Thermistor troubles

  Diagnosis, repair, autopsy of a dinky Daikin part

  The Daikin heat-pump system had been pretty rock-solid for close to six years since its installation, handling the deepest cold of winter without assistance from the resistance-heater, along with a modest amount of summer cooling / drying.  Along came the summer of 2018, a bit warmer and wetter than those previous, which led to more need for air-conditioning.  My usage pattern was pretty simple: run the system for a couple of hours a day with the HRV on high-fan exchanging air at the same time, to make sure the "fresh" incoming air got *dried* before getting sent around to the rest of the house.  I wasn't even bothering to use the condensate pump -- a bucket under the drain pipe would handily catch that much extracted water and get dumped out later when everything stopped dripping.

Except that one afternoon, I went to start up the system and got the somewhat dreaded flashing green LED on the controller.  Huh, had I forgotten to re-enable the data line or something?  Tapping the "menu" button brought up something I hadn't seen before: error code C5.  I still had all the service and repair PDF manuals from the research/learning days, of course, and it took a while to remember after all this time which was the right one to look in ... but bottom line, C5 meant the thermistor on the gas pipe at the inside coil had gone on the fritz.  The system refused to start up at all in these conditions, not even in some kind of less efficient "limp mode".

Since the house naturally stays so cool by itself it wasn't an emergency, but the CO2 level was gradually creeping up and there was no way I was going to bring the disgusting soup that passed for outdoor air inside, so I had to start digging into this.


[Images are linked to larger copies.]
Location of bum gas-pipe thermistor It was pretty obvious where the thermistor sits; inside a tightly wrapped-up bundle of foam a little way in from where the lineset pipes enter the unit, with a wire running into it.  The other end of the wire could easily be traced up to the control board, where it plugs into connector X17A and is called R2T in the schematics.  However, the controller/thermostat upstairs refers to this as sensor "Th3" in the diagnostic display.  Are we totally confused yet?!  Anyway, the upstairs panel was showing that temp value as "--" instead of the 60F or so it should have been in here, so it was clearly out to lunch from the point of view of the unit.

Thermistor R2T and a custom boss it slides into It took a bit of tight-space work to cut loose the foam and glue around this thing, but eventually it was all exposed.  The thermistor pod goes into a custom-fit boss that's brazed right onto the pipe, ensuring the best possible thermal contact and the sensor sitting at the correct angle around the pipe, the same as where a traditional TXV bulb would clamp right at the "sweet spot" gas-above-liquid level.  A small spring clip inserted along with it keeps the thermistor module pressed evenly toward the pipe itself.

Stupidly long wire, probably for different models Pulling the wiring bundle apart revealed the full extent of the stupidly-long wire going from the sensor up to the control board.  The same one is probably also used in different models with longer distances between the coils and electronics.  Pulling the little connector at the board and getting an ohmmeter on the sensor showed somewhere around 160 ohms, a far cry from the 24Kohms or so it should have been around the basement temperature.  So, at least partially shorted somehow.  The control-board side seemed fine, delivering a predictable 5 VDC across the connector which I assume comes from a voltage-divider sense circuit. 

  Then things started to get weird.  While watching the meter I started wiggling parts of the sensor and wire around, to determine if anything was physically flakey about it.  When I happened to tap the sensor bulb against something, suddenly the resistance changed and shot up to somewhere around the right level!  Now I couldn't even get it back into the shorted state with more wiggling and flexing and banging.  It was like mild impact had suddenly, magically fixed the thing somehow.  I plugged it back into the electronics and now the upstairs panel read "Th3" at 70-something F, slightly warmed above basement ambient from my hands fooling with it.

  *WTF*.

I began to suspect that some bit of substance had gotten loose inside the thermistor module and I'd just knocked it out of the way, but this wasn't something I wanted to rely on long-term and I was also starting to think about how to chase down Daikin parts.  It was also rather dismaying that such a small thing was causing total system nonfunctionality.  What if this had happened an hour after I left the house for a long winter roadtrip??  I'd be two days down the road and the freeze-alarm would suddenly be landing on my phone, which wouldn't exactly be good news.


