Professional Forensic Analysis of a sketchy Pro-Mini (ATMEGA328P) - is it Counterfeit?

Part 1 on the pro-mini analysis: • Sleep Mode breaks on c...
Part 3 Test Sketch to see if yours is real: • Is Your Cheap Pro-Mini...
So based on that, there were a lot of suggestions that maybe the ATMEGA328 was still genuine, but somehow failed QA and ended up on some kind of great grey market... well I decided to reach out to Sensible Micro Corporation to conduct a full forensic analysis on the part! Check them out here: www.sensiblemi...
My Tindie store - I have a board there with genuine ATMEGA328 :) www.tindie.com...
Thanks to all the Patrons for dropping a few bucks in the tip jar to help make these videos happen!
/ kdarrah
Twitter: / kdcircuits
For inquiries or design services:
www.kdcircuits...

[@chiparooo]

Very interesting! Thank you Kevin and Chris for sharing.

[@Maxtr0sity]

This is really educational, even for someone who doesn't understand the full inner workings of the Atmega. Learned a lot!

[@SimjetAU]

That was a really interesting video Kevin..thank you...our production has been stifled twice now with us having to change 2500 chips by hand...because of this issue

[@Kevindarrah]

whoa! can you share what chip it was?

[@SimjetAU]

@@Kevindarrah The first bad bunch was a LM741 that was around 4 years ago. The second one was just late last month we just got a new production batch of pcbs in and in making them up we found big issues...The onboard temperature sensor was reading negative values and all were the same so out of 3000 board there were 2500 with that batch number. There are around 4 batch numbers we have found as we use the temp sensor in another pcb as well but it was the one batch number that were duds...QT18B20...we ahve used probably 30,000 so far at least and no issues until last month. So the assemblers sent us another 2500 so we have to change them all. Our assembler in China is a high end company who makes LCD as well as does our assembly of all of our production pcb..they are beside themselves especially as this has happened twice now. Its a big drama when it happens in a production enviroment

[@SimjetAU]

As a side note the duds for the 741 had a poor copy of the TI logo and the QT18B20 we found they had the same symbol on them but no pin 1 indent on the dus but were all ok on the ones that worked other than that the printing style and depth were the same and looked the same except for the batch numbers

[@Kevindarrah]

wow, you would not believe how funny this is! As we speak right now, I'm working on a project using the LMT01, which is what I switched to after a bad batch of 18B20's... I feel like that part is just a pile of crap now and the whole market is saturated with fakes...

[@SimjetAU]

@@Kevindarrah We learnt after the 741 incident that we now get about 10 random pcbs from their production lots of each pcb they do for us sent over and we test them before we give the go ahead to ship them to us. This batch obviously they didnt spread the samples across the 3000 lots of each pcb they produced as we did not find any issues with the samples

[@stevetobias4890]

That was a super cool video. It is really good to know there is a company out there that can inspect chips to such a degree. I am really blown away. Thank you so much for posting this video. Stay safe mate, cheers from Down Under.

[@Kevindarrah]

Thank you!

[@johnmccormick2883]

Crazy stuff for sure... would’ve liked your closing comments but I guess the conclusion is clear. Great video ... thanks for sharing.

[@TheEmbeddedHobbyist]

I think this points to how easy it is to set up a wafer lab plant. We had a custom wafer made and after about 10 years the plant was shut down and we got a onetime buy to keep us going. But we then found a small start up that could almost get to what we needed. These were sensors on a die that was directly exposed to liquids, and had to have some extremely consistent properties and a very tight process in wafer build to work. They only had to work for a few days, but within a extremely tight specification with minimal calibration and then thrown away. But they did it. I think that uC internals are now just like cut and paste, a memory here, an ALU there and with almost off the shelf designs. The computer will then do the die design with little input from us. It’s been going on so long with high value chips! that it’s now getting to the point where they can knock out parts so quick N cheap that nothing is now off limits.

[@McTroyd]

Very interesting. They spent some time on the chip packaging, but even at the X-ray level the suspect fails comparison. The infuriating thing is, the copycat could have saved themselves a world of trust issues, branding it differently and advertising as "compatible with ATMega328P", especially since it mostly works as advertised. Goes to show why it's worth spending the extra couple of bucks and sourcing from a reputable company (my Arduino came from Sparkfun). Better yet, it would be awesome if everyone would go to arduino.cc periodically and buy a board to support those guys.

