Let’s Play Spot The Fake MOSFET

Recently, the voice push to talk circuit in [Ryan]’s BITX40 radio was keyed down for a very long time. Blue smoke was released, a MOSFET was burnt out, and [Ryan] needed a new IRF510 N-channel MOSFET. Not a problem; this is a $1 in quantity one, but shipping from Mouser or Digikey will always kill you if you only buy one part at a time. Instead, [Ryan] found a supplier for five of these MOSFETs for $6 shipped. This was a good deal and a bad move because those new parts were fakes. Now we have an opportunity to play spot the fake MOSFET and learn that it’s all about the supply chain.

Spot the fake

To be fair to the counterfeit MOSFET [Ryan] acquired, it probably would have worked just fine if he were using his radio for SSB voice. [Ryan] is using this radio for digital, and that means the duty cycle for this MOSFET was 100% for two minutes straight. The fake got hot, and the magic blue smoke was released.

Through an industry contact, [Ryan] got a new, genuine IRF510 direct from Vishay Semiconductors. This is a fantastic opportunity to do a side-by-side comparison of real and counterfeit semiconductors, shown at right. Take a look: the MOSFET on the left has clear lettering, the one on the right has tinned leads and a notched heatsink. [Ryan] posed the question to a few Facebook groups, and there was a clear consensus: out of 37 votes, 21 people chose the MOSFET on the left to be genuine.

The majority of people were wrong. The real chip looked ugly, had tinned leads, and a thinner heatsink. The real chip looked like a poor imitation of the counterfeit chip.

What’s the takeaway here?  Even ‘experts’ — i.e. people who think they know what they’re talking about on the Internet — sometimes don’t have a clue when it comes to counterfeit components. How can you keep yourself from being burned by counterfeit components? Stick to reputable resellers (Mouser, Digikey, etc) and assume that too good to be true is too good to be true.

85 thoughts on “Let’s Play Spot The Fake MOSFET

  1. Usually going for the device with the better looks (mold quality, unused look, labelling quality including crisp and consistent fonts, non-chinglish description and overall material impression) is the way to go. Real takeaway here? Vishay, get your sh!t together and get proper labelling gear for your $1 a pop devices. If you want customers to spot fakes and use YOUR genuine parts, set the standard in product quality instead of being subpar to some russian backyard chip flippers.

  2. I agree about the shipping. Its usually not the actual price of components that makes you search elsewhere, its the ridiculous cost of shipping. I don’t understand how I can buy something from Hong Kong these days with practically zero shipping yet it costs so much to send something within the same country. Yes, I realise the item is coming form a different country where basic costs are different but the last leg of its journey in the destination country is almost identical. Something doesn’t add up to me.

    1. Due to an old treaty intended to keep international correspondence cheap small packages ship to the US from overseas are shipped at a loss within the US. Combine this with cheap shipping out of China and tiny devices shipped from China are effectively massively subsidized.

        1. And here I thought it was for enabling growth in the export sector thus enabling growth in other economical sectors having positive effects on the manufacturing sector while expanding the market for Chinese products.

          But of course it is for evil dumping of “cheap” stuff by the rus… chineeze. Yeah.

    2. Not to dive too deep into global politics here, but the answer to your question is that the postal systems are being abused. The Treaty of Bern (1874) created a bilateral agreement where the delivery half of international mail is basically done for free by the destination country (we deliver your’s, you deliver ours). Certain countries have reduced their international shipping prices to effectively zero in order to encourage exports.

      Now we have a situation where I can order something from China to the US and have it delivered for less than the cost of the envelope if purchased locally.

        1. Never – there’s no leverage. The direct options are to cut off international mail delivery or to reduce internal mail rates to zero. Alternatives are to start trade wars in other areas. The U.S. frittered away it’s dominance in semiconductor mass manufacturing which made this all possible in the first place.

          1. I have noticed, that Canada sits on the packages coming from China, long enough to pass the 60 day (or whatever) guarantee from the seller, to make it long and tedious, so people would not buy from China.
            Smart but evil.

