Deep-Sleep Problems Lead To Forensic Investigation Of Troublesome Chip

When you buy a chip, how can you be sure you’re getting what you paid for? After all, it’s just a black fleck of plastic with some leads sticking out of it, and a few laser-etched markings on it that attest to what lies within. All of that’s straightforward to fake, of course, and it’s pretty easy to tell if you’ve got a defective chip once you try it out in a circuit.

But what about off-brand chips? Those chips might be functionally similar, but still off-spec in some critical way. That was the case for [Kevin Darrah] which led to his forensic analysis of potentially counterfeit MCU chips. [Kevin] noticed that one of his ATMega328 projects was consuming way too much power in deep sleep mode — about two orders of magnitude too much. The first video below shows his initial investigation and characterization of the problem, including removal of the questionable chip from the dev board it was on and putting it onto a breakout board that should draw less than a microamp in deep sleep. Showing that it drew 100 μA instead sealed the deal — something was up with the chip.

[Kevin] then sent the potentially bogus chip off to a lab for a full forensic analysis, because of course there are companies that do this for a living. The second video below shows the external inspection, which revealed nothing conclusive, followed by an X-ray analysis. That revealed enough weirdness to warrant destructive testing, which showed the sorry truth — the die in the suspect unit was vastly different from the Atmel chip’s die.

It’s hard to say that this chip is a counterfeit; after all, Atmel may have some sort of contract with another foundry to produce MCUs. But it’s clearly an issue to keep in mind when buying bargain-basement chips, especially ones that test functionally almost-sorta in-spec. Caveat emptor.

Counterfeit parts are depressingly common, and are a subject we’ve touched on many times before. If you’d like to know more, start with a guide.

[MCUdude], thanks for the heads up.

53 thoughts on “Deep-Sleep Problems Lead To Forensic Investigation Of Troublesome Chip

  1. It’s CCC “Cheap-Ch**a-Cr*p”. Using stolen IP, it’s a “mostly compatible” Atmel Clone, on an ancient process, using stolen tech, using stolen tools, using stolen processes, using stolen masks, blah blah blah…

    The “CCC” manufacturers are good at stealing, cloning, etc.

        1. Good for them! They are without a doubt the vendor I most enjoy dealing with. They started in the 70s, selling “real” parts to hobbyists, when the major industrial distributors wanted nothing to do with small quantities and the only other alternatives were surplus or Radio Shack.

      1. Unfortunately 99% of Manufacturers and suppliers are not interested in selling anything less the value of your house or more worth of parts at time, so direct to any smaller company or maker just isn’t going to happen..
        I’m probably exaggerating the percentage, as it does seem to be getting less ridiculously impossible to source ‘reasonable’ batches. But the issue does still exist…

    1. Well, it’s obvious in this case that we’re NOT dealing with “stolen masks.” The chips are substantially different.
      We’re more in the range of “Duplicating unadventuresome 10-y old tech with modern tools is pretty trivial.” Even if you update features. Especially if you’re willing to ignore some of the more obscure specs.
      I mean, there are Open Source AVR Cores for FPGAs that were (more or less) some college student’s thesis project, polished off in a couple of year with minimal staff. How much easier for a large semiconductor company to do essentially the same thing?

  2. Counterfeit chips, transistors, mice, gamepads, clothing, shoes, etc,etc…Try scratchbuilding an rf project – better stick to parts from digikey or mouser or some other reputable supplier but then that seriously limits selection.

  3. I got lot’s of those boards. I don’t buy CCC any more they don’t respect rights. There is no such thing as a private company in China. So it’s the government doing the stealing.

  4. >Atmel may have some sort of contract with another foundry to produce MCUs.

    Their datasheet would have accounted for that. Under Power down current, there are max. values given.
    It wouldn’t be 2 orders of magnitude higher. They would have disqualified that fab/process.

  5. There are dozens if not hundreds of threads on various forums where someone is essentially asking:
    “I bought X from a no name dubious “distributor” for a suspiciously good price. The part arrived, but – the parameters don’t match/it shows signs of reuse/it doesn’t work at all… Is it possible that I have been scammed? Or are normal proper Microchip parts supposed to explode?”

    At this point, with all of the info available, we’re into “Shame on me” territory.

    The remedy is simple – DO NOT BUY FROM DUBIOUS SOURCES. By buying stuff from them you are supporting this sort of crap. Don’t. Just don’t.

    While I’m sure that there’s the odd mishap in the supply chain of a real distributor, it’ll happen far less often than from the online equivalent of a guy in a trench-coat that’s standing behind a newsstand and offering genuine Rowlexx watches for 10 bucks.

      1. That’s a pretty poor argument and apples to oranges comparison. Those were not defective/counterfeit chips but a vendor planting a “bomb” in the Windows driver to brick a product of someone else.

        I don’t see how that would be responsibility of the chip distributor (never mind that reputable ones wouldn’t sell you a counterfeit FTDI chip in the first place).

    1. The irony is: I’d be willing to buy original-but-used chips for a lower price. But these fakers in China are all into relabeling the chips as if they are new. So you can hardly find any original-but-used chips on EBay anymore!

    2. If all you need is, to know the time, the “Rowlexx” could be sufficient :-)
      If the cloned 328P behaves otherwise equal as the Atmel part, it can be OK for many applications for 1/4 of the price.
      Of course it would be helpful, to know the differences. In that case I have to test, because I have a Chinese Arduino Nano and I planned to use it’s standby feature in a battery powered Sensor application. OTH: For a price of €2,- instead of €20,- I can add a switch or a shut off transistor.

