The "service mode" on the control panel also showed the outdoor evaporation temperature wandering around at a fairly ridiculous low level, like -20F in maybe a 40 degree day.  Normally the system runs the outdoor coil about 10 - 15 degrees under ambient, which is enough to pick up heat from the outside air and bring it inside.  This is a little nonintuitively puzzling -- if the outdoor coil is so much colder than the air around it, why wouldn't it be picking up *plenty* of heat to transfer in?

Well, the properties of refrigerants aren't that simple, and the numbers I was reading were actually quite misleading.  To transfer heat efficiently, refrigerant has to change state evenly across the entire area of an exchange device, aka a coil with fins on it.  If liquid refrigerant flow is low and it tries to expand into a given volume of piping, it can all flash to gas immediately and then spend the rest of the journey through the exchanger fairly uselessly at ambient temperature, picking up almost nothing.  So it might get really cold early in the process, but there just isn't enough mass of it flowing to maintain that state-change process through the entire length of piping.

In other words, enough of the "magic juice" had leaked out, leaving the system valiantly struggling to perform but just not able to deliver.  I eventually learned to recognize the "ridiculous superheat" as the definitive symptom of needing more refrigerant, and after a couple of expensive technician visits involving weighing charges in and out, I knew about how far down it was when the symptoms started.

So we start with Technician #1, who seemed fairly impressed and amused by the things I already knew about my own system.  During his visit he took some phone pictures of my hacks and the whiteboard diagrams on my kitchen wall, "for the guys".

The tech *almost* thought he got a hint of detection near the bottom of the coil, but after taking the lower front cover off he still couldn't find anything solid.  It did provide another good look at the high-side thermistor, where I could confirm that its wiring enters from below and is thus less likely to get water seeping into the capsule.

    Dammit.

This was already getting super-old.

That was unfortunate in many ways, but one of those was that I had to basically start over and train up another guy on what my system consisted of and what had gone before with it.  He came up to speed soon enough, and along with more fruitless leak-sniffing, we agreed to try a couple more things to proactively seal the system better.  With it once again emptied of about the same remaining amount of refrigerant as before, I suggested opening the flare fittings for a good look and possible improved sealing with a bit of lubricant on reassembly.

Now, these were the same flares that had been sketchily installed in the first place, the larger of which was visibly leaking the day after and all the installer did on the callback was crank the nuts down tighter.  These had never been disassembled since, on the assumption that if they weren't leaking in any obvious way, they were best left alone.  But Technician #2 seemed confident that because they were always suspect, they could be opened, inspected, and reassembled in a more leakproof way by applying the Nylog that the original guy should have used in the first place.

The surfaces actally looked fairly good, so he spun a little Nylog around the threads and the back of the flares and buttoned things up again.  He resisted the concept of putting anything on the *mating* surfaces, even though Nylog is supposed to be safe for systems if a little bit of it gets into where the refrigerant circulates.  It's basically a super-thick polyolester oil.  Still, color me skeptical of the sealing ability of a raw metal-to-metal junction.

What's ironic is that now they actually make crush seals for flare connections now, for exactly this issue, but the guy didn't offer to install those at the time or possibly even didn't know about them yet.  Next time these get opened, if ever, maybe that's something to do as well.

We didn't pressure-test, but he figured another evacuate-and-hold cycle would be enough of a reasonable leak test.  Naturally this also requires the test hoses to be in good shape too, as they become part of the evacuated system as long as they're hooked up.  There was no obvious pressure rise over a reasonable time, so he went ahead with refilling.  I showed him my charge calculations based on the spec per foot of lineset and indoor coil addition factor, and he actually overshot a little and put closer to 8 pounds in the system.  It actually seems quite happy with that much in it; there's a monster accumulator on the suction side that can buffer a lot of excess.

To facilitate longer-term diagnostics, the guy suggested putting some fluorescent dye in the system which might help locate a tiny leak over time.  He assured me that it wouldn't harm the system at all and wasn't any kind of leak *sealer*, which is still the subject of raging debate across the industry.  A simple tracer visible under UV light might give a better chance of letting me eventually find the leak, so I was on board with that.  He understood that I would be perfectly comfortable continuing with my own diagnostics, and that an appropriate light source would be inexpensive.

Nylog was also applied to the access-port threads and flanges as they got buttoned up again, to make sure the caps were sealing right.  Those are another direct metal-to-metal contact in this case; there are no O-rings inside the caps.  Again, this seems like an inherently flawed idea, as the tiniest scratch in one of those surfaces could easily leave a leak path.