        2. In The Netherlands, the postal services just deliver the package to a pickup point, instead of delivery to your home.
          That sucks sometimes, because those pickup points are mostly regular shops, with 9-5 working hours, so most often, you’ll have to wait for the weekend to pick your stuff up.

      1. Not completely true.
        In the case of China’s ePost services we owe it all to ebay / Amazon and their Lobbyists.
        They made sure that the ePost system was accepted by the members of the postal union.
        Then it took Chinese sellers about 24hrs to figure out how to use this system without selling on ebay / Amazon.
        It’s a bulk system where you pay almost nothing (by western standard) when handing in large quantities of small packages.
        They even have a complete choice what carrier to use!
        There are shops for DHL, Deutsche Post, NLpost, UKpost, etc. in Shenzen for example. If you have 100 packages for Germany you just hand them to Deutsche Post directly. In China!

    3. True but there’re distributors shipping for free, too. I find myself constantly using Arrow to source parts in Germany with free express shipping, rather than Farnell, Digikey or any of the other ones with local branches due to their crazy shipping costs if you just need a few parts. The only downsides of Arrow are the huge amount of garbage they produce with their packaging as well as their international shipping incompetency; which did cost me a decent chunk of money twice already because they messed up the declaration triggering the transportation companies to step in with their premium customs handling service…

  3. I noticed something interesting in the dollar store the other day– once you go below a certain cheapness threshold, products start to look fancier and shinier because they have to have the “it’s pretty” factor to convince you to trust the low prices. The super high end and super low end stuff is where it makes economic sense to make it look super shiny and new, and midrange items have a dip in “straight out of the box” shininess. Someone else probably named this obvious fact before me, but if not, I christen the phenomenon “Ginger’s dip.”

    1. “Fisher-price toys” is what I call those shiny “Pretty, curvey, cheap plastic-y and/or so-called thin-n-lightweight” things… that or “peasant grade”, however peasant grade is designed to break and not designed to look like a toy you’d give a kid:

      Apparently (I hardly remember though) when I was just a toddler, I’d been given a fisher-price tape+radio and just over an hour there was splinter remains of said “Tape+Radio”,
      Given a plain generic dark grey radio+tape recording and playback machine (Stereo, plain square edged ghetto-blaster style)…. I’d treat is as such: a machine.
      I’d used it for its purpose, kept it clean, polished and I made sure to do nothing that could damage it. I actually remember the dark grey stereo oddly enough, had it from around 3-ish yrs to late 6years old, then that GVT-issue happened lol.

  4. I took a different appearance tactic. Which one looks cheaper to produce? Assuming that imitators are not going to spend a lot of money making the counterfeit, the one that looks like it takes additional steps should be the genuine one. It may be “ugly”, but notching the heatsink and tinning the leads are extra steps that cost more money in production, so it can be assumed fakers would skip those steps to increase profit.

    1. I’m not an expert either, but I immediately assumed that the tinned leads would be more expensive to produce and therefore were the real part. It is very unusual for the Chinese copy to have clearer lettering than the real thing though.

        1. It may also hide rust under the tin. For me it is a minus as the thinned pad may not fit the hole compared to the untinned, and heating all the three leads at once it is hard or can damage other parts.

          1. The tin does not expand the dimensions of the legs equally. Most of the tin will be int he flat centre parts of the pins. If you use round holes and follow the manufacturer’s datasheet for sizes then tinning won’t make a difference.

    2. I mostly agree, but tinned leads means (for me) that this component has been salvaged from a board (which is quite common from untrusted chinese suppliers because of the cheap labor cost).

      1. That wouldn’t make it counterfeit though. Sure it may not meet life expectancy anymore, but it should still perform as original, … assuming it wasn’t originally counterfeit.

    3. Exactly!
      I was quite surprised the majority was wrong. Other differences pointing to the right one being the genuine one are the imprinted “99” and the better/shinier looking heatsink (the left one looks oxidized or sth.).

