Thermistor drift: why the display and the cabinet disagree

What the sensor is, and why it drifts instead of failing

The thermistors used in refrigeration are negative-temperature-coefficient devices, which means resistance falls as temperature rises: cold reads high in ohms, warm reads low. The control measures that resistance, looks it up against a curve stored in firmware, and calls the answer a temperature. A hard failure is the easy case — a sensor gone open circuit or short circuit produces a value that sits nowhere on the curve, and most controls notice and complain about it. Drift is the difficult case. The bead ages, moisture works past the seal over years, a lead gets stressed where it leaves the body, and the curve shifts by a few percent. A few percent is several degrees at the cabinet, and it is still an entirely plausible number, so nothing in the machine objects to it.

• Cold reads high resistance, warm reads low. That one relationