Ever since I got it there has been something slightly funky with the control system on this one air conditioner. Never bothered too much about it at the time given the price of it and being a unit from the early 90s and that it was clearly an electronic issue rather than mechanical issues with the refrigeration side of things.
The issue has always been with that fancy pants soft touch control panel - which obviously has an unnecessarily complicated layer of electronics behind it for a device of this type. This sort of control system seemed to be the late 80s/early 90s version of the obsession with everything having to have a touch screen these days.
As Montgomery Scott said, "The more they overcomplicate the plumbing, the easier it is to stop up the drain."
If switched on without pressing anything it should just run in fan mode - cooling or heating modes are enabled by pressing the buttons under the striped grey areas first or second from the left. Next one cycles through fan speeds, then turns on/off an ioniser, then the timer.
What it had taken to doing however was after a completely random period of time between zero seconds and about six hours is that it would drop out of cooling mode and then refuse to acknowledge any keypad presses aside from the power button or timer button (or the ioniser - but that's a completely independent subsystem which is live whenever the AC supply is on). Sometimes power cycling the unit would kick it back on into cooling mode, but not always. Day before yesterday it decided to play up once too often so I started investigating. Back right when this first started I checked out the keypad itself and ruled it out as a possible culprit. A sticking button could give similar symptoms but would also knock out the power button functionality. Plus they all checked out individually anyway. The issue was clearly with the brains of the machine.
This is the control PCB.
For something that needs to turn on/off a compressor, a heating element and switch between three fan speeds this really is unnecessarily complicated. The timer functionality is a totally separate physical unit so that's not even on this board. Likewise there's no need to run the condensate pump off the board...the switch contacts on the float are more than rated for the full load of the whole unit, so just let it handle itself and use the overflow one to cut power to everything aside from the warning light. Initial checks didn't reveal anything visibly showing signs of distress or any bulging caps. The wire you can see hanging off down by the transformer was the receiver for the remote control - I pulled that off first in the off chance something was interfering with it. Second step was to obviously pull and reseat that large socketed chip.
Here's a close up of the markings for those of you playing along at home.
Given this chip is basically running the whole show it was an obvious first port of call - sadly didn't make any difference. I also went over the board with a good magnifying glass to look for any dry joints or possibly cracked traces...Zip turned up. In fact the quality of the soldering and the traces on the board is really very good and I can't fault it at all.
Helpfully there is a simplified system schematic on the inside of the rear cover. As none of the connectors on the PCB are labelled this is appreciated.
Given the unit is thirty years old, my obvious next port of call was the DC smoothing capacitors. Looking at the voltage rails (5V and 24V) they were both slightly on the low side.
Smoothing capacitors are 1000uF 35V...and of course I didn't have any of those in stock. I did have two 470uF ones in a suitable voltage rating though so in the spirit of experimentation I tacked those on in parallel with the existing ones to see if a bit of extra capacitance would make any difference...if they'd just gone low value it probably would have gone a long way to help. Not tidy by any stretch of the imagination, but for testing purposes it'll do just fine.
It fooled me into thinking I'd fixed it as after doing this it behaved itself for about six hours...before doing exactly the same thing again.
Fair enough. My intention is basically to do away with this overcomplicated "brain" and stick a proper temperature controller in. I've got one in a box somewhere which can handle heating/cooling (this thing could only be set to one or the other - and just has a 0-9 numbered thermostat) and has proper adjustments for hysteresis, anti-cycle timers etc...I just need to figure out where the heck it is. The original PCB will stick around to handle the evaporator fan speed control, but that will basically be it. Control over the compressor, condenser fan and heating elements will be handed over to the new controller.
I wanted it back up and running today though...and I'm not one to be beaten by something like this.
The thermostat on this thing is just a mechanical make/break switch, and that quickly got me to thinking that I could very easily bypass the brain. I'd lose the anti-cycle timer, but to be honest there's a decent deadband on the thermostat I don't see that being an issue.
A little bit of poking around with a meter found that there's an always on 24V feed to all of the relay coils, and it's the earthy end that's switched by the control logic. Sure enough, if I earthed the low side of the compressor relay coil it and the condenser fan motor both kicked in, even though the unit wasn't in cooling mode.
Didn't take long for a plan to form. Even though this unit doesn't (and as far as I can see never has had) an ioniser unit fitted, the button and relay are there for it. This function being "always on" seemed to escape from the lockup issue...with the relay clicking away merrily when the button was toggled. This meant that I had an easily accessible switchable ground available to me.
That's us halfway there. I then pulled the thermostat connector off the board and spliced a couple of wires into that - wiring it in series with a wire to the switched ground I mentioned above.
Result is that the ioniser button is now a "compressor enable" button, and when that's turned on the thermostat makes it cut in/out as originally intended.
Only other thing I had to do was to physically unplug the heating element power supply as if it jumps into heating mode it will turn the element on (as in heating mode it interprets the stat being open as a cool for heat - whereas it's the other way around for cooling) because the thermostat is no longer hooked up to the PCB. Dead easy to do as it's got a socket up front.
I've not gone any further by way of making it pretty or securing the wires to the PCB with hot glue or anything as this only needs to last until I track down the proper controller, then quite a bit of re-engineering will happen. Hopefully see it going for a good while yet, and actually with better functionality as I'll be able to dial in an actual desired temperature rather than just picking a number. Plus overnight I'd really prefer the room to be a degree or two cooler than the inbuilt stat allows for...new control will resolve that. Worth noting that the wiring on there is just 24V, nothing on the mains side has been touched, and I've been careful to keep the tape from bridging the isolation gap.
Still not sure what the original problem is, and there's very little on these units on the web that I've been able to find. As best I can tell though there's something amiss with that big IC...Without another unit on hand I can rob one out of to test it's pretty hard to tell really.
It's a total bodge, but it's doing what I bought it to again, and soon will be upgraded rather than bodged.
Also in fairness to the unit...It's 30 years old and has spent a good portion of its life on an equipment rental company's books...so it's not done bad at all really! May 1991 is the date on the QC sticker on the inside of the back panel.
