How Good (Or Bad) Are Fake Power Semiconductors?

We all know that there’s a significant risk of receiving fake hardware when buying parts from less reputable sources. These counterfeit parts are usually a much cheaper component relabeled as a more expensive one, with a consequent reduction in performance. It goes without saying that the fake is lower quality then, but by just how much? [Denki Otaku] has a video comparing two power FETs, a real and a fake one, and it makes for an interesting watch.

For once the fact that a video is sponsored is a positive, for instead of a spiel about a dodgy VPN or a game involving tanks, he takes us into Keysight’s own lab to work with some high-end component characterization instruments we wouldn’t normally see. A curve tracer produces the equivalents of all those graphs from the data sheet, while a double pulse tester puts the two transistors through a punishing high-power dynamic characteristic examination. Then back in his own lab we see the devices compared in a typical circuit, a high-power buck converter. The most obvious differences between the two parts reveal something about their physical difference, as a lower parasitic capacitance and turn-on time with a higher on resistance for the fake is a pointer to it being a smaller part. Decapping the two side by side backs this up.

So it should be no surprise that a fake part has a much lower performance than the real one. In this case it’s a fully working transistor, but one that works very inefficiently at the higher currents which the real one is designed for. We can all be caught by fakes, even Hackaday scribes.

12 thoughts on “How Good (Or Bad) Are Fake Power Semiconductors?

  1. Not as bad as I was expecting to be perfectly honest.
    If only they’d release it under a different name with the true specs, people would probably use them in actual designs.

    1. It sounds like they do exactly that.

      The problem is that they don’t make dice that meet the spec of the part they’re faking, and hope that nobody can tell it’s actually Folger’s Crystals.

  2. After seeing the video, I don’t think the Keysight “ad” was intrusive, and necessary to properly validate the comparison.

    (Besides, most Hackaday readers couldn’t afford that gear anyway.

  3. Do they not teach statistics in engineering anymore?

    Test 1000 of each one and then show me your results.

    These measurements mean nothing for the fake parts, unless I’m going to use THOSE SPECIFIC fake parts that got tested.

    The QC on official parts mean the datasheets can be trusted.
    Which is literally what you are paying for when you buy most real parts.

  4. The principle about fake parts is: is it western or Japanese + expensive + popular + easy and cheap to counterfeit? If that’s the case then you will most certainly never find the original part when shopping on Amazon/Aliexpress/Ebay, especially from far east sellers. You bought that very low noise audio or instrumentation opamp, or a power mosfet? Well, you 99.9% bought a relabeled 50 cents common part with totally different ratings, probably even a reject one. In some cases you get completely different parts, like bipolar transistors relabeled and sold as jfets, or even empty plastic “chips” with pins but no die inside. You’ll never find a genuine NE5532, LM833, AD*, NJM* or BF245, 2SK170, etc. on Aliexpress; can bet all your horses on that.

    Sometimes the fake is easy to spot; that’s the case of a Chinese seller on Ebay who years ago attempted to sell AC128 labeled transistors in TO5 case (!), probably because everyone wants them in their fuzz pedal to obtain the famous Jimi Hendrix sound, but sometimes they’re near perfect copy and much harder to recognize until they’re put in a working circuit.

    (hint1: Ge transistors sound great because they’re germanium; just stick the cheapest you can find and tune biasing accordingly; you won’t hear the difference- hint2: soviet fabs stopped production much later than the US or EU, therefore if you buy them from former soviet countries, they’ll be often military grade much newer and less prone to leaking, that is, much better parts at a much lower price).
    (trivia: I actually have some Ge transistors in TO5 case that sounded absolutely fantastic in a fuzz pedal, but they’re grounded base surplus military parts, probably RF, relabeled with an impossible name by their manufacturer)

    Bottom line: want to build a 555 flasher or an Arduino thingy, or play with cheap modules? No problems at all, just go for the cheapest, but if you shop for parts that are important, just ignore Amazon, Aliexpress and Ebay and go straight to the usual well known sellers (Mouser, Digikey, Reichelt, TME, Conrad, etc).

  5. Sigh. I wished we would live in a world in which the specs mattered, rather than price and heritage.
    There are quite a few occasions were a clone chip was “better” than the original.
    The U880 comes to mind, an East German Z80 compatible. It had a few minor bugs (carry flag not set for outi), but otherwise was running very stable.
    According to what I’ve read, it had more tolerant timings than the Zilog part because the East Germans had believed in the written specs mentioned in the public Zilog dataheets.
    Or, let’s take Soviet era parts. Many ICs were being produced in white ceramic chassis with gold-plated pins. Compared to your average western parts, these were high class parts. Also, the wiring style (threading technique). Soviet PCBs had well done wiring that involved a lot of skill.
    Of course, to be fair, many of those Soviet devices (computers, measuring devices) had not been used by hobbyists back then. They had been designed with schools, universities and government institutions in mind, rather than citizens.
    East German hobbyists had similar issues accessing equipment and parts, though, so it wasn’t an unique situation.
    In retrospect and in reverse, I’m a bit sad that we West Germans didn’t have U880s for sales in our stores back then. A lot of our hobbyists surely had accepted them as an alternative.

    1. “I’m a bit sad that we West Germans didn’t have U880s for sales in our stores back then. A lot of our hobbyists surely had accepted them as an alternative.”

      I doubt that. In 1985, the price of a western Z80A had already dropped below 5 DM (single unit, incl. VAT). There would have been no way for RFT to offer an UA880D at a competing price.

    2. “… Soviet era parts. Many ICs were being produced in white ceramic …”
      These were military spec parts. The stuff that surfaced via Fleabuy in the 90s was often “harvested” from abandoned soviet military bases.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.