Coil Freeze Protection Air Systems

A split coil is the air side's most expensive five minutes. Water in the tubes freezes, ice takes about nine percent more room than the water did, and a copper tube that was fine at 6 AM is weeping onto the mechanical-room floor by 7 — or into a ceiling two floors down. Every air handler with water in its coils and outdoor air at its intake carries this risk, and every one of them ships with a stack of protections against it that most people only meet when one of them trips. This page walks that stack: what is actually freezing, what sets how exposed a given unit is, and what each layer does. One question, start to finish: why do coils split, and what stands between a cold morning and a split coil?

What Actually Freezes

Start with the quantity, because it is the one most often got wrong. The hazard is not the supply-air temperature. A unit designed to deliver 55 °F air cannot freeze a coil with 55 °F air — the coil freezes on the air that reaches its face, which in winter is the mixed air, or past a preheat coil, whatever that coil left behind. A healthy-looking discharge temperature on the graphic says nothing about what the first row of tubes felt at 6 AM. Watch MA-T, and watch what is in the tubes.

What is in the tubes decides whether there is a hazard at all. Freeze-burst protection — glycol, preheat, drainable coils, circulating pumps — is written for water and steam coils, because those are the coils with something inside them that expands. A hot-water heating coil is protected only while hot water is moving through it; the classic split is a modulating valve throttled nearly shut at light load, the coldest air of the morning meeting the slowest water of the day in the same tubes. A chilled-water coil sitting idle through winter is worse off: no heat source exists to rescue it, so it survives on glycol, on being drained, or on never seeing freezing air. A steam coil freezes its own way — a throttled valve pulls the coil below atmospheric pressure, the vacuum holds condensate up in the tubes, and sub-freezing air does the rest; the Coil Freeze Risk Checker's steam notes walk that mechanism and the face-and-bypass arrangement that defeats it. And a packaged DX rooftop with a gas furnace has no water coil to burst at all — its cold-weather concern is an evaporator icing over and losing airflow, a compressor problem rather than a burst problem.

One more framing before the numbers, because it explains why a southern sequence and a northern one can both be right. Cold air is a hazard in both climates — on opposite sides of the duct wall. Up north, air near freezing crossing a wet coil bursts the coil. Down south, 55 °F supply air inside a duct running through an attic at an 80 °F dew point sweats the outside of the duct, and the margin is thin enough that a southern job specifies the vapor barrier as carefully as the insulation's R-value. Two cold surfaces, two failures — and two different quantities: the coil-face temperature up north, the duct skin against the attic dew point down south. A tech reading a spec written for the other climate does well to notice which surface it was written against.

Exposure Has Three Terms

How exposed a unit is multiplies three things: the coldest outdoor air the design plans for, the outdoor-air fraction the mixing box runs, and a wet coil in the path. Climate sets exactly one of them.

Designers size winter equipment to a published heating design temperature — the 99.6 % figure, the outdoor dry-bulb that only a few dozen hours a year fall below. Those figures put Miami near 46 °F, Houston just under freezing near 28 °F, Atlanta near 19 °F, Boston near 8 °F, and Minneapolis near −15 °F. In Miami the coldest air a designer plans for is roughly 14 °F above freezing — coil burst sits outside the design envelope. In Minneapolis it is roughly 47 °F below, and everything on this page is load-bearing every January morning.

The second term is independent of climate. A 100 %-outdoor-air unit — a makeup-air unit, a lab's exhaust-makeup unit, a dedicated outdoor air unit — has nothing to dilute the intake with: whatever the design temperature is, that is what hits the coil, and 100 %-OA units are the ones the freeze-protection literature singles out as particularly at risk of bursting heating and cooling coils and flooding the building. A mixed-air unit at a 20 % minimum is diluting one part of outdoor air into four parts of 75 °F return. Put the two side by side on their design mornings: a 100 %-OA unit in Atlanta sees 19 °F at the coil. A minimum-OA mixed-air unit in Boston sees, in °F, 0.2 × 8 + 0.8 × 75 = 1.6 + 60 = 61.6 °F (in °C: 0.2 × (−13.3) + 0.8 × 23.9 ≈ 16.5). The Atlanta unit is far more exposed, and the map had nothing to do with it.

The third term is the wet coil, and it is where the first two meet. Below 100 % outdoor air, the two streams do not blend the instant they meet — cold outdoor air is denser and tends to stay stratified, so a cold layer can ride along the bottom of the plenum and freeze the bottom rows of a coil while the average mixed air reads comfortably warm. The Coil Freeze Risk Checker's note on coils that split while the trend looks fine puts a field number on that layer; the point here is that the OA fraction gives you an average, and coils do not freeze on averages.

