How to Read a Surface Temperature Honestly

Why this matters

A surface temperature is the easiest reading in the trades to take and one of the easiest to take wrongly, because nothing about a wrong one looks wrong. The instrument answers instantly, to a tenth of a degree, whether or not it was pointed at the thing you meant, coupled to it, settled on it, or comparable to anything. Techs condemn bearings, clear failing connections, and argue with each other over numbers that were never measurements in the first place.

The steps below are ordered by what you lose if you skip each one, worst first. The first four are ordered by how far each can move today's number, from total to a few degrees. The last two are last because they cost you something other than today's number, not because they are optional.

Before any of it: what the surface can do to you

Each of these attaches to the specific action it belongs to, because the hazard changes with the step, not with the article.

Reaching a probe toward a rotating machine. Do not put a hand, a probe, a lead or a sleeve inside the plane of a guard while a shaft, coupling or fan is turning. Take the reading non-contact from outside the guard, or stop the equipment and isolate it under 29 CFR 1910.147 for the mechanical and stored energy before contact.

Touching anything hot. A surface hot enough to be interesting is hot enough to burn through a glove on contact. Use a probe with an insulated handle, clamp or magnet-mount it, and let go of it while it settles rather than holding it in place.

Opening an enclosure to reach a surface. Live parts exposed means energized electrical work under 29 CFR 1910.333(b)(2). Establish an electrically safe working condition where the task permits it, and prove any companion meter live-dead-live per NFPA 70E-2021, 120.5.

Removing insulation to reach the pipe underneath. This is the inhalation route that gets skipped. Pipe and equipment insulation of unknown age, especially in buildings that predate the 1980s, may contain asbestos, and cutting, tearing or abrading it is how the fibers get airborne. Do not cut or abrade unknown insulation to reach a surface. Treat it as presumed asbestos-containing until it is sampled, use an existing access point or measure elsewhere on the same run, and where the material has to be disturbed, that is regulated work under 29 CFR 1910.1001 in general industry or 29 CFR 1926.1101 in construction, with respiratory protection and containment. Gloves and glasses do nothing for an airborne fiber.

Opening a pressurized or hot fluid path to insert a probe. Do not. Use an existing thermowell, or strap the probe to the outside. Where a plug genuinely has to come out, relieve pressure and confirm zero at a gauge first, and remember a hot closed system holds both pressure and scald energy.

Step 1: name the quantity you actually need

What you lose by skipping it: the entire reading. Every later refinement is applied to the wrong thing.

Decide before you aim: are you after the surface itself, or is the surface a proxy for something inside? The outside of an insulated line is not the fluid. A bearing housing is not the bearing, and runs cooler than it. A cabinet wall is not the component behind it. A pipe's top is not its bottom on a partially filled line.

If the surface is a proxy, say so out loud in the record, because the offset between the proxy and the real quantity is not fixed. It moves with load, with insulation condition, with ambient, and with airflow across the surface.

Then pick the route. Non-contact reads what the optics saw, over an area, subject to emissivity and reflection. Contact reads what the probe reached thermal equilibrium with, at a point, subject to coupling. Neither is better; they fail differently, which is why the two together are worth more than either twice.

Step 2: fix the emissivity or fix the coupling

What you lose by skipping it: tens of degrees, growing with how far the target is from the room.

Non-contact route: bare bright metal reads badly and reads worse the hotter it gets. On a target well above its surroundings, a default emissivity setting on a low-emissivity surface can be off by tens of degrees, and setting emissivity correctly on an instrument with no reflected-background compensation can overshoot by more than the original error undershot. The mechanism and the computed figures live in the sibling article on emissivity; the field action is to put a strip of matte tape or a spot of flat dark paint on the target and read that instead.

Contact route: coupling is the same problem in different clothes. A probe tip touching a rough surface at a point is separated from it by air, and air is a good insulator. Use thermal compound or a thin thermal pad, press the tip flat rather than at an angle, and shield the tip and the first inch of the probe from moving air, which will drag the reading toward ambient the whole time you are waiting.

Step 3: make the target bigger than what the instrument sees

What you lose by skipping it: tens of degrees on a small target, and it always slides toward the background.

Non-contact instruments average whatever fell inside the optical spot. If the target is smaller than the spot, part of your reading is the wall behind it. Aim for the target being at least twice the spot diameter, which almost always means walking closer rather than changing a setting.

Contact probes have their own version: a probe applied to a thin, low-mass surface pulls heat out of it and cools the very spot it is reading. On sheet metal or a small fitting, expect the probe to drag the surface down and expect the drag to be worse with a heavy probe.

Reading through anything at all, mesh guards, glass, plastic film, is not a reduced-quality reading. It is a reading of the mesh or the glass. Find an opening or use contact.

Step 4: let it settle, and watch it settle

What you lose by skipping it: usually single digits to ten degrees, biased toward wherever you started.

Contact probes take time. Log the reading every 15 seconds rather than taking one and walking away, because the shape of the approach tells you whether it settled or you got bored. Settled means two consecutive readings the same.

