Cheap Power Supplies With Fake Chips Might Not Be That Bad

We all know the old maxim: if it’s too good to be true, it’s probably made with fake components. OK, maybe that’s not exactly how it goes, but in our world gone a little crazy, there’s good reason to be skeptical of pretty much everything you buy. And when you pay the equivalent of less than a buck for a DC-DC converter, you get what you pay for.

Or do you? It’s not so clear after watching [Denki Otaku]’s video on a bargain bag of buck converters he got from Amazon — ¥1,290 for a lot of ten, or $0.85 a piece. The thing that got [Denki]’s Spidey senses tingling is the chip around which these boards were built: the LM2596. These aren’t especially cheap chips; Mouser lists them for about $5.00 each in a reel of 500.

Initial testing showed the converters, which are rated at 3 to 42 VDC in and 1.25 to 35 VDC out, actually seem to do a decent job. At least with output voltage, which stays at the set point over a wide range of input voltages. The ripple voltage, though, is an astonishing 400 mV — almost 10% of the desired 5.0 V output. What’s more, the ripple frequency is 18 kHz, which is far below the 150 kHz oscillator that’s supposed to be in the LM2596. Other modules from the batch tested at 53 kHz ripple, so better, but still not good. There were more telltales of chip fakery, such as dodgy-looking lettering on the package, incorrect lead forming, and finger-scorching heat under the rated 3 A maximum load. Counterfeit? Almost definitely. Useless? Surprisingly, probably not. Depending on your application, these might do the job just fine, especially if you slap a bigger cap on the output to smooth that ripple and keep the draw low. And keep your fingers away, of course.

Worried that your chips are counterfeits? Here’s a field guide for fake chip spotters. And what do you do if you get something fake? A refund might just be possible.

33 thoughts on “Cheap Power Supplies With Fake Chips Might Not Be That Bad

  1. There is a lot more to DC/DC converters than what was considered in the video (which is of course fine).

    Another issue with fake chips is that you’ll often get vastly different behavior & characteristics over separate production batches. You might get a couple of these that “work well enough” for what you’re going to use them, then you’ll purchase another set and suddenly the world looks different.

  2. The important question when using dodgy parts like this is “When the pass transistor fails short and dumps input voltage straight to the output, how much will I lose?”

    If the only thing at risk is an equally counterfeit $5 Arduino and some blinky LEDs, then go ahead and gamble. If it’s going to burn your house down, spring for genuine parts.

    1. “When the pass transistor fails short and dumps input voltage straight to the output, how much will I lose?”

      In the immortal words of a famous police inspector (in what *was*, a once great city): “you gotta ask yourself one question – do I feel lucky ?”

  3. I have had such LM2596HV clone fail while (seemingly) entirely within the datasheet specs. It failed in short input-output, but fortunately a separately installed fuse and TVS saved any downstream devices from breaking.

  4. If I’m not mistaken, of all these little buck/boost boards the ones based on the LM series chips are the worse of the bunch. Folks mostly get them because they recognize the LM brand.

    The ones based on XL chips are better and often cheaper. I think the XL6019 based ones are the best of the bunch, and do buck-boost to boot.

    Feel like it’s not very fair to pick the worse of the bunch to test, people will get the impression that these little boards are worse than they actually are. Still just cheap boards from china though, don’t expect miracles.

    1. The Teledyne LeCroy t3dso1204 @04:32 is also suspiciously exact the same as a Siglent. Except for the price tag. The Teledyne costs USD1170 at digikey, while the Siglent is USD700.

  5. Started to watch the video, then the comment voice say
    “the input voltage must be higher than the output voltage”
    and then “the output voltage will be lower than the input voltage”…
    and then I felt a little tired and went to take a nap !

    Tbh I bought two of those circuit from eBay, they work as required, output on other eBay stuff…

    1. With you there…but after 40 years, I am running low on 7805s…

      Nice thing about transformers and South Africa’s collapsing power grid, I’m not losing any power supplies like the rest of the folk… :-)

      1. +1

        Last year I’ve built an old school PSU for my Raspberry Pi 2 that way.

        I’ve used a transformer from a broken radio alarm clock, four 1N4007 diodes for the bridge rectifier and an 1,5A (or 2A) version of an 7805. And a large DIY cooler made of copper (old PCB materials).

        The Pi ran rock solid on that thing (under low load, it was meant as a project).
        It ran without any noise (clean video/audio); much better than if I had used a Chinese power brick full of ripple.

