2
\$\begingroup\$

To practice my SMD soldering skills and to try out stuff (self taught hobbyist level) I purchased various ATMega uC chips from Aliexpress. I know you never know what you get that way (usually what you pay for!), but I was wondering if there was a method to check if they are fake or not, now that I have them. That's why I've been watching the video from Noel's Retro Lab, after what I purchased 2 of each chips from Farnell to compare the resistance between Vcc and Gnd pins with the Chinese ones. Here are my findings, in Mohm:

            Farnell #1  Farnell #2  Ali #1  Ali #2
ATMega328PB 9.37        12.2        10.5    12
ATMega328PU 11.3        13.2        21      1.79
ATMega88PA  11          15          2.8     2.8

The ATMega328PB measures are consistant between Farnells and Alis chips, but the other two models seem "in the cabbage" (as we say in French) with random values (still, it's in the range of Mohms). Are those necessarily fake? Another interesting finding is that all the models from Farnell share a similar resistance, between 9 and 15Mohm. I'm curious to know if it consistant among all ATMega chips, or is it a coincidence? Anyway, does my measuring make any sense? What would be the next step to test those chips? Thanks!

\$\endgroup\$
11
  • 7
    \$\begingroup\$ Why waste your time - just buy from a bona fide source. \$\endgroup\$
    – Andy aka
    Nov 13, 2020 at 12:08
  • 6
    \$\begingroup\$ Measuring "resistance" of a semiconductor device will tell you very little. One or two leaky protection diodes will probably determine the reading. \$\endgroup\$
    – Transistor
    Nov 13, 2020 at 12:30
  • 4
    \$\begingroup\$ Different silicon revision might be made with a different process or technology at a different factory, so the resistance measurement is meaningless. Are you sure about what the markings on chip indicate about silicon revision? \$\endgroup\$
    – Justme
    Nov 13, 2020 at 12:36
  • 7
    \$\begingroup\$ As an IC designer I can confirm that such a resistance measurement is meaningless that resistance depends on so many things making the "test" meaningless. If you want a real test, make a program that stresses the processing capability of the microcontroller and check at what clockspeeds that can run without errors over different supply voltages. Make sure each chip is tested at the same temperature also. Complicated? Yes. I buy all my Arduino (containing ATMega328 mcu) from Ebay and don't care if these are "fake chips" or not as long as they work for me which they do. \$\endgroup\$ Nov 13, 2020 at 13:29
  • 1
    \$\begingroup\$ @Justme the markings of the chips are : Aliexpress : MEGA328PB -U 2021D TW 2021BH ; MEGA328P U-KR 35473D 1950AN6 ; MEGA88PA AU 1939 ; Farnell : MEGA328PB -U 20120 TW 2012R82 ; MEGA328PB U-TH 354730 2012SMH ; MEGA88PA U-TH 35465F 1939431 ; I did not find good references online, could you point me some? \$\endgroup\$
    – Nico
    Nov 13, 2020 at 15:16

2 Answers 2

1
\$\begingroup\$

I'm curious to know if it consistant among all ATMega chips, or is it a coincidence? Anyway, does my measuring make any sense?

In this case it's probably just coincidence. All your resistance measurements are so high that they only represent leakage, which could vary for a variety of reasons even in genuine chips (which we can see from your numbers).

Resistance measurements can be useful as a first step in determining whether suspect chips are likely to be at least similar to what is marked on them - but not the way you are doing it. A better way is to put the meter on 'diode test' and try to measure the presence of protection diodes. This can help to identify Vcc and Ground pins, and whether other pins are at least connected to something inside the IC.

CMOS ICs usually have protection diodes on inputs and parasitic diodes on outputs, whose forward voltage drop can be measured between I/O pins and Vcc/Ground with 'reverse' polarity. You can do the same to determine which pins are Vcc and Ground, where the 'diode test' voltage is usually significantly lower. You can also compare the results to a known good chip (if you have one) and be suspicious of any large variations. Do not make any assumptions regarding pin identification until this test is completed!

What would be the next step to test those chips? Thanks!

