Fake chips, I got stung
linuxjedi.co.uk
linuxjedi.co.uk
Fortunately I was able to make do with the genuine part, though not trivially: there was a weird clock termination difference. It's never easy, is it?
(not even kidding!)
Don't joke, they exist.
Though I won’t say I was too sad when we got to use a 5GHz touch screen Tektronix instead! It ran Windows XP and was generally super awesome to use.
Also, of course: "instead of bobcat, package contained chair".
Penicillin is probably the best and most well known example.
Reminds me of reading about "Fogbank"...
https://en.wikipedia.org/wiki/Fogbank:
"In March 2007, engineers devised a manufacturing process for Fogbank. Unfortunately, the material turned out to have problems when tested, and in September 2007 the Fogbank project was upgraded to "Code Blue" status by the NNSA, making it a major priority [...]
The experience of reverse engineering Fogbank produced some improvements in scientific knowledge of the process. The new production scientists noticed that certain problems in production resembled those noted by the original team. These problems were traced to a particular impurity in the final product that was required to meet quality standards. A root cause investigation showed that input materials were subject to cleaning processes that had not existed during the original production run. This cleaning removed a substance that generated the required impurity. With the implicit role of this substance finally understood, the production scientists can control output quality better than during the original run."
I have also spent thousands of dollars of my time so I wouldn't have to replace a $3 part.
Some years ago at an IETF one of the talks was about an outfit whose hosting or cloud or something pricing change was designed to prod people towards IPv6. Something like, on contract renewal you could keep the same config, same price, for the next period, but you only get IPv6, or, you could decide you must have IPv4 still, but that's a separate product now, so you need to get your finance people to sign off on the revised billing with the small extra cost for the v4 addresses you'll need as they're no longer baked in. One reason to do this was because v4 addressing does actually cost money, but they also expected it would itself drive migrations to IPv6 and they were correct.
Because "We should fix our thing to work with IPv6" sounds potentially like a fun learning experience whereas "We should talk to our finance department" sounds miserable so they're obviously gonna put a few hours into making IPv6 work rather than take the "easy" option.
IIRC They did have customers who paid for IPv4 after the changes, but the vast majority went IPv6 only.
That was some repair! Bravo! And fantastic write-up - technical diagram included.
But then again if it's already dead you can't kill it more.
I live by this, and try my hand at fixing everything that I would otherwise junk. Some of the time, I even succeed!
I have a digital mixer for music recording that is also an audio interface for my computer. I had a somewhat known problem where 12 out of the 16 channels stop working.
Turns out what’s going on is the mixer is ‘gating’ the audio signal because the digital gate effect (that is built in) is claiming that it is receiving an input such that it will attenuate the audio signal.
I don’t even know how to diagnose a problem like this. It’s awesome watching someone who does!
The originals often have pretty bad markings on them, actually. Off-center, not aligned, not evenly printed. Too perfect is suspicious.
Also look at the pins. Sometimes they dip the feet into solder to make them look unscratched and shiny. I have avoided buying 'New Old Stock' ICs because I could tell the pins had solder on them.
A couple of times I found the IC was of the type advertised (after I removed the paint). I guess they just added a fresh coat of paint to make them look new!
A civilian part might be very close (or the exact same part with a different model number) and even better in some cases but as it was never tested you don’t know.
Many of the parts aren't really any different than what you'd find inside consumer goods. They cost more because they've been tested and certified.
If the components don't work in a military situation, people might very well die. Sometimes a lot of people.
A counterfeit microprocessor in a dollar store Furby clone probably isn't going to kill anyone if it fails. The same microprocessor in a surface-to-air missile? Well...
I knew Furbies had the same killing brains of missiles! It's in the eyes.
You are thinking of Home Alone 3 which is a cynical money grabbing shadow of the original two films.
Personally small soldiers is my favorite missile chip in toy genre movie.
Sure, but if the components do work, people might very well die. Sometimes a lot of people. Probably different people.
Military (and aviation) want uniform, reliable parts and to be able to do post mortem investigation of failures where they lessons learned can be applied to the installed base and future production.
You can't do that very well when you don't even know who made your ICs to ask them what happened during production or to improve their processes.
TBF, that's also true if they do work.
If you can depend on depend on your machine gun to jam if you fire it for more than 3 seconds, then you just don't fire it for that long. But if you have no way of knowing how many rounds you'll get out of it (if any) before it jams, you'll just stop using the thing.
Further, a lot of the procurement rules devolve down to "is this supply chain a strategic weak point?" We don't buy our HMMWV tires from Russia not out of quality concerns, but to deny them that weapon. So the testing is done to verify provenance, rather than quality.
If you're making a chicken feeder IoT device, the amount of marginal QA cost you can tolerate is low (in fact, you're probably tempted to use the counterfeits...) and the only adverse outcome is a slightly hungry chicken. Medical devices have a much higher risk, but not as much money or time as military (a bottomless pit of both), so the QA level is higher there, but probably not as much as defence.
