Choosing IGBTs, and adventures buying FGY75N60SMDs from Aliexpress

[Hydron]

As part of the design process for my (under construction) QCW coil I took a long look at the different TO-247 (and similar) IGBTs available, and ended up deciding to buy some FGA60N65SMD parts from Mouser, with FGY75N60SMDs coming in as a second choice, mainly due to significantly increased cost. The criteria and pros/cons I used to decide this are quoted below from one of my 4hv posts (see http://4hv.org/e107_plugins/forum/forum_viewtopic.php?180799.0#post_180879 for full thread context):

I have just ordered a bunch of FGA60N65SMDs myself after quite a lot of comparison and evaluation.

I had the following wish list when picking IGBTs:
- Reasonable Qg to make driving at 350kHz for 10-15ms at a time not too difficult
- Fast, with low Eoff (will be some hard switching - OFF only - in QCW)
- Cheap (am running multiple parallel bridges, so needed a few!)
- Low Rjc to keep die temperature swing down

After hours of searching I ended up back where I started - overall the standard choice of 60N65s(/60N60s) were the best option, and are well tested. Second pick was the FGY75N60SMD, with the following pros/cons compared to the 60N6x parts:

Cons:
- about 50% more expensive
- more Qg (though not excessive)
- slightly higher Eoff per amp, though significantly lower fall time, so a little dubious about how much difference there really is here
- no mounting hole for those who need this (am using clips so I don't care)
- no 650V rated option, only 600V (though I suspect the 600 vs 650 is mostly packaging related, and can likely be disregarded)

Pros:
- no mounting hole to reduce heatsink contact area
- lower Rjc thermal impedance, especially for the co-pack diode (which has less than half of the 60N65 value). Diode is only important if using freewheeling or phase shift QCW
- maybe easier to find?

I was very close to getting the 75N60s, mainly due to the thermal impedance stuff, as I plan to use the freewheeling diode quite a bit for phase shift QCW modulation. In the end the 60N65 seemed likely to be good enough, and if I have issues then there's still an upgrade path!

As a heads up, stock seems to be a bit scarce of some of these parts - Mouser had the best price on the 60N65s so I bought from them, but they only have 11 left now and a 1 YEAR lead time on more!

The Mouser stock of 11 mentioned is now completely gone, but as suggested by loneoceans, it's also worth checking the other more wholesale orientated distibutors like Arrow/AVNET too for this sort of thing (I would have saved about 10% had I done so).

This brings me to the Aliexpress part of the story - having ordered a bunch of legitimate 60N65s, I thought I'd look on Aliexpress to see if there were any suspiciously cheap 75N60s to buy as a second option

Playing Aliexpress roulette with the larger 75N60 parts vs 60N60/60N65s seemed to have a few pros/cons:

- Fewer sellers of the larger parts mean it's easier to sort through the options, but less choice of sellers/quantities/prices
- There aren't any really cheap "Power TO247" parts that could be relabeled for a counterfeit 75N60, compared to the much more common TO-247/TO-3P packages used by the 60N6x parts. Anything that actually has a real IGBT die inside should have a current rating at least as high as a 75N60, and there's probably not much to be gained by swapping something else for what is one of the smaller/cheaper Power-TO247 parts.

Sorting the options down to sellers that:
a) Had a picture of the real part, not something with the Fairchild "F" missing or just a completely different package
b) Offered singles or small lots (up to 10 parts) at a reasonable price & shipping rate
Gave 2-3 options of sellers to purchase from.

In the end I chose these: https://www.aliexpress.com/item/10PCS-FGY75N60SMD-600V75A/32327762387.html because the seller had much more of a history than the others, with many positive reviews to go with a few bad ones complaining of fakes etc (of different products, not the 75N60s). Price ended up being 1.75 USD each, or about 30% of what I would have paid from the normal sources (e.g. Mouser, RS etc) - cheap but not so insane that it MUST be fake. US large quantity prices (without ~20% EU sales tax) are much closer to what I paid than what I could get buying 10 in Europe.

The 10 parts I ordered (1 lot) shipped within a few days, with the post taking about 2 weeks. When they arrived I noted the following:
a) All 10 were just thrown in a single (non-antistatic) bag inside the padded postal envelope. Some legs had minor bends and the epoxy of the packages had a fair number of light scratches in it, possibly from shipping with no protection from the other parts in the bag.
b) The packages exactly matched datasheet information with the exception that the legs were longer than specified.
c) No evidence of any sanding/grinding of the package for re-marking. All expected tooling marks were present, including moulded numbers/letters in one of the recesses.
d) Laser engraving of part number exactly matched that of other Fairchild/ONsemi IGBTs I have, e.g. shape of letters, where the laser started/stopped (see pic)
e) Testing all 10 using a multi-meter showed expected behavior of an IGBT - 1 diode drop when measured backwards, open-circuit in forward direction until the gate is charged at which point they showed approx. 1 diode drop in that direction too.

