The Mini-Split That Lost Its Charge in a Month
Right the first time, and back anyway
This is a case where the first callback tech found the leak, in the right place, and the system came back three weeks later with the same complaint. The diagnosis was correct. The repair was the kind that makes a correct diagnosis look wrong. The interesting part is why being right about where the refrigerant was going was not enough, and what the install record was missing that would have stopped the whole thing on day one.
The timeline
Install, week 0. A 12,000 Btu/h single-zone R-410A ductless heat pump in a bedroom addition. Line set 15 ft, inside the length the outdoor unit's factory charge covers, so no refrigerant was added. Flares made with a flaring tool and tightened with two wrenches by feel. The install sheet read "evacuated, 500 microns" and had no line for a nitrogen pressure test, because the checklist did not ask for one.
First callback, week 4. Homeowner: "It runs all the time and the room is never cold." The tech found low suction pressure, the indoor coil cold only on its first few passes, and a dark, oily film on the insulation sleeve behind the indoor unit at the 1/4 in liquid flare. Oil travels with escaping refrigerant, so an oily joint on a month-old system is a leak until proven otherwise. Bubbles on the flare nut confirmed it. He tightened the nut about a quarter turn, the bubbles stopped, and he added 6 oz while watching suction pressure come up toward where he expected it. Ticket: "Leak at indoor liquid flare, tightened, recharged."
Second callback, week 7. Same complaint.
Why the first repair could not hold
The first tech's diagnosis was exactly right: the leak was at the indoor liquid flare. What he could not know from the outside was why that flare leaked. A flare leaks for one of three reasons: it is under-torqued, the flare face is defective (off-center, split, scored, burred, too short or too long), or the nut and flare were damaged by overtightening. Tightening fixes only the first, and he had no way to know which he had, because nobody had recorded what torque the flare was set to at install.
When the face is the problem, extra torque crushes the soft copper lip until it seals against the cone for a while. The lip thins and work-hardens under the extra load, and a few weeks of heating and cooling cycles open it again, now worse than before. The tightening converted a leak into a delayed leak and made the flare harder to judge.
The top-off made the second visit harder too. A mini-split's charge is set by weight, and adding refrigerant by watching suction pressure on a system with an unknown remaining charge leaves the total unknown. The second tech now had a system whose charge nobody could state.
The second visit, in order
Isolation first. On this model the interconnect cable between the outdoor and indoor units carries line voltage to the indoor unit as well as communication, which is common on ductless equipment. The tech switched off the outdoor disconnect, locked and tagged it, and proved the indoor terminal block dead with a tester checked on a known live source before and after, per NFPA 70E Article 120 in the current edition. A tech who isolates only "the outdoor unit" and then lifts the indoor unit off its plate is reaching into a live terminal block.
Recover and weigh. A pump-down into the outdoor unit would have left the total charge as unknown as before, and the goal was a known total. The tech recovered the whole charge into a tank on a scale: 22 oz. The nameplate charge for this unit with the included line length is 35 oz. Adding the 6 oz from the first callback, the system had held 35 plus 6, or 41 oz, and gave up 41 minus 22, or 19 oz, in seven weeks.
Lift the indoor unit and inspect. With the indoor unit tilted out on its mounting plate and supported, the 1/4 in flare came apart. The nut threads were fine. The flare itself was off-center: the lip was visibly wider on one side, with a hairline crack on the thin side and a bright, flattened ring where the second tightening had pressed it. An off-center flare happens when the tube is cut out of square or not centered in the bar, and the thin side is where it fails. That flare had probably seeped from the first week.
Check the other three flares. One bad flare from a crew that tightens by feel says the other three are unproven, not good. All four were undone. The 3/8 in gas flares at both ends and the outdoor liquid flare were concentric and clean, with no oil. A flare that has been torqued and then loosened has already formed against one cone, and this maker's manual said to re-make any opened joint, so all three were cut back and re-flared along with the failed one.
