How Humidity Changes What You Measure

Why this matters

A shop condemned a motor in August on a 42 megohm insulation reading, ordered a replacement, and the customer deferred. In February the same tech read 400 megohms on the same motor and quietly cancelled the order. Nobody was incompetent and neither reading was wrong. The winding did not improve; the air did. If you cannot say which of your readings move with the weather and which do not, you will replace parts that were fine and pass parts that were failing, and you will never know which one you just did.

Before you take any of these readings

An insulation resistance test applies 500 to 1,000 volts direct current to a winding. This is electrical work on a circuit, so the isolation standard is 29 CFR 1910.333(b)(2), not the mechanical lockout standard: open the disconnect, lock and tag it, and prove the conductors dead with the live-dead-live sequence at NFPA 70E-2021, 120.5, checking your meter on a known live source before and after. Disconnect the motor leads from the starter so you are testing the winding and not the whole circuit, and disconnect any drive, surge capacitor, or winding thermistor before applying test voltage, because the test voltage will damage them.

When the test ends, the winding holds a capacitive charge that can put you on the floor. Leave the tester connected and let its discharge circuit run, then verify zero volts across the leads with a meter before you touch a conductor.

Where a reading genuinely has to be taken with the equipment energized, such as a flame rectification current on a running control circuit, that exposed live work is permitted under the troubleshooting exception at 29 CFR 1910.333(a)(1), which covers testing where de-energizing would introduce additional hazards or is infeasible. Use a meter and test leads rated for the category of the circuit, CAT III for branch-circuit and distribution work, and inspect the leads before every use. An under-rated meter across a line-voltage fault is the failure that kills.

Start with what humidity does not touch

Most of your instruments are indifferent to it, and knowing which ones frees you to stop hedging every number.

A thermocouple or RTD reading. The junction reports the temperature it is at. Humidity changes nothing about that, with one exception worth naming: a junction or extension lead that is actually wet introduces a shunt path and, on a thermocouple, a second junction where dissimilar metals meet in water. A dry probe is a dry probe.

A pressure or draft reading. A manometer reads a pressure difference. Water vapor's partial pressure is already part of the total pressure on both sides, so it cancels.

The resistance of the conductor itself. Copper resistance is a function of temperature, not of the air around it. A winding resistance test that reads 2 percent high is telling you about temperature or about a connection, never about the weather.

A combustion analyzer's oxygen reading. These instruments condition and dry the sample, so what comes out is on a dry basis and ambient humidity has already been removed from it. What can move is the analyzer's own behavior if its water trap or filter saturates, which is a maintenance problem rather than a physics one.

Air density, mostly. Humid air is genuinely less dense than dry air at the same temperature and pressure, because a water molecule is lighter than the nitrogen and oxygen it displaces. The magnitude is small: hot saturated air runs on the order of 2 percent below dry air at the same temperature. That is real, and it is a fraction of the error that altitude and temperature introduce into the same calculation, so correct for those two first and treat humidity as the third-order term it is.

The three routes by which humidity does get in

Route one: a conductive film on a surface. This is the big one and it drives most of the surprises. A microscopically thin layer of adsorbed water on an insulator surface, especially over dust, salts, or oil residue, is a parallel conduction path. It does not degrade the insulation; it bypasses it. Every reading that depends on very small currents through a very high resistance is exposed: insulation resistance, leakage current, flame rectification signal, control board tracking, high-voltage ignition components.

Route two: evaporation from a wet surface. A wet surface is not at air temperature. Evaporation cools it toward the wet bulb, so an infrared thermometer aimed at a sweating duct or a damp slab reads the film, not the material, and reads it low. This is also why a surface that has just been cleaned reads cold for several minutes.

Route three: the vapor is part of what you are measuring. Anything reported as a fraction of the air, or computed with a constant that assumes a standard mixture, has water vapor in it. Airflow converted from velocity pressure, and any measurement of the air's heat content, both belong here.

The case: a motor that degraded and then recovered

August visit. Belt-drive fan motor, insulation resistance test at 500 volts direct current, one-minute reading 42 megohms. Winding temperature measured at 30 C by contact probe on the frame after a cool-down. Outdoor conditions were around 78 F and 70 percent relative humidity, and the mechanical room was open to a damp crawl space. The reading was written down as a bare 42 megohms with no temperature and no humidity note.

February visit, same motor, same tester, same technician. One-minute reading 400 megohms, winding at 20 C, dry heating-season air in the same room.

