How to identify a counterfeit component

[HansGruber]

I am having some issues with a motor control PCB recently, so I grabbed a old working motor controller and started looking at each component and comparing them. I am not sure with one MOSFET I found, they are both supposed to be onsemi FDMS86540 N-Channel MOSFETs: https://www.onsemi.com/pdf/datasheet/fdms86540-d.pdf. I can't find the manufacturers description of the original markings on the package, maybe someone has a idea ?

They are both 5mm wide but one is 6mm long and the other one is 5.5mm long (without pins). Is this just a production variance from a old batch or is this really what I expect it to be. Is there a way to measure this in-circut or maybe after desoldering them ?

[magic]

Your datasheet says 6mm and an older one from Fairchild says 6mm too, and your part must be newer than that because it has the ON logo.
https://html.alldatasheet.com/html-pdf/475048/FAIRCHILD/FDMS86540/2958/6/FDMS86540.html

Not sure why there is no F in the part number and DMS86540 doesn't seem to exist.

Did anyone work on this board before?

[HansGruber]

You can't see the F because of my low quality picture, it's hard to take one with all the conformal coating, I made another picture where you can see it a little bit more. I just remove a little bit of the conformal coating, otherwise it's not been worked on. I would have assumed that the right component on the picture of my first post is the original component (because it's working). But that's actually the one that's 5.5mm long.
The onsemi logo on the right component looks better to me, also the general alignment of the marking font, the left one looks a little crooked and the onsemi logo does not seem so... crisp ? It's just a feeling and hard to proof.

[HansGruber]

I found a dimension that I can measure more clearly, it's from the most recent datasheet. It's supposed to be the "height" of the component. I can measure the same at the right component (~5.6mm) of my first post, the left one is ~6mm.

EDIT: I wrote a E-Mail to onsemi, I wonder how my chances are for a real answer.

[strawberry]

measure DS on resistance with miliohm meter. it must be less than datasheet value at room temperature
dont know how to cheat on resistance to get smaller die
gate capacitance. smaller die will give less capacitance

[HansGruber]

Just as a general question: how likely is it that the manufacturer will change the package size with the markings on it ? I've collected several boards from the past months and years (see attachment). The images are a little distorted because of different microscope zoom and resizing the image etc.
But everything in the past looked similar, except for the "new" batch. Sadly I realized I don't have the equipment to measure it.

[Haenk]

The 6 month old part looks coated, I'd suggest trying to use some solvent to rub it away - if the black color does come off, it's coated and likely fake.

[magic]

I just remove a little bit of the conformal coating, otherwise it's not been worked on. I would have assumed that the right component on the picture of my first post is the original component (because it's working). But that's actually the one that's 5.5mm long.

So we can assume that the board contains exactly the components that have been placed on it by the manufacturer. Question arises: who made this board and where did they get these transistors from?

Did you order the board from China?
Could the manufacturer source components from dodgy vendors in China?

The onsemi logo on the right component looks better to me, also the general alignment of the marking font, the left one looks a little crooked and the onsemi logo does not seem so... crisp ? It's just a feeling and hard to proof.

If the logo or markings are rotated and not exactly horizontal (hard to tell from the pictures, but maybe you are right about the newest one) then yes, it's a common thing on fake parts and it generally doesn't happen on quality parts.

[HansGruber]

The 6 month old part looks coated, I'd suggest trying to use some solvent to rub it away - if the black color does come off, it's coated and likely fake.

I would assume that's just the silicon based conformal coating. It's a little unfortunate that they are all mixed of different conformal coatings (or none), makes it hard to compare them. My main argument was just the general marking, which looks similar on all except the new one.

So we can assume that the board contains exactly the components that have been placed on it by the manufacturer. Question arises: who made this board and where did they get these transistors from?

Did you order the board from China?
Could the manufacturer source components from dodgy vendors in China?

I think our electronics manufacturer could not get it from a reliable source and just ordered something in China. Onsemi actually answered me, they want a picture of the sticker with the LOT-number. I hope they still have the roll somewhere, maybe it's traceable.

[wraper]

Just as a general question: how likely is it that the manufacturer will change the package size with the markings on it?

Often they don't change anything. Just produce the same part at factories in different parts of the world, and then they have some significant differences in package and marking. Sometimes differences are shown in the datasheet, but usually not.

[mikerj]

Your datasheet says 6mm and an older one from Fairchild says 6mm too, and your part must be newer than that because it has the ON logo.

