EEVblog  

Hi Dave, I know all about this. I reverse engineered this years ago to a pleasingly ridiculous level of detail. My motivation was to move the controls on to an app on my phone. I succeeded! My electricty bills were lower, temperature control was more comfortable, and it was just convenient. Key presses (outgoing): So the pushbuttons are tactile (and not membrane) because the OUTDOOR controller expects to see a specific resistor value for each unique pushbutton. By pushing a button, a different resistor is "presented" to the pull up in the outdoor controller. You can test this by measuring the resistance between common and "aux" (wires disconnected). You will find the ON/OFF button is a dead short, so this will appear as close to zero volts as possible at the voltage divider outdoor for THAT ADC to read. you'll find the resistance values in the indoor pad are spaced about 200~400ohms apart - membrane switches cant hit that target range. So the resistor value for the "heat/cool" button must have been the next higher value (closest to 0 ohms ...being on/off). I didn't have any DIP switches on mine but you do - so I believe all they must/could do is disable or otherwise modify the resistance valves for certain key presses so they "map" correctly to the specific outdoor unit. So there is analog multiplexing going on, but no digital multiplexing! the onboard micro wouldn't be doing much in this respect (other than maybe "sniff" the same voltage value that the outdoor unit receives (slight deviation due to wiring run). I can see why they would implement it like this - you could add multiple panels in parallel and have both able to drive the outdoor unit. Communication (incoming): This happens over the power line! there is a trigger and timing logic on board, and a parasitic (but efficient) switching power supply for the onboard logic (and LEDs) on the indoor control panel (if you draw too much current, voltage sags a bit much, and it messes with the trigger circuit). What you see on the scope is a bitstream that is processed by the timing/trigger logic... you don't need a micro to do this. On my actron unit indoor wall-mounted panel, there is no micro, just a series of daisy chained 74HC595 Shift registers fed by the trigger/timing circuit. The parallel outputs of the shift registers were tied directly to the various LEDs and 7 segment displays. Once this was apparent, It was straightforward figuring out which bit in the stream mapped to which LED/segment on the display! It appears in the model you have, the micro is doing little more than what the shift registers would have done. Again, this implementation allows multiple panels to sit on the same circuit. Armed with the above, I was able to program an ESP8266 to 1) Read incoming bitstream (sniffing the data getting clocked into the HC595's) 2) send this data to a Blynk application on my mobile 3) Receive touch commands from my Blynk app and convert them to virtual button presses (I used a 10k digital pot to masquerade as button presses). 4) I could use a second digital pot to "route" in a remote sensor (controlling temperature in the bedroom, rather than the hallway!) 5) Implement elaborate schedules (eg for comfort, raise set point by one degree at 3am, but ramp back down at 9am) The beauty of the solution was that it was completely transparent - both the indoor and the outside PCB's had no clue a man in the middle existed :)

I think repair videos are some of the best content you make - Thanks Dave! Cheers

The "run" flashing probably means that the compressor can't switch on again, because it had just been switched off, and that's how it works in refrigerators, it can't turn on right again after turning off, probably because the coolant needs to re-condense or something like that.

The switched resistive ladder was very common on early car stereos with removable front panels back in the 90's. Is an easy method to bring all the switches through a single connection to the main unit.

"Cool/Elec" would be cooling-only with (optional) electric heat, which would apply to any configuration that has heating that isn't a heat pump (i.e. doesn't use the outdoor compressor) - that includes no heat at all.

This video ticked all the boxes!!! And, I leaned a TON!!! The idea of a push button array that is tied to the ADC line and a resistor network is brilliant. So imagine you have a 3 cent micro with limited I/O and you needed to feed in 10 button inputs. Well, no worries! Use the ADC input, the 3.3v rail, buttons and resistors and there ya go. I NEVER would have thought of that! The the EE's out there. Is that a common way to implement push buttons when you have limited I/O?? Thanks Dave!!!!! Glad you got it fixed! This positive outcome video really helped my bout of anxiety I've been dealing with the last 2 weeks. So, thank you!

I spent ages looking for the ox line that Dave kept mentioning.

