Fail Of The Week: How A Cheap USB AC Adapter Might Indirectly Burn Your House Down

This Fail of the Week will remind our readers that every project they make, no matter how small they might be, may have big consequences if something goes wrong. Shown in the picture above is an oven that [Kevin] tweaked to perform reflow soldering. The story is he had just moved into a new place a few weeks ago and needed to make a new batch of boards. As he had cycled this oven many times, he was confident enough to leave the room to answer a few emails. A few minutes later, he had the unfortunate experience of smelling something burning as well as discovering white smoke invading his place.

The oven control board [Kevin] designed is powered by USB and uses simple ON/OFF temperature control. The temperature inside the oven is measured using a MAX6675 K-type thermocouple amplifier. After having thrown the oven outside to contain the risk and let it cool down, [Kevin] started to investigate what may have been the cause of the incident. He discovered that his temperature reading was not correct, and that swapping the USB AC adapter with another resolved the problem. The chip reading the thermocouple needs a solid 5V reference for accurate readings. The mains adapter he was using is non-branded and he guesses that the 5V rail is super dirty. We’d love to see an oscilloscope screenshot… perhaps encouragement in the comments section will convince him to publish a follow-up?

He plans to implement a simple over-temperature detector on a completely separate circuit using a mechanical relay to cut mains if things heat up too much.


2013-09-05-Hackaday-Fail-tips-tileFail of the Week is a Hackaday column which runs every Wednesday. Help keep the fun rolling by writing about your past failures and sending us a link to the story — or sending in links to fail write ups you find in your Internet travels.

41 thoughts on “Fail Of The Week: How A Cheap USB AC Adapter Might Indirectly Burn Your House Down

  1. The fail isn’t the cheap adapter, its actually assuming it’s clean, and failing to test it for critical applications like this. At the very least, he should have thrown some filtering caps on the input. And since it’s a bulky non-switching supply, it’s probably not even 5v at low current draws. Still the user’s failure, not the adapters.

    1. Not to mention that a mechanical normally closed temperature switch in line with the power would have eliminated this possibility. They can be bought for a dollar or two online or pulled out of a microwave oven.

    2. +1, I wouldn’t assume _any_ power supply outside of expensive lab/bench ones was perfectly smooth & stable, and we’ve had one of those go faulty and blow shit up before now.

      +1 also for John’s hard cutout, a dumb thermoswitch may be unglamorous for a project like this but it’s a backup against something going too far wrong. You could wire it to kill the power until the user resets it, or even to throw itself across the supply rails and take the fuse out if stuff is really safety critical.

    1. It’s easy enough to do and it’s often the bits that you didn’t do yourself that bite you with some assumption. There are plenty of thermal fuses or bimetallic strip type thermal cut-outs that can be used for this purpose. If you have them wired into the incoming power feed, they will also protect you from things like a broken SSR, faulty code, faulty microcontroller etc.

      If you’re not regulating your voltage rail yourself, it might be worth using a separate precision voltage reference at roughly a volt lower than your rail. I use a 4.096V reference voltage chip when running 5V stuff which has the added bonus of making each count of a 12-bit ADC correspond exactly to 1mV on the input.

    2. One alternate suggestion: use a timer to check and make sure you haven’t left the oven on for an amount of time that’s absurd. If you expect to hit 225 C in 3 minutes (maybe a bad example – I use the skillet method instead), something is very wrong if it goes much beyond that point. Usually there’s an unused timer that can do this and it doesn’t cost anything extra other than the time to code it.

  2. While the temperature control on the toaster oven might not be that
    great, setting it to tiny bit above 260C (500F) would be a secondary
    safe guard to limit the upper temperature.

    I don’t think thermal couples wires are supposed to be soldered directly
    onto a PCB.

  3. I think it’s a good lesson for all of us. Things can go wrong, and when we are dealing with something with a safety issue, we need to make sure there are sufficient failsafes in place. One thing that could be done was to write a timeout in the software to make sure that the temperature is reached by a certain time (sorta like a watchdog timer), and then shut down the whole system if that timeout fails.

  4. Looking at the label on the suspect PSU it seems to imply that the unit is unregulated and will supply 5V +/- 5% only when loaded with between 270mA and 330mA. If greatly less than 300mA is drawn then the voltage may rise to an unacceptable level. Placing a 5.6V 3W Zener diode across the supply would limit the voltage to a safe level.

  5. Anybody who thinks this is a problem only hackers have, you’re incorrect. In my professional life I’ve had to build temperature stabilization systems for sensors using TECs. The nice thing about them is that they’re reversible, so you can maintain temp even if the outside temperature drops below what you expected. Great, except for the fact that if your logic breaks while in heating mode you’ll cook your sensors in about two minutes. Despite this being a quite obvious failure mode, I still have to fight tooth and nail to get hardware thermal cutouts installed. Everybody’s convinced that it will never happen to their system.

  6. I’ve don’t recall seeing a power supply that lists its rating in this format:

    “5.00 +/-5%, 300mA +/-30mA”

    The +/- on the current rating immediately jumps out at me as peculiar. So I would say, unless I knew otherwise, that really reads as “it will maintain 4.75-5.25V when, and only when, 270-330mA is drawn from it”. In short, it’s probably not a regulated power supply.

    [Kevin] seems to acknowledge this possibility:

    “Can you guess which is the bad one? It’s the one on the right… the one that is not a switch mode power supply, the one that has no agency markings!”

    Well if it’s not a regulated power supply, then you really can’t call it “bad” if you’re simply misusing it! Also odd:

    “I can’t imagine how crappy that voltage must look like on a scope. I imagine there’s a lot of ripple and high frequency content.”

    First he says it’s not a switch mode supply, but then suspects high frequency content typical of a poor switch mode supply. Forget the scope, how about just checking it with a voltmeter? Given the label I would have done that before ever connecting it to my circuit.

      1. Yeah, but most of the “ripple” will be 50 (or 60) Hz. The thing is, those unregulated supplies simply have a transformer, diode bridge and a capacitor. If you’re building a 1000W version, the transformer will be bulky and have a low output impedance: The output voltage does not change significantly with current draw. Scale the thing down a few notches and the impedance goes up: the output voltage changes a lot with the current draw.

        In short: these things will happily provide 7V at 20mA.

  7. Hi, i think I know and understand all these hardware and software interlockes but can i suggest that with home made setups like this a bit of calibration is also done.
    Very simple, just put prob in kettle if it’s about 100c your good. Now onto a piece ice about 0c your good. Hold it between your finger, or some where similar, your got about 34c, same for room temperature.

    If you want to be fancy, buy a proper temperature probe and measure the ice and kettle at the same time. Then compare your to your homemade one.

    Simple but can identify problems very quick.
    Including rate of change and non linearity, direction of change. Leave the bought probe and your one in the kettle or ice for 5 min and you have now got an idea of stability.

    Just a suggestion!

      1. Totally agree.
        So to clarify for a quick check, ice from the freezer, that’s beginning to melt, and a boiling kettle. SHOULD be at about 0 and 100 c.

        Please note “should” and “about”.

          1. This is why I said ice+liquid water, and wait for an stable reading. If all the ice melts, add more ice. If the liquid water freezes, add more water. While you have both phases, solid and liquid, you are at the freezing point, about 0C depending on atmospheric presure, impurities in the water, etc… The same for boiling water. You have liquid and gas phases, with the same caveats, but enough for this calibration.

          2. I’ve seem far too many people put the probe into the melting water, but press the probe against the ice.

            Most people don’t realise that ice can easily be at -20C, they get ice = 0C stuck in their heads.

  8. I used two diodes to allow a simple 12V-7805->5V power rail OR a USB->5V power rail to be used in a recent build, and wondered why I was getting clicking on the piezo attached to the micro when powered from USB. It turned out the diode was not giving a ~4.5V (i.e. 5V minus the diode forward voltage drop) rail as expected, but a 5.5V rail – it was rectifying the lumpy, noisy waveform coming out of the dodgy chinese 5V USB wall wart being used and spikes were still making it through, past the filtering caps, to the piezo, making it click and pop. Lesson learnt.

  9. Dumb question, since this is modification of a toaster oven, what happened to the normally built in overheat protective device? Was it too low a value for the temperature you’re trying to achieve? I used to service lighting for special effects and no matter how many times I told customers that they had to adhere to the duty cycles (10 mins on, 10 mins off or any variant of equal time) they’d run those cheap helicopters and high power systems constantly. One time it caused a fan to melt down and clear a club, and this was just days after I did the repair and placed that warning on the repair documentation.

    Lesson learned that day, people won’t listen, install safety devices. I did it on every high temperature lamp system thereafter.

    Never fret. Your place is still intact, your ego may be bruised, but you’re learning from your mistakes. Just remember SAFETY FIRST!

    1. Couldn’t the advice “10 mins on, 10 mins off or any variant of equal time” insure someone could actually exceed any stated duty cycle? I never have seen duty cycle specified in anything other than percent. For example a 20% duty cycle would mean two minutes on, 8 minutes off in a 10 minute period. I had always assumed that when an electric oven was hacked to be used as reflow oven the over temperature device was replace with on that would allow a temperature high enough to allow reflow, and low enough to prevent a fire. No doubt many hackers simply bypass it, thinking ‘Ill be there watching the process so it should be OK, until they forget. Given any replacement device would have a best guess high limit value it’s not a set it and forget it item, and would give a false sense of security.

      1. I’ve seen duty stated as time on a lot of things myself. To name a few: A weller soldering “gun”, a lincoln arc welder, a spot welder, a heat gun or two, some intermittent duty electric motors. I’ve actually seen more items with intermittent on/off duty listed as time than percentage. Of course RF duty cycle and signals is another story.

        1. Going a step further, I’ve seen a lot of products that document a percent and don’t specify a time limit. I can never be sure if I’m going beyond the limits because the actual limit is undocumented. The same thing can happen when time-in-use is stated, but resting time is not. Shopping for a welder a few years ago was very challenging – I saw both errors listed on a lot of Lincoln and Miller units.

  10. Redundancy. Every coffee maker I ever worked on that stopped heating did so because a thermal fuse in the heaters primary power loop blew, preventing a fire. Another common technique is dual TC’s and if they are out of agreement the controller faults out.

  11. It’s taken me a while to see this but better late than never. As in preventing any Hackerfolk/others from becoming “The Late”

    Smoke Operated Relay= “Less Chance” of Being a Darwin-By-Fire.

    Simple in concept and sadly not yet an off the shelf item the idea is to connect a Smoke Detector to your Power Relay. Smoke Detection triggers *POWER OFF*

    Then again- Life’s a pass/fail IQ test.

Leave a Reply to rjCancel 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.