How to Read a Temperature Difference Across a Component
Why this matters
Most bad temperature-difference readings are not bad because the tech read the meter wrong. They are bad because the two ends were not comparable: different instruments, different depths in the stream, taken minutes apart on a system that had not settled, or taken inside a region where temperature had stopped tracking heat at all. The number that comes out of that looks exactly like a good number. It gets written on the ticket, quoted from for a year, and drives a parts recommendation that the equipment never needed.
This is a measurement procedure, not an interpretation guide. What the number means once you trust it belongs to a sibling card.
Lead with the isolation, not the probe
Every one of the placements below puts a hand or an instrument near something that will hurt it.
- Hot fluid under pressure. Never break a fitting, loosen a plug or open a well on a live pressurized system to insert a probe. A hydronic or domestic-hot-water line at operating temperature releases flashing hot water at the moment of loosening. Isolate the section, relieve the pressure under 29 CFR 1910.147, and confirm zero at the point you are opening rather than at a gauge upstream of a closed valve. If the reading cannot be taken from an existing port or from the pipe surface, it waits for a shutdown.
- Hot surfaces. Take a non-contact reading at a standoff before deciding anything is touchable, and select hand protection rated for the temperature you just read under the hazard assessment at 29 CFR 1910.132(d)(1).
- Anything inside an electrical enclosure. De-energize and lock out under 29 CFR 1910.333(b)(2) and prove the conductors dead with the live-dead-live sequence in NFPA 70E-2021, 120.5 before an instrument or a hand goes in. A probe lead is a conductor, and it will reach across a gap that a finger would not.
- Rotating equipment. Probes, leads and sleeves all get pulled into fan wheels and couplings. Route leads away from the rotating plane and clear of any inlet before you energize, not after.
The record you are filling in
The deliverable of this procedure is not a number, it is a record with six fields. A reading missing any one of them cannot be defended later, because there is no way to reconstruct whether the two ends were comparable.
| Field | What goes in it |
|---|---|
| 1. Two locations | Exactly where each probe sat, relative to every port and bypass |
| 2. Both raw temperatures | Both ends, not the difference |
| 3. Instrument and method | Same type at both ends, with resolution |
| 4. Clock | Minutes of steady operation before the reading, both ends taken together |
| 5. Flow | The flow rate or fan condition at that moment |
| 6. Saturated or not | Whether either end sat inside a phase change |
Fill it in in that order. Each field below is the reasoning for why it is a field, and what a reading is worth without it.
Field 1: the two locations. "Across the component" is ambiguous on almost every real assembly, because there is usually something between the port and the component. A bypass, a mixing valve, a second circuit tying in, an economizer, a bleed, or a section of the same pipe that runs somewhere else first. Walk the physical path from one probe to the other before you commit, and write down what is between them. Skipping this produces the most expensive class of wrong reading: a difference measured across a component plus a bypass, which reads far smaller than the component's true difference and gets diagnosed as poor heat transfer.
Field 2: both raw temperatures. Record 160.2 and 141.6, not "18.6 F split." Six months later the raw pair lets anyone recheck against ambient, against saturation, and against the operating setpoint. The difference alone lets them do none of that, and a difference cannot tell you whether the whole system has drifted.
Field 3: same instrument, both ends. Two different instrument types disagree by more than the difference you are trying to measure on a great many jobs. An infrared thermometer reads a surface and is thrown badly by low emissivity on bright metal; a strap-on thermistor reads pipe metal under whatever insulation you did or did not restore, and note what that instruction actually asks you to do: disturbing thermal system insulation on older hot piping is a presumed-asbestos act, so treat it as PACM until it is proven otherwise, control it under 29 CFR 1910.1001 in general industry or 1926.1101 in construction, and use a NIOSH-approved respirator under a 1910.134 program rather than a dust mask, because the route is inhalation and gloves do not address it; an immersion probe in a well reads the fluid. Mixing them puts the instrument difference straight into your answer. Use one instrument moved between two points, or a matched pair, and note the resolution: a 1 F resolution instrument reporting a 3 F difference is reporting noise.
Field 4: the clock. Two ends taken four minutes apart on a system that is still coming up is two readings of two different states. Let the system run until both ends stop moving, then take them together, and record how long you waited. On equipment with real mass, that wait is minutes rather than seconds; the sibling card on thermal mass gives the way to estimate it. A reading taken during the climb always understates the eventual difference on the heating side and overstates it on the cooling side.