Once you have verified that Vcc and Ground are on the correct pins, power up the chip and check that other pins are inputs (high impedance) or outputs (stuck high or low, or pulsing) according to what the datasheet says they should be at startup. Then create a circuit that tests the chip's functions. For most MCUs connecting them to a programmer and reading the device ID should be enough.

I purchased various ATMega uC chips from Aliexpress. I know you never know what you get that way (usually what you pay for!)

If you want to know what you are getting then buy from reputable suppliers. If you are manufacturing a commercial product any price difference will be outweighed by the greater confidence, and for small quantities it usually doesn't matter (and the time wasted trying to get a fake chip working is not worth the cost 'saving').

When purchasing new ICs I only deal with companies such as Farnell, Mouser and Digikey, who have impeccable reputations and guarantee their supply chains. However sometimes I need a 'vintage' chip that is no longer in production, and then I am forced to buy from eBay. I usually deal with a small number of vendors who supply genuine surplus and recycled chips. Others have suspiciously low prices and parts that appear to have been scraped and remarked. With these you never know what you will get.

Occasionally I might get a remarked chip which is completely different because it had a 'similar' part number by coincidence, and the refurbishers were too lazy or ignorant to separate it from the rest. However most of the time they are least roughly equivalent to what is marked on them, but might be NMOS instead of CMOS (which is easily detected by measuring supply current) or a lower spec part. This is unfortunate because I care more about having a chip with known specs than a 'premium' fake, and even prefer chips that look shabby because then I know they are genuine pulls. Vendors of 'fake' chips should realize that they while they might con a few suckers by remarking them, they lose sales overall - so 'prettying up' chips to make them more attractive is a waste of time and effort.

Another way to get cheap parts with higher confidence is to buy complete modules such as Arduino clones. Millions of these are sold on eBay and through retail electronics shops, so if they didn't work the manufacturers would quickly get into trouble. Whether they are 'genuine' Atmel/Microchip MCUs or not is irrelevant if they work properly. I have purchased hundreds of these and other modules, and not had a single dud.

\$\endgroup\$
1
  • \$\begingroup\$ And note that it is possible for a chip to differ in physical construction between production batches, even from same manufacturer. Like CPU stepping (revisions) i.e. infamous Pentium FDIV bug; manufacturer implements a minor improvement, puts it into production. Only difference we see is a different date code and/or other (usually undefined) digit(s) on package. And if the chip is produced at more than one facility - who knows how different they would measure. Functional testing beats static testing. \$\endgroup\$
    – rdtsc
    Nov 23, 2020 at 13:47
0
\$\begingroup\$

Found answers in the series of video from Kevin Darrah: https://www.youtube.com/watch?v=PlGycKwnsSw and https://www.youtube.com/watch?v=eeDC1m7ANJI In the first video, he describes how he used low power sketch to measure power consumption: on a supposedly genuine one, it should be less than 1 microamp. In the second video, he uses a signature reading sketch to compare genuine chip signatures with suspect ones.

\$\endgroup\$
5
  • \$\begingroup\$ While a YouTube link is great, it could disappear at any time. Best to quote from it here and give attribution. \$\endgroup\$
    – rdtsc
    Nov 19, 2020 at 17:50
  • \$\begingroup\$ This doesn't seem like a reliable method in general. There is no guarantee that a "counterfeit" chip would behave any differently from a "genuine" chip. It might work once, for one particular batch of parts. \$\endgroup\$ Nov 19, 2020 at 19:09
  • \$\begingroup\$ Thx @rdtsc, I edited with more details. \$\endgroup\$
    – Nico
    Nov 20, 2020 at 16:45
  • \$\begingroup\$ @elliot-alderson what would you suggest? \$\endgroup\$
    – Nico
    Nov 20, 2020 at 16:45
  • 1
    \$\begingroup\$ You need to start with your assumptions about who is doing the counterfeiting and what their technical skills might be. If someone is getting chips from the original manufacturer that fail some sort of production test then it will be virtually impossible detect this, unless you are capable of running the manufacturer's full test suite. If the chips have been cloned by another manufacturer then looking at SEMs of the dice might tell you. At the end of the day, you must use a trusted supplier. \$\endgroup\$ Nov 20, 2020 at 19:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.