There are these boutique one-man contracting shops, which are closed most of the time in places like Shanghai and Shenzhen. They are contracted to do R&D for manufacturers, and deliver on firmware, product design, software, etc. but again are one-man shops which are closed much of the time.
I am told this is how the pipeline of information works between state sponsored cyber attacks on big tech companies, and their Chinese competitors.
Talking about Defense contractors, I've heard stories from govvies I know about asian dudes following them around the DC area, constantly catching them looking over their shoulders at coffee shops.
That's not just a China thing, nor is it suspicious all by itself. I made a good living in the US doing exactly the same thing for years.
The specific shops you're talking about may have been nefarious (I don't know), but the mere existence of private contractors is not inherently suspicious. They're pretty common.
These one man shops deliver to Chinese firms (Like Huawei) hardware designs, firmware source and also other misc. software. All in one go, on contract under the auspice of "outsourcing R&D". Wayyyy too much for one person to deliver on, especially for a small office which is mostly closed.
I am not sure if I am mixing up anecdotes, but my source has mentioned that the materials provided often contain the same firmware bugs as a similar competitor's product does.
mil-spec parts need to meet specific standards, and since there are equivalent parts that don't have to meet these standards, the possibility and incentive for mislabelling parts is quite high.
What educational background and/or practical experience route would someone have to get into the field? Is this a “requires electrical||mechanical engineering degree” role or can it be learned via apprenticeship?
There are tons of eBay sellers who bulk buy from China and then resell with a huge markup but with the convenience of quicker delivery - and the slightly fake sense of being more reliable as they are UK based.
In my experience, most "fakes" are crappier parts marked up to be better, or chips that have failed QA. Rarely do you get a whole clone, although for some very high volume or simple, old parts, it's a bit more common.
If you like this, check out [zeptobars](https://zeptobars.com/en/), they post high quality dieshots, and sometimes fakes.
IR2113/IR2112 die: https://www.youtube.com/watch?v=e6DfBuPwAAA
And often the prevalence of the high volume parts actually makes it nearly impossible to find things using the originals.
For example, I once wanted a real FT232 breakout board (or at least a clone that can do most of what the original can do) to use some of the bit-banging features of the chip. But the most popular clones of that chip are very much reduced functionality variants that work as a USB-to-UART but don't do most of the other functionality. This was probably not helped by the fact that the original chips, and early more faithful replicas had onboard flash and allowed a degree of programming, but then the manufacturer of the original chip embedded malware in their Windows driver that detected and wrote bad data to the clones flash (which would have broken many products sold to consumers with them - technically fixable by removing the driver and re-flashing, but probably in most cases just becoming e-waste). So the clone manufacturers responded by simplifying: removing the flash, and removing most of the programmable functionality, and focusing on the UART part that was the most common use case.
Now, no matter how hard you try pre-purchase to make sure something is a full FT232 implementation (including asking for assurance about the chip up front) on a marketplace like Aliexpress or eBay, you are far more likely than not to got a limited functionality clone.
There are a couple of vendors who do this in the UK. They sell counterfeit crap while complaining about it loudly. I bet this was littlediode. Got into a large argument with them over sending out crappy clone / reject bin transistors a few years back.
Edit: at the same time I would like to buy a hell of a lot of 2N5458's and relabel them as MPF102's and retire. The MPF102 was an unbinned garbage JFET with wide characteristic spread. The 5458 is a binned one within the range of an MPF102. Same process. Of course though all the crap schematics out there on the Internet demand the MPF102 because that's what Radio Shack sold and no one knows how to substitute parts!
[1] https://www.yodel.co.uk/yodel-services/yodel-uk-delivery/int...
;)
They all have traceable lot codes and and entire chains of custody that go directly back to the manufacturer. And during Covid we were checking.
That traceability is the reason people buy from them instead of brokers. If you could prove what you are saying it would be a huge scandal.
How do you qualify them as being fakes?
I'm sure that even they get scammed from time to time, but I've never had a problem with my orders (or at least, none that I've noticed). Their prices are higher, but I consider their markup well-earned.
I'm really surprised by the length people will go to, in order to scam others. Someone took the time to do a fake plastic chip, or reprint a silk screen on an chip with a matching pin-out. My favorites are fake USB drives and "fake" GPUs. The amount of time that has to go into finding old GPUs, faking a driver and relabeling seems like it would exceed the value of the scam, but apparently not.
Also, there's this interesting exchange I had with an eBay seller a couple of years ago. I purchased a bunch of NiMH batteries. Because I test every component I buy even for my hobby projects, I immediately noticed that they were 1/10th the capacity that they should have been, and were marked as.
