What a Heat Exchanger Is Actually Doing

Why this matters

A coil that "still gets warm" is the most expensive kind of half-working part in the trades. It passes the eyeball test, so the tech chases the thermostat, the pump, the charge, the blower, and the customer pays for three visits before anyone measures what the exchanger is actually moving. The reverse mistake costs just as much: condemning a sound coil because the air off it felt weak, when the real problem was that only two thirds of the design flow ever reached it. Both mistakes come from the same gap. Nobody wrote down what the exchanger was doing in units of heat.

This card gives you the worksheet that closes that gap, and the reasoning that makes the finished sheet mean something.

Three levers, and only three

Every heat exchanger ever built moves heat according to the same product: how much surface it presents, how easily heat crosses that surface, and how big a temperature difference is pushing across it.

• Area (A) is fixed at the factory. In the field it only ever goes down: bent fins, blocked passes, a bypass path where air or water gets past the surface without touching it.
• Overall conductance (U) is how readily heat crosses from one fluid to the other, per unit of area and per degree of difference. It is not a property of the metal alone. It is the metal plus the two fluid films clinging to each side plus anything deposited on either face.
• Temperature difference is the driver. No difference, no transfer, regardless of how good the first two are.

Multiply them and you get duty: Q = U x A x (mean temperature difference). U and A are almost always quoted together as UA, because you cannot separate them from outside the equipment anyway, and UA is the number that tells you whether the exchanger itself has degraded.

The practical consequence is that "the coil is bad" and "the coil is starved" are different faults with the same symptom, and UA is what separates them.

Both sides have to balance

The second idea is simpler and does more work. Whatever heat leaves one fluid enters the other, minus a small loss to the cabinet and the surrounding air. That gives you two independent ways to measure the same duty, which means you can check yourself.

For water in a closed loop near normal building temperatures:

Btu/h = 500 x gpm x (temperature drop in F)

The 500 comes from about 8.33 lb per gallon, 60 minutes per hour, and a specific heat close to 1.0 Btu per lb per F. For a glycol mix, both the density and the specific heat move, so the multiplier drops; pull the properties for your actual fluid and temperature from the fluid data rather than reusing 500 on a glycol system.

For air at sea level and ordinary densities, sensible heat only:

Btu/h = 1.08 x cfm x (temperature rise in F)

The 1.08 comes from roughly 0.075 lb per cubic foot, 60 minutes per hour, and about 0.24 Btu per lb per F. At altitude the air is thinner and that constant falls with density, so a high-elevation job needs the corrected value. This form also ignores any moisture removed or added; if the coil is condensing, air-side sensible math will read low against the water or refrigerant side, and that gap is not an error, it is the latent portion.

Before you take a single reading

Every measurement below puts a hand or a probe near something that can hurt you, and the control for each hazard is different.

• Opening a cabinet with a fan in it: the blower can restart on a call while you are inside it. Lock and tag the disconnect for the mechanical and stored-energy hazard under 29 CFR 1910.147, and let the wheel come to a full stop before your hand crosses the plane of the housing. On a construction site the electrical isolation counterpart is 29 CFR 1926.417.
• Metering anything energized in a panel or at a control board: electrical work is carved out of 1910.147 and lives at 29 CFR 1910.333(b)(2). Prove your meter on a known live source, prove the circuit dead, prove the meter again, which is the live-dead-live sequence in NFPA 70E-2021, 120.5.
• Touching hot piping or a boiler-side sensor well: heating water above 140 F scalds on contact and flashes to steam if the line is opened under pressure. Use a strap-on or infrared reading rather than a bare hand, and do not break a joint to install a well until the section is isolated, drained, and reading zero on a gauge you trust.
• Any refrigerant-bearing path: do not open it to install a probe. Liquid refrigerant freezes skin on contact and heavier-than-air vapor displaces oxygen in a low or confined space. Read surface temperature on the tube and take pressures at existing service ports only.

The temperature difference shrinks as you travel through it

Techs often reason as if one temperature difference applies across the whole exchanger. It never does. The two streams change temperature as they travel, so the gap between them changes too, and the driving force is the average of that changing gap.

The standard average for a steady single-pass unit is the log mean temperature difference, LMTD. You take the gap at one end and the gap at the other, then:

LMTD = (larger gap - smaller gap) / ln(larger gap / smaller gap)

Two things follow. First, running the streams in opposite directions (counterflow) holds the gap more nearly constant along the length than running them the same direction, which is why counterflow units reach a closer approach with the same surface. Second, LMTD as written applies to a single-pass unit; a multi-pass or crossflow coil needs a geometry correction factor from the design data, so treat an absolute LMTD from a finned coil as approximate. The ratio between two LMTDs on the same coil is far more trustworthy than either one alone, because the geometry correction is common to both and largely divides out.

One layer usually rules

U is the reciprocal of a stack of resistances in series: the film on the first fluid, any deposit on that face, the tube wall, any deposit on the other face, the film on the second fluid. Series resistances add, so the largest one sets the total, and improving anything else barely moves the number.

