Inside A Fake LM358

[IMSAI Guy] got some fake LM358 op-amps. Uncharacteristically, these chips actually performed well even though they didn’t act like LM358s. [IMSAI Guy] did a video about the fake chips and someone who saw it offered to analyze the part compared to a real LM358 to see what was going on. You can see it too in the video below.

A visual inspection made it obvious that the chip was probably a fake. X-ray analysis was a little less obvious but still showed poor quality and different internals. But the fun was when they actually decapsulated the part.

The die on the fake was labeled C665 on the die. The die was significantly different, but it isn’t clear why it wouldn’t be a more direct knockoff. It doesn’t appear that the device is cheaper to produce and it wasn’t a proper die rejected by a regular manufacturer.

If you want to see how the “C665” works, the original video that started it all appears below. The video shows the slew rate response was faster than a real LM358. The step response didn’t make a nice slope like the LM358 either. We hope he (or someone) will trace out the circuit and see what the circuit actually looks like.

Sometimes fake parts are good enough for hobby projects and sometimes not. Want to spot a fake? There are several telltale signs.

23 thoughts on “Inside A Fake LM358

  1. The thing that bothers me about fake components (other than someone scamming you) is that while it may perform to spec, and the way you expect it to perform when used properly, it may not perform the way it would when used out of spec, or for some weird solution.

    Engineering is full of solutions that misuse(?) parts and get a cool result that no one expected but lead to further explorations and learning. A beginner might have a hunch about how they could use a device which is fundamentally right, but because someone scammed them to make a quick buck, that experiment fails and the experimenter is just left thinking that their concept didn’t work out.

    1. The only thing you should expect on is that parts work to spec.

      Anything more than that is sheer dumb luck and you’d have to be a fool to expect that to happen reliably and repeatably, heck the chip you’ve abused might not perform the same way reliably twice in a row.

      There’s zero guarantee that even the next genuine chip you take out of the packet will exhibit the same behaviour *precisely* because it’s out of spec.

  2. In the 2nd video, I found the “ski slope” @3:30 suspicions. First because I don’t know why it would be there, and second because it’s duration is the same as the rise time of the LM358.

    The Opamp @ 04:48 also has this same ski slope, and it disappears when the LM358 is removed. So my guess is it’s some kind of interference caused by the “test PCB” or wiring. Maybe a lift of the GND level somewhere.

    Also, If I notice that any of the electronic parts are black-topped, I always give negative feedback. Regardless of whether the part works. I just do not like being lied to. I can deal with mentioning “lm358” (or whatever) in the title for “search result optimization” (sigh), but if it does not clearly state in the description what part it really is, then it is instant negative feedback.

    1. Happened to me. I purchased a bunch of ATTINY4313 (which is the ATTINY2313 but with double the flash, eeprom, and ram storage). I received them and tried to use them. Flashed fine, but just didn’t work.

      Then I had a better look at the chips and then noticed that they were remarked chips. But I could flash them fine, verification succeeded with my program that was 3KB, also verified that it wouldn’t flash and verify in an actual ATTINY2313. The chip id was correct for an ATTINY4313, and everything just seemed right, except the markings on the chip.

      Then I finally thought about checking the fuses, and that was where the problem turned out to be. The fuses were not at factory default but had been reprogrammed.

      So the chips turned out to be used chips, probably erased flash and eeprom, but not erased fuse bits. Then they were remarked to look as new.

      After programming the fuses to what I needed (basically the default), the chips worked like a charm. No hitches, nothing, flash was 4K, everything was fine.

      My guess is that they were remarked because Atmel recommended to only use ics of max 5 or 6 years old (I can’t find it anymore). I guess these were older and used chips, and remarking them would make them more sellable.

      There’s still a small chance that they were actual working counterfeits, with a different core emulating the ATTINY4313. I don’t really know, never decapped them.

      1. I bought 4 identical “Atmel” parts (something in the 8051 family). Weird thing is they all had been blacktopped and with identical laser markings, yet all 4 had radically different 40-pin dip leadframes: The dimple at pin 1, the slot at the end of the package, the package mold, and the pins themselves. The silicon ID read as Atmel, and they nominally worked, so I’m not sure what the story was. Acetone q-tip definitely was removing the black overcoat, but the sanding meant you could never read the original part number.

      2. The device ID is actually stored in flash too, usual programmers does not allow you to write into that memory region though. So my guess would be that these were ATTINY2313 with ATTINY4313 ID flashed and repainted markings.

