Faulty ESP8266s Release Smoke, Then Keep Working?

[Ray] is in a bit of a pickle. All appeared well when he began selling an ESP8266-based product, but shortly thereafter some of them got hot and let the smoke out. Not to worry, he recommends ignoring the problem since once the faulty components have vaporized the device will be fine.

The symptom lies in the onboard red power indicator LED smoking. (Probably) nothing is wrong with the LED, because upon testing the batch he discovered its current limiting resistor is sometimes a little bit low to spec. Off by a hair of, oh, call it an even 1000x.

HAD - HotESPY3Yep, the 4700 ohm resistor is sometimes replaced with a 4.7 ohm. Right across the power rail. That poor little LED is trying to dissipate half a watt on a pinhead. Like a sparrow trying to slow a sledgehammer, it does not end well. Try not to be too critical, pick ‘n place machines have rough days now and then too and everyone knows those reels look practically the same!

The good news is that the LED and resistor begin a thermal race and whoever wins escapes in the breeze. Soon as the connection cuts the heat issue disappears and power draw drops back to normal. Everything is fine unless you needed that indicator light. Behold – there are not many repairs you can make with zero tools, zero effort, and only a few seconds of your time.

[Ray] also recommends measuring and desoldering the resistor or LED if you are one of the unlucky few, or, if worst comes to worst, he has of course offered to replace the product too. He did his best to buy from authentic vendors and apologizes to the few customers affected. As far as he knows no one else has had this problem yet so he wanted to share it with the community here on Hackaday as soon as possible. Keep an eye out.

If you have never seen smoke ISO9001-certified electronics repair before, there is a short video of this particular disaster upgrade caught live on tape after the break.

83 thoughts on “Faulty ESP8266s Release Smoke, Then Keep Working?

      1. That is true although for the price, i don’t think they would do Q.C. Its a rush to make and sell them.There are so many variations popping up now.
        I can have 5 of them for 15 dollars.

        At least this was shared.Other users can at least test if this is the case with theirs.Maybe sticky post to forums?

    1. The manufacturer probably has tested these modules, but because the smoking only happens after like 10 seconds, if the testing procedure is less than 10 seconds they won’t notice it easily.

      1. For a module like this the testing should include wifi performance (which probably would take longer than 10 seconds) as well as power consumption. Not to mention the calibration step, which probably isn’t happening either.

        1. Well, current/power consumption might not be part of their testing procedures; or they might used a much lower voltage (like 2.2V or below) instead of 3.3V for testing, in which case there would be no symptom. Not that I know what happened, just thinking about all the possible reasons it fell through the crack.

          1. Lol, test with 2.2v? Why the hell would you do that and how can you test without testing? Doesn’t testing mean running it with the required voltage?

            But anyway, I think testing of such products is probably done once per few hundred and then only briefly.

          2. @Whatnot: what do you mean by ‘required voltage’? The module’s operating voltage is 1.7V to 3.6V. There is no particular voltage that’s the required voltage, although 3.3V is pretty standard.

        2. “The module’s operating voltage is 1.7V to 3.6V”

          Indeed so a reasonable test would be to run the module at 1.7V and at 3.6V. I wouldn’t expect every single module to be tested for absolute maximum and absolute minimum but certainly testing should be done at both extremes of operating limits if you’re going to do it at all.

    1. Per-unit calibration of PCB antennas has to be done after it is placed in an enclosure, as even plastic enclosures change the behavior of the antenna a lot. Though despite this, ESP8266 has given very good wifi performance for me. Maybe it has some automatic antenna tuning, like some mobile phones.

    1. It doesn’t have to be counterfeit, but it shows the boards are without any functional test which would cost more. Even testing a random sample from a batch would have detected this.

      I have seen parts mislabeled from Digi-key. It just a matter if the person loads in a reel reads 4R7 as 4K7 after a long day.

      It can happen for a real product too. Our North American CM do flying probe test for component values whenever they can. For R & C parts with values too high or low to measure in circuit, we are told to have them on their own with test pads. All that and full suite of continuity/functional tests to catch mistakes/productions issues as our products cost too much to be thrown away.