[@NivagSwerdna]

Wow! Just wow! I have been suffering from counterfeits recently... first serial to TTL converters and more recently fake CPLDs and memory... for years the cheap modules/parts from China were fine (or fine enough) but over the last 12 months or so it has become very unreliable. Very interesting video, thanks both!

[@TradieTrev]

There's the quote from Oscar Wilde "Imitation is the sincerest form of flattery that mediocrity can pay to greatness" that rings very true when you've got some coming in and mixing their stew so to speak.

[@AdilKhan-jp7hn]

Kevin, This is truly an eye opener for me. Excellent video and information. Thank you so much to bring this darker side of silicone out in the open, it really opened my eyes.

[@timjohansson4304]

Very intresting and good content nice of you to put it up!

[@howarthcd]

The cloning is not something to commend, but you have to admire the effort that has gone in to cloning such a device. There are some clever people out there who are wasted on this kind of activity.

[@user-qf6yt3id3w]

Back in the days of the USSR the Russians were so infamous for copying Dec chips that the Dec engineers put a message in Russian on the CVAX die en.wikipedia.org/wiki/VAX#History " Sundry graphics were etched onto the CVAX microprocessor die. The phrase CVAX... when you care enough to steal the very best was etched in broken Russian as a play on a Hallmark Cards slogan, intended as a message to Soviet engineers who were known to be both purloining DEC computers for military applications and reverse engineering their chip design.[18][19] " There's a photo here: micro.magnet.fsu.edu/creatures/pages/russians.html

[@MrGeekGamer]

Xiao Bao... sounds legit.

[@Kevindarrah]

lol, yea think. that translates to small package or something

[@peterwooldridge7285]

Interesting....thanks

[@Ed19601]

I know little of chip making but i wonder: when you are forging a chip, wouldnt it be just as easy to make a correct chip as it would be to make a faulty chip? Or is this a case of: "we made proper 328's but there is demand for 328p's so we just put that on the label

[@oberoiddroidus2114]

Where is pin 1 on the board? Did anyone mapped the functional regions of this chip? I would like to know more about its architecture.

[@Korpaksson]

Well that explains why all my ProMinis bought in China crashes when voltage drop below 3V. While in theory they shall run fine with lower voltages with 8MHz and slower crystals.

[@Kevindarrah]

hmm, maybe I should try that out here as well.

[@englishrupe01]

@@Kevindarrah Please do, Kevin and let us know! But i gotta say, wow! This is a really cool exploration, thanks! I have just ordered my first five boards from China, and now worried they wont do the deep sleep i got them for! Thanks for this!

[@avejst]

Impressive find A good detection Thanks for sharing👍😀

[@sanityd1]

Given the price of aliexpress arduinos, I always assumed this was the case.

[@fredericbenzaquine3510]

you surely have some resources Kevin, thx a lot for the insight, I never imagined this could be possible ! And even if this is no big risk regarding our little insignificant DIY projects, I'm mostly concerned with giving my money to REAL engineers who worked hard designing those chips, I'll give it a more careful look next time I order.

[@Kevindarrah]

thanks - yea the results were surprising to me as well!

[@ClaytonDarwin]

Good info. Thanks, Kevin.

[@38911bytefree]

10 years ago I couldnt meet the nanowatt figure on a pic16F684, but they were off like 100 times. I made all the changes possible, talked to microchipp assitance, read all the guidelines ... NOTHING. They even fake mixed signal chips and perform pretty good actually, they are only weak when input comes really too low. At work we have been fooled a couple of times with thos chips, in big quantities. Also, we worked with MSP430, board assembled in China. They performed good, no probs. One time a TI representative visited us, we showed the actual board and got lost inmediately, because the coding on the chip. He had an internal web to put this code to get the plant, customer et ... and eh couldnt get any valid data out of this code (bizarre). The contract manufacturer was a big one in China, that work in "big" quantities and they are not supposed to put fakes or trade in parallel markets for pieces from another customers. TI was clear about that, this could be a batch that other customer did not use but the contract manufacturer MOST TELL YOU about the situation. TI wont accept any liability in this cases. Only part that were bought to a TI dealer and put in the board.

[@MCPicoli]

How much this kind of report usually cost?