      1. The 99 stamp is the only part of the mold less than a year old, so it’s crisp (It’s an interchangeable insert for production date, production run, whatever). The rest of the markings are in the mold since it was made, maybe 20 years in production, so they look ugly. In the counterfeit, they are printed/stamped in the part, so they can use the same mold for all the parts with the same casing.

    4. That’s exactly it. Things like notches, polished heat-sinks, embossing and tinning cost money. Counterfeiters will only create shininess as long as it doesn’t cost extra or is required for the perfect illusion, like when rebadging a cheaper part. I can very well imagine that this MOSFET was only coated with epoxy and then imprinted, thus yielding the bland looks…

  5. Admittedly the tinned leads and poor markings had me at first sight, but the engraved “09” dissipated most doubts. Fake transistors and chips are always sanded before being relabeled (so you often see rounded corners). The one on the rignt has engraved markings which would be hard to cancel without ruining the “09” marking.

  6. Not sure why the conclusion is that its a fake and not a defective part, mislabeled part or one of failure, especially since one of the commenters mentioned their true parts looks like the one that is supposedly fake…

    1. …which shows a somewhat shiny generic image with notched tab but no stamping on the top. The technical drawing doesn’t show notches at all and there’s not a single mentioning of “tinned legs” in the entire document. Also, there are two real images from different factories on the bottom: One with tinned legs, one without. One with clear (but light) laser-etched labelling on a matte body, one with thick print in shitty quality on a glossy body. One has a round notch on the tab, one a square notch. One has a shiny metal tab with rounded edges, the other one is dull and probably oxidized, but not grinded down. If you’d show people these two pictures and ask which one is fake, I bet 90%+ would choose the part from the Wan Hung Lo factory in Xi’an.

      Let me conclude that once again: If you make products that look like fakes, don’t expect the customer to be able distinguish them from real fakes. And if those fakes damage your reputation, deal with it.

  7. We have gotten counterfeit components Jameco.

    The last one was in a LM317K in the TO-3 package (which is obsolete*, so a good target to counterfeit ), we received a mix of real with counterfeit.

    Counterfeit parts would fail open at around 100mA.
    But you could tell the difference by the weight, the real parts where about twice as heavy.

    Unfortunately, the parts got shipped to the contract manufacturer and we did not discover that they where counterfeit until electrical test.

    Jameco, finally, after much back and fourth (months), acknowledged that they where counterfeit and offered to replace the parts… If we returned the counterfeit ones… Which where already soldered in assemblies… mixed with real parts…

    So another contract manufacturer run to unsolder all the parts (around 500), determine which where counterfeit (by weight), ship them to Mouser, get replacement parts, double check that they where real, and then solder them back into the assembly.

    Honestly, these things happen…
    But the difference is how the company handles the customer service end of things, and thus, Jameco is no longer in our database as a “preferred” vendor. ;)

    *The assembly is old, like 20+ year old design, but management did not want to spend the time and money to re-layout the board to accept a more modern part, because, this was a “last run”.

    In the end, I think it would have been cheaper to layout a new board for a different part. ;)

  8. I have to agree that I was wrong. I think it was the notch that I thought was wrong. Most of the ones I have seen in older electronics vary between the two but for whatever reason my brain said “notches bad” lol. I’ll have to look thru my parts bin later and see what all I have. These posts are fun :)

  9. I initially thought to myself: Either one could be fake, must this be a trick question?
    At first glance:
    The left hand one looks like it was made in the 1970s-80s era and the other part looked like it was a more recent part but not too recent, possibly 2000-ish as though they’re going through some machining changes.

    The one on the right is the one I’d trust the most as it has an engraved date code 99 and the casing looks like a transition period model hence my first guess above.

    On the other hand the left one has slightly slanted text (not uncommon for 70s and 80s components) so I’d give the thing a fail for aging alone as technology improved up until around 2010-ish when things are more designed to blow up in peoples hands and look pretty whilst doing so.

    I’d never guess either one to be fake… at least until I check the datasheet to be sure. I’ve seen a lot of I0R (International rectifier) mosfets in a lot of OK to decent quality PSUs and in some high quality PSUs.