    *Argh*.

Still with the Covid-driven apparent impossibility of finding someone to come out and do the $800 "refill dance", in despair I simply shut off the outdoor unit and started heating on the 3 kW "backup" resistance element.  Good thing I had done the secondary thermostat integration to let that work independently, because now I really needed it as "emergency heat".  While that is *just* enough to heat the house through most of the winter, it is decidedly the expensive way to do it, as there's no benefit from the COP of the heat pump *and* it has to run longer to produce the same overall BTUs inside.  With the refrigerant side of the picture out of commission again, I was almost resigned to having to go through the entire winter on the resistance coil and who knew when I'd be able to ever get the heat pump fixed.

And the kicker was that I *still* couldn't find any evidence of an identifiable leak point, even scanning all over everything with the UV light and trying to sniff key points with an ancient leak-finder I'd had for years and filter out all the false alarms.  That was only really a gas density detector, not specific for refrigerants, and would start indicating on just about anything including dust in the air.  At one point I slit into the lineset insulation at a couple of points to see if I could detect anything along the piping, and the wheezy old sniffer started alarming on the *outgas* from the cut foam cells. 

By now I knew that all the system really needed short-term was a simple top-up of R410A, about three pounds worth.  But that wasn't something I could just hop down to the hardware store and buy.  Sale of refrigerants is quite restricted, to certified industry professionals, and a bit of specialized gear is also needed to fill a system.  One of my nearish neighbors works in the HVAC industry and has his "ticket" to legally work on this stuff, and at some point we got to talking about my system.  He offered to wander over and take a quick look, and after another couple of superficial and negative sniff and bubble tests, he came up with a couple of useful ideas on how to proceed.

One was that he could bring his gear over and add the three or so pounds of refrigerant needed, which could hold me over for a while if the leak rate remained steadily low.  This is called a "gas-n-go", and while it isn't the most environmentally friendly approach, is done all the time in the industry.  For small systems, escape of a pound or two per year of HFC refrigerants isn't considered so big a deal -- especially compared to some of the allowable leakage from large industrial systems.  The right answer, of course, is to fix the leak, but if it's proving just about impossible to *find*, we run into the limits of practicality and labor cost.  We geeked a bit about all of this, and he seemed somewhat impressed that I knew what I did about the system and what to add and get the system charge back to nominal.

Since I'd never examined one of these tanks carefully, it seemed worth recording the text that they generally have on them.  The pressure inside is a direct function of temperature, and somewhat counter-intuitively, at higher temperature/pressure more of the refrigerant is in liquid phase and thus incompressible, which is why storage of a full one at too high a temperature is super-dangerous.

One might ask, how does any air get through once it's that frosted up?  Well, just the turbulence of the fan straining to pull air through can still let the coil extract more heat from the air, but once the sensors start detecting that there's too *little* superheat, a defrost cycle is done.  The system actually isn't that smart about it, and likely includes some simple blind timing into that determination -- on dry days where there's no frosting at all, it still does defrosts every so often even when there's no need for it.  I can basically "hear" how blocked up the coil is by the louder roaring that the fan makes, without looking out the window. 

Oddly, even in that pathological under-fed state the suction pressure never reached the cutoff point and the system never threw an error, it would just keep trying even if it was just wasting energy for very little heat by then.

This idea took root and grew, and along with possibly getting myself EPA 608 certified, as it's known, I could already buy HVAC tools on Amazon without even having that status.  After observing my various service visits and a few Youtubes I knew I could easily and safely handle this stuff, so why not?  If a bit more money and passing a test involving some science were all that stood in the way, it was a project well worth tackling.  I started making a list of what I'd need as permanently owned tools to work on the system, and easily found a lot of good EPA study guides all over the internet for free.  While the neighborhood guy had recommended the two or three evening training course that he'd taken over at S.G.Torrice, it became clear that I could just learn this stuff on my own and only bring in an external agency to proctor the test when I was ready for it.

And to hasten this process along, this latest top-up only lasted about a month before the system was wheezing again and giving me "the finger" of premature evaporation.  Either the leak had gotten worse, or my neighborhood buddy had done something wrong when he unhooked and closed up the fittings again.  I didn't feel right about pestering him to keep coming by with his pink bottles of salvation, and at this point, *fuck* if I was going to keep paying for $800 "professional" truck rolls.  I was determined now: I'd go ahead and get myself able to pursue this whole mess on my own, and began working on several fronts to make that happen.

Basically, every valve simply opens its port into the common chamber in the middle.  The gauges connect directly to the respective high and low side hoses.