The repair
Cut back and re-flare. The tech cut the indoor liquid line back past the damaged section with a tubing cutter, deburred with the tube pointed down so chips fell out rather than in, slid the nut on first, and made a new flare with an eccentric flaring tool, setting the tube protrusion with the tool's gauge. The flare was checked with a flare gauge and by eye for an even, smooth lip with no cracks. Where the maker allows it, a drop of refrigerant oil goes on the back of the flare and the nut threads, never on the sealing face; this maker's manual allowed it. See related: How to Make and Torque a Flare on a Mini-Split Line Set, which owns the flaring method and the inspection.
Torque to the maker's table. A torque wrench with the correct flare crowfoot and a backing wrench on the indoor unit's flare fitting. This maker's table gave 14 to 18 N-m for a 1/4 in flare and a separate range for the 3/8 in; the tech set the wrench inside each range and recorded the setting. The backing wrench matters: torquing against an unbacked fitting twists the copper inside the indoor unit, which starts a second leak where nobody will look. Torque values come from that maker's table, not from a chart for a different brand; see related: Mini-Split Flare Torque and Flare Size Field Card.
Standing nitrogen test. Through the gas service port, with the outdoor service valves still closed so only the line set and indoor coil saw nitrogen, pressurized through the regulator to the test pressure in the installation manual and no higher. The tech stood to the side of the indoor joints as pressure came up, because a flare that is going to blow off does it during pressurization. Every flare soaped clean. The gauge was read at the start and after 30 minutes with the line set and the room at a steady temperature, and showed no drop. Ambient temperature went on the ticket with both readings, because a nitrogen charge rises and falls with temperature and a reading taken in morning shade and afternoon sun can fake a drop or hide one. See related: How to Run a Standing Nitrogen Pressure Test on a New Line Set.
Evacuate and hold. Nitrogen released outdoors, then both service valves opened so the outdoor unit, emptied by the recovery, was evacuated with the line set and coil through core-removal tools to 400 microns, isolated, and held: it rose to 470 in 10 minutes and leveled. Evacuation and Vacuum Decay Verification owns the pass rule; this one passed.
Weigh in the full charge. 35 oz, the nameplate charge for the included length, weighed in from a cylinder on a zeroed scale. The 22 oz recovered went to the recovery cylinder log rather than back into the system, because its weight was known but its history included a top-off from an unknown cylinder. Every ounce in and out was logged per The Refrigerant Added and Leak Record SOP.
Restore and verify. Interconnect and indoor unit reassembled, panels on, outdoor disconnect restored with the tech standing clear of the outdoor fan. After 20 minutes in cooling, the indoor coil was cold across its full face and supply air at the vanes read 52 F against a 75 F room.
Confirming it held
The tech booked a 30-day check, and asked the homeowner to call if the room stopped holding temperature before then. At 30 days the indoor flares showed no oil on clean insulation, the bubbles stayed silent on all four flares with the system running, and room temperature held. Because the charge had been weighed in from a known zero, a weighed recovery at a future failure would now tell exactly how much was lost.
What the install checklist changed
The original install was not careless in the way the customer would assume: the line set was the right size, the flares were made with a real flaring tool, and the vacuum was pulled to a gauge. It was missing two numbers, and either one would have caught this flare before the crew left.
A standing nitrogen test reading is now a required line: start pressure, end pressure, duration, and ambient temperature at each reading. An off-center flare with a hairline crack loses nitrogen in a 30-minute test far faster than it loses refrigerant in a running system, and it does it while the indoor unit is still off the wall.
A torque value per flare is now a required line, recorded from the wrench setting. A recorded torque would have told the first callback tech that tightening was not the answer: a flare that leaks at the maker's torque has a face problem, and the only repair is a new flare.
And the callback rule changed: a flare that leaks is cut back and re-made, never tightened past its recorded torque, and the charge is recovered and weighed in rather than topped off.
When the call goes differently. If the oil had been at a brazed joint inside the outdoor unit instead of a field flare, the fix belongs to the maker's warranty process, and the install crew's flares would be off the hook. If the line set had run past the included length, the weigh-in would add the maker's per-foot adder from the long line charge table, not just the nameplate figure.
References
- See related: How to Make and Torque a Flare on a Mini-Split Line Set
- See related: Mini-Split Flare Torque and Flare Size Field Card
- See related: How to Run a Standing Nitrogen Pressure Test on a New Line Set
- See related: Evacuation and Vacuum Decay Verification
- NFPA 70E Article 120, current edition