        1. Yes I would, but not at the price. China does not mean cheap. China means manufacturing to a cost. If the cost is low you get what you pay for. If you want some high quality telescope optics, or a PCB with 12 layers and 2mil trace tolerance from China then you still get what you pay for.

  1. Reminds me of a problem on the shop floor when we discovered that the assemblers were using 4k7 and 270k resistors interchangeably (yellow-violet-red versus red-violet-yellow). We solved the problem by altering the value of one (I think to 220k or red-red-yellow).

    1. Let me put on my parachute pants, y’all can’t touch this, it’s GRAMMAR TIME!

      According to Oxford “Worse to Worst” is a popular American bastardizing of the phrase.

      Both idioms exist, they mean slightly different things.

      Spoiler – I’m secretly an ancient dragon that lives in Canada.

      http://dictionary.cambridge.org/dictionary/british/if-the-worst-comes-to-the-worst#if-the-worst-comes-to-the-worst__1 <– Apparently "Worst to worst" is the traditional British idiom. "Worse to worst" is American. Maybe we can blame the metric system.

      http://grammarist.com/usage/worse-comes-to-worst/ <– For centuries the idiom was "Worst to Worst", though if it makes you feel any better, since the turn of the millennium the nonsensical "Worse to Worse" is also taking over.


      Well, New York Times wrote a whole article about it.

      http://www.nytimes.com/2011/02/06/magazine/06FOB-onlanguage-t.html?_r=0 <– "Idioms are like barnacles on the ship of language."


      "The idea encapsulated by the original idiom is that of a worst-case scenario brought to life. “Come to” here means “result in” (as in come to nothing), so the phrase describes the worst thing in theory turning into the worst thing in actuality."


      "“If the worse came to the worst, I could but die.” This subtle change of one letter changes the interpretation of the phrase. Now it marks a downward progression, not just from bad to worse, but from the comparative worse to the superlative worst. The close phonetic resemblance of worse to worst helped the reinterpretation take root, especially because the final “t” of worst can get swallowed up before the “k” sound at the beginning of comes in everyday speech."

      So, looks like "Worse" is a stowaway trying to get a free train ride to the big city.

      Doesn't matter, he succeeded and now we're stuck with it:

      "Both the worst/worst and worse/worst variants are perfectly acceptable in standard English. If worse/worst seems more logical to you, then by all means use that opportunity to make idiomatic language a little more reasonable."

    2. I’d like to make a quick pitch for “When worst comes to worse” as in “when the baddest thing inconceivably becomes even more bad”. I feel like it’s underrepresented everywhere, and yet it’s my favorite variant.

    1. Yup, that’s right. It’s about 0.2 mA. So the LED is pretty dim. Given this is just to indicate power, it’s a good idea to set it to draw as little current as possible. To be honest I don’t think this LED is very useful anyways, especially if you are doing low-power projects.

    2. Not sure you would see much light 3v3 supply and 1.8V LED forward voltage and 200uA (4K7 resistor) – seems pretty unlikely, however I would need to experimet to find out.

        1. Agreed. 4.7k was a good call on their part. It’s not obnoxious but lets you know if there’s power. Plus the blue LED lets you know it’s working and doesn’t keep you up at night.

    3. To put this in perspective, have you ever run a common T1-3/4 through-hole LED at a mA or two? The light is easily visible even at that low power, right? Now think how much smaller that SMD LED is. You’re lighting something up nearly as small as a grain of salt, and it takes correspondingly less power to achieve the same brightness over the smaller area. 200uA makes a lot more sense when you think about it that way.

    1. Oh and on a side-note: I hear the first USB3.1 motherboards are reaching the channels, so the chips are ready and soon we’ll have USB3.1 devices, twice as fast as USB3.0 and with a plug which you can plug in either side, up or down.

      And more power too than 3.0. So get ready tinkerers

    2. Wait a minute, how much current the USB port is capable at outputting is not the same as how much the circuit actually draws. Just because USB 3.0 can output twice as much current does not mean the circuit will break faster.

      1. But the thing is broken, the resistor burns out because it’s the wrong one, and surely pumping twice as much amperage in it would speed it up.. Basically the only real resistance is the limiting of the powersupply.in this faulty configuration.

        1. V=IR? With USB the voltage is still regulated: the amount of current doesn’t magically go up without something happening to the resistor first.

          Resistors don’t change that much as they burn up.

          1. Perhaps I am misreading, but I am not sure what you are saying here? The USB standard mandates that the host limit current, and this is actually the limiting factor in the case of this mis-configured resistor.

            USB devices initially may draw one unit load, which is 100 mA in the USB 2.0 spec and 150 mA in the USB 3.0 spec. Since this device likely does not request any more power, the LED’s resistor is dissipating a nominal ~0.5 W for USB 2.0 and 0.75 W if USB 3.0 were used.

          2. This presumes the USB source is is intelligent which it is almost certainly not.

            The whole of USB spec is an electronic honor system. Devices are supposed to request power and only take it when approved. But the power rail is usually just a power rail and you can draw whatever you want.

            In this case, the LED is across the VCC and the Gnd of the ESPy, who knows what the (already off-spec) chip itself decides to do with the power it gets from USB.

          3. Yes that formula applies until the circuit (LED and resistor) start to burn, at that point the circuit isn’t following the rules of what it says on the resistor anymore, that’s the whole reason there is smoke.

            To boil it all down: You should be looking at the law for calculating wattage to determine how much the circuit has to dissipate in heat and what capability it has to do so given the surface area in a case like this,

            As for the USB specs, it is simply a reality that at least half the USB connectors in the world will have no intelligence to them at all. And even if they do and USB3.0 is only 50% more amperage, it will still burn out things quicker if the parts are so far out of specs that they burn.

        2. First off read the actual USB specs and do not speculate. RPi designers made that mistake.

          USB voltage is highly unregulated and the voltage drop at each point is dependent on the internal resistance of wiring, connector, PCB, PTC/protection circuit, diode(optional?) and loads (peripherals).

          >the voltage supplied by low-powered hub ports is 4.45–5.25 V.

    1. No, it’s delivered through VCC (the power indicator LED does not go through any GPIO). This is a very rare issue — I’ve bought a couple hundreds of these and this is the first time I’ve seen the issue. Just a bad batch.

      1. that’s good, something less to worry about then. so far i’ve been really lucky with these things, i’ve got about 80 of them in use and only had a couple that didn’t perform as well, i noticed a couple had blobs of solder on the antenna.
        definitely favourite device of 2014

  2. You could assume the self test would take less than a second but it’s crazy they don’t pick this up, specially across a whole batch.
    Interestingly, if you look at the some of the videos of the QC facilities for tablet/phone manufacturers, one of the first tests is the current consumption of the device under test. They have a chart/checklist that characterises (sp?) the current and current slew across in different operating conditions against known faulty components. They don’t even have to open up the case.

  3. Oops, the forum dropped part of my post… So again…

    I have some 0805 LEDs where they are too bright to look at with 5V and 5K. Red drops 1V8, Green 2V4 and Blue 2V7. So all these are 1mA for LEDs since the mid 90’s.

  4. Oops, the forum dropped part of my post… So again… (it was the less than and greater than symbols :(

    I have some 0805 LEDs where they are too bright to look at with 5V and 5K. Red drops 1V8, Green 2V4 and Blue 2V7. So all these are less than 1mA.
    I haven’t used greater than 1mA for LEDs since the mid 90’s.

  5. I thought this might be why the ESP8266 was browning out my CP2102 USB-serial adapter when run off the 3.3v supply. (everything works fine on a separate 3.3v supply)
    But I didn’t recall seeing a red LED on the board.
    Lo and behold, 8K? resistor and NO RED LED and nicely flowed empty pads. So I suspect they figured something was up and started leaving them off … the blue LED had a 2K2 as mentioned by someone else.

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.