At the moment of cracking those caps, I could hear a tiny little "pfft!" as a tiny volume of trapped refrigerant under the cap escaped.  While this was a possible red flag, I figured that no Schrader valve is perfect and over a long time, some gas would get past it and accumulate behind the cap seal.  I could also see some remains of splashed dye from the injection process, but not visibly past where the shiny area of the cap-to-port sealing surface was.

I then carefully purged the rest of the hoses and manifold, listening for when the sound of escaping gas changed to know that refrigerant had reached another one of the ball valve ends.  I used a little more system content to let the gauges cycle up and down a couple of times to try and clear the air out of their bourdon tubes, letting that little bit of vapor pressure back out to the atmosphere to bring them to zero.  This is still within the "de minimus" practice of allowable venting, just to establish a solid column of refrigerant gas and nothing else through the entire kit.  Normally that would come from a supply tank, but I didn't have that yet.

One key thing to note here is that the gauge set reads at the lineset on the "outside" of the service valves, and my voltmeter box reads *inside* the unit on either side of the compressor.  This could be distinctly advantageous over what most techs have, where they have less insight as to what's going on in the outdoor unit itself.  Presumably the Daikin "service checker" interface and PC app would have those figures available too.

It was also clearly good to "exercise" the service valves, as turning them felt quite stiff at first and then loosened up as I worked them in and out with the hex wrenches.  It was likely redistributing lubricant over the O-rings where it had been static for six-plus years, possibly leading to a better seal around the valve plungers in general.  Bypass of those could be another potential leak point too.

The next thing to observe was heating mode, for what it was worth, even without quite enough refrigerant present to make it actually work.  Observing pathological failures is just as instructive as observing proper operation.

According to the voltmeter box, the real compressor-side suction was down to around 50 PSI and slowly hunting all over the place, as the system strove to find a balance.  Along with that, the "icy finger" would grow and then retreat along that first run of outdoor coil, over and over.  You'd think it would just give up and throw an error after a while.

I finished my playing with around 150 PSI in the gauge set, which rose to around 200 once inside at normal room temperature, and there was clearly a little liquid present in the sight glass if I turned the whole set upside down.  A saturate mix!

So the next thing was to finish studying up and schedule my 608 test.  This process was rather fascinating, because there was a lot I didn't know about other types of refrigeration systems and correct handling of their contents.  Industrial low-pressure chillers using R-123 or old R-11, for example, run at both above and below atmospheric pressure, and invariably suck in a little bit of ambient air -- and have special "purge units" to collect and eject it automatically!  I learned what those little dead-end copper tubing stubs sticking out of small refrigerator compressors and such are for.  I delved further into what recovery machines can and can't do.  There are all kinds of arbitrary facts on what year certain restrictions kicked in, and how much negative pressure a given size system needs to be recovered to, and even how much a responsible technician or company can be fined per incident of illegal venting to the atmosphere.  Much of it is interesting science; some was just arbitrary cramming that I've already forgotten but could easily look up in the material I collected.  And all of my study material came for free with a little google-fu.

Taking the test was a bit of a saga, leading to one total fail for completely unrelated reasons at the company my buddy had recommended, and then a successful test session at their competition.  And I *aced* the test, only two wrong out of 100, handily earning my "Universal" certification to work on all four types of equipment.  The same shop where I took the test accepted the temporary PDF printout of what my card would look like when I went back the next week, and I drove triumphantly home with a new pink jug of R410A and gave my system the three additional pounds of juice that it had been craving.  Now I could at least cater to its drinking problem without outside help, while continuing to chase the leak conundrum.

As more time went by it was clear that the leak was still present, albeit a little slower than after my buddy had taken his stab at it, so there was some maddeningly variable factor still in play.  Even armed with a better sniffer by now, I was still basically getting nothing on my leak-hunting efforts.  I got a few one-off sporadic readings around parts of the indoor coil, and couldn't quite determine if the UV light spotted a tiny hint of bright green?? buried deep in one section of the finning.  Probing the indoor coil is easier since the air around it can be made absolutely still, and refrigerant tends to drift downward in air.  But those hints weren't consistent, just like for the first tech who visited this thing when the problem first began.  The primary suspects were still elsewhere, however, and one of the purchase cycles from Amazon included the gear necessary to change out the access-port Schrader cores.  The guys responding to my thread at HVAC-Talk agreed that in a system eight or nine years old, the original valve cores could easily be suspect.

Nothing visibly moved or leaked while doing either side, so the swap had evidently succeeded.