Warming the sensor, reading acceptable value I put it back on the meter and warmed it some more, watching it behave exactly like an NTC thermistor should -- my hand got it down to 14k or so, and letting it cool back to basement temp had it around 23k.  Response was fairly slow; the sensor has some amount of its own thermal mass.  But with it now behaving right maybe I could exchange that house air, so I shoved the sensor back into the gas-pipe fitting, halfass slapped the cover back onto the unit and fired it up.  I figured I could run until it failed again, even if lack of a tiedown around the sensor might make its measuring a little less accurate.

Diagnostics: sensor is way off again The system would now run and cool, but soon after startup I observed that the Th3 temp had shot up to some ridiculous value -- the short was back, at least partially.  But somehow, with the unit being told that its coil outlet was basically on fire, it kept running long enough for a decent air exchange until I shut it down.  Opening up the air-handler again and banging the thermistor against the edge of it immediately broke the short and restored the right resistance value, so for the moment maybe I had a klunky way to deal with it temporarily until I could get a replacement part.

How it could do a superheat calculation in this state, I have *no* idea.


  I limped it along like that for a solid week.  Meanwhile, a lot of phone-tag was going on.  Given my prior great experience with N.E.T.R, I called them.  Their "A tech" that I had totally geeked out with before, Joey, was long gone from there by now, and the *one* guy they still had who knew Daikin worth a damn was on his way to the Cape for a big multi-day job at an air force base or something.  I was like "what am I supposed to do, swelter for a week?"  They said there was nothing to be done, and the idiot service front-desk lady kept saying "you need a Service Checker, and we don't have that or the necessary skill here this week".  I kept telling them they didn't need a service checker because I already knew what was wrong with it, I needed a *part*.  No, they wouldn't believe that, they said the "service manager" told them they can't even order that for me.  I asked to speak to him, and eventually did on a callback; he was actually quite nice and apologetic and did remember Joey sort of fondly ... and confessed that other companies had come into the area over the years and basically taken most of the VRV business away from them other than a couple of remaining commercial customer accounts.

Again, this unit isn't a typical residential type, it's more light-duty commercial.  Almost all inverter-drive systems use electronic valves and full sensing and are essentially "Variable Refrigerant Flow" now anyway, residential included, but the peculiarities of working with what Daikin actually calls VRV and/or Skyair are obscure and weird enough that a lot of outfits just can't afford the special training.  He felt sympathetic to my little plight and threw out a couple of names of other companies that might be able to help, but they turned out to either be clueless or simply refused to deal with homeowners.  After hearing "you need a licensed HVAC technician" enough times the unbelievable arrogance was starting to really irk me; I probably knew more about this system than most of the supposedly trained local techs.  Having them go through the same diagnostic process for the same conclusion would be a total waste of time and money.

Just about when I was about to start calling Daikin directly, one of the service supervisors at a place which he apologetically said only does commercial contracts, nonetheless finally turned my day around by suggesting that I just ask F.W.Webb.  He reminded me that they're a major Daikin supplier in the area, and thus could order parts, and they're happy to sell to homeowners.  I hadn't even thought of that, despite this system not only having been *delivered* by Webb in the first place, I had been to their shops on my own several times for other HVAC and plumbing stuff!  This thing just seemed like it would be too specialized, but not so.  A guy at Webb was able to look up my model and the part number for "thermistor, gas" and order one COD, no problem.

I talked to Daikin later anyway, and their tech-support department is quite good once you get hold of the right people who actually know stuff.  I explained my whole predicament and procedure and they totally understood, seemed glad that I was on top of the situation, and also warned that these thermistors actually do fail as a somewhat common-pattern problem.  Both of them, in fact, on gas and liquid pipes, and on a variety of similar units including minisplits.  Why?? I asked; they didn't seem to know, but said they considered them almost a "maintenance part" or expendable.  I told them that once the new thermistor arrived and the repair was done, I would do a full teardown on the failed one as best as I could and get them the results, and they encouraged me to send email once I had that ready. 

