Beware fake parts
There are a lot of counterfeit electronic parts in circulation. It's especially a problem with parts that are obsolete, hard to find, or expensive; sold privately in places like eBay or AliExpress at unusually low prices; and shipped from the People's Republic of China. However, just this week I had to deal with a batch of fake TL074 op amps from a well-known US distributor. Here are some notes that might help others avoid, or at least debug, similar problems.
The TL074 is a very popular quad JFET-input op amp chip, used in many synthesizer circuits. I use this type in the Leapfrog VCF, Fixed Sine Bank, and Middle Path VCO (and the closely-related TL074B in a few other products); basically everywhere I need an audio op amp. Although this is a good-quality chip when it's genuine, it's not a particularly expensive chip. It might retail for US$0.35. It's still in production, available from more than one major manufacturer, and not normally in short supply. One wouldn't think the TL074 would be a big target for counterfeiting.
The conventional wisdom is that counterfeiters will go for stuff like the out-of-production µA726, which is used in a lot of classic synthesizer circuits, has no good modern replacement, and might be worth over US$100 to a hobbyist today if genuine. That kind of target provides a lot of margin to cover faking the packaging, putting some kind of chip in the package that will fool the simplest acceptance tests, and so on. "Black-topped" chips are notorious: a cheaper chip, or even an empty DIP package with no chip inside, may have its markings on top covered by textured black paint and then new markings applied to make it look like a more expensive chip. Another dishonest practice sometimes observed is to sell chips that are really the type they're claimed to be, but are used (possibly non-functional) chips pulled out of old equipment and dressed up to look new; and it's hard to imagine a chip that sells for US$0.35 being worth that effort.
So I wouldn't normally think I needed to be as cautious about TL074 op amps as I might be with, say, AD633 analog multipliers (much more expensive and rare). Ken Shirriff did some reverse engineering on a counterfeit AD633 recently. But there has been a supply crunch on TL074 chips recently. Very likely the global pandemic is at least part of the cause. I recently ran out, and my usual suppliers were all out of stock. I've heard from other people also having trouble sourcing them. For a while it seemed like there just were no TL074 chips to be had; and I needed some more to continue producing my own products. So I went looking at alternate suppliers, distributors who wouldn't normally be my first choice. RS Components said they had this item, and I tried to place an order there, but they rejected my credit card, so after a couple of tries I decided to keep looking. I eventually went to Jameco - even though I had misgivings about dealing with them.
Jameco Electronics is a popular distributor of parts to hobbyists, based in California. Almost 30 years ago, in the early 1990s when I was a teenager, I had very little money to spend, and the Net wasn't a thing available to the general public, Jameco sold me a shoddy drill press that I couldn't use and couldn't get refunded or replaced. I decided I didn't want to go back to them again. But now in 2021, they seemed to have the chips I was looking for, nobody else did, and I figured, okay, maybe they deserve another chance. After all, it's been a very long time. The point I'd like to make is that although I had a personal reason to avoid Jameco, and they're certainly a lower tier than Mouser or Digi-Key, Jameco is not some fly-by-night deep-discount seller on eBay or AliExpress. Hobbyists have been using and trusting Jameco for many years and they're not an outfit one would normally expect to catch selling counterfeit parts.
There were a couple of warning signs early on. For one thing, although I paid extra for a higher-tier UPS shipping service that I thought would include Customs brokerage service at no additional charge, I still ended up having to pay an additional charge for brokerage, talk to UPS customer service on the phone, and wait an extra day for the delivery. I've got an account set up with UPS so that brokerage charges will be automatically charged to my account, and that has worked fine for other packages; but for some reason (and I will probably never know why) it didn't happen on this one package, this one package had to be paid manually, and there shouldn't have been a brokerage charge on this package anyway. The point being: delivery of the package was irregular, not what I expected and not like other packages. But all that could easily have been UPS's fault, not Jameco's, and it shouldn't necessarily bear on the genuine/counterfeit status of the chips inside the package.
I unboxed the package live on my Twitch stream and noted another irregularity: the chips were packaged just in their anti-static rails rubber-banded together. The rails weren't in a shield bag. I don't think I've ever seen a major distributor ship DIP ICs without a shield bag before. I myself have a big enough pile of those bags sitting here that getting rid of them is a bit of a waste disposal issue. So, fine, I can put the rails in a bag of my own, that's not a big problem. Maybe that's just how Jameco does things. And these chips are not particularly static-sensitive anyway, and they were in anti-static rails which should be good enough to protect them. Not a big problem. But: it was another irregularity.
