Around these parts, we generally celebrate clever hacks that let you do more with less. So if somebody wrote in to tell us how they used multiplexing to drive the front panel of their latest gadget with fewer pins on the microcontroller than would normally be required, we’d be all over it. But what if that same hack ended up leading to a common failure in a piece of consumer hardware?
As [Jim] recently found out, that’s precisely what seems to be ailing the Meaco Arete dehumidifier. When his stopped working, some Internet searching uncovered the cause of the failure: if a segment in the cheap LED display dies and shorts out, the multiplexing scheme used to interface with the front panel essentially reads that as a stuck button and causes the microcontroller to lock up. He passed the info along to us as a cautionary tale of how over-optimization can come with a hidden cost down the line.
Judging by the thread from the Badcaps forum, the problem was identified last summer. But unless you had this particular dehumidifier and went searching for it, it’s not the kind of thing that you’d otherwise run into. The users start by going through the normal diagnostic steps, but come up short (no pun intended).
Eventually, user [CG2] resorts to buzzing out all the connections to the two digit seven-segment LED display on the front panel, and finds a dead short on one of the segments. After removing the display, the dehumidifier sprung back to life and everything worked as expected. It wasn’t hard to identify a suitable replacement display on AliExpress, and swapping it out brought the appliance back up to full functionality.
Now to be fair, a shorted out component is likely to cause havoc wherever it might be in the circuit, and as such perhaps it’s the lowest-bidder LED display with the unusually high failure rate that’s really to blame here. But it’s also more likely you’d interpret a dark display as a symptom of the problem rather than the cause, making this a particularly tricky failure to identify.
In any event, judging by how many people seem to be having the same problem, and the fact that there’s now an iFixit guide on how to replace the shorted display, it seems like this particular product was cost-optimized just a bit too far.
It’s a bit of a shame as this is otherwise the best dehumidifier on the market.
I imagine it is a bad batch of LED display components. The Meaco Arete dehumidifier is one of the most respected brands, certainly by reviewer recommendation. Though that could be strong marketing. An interesting failure mode to add to the memory.
Well yeah I have the same one and it’s by far the best dehumidifier I’ve ever had. I doubt it’s deliberate, and more the engineers just being clever to a fault that’s not apparent till a few years later.
They certainly don’t skimp on online reputation management.
I think that the TM1638 chip might apply too much negative voltage over the LEDs when it is scanning the display and keyboard with the shared pins.
It seems with the common cathode displays, SEGx go to the anodes and GRIDx go to the cathodes. The SEGx are also used to scan the keyboard. When the display is off (GRIDx is high) and SEGx is low, negative supply voltage would be over the segment LEDs.
Looks like most 7-segment displays specify 5V as the absolute maximum reverse voltage. The TM1638 appears to be supplied 5V from a simple zener based regulator, so I wouldn’t be surprised if it is a bit over that.
Exactly. The relatively high defect rate suggests the issue isn’t a faulty component, but rather that the component is being driven outside its specifications. You’re probably right — a replacement display would likely fail again after some time.
If you really want a low pin count interface for your uC…
It’s quite doable to connect both a LED and a single button to a uC pin, and then use morse code both for input and output.
I played a bit with this idea and even spend a few hours learning morse code especially for this. I thought it would be useful for setting some configuration data, for example network adresses in an RS485 network. But in the end, using a simple serial port (with just two wires, Rxd, TxD (and GND) combined with a terminal emulator (console, screen, putty, etc) is much easier to use and has more capabilities. So my morse adventure never got very far (I think I learned 6 or 8 letters).
I’m still a bit curious about morse code though. How many hours of practice does it take (on average) to get your morse code to a usable level?
(Also, if you ever want to learn morse, do not start with a slow speed, and later increase the speed. This programs your brain wrongly, and you have to unlearn the “slow speed” later. You have to start at full speed and start committing the “beep patterns” into muscle memory without being concious of the individual dits and dah’s. )
RE: How many hours of practice does it take (on average) to get your morse code to a usable level?
Hours, not sure, but around two weeks of daily training is plenty to master entire thing. It is not tremendously complicated, and practice makes it solid. (as a side note, morse code is far from being “retired”, merely pushed back to the background of niche communications).
…It depends.
In jr. high school, I learned it Morse well enough in a couple of days to be able to type morse code using a regular keyboard with one finger on the dots and one on the dash without thinking about it, as well as read similar “secret” messages sent to me at normal reading speed. But that is visually/ “reading it” which almost no one does- most agree that learning should be done by ear only, and at a speed above the “counting dots and dashes then processing that into the letter” threshold as you’ve noted. Typically 20-25 words per minute.