        1. Also running my Pis from linear supplies (actually I’m running everything I can from linear supplies), but I’m sticking to aluminium heatsinks. Very easy to get a piece of off-cut aluminium shaped for a small case, with only a vice, file, saw, and drill.

          Luckily we still have a local manufacturer that does custom one-off mains toroids up to 500VA, with or without shielding, but I do have a healthy supply of small PCB mains transformers boxed-up.

  6. Before concluding it’s fake chip, the demonstration is wrong. The ripple frequency can come from a bad circuit itself (the C/L on the output might not be what’s required in the application note). Or the chip can be genuine but de-classed (typically, the factory making the real chip did not throw the chip that doesn’t pass the quality check, but sell them on a second market).

    There is no gain in faking such chip since the vendor never specify the DC/DC component on these modules. So if a cheaper equivalent exists that does the job, he’ll likely use it. Said differently, a non branded chip would sell just as well, so why take the risk to counterfeit a chip marking?

    1. You’d think so, but people make and sell fakes of even such (literally) cheap as chips things as the ancient and crappy LM358 operational amplifier. 40 cents on Mouser, and yet if you buy them on eBay and other places you end up with relabled floor sweepings that don’t even have the same pinout as the LM358.

      If somebody will pay for it, somebody else will make a fake and sell it.

    2. “the vendor never specify the DC/DC component on these modules.”

      Actually, they do. The module has LM2596 printed all over it, and the sellers advertisement on Amazon says “LM2596.”

      So, yeah. Fake and using known name and brand to get customers.

      1. If it were second source, then it’d have some other company logo on it. It has a (crummy) copy of the original logo. In fact, the video shows two different ones.

        Second source is a licensed copy that meets the specifications of the original. Fake is fake.

        1. Ah, indeed, didn’t watch the video long enough to see that this one has a fake logo. Many of the LM2596 chips floating around have just the part code without any logo.

          Second sourcing does not need a license if you don’t copy the silicon design. Part numbers are not protected property. And there tend to be functional and specification differences even between reputable manufacturers.

    3. For what it’s worth, the company that makes the LM2596 does not sell the chips that don’t pass testing. Our yield on that part is very high. I don’t know the exact number but I do know that parts in that category are expected to have a 99% yield or product engineering gets involved, and the parts that fail are mostly unusable. Yes I work for them.

      A lot of cheap buck regulators have inadequate ceramic caps on the output, and ceramics derate with voltage, so if a 0.47uF is providing just barely acceptable ripple at 5V, it will definitely be inadequate at 10V and laughable at 30V output. But the board manufacturer doesn’t have to be particularly honest about it: they can say it has 4% ripple at 5V, and leave you to find out that you have a big problem at the 30V output you need. I’ve seen plenty that have the output cap voltage rated at exactly the quoted output voltage, so if you were to operate at that, the ripple would exceed the cap’s voltage rating.

    4. “Before concluding it’s fake chip, the demonstration is wrong. The ripple frequency can come from a bad circuit itself (the C/L on the output might not be what’s required in the application note).”

      Sorry, but the LM2596 (if it’s genuine TI) has an internal monolithic oscillator that sets the frequency as constant, with the pulse width largely determined by load current and input-output differential. The external L-C has little authority over the period.

    1. Same experience. I once used such a cheapo PSU for an Arduino Uno project that involved an DDS generator. The result was horrible, the SSTV image it generated was totally messed up.

      When I switched to using the original wall brick that was sold with my Huawei smartphone, everything worked just fine.

      So yes, I’m outing myself as a “permanently stuck in the past” when it comes to power supplies.

      It’s not that good switching PSUs are impossible to build, but they’re hard to make out.

      And that’s the main problem, I think.

      See, a transfomer based linear PSU, – no matter how cheaply made -, doesn’t generate RFI (noise) like a jammer.

      Sure, it has its own limitations, like providing good efficiency, a single, fixed AC input voltage (220v or 110v), power stability and enough current (amps).

      But for what it is, it’s very friendly and clean by design.
      Like a light source/solar cell combo (DC alternative to transformer PSU).

      Sure, proper switching PSUs (lab grade) can be made, but that involves quality parts.

      A set of multiple filter banks and caps is to a transistors PSU to what a big, expensive transformer is to a linear PSU.

      And that’s the paradox here.
      A switching PSU must painstakingly remove what a linear PSU doesn’t even generate in first place.

