Flaky ATmega328Ps from China
I recently ordered some ATmega328P-AU from a Chinese seller I deemed to be trustworthy, because I had only good experiences with them so far. They arrived in a piece of cut tape reel and looked legit at first glance. After having some troubles trying to program them I took a closer look and noticed they look a little odd.
The photo below shows an ATmega328P from a local, reputable distributor on the right and on the left is a chip from my recent order. I've drawn some symbols to illustrate that the marking isn't in the same place on the Chinese chip. Have you ever noticed something like that? The marking is also duller and harder to read in bad light conditions. I don't know how many factories Microchip has in Thailand or if they change their setup over the years (one is from week 26 of 2018, the other from week 52 of 2019), but it doesn't feel right to me that they appear so different. The relative bright "spots" on the right chip (I had a light source right above them) may also suggest, that its surface is much rougher than the one's on the left.
Anyway - I've now tested four out of the ten I got and they all behave the same:
- I can read them successfully every time with avrdude. Device signature is correct, fuses are in factory setting.
- Uploading simple programs which don't use up a lot of memory, like a blink sketch, works without issues.
- Larger programs, which take up more memory, fail to upload with growing propability, the larger the programs get with a verification error on byte 0x0000. A ~20kB program with MySensors plus some additional libraries fails about 80 - 90% of the time.
- When the program eventually uploads successfully, it seems to be fully functional. SPI, I2C and general I/O pins are working.
- In a circuit with only the MCU (3.3V, 1MHz with CKDIV8) and neccessary passives, I measured an active supply current of about 0.8 mA and about 4 µA when in power down mode. The former seems to be a little on the high side but still acceptable, the latter is perfectly within specs (WTD enabled).
I've swapped out all wires and the USB cable, used two different USBasps with and without slow SCK jumper and a genuine Arduino Uno with ArduinoISP, different sketches, programming at 5V and 3.3V with baud rates between 9600 and 115200. However I don't have this issue with MCUs from other sources. Unfortunatelly I'm not (yet) willing to invest in a decent Oscilloscope for the little bit of electronics tinkering that I do, so I don't think I can do much more to investigate the programming issue further.
I don't think I got sent counterfeits - essentially, they are working. It's just odd that they look different and, for some reason, don't behave well when they are being programmed. Maybe rejects? Resurfaced and relabeled? Any thoughts about that?
You can't trust China these days... As many DIY people a few years ago purchased a lot of electronic stuff from China, the counterfeit industry scaled the growth too
If you can take the "original" and program and it's working flawless, and those cheap Chinese ain't working, then assume they are counterfeit...
I totally agree with you that a lot of electronics from China is of dubious quality and lots of ICs are forged. I just didn't hear of a "well made" ATmega328P clone yet. I read that they either work fine or not at all. So, if the chips I got are counterfeits, they are made pretty darn good! Once they are programmed I can't tell a difference between a device with a Chinese or a "genuine" chip. I currently have a node with one of those chips from China running for about a week and it works as well as one with an ATmega from my local distributor.
I really don't want to be "that guy" who spreads the rumor of almost perfect Chinese AVR clones if they aren't clones (because I can't prove it), which is also why I decided against posting in the other thread you linked. At least I could rule out that they are relabled LGR8F328Ps (different pin out etc.) - and this is the only reverse engineered "ATmega clone" I know of. I think it's more likely that they are rejects from the factory or something like that. This could explain why they are so flaky, but otherwise work.
In any case, if I didn't have the programming issues, I hadn't noticed that there is a difference in the position and appearance of the marking on two different batches and I was primarily curious if you guys noticed similar variations.
@BearWithBeard I already got flaky 328p, a long time ago when I started with MySensors. And I decided to never buy my mcus from "unreliable" source. Especially now as I'm using advanced mcus.
I never digged what was the exact problem because I thought it was just waste of my time compared to original price. And it already wasted my time because I was searching if fault was on my side..
So I can't tell you why, but counterfeit, or out of specs, is not worth it for prod.