Non-contact readings appear instantly and still need a settling step of their own: the tape or paint you applied in step 2 has to come to the surface's temperature first, which takes at least a minute on a light surface and longer where a still layer of air sits over it.

Step 5: write down the conditions with the number

What you lose by skipping it: the next visit, not this one.

Six fields, and the number is only one of them: the target and the exact spot, the surface condition, the emissivity setting or coupling method used, the ambient and what surrounds the target, the distance and instrument, and the reading.

The reason this is worth the ninety seconds is that surface temperatures are almost never interpreted alone. They are interpreted against a previous reading, and two readings taken with different instruments at different distances on differently prepared surfaces are not comparable, however carefully each was taken.

Step 6: take a comparison reading on the same visit

What you lose by skipping it: the interpretation.

An absolute surface temperature answers almost nothing. Compare against the same component's opposite side, an identical component in the same service, the same point on a previous visit, or the ambient. Take the comparison with the same instrument, at the same distance, under the same surface preparation, within the same few minutes.

The whole thing on one component

A pump motor with a complaint of noise. The target is the drive-end bearing housing, a painted flat about 2.5 in across.

Step 1. Quantity needed is the housing outer surface, used as a proxy for bearing condition and interpreted against the opposite housing. Noted in the record as a proxy, because the bearing itself runs hotter than the housing it sits in and the gap between them is not a constant.

Step 2. Surface is painted, so emissivity is already high. Instrument set to 0.95. Tape check: taped spot reads 168 degrees F, bare paint immediately beside it reads 166 degrees F. A 2 degree F difference on a target this far above ambient says the paint is behaving as assumed, so no correction is needed and the bare-paint readings are usable across the rest of the machine.

Step 3. Instrument is 12:1. To keep the spot at or under half the 2.5 in target, the spot must be 1.25 in or less, so the distance must be 12 x 1.25 = 15 in or less. Reading taken at 12 in, giving a 1.0 in spot, comfortably inside the target.

Step 4. Contact probe with compound, magnet-mounted, logged every 15 seconds: 158, 164, 166, 166 degrees F. It rose 6 degrees F between the 15 and 30 second marks, 2 more by 45 seconds, and held at 45 and 60. Settled value 166 degrees F, agreeing with the non-contact 166.

A tech who read at 15 seconds and left would have written down 158 degrees F. Hold that number for a moment.

Step 5. Recorded: drive-end bearing housing, top of flat, painted, 0.95 setting, tape-verified, ambient 78 degrees F, 12 in with the shop's handheld, 166 degrees F, contact-confirmed.

Step 6. Opposite-end bearing housing, same instrument, same distance, same painted surface, 142 degrees F. Ambient 78 degrees F.

Now interpret, using rise above ambient rather than raw temperature, because a machine in a hot room reads hot at both ends and the raw numbers hide the asymmetry:

• Drive end rise: 166 - 78 = 88 degrees F.
• Opposite end rise: 142 - 78 = 64 degrees F.
• Ratio: 88 / 64 = 1.375, so the drive end is running about 37.5 percent more rise than the opposite end.

That asymmetry is the finding. Not 166.

And the 158 that a hurried reading would have produced: rise 158 - 78 = 80, against the same 64, gives 80 / 64 = 1.25, or 25 percent more rise. Both comparisons are computed the same way against the same corrected opposite-end figure, so they are directly comparable, and the settling shortcut alone moved the reported asymmetry from 37.5 percent down to 25 percent. On a machine being trended visit over visit, that is the difference between a rising line and a flat one.

What invalidates the comparison

Step 6 assumes the two ends are comparable, and often they are not. Different bearing type or size at each end, a coupling or belt load on one end only, a shaft seal or packing gland generating its own heat, one end sitting in the motor fan's discharge airstream while the other does not, or a hot pipe passing close to one side. Any of those and the two ends are not the same experiment.

When that happens, do not abandon the comparison, change it. Compare the drive end against the same drive end on the last visit, taken with the same six fields recorded. That is what step 5 was buying.

The failure mode this whole sequence exists to prevent

A tech takes one non-contact reading off bare metal at arm's length through a guard, gets a number, compares it against a temperature limit they remember, and makes a call. Every one of the six steps got skipped at once, and the resulting number could be twenty or thirty degrees off in either direction, against a remembered limit that may belong to a different component in a different service.

You will spot it in your own records by what is missing: a reading with no ambient beside it, no comparison point, and no note of surface condition. When you find one in a file, treat it as a note that somebody looked, not as a measurement, and take a real one.

References

• Instrument manufacturer documentation for distance-to-spot ratio, emissivity settings, probe response time, and stated accuracy
• 29 CFR 1910.147 for mechanical isolation and stored energy; 29 CFR 1910.333(b)(2) for energized electrical work; NFPA 70E-2021, 120.5 for live-dead-live proving
• 29 CFR 1910.1001 (general industry) and 29 CFR 1926.1101 (construction) for asbestos-containing insulation work
• See related: What Emissivity Does to an Infrared Reading; The Stratification That Fools a Single Reading