Seven days and seventy fucking dollars later, the new part was in my hands.


New thermistor looks rather different The new sensor [bottom] had a rather different package, hopefully better sealed than the old one.  The fact that the rubber wasn't really bonded to the old sensor's copper tubing and I could wiggle a small gap around it led me to my first theories as to why this had failed.  I crocked the new sensor into the unit temporarily enough to run, in case I needed it back out for any comparisons, and planned some general fixups around buttoning everything else back up.  Meanwhile, it was autopsy time.

Slicing the capsule open Step 1: slice the housing open, carefully and completely enough to bend the tubing away from the rubber inside without disturbing it.

Hairline crack through encapsulant Under the magnifier I could immediately see what the problem likely involved: a couple of hairline cracks into the rubber, appearing to go most of the way through the center.  That's the thin bit with white showing through from below, not the remains of my slit with copper dremel-dust in it.  [Use the big pic.]

Missing rubber On the other side, some of the rubber seemed to be flaking away to reveal the side of the thermistor package itself.

Carving rubber away Another long crack through rubber
I took a surgical scalpel and started gently carving the rubber away.  I needed something extremely sharp to cut without distortion, and I was working under the scope most of the time in case I got down to any different materials.  The shape of the harder thermistor pod started to gradually emerge from the rubber, and *another* long internal fissure showed up in the process.

Hard plastic package Finally I was down to the little thermistor "torpedo" and whatever remained around the wires.  Now I could see a substantial crack in the thermistor envelope itself.  This thing was a real mess.

Thermistor leads through cracked area Having found nothing looking like conductive material yet, I carefully broke the torpedo open along that crack.  A lot easier than trying to shave down the encapsulant by layers, which turned out to be some kind of brown resin and somewhat brittle.  Here I found a little bit of soft solder where the wires were bonded to the thermistor leads, that had simply pulled apart ... probably that early RoHS crap ... and while not as much crystalline evidence of corrosion as I thought I might find, not exactly clean and clearly an area where a foreign substance could have gotten in around the leads.

  The substance?  Water.  In contact with metals, and with a voltage applied across the space.  What do you get with all that?  A battery, or electrolysis.  Everything pointed to water ingress being the problem, including an additional and very interesting piece of evidence I had turned up in the testing process.

Before it got cut open, when the sensor was rapped back into "fixed" state and left alone, the short did not reappear when ohmed out later.  If the meter was put on the "diode check" mode which would send a higher voltage to the sensor, after a minute or two the value would suddenly drop from "infinite" down to something less than 1K all by itself.  Knocking the sensor around a little would make it fluctuate or open up again, but with the meter applied [in either direction!] it would soon settle back down to the shorted, malfunction state.

Which explains how the old sensor could start "correct" installed back in the unit but start giving insane readings soon afterward -- the 5V sense supply was acting like my meter, putting a voltage bias across the thermistor and "growing" the short back across the leads.  I couldn't really see physical evidence of that here but the relevant space was basically microscopic, so tiny hairs of rapid electrolytic growth through the width of the crack in the resin is entirely feasible.  There was likely a tiny bit of water in here with enough ions trapped in it to act as an electrolyte, and impact would disrupt it just enough to need a new conduction path established once enough voltage hit it again.


Testing both halves I put the meter onto both halves of the sensor at once, now reading a correct thermistor value from the sensing part on the left.  With the crack now completely opened up any trapped water was now gone, so my meter didn't produce any changes after waiting a while.  For yucks I put a drop of water on the thermistor half, and the value quickly fell to a couple of K or thereabouts and changed as I prodded the water around on that surface.  So clearly, it still became wrongly conductive when wet.  Proof enough for me.

But how had this happened in the first place, besides a badly constructed sensor?  [And more importantly, where the F was Goodman OEMing this garbage from, and how many more of these were already fielded waiting to fail?]