I later noticed that the anti-static rails on the genuine TL074 chips I still had on hand from another distributor, were printed with the words "TEXAS INSTRUMENTS STATIC DISSIPATIVE," and some numbers and recycling information, while the rails on the chips in the batch from Jameco had no printing. But that's also not particularly damning. Texas Instruments probably has many variations on the details of the anti-static rails they use, and chips bought from a distributor might be repackaged from the way they came from the factory, without anything untoward having occurred.
However, when I took the chips out of the rails to examine them, the red flags started going up. Here's a photo of one of the chips I got from Jameco last week (bottom) compared to a believed-genuine chip I got from Mouser in October 2020 (top).
Several differences are obvious between the two chips but most of those differences could be innocuous. The chips are slightly different shades of grey. The notch at the end (indicating pin 1) is a different size and shape. There is a round indentation on the lower chip, also indicating pin 1, but not on the upper. Both sets of pin 1 markings are within specification and I've seen those kinds of markings on genuine chips; as far as those points go it could be that the two chips were made in different factories with different moulds. TI has many different factories. The typeface is different in the etched markings (wider on the lower chip) but that could also be a factory-to-factory variation.
There are differences in the leads and plating: the upper chip has narrower leads with slightly oxidized, once shiny, tin plating. The lower chip has brighter, unoxidized, matte tin plating. The oxidation difference could simply be because my older chip is older. I haven't been able to find decoding information for the date and lot codes in the first line of the chip markings, which would allow me to figure out how old each chip is (or claims to be, in the case of the fake one).
The mould indentations (circular features toward the right end of the chips in the photo) are different but neither of them looks suspicious. It's considered a dead giveaway of blacktopping if a package is textured inside those indentations, because the normal manufacturing process makes most of the surface textured but the mould indentations smooth, whereas the paint-like blacktopping material will leave the entire surface textured including the mould indentations. But both chips are smooth inside the mould indentations: no indication of blacktopping. Similarly, on the bottoms of the chips, one has one and the other has two mould indentations, but the mould indentations themselves look fine - smooth surface and although these doesn't show up well in my photo, they both have crisp text markings inside the mould indentations, which is another thing counterfeiters sometimes get wrong.
I tried wiping both chips with acetone, because that's another popular test for blacktopped chips. Real ones are supposed to be made of plastic that resists acetone, and marked by laser etching. When you wipe the acetone over the surface, the markings briefly disappear (as the liquid fills in the roughness created by the laser) and then seem to "pop" back into existence as it dries. If the surface or the markings smear away (because they're paint, dissolved by the acetone) that's an indicator of blacktopping. But both chips passed that test.
The biggest warning sign on visual examination had to do with the Texas Instruments logo in the upper-line marking. On the real chip from Mouser it's a fairly accurate representation of the map outline of the state of Texas, filled in, with a readable letter "i" superimposed, and the entire logo is about the same height as the text. On the fake chip, the logo is taller, and crudely formed: bigger rectangular notch at the West, no sharp point at the South, the letter "i" is only approximated, and there are squiggles around it. Texas Instruments has used a number of different variations of their logo, and there is a not-filled-in version they've used on some chips which looks a little like this one. But the one on the chip looks more like it's "Tianjin Instruments." Somebody unfamiliar with the map of Texas probably drew it with a joystick on an industrial laser-etching machine somewhere and called it close enough.
I have since found a report from user ChristianBloch on the Look Mum No Computer Web forum, who seems to have gotten fake TL074 chips on AliExpress that visually resemble the ones I got from Jameco.
I tried putting the suspect chip into one of my Middle Path VCOs as U2 (triangle buffer and sawtooth and pulse shaper for channel A) and tried the different features, with the frequency set to a few hundred Hertz. At first I thought it was working - all the outputs coming from that chip looked the right shape on the oscilloscope screen. That suggested the suspect chip was at least a quad op amp of some sort, doing the things in the circuit that an op amp should do, and I almost decided "I'm being silly, this chip is fine, a reputable distributor wouldn't be selling fake ones" and dropped the whole thing. But, I decided to run a few more tests.
I tried using a multimeter's "diode test" in both directions between each pair of pins on each chip and here's a photo of my handwritten notes from that. This is the kind of test sometimes used for identifying an unknown chip.
