Counterfeiting of chips is such a big problem that the US DARPA has a major program to develop tiny cryptographic chips that can be embedded inside chip packages to prove their authenticity. It's called the SHIELD program (solicitation number DARPA-BAA-14-16 if you want to ask for money).
I'm not an EE, but I imagine that a much better solution would be a universal device where you can connect a chip and run a publicly available test suite against the chip to confirm that is performs as expected.
Like @wyager, I'm extremely skeptical of a program like SHIELD.
They think that the embedded chip (which they call a dielet), which responds to a crytographic challenge from a handheld device on an assembly line, can avoid testing, which is much more time consuming and complicated. To test, you need to connect the device's pins to something and power it on. These devices are for military systems: you'll need to test lifespan and performance under a variety of environmental conditions. I don't know if the dielet idea will work out, but it's specifically to verify authenticity without having to test.
A public test suite wouldn't reveal features introduced by a hostile foreign government, such as a security backdoor that only activated in response to a secret key too long to brute force. Presumably for DARPA this is a worry.
Agree a public test suite would be good for other fake component risks though.
Both. Counterfeiting, or substituting recycled parts, can be for economic reasons, or can be a hostile government that managed to substitute its hardware into the supply chain, for sabotage or to create backdoors.
The trimmed resistors are the four brown rectangles with notches (one is shown in atomlib's pic).
The silvery surfaces are aluminum - seems a single layer of metal process. Some large plates may be capacitors.
In analog circuits, MOS transistors are often fairly big, to carry larger currents, sometimes in the tens or hundreds of milliamps. There are a few such large devices, where you can see an interdigitated structure: think two hands open flat with the fingers interpenetrating but not touching. One hand is the MOS source terminal, the other is the drain terminal, the space between is the gate that controls current flow in the channel between the two first terminals.
Here is image Mikhail Svarichevsky (author of that blog post) posted as answer to this question in comments on Habrahabr http://i.imgur.com/vC5VkH0.jpg
The interesting thing here is that the part has been substituted with another part that's comparable in quality and only slightly less expensive. The margins for counterfeiting this part can't have been very good.
It's pretty common to counterfeit electronic parts. I've come across counterfeit op-amps, exotic transistors, power transistors, ordinary transistors, capacitors, and even fake resistors which is amazing. It always just makes me wonder what is wrong, collectively and institutionally, with China? Making counterfeit goods of all kinds is apparently a common line of business in China. Is it just because _everything_ is made there so counterfeiting has proportionately shifted there as well?
While most of the discussion here is technical from engineering and business perspectives, your question digs more at the root of the problem: this is a cultural and institutional problem.
I live in China: people pinch pennies (or mao) here without taking into account the overall cost. This is why you have a large market for imported goods; even amongst the most nationalistic Chinese consumer, there is an assumption that imported goods are better because domestic products are made by cutting corners.
There's a strong suspicion of things that are priced cheaply but, bizarrely, people still buy cheap goods but are resigned to the product failing rather quickly. This just doesn't apply to electronics, but also to clothes, vehicles, etc.
There's a very strong tension between value-pricing and product quality. That is, there is a demand low prices but assume that the product quality is poor. On the other side of the spectrum there is a demand for high-quality "luxury" goods too. Thus pricing your product too low, even though it may be a quality product, basically indicates that it is a low-quality product.
Part of the myopia here is that people will cut corners without figuring out the benefit/risk ratio. That is, is it worth losing a contract to improve your bottom-line by a negligible percentage? Culturally (and I'm speaking in a broad generalization here) there's a desire to "win" at a contract... to make the contract be better for yourself than what's on paper. So even if the contract spells out your profit margin, it's a matter of pride to be able to improve on it (aka, "sticking it to the man"). With China's large population and widening gap between rich and everyone else, there's a perception that you have to get rich or die trying; no one else is going to help you... certainly not the government.
China's future could be a lot brighter than it is; the problems aren't technical.
Fake resistors? How is one supposed to fake a resistor? There are no manufacturer markings on them, just (easily testable) resistance/precision indicators...
Resistor is more than resistance. There is also other characteristics like power capacity, accuracy, tolerance, frequency responses, temperature coefficient etc.
The cheaper resistors are typically very loose for most of these characteristics. A tight characteristics means a high level of quality control at manufacturing, and this is usually lacking for the fakes.
For Op-Amps, the interesting spec is their frequency response as the op amps are not typically used in DC situations.
Especially if you get a batch that was manufactured in the original factory, and the faulty one put on the grey market. Indistinguishable product, but guaranteed and tested to be out of spec (might be a important parameter, or not).
Best interview question for EEs, IMHO: hand him/her an ordinary through-hole resistor, and say, "Tell me everything you possibly can about this component."
As a hobbyist, I'm curious: other than what the markings say (value and tolerance) and an estimate of power capacity, what would you expect the interviewee to be able to say about the resistor? Assuming you're not interviewing in a lab.
- Can they tell me the value and tolerance of the resistor by reading its color code?
- Based on the physical size/package, what's the resistor's power rating? Estimate if you don't know. Does pulsed versus continuous power make a difference?
- What kind of resistor is it? Metal film, carbon film, carbon composition, wirewound, ...?
- What's its temperature coefficient like?
- About what does it cost in small and large quantities?
- What kinds of parameters would you expect to see in a SPICE model?
- What are its noise characteristics? Johnson noise, flicker noise? Say I handed you a metal film resistor -- when would you want to use it instead of a carbon composition resistor?
- What does the resistor look like at RF? If it's a wirewound resistor, or one that was constructed by cutting a spiral groove into a film or substrate, what effect might that have at high frequencies? Estimate what its equivalent AC circuit might look like.
- Any concerns about quality, aging, and reliability?
- How about voltage coefficient of resistance?