  10. This one was too easy, since I have used similar IRF parts before. The one on the left looks like the standard chinese manufacturing (very ‘clean’ and generic looking with universally matt coated leads, and laser etched markings). You know you are getting a part with some history behind it has been stamped and tinned since no one in their right mind does that anymore.

  11. I have tons of fake MOSFETs from Chinese sellers, usually I just measure the Rds(on) first and if it’s much higher than the datasheet says I have to open a dispute on Aliexpress and ask for a refund. But the good ones weren’t necessarily more expensive and sellers that sold me FETs that matched the specification sent me obvious fakes of other FETs, so it’s hard to tell them apart before they arrive.

    1. Quality control testing on receipt for critical parameters seems the safest bet: expensive for manufacturers but not so bad for hobbyists working with far fewer parts, provided we have the measurement tools and techniques.

  12. If you want to see fake mosfets, open up an e-bike controller. I just bought two 18-fet controllers for $40 each. There’s 18 fets in it that are all supposed to be the same IR part, but at least 3 different types of markings.

  13. Not all parts are fake, but you can get a run for your money with genuine parts too. I got a run of boards out of china, (built for us) that used FTDI chip parts. 132 out of 148, did not work correctly. Every kinda failure you can think of happened.

    What happened in the end. The purchasing agent bought what they thought was good genuine FTDI parts. Well they were but not “prime” virgin parts. What they got and used on our boards was re-programmed parts. The programming place un-tape the parts programmed they and re-taped reeled them. They most likely put them back on the FTDI reels too. And resold the excess back into the supply chain.

    What I found. BTW all working parts had the 0403 VID vendor code.
    1. Different PID, not 0000 these were genuine FTDI parts.
    2. Different vendor name. Found all the vendors on the internet.
    3. Different device descriptor. Found Opus Canada hospital university. Good way to keep out UN-authorized equipment.
    4. Other device driver required. Not the default.
    5. External crystal checked off. No worky, until a crystal is installed.
    6. Default current lower and higher than 90ma.
    7. Device serial numbers way out of normal. Like “UUT###”.

    All these parameters can be changed using the ft_prog program from FTDI. BTW, the fake FDTI chips are still milling about in china. So be weary.

  14. Most expensive part in those beefy mosfets is the silicon. Check the prices for 1A and 10A mosfets with same voltage ratings from reputable dealers, they differ by the order of magnitude for the same looking (and genuine) parts with different current ratings. Fake ones can look very nice, but they will blow up at much lower currents than they’re rated. Search for “fake transistors” on google images to see the difference between fake ones and genuine.

  15. Why are fakes assumed to be a homogenous group?

    I’ve had ouright duds or empty casings.

    I’ve had real scavenged parts, drops from a P&P machine, and cut leads.

    I’ve had new and improved parts sold as old ones. (Everyone knows what an 5532 is but no one knows about the millions of cheaper and better op amps that are pin for pin compatible.)

    I’ve also had straight up “midnight run” parts.

    I’ve had 2nd tier OEM copies of popular parts

    And the list goes on.

    What in this group is fake?

    It’s clear on the most obvious sides, but the middle ares are much more grey.

  16. The entire premise of this article is foolish and superficial. Only a fool judges a book by its cover.

    Experts, such as Ben Krasnow, identify a device by it’s behaviour not by it’s appearance. It’s performance curves, as Ben demonstrates in the video from this recent HAD article. http://hackaday.com/2017/07/03/tiny-light-bulb-flasher-vies-for-worlds-record/

    As a general rule once you have audited a supplier you still need to regularly check the quality of their products and it doesn’t mater if it is silicon or hamburger meat, the quality control principles are the same.

  17. I question whether the failure was at all related to the fact that the replacement part was a fake. This looks more like either a design fault or a wrong component value in the circuit that biases the MOSFET.

    Note that the original part failed, then the counterfeit part also failed, apparently in the same manner. This can happen in MOSFETs when you’re using them as switching devices and don’t give them enough Vgs to completely turn them on. This puts the MOSFET in a resistive regime, which will cause the device to run hotter than if it’s properly driven. It IS possible that the fake has a higher Vgs threshold, which would account for running hotter, but then, why did the original part fail?