Sure is, reliable source always worked for me.
huex last edited by huex
as chineses, i tell you ,the Official 328P in china market have made in china , Malaysia different manufactured ,price not same, manufactured in China only 6RMB[1dollar],Malaysia 8RMB [1.2dollar], so i think if you can run mysensors ok but only Different appearance，this is Official product.....as i know fake 328p can't run mysenors such as LGT ,but LGT different with 328P, it's only can compatible 328p some code ,but not same For example 16MHZ Built-in[also 32MHZ] ,different memory register,more hispeed adc 。and it can‘t run mysensors code , but very cheap and have better speed 、adc 、USB , it only 3RM[0.5dollar]in china ....., if only alone use, more use LGT...., if mysenors can run LGT, this will be good news ,!
regardless if those are counterfeit or they are production rejects, selling those as "good" and "original" is just not appropriate - this leads to that noone will buy from the "grey area"Chinese suppliers...
And not to prejudiced against China or Chinese people, there are good and untrustworthy people in all countries
BearWithBeard last edited by
Yes, this is indeed a problem. China's bad reputation for scams and counterfeit products is of their own making. It's hard to build trust if you always risk running into dishonest and fraudulent sellers. My case is just another example amongst countless others.
I just thought about all those thousands, if not millions of Chinese Arduinos we all buy all the time. And most of them work perfectly fine, so there has to be a legitimate source for genuine ATmegas in China. What could go wrong if I buy the bare chips from the same stores that sell those Arduinos? I expected the MCUs to either work completely or not at all - I either get them for a bargain or get my money back.
And btw, I indeed got my money back. I had a chat with the customer service and they offered me a full refund without hesitation, I didn't even have to ask for it. I spent 8 EUR (~8.5 USD) for 10 pieces (includes coupons). A reputable distributor would have charged about twice as much for this quantity, not even counting in shipping costs, potential small order fees or customs. Worth the extra expense? Propably, but I'm not regretting that I tried my luck either.
huex last edited by huex
emm...i don't say about Which country and which people，only told market info, i buy most form taobao and ask saler easy to get atmeal fake or different manufactured,forum most post about this fake , from alipexress? if you want get from china mouser and Digi-Key will be good choose for you,or use express transfer service with good B2c saler like szlcsc .... The third party is hard to trust.......same for me to buy amazon us or de .....
@BearWithBeard Maybe static electricity or something corrupted the chips during transport? I suggest you use a high voltage programmer to thoroughly erase the chips and then reprogram the fuses and everything else using the high voltage programmer. If you don't like the chips as they are, then you've got nothing to lose by trying.
BearWithBeard last edited by
@huex Yes, I got them on AliExpress. Sites like taobao and such, who mainly target locals, aren't an option, since I can't read nor write Mandarin. It's also not possible to order from the Chinese branch office of Mouser / DigiKey if you're not living in China. LCSC is great though.
@NeverDie ESD damage is a possibility, I guess. I will look into high voltage programming and see if it helps, but this has to wait for a while. Thanks for the suggestion!
Those days while I worked for Nokia (which is at least 10 years ago), and I were at our factory plant in Beijing, some of our local chinese Nokia guys, bought Arduino stuff from Taobao and at this time, Taobao was much cheaper and quality was perfect
@BearWithBeard Nordic Semiconductor had (still has?) a free service whereby you can send them a chip and they would x-ray it and then tell you whether or not it was genuine. You don't get the chip back afterwards. I don't know whether Atmel has a similar free service, but if you really, really want to know for sure, you might ask them.
I think you can drive yourself a bit crazy looking for subtle differences in packaging that might signal a fake or not. Unless there's a flat out blatant marking error (like mispelling the company name or something like that), it probably isn't a good indicator. At least, in the case of Nordic chips, it didn't seem to help. What proved decisive with the Nordic chips was looking at the power utilization during Tx and Rx. That worked incredibly well at not just spotting fakes, but even identifying exactly which fake it was. Ironically, in the end I ended up preferring a certain strain of fake over the official Nordic chip.
@NeverDie Nah, I'm fine. It's always helpful to get opinions and views from others, especially if they are more experienced in the matter, if only to set one's own viewpoint straight. I'm grateful for all you guys replies, but I don't feel like bothering Microchip with this.
Regarding Nordic - besides the circulating copies / clones (like the SI24R1), I guess it's also not helping that they are fabless, unlike Atmel / Microchip, in which case it seems totally reasonable that the various fabs may have slight marking variations. I just didn't expect that from those ATmegas which seem to have been produced in one and the same factory.
But it's certainly a neat touch from Nordic to offer a free test.