The interviewee wouldn't be expected to go into much detail from memory, but they would definitely get extra points by just mentioning some of the factors. It would be fine if they answered the last question by saying something like, "There will be a certain change in resistance with applied voltage. It's going to be small and will almost never matter, but for precision applications I'd want to look it up on the data sheet." Basically I'm looking for an acknowledgement that there's no such thing as a truly linear or "boring" part.
That reminds me of a common practical joke from EE lab in college (~30 years ago): take the "right" value of carbon resistor (probably 1/4 watt, I forget how resistance mapped to time delay), plug the leads into an ordinary 120v power outlet, and walk away.
As the current flows through it, it'll warm up a little. Because carbon resistors have lower resistance at higher temperatures, as the resistor warms up, it'll start conducting more current, which will make it warmer, until a few seconds or minutes later (depending on resistance value, ambient temperature, etc.) ... "bang!".
All real components have real impedance, not just resistance. So at a minimum you'd expect a new EE grad to at least model a resistor as a combination of resistance, capacitance, and inductance, and explain what contributes to each (i.e. a wirewound resistor has more inductance than a bulk metal foil, or something). For veteran industry EEs you'd also expect them to mention things like microphonics, self-heating, thermocoupling, etc.
For commodity resistors, sure. There are high-performance resistors though, and they are usually packaged with a manufacturers mark (e.g. some of Vishay's line).
Also, the resistance of the resistor is only part of what I care about when I specify the part. Temperature co-efficient and the technology behind the manufacture can matter quite a bit -- and those are harder to test.
In addition to what has already been stated, components have additional concerns about environmental tolerance and long-term reliability. If two chips test the same on a test bench, that just means there is no difference at this point in time and under the conditions of the test bench. It says nothing about vibration susceptibility, moisture susceptibility, EMI susceptibility, ESD sensitivity, operational and non-operational temperature ranges, or a host of other factors that affect reliability and manufacturability.
Exactly this. There are many cheap (I mean cheap, not merely inexpensive) versions of "industrial" electronics on eBay. Most of them will probably work just fine in an office environment at room temperature. But put them in a factory where the temperature near the doors in winter can be below freezing and up near the roof in the middle of summer it's well over 100F, then your cheap eBay-purchased sequencing timer starts having hiccups, dumping hot asphalt on the floor.
exactly. We use them e.g. for amplifying signals where low noise matters. Never compared against the AD device but did compare them once against your standard 741 and the difference is obvious, eben simply by looking at the signal on an oscilloscope.
According to the blog post fake opamp has higher noise levels and higher offset voltage. But I would love to see it tested on audiophiles, I'm sure they can't hear the difference.
I was being facetious; but my point was that for audiophiles, you shouldn't need to be disassembling components in order to tell the difference between systems.
Sure, if you can notice a difference, then go ahead and track down where it is coming from. But if you start from the wrong end (i.e. the 'AD744 has higher noise and higher offset voltage') then you should stop and question your actions, IMO.
Except that it's only one component of many in a system. I doubt there would be a measurable performance difference between a standard formula 1 car, and one which has had it's carbon fiber steering wheel swapped out for an aluminum one. But making a lot of small, nearly immeasurable changes can result in noticeable differences.
Certainly, in the case of racecars if two components differ only in weight, we can infer in what ways this should affect the performance. With audio it's not quite as clear (to me). Is there a measurable difference in the signal between using copper and silver wire? I don't know. But if we know that one component has measurably higher noise, that seems like it has the potential to result in a "cleaner" signal.
However, without going too far down the rabbit hole, I generally agree that in the end the important question is, "does system A sound better than system B".
Nope, we are talking about replacing the engine for one of a completely different model.
Those amps and a very small number of resistors and capacitors are the first stage of amplification of an equipment. They are the biggest source of noise, by a couple orders of magnitude. If you can't tell the difference once you change them, it's because the difference is not relevant (and for audio systems, it's completely irrelevant).
It's hard to notice subtle things without an A/B test, it might just be that quiet parts of certain songs just don't sound as good.
For something like this you would want to measure the noise levels, not make assumptions about audiophiles being wrong; they're only wrong a significant fraction of the time.
There's no reason for an audiophile to test if the part is genuine. They would need to buy a "genuine" part to compare against, but that presumes knowledge about the original part being fake.
The difference in noise is readily measurable. I've done op amp based design for commercial products, including an audio product for my side business. (n.b., it's not targeted at the "audiophile" market). I've measured the noise of dozens of op amp types for a specialized application. Impressively, the measured noise performance typically doesn't stray very far from datasheet values. You can measure it with a few dollars worth of parts, and the audio input of your PC.
Now, it's not a slam dunk that every application should use the quietest chip. Noise trades against some other factors such as current consumption. And of course from an engineering standpoint, cost matters too.
Small parts in complex systems may make drastic outcomes (e.g. Heartbleed bug). In many cases there are so many small components that perceiving such a "difference" is a non-trivial task. Thus, not being able to identify from whence the difference sprang in the context of a complete circuit does not entail a lack of any difference. tl:dr: Difference is difference: a tautology; or rather: no difference != difference
For audio yes. For the stability of your circuit it can cause mayhem if you are doing wonky things. Often can be solved by adding extra capacitance, but can be quite tricky to do (no room on PCB, capacitance was already large, ...)
They also may use lasers to trim and tune the mostly-crystal oscillators used in timing circuits, such as the small quartz "tuning fork" found in many digital watches.
Decent optical microscope is more than adequate for photos like that; electron microscope is for gate-level.
Zac Brown and Adam Laurie are doing awesome stuff (with cheap equipment, relatively) in chip teardowns; basically the stuff Chris Tarnovsky (flylogic, now ioactive) does with $5mm stuff, they do 90% with $5-10k.