Why the Hand Math Runs Warm on a Cold Day

The mixing arithmetic from Air Handlers — %OA × OAT + %RA × RAT — weights the streams by volume: the damper or CFM share. The coil does not care about volume. It cares about how many pounds of cold air arrive per pound of warm, and cold air is denser. At 0 °F and 40 % RH a pound of dry air occupies about 11.6 ft³; at 75 °F and 50 % RH it occupies about 13.7 ft³ — an 18 % spread. So a 20 % damper share on that morning is not 20 % of the air by mass. It is 22.8 %.

Run those two streams through the site's moist-air engine on a dry-air mass basis and the mixed state lands near 58.1 °F, where the hand arithmetic says 60.0 °F. The hand number reads about 1.9 °F warm — and it reads warm on exactly the cold mornings when the margin matters, against a freezestat band that lives at 35–37 °F. That gap is a footnote on a 60 °F mix. It is a different conversation on a 39 °F one. The Air-Mixing Calculator runs the mass basis and says why; the hand form is still the one to do on the roof — just know which way it leans.

The Protection Stack

Here is the stack, roughly in the order a unit reaches for it — the same order the Coil Freeze Risk Checker's reference row lists. Each layer exists because the one before it can fail.

1. The freezestat — the hardwired trip. A vapor-charged capillary element, 20 ft or so of it, serpentined back and forth across the coil face on the leaving-air side, that trips when any 8–12 in. of it sees air below setpoint. Not averaging — the opposite: it reacts to the coldest spot it crosses, which is the whole reason it can catch a stratified cold layer the MA-T average misses. It is wired as a normally-closed contact in series with the fan's safety string — to the starter, or to the drive's safety input on a VAV unit — so a trip stops the fan without asking the controller's permission, and the sequence then closes the outdoor dampers and drives the heating valve open. Mount it on the leaving side: on the entering side a short length of tube sitting in a cold stratified layer trips it while the coil itself is in no danger. Setpoint: stats ship set around 35 °F, adjustable, and are typically applied between 35 and 37 °F — a national device band, not a regional one. A house setting of 38 °F sits a degree above it: a deliberate early trip that buys a degree of margin and accepts a few more nuisance trips on marginal mornings. Write the number down; it is a choice, not a code line. And insist on manual reset: an auto-reset stat cycles the fan all night while the coil keeps icing, and nobody finds out why until the puddle.
2. The mixed-air low-limit override — the software floor. Before the stat ever has to act, the controller should have seen the dive coming: when MA-T sags toward a low limit, the override drives the dampers back toward minimum whatever the free-cooling logic wants. Economizers teaches it as the floor under free cooling, and that is where to read it. Two things belong here instead. First, it sits above the hardware stat on purpose — the DDC Workbench's air handler stacks a software discharge low-limit at 41 °F over its hardwired stat at 38 °F — its mixed-air override is a separate modulating loop that starts throttling the dampers at a 50 °F floor — so software acts first and the stat stays the backstop rather than the first line. Second, it reads the same mixed-air sensor as everything else, so a single-point sensor parked in the warm half of a stratified plenum defeats it quietly; an averaging element across the plenum is the fix, and the sensor-strip widget in Air Handlers hides a war story about a sensor that read the coil instead of the mix.
3. Pump on freeze, valve open on trip. The strongest protection a water coil has is hot water moving through it. So the trip — or a mixed-air threshold, or outdoor temperature alone; sequences differ on the trigger and the spec says which — starts the coil's circulating pump and drives the heating valve wide open, dumping heat into the tubes the moment the hazard shows. The 100 %-OA designs make it routine: run the cooling-coil circulation pump whenever the air leaving the preheat coil is below 40 °F. A valve that only cracks open is the hazard, not the protection — a trickle of hot water through a coil in freezing air freezes before it makes it back out.
4. Glycol at burst strength. For the chilled-water coil idle all winter, where no heat exists to rescue it, the answer is in the fluid. Freeze point is where slush first forms; burst point sits far below it, because slush still flows and absorbs expansion — a 30 % mix can sit in 0 °F air all winter without splitting, which is burst protection, not an operating condition. The Coil Freeze Risk Checker carries the freeze-and-burst table for ethylene and propylene glycol by concentration. Two field facts travel with it: glycol protects tubes, not trips — the stat reads air, so a glycol loop trips exactly as often as a water one — and loops drift lean as they get topped off with water, so a refractometer reading of the actual loop outranks the number on the fill tag.
5. Drain what is truly idle. An empty coil cannot split. A chilled-water coil that will not run until spring can be isolated, drained, and blown out with compressed air — the blow-out matters, because a coil drained by gravity alone still holds water in the low return bends, and that water splits the bend. It costs a refill and an air bleed in the spring; it costs nothing in glycol's heat-carrying penalty.
6. Winter-safe minimum positions. The quiet layer under all of the above: a minimum-OA share that keeps the calculated mixed air comfortably warm by arithmetic, so the routine morning never gets near the bands above. Air Handlers works the number — one part outdoor to four parts return holds the mix near 60 °F even at 0 °F outdoors. Three things eat into that margin, and all three appear on this page: the mass basis (the hand number reads warm), stratification (the bottom of the coil is colder than the average), and the damper itself — a commanded percentage is a blade angle, not a flow fraction (flow is not linear with stroke, which is why the floor gets proven with an airflow station or a traverse, not by commanding a number), and the flow a fixed angle delivers moves with the pressure across it: a VAV fan turned down for a January load, wind and stack effect on the intake. A minimum proven on a mild afternoon is not necessarily the same fraction on the design morning. The margin is real. Just do not spend all of it.