Dead-end pit in foam wrapper The vapor line is the one that gets "beer-can cold" in air-conditioning mode, so it runs quite wet.  The sensor is wrapped up in a foam blanket which has a little recessed cutout to accomodate the boss and the sensor, and eventually that pit will fill up with water.  With an imperfect seal at the wire end of the sensor and internal hairline cracks, gravity and/or capillary action would inevitably let that water seep in.

Gas pipe has a slope Not helping matters is the fact that the pipe has a slight slope here, and as built the sensor wire arrives from the high side.  That's probably the primary arrival path of the water, and the combination of positioning of all these parts provides no opportunity for the sensor and its surroundings to drain.

I'll also note that this area is just inboard of where the lineset got brazed on during installation, so the whole assembly may have gotten rather hot and could have contributed to early degradation of the sensor.  The original foam wrapper is definitely discolored toward the outer end, a possible telltale from that high-temp event.


  So a couple of weeks of air-conditioning, even just a little bit a day, left the whole assembly soaked a good amount of the time.  This probably would have never happened in heating mode -- in fact, the gas flow running upwards of 150F in winter would probably have helped drive any water back *out* of the sensor cracks.  Maybe something like that even happened over a couple of seasonal cycles and I never knew how close to failure it could have come.  But this year, with more A/C use and leaving everything to stew in its own soggy juice longer, the water finally found its way to a critical area.

No wonder these things periodically fail.

Now the big question: if the new sensor has actually been re-engineered to be more leakproof and defect-free.  It visually looks better, but I'm not drawing any conclusions.  Obviously it and its wiring have to be hermetically sealed against this sort of environment near the active coil, or anything metal inside will corrode away.  This is not rocket science -- my car, for example, has several thermistors that have been sitting in harsh outdoor environments for over a decade and *those* aren't failing.  Why?  Because Denso knows how to build sensors that last.

As I considered how to rework the air-handler wiring in general, I decided that I'd mount the thermistor the other way into the boss, bring the wiring in from slightly downhill and give it a tiny drip loop, and not be so anal about entombing the whole mess in tight insulation.  That would at least give it a fighting chance to resist water ingress and dry out better between A/C runs.  In the end the insulation was just a short length of slit-tube pipe foam, clipped over the length of pipe from above with the slit *left open at the bottom* with the drip-loop hanging out through it.  This section of pipe comes right after a big manifold structure that is completely uninsulated and sits in the airflow, so everything around there in the box is probably about the same temperature anyways.

The liquid-line thermistor is mounted at a steeper angle, and its wire arrives at the sensor itself from below.  It can sort of be seen in the "tech butt shot" about 60% through section 05 of the main story.  Interestingly, it's a different part number but with the exact same electrical characteristics.  I now fervently hoped I wouldn't have trouble with that one, but perhaps laying in stock of a replacement couldn't hurt?  Really, the ruinous pricing of these things would be the part that hurt, but even with the system still possibly under ten-year warranty, getting this fix done my way was a heckuvalot easier than trying to get a capable HVAC company to even give me the time of day, let alone send a tech out in a timely fashion who would then screw around with a laptop for an hour to confirm a problem I already found.


Fan pulled out for inspection As long as I was in here, I decided that part of rerouting the wiring would make it easier to pull the fan assembly out without disturbing any of the other stuff.  That sucker is heavy!  And it just hangs off that separator plate below the control box.  It also runs wet, and after a while the end of the ECM housing is covered with condensation which then doesn't dry off for a long time.  Clearly, that's more electronics that are getting wet down in A/C mode, and those are not only carrying high voltage but rectifying it into DC and holding it across big capacitors.  Made me wonder how often they fail for water-related reasons too.  Most installations probably run a lot more cooling than I need to, but I would try to keep an eye on this.

Neatened-up wiring With the sensor and EXV harnesses tucked out of the way and the fan connectors on their own dedicated run, everything went back together much more neatly and more accessible for next time.  A minor headache was the ground wires; both the coil frame and the fan motor need their own direct electrical connections to the box, because they tend to be isolated by the plastic drip tray and rubber bushings respectively.

_H*   180809