On a finned air-to-liquid or air-to-refrigerant coil the air-side film is normally the controlling resistance by a wide margin, which is why fin surface exists at all, and why air-side blockage hurts more than an equal thickness of scale inside the tube. That generalization has real exceptions and they matter in the field: a tube carrying oil-logged refrigerant, a hydronic circuit running at a trickle in laminar flow, or a heavily scaled water side can all move control to the inside, and once control moves there, cleaning the fins does nothing measurable. The sibling card on fouling works through how to tell which side owns the resistance.

The worksheet, filled in

A hydronic heating coil in an air handler, complaint of "not enough heat in the north zone." Balancing report and equipment schedule are on site. All readings taken in one steady 10 minute window with the unit at full call, using the same instrument for every temperature so any offset is common to all four readings.

Recorded

Water side: 500 x 6.0 x (180 - 160) = 500 x 6.0 x 20 = 60,000 Btu/h.

Air side: 1.08 x 1,600 x (92 - 60) = 1,728 x 32 = 55,296 Btu/h.

The two sides differ by 4,704 Btu/h, which is just under 8% of the water-side figure. On a dry heating coil with cabinet loss and a single-point air reading, that is normal disagreement, not a fault. Take the water side as the better number here because a balancing valve readout and two well temperatures are less error-prone than a discharge traverse.

Delivered against scheduled: 60,000 / 78,000 = about 77% of the scheduled duty. That is the customer's complaint in units.

Now separate the exchanger from the flow. Both flows are at their scheduled values, so the shortfall is not starvation. Compute UA both ways using the same counterflow LMTD form.

Measured, counterflow: gap at the water-inlet end is 180 - 92 = 88 F, gap at the water-outlet end is 160 - 60 = 100 F. LMTD = (100 - 88) / ln(100 / 88) = 12 / 0.1278 = 93.9 F. UA = 60,000 / 93.9 = about 639 Btu/h per F.

Scheduled, same entering temperatures and flows: water drop would be 78,000 / (500 x 6.0) = 26 F, so leaving water 154 F. Air rise would be 78,000 / 1,728 = 45.1 F, so leaving air 105 F. Gaps are 180 - 105 = 75 F and 154 - 60 = 94 F. LMTD = (94 - 75) / ln(94 / 75) = 19 / 0.2260 = 84.1 F. UA = 78,000 / 84.1 = about 928 Btu/h per F.

The finding: UA is at roughly 69% of design (639 / 928) while capacity is at 77% of design. Capacity falls less than UA because as the coil transfers less, the two streams stay further apart, and the larger mean gap partially compensates for the lost conductance. A tech who tracks only capacity will consistently underestimate how much surface performance has been lost, and will keep cleaning a coil back to "good enough" while it drifts.

Both UA figures were built from the same counterflow form on the same coil, so the geometry correction is common to both and the 69% ratio stands even though neither absolute UA is exact. That symmetry is the reason the ratio is the reportable number and the absolute UA is not.

With flows verified at design and UA down about a third, the fault is on the surface: fouling, fin damage, air bypassing the coil face through a gap in the sheet metal, or entrained air sitting in the upper tubes. Pull the coil face and look before quoting anything.

What a closed balance still cannot tell you

The sheet is powerful and it has hard limits, and each one has bitten somebody.

A balance that closes proves consistency, not correctness. Flow enters each side's calculation exactly once, and it enters separately. If you took gpm from the pump schedule instead of a balancing readout and the pump is actually delivering three quarters of that, your water-side duty is overstated by a quarter while the air side is untouched, and you will read the resulting mismatch as measurement scatter rather than as the flow error it is. Measure or read flow at the device; never inherit it from a drawing.

The sheet is a snapshot of one operating point. A coil that only starves at design load reads clean on a mild day, because at part load the reduced duty is inside what the degraded surface can still pass. If the complaint is weather-dependent, the sheet has to be filled in during the weather that produces it, or with the load artificially driven up.

Small temperature differences carry big percentage errors. A 1 F instrument offset against a 20 F water drop is a 5% error in duty. Against a 4 F drop on a lightly loaded chilled-water coil it is a 25% error. That is why the same instrument is used for every reading on the sheet, and why very low delta-T readings should be treated as directional rather than quantitative.

Duty is not the same question as sizing. The sheet tells you what the exchanger moved. Whether that is enough for the space is a separate calculation against the load, and a coil at 100% of its schedule can still leave a room cold if the schedule was wrong.

References

• ASHRAE Handbook, Fundamentals volume, chapters on heat transfer and on the log mean temperature difference method
• 29 CFR 1910.147 (mechanical isolation and stored energy) and 29 CFR 1910.333(b)(2) (electrical work); 29 CFR 1926.417 for lockout and tagging of circuits on construction work
• NFPA 70E-2021, 120.5, for the live-dead-live instrument proving sequence
• Manufacturer coil selection data for the geometry correction factor applied to multi-pass and crossflow units
• See related: How Fouling Changes Heat Transfer; Why Approach Temperature Matters; How to Think About Capacity Versus Output