Love the way Dave's curiosity and enthusiasm lead to a deeper than necessary analysis when all he needed to do was replace the switch. Those HVAC controls are more complex than one might imagine. Throwing a micro into anything breeds complexity. I have a programmable thermostat that has more configuration options than Windows (almost). I'm also torn between storing parts in umpteen project boxes - some of which I will surely never get back to - and sorted parts bins. EEVblog = entertainment for geeks for sure. Thanks Dave, and Happy New Year!

"these switches fail rarely" NOPE! they fail ALL the time! I'd bet that the other switches on that panel are ALSO bad and need replacing.. they do CORRODE... OFTEN!

"It worked a decade ago".

I use to do this with tiny arduino projects. Read the keyboard off a one ADC line and have the switches tied in parallel with highly different voltage dividers. Basically it's because you can use one ADC pin of the micro to read any numbers of switches, but the downside being that multiple simultaneous switches could give a false reading. If you do the calculations right and not have that many switches it's Okaysh. In the software part I'd then compare the read ADC values within tolerances, using something like switch case. It's a lot cheaper than doing the serial-parallel and vice-versa dance but comes with the risk of a dodgy switch shitting up the whole system. Might also be the only way to do it sometimes due to the size of the PIC. By the looks of that ds1307 battery placement, this was not thoroughly thought through

I have the DP20003 (The one that goes up to 5600V) and that probe is a total beast! I love it. I had to design a 1KW pulse transformer that would output around 1KV RMS Sine wave several micro seconds. One of the prototypes overshot it to around 2.5KV - 2.7KV. The probe held up incredibly well to that mistake. I also love the accessories that it comes with - all high quality silicone once.

I built my own ACU controller in '84/85 using the MCS-51. It sat, controlling computer multiple room air conditioning units for years. When the building was demolished, and I retired, I brought the unit home. It is now controlling the ACU in my home machine room. It has now operated for over 35 years without a problem. This is almost as long as my marine watches bell controller has been running. I used hall-effect switches.

When you think about it, the air con tech would have simply replaced the whole unit and put the old one to trash... for a 10 cents defective tactile switch... doh!

My first indication was to smash it. But the GEEKS we are take it apart and engineer a better one. Love you dude!

That kind of resistor ladder control was very common in VHS vcrs for pannel buttons as for mech position control. In my experience when the manufacturer puts the more buttons in a ladder the less reliable is the machine (and the accountants of the company are happier as the costs goes down...).

I miss the teardown of the switch. Want to see its guts failing.

My junky Yihua hot air station just started doing a very similar strange thing. Gotta fix it. The fun part about these problems is they creep up on you and we don't always notice it at first, so it takes longer than normal to get around to fixing it.

Excellent real world problem with explanation and fix. Thanks for sharing.

At 11:47 I think it´s not a dropper resistor, it´s an inductor

Gamers know the phantom double click effect when the mouse switch contact worn out.

31:18 No Dave, it is not rare for these type of tactile switches to go faulty and present intermittent resistance values. If you have a dodgy board with these switches on it you can bet the switches are to blame.

How to procrastinate like a pro, Turning a 5 min switch replacement into a huge saga

I've been watching you more and more, I'm a studying EE junior and the teardown and explanations you make remind me why I love to learn from your videos, so much fun!! thank you

13:58 "Because the shaft length is going to matter" - Dave 2020

When you are a you tuber, you are always thinking: could this be a video?

The mechanical issue with the switch design is using soft key pad together with longer button shaft. If the button is not pressed at the exact center this causes the shaft to wiggle and destroy the contacting surfaces inside the switch. They probably should have made the guiding hole in the front panel plastic for the shaft to stay exactly at 90 degrees to the board

My kingdom for an old-fashioned thermostat with a bi-metallic coil and mercury bulb switches. Indestructible.

I have a whole lot of the "normal" length buttons, then I buy some really long ones like 2 cm, and just cut them down to the needed length for repairs.

I would like to see what the problem was. Could you please make a short teardown video on the switch!

Unfortunately this is a common type of fault for this sort of thing, for which the solution is "Field service call, $300, replaced panel, parts cost $300". I had a control box fail on my furnace years ago. The HVAC guy said it'd be about $600 to come out and replace it. I pulled it and found it had a bad solder joint. $0 repair cost. Worked fine for 20 years after that.