I contacted the seller to return them and get a refund. They told me "honest mistake" and then began a negotiation. "How large of a refund would you accept?"
Since the batteries were useless to me, I said "100%". In the end, I did get a complete refund and didn't need to return the batteries.
Fun thought: since eBay often rules in favor of buyers, it would be fun if someone would 'abuse' this: buy a good # of often-faked parts from all sellers that offer them, test extensively, and open disputes for any fakes received. That should (in theory) cause a good deal of $ losses for any seller that sold fakes.
Maybe repeat regularly, and it might remove some fakes from the market?
Back when everyone was making Windows 8 tablets with a quad-core Atom and a $350 price tag, I ordered one of theirs. It got good reviews in the Maximum PC comparo of the time, and it was available with a then-cavernous 64Gb of flash.
In week 1 the Wi-fi wouldn't keep connection. By week 2 it had fully bricked. I initiated the RMA process, and the chat monkey asked "If we refund you $50, would you keep the device?"
I ended up getting the Asus version which had a nifty little clip-on keyboard, and it at least lasted the warranty term before the touchscreen cracked.
I did end up getting a partial refund for about the price of a replacement OLED screen.
If you're going to commit fraud, at least be honest about it.
Second, being honest about commiting fraud is counter-productive. Comercial fraud, even though illegal, is still a commercial enterprise at its core and its goal is the same as of any other comercial enterprise: to bring in profit, the more the better. Fraud is not about sending a message on ethics, or showing off someone's character, or helping the marks to keep their money, it's about getting as much money as possible at little cost as possible (with the costs of dealing with legal persecution accounted).
They make fake skateboard bearings. These are like 10 cents a piece in bulk.
2. If the customer doesn't use the product right away, having it look close enough means they might not notice its counterfeit until after the N day return window has expired.
You say on your end it was real chips, and that the other guy is faking it by claiming they're plastic chips.
AN-002: Replacement Notes for Obsolete Versions of 6502 8-bit Microprocessors https://www.westerndesigncenter.com/wdc/AN-002_W65C02S_Repla...
They responded to JTAG correctly, but none of the other documented bootloader modes, which was a dealbreaker. I don't think they were actually fake, but certainly defective or at least incorrectly documented.
Every single time I've tried to use an ST device I get bit. I'm sure it's just me being unlucky, but I avoid their parts now when I can.
Like the author of this article, I had some performance problems with the part, but I was sure I had a fake when I looked at the printing on the package of the chip.
Are there so many original 6052s around that it's cheaper to just relabel those? (I guess cheap 6502s were really everywhere in the 80s/90s)
Say if I find a batch from a scalper but I am not sure if it is genuine.
But: shipping, and low $ per part doesn't make it worth the effort. If you're buying a $3 cpu it just isn't worth it to re-route through a testing service for several additional $ per pc. + shipping.
Now if talking about say, $100+ / piece parts, sure. But then the testing services would have to be extensive, professional, and equipped accordingly. Above my pay grade, so to speak.
But yes it would be nice if such service(s) existed.
Now my imagination is running wild...
I haven’t used DangerousPrototypes for probably 5+ years, but they’ve got a decapping service that you send parts and they send back a die shot.
Even though hobbyists have setups for decapping or de-packaging, it's going to cost a lot to actually do this as a business, most of the techniques are dangerous and use toxic materials.
Depending on the sensitivity and risk to a particular product, it's not uncommon to have a suite of testing done on a sample from every received shipment of a particular good, or set of goods. Testing typically goes beyond counterfeit detection to being within manufacturability and usability specs (e.g. solderability, lot code validation, date code validation, etc.)
Usually priced out of the hobbyist range, think 2-3k to do basic suite verification on a sample of 5 components. That would not include de-capping and die examination.
Fun fact: the ARM is largely based on the 6502. You could argue that the 6502 was a RISC chip before RISC was a thing!
The Page 0 addressing mode basically let you use the first page of RAM more or less as 256 extra 8-bit registers (or 128 16-bit registers, etc.), as those locations could be accessed faster and with fewer instruction bytes than locations elsewhere in RAM.
A fair number of languages were implemented on 6502-based machines by using multiple Page 0 locations to simulate wider registers.
Well, it is a touch hyperbolic, yes, but the 6502 could at least be considered a proto-RISC.
But I suspect you're right if it came to actual manufacturing of counterfeits.
Or maybe more like 阮 which means "Ruan" (maybe that's a person's name?).
Is that word plausibly used to refer to devices to mean "it works" or "it passes tests"?
> The logographic character you're referring to appears to be a Chinese character, but it's not clearly written which makes it difficult to identify with certainty.
(There was more to the response, but it was more about why the character might be written there)
Note that in some cases like with TI, this is viable, even if making small orders.
You can easily figure it out as the date code on the chips is unique to them and they still sell them on eBay.