the marking areas (is it painted or polished, or molded in?).  Leads look straightened.  Way too many scratches on the plating.  Not sure if they're successfully remaking matte tin plating these days but at least it's not hot-dipped.

All power transistors from questionable sources can be assumed fake

All components---FTFY.  I doubt it's even unique to semiconductors, there's just a little less value in commodity chip or leaded components and without as much marking (logo, date code, etc.) we don't notice.  You can't really clone a resistor with it still being apparently a resistor, but you can play the same sourcing shenanigans that affect semiconductors -- if there's no traceable path or accountability back to the supplier(s) and manufacturer, assume that break in trust has been abused.

Tim

[masterx81]

They are from ebay... i can't find them on relaible sources as seem a discontinued part.
Good test the one on the clear part. On the original device, the clear part is a bit under the level of the surface, while on the new part is a bit higher. And yes, if i scrape, the clear part came away...
Not good...
There are out there some versions already with the onsemi logo... But always from not reliable sources.
There are on the market more counterfait parts than original

[masterx81]

And worst part, on the onsemi site i can't find a similar part. The suggested replacement parts have all more capacitance on the gate, and are slower, and also less maximum power.
Seem discontinued by 2022, after the onsemi acquisition, maybe this onsemi variants are genuine old stock?
https://www.ebay.it/itm/364338821992?mkcid=16&mkevt=1&mkrid=711-127632-2357-0&ssspo=qp9ek5v3tkc&sssrc=4429486&ssuid=GzrqTU42TZa&var=&widget_ver=artemis&media=COPY

Why it's so difficult find a replacement part 😭

The fun part is that this fake items costs as much as the original parts at the time.

[T3sl4co1l]

APT35GN120L2DQ2G looks to be a similar part, and probably a contemporary of it. Still available.

If you can fit a smaller part, NGTB40N120FL3WG is probably comparable enough.

Tim

[masterx81]

Yes i was watching too the NGTB40N120FL3WG, that seem quite similar.
I was watching also some IXYS devices, but any of them seem to have something off.
The APT35GN120L2DQ2G have the same package, but seem a bit under the characteristics of the original part.
By now, seem that the NGTB40N120FL3WG is the best bet. I not think that i have problems adopting the smaller TO247 package, was a problem if i had to monut something bigger....

[masterx81]

Bought all the semiconductors from digikey....
Hope to not have problems...
By now i think that 90% of the FGL40N120AND on the market are counterfait items. Would be nice to test them to death for see if they are at least similar to the original parts... They costs almost the same as a genuine part.
Thanks for the help!

[T3sl4co1l]

You're being too picky -- the variation within a single type is as wide as the difference between these, I think; with the main exception being the APT's higher switching energy (which would be something to monitor, check that the heatsinking situation is adequate, maybe beef it up).  But it's hard to say without the application known, for example the peak collector current rating might matter, I don't know.

I wouldn't be surprised if most of the parts sold are salvage: likely good to ratings -- but specifically, are not new parts carrying any sort of guarantee, and may have latent failure mechanisms like thermal cycling or ESD damage.  Why bothering to remark them, who knows, maybe one shop does that out of habit, maybe another shop does that to launder them (new swapped out / randomized / updated date and production codes), but clearly they got "almost the same as a genuine part" value out of you so the process works.

Still others might sell relabeled cheaper parts (might still work, but not at ratings), completely unrelated types, or complete duds.

They're not after repeat customers; maybe that happens from time to time, but the world (or western, or just US or EU or..) market is so vast that they have an essentially unlimited pool of customers to sell to.

Tim

[masterx81]

I've choose the NGT as seemed more similar than the ATP... It was the primary parameter. Hope that they work... Space is not a problem, also the fixture of the heatsink is not a problem.

Now, i have a 120v 30a 300w dc load, i can generate any voltage with a variac, scope, etc, what tests (maybe, non destructive  )  i can try on them to see if they are something good?

[suntime]

Hi,

Original here:
https://www.lcsc.com/product-detail/IGBT-Transistors-Modules_onsemi-FGL40N120ANDTU_C11754.html

[masterx81]

Hi! How i can know that they are genuine parts? Also the one bought on ebay had the description "Original" 

[Phil1977]

A "semi-destructive" test would be to check the thermal resistance from the junction to the backplate.

Mount the device to a decent heat sink and drive it to the power dissipation of the absolute max ratings. If it´s original it will survive this test, if it´s fake then chances are good it destroys itself.

Unfortunately it´s quite hard to access T_junction of an IGBT by measuring some temperature dependent characteristic. For bipolar transistors or diodes you can just measure the forward voltage, but AFAIK most IGBT parameters are not so well related to T_j .

[masterx81]

A "semi-destructive" test would be to check the thermal resistance from the junction to the backplate.

Mount the device to a decent heat sink and drive it to the power dissipation of the absolute max ratings. If it´s original it will survive this test, if it´s fake then chances are good it destroys itself.