    Also, if the parts only fail after running for a while, this is most likely due to the part dissipating heat faster than it can be removed. This can of course be caused by either dissipating too much heat (as described above) OR not removing the heat when necessary. Many people think “it’s a 10 A part, so I shouldn’t need a heatsink if I’m only pushing 1 A through it”, but this is bad engineering. For a radio transmitter, even if the expected duty cycle is 50%, there’s no excuse for designing it so that it can’t be run at 100%.

    1. After taking a look at the circuit and datasheet, I’m in agreement with you. This is a bad design with respect to the FET. Looking at the datasheet reveals what really happened. Vgs threshold could be as high as 4V. This circuit only allows for a *max* of 5V if RV1 is fully turned up. Then take a look at the plot on the datasheet for Rds ON vs temperature with a Vgs of 10V (how the device was characterized) and see what happens…. the heat in that FET just kept on rising until smoke. Not even a current limiting device in there for biasing…. FETs like these can take a lot of abuse, but they will fail at some point after they burn your fingers and melt solder on the pads.

      1. To Be Fair, the BITX40 is designed as an SSB transceiver, and I’m using it at a much higher duty cycle with a larger heat sink than what it came with. The focus of the article was that of the counterfeit pieces, not the original cause of the failure which was me bumping up the PA voltage from 13.8v to 20v to get more power out of it. Fine for SSB, not for digital modes that are 100% duty cycle. Not a design flaw- It’s a modified radio. At 5w output with 13.8v it works great, although I do have to watch the duty cycle until I get a bigger heat sink.

        1. geocrasher: as guest points out, this wasn’t really the appropriate part for this application anyway; either a MOSFET with a lower Vds threshold or other measures should have been taken to make the circuit more robust.

          Whether or not this had anything to do with the failure, it was at least implied in the article (“This was a good deal and a bad move because those new parts were fakes”). This was the part I was disputing.

  18. Back in the 60’s, I worked for the special products division of Ferranti Packard. They had an incoming inspection department that sampled and qualified just about every electronics part that came into the company. I doubt that non-mil-spec outfits do that any more.

    If someone sends me some good and fake MOSFETs, I’ll put them on our curve tracer and see what the differences are in the characteristic curves.

    Peter, Syscomp Electronic Design: http://www.syscompdesign.com

  19. I recognised the real part right away because I see IR parts all the time in power supplies. They laser mark the real parts because it’s actually cheaper in large quantities than pad-printing them, and it also allows the application of lot numbers in the same process step.

    1. Also very common to sand off the original serial number, and print a bogus one over top like this article. Many larger FETs will use a proprietary pin sizing and spacing to help identify bogus knockoffs. However, we always check the casting marks on plastic encased parts…

      I guessed right too, but assumed the 90’s packaged part being compared is so old it may have been recovered from a recycling source. Authentic parts recovered from equipment is very difficult to QC, and it is very common for manufactures to no longer pre-tin leads on parts.

      We always sample parts from a new seller, and when possible test them to failure prior to buying a roll. Having a golden-standard sample part to compare electrically is the only true way to know for sure as even supply houses (digikey/Newark) will get bad spools and or production lots once in awhile.

      How to use an Octopus Curve Tracer, and a very simple way to spot out-of-spec parts without destructive QC:

  20. Let me understand this. We have a device that cooks a legit part because he holds the key down “too long”. Ok, so we have some halfwit who designed the thing to begin with, but that to go and replace the part with a new one after the old one melted. Um, am I the only one seeing the futility in this? Before even starting to argue about real vs fake, I would be more concerned as to why the thing croaked in the first place.

    1. “The focus of the article was that of the counterfeit pieces, not the original cause of the failure which was me bumping up the PA voltage from 13.8v to 20v to get more power out of it.”