A factor of nearly ten between two readings on a machine that ran continuously in between. If you take either number at face value you get a different repair recommendation, and both recommendations are wrong.

Correcting both ends before you compare

Insulation resistance falls sharply with temperature, so no two readings at different winding temperatures are comparable until both are brought to a common reference. IEEE 43 uses 40 C as that reference and provides a correction curve; the field approximation is that insulation resistance roughly halves for every 10 C rise, which is close enough for a comparison and not close enough for an acceptance decision.

• August: 42 megohms at 30 C, corrected up one 10 C step to 40 C, so divide by 2. About 21 megohms at 40 C.
• February: 400 megohms at 20 C, corrected up two 10 C steps to 40 C, so divide by 4. About 100 megohms at 40 C.

Both figures are now on the same basis, which is the point of doing it to both. Comparing the corrected August number against the uncorrected February number would have produced a factor of 19 and a much more alarming story out of the same two measurements.

Corrected against corrected, the ratio is 100 divided by 21, about 4.8. Temperature accounted for roughly half of the apparent ten-fold swing. The other factor of about five is still sitting there unexplained, and it is route one: a conductive film across the winding end turns and the terminal box surfaces in a damp room, shunting the test current around the insulation.

The reading that did not move

The reason to know this is that there is a measurement immune to the film, and it was taken on both visits.

The polarization index is the ten-minute reading divided by the one-minute reading on the same test. Both halves are taken at the same temperature, in the same room, through the same surface film, so the film's contribution largely divides out. It is a ratio describing how the insulation's absorption current decays, which is a property of the insulation itself.

• August: 95 megohms at ten minutes over 42 at one minute, index about 2.26.
• February: 900 over 400, index about 2.25.

Two visits, two very different absolute readings, and an index that did not move. That is the winding telling you it is unchanged. The spot value was tracking the weather; the ratio was tracking the insulation.

One caveat that belongs with the number rather than three paragraphs away: IEEE 43 notes the polarization index loses its meaning when the one-minute reading is already very high, because at that point you are dividing two large numbers dominated by instrument and surface effects. Where the one-minute value is high, the standard directs you to judge on the one-minute value itself rather than on the index.

Where the same three routes bite elsewhere

Flame rectification. The flame signal is measured in microamps and the return path runs through a ceramic insulator on the flame rod. A damp, dusty insulator leaks that tiny current to ground. The signature is the giveaway: intermittent lockouts on humid mornings, clearing by mid-morning, with a clean flame and a clean rod. The fault is on the insulator, not the rod.

Airflow from velocity pressure. The conversion uses a constant that assumes standard air. Correct for temperature and elevation first, since those dominate; add the humidity term only where you are working in hot saturated conditions and the second decimal place matters.

Infrared surface readings. Aim at a dry, non-reflective spot, or accept that you are reading a water film cooled by evaporation. If the surface has to be wet, use a contact probe.

Relative humidity sensors themselves. Relative humidity is defined against saturation at the sensor's temperature, so a probe that is warmer or colder than the air it is sampling reports a value for a temperature the air is not at. Give the probe time to equilibrate in a duct before believing it, and use dew point rather than relative humidity when you need a number you can carry to another location.

What to write down so next season's reading is comparable

The whole case above was created by a ticket that recorded one number. Three extra fields would have prevented it, and they take under a minute.

For any high-resistance measurement, record the reading, the winding or component temperature, the test voltage, and the ambient conditions in the room, plus the ten-minute value if you took one. For any air-side measurement, record the dry bulb, the elevation, and either the relative humidity with its temperature or, better, the dew point.

Then verify the record does its job by using it: before you condemn anything on a trend, pull the prior reading, correct both to the same reference, and confirm you are comparing corrected to corrected. If the earlier ticket does not carry the fields you need to correct it, you do not have a trend. You have two unrelated numbers, and the honest move is to say so and take a fresh baseline rather than build a recommendation on them.

References

• IEEE 43, Recommended Practice for Testing Insulation Resistance of Electric Machinery, for the 40 C reference, temperature correction, and the polarization index and its limits
• 29 CFR 1910.333(b)(2), OSHA general industry, for de-energizing and verifying circuits before electrical work
• 29 CFR 1910.333(a)(1), OSHA general industry, for the troubleshooting exception permitting energized testing
• NFPA 70E-2021, 120.5, for the live-dead-live proving sequence with a meter of appropriate rating
• See related: Moisture Meters and What Their Readings Actually Mean; What Dew Point Predicts; How to Read a Flue Gas Result