So far so good!

For thoroughness I wanted to do a couple of extra tests. The first was to sacrifice a part to mechanically decap and check that die size was realistic. I was very happy to find that searching for "FGY75N60SMD die size" gives the bare die datasheet: https://shop.micross.com/pdf/fairchild/igbt/FGY75N60SM_IGBT_DIE.pdf as the first result; exactly what I needed to comprehensibly confirm that the parts I'd ordered were real, or at least QC rejects, rather than a completely different part. Grinding down the package revealed a large IGBT die of the exact dimensions given in the bare die datasheet, along with a smaller diode die. The diode die was much thicker than the IGBT, which also meets the expectations of a "Field Stop" IGBT die, which is significantly thinner than the original wafer (75 microns in this case).

I also used a "Megger" to test the parts for OFF-state leakage at up to 650V, getting much less than the datasheet value of 250uA (cannot remember exact value, but may have been in the single digit uA).

Finally I put them in a half bridge and ran them up to ~150A peak current. No smoke was emitted, and the gate charge times were approximately double those of the FGH40N60SMDs I was comparing to, which matches the gate charge increase between the 40N60 and 75N60 parts. Soft switching times were significantly slower on the larger parts, but hard switching was much closer, so everything seems correct.

I have subsequently bought 20 more parts from the same seller, and they seem physically identical to the first lot, and multi-meter test passes as well, though I have not done any other testing. Obviously I cannot tell whether the seller will continue to supply real parts or not, but at under $2 each (less when the seller has a sale) they might be worth the risk!

Hopefully the info about what I bought and the tests I did will be of use to others tempted by cheap deals on Aliexpress/Ebay - I know there are a lot of fakes out there but it seems that I got lucky this time!

Shown in the scan below are a number of the 75N60 parts (included the one I ground down to show the die), along with some 40N60 and 60N65 parts for comparison.

[Mads Barnkob]

Thank you for the summation of this though-rough is-it-a-fake test, considering he has 997 lots left, there is plenty of supply left, could it be that he got a very large whole sale discount on buying 10000?

Also, do you have two users on AE? Since there is only these two buyers listed and it fits your buys exactly?

E***.
UK
2 Lots
06 Dec 2017 00:57


T***.
UK
1 Lot
20 Nov 2017 03:25

[Hydron]

Yes, signed my girlfriend up for the second order to get the Aliexpress new user bonus. Though it may have been a mistake to introduce her to a new shipping site!

[futurist]

Yes, signed my girlfriend up for the second order to get the Aliexpress new user bonus. Though it may have been a mistake to introduce her to a new shipping site!

[T3sl4co1l]

Reasonably well tested, cheers!

I point this out from time to time:

- If you need a thing, buy it from a reputable supplier.  You'll get it in a few days, and it will work (well, to the extent that the circuit it's in will work, but that's a separate issue ).
- If you purchase the cheap probably-counterfeit option, don't freaking come to us and complain that you got shafted.  Of course you did, you're too damn cheap to buy the real thing!  And that means you're not worth our time to hear you whine about it. 
- Or if it's a risk you want to try, and you want to be wise about it, then you need to perform a receiving inspection and qualification test.  Just like the above!  Downside: you'll spend somewhere between a few hours (if you have everything handy to test with), to several months and $100k (for a production tester), to set up that testing.  In that time, how many dollars of labor have you just spent saving those precious pennies?  More often than not: "you just played yourself, fool!"

But alas, nothing new under the sun; "penny wise and pound foolish" was already old in its day.

So... given all that, I wonder how most sellers would react, if you make it clear that you will be doing this testing, and withholding payment (if net or escrow) and ratings, contingent on that approval.  Assuming you aren't pressed for time, would they keep sending you stuff until it gets approved?  Would they disappear after one or more failed attempts?  Would they decline your offer in the first place?

Tim

[Hydron]

Agreed about using reputable suppliers for commercial production - even for hobby work the time and effort to test stuff bought through non-official channels adds up.

I mainly got the parts from Aliexpress as a fun experiment - I already had some other IGBTs purchased from Mouser to use if the Aliexpress parts were fake (I may still use them depending on the difference in switching times/behaviour between the 60N65 and 75N60 parts).

[Laci]

I ordered 10pcs IRFP460 from ebay a month ago and they arrived this week.So far,all of them seems to work.They were really cheap,6.35$ for 10,so 0.635$ for a single one.The shipping was also fast,unlike the last time,when it took around 5 months to arrive for some other components...The only negative aspect is that their legs were soldered,so probably they are used,but they are working.

[panjaksli]

Don't buy FGY160T65, 20$ per 5 pieces.
They're definitely fake, the die is even smaller than those fgy75s have.. (all 15 I bought)
They should be 8x6mm, instead they are only 4x5 or so.