Field 5: the flow at that moment. A temperature difference without a flow figure beside it is uninterpretable, and that is the single largest cause of a confidently wrong conclusion in this whole area. Even an approximate flow, honestly labeled as approximate, is worth far more than none.
Field 6: saturated or not. If either end sits inside a boiling or condensing region, the temperature there is being set by pressure rather than by load, and a difference that straddles the saturation boundary is not a load measurement. Note it and treat those two numbers differently.
The completed record on a live coil
A hydronic heating coil in an air handler, complaint of insufficient heat. Both ends read with the same immersion probe in existing wells, 0.1 F resolution, after 12 minutes of steady operation with no setpoint changes, no bypass in the loop between the wells, and neither end anywhere near saturation.
| Field | Water side | Air side |
|---|---|---|
| Entering | 160.2 F | 62 F |
| Leaving | 141.6 F | 104 F |
| Difference | 18.6 F | 42 F |
| Flow | 6 GPM, from the balancing valve reading | 1,000 CFM, from the fan table at the recorded speed tap |
Now use the two sides as a check on each other. Both should be describing the same heat, and the standard forms for standard conditions are 500 x GPM x delta-T for water and 1.08 x CFM x delta-T for sensible air:
- Water side: 500 x 6 x 18.6 = 55,800 Btu per hr
- Air side: 1.08 x 1,000 x 42 = about 45,400 Btu per hr
Those do not agree. The air side is 45,400 / 55,800 = 81 percent of the water side, a 19 percent gap, and that gap is the finding. The same heat left the water and entered the air, so one of the four inputs is wrong.
Rank the inputs by how much you trust them. The four temperatures came from one instrument, in wells, at 0.1 F resolution, after a settled 12 minutes. They are the most trustworthy numbers on the page. The 6 GPM came from a balancing valve with a published curve. The 1,000 CFM came from a fan table, which assumes an external static pressure nobody measured. The least-certain input is the airflow, and it is the one to test first.
Solve the air side backwards from the water side. If 55,800 Btu per hr is real:
implied CFM = 55,800 / (1.08 x 42) = 55,800 / 45.36 = about 1,230 CFM
That is 23 percent above the 1,000 CFM assumed. The next test is a real airflow measurement, and if it lands near 1,230 CFM the coil is transferring what it should and the complaint is a distribution or load problem, not a coil problem. If a real airflow measurement lands near 1,000 CFM instead, the disagreement is genuine and the water flow figure comes under suspicion next.
What a tech who skipped the heat balance concludes. He records a 42 F air rise, decides the coil is fine, and leaves. Or he records the water side, decides 18.6 F is a small split for a heating coil, and quotes flow work. Both are one-sided readings of a two-sided device, and the check that separates them cost one multiplication per side.
What would change the answer here. Two conditions invert it. First, if the air is picking up or losing moisture across the coil, the sensible-only 1.08 form understates the air-side heat and the two sides legitimately will not balance; on a wet coil you need the enthalpy form instead. Second, if the coil is in an airstream with any leakage, bypass or unmeasured outdoor-air introduction between the two air probes, the air-side numbers describe a different air mass than the one that crossed the coil, and no amount of instrument care fixes that. Field 1 is what catches it.
The error the record cannot catch
Six good fields still will not save a reading taken in a stratified stream. Air in a duct and water in a large header do not mix instantly, and a probe at one insertion depth can read several degrees away from the bulk average, consistently, with no sign anything is wrong. The reading is repeatable, which is exactly why it is convincing.
Two habits catch it:
- Traverse rather than spot-read where the geometry allows. Take readings at several depths or positions across the same plane and use the average. If the spread across that plane is larger than the difference you are trying to measure, a single-point reading was never going to work.
- Read downstream of a mixing feature, not upstream. After an elbow, a fitting, a fan or a bend, the stream is better mixed. Immediately downstream of a tee where two temperatures join, it is not mixed at all, and that location produces the most convincing wrong numbers in the trade.
The tell is a heat balance that misses by a consistent amount in the same direction every time you check it, on a system where the flow figures have already been verified. That pattern is a probe-placement problem, not an equipment problem.
References
- ASHRAE Handbook Fundamentals, sensible heat rate relationships and measurement practice
- 29 CFR 1910.147, control of hazardous energy including stored pressure and hot fluid
- 29 CFR 1910.333(b)(2), de-energized electrical work; NFPA 70E-2021, 120.5, live-dead-live verification
- 29 CFR 1910.132(d)(1), PPE hazard assessment for hot-surface contact
- See related: What a Delta-T Is Actually Telling You; The Thermal Mass That Explains a Slow Response