A couple of weeks back mid evening I noticed that our fridge was a bit warmer than usual. A few tests showed that the thermostat was off and the control knob inside the fridge had no effect. I wired a lead to the compressor and ran it for a while to get the temperature down  and decided that I couldn’t leave it on or off over night and the quickest solution was a timer. I had a circuit board from a few years back that was designed to be a basic heater thermostat so it was simple enough to program it to cycle on/off. This ran for a couple of days while part time I experimented with an improved program.

The experimenting has been very interesting. It’s a vertical standing fridge-only unit, quite basic with a 70 watt compressor and mechanical thermostat with a capillary sensor behind the lining at the back. The evaporator (cooling) plate/coil is behind the rear plastic wall. This means that when cooling, ice forms on the back wall. There is no heater or fan inside the fridge so to defrost it must be warm enough for long enough for the ice to melt and run off through the drain. The original thermostat effectively monitors the temperature of the evaporator and not the air inside the fridge. This provides the time delays to prevent rapid cycling when the door is opened but does not maintain the temperature very accurately. The evaporator goes below 0-degC when the compressor is running and must go above 5 deg-C for some time to defrost the back wall of th fridge.

Using a digital thermometer with the sensor inside the fridge showed that when the door was opened even for just a few seconds, all the cold air fell out onto the kitchen floor and the ambient temperature inside rose by 3 deg-C or more. This dropped slightly after a time but remained higher than before the door was opened.

My first thermostat used a thermistor sensor and the program turned on at about 4 deg-C and off at 2 deg-C. It had timers for minimum and maximum run times and a minimum off time between compressor runs. A later version included a delay and recheck of the temperature so that brief increases would not run the compressor. Apparently, it’s best to maintain perishable foods between 3 to 5 deg-C. The milk label states 2 to 4 deg-C

I decided that if the sensor had some mass it would not respond as quickly to brief temperature fluctuations so I sealed it into a thick wall brass tube with a couple of layers of heat-shrink.

This worked well but did not handle defrost and as the ice layer built up the compressor worked harder to keep the temperature down. So I added a defrost timer that would raise the target temperature every 24 hours or so of compressor run time to allow defrost. In effect warming the whole fridge similar to what the original mechanical thermostat did. This didn’t seem like a very sensible way to do it as it had to be too warm for too long to clear the ice.

The next thought was that if the temperature was always between 2 and 4 deg-C it should be possible to defrost without raising the temperature. So I added a small fan on the bottom  shelf blowing air up the back wall where the ice formed. This worked perfectly. A light layer of ice develops when the compressor is running and clears when it is not. The compressor run time is reduced and it runs less frequently. The fan runs all the time and probably also keeps the temperature more even throughout the fridge.

Now I have a thermostat that works well, probably maintains the temperature more accurately and possibly uses less power due to the fan addition. Now, I could just buy a replacement for the old mechanical thermostat and we’d be back to an original fridge; but where’s the fun in that. If I keep the smart thermostat I have to find a way to fit it permanently with the least work and least modification to the. I don’t want to drill holes or do anything that will compromise the insulation etc. There are existing wires for the thermostat and light. I may be able to use these for light, thermistor and fan by putting the 12V power supply and thermostat pcb with the compressor and running 12Vinto the fridge. I can easily replace the light with an LED. With the fan and reasonably accurate temperature control, it doesn’t really need the temperature adjustment inside the fridge.

The current hardware consists of a plug-pack 12V supply to the PCB which has a PIC12F629 driving a relay switching the compressor. One LED indicates status.I would like to improve the thermostat pcb to include a 3 LED indicators, push buttons for up/down temperature adjustment and a port that can be used to output information for logging or display. We’ll see how it goes.