[@christorrioni6238]

Hi Marcelo, our standard reporting is free if our customer is also purchasing the components from us. Our standard reporting includes external visual inspection, dimensional verification to the OCM datasheet, part marking inspection and surface integrity analysis to confirm the top of the chip has not been remarked or tampered, X-Ray analysis, low power electrical testing and pin / BGA inspection to confirm not previously used or damaged. Decapsulation is the only item we charge for which is $50 and most customers normally just have us roll that cost into the price of the components they are purchasing. Here is a link to our lab page: www.sensiblemicro.com/electronic-components-inspection-test-lab

[@chosen1one930]

Are they using sulfuric acid to chemically show the die?

[@hansibull]

Thank you for sharing this! Very interesting! It would be even more awesome if we could get to know more about this counterfeit 328P die. Why is it smaller than the original die? Is there any other difference between them except the power consumption? And who is Aries? Have any of you guys been able to find their website?

[@Kevindarrah]

yea that would be cool - I can't find any info about them.

[@Kevindarrah]

@@billiejean3748 nah I don't think that's them - there's a company in Shenzhen that's more likely: szaries.en.alibaba.com/company_profile.html

[@billiejean3748]

@@Kevindarrah yea the logo dosen't match like I previously thought, my mistake. It would be interesting to know how the fake was created and why it couldn't sleep as well.

[@campbellmorrison8540]

Im amazed its worth the cost of silicon facilities etc to copy these devices. I mean honestly how much does an original cost in huge quantities. Im obviously missing something regarding production cost optimization. Mind you it amazes me the effort put in to copying quality watches, just add a little more effort and they would be amazingly good watches but no, its all about saving 1c

[@mrlazda]

So much trouble for something that can be tested simple by measuring number of cycles it take to complete mul instruction. Chinese version of AVR core need (called MAVR) just need 1 cycle and original AVR take 2 cycles

[@PainterVierax]

interesting ! I just found an Olimex's blog post about this.

[@mrlazda]

@@PainterVierax Chinese version of AVR are less known then their version of STM microcontrollers ( I call them versions because they are not 100% identical and according to law they are "legal"). For me was biggest surprise when Keil putted Chinese version (GigaDevice) of STM as supported device separate then STM.

[@prathambumb5593]

Excellent 👍🏻

[@ToBeDefined85]

The video title should be changed or the video description should also name Arduino Nano and Arduino Uno since the same µC is also used on those very popular arduino boards and it wouldnt be presumptuous to claim they used the same counterfeit chips on those boards. Great video !

[@luizcassettari]

Wooo that's really crazy, I want some to make some random tests!

[@dave-in-nj9393]

it just goes to show that if you have the chip making machines, you can make whatever chip you want. Sensible Micro can take a chip apart, so can any chip maker. We all know that EVERY manufacture takes apart other manufactures chips to learn their secrets and often improve on them.

[@datapoint6859]

The images of the good chip show pins 3 and 6 are missing, pin 4 is missing on the dodgy ones. The x-rays at 6:00 do not show these missing pins, but then we hit the decap section and they're back. What gives?

[@Kevindarrah]

I'll check with Chris and see what happened there - may just be a side effect of the decap process

[@christorrioni6238]

Yes, during the destruct process wire bonds can get removed. Xray images are always taken first.

[@datapoint6859]

@@christorrioni6238 Neat, thanks for clearing that up gents (no idea why I didn't see a notification on this)

[@nerfnerd8404]

Excellent findings

[@sdriza]

Awesome video!

[@Asyss_Complex]

That was very interesting to watch! Although I don't buy my electronic components from Ebay or similar 'cheap product' sites, it makes me wonder what is the ratio of fake chips vs real ones in the market. And a better question yet, how easy is it to make a clone chip? Thanks for the video! - Cristian

[@Kevindarrah]

yea, don't even want to think about how many lost hours, days people have spent trying to get this stuff to work

[@superdau]

So we can pretty much expect that all those ebay/aliexpress Arduino unos/minis/nanos will use that chip and not a genuine one.

[@Kevindarrah]

maybe and makes you wonder what else...

[@Asyss_Complex]

I'm wondering that too. It must be a good business for the sellers.

[@erynn9770]

Not only that. The very first Raspberry Pies were manufactured in China, too (producing in the UK was far too costly due to insane tolls). When XRaying them it was discovered that the manufacturer "accidentally" had used cheaper parts than specified (and paid for). I wonder how many small-production-run devices of "Western" companies have similar accidents that are not noticed

[@Mr30friends]

@@erynn9770 its so funny that its just impossible to take china out of the supply chain in anything that has electronic components. Its even funnier that most of those stuff are pretty much manufactured automatically and the western companies placed their manufacturing lines there decades ago and gave away their know-how to save pennies. Too late know I guess. Can't make anything out of china that won't cost 4 times more than what it would cost to make it in china.