      But that’s just me. Maybe the switching PSU gets a successor, eventually. Let’s hope for it. 🙂🤞

  7. i don’t really know anything about power supply design but i looked at the output from my low-cost bench supply once and i was astonished by the huge transients. i guess “ripple”. it is way beyond what i could filter out with the ubiquitous 100uF electrolytic capacitors, unless the load is very low.

    kind of amazing how useful such an awful device is, but knowing it is awful sure is a bummer

  8. I had bought some of these through a stateside (Flori-dah) vendor. Not an authorized one, but they should have known that those were not LM2596. Tested then, saw the 52KHz switch rate , and then knew that hey were bogus. The other sign was high temp at only 2A load. Along with a buddy, we said “screw it” and pulled that right out of our Open-Source Project. Not worth risking the rest of it on a cheap/fake regulator.
    The part itself was labeled “LM2596”. Yeah, right, believe everything you read….

  9. I’m surprised people still use the old simple switchers – there are much better modern parts that are more efficient, and more forgiving on support components. With that said, I don’t think anyone (hobbyists/hackaday readers, and definitely not pros) buying buck boost modules from ebay or amazon (or aliexpress) are putting these in mission-critical apps, or expecting them to work right at the margins of spec – they’re just cheap ways to step down voltage in situations where a linear reg would just get a bit too hot.

    I have a few of these and a red one that has 3 pots (CC/CV). One of those switches at 52kHz (fake), the others at about 148kHz, so I guess they’re ok. After replacing the inductor with a higher value one, and changing the output cap, it’ll just about handle 2A – I wouldn’t run an original LM2596 at 3A anyway. So fake, yes, useless? not really, but unless you want to mod the boards yourself, might as well just get something from MPS, TI etc..

  10. Even a guaranteed TI part would get crazy hot with poor heatsinking and those boards are woefully underheatsinked.

    Some listings (used to?) at least acknowledge that and state that you must add an extra heatsink if you want to use it at high power or over some specified current.

    The ripple problem could be any combination of capacitor, inductor or diode being cheap or badly specced, even the PCB layout might have an effect.

    Are they still useful boards?
    Yeah, with a few caveats, as long as you don’t want to push them to the limits of the specs.

    Is it worth fixing the problems?
    Probably not, if you need to hit max ratings it’s probably better to design your own (I’d probably also question your choice of LM2596 and suggest a higher rated part purely because headroom is good to have)

    What I *would* like to see is someone follow the datasheet to the letter* , build up a board using one of these chips and then do an evaluation to see how it performs.

    * Never forget, you really have to know how to read and understand a datasheet, semiconductor manufacturers almost always have a *ton* of weasel words in their datasheets, E.G. TO-220 MOSFETs with claims of current ratings that would melt the tab, never mind the Drain/Source leads, in a millisecond and can’t possibly be met in the real world unless it’s cryogenically cooled (and even then it’s dodgy).

  11. So this kid tests a Chinese $0.85 cent known-fake (to most people) LM2596 based buck-converter breakout board with a lovely pair of $1.6K USD each (new circa 2011) Keithley Instruments DMM6500 6.5 digit multimeters [1] and a similarly expensive Keithley bench power supply. The Digital Storage Oscilloscope (DSO) he uses is the $1,170.00 200 MHz/1 Gs/s (max) 4 channel DSO from Teledyne Technologies (c. 2012 now called Teledyne LeCroy) [2] T3DSO1204 [3]. Not exactly “cheap” either. All the test equipment new totals around $6,000.00 USD!

    The LM2596, like all original LM series chips was manufactured by U.S. based National Semiconductor who in-turn was bought-out by Texas Instruments on 23-September-2011.[4] You can see National Semiconductor’s official logo at [5].

    I love how at around 7 min 55 sec (475 sec) into the video [6] he tries to handle the fake LM2596 board and gets BURNED ;-) I was waiting for that!

    Remember: the LM2596 buck switcher on these cheap Chinese breakout boards is FAKE! Be very careful referring to the real LM2596 datsheet when working with the fake LM2596 parts. But, if you know the basics of how simple switching power supplies work, you will probably be OK.

    * References:

    1. Keithley DMM6500 Digital Multimeter, 6.5 Digit, 10 A, 1 MS/s $1,595.00 each

    2. Teledyne LeCroy

    3. Teledyne LeCroy T3DSO1204 DSO, $1,170.00 each new today (likely Made in China, looks like Sigalent?)

    4. National Semiconductor

    5. National Semiconductor Logo

    6. Ouch – HOT!

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.