Having looked at aircon from the perspective of an HVAC tech for quite a while I started this vid just thinking "why... That's something that just gets replaced in a couple minutes whenever there's a failure." But it was quite interesting to explore it from a more re-use recycle attitude. Learned alot :)

The flashing "run" is likely compressor lock-out which protects the compressor motor from being restarted under load of the discharge line pressure. Usually about 5 minutes for pressures to equalize across the metering device. Some systems in the US use communicating thermostats, but most are still based on switching 24 vac to different lines. Mostly signaling these days instead of inline switches.

This is the sort of device that fails and we just hold on to it swearing every time we try to make it work (indefinitely.. 15 years or longer). Nice video. You now have a new subscriber.

Nice repair vid. I like this sort of vid. Replace a simple component to breath some more life in to a product.

At 5:46 the white wires are actually already labelled with text printed on them. We can see "two" on the wire going to SENS 1 and "three" on the wire going to AUX. So quite a nice cable.

I never get 0.000VAC when measuring the voltage between two different ground points in my house. I usually get a couple of volts, even if both things eventually end up at the same ground/earth bar in the breaker box (our house has old school IT-mains network (Isolated Terra)). How did EEVBlog do that at 19:19? Did he secretly press "Lo Z" on his NM786 DMM? EDIT: Dave actually went directly to the ohms range while probing the to points with unknown voltage differential. Will a good quality DMM handle that if it is 230VAC (high Amp) differential between the pins? Nevertheless, 0,000 ohms resistance seems funny. Too low.

HAPY NEW YEAR .

Awesome video Dave. While others are complaining I enjoyed the whole video and that dmm is cool as I wouldn't mind getting on for my birthday next month. Never can have too many digital multi meters

It's flashing "run" because it's still in "time out" mode from the previous run. A good thermostat will not restart the HVAC for a few minutes after the previous run. This can be configured in some of the more advanced digital one.

I like to keep parts in project boxes as well but I ALWAYS put away some parts to keep in my general parts storage so that I don't have to work through a bunch of project boxes when I need something.

We (electronics techs/engineers) put up with faults which we repair for customers all the time... This is the same thing as the old adage that mechanics' cars are the worst cars on the road! :)

Dave, replace the friggin’ switch already.... 😂 I’ve got my own fixing to do today.

A little bit of KF F2, some pushes, and switches can works again. The analog input for multiplexing buttons were also used on TV (i had to use a small mosfet to interface to that!).

Had the exact same problem with a small radio. Assumed it was caused by switch bounce.

Dave, just a thought. If you set it at a certain temp, and just leave it, you don't have to turn it on or off. :)

My CTEK charger had some similar issue, had to puch, "viggle", yell, curse, give up and change it and now it works flawless. My switch did multiple oscillations so it jumped all over the choices of charging programs.

Just a guess, but it seems like the packet is incoming to the thermostat for the micro-controller to pick up on to know what state the AC system is in so it can show if it's in run, heat/cool, etc. The device generating the packet could probably then sample that same line after it sends the packet looking for a resistance value to determine any requested state changes. That way you can get 2-way signaling over a single wire. If that theory is right, it should be easy enough to test. Pick a function that is repeatable like change the fan speed, see what resistance is put across the AUX line for that, then just put an external resistor across the aux line and see if you can trigger that change without using the thermostat buttons.

This looks like an infra-red remote style protocol, signalling with variable pauses between pulses. 16 bits in a "packet". A zero is a pulse with no delay, a one is a delay then a pulse. By freeze-framing your video I've seen the following messages (might be bit reversed / backwards): 0101-1000-0010-1001, 0101-1000-0000-1001, 0000-1000-0000-1001 "Thats aircon on", 0000-0000-0000-0001 "Thats aircon completely off".

My dehumidifier seems to have this problem. The power button doesn't work and sometimes some other random button gets it to power on. Time to take a second look. Thanks! Also, Atari 7800 ProSystems have tactile switches that fail and they look like oversized versions of that. Power always fails first since it's the most used. I've seen it blamed on factory flux from wave soldering but I seriously doubt it. I've disassembled one of the lesser-used switches, cleaned it out, and hand-soldered it in place of the Power button and had it fail too while the slightly different replacements people sell seemingly never fail.

I've had this tactile switch issue several times (and repaired) in the past with car battery chargers and computer monitors..

Ford used to do the resistance thing for cruise control buttons. Generally worked fine, but were unreliable on the first generation Taurus. Ours would sometimes accelerate when you pressed the "off" button.