  21. Maybe it is worth pointing out, that Vishay is also not the one to sell the original part.

    The name already suggests it: IRF=International Recfifier
    http://www.irf.com/product-info/datasheets/data/irf510.pdf

    Vishay simply cloned this device and sells it as a second source.

    In some sense the chinese cloning is often done in a similar vein. However, these companies often lack brand recognition (rightly so) and have zero quality assurance. So copy the brand as well…

    1. BRIAN: [cpldcpu] makes a very good point. A Vishay manufactured part SHOULD look different from an IR-made part. What evidence do you (or [Ryan]) have that the part marked “IR” is a fake???

    2. As far as I can tell, the genuine IR part is no longer available. Infineon bought IR and I cannot find data for IRF510 on their website. (Admittedly I did not spend a great deal of time on this.)

      Fake parts are a problem in other industries. A few years ago Vauxhall managed to get fake copies of their own braking parts into their official supply chain.

  22. 21 out of 37 isn’t a clear consensus, it’s two people away from a 50:50 split. It’s less than 57%- at that small a sample size you could probably get the opposite results if you asked another 37 people. I’d roll my eyes at that percentage being called a “clear consensus” on a shampoo commercial.

  23. Probably even more common are the chinese DC-DC boost and buck converters you see on amazon and ebay.

    I compared the chinese “LM” parts to legit LM parts purchased direct from TI.

    100% of them use counterfeit LM2596 / LM2577.parts. ONE HUNDRED PERCENT.

    It’s pretty easy to tell the fakes though – you can’t build complete DC converters with genuine parts for what they are blowing them out on ebay and amazon for. You can’t get the genuine part that cheaply even in quantity 100,000.

    1. Yes, of course, it’s just to cheap. But the actual question (for hobbyists!) is: Do they work? If there is a general rule of thumb like “works for specified current divided by 5” a lot of people won’t care where the IC is from, as long as it is cheap and does his job.

      Btw, has anybody experience with these extremly cheap AVR (just the chip or soldered to some “Arduino”)? Are they fakes or maybe genuine but out of spec parts somebody retrieved from a dumpster? The prices are really interesting, especially because you get a “big” AVR soldered to a breakout-board basically, but what about the quality? Debugging µC-stuff can be a real PITA if it’s actually some part of the IC that doesn’t work right.

      1. Indeed. I’ve ordered a few AT90USB1287 just to find out that they all were recycled and also the ones I’ve actually tried did not work at all; I’ve wasted a lot of time on those guys up to the point where I’m sick of using the AVR stuff at all and am probably going to switch to Cortex-M MCUs which are a lot trickier to program but much easier to handle and more capable…

        1. LM2596 and friends are not for power supplys. They are for low voltage DC-DC, like powering your RaPi in your car from 12V. I may purchase cheap DC-DC modules but certainly not cheap power supplys or other stuff connected to the mains!!

          1. You are, of course, correct. You also completely missed my point.

            My point, that I attempted to convey with a lighthearted tone rather than to spell out in painful detail, is that saving money on some small part of a project is false economy when that part has the potential to damage or destroy significantly more expensive parts of the project when/if it fails.

          2. DKE: I understand your point. It’s effectively something to consider. But at least with low voltage there is less risk for life beeings than with mains stuff. The risks for (expensive) electronics are however a different thing that shouldn’t be forgotten too.

      2. the fakes let out magic smoke at power levels far below the genuine part spec. as long as you never use them anywhere near their datasheet specs, you may be ok.

        however the fakes switch at frequencies far, far lower than spec. this may cause interference issues in your devices if you are expecting switching frequencies of the authentic part. swap the real part in on the chinese DC board, and the switching is the correct frequency.

  24. Does he actually _know_ if EITHER is fake? The superficial packaging differences the he spotted might be normal difference between two different facilities of the same manufacturer. Or perhaps Vishay uses different packaging sub-contractors.

    To spot a real “fake” you might need to look at the silicon die itself. But even *that* might not work. Thirty years ago, a LOT of DRAMs were were manufactured by one company but packaged and sold by another (having production problems, etc). [This is from my memory, but I recall that Fujitsu made a lot of parts for Intel.)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s