Unfortunately it´s quite hard to access T_junction of an IGBT by measuring some temperature dependent characteristic. For bipolar transistors or diodes you can just measure the forward voltage, but AFAIK most IGBT parameters are not so well related to T_j .

It's rated 200w at 100*C, i can preheat the whole heat sink, and push 100Vdc at 20A. I think that i can leave it in this state only for few seconds... And eventually be prepared to some nice fireworks 

[Phil1977]

Don't go to far, 100V @20A are 2kW 

If you have the possibility it would even be better to go to the 500W / 20°C value. If a fake has a bad thermal connection or even a smaller die, it will most probably fail there.

[masterx81]

Whops 
I'll try them 

[wraper]

It has a very weird surface on the front side, as if it was painted over by screen printing. I suspect they polished the surface entirely prior to applying new parking and then painted outer area. Try rubbing with a strong solvent like acetone, or alcohol if you don't have anything stronger.

[T3sl4co1l]

LCSC is a generally acceptable distributor, if not a first-tier affiliated one, I'm not sure.  I would be a bit curious where they're getting those from.

It seems to me, a lot of older parts either remain in stock, or even production, in foreign markets.  Which shouldn't make much sense, these are global manufacturers and distributors.  And it shouldn't be due to secondary markets (surplus and salvage), but I don't what all sources they pull from, and maybe they do that from time to time.  I assume they're either a primary distributor, or the lion's share of their business is.  But I haven't looked into it for sure.

Might be able to ask them about date of production and sample lot codes for their stock, and in turn contact the manufacturer to confirm such.

Or just check if onsemi lists LCSC as an authorized distributor.

There are many ways to convey to others, the cooperation, good faith, and trust, that given parties share; the independent verifier pattern (A asks B for information on C, A asks C for confirmation about that information) is such an example.  A manufacturer listing their authorized distributors puts them in a liability position: it would reflect poorly on them if the distributor they had authorized, behaves badly, probably among many other things in their supplier agreements.  (The agreements themselves, are private and proprietary, probably for good (internal) reasons -- marketing secrecy, etc., so they can't post their agreements in full to broadcast that cooperation.)  Betraying that trust would most likely have them pull their products at that distributor, a very visible signal to customers.  In either case, both parties are likely to have some representation about each other, that can be found on the website or by contact.  It can take some digging to find things (or where and how to ask about verification like lot numbers and stuff, and some time to await their response) so poke around a bit and see. It's a good research exercise

Tim

[wraper]

According to onsemi website it's in last shipments status.

[ArdWar]

Oh so the left one is indeed just laser bloom I guess

It seems to me, a lot of older parts either remain in stock, or even production, in foreign markets.

For those in Asian countries you'll quickly get used to it. New stocks of old parts that somehow continue to exist way past a NOS supply would realistically provide. It's usually prevalent in repair or word-of-mouth hobbyist (eg. audio) communities.
Otherwise when would you realistically find the last Toshiba TO3 2N3055/MJ2955, or ST TDA2030, or Philips brand CRT TV driver. Some of those stopped production two decades ago.

[suntime]

Hi,

I have ordered many components from LCSC and have never had any problems.
The image of the components that you see on their website and that are taken by them is what is sent to customers.

Regards.

[Haenk]

After googling around for a bit, I can only assume that these are "real" ones. However likely in used condition, sanded down, coated and relabeled. The used parts likely came with long legs, easy to cut away to an unsuspecting "unused" length. The rasterized surface you see is from an UV printer. Why? They can charge more for "new" parts than for used parts - even if it is only a Dollar, that adds up...
So likely fine for a repair job.

[wraper]

After googling around for a bit, I can only assume that these are "real" ones. However likely in used condition, sanded down, coated and relabeled. The used parts likely came with long legs, easy to cut away to an unsuspecting "unused" length. The rasterized surface you see is from an UV printer. Why? They can charge more for "new" parts than for used parts - even if it is only a Dollar, that adds up...
So likely fine for a repair job.

I don't get how you came to the conclusion it should be the same part. Nothing prevents them relabeling a different component with worse ratings that may even work on a first glance.

[masterx81]

The legs are normally long (as a new device), and not seem welded.
Before the power test, i've tried a thing bought time ago on Aliexpress. The scary 3.5kv breakdown tester. Never used before. I've tried it at minimum current, and top at around 1280v. At least the voltage rating seem spot on, but i not think that the device is happy 
Tried also with the LCR the gate-emitter capacitance, and is 3.9nf at any frequency (up to 100khz), but the test on pdf is taken at Vce=30v, Vge=0v and at 1mhz, so i not know how can be compared. In any case isn't too much distant from the 3200pf on the pdf...
I was thinking that the on delay time can be measured quite accurately. I have a 200mhz scope (siglent 1204 x-e), somewhere i have something that can output a pretty sharp square wave... The original part have a specified 15ns Td(on), but with 600vcc, 40A, Rg=5ohm, Vge=15v, inductive load at 25*c. The problem is how recreate this conditions...