The point here was to infer the nature of the circuits inside the chips; and the important result for me wasn't the specific voltages but the fact that the pattern is different. The set of measurements on the left (from the old batch of chips) shows a forward voltage from every op amp output and each of the two power pins, to all other pins. That might well result from protection diodes built into the circuit, and the kinds of stray diode junctions we expect to see in a classically designed analog chip. The set of measurements for the suspect chip on the right shows some of the symmetries we might expect of a quad op amp chip (in particular, note how the pin 1 and pin 14 columns are basically the reverse of each other) but there are a lot more pin pairs with no forward voltage, and it's clear that the two chips contain different circuits with diode junctions in different places. So this is building up the picture that it's a quad op amp chip but it's not a TL074.
I put the suspect chip in the Middle Path again, turned up the frequency, and scoped the pulse output, and here's the result. As I said on Twitter, it's a smoking gun.
Old chip at left, new at right. There's just no way the new chip can be a genuine non-defective TL074 and do that.
The left trace, from the believed-genuine chip, is a sharp square wave, even at 20kHz. The rise and fall times are under two microseconds. The right trace, from the fake chip, is rounded off and slew rate limited. Op amps normally have a "slew rate" representing the maximum speed at which they can pump their output voltage up or down. On large step signals they usually work by switching a fixed current around to charge an internal capacitor up or down, which causes the voltage to go up or down in a straight line. Those straight-line slopes on the voltage are pretty much classic op amp behaviour. The problem is the steepness of the slopes: the fake chip's slopes are much less steep than those of the real TL074.
The test circuit used for those scope shots isn't ideal. I just plugged the chips into my MSK 013 prototype and scoped the output because that was convenient, but in fact that circuit is using the op amps as comparators (on the one hand: we're traditionally told not to do that; on the other hand: it works), and the output I'm measuring is not the direct output from the chip but the module output downstream of a Zener-diode clipper. So in this circuit the op amps are being pushed pretty hard both for switching time and output current, and I think that's why we see effects like the rounding at the top of the pulse in the right-hand trace. Very likely that chip, whatever it is, isn't able to drive as much current as the TL074 would, so as the clipper starts to kick in, it can't drive the voltage up much further and the pulses get rounded. I think the comparator application and Zener clipper are also the reason that if you do the math from the voltages and times measured by my scope, it appears that even the genuine TL074 is not hitting its full specified slew rate here.
But even with those caveats this measurement makes it quite clear that the chips in the batch from Jameco were not genuine, normally-functioning, Texas Instruments TL074 op amps. I sent my photos and scope shots to Jameco customer service and although they didn't directly acknowledge that the chips were fake, they agreed to give me a refund under their guarantee.
Before packing up the chips to send back, I did a live stream in which I tested them a little more thoroughly for slew rate, in a more proper slew-rate testing circuit; and I screwed up by using the wrong format of USB key saving the screen shots and I ended up losing all the measurement results. Between that and the mistakes I made along the way while trying to do the measurements on-stream I ended up with very little to show for the effort, and it's kind of embarrassing and I'm not sure I want to link to it directly, but at least for the duration of Twitch's archiving you'll be able to find it on the "previous streams" page of my Twitch channel.
Please note that I no longer have the fake chips. As of this writing they are en route back to California, and I cannot do further tests, send samples to people who want to decap them, etc. I will delete comments from readers who fail to understand this point.
Mouser claims to have gotten TL074 chips back in stock as of yesterday. They haven't shipped my backordered batch yet, but I hope they will today. I'm out of pocket for the brokerage fee and the shipping of the bad ones back to Jameco, but they say they will refund the original payment I made to them for the chips and shipping to me, and assuming they do that, I end up not really losing much. And it may be another 30 years before I order from Jameco again.
It's a good thing I was cautious enough about chips from a non-preferred supplier to take a close look at them and run the tests, because it would've been bad if I had shipped those fakes to my own customers in kits or assembled modules. This goes to show part of the value of buying a full SDIY kit from me instead of trying to do your own sourcing with the "PCB and panel" products available from some other SDIY vendors. With a full kit, you know I've already done the sourcing work and I've made some effort to be sure the parts are genuine.
The take-aways for hobbyists are that even a popular distributor may be selling counterfeit parts; even inexpensive easy-to-find parts may sometimes be targets of counterfeiting; and counterfeiting may go all the way back to the original factory (brand-new chips that really perform the right general function, falsely labelled right from the start) instead of classic blacktopping or used chips sold as new.