Here is where each layer lives on the unit. The three exposure terms are numbered along the air path; the two low limits — software and hardware — sit where they read, which is not the same place.

Why the Sequence Looks Different by Climate

The economizer sequence genuinely does look different down south, and the energy code is why. It does not ask for an airside economizer at all in the very-hot zones — 1A and 1B on the climate-zone map — because outdoor air there carries too much heat too much of the year for free cooling to pay back. Where an economizer exists in a humid zone, it locks out earlier: the fixed dry-bulb high limit sits at 65 °F outdoors in zones 1A through 4A, against 70 °F in 5A and 6A and 75 °F in the dry, marine, and cold zones — and differential dry-bulb changeover is not permitted in 1A through 4A at all, because comparing two thermometers admits humid air that costs more to dry than the free cooling saves. (Those are the 2013 energy standard's figures, confirmed against the 2018 model energy code; editions move, so read the one your jurisdiction adopted.)

Now the trap, because it is the one that reads perfectly well and is wrong. All of that constrains the warm end of the economizer band — when free cooling must stop. The only cold-end number it does write is for DX units: with the compressor running, the outdoor damper stays open and is not to start closing — to keep the evaporator from icing on minimum compressor run time — until the leaving air is below 45 °F. That is a refrigerant-coil rule, not a water-coil one; for the water coil the code says nothing at the cold end at all. So a southern economizer running in a narrow mild band is a true description of where it happens to operate. It is not the code forbidding cold-weather economizing in the South, and a sequence written as if it were has misread the standard.

Is the hardwired stat required? The honest answer is that it lives in this job's specification — and the pattern is worth knowing before you go read it. No model building code mandates a coil freezestat: the mechanical code's hydronic chapter has no freeze-protection section, and the energy code's one freeze-protection line regulates the controls on freeze-protection heating — shut it off above 40 °F — an energy rule, not an install mandate. The public federal guide spec (UFGS 23 09 93) puts a freezestat in every hydronic-coil air-handler sequence, unconditionally, hardwired to the fan starter with manual reset at the device; the word climate appears nowhere near it. The flagship sequence guideline (ASHRAE Guideline 36) treats the device as optional, layered over a mandatory software freeze sequence — "if a freeze-stat is present." And owner standards split by latitude in a telling way: cold-climate owners spell out every trip action; warm-climate owners simply never mention the device. Silence is not a rule. It is not permission to omit the stat and it is not a requirement to fit one — it is the absence of an answer, which means the answer is on this project's drawings and in its sequence. No code, guide spec, or guideline we can find waives or mandates the device by climate zone; practice varies, and "they don't use them down south" is a claim no source makes.

One more reading of the map. Houston's design temperature is below freezing too — worth remembering before anyone says it doesn't freeze here. February 2021 put Texas far outside its design envelope, and a balancing firm there documented a school district that set its freeze protocol aside, shut the central plant down, and flooded. The map sets one term of three. It does not zero the product.

Reading the Trip

The freezestat tripped at 6 AM. Before anyone resets it, treat the trip as evidence — the stat is the one witness that was awake. Pull the MA-T trend against OA-T and the damper command for the hour before the trip: a mix that tracked outdoors with the dampers commanded to minimum is a damper that was not at minimum — a slipped linkage, a seized actuator, a return damper stuck half closed — and the stat did its job. Find the element and note where it sits: leaving side of the first wet coil, serpentined across the whole face, is right; a short run on the entering side, or a coil of it parked in one corner, is a nuisance-trip machine. Check the heating valve and the pump at the time of the trip — a valve cracked open at light load is the classic split, not a defense. Then take the numbers to the Coil Freeze Risk Checker: place the mix against what is actually in the tubes — water, glycol at a measured strength, steam condensate — and read the margin. A trip with glycol in the coil was still right; the stat reads air. A stat set below the fluid's freeze point can only trip after ice is already possible. And a calm verdict on a calculated average is never a reason to skip a layer of the stack: the tool cannot see stratification, circuiting, or the cold corner the sensor misses — which is why the stack is a stack and not a single device.

What this page is — and isn't. A map of the hazard and of the stack built against it, so a trip reads as evidence instead of a nuisance and a protection design can be judged for what it is missing. It is not the protection, and neither is any number on it: the freezestat band, the software-limit gap, the burst points, and the design temperatures are patterns to recognize, not settings to copy. The coil manufacturer's data, the fluid maker's chart, and this job's own sequence and specification govern — and when the cost of being wrong is split tubes and a flooded floor, nothing on a web page outranks them.