[@PainterVierax]

@@Mr30friends Well, 4 times more is the real price we should actually pay. It will make the whole electronic market more sane for both manufacturers and consumers.

[@SOSSTSE]

Thank You

[@andymouse]

Aries the RAM (not memory ! )... ..the picture is a goat's head, would it not be possible to ask Atmel if this design is used in their lithography and if the answer is no then the chip is manufactured elsewhere and if there is no license then its a fake ?...great vid...cheers.

[@andymouse]

@@billiejean3748 the logos different, maybe this lot are getting ripped off too.

[@billiejean3748]

Yea I realised that, my mistake

[@andymouse]

@@billiejean3748 well thought out though.cheers.

[@M0UAW_IO83]

Aries and Atmel rings a bell for some reason, can't puitmy finger on why though

[@Kevindarrah]

would be really interesting if you have anything to share - I can't find any trace of this company anywhere

[@Fpvdrone1325]

This explains my failing pcbs. I got a batch of pcbs made at jlcpcb and using lcsc atmega328 chips and my failure rate is through the roof. Anyone else have this experience with lcsc chips?

[@Kevindarrah]

ouch, I didn't think LCSC would not have genuine parts... I'm thinking about creating a self-testing sketch to identify these chips

[@VLS-Why]

Very cool!

[@AshenTiger]

The saga continues!

[@Kevindarrah]

yea, and might even make a part 3 with a special self-test sketch to ID the clone

[@cokegen]

Interesting ... it blows my mind that people (china) are actually re implementing chips like these.

[@Flapjackbatter]

Most interesting.

[@ArnaudMEURET]

Im not sure why you call these counterfeits, they’re not using the brand, not posing as Atmel. They made a mostly compatible chip. Not pretending it’s an Atmel product. ??

[@Rocket_Try]

That's the kind of fun video no one understands I find fun.

[@Kevindarrah]

me too :)

[@irukard]

Nice :)

[@fleetinglife2984]

i am sorry for all this，but can do nothing，my country is sick。。。。

[@ThePopolong]

I don't understand why for people still use ATMega, they are expansive, not powerful and not efficient at all. you can found far more competitive ARM or, better, RISC-V microcontrollers for less than 1$ now. They generally works well with Arduino microPython, or other standard development tools.

[@jolangse]

There's a lot more to part- and architecture selection of a controller than price-per-unit and "getting the most horsepower". The PIC16, STM8 and even the 8051 are still widely used, not because they are cheap or powerful (none of those are) but because they are appropriate for the total design and cost stack-up.

[@FixDaily]

Well, it's the same as cracking software, it gives no profit to the companies but improves access to most people, so it supports human and technology evolution. Chinese people doesn't earn like Atmel's people, that's why they can afford lower costs. And most likely their chips are made in china anyways. I'm not against china, i'm against certain actions china does to their people and to the world, but they are making the technology evolution affordable (even if they clone stuff), that's for sure. So why to pay more if you can have a copy for less that works almost the same way, right? Silicone, the etching process itself, epoxy and bound wires doesn't cost that much, the overall cost relays on "administrative and scientific matters" - so let's keep the money machine active for those Atmel guys. Also china gives no warranty, so less loss.

[@SianaGearz]

Well i'd for sure like to always know what i'm getting. If the product was tested and engineered with one chip, and you silently get another, and it causes issues in operation, well that's no good. This is why counterfeit chips and deceptive top markings are a problem. As to having a more or less compatible third party chip, well that's certainly just fair.

[@Kevindarrah]

yea, would at least me nice if they marked it something else, so you know it's a clone... like for some basic things this might be ok.

[@ElectraFlarefire]

I'll agree with Kevin. If you are going to make a compatible chip, mark it as such, sell it cheap and let people know what they are getting. There'll still be a huge market for them. A lot of the Chinese blue/black pill boards come with GigaDevice gd32f series chips(My understanding is it's a licanced ARM core too, but I could be wrong), pretty much compatible with the STM version, but you at least know what you are getting. Copying a chip's compatibility is a grey area with arguments on both sides.. Selling that chip as the original is just plain dishonest.

[@bardenegri21]

Super cool!