I searched and found a few related posts but none with exactly the problem I was facing... external air temperature had been reading a bit higher every day. I know it’s Spring here but this was ridiculous - 34 one day, 43 the next, 51, and then finally over 60 degrees (C) as you see in photo 1. The problem with this is that the ventilation system becomes reluctant to admit outside air into the passenger compartment. Plus, on cold mornings (when it’s actually 8 degrees C) the system gets confused by the erroneous 26 degree reading and gives a blast of cold air, when warm would be nice.
I first tried cleaning the doorframe plug, reasoning that the temp sensor signals from the mirror would pass through there. I pulled off the tweeter cover and unplugged the mirror - reading was ‘- - -‘. I removed the mirror glass by tilting it in and down, then used a long thin screwdriver to release two of the clips. Then I discovered that the temp sensor has been living underwater (photo 2). It’s also pretty funny that the mirror motor writing begins “VW AG”.
I identified the temp sensor wires in the mirror plug and measured a resistance of 2.2k ohms - not a short circuit or open circuit by any means. Perhaps it was working correctly and there was a short somewhere else, or perhaps the body computer was faulty? I used a resistance substitution wheel (could use just a resistor) to substitute a 10k ohm resistance (in place of the sensor) and got a reading of 25 degrees C. This seemed to suggest that the body computer and wiring was fine, and that the sensor was somehow reading too low (i.e. a connection that was “too good”; hard to imagine!)
At this point I considered getting the car fixed under warranty, but there were two flaws with that idea; firstly the dismantling required for replacement of the mirror wiring loom (the sensor is part of the wiring; cannot be unplugged) - the chance of something getting broken, damaged, or refitted incorrectly seemed quite high - even removing the mirror glass was a slow and delicate task. Secondly and more importantly, I wondered about the accuracy of the sensor even when it had been working correctly; it always seemed to read a little high and I had read here of others having the same. Arguing with the dealer about how accurate it was didn’t seem constructive. Thirdly, two trips to the dealer would be needed, with no courtesy car... probably at least a day would be wasted getting there and back.
Digging the resin out of the sensor and removing the thermistor itself produced a reading of 8.5k ohms. So, its embedding in resin had been reducing the resistance to 2.7k ohms. I have never seen damp resin that has a resistance of 5000-6000 ohms or so; truly bizarre. If something’s wet, it’s usually only a few ohms, and if a connection’s corroded, it would increase resistance, not decrease it.
Connecting the thermistor by itself produced a reading of 24 degrees - obviously better than 55 degrees, but my second concern was now proven, as it was actually 19 degrees outdoors that day. I used another thermometer for comparison, plus the weather app on my phone. I was careful to keep the thermistor shaded, but it still read consistently five degrees high.
The solution was to rummage around in the junk box - you know, the kind of stuff people tell you to throw away - where I found a blended air temp sensor from inside the heater box of an Alfa Romeo 164 that I stripped for parts about eleven years ago. I measured the sensor at about 10k ohms and therefore believe it is a standard 10k ohm thermistor (the resistance of which is usually quoted at 25 degrees C). I extracted this thermistor and connected it, getting a correct reading of 19 degrees C.
I wondered about the linearity of the sensor - would it be compatible with the rather-different body computer of the Giulietta - so I devised a test involving a gel pack from the freezer. By the time I’d managed to wrap the gel pack around two sensors and get stable readings, the car and the separate thermometer gave matched readings about 5 degrees C, suggesting that at least in the useful range, the sensor seems accurate enough.
All that remained was to seal the sensor with epoxy resin, re-mount the mirror housing with its three screws (it holds the sensor in place), and - crucially - I drilled a drain hole in the underside of the mirror housing, so that water cannot submerge the sensor in the future. Did the same to the passenger door mirror and was surprised by how much water ran out of there, too. This will greatly help with drying the mirrors after washing the car - I usually have to use a leaf blower to get the sitting water out and it’s obvious this wasn’t always effective.
The final step, probably optional, was to connect MultiECUScan and clear fault codes from both the body computer and the climate control ECU - caused by unplugging the sensor, the body computer actually reads the value, and passes it along (or not) to the climate control ECU. So both will register a fault if the sensor is not present.
Also, using MultiECUScan to display parameters of the climate control ECU allows a cross check of the external temp against the four cabin temp sensors (blending air, footwell vent air, left side and right side) which should all be about the same if the car is parked in the shade without the engine/heating running. A good test, perhaps, to decide whether to replace an original sensor, even one that isn’t dramatically over-reading from being waterlogged!
This all seems to have made a pleasant difference to how well the climate control works - subtle heating is more effective than before, noticeable particularly in the morning. And, the fan no longer works overtime in the hotter afternoon weather. I’d always thought the slightly-high readings were from being parked in the sun, etc., but I now know that the readings can be accurate if the sensor is correct.