  3. I breathed on my knockoff lm358s and they had a wobbly.

    In one design, I use the LM358 as a low cost, very effective, thermocouple amplifier. Although the 358 may have a significant offset, the thermal drift is pretty low, and the circuits are all individually calibrated, so the offset is negated. More than good enough when looking at temperatures of 1000 degrees with a K type probe.
    The usual TI or onsemi products are rock sold. However, just breathing on the knockoffs cause an output error signal of hundreds of degrees. Quite impressive. Not thermally stable at all.

  4. Very cool seeing them side by side like that.
    A while back I got some oddball chips, LM3124’s, I was working on a sequential LED display and the minimal parts solution was to use one of them. I tied the input to a simple RC circuit, being careful to keep the R part large, and the last output did not drive an LED, it drove an opto triac that shorted out the cap, starting the sequence over. It worked well. Anyway, I got the chips from China and I think I got 10 of them for $2 or something like that. My first thought was they have to be random floor sweepings and not the real deal, but I stuck each one in my little demo on protoboard and son of a gun they all worked. What I am wondering if they copy the original or what. I guess out on a limb they could even be real but I doubt that.
    I should mention my set up did not test “features and exact specs” but just crude functionality.

    1. Frequently I get obvious re-badged pulls at that price. Then bin them.

      For my use case and with what I’m buying that’s actually what I want. I’d actually pay extra if they didn’t wipe the original markings.

  5. I noticed the bottom bond pad’s weld on the knockoff just about contacted the trace next to it. In fact it looked like all of the bond welds were toward the top side of the pad.

    Makes me wonder how many rejects were due to poor bond wire alignment and how many early failures it would cause for those that pass initial testing… assuming there is even initial testing.

    1. Fakers probably have this line of say 25 dies each one roughly replicating the functions of some popular family of opamps, then according to demand they put them into a case and print them with what is being sold most at the moment to obtain something similar but that will perform quite poorly compared to the originals, assuming it works at all. Having a restricted line of production lowers the costs significantly, so even faking a uber cheap thing like the lm358 can turn in a profit.

    2. Because you buy excess, surplus, bankrupt or salvaged parts for pennies on the dollar and any 8 pin chip with ground off markings, ready to relabel, that’s sat on your shelf is cheaper than the genuine part you need to order in and manage stock levels of

  6. So here’s what’s going on: this is a legitimate chip, just not made by TI. There are a lot of companies that make “lm358” chips of their own, that are designed to be compatible to some degree with the original lm358.

    Take a look here:

    Lots of different lm358’s, each with their own datasheet. many likely won’t perform as well as the ‘genuine’ part, but for some designs that’s not needed, and let’s you save a few cents. check the actual datasheet (not just TI’s lm358 datasheet) to know.

    what becomes a problem is when random no-name resellers of chips on aliexpress or amazon or whatnot have a listing that just specifies “lm358” or such. you don’t know really what specific chip you’re getting. especially when you’re trying to just find a few super cheap chips. If it’s cheaper than you can find from the official resellers (think digikey, mouser, lcsc, and such) then it’s probably a functionally similar substitute.

    If you need the real chip, buy it from an official seller that specifies the manufacturer and gives a datasheet. If you buy from a no-name source that has no information beyond ‘358’… don’t be surprised when they sell the cheapest thing that works well enough to pass.

    Not to be that guy… but this is kinda a well known thing in the electronics world at this point. Making a headline and article about “FAKE lm358” seems rather like clickbait. Or maybe someone just didn’t do their research before writing an article. Or maybe that’s on whoever made the youtube vid. Who knows. Either way, it surprises me not at all that the die was different than a genuine chip. I don’t see the need to dramatize that rather than actually write something informative about why that’s the case.

    1. “but it isn’t clear why it wouldn’t be a more direct knockoff” cause the fake semiconductor industry consists of sanding off part numbers of real chips, and re-etching them with the part numbers of more expensive chips. At least for legacy + simple chips like this where there are many that do about the same thing out there now.

    2. There’s a good point in there. If you actually need a cheap pin-functional alternative find it on LCSC, grab the spec sheet, then search out the part in qty 10 on Aliexpress. You can probably find it with the real markings.

      1. true in my experience. for example, i just bought a dozen tl074s from aliexpress and they’re definitely not genuine TI parts, they don’t even behave like fet-input op-amps. completely fake.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.