Those packets look like timed differential packets... though seems 18 bit which is a bit odd... the difference between 0 and 1 is the size of the gap between pulses. Kinda similar to how single data line addressable leds work.

The resistor divider allows the entire keypad to work off one pin of the MCU, which lets them get away with a smaller pin count device. Looks like most of the pins on that STM is taken up with the LED outputs. The data signal looks like a pretty simple pulse timing protocol -- 1T for a zero, 3T for a one. It's a bidirectional bus, so the master sends the keypad some state information, after which the keypad gets a chance to reply which, if any, buttons it currently has pressed. If there is no master, the keypad just waits patiently.

I think you are lucky when your aircon fails in this mode. My experience with wall-mounted aircons usually goes like this (have had this happen at least 10 times in various offices): The indoor unit has an output for the condensed water. Because nobody wants to run a pipe along the wall from the unit down to floor level, usually a small pump is mounted which pumps the water up through a tiny hose, and then it goes above the drop ceiling and off to some sewer line. These pumps invariably fail after time, or the tubing gets clogged. Now the water starts dripping down on whatever is placed under the unit. Of course there always is storage shelving there. I always surprises me that there is no standard facility in these installations to sound an alarm and stop the aircon when excess water is in the unit and is not being pumped or drained away.

Funny enough some manufacturers go the extra mile and put a transistor to each switch to make the contacts even less susceptible to contact problems, and I guess there are even 20mA loop versions of these things

The silkscreen saying PB__ tells me that each pushbutton is on an individual pin of IO register B from the MCU. The changes in resistance on the aux line might just be some issues with the voltage induced from the Multimeter through some clamping diodes.

For buttons with incorrect shaft lengths, an idea is to 3D print a shim to make the button longer. If you have a 3D printer.

Perhaps the constant signal thing is a failsafe heartbeat. If the panel fails for some reason, no signals are sent to the aircon unit and so it shuts down. Edit: Scratch that, just watched more of the video :facepalm:

You could try to fix the button with the alcohol or some other dissolvent, like gas, or contact cleaning stuff. Just putting a drop on the shaft could be enough

How much E-Waste could be saved if more people knew how to do relatively easy repairs like this . Thanks from NY

A squirt of contact cleaner - good for another twenty years.

14:40 In such case I just take or buy very long tactile and cut/grind them to the needed length.

Largest pronunciation difference noted to date, from your deck' le to my (US of A) dee kal'. Wonder when that happened?

I fixed the key fob for my car alarm by spritzing some deoxit into the tactile switches. It's not that uncommon that they have a little trouble. Granted it was 20 years old, but still....

Your aircon control panel has turned into a HAL..."I don't think so, Dave"!

Flashing run = probably because you cycled the AC on/off/on - might be a cooldown period as compressors don't like to start shortly after stopping. The pressure built in the system makes it very hard on the motor. (Might not be the case that you are controlling a compressor in a large building, more likely you are calling for a solenoid valve to open allowing chilled water into a heat exchanger...but the protection behaviour will be built into the controller).

Dunno about down under, but in the states control circuits like that are 24V AC. (I just play a HVAC tech on TV...)

The digital signal looks like a 'bits' to me. In OFF mode will always look the same, but in ON mode it is sending several bits of data to determine heat/cool, fan speed, running/standby, etc.

This is where 3D printers come in handy, can't find the exact switch then print a new button/adapter to suit a switch you do have. That said pretty sure the last time I looked most suppliers like RS etc. normally offer a wide range of button lengths.

DeOxit worth a try?

The other people in the building will be knocking on your door because their switch is bad too. Do you have enough in stock?

I'm happy to see the DMM models continuing to evolve, but I'd rather see a ESR capable capacitance meter added to the lineup.

"The shaft length is going to matter." 😜

I love "Dodgy Brothers", never heard that one before... being Kiwi I find Aussie slang is so cool!

Love your videos, learnt a lot and had a lot of fun. Don't care about those commenters who ...

Airco on/off button works eratically. Half an hour of analysis later (using various expensive tools): replaces switch. Fixed.

oohhh, that old CRT :)

Love the comedy company reference

I'm thinking those bits coming in is just a mask for what LEDs to light up on the control panel....