  6. BTW I wouldn’t trust the more expensive popular chips e.g. AVR parts as it means more profitable for them to sell you a “fake” one. Laser engraving is just too cheap these days for them to substitute with a clone and put in a different part number.

    At least in this case you are getting something that is functional similar.

  7. It’s fairly simple to source current parts from reputable suppliers. The problem I’ve run into is sourcing authentic new old stock for parts that have been discontinued. Ebay is full of them and 90% are fake.

        1. You are looking at it all wrong.

          It’s not a comment system at all. It’s therapeutic. It allows those with strong opinions to express them, yet makes it unlikely that others are going to see them.

  8. A bit similar to the “Blue Pill” boards which used to have the STM32F103C8T6.
    A few years ago you had some reasonable chance that there was an original STM chip on it.

    There is an 8 page thread about this on:
    https://www.eevblog.com/forum/microcontrollers/cheap-bluepill-very-likely-it-has-fake-stm32-right/

    Which links to Richi’s lab with die pictures of some 8 different “variants”

    https://www.richis-lab.de/STM32.htm

    If you buy a “blue pill” from Ali / Ebay / China it has the STM prefix, but you never know what hardware is inside the chip, as they tend to put on “the cheapest of this week”, and as there are differences and incompatibilities between these chips, It pretty much means the “Blue Pill” is dead. Several people have resorted to making their own PCB for the Blue Pill and soldering the boards themselves with parts from known sources, which also sort of defeats the purpose of the blue pill.

    Quite a shame, as I very much liked this breadboard and matrix board friendly minimal development board.

  9. Why exactly do they bother counterfeiting this anyway? Seems like they can get away with any kind of infringement they want, so why not just use their own part number, call it a “compatible” version, and still sell just as many because 99% of the stuff that uses that is fine with the fake?

    Just sell it as the CN328 or something, everyone will know what it is.

  10. There is a reason why manufacturing is in China: to make $$$ for US companies. And as long as that is true (US import tariffs non-existent, mostly) this will remain true.

    So, don’t blame China. Look at your politicians in last 40 years. And you will also realize that this offshoring has destroyed American manufacturing and caused a wipeout in middle class in the US.

    Bon appetit.

    1. I just now finished watching this movie called “Heroes For Sale” from 1933. It shows how the automation of labor forced men to be without income. The US managed to recover the depression, maybe things change for the better.

  11. I wonder If Shenzhen Aries Technology would have sold their chip as Ariesduino with their internal part maybe called AR87L21D for .35 cents per MCU ,the community would have been glad to add them to the Arduino community . why all the tricks, just be candid and say the truth , and stick that parts number on front with big zodiac sign . looks like nice work on the silicon and no copies. I like the logo as well , I estimate 10-20 or so college graduates took to make this silicon a reality. US college graduates should do silicon too. [ However in my work I find less and less workers cares about my new circuits and ideas ,and more and more talks about 401K plans ]. And work harder on that sleep down power state ( more headaches with mosfets and metal-interlinks.) . But personally, no matter what I would still buy the Atmel part and pay the extra buck , because trust always cost $$$ , and honesty is valuable commodity.

    1. Pff. “Reputable”.. The people behind companies constantly change. In the end, sometimes the only thing a company has in common with its 5, 10, 20, 50 years older self is the name. 🙄

      1. A company does not need to be more than 5 years old to be determined as reputable or not. If you believe otherwise, than you must be referring to sellers on eBay, the fly by night web stores, or joe schmoe from that one hit wonder kickstarter.
        On the other hand, Adafruit was well established as reputable within the first few years even though they were a very small and non-typical company for electronics. While still considered small, the general population built significant trust into them because they earned it.
        Chinese knockoffs are very common today and one reason why is because people don’t want to pay for quality anymore, they want everything as cheap as possible. It is very common for people to buy 5 of something from Aliexpress because, even though they only need 2, they expect 1 or 2 to not be functional. Spend a couple bucks more buying from a reputable place, and things magically work….correctly…the first time….and for years….without bandaiding code….or making mods……etc,etc, etc

  12. So many commenters, but all the same opinion.. 🙄
    Let’s see things from the bright side also.
    These “compatible” chips ensure a little bit of
    independency from the monopoly of a single manufacturer. Also, I think it’s generally wise to test the product you make thoroughly before you begin selling it, instead of blindly trusting a popular brand/manufacturer. To give a recent example, several big HDD manufacturers didn’t mention the use of SMR.

    PS: Thanks fir the article, it was interesting to read. 😎

  13. Thanks for the great detective work and analysis…Iʻm prototyping a data logger for a farmerʻs co-op here and was beating my head on my bench over this yesterday… “HTF are folks getting 1-2uA?”. All of my “ The Simple” 3.3V, 8 Mhz Pro Minis from Amazon all “sleep forever” at 112uA on a board stripped down to only a couple of discretes.

    1. After a week of fuddling and Iʻm now getting roughly 1.5uA sleep mode Pro Mini power consumption with a DS3231SN , micro SD and a few sensors. Same performance from Pro Minis from 6 different sources including US vendors like Sparkfun and Adafruit. New and old “ junkbox “. If theyʻre counterfeit, theyʻre pretty consistent ( or all from the same foundry) . Part “numbers” are the 328PU-TH and 328PU-KR. Do I even have any boards with “legit” 328s? Maybe…donʻt know anymore..All of these have the correct signature. Do they work? So far, so good. I just need 100 but with lead times from Microchip into late next year, theseʻll have to do until then.

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.