Note that the OP said 6mm without pins.  The datasheet gives a dimension of 5.65-5.85mm for the body, so neither 5.5mm or 6mm is in tolerance.  OP, did you use some precision measurement such as vernier calipers to get these dimensions?

[wraper]

I would remove transistor from PCB, then first measure R_DS(on), if it's fine then also drain-source voltage. Maybe compare gate capacitance measurement with a known good part (or meet all measurement conditions and compare it with datasheet). Fake power transistors are rarely up to spec and usually have a much smaller die.

[HansGruber]

Your datasheet says 6mm and an older one from Fairchild says 6mm too, and your part must be newer than that because it has the ON logo.

Note that the OP said 6mm without pins.  The datasheet gives a dimension of 5.65-5.85mm for the body, so neither 5.5mm or 6mm is in tolerance.  OP, did you use some precision measurement such as vernier calipers to get these dimensions?

I am just bad at measuring, I've attached my new measurement. Maybe it's clearer now.

[strawberry]

that new PCB is counterfeit 100%
poorer quality PCB (made in China)
vias and traces dont match
and all other parts dont match either
if they were honest they put similar spec transistor (made in China) with fake ON semi markings ...

[mikerj]

I am just bad at measuring, I've attached my new measurement. Maybe it's clearer now.

It's clear the newer MOSFET is unlikely to be a genuine part!

[wraper]

I think it's actually a different package. Old one is PQFN8 5X6, new one should be similar to DFN 5X6E (ALPHA & OMEGA semi) http://www.aosmd.com/res/data_sheets/AOE6930.pdf which has no protruding terminals. But Infenion makes both type and calls both with the same name ThinPAK 5x6 SMD although they are obviously different.
With protruding terminalshttps://www.farnell.com/datasheets/3629119.pdf
Without themhttps://docs.rs-online.com/f5c7/A700000007497990.pdf

[wraper]

Actually ON also uses both packages and calls both PQFN8 5X6, 1.27P
https://www.onsemi.com/pub/Collateral/483AE.PDF
https://www.onsemi.com/pub/Collateral/483AF.PDF
Although more specifically one is CASE 483AE, another is CASE 483AF. However the datasheet you provided says specifically PQFN8 5X6, 1.27P 483AE in the drawing. But in ordering information ON datasheets just says PQFN8 5X6, 1.27P.  IMHO the right question you should ask manufacturer is if particular transistor may come in anything other than CASE 483AE.
From this there are at least 10 variants of this package https://www.onsemi.com/pub/collateral/brd8011-d.pdf

[HansGruber]

The most recent (May 2015) datasheet (from the official onsemi website) states that the used case is 483AE, like wraper said. Based on the reel sticker of our components, the MOSFETs were produced week 42 in 2020 (so they had 2 years to refresh the datasheet). I am in contact with onsemi, if the reel itself is genuine. If the reel is genuine, I will ask them about the different case, if it's possibly a different one.

[HansGruber]

Waiting for the answer of onsemi made me bored, so I made a poor man's die analysis, I hit it with a hammer and chisel to professionally remove the package. Except for the package, I also remove the copper "hat" that was sitting on top to connect the lower source/gate pins. The "new" part got a unlucky hit to the top right, it would normally look different.  Overall it looks different to me, but I am really no expert.
I did it to 3 different ones, just to be sure they don't look different purely because I hit one harder than the other one.

EDIT: onsemi just confirmed it's fake, the LOT number does not exist

[magic]

This is the professional way
https://www.richis-lab.de/Howto_Decap.htm
https://www.richis-lab.de/Howto_Decap_Ofen.htm

BTW, no fancy gas torch is strictly required - even a lighter may work.

A bunch of fake IRF3708 from that guy:
https://www.richis-lab.de/Transistoren_FET_pwr.htm

[strawberry]

22 ohm resistor
new 5% old 1% even not necessary precision there

it s clear that board was made to be close to original (and work for 1 month to not receive refund) for some dumbass to not notice any difference and think he got lifetime deal
even original may hide some design flaw but this new board may not be even as good
less via count less current capability

[wraper]

22 ohm resistor
new 5% old 1% even not necessary precision there

it s clear that board was made to be close to original (and work for 1 month to not receive refund) for some dumbass to not notice any difference and think he got lifetime deal
even original may hide some design flaw but this new board may not be even as good
less via count less current capability

Do you net get it's their own board? They had a problem with counterfeit components.

[strawberry]

they didnt stole it , right?