@11:47 It's a buck converter not an LDO. LM2594 with the Diode D2 and the Inductor L1 (not a Resistor). It is a switching supply despite the commentary.

I wonder if those pulses actually come from the AC unit itself rather than the thermostat. It might be the AC unit reporting what mode it's in, basically to give confirmation that the mode you set actually went through. I hate that some hvac units over complicate the thermostat though. Makes it harder to replace with something else if you wanted to. Most residential ones are simple as they typically just use dry contacts to set whatever mode you want. Ex: yellow is A/C, white is heat, red is +24v etc. I ended up using a relay controller and some Dallas temp sensors for mine so I can control it via a web page. No app shenanigans just a basic web page I can access from any web browser.

PWM data. Look like 16 bits. The width of the bits is what define the 1 or 0. The protocol is most likelly very stupid, like 2 zone bits (1-4), 2 modes bits (off, fan, cool, heat), 2 bits fan speed and so on. My heatpump remote is like that, it just send all the settings every time you press a key, and also have another mode/data format for the remote temp sensing.

The weird thing is so many English words. Usually with the use of iso standard symbols they can be used almost without language at all. It almost resembles a binary clock. Insane design outside, regular insides. I've got tact switches in one of my monitors and the whole set of them randomly activate unwanted functions. That'll give me something to fix next week /month /year /decade. (when I get around to it or run out of procrastination-fu, though I've got A LOT of that!)

And replace all the switches one by one is a bit of challenge.If were through hole the desoldering of them would be a real nightmare,smd switches also is not the easiest thing to deal with.So put it all back in place and leave it as it is.

I dunno dave I thought it was more sillybuggers than dodgybrothers but then I can't speak upside down.

At 8:51 Those LEDs at LED38 and LED 37 mounted vertically and why ? Is it planned or an accident or ????

Think about it?? Why would they use a resistive ladder derived voltage to talk to the remote air con controller when the resistance and therefore voltage would change because of the length of the cable, the gauge of the wire and electrical noise. You should have realised that much earlier in the diagnosis. Great repair but in 30 years I’ve never seen a high resistance switch of that type so good spotting.

This might explain the weird behaviour of my Hi-Fi buttons

Daww, didn't take the switch apart.

My Hitachi clock radio is having a similar button issue, will probably get round to fixing at some point, but at the moment a firm press is the fix.

I'm not sure if HVAC is handed differently in Australia, it probably is, but in the US the HVAC system is tied in with the fire protection system, so there might be the possibility of triggering a trouble alarm. Won't set off the fire alarm obviously but it would log an issue with HVAC and may potentially phone home to whatever company is monitoring the FSS.

That was fun.

Just a thought... It could be remote analog sensing, but with constant voltage and measure current. I would imagine long cables in a voltage divider means high impedance driving the cables and therefore pick up more noise. So what if you measure the aux wire with a current prove instead of a voltage probe? Also, the data, could that be noise coupled from for example modulating data on the power supply lines? Let the unit be driven be the capacitor when the supply goes to zero for a data bit?

I've seen faulty switches on an old school tape deck incorrectly put the unit into record and overwrite the tape because of this issue.

You kept saying that you would use the internal sensor. Which is dumb. The internal sensor is that blobby thing that you pointed out. The biggest problem with using the internal sensor is that it is wherever the control panel is. The control is placed for convenience, not for optimal temperature sensing. What makes it even worse is that it is housed inside of the control panel so it is slightly warmer than the ambient air temperature. (It is at the bottom of the control panel to negate this effect a bit.) An external sensor can be placed anywhere in the room. In particular you can place it in or near the return where you can get a more accurate temperature reading. All of which makes a remote sensor better. The only reason to use the internal one is because it is cheaper to install. (or if the external sensor is broken and you are waiting for a replacement.)

Have you consulted the mighty GOOG to see if there is a schematic online somewhere. I’d be surprised if it isn’t. That would save the reverse disintegration. I bet the analog “protocol” dates back to a controller with discrete xistors and relays.

The big resistor at the top is a hearer to create air movement inside the unit to get air past the thermistor ( A fan with no moving parts )

Dave , if you send this faulty one and one of good ones - I will be able to analyze on SEM+EDX what's the difference, where 1 cent was saved by manufacturer :-)

Air con service would have charged 500 € for new unit.