The dew point, always. Superheat is measured on the suction line, where the refrigerant has finished boiling and is entirely vapor, and the dew point is the temperature at which that last of the vapor would begin to condense at the pressure on your gauge. Subcooling is the mirror image: it is measured in a liquid line, so it reads the bubble point. On a single-component refrigerant like R-32 or R-22 the distinction does not exist, because the two points are the same temperature. On R-454B they are roughly 2°F apart, and swapping them costs you about 2°F on every reading before you have made any error of your own — superheat would come out about 2°F high, subcooling about 2°F low.

Glide is how far the saturation temperature moves while a blend boils or condenses at constant pressure, and it exists because the components of a mixture do not change state at the same temperature. Of the four here, R-22 and R-32 are single compounds and have no glide at all. R-410A is a near-azeotrope: its bubble and dew columns differ by 0.1 to 1.0 psig, about 0.2°F, which is small but is not nothing and is carried here rather than flattened. R-454B is a genuine zeotrope of R-32 and R-1234yf; Chemours states its glide as ~1.1 K (~2.0 R), and the published data works out to about 2.3°F at both 40°F and 100°F saturation.

Because on a fixed orifice — a piston or a capillary tube — nothing is regulating superheat, so it moves with how hard the system is working. The widely published approximation is [(3 × indoor wet bulb) − 80 − outdoor dry bulb] / 2: at 64°F wet bulb indoors and 96°F dry bulb outdoors the target is 8°F, and at 71°F and 93°F it is 20°F. Both of those are worked examples from published sources, re-derived here rather than copied. It is an approximation to the manufacturer's own charging chart and it should be treated as one — outside roughly the 5°F to 30°F band those charts cover, it has run past its useful range. On a TXV system none of this applies, because the valve holds superheat steady and the charge shows up in subcooling instead.

Most residential outdoor units ship factory charged for 15 feet of line set, and the published rate for a 3/8 inch liquid line is 0.6 ounces per foot beyond that — Goodman's TP-107 long line set guidance and Trane's installer literature both give the same figure. The trap is which length you multiply. Goodman subtracts the 15 feet the unit is already charged for; Trane's weigh-in worksheet multiplies the whole run, because it assumes the system was evacuated first and the entire charge is going in fresh. On a 50 foot run those two methods are 9 ounces apart, so read which one your manual is describing. Either way it is a starting charge, not a finished one: both manuals require a final adjustment by superheat on a fixed orifice or subcooling on a TXV.

A restriction between the condenser and the evaporator. The two readings look contradictory at first — the evaporator is starved and the condenser is backed up at the same time — but that is exactly the signature of refrigerant that is present in the system and not getting through: a plugged filter drier, a kinked or crushed liquid line, or a metering device underfeeding. It is worth knowing this one specifically because the starved evaporator on its own looks identical to an undercharge, and adding refrigerant to a restricted system raises head pressure without fixing anything. The other three combinations are: high superheat with low subcooling for undercharge, low with high for overcharge, and low with low for a metering device overfeeding.

Close, and where they are not, we say so. R-22, R-32 and R-410A — including both of R-410A's columns — agree to the printed 0.1 psig at every row across three independent publications: Chemours' Pressure-Temperature Guide for A/C, Arkema's Forane chart, and the Daikin-branded A2L chart PM-A2LPTC-USA_04-24. R-454B is the exception. That Daikin chart, whose values its own footer credits to Weitron, runs 0.4 to 0.8 psig above Chemours on the dew line and up to 4.6 psig above it on the bubble line, and its columns imply about 2.6°F of glide where Chemours states ~2.0°R. We publish Chemours — the company that makes the blend, working from NIST REFPROP — and state the divergence rather than splitting the difference. All values are PSIG at sea level; above sea level the real saturation temperature is lower than any sea-level chart returns.