Think of it like any other language, though. The comment above saying two weeks is probably right if you have two straight weeks to dedicated to it. I dunno what your schedule is like, but that is impossible for most adults. But like High School Spanish, it is use it or lose it. I still mostly use CW exclusively on ham radio, and at my rapidly advancing age, I can muttle though but it comes and goes.
There is almost a century of fascinating data on the best way to train people up on Morse Code beginning with landline railroad operators up until a couple of decades ago when the US military dropped the requirements for radio operators. This seems to indicated that most people can do it eventually, but some people … just can’t. When I got back to it later in life I was doing practice about 15 minutes every morning, and got up to speed in a couple of months. I used LCWO (learn CW online) which allows you to do Farnsworth and/or Koch methods. Oh also, sending is wayyyyy easier than receiving, an afterthought really. like 90% of your time should be spent listening and receiving. sorry this is getting long, one last thing- receiving random letter groups (like call signs) is obviously harder than plain text/ normal speech. Your copy doesn’t have to be great for regular sentences- you can fill in the blanks yourself. If you are copying numbers (like IP addresses) or a callsign, your copy has to be 100% perfect. Something to think about how amazing the landline railroad operators were back in the day.
You more-or-less just described the freqmite. Worked great.
You don’t need to be a morse virtuoso to use it.
Sounds like a feature to a manufacturer who wants to sell you a new one every so often.
Resist the temptation to attribute to malice what can be explained by incompetence.
Impeccable timing – I sent this back to the manufacturer yesterday to solve this exact issue, as it’s still under warranty.
Good to know it’s a fairly simple fix if this happens again!
I’m not sure there’s much of a lesson here about design – as the OP notes, you can’t in general design something to work without working components. Like, if an interface includes a display, it’s probably because you need a display to operate the device, even if the buttons are still working. Other than including a second, redundant UI, what could the designers have done better? Would using more pins on a bigger MCU have meant it was fine if the display failed?
It sounds like the moral is more about using excessively cheap parts. Who expects a 7-segment LED to be the weak link in a piece of turbomachinery? The whole issue probably only came up because the manufacturer shaved one cent too many off the BoM.
This, but also the firmware could have been designed with more failure detection. If it detects (what it thinks is) a stuck button, it could flash an “error” indicator, and perhaps operate in some default, non-UI, failsafe mode, so the unit could still be used, at least somewhat, until a repair or replacement can be effected.
“Would using more pins on a bigger MCU have meant it was fine if the display failed?”
Well… yes! In this case, because of how they multiplexed the display and a button together, a single dead segment on the display caused the button to stop working (and the firmware to lock up, but that could have been fixed with better firmware).
If the button was on its own pin, the device would have operated perfectly fine, just with a dead display segment. You could still use it no problem, it’ll be obvious to the user what the display should say. But a button not working would likely cripple it.
Repair thread also here: https://www.badcaps.net/forum/troubleshooting-hardware-devices-and-electronics-theory/troubleshooting-appliances-and-other-household-electronic-stuff/3331121-meaco-arete-one-deye-dehumidifier-not-working
There are problems with these LED display modules, the internal connections fail either short or open. Just low quality workmanship.
“such perhaps it’s the lowest-bidder LED display” … Had to smile, when fix was to go to AliExpress for a replacement…
Anyway, you know what needs replaced the next time …
Had an old Realistic receiver/amp with a PLL and an indicator light for when said PLL was locked in. When that indicator’s mini incandescent bulb burned out, the PLL stopped working too. Wire in a new bulb, PLL back in business.
Realistic gave you a circuit diagram in the owner’s manual, gotta give them credit for that. All I can recall is that the bulb was directly connected to the PLL.
Does a dehumidifier even need a microcontroller? I’ve never had a dehumidifier, but I remember my uncle and aunt having one long before computers and microcontrollers were household items. Our son and daughter-in-law bought a new stove a couple of months ago though which had a computer and digital display, which a stove does not need. When it was only something like a week old, the computer began crashing, and they had to unplug it for a few seconds and plug it back in to reboot the computer. At first is was just a couple of times a day, then got to where the whole thing was useless. I’ve brought a dozen products to market with microcontrollers myself; but I am against misuse of technology, which I believe putting microcontrollers in appliances is.
I don’t get it either. Where is the “Keep It Simple Stupid” mentality for good solid, long lasting equipment/appliances???
Good point. Why do jet airliners need microcontrollers when it’s been demonstrated that all that’s necessary is a stick and a piece of string?
I’ll get off your lawn.
They don’t need them, and pilots have become too reliant on them.
The datasheet for the display driver IC says that it’s built to multiplex the display and buttons. It’s an absolute freak accident that these LEDs fail shorted. No engineer would have thought up that error condition or probably not see it in pre-production. In this case I wouldn’t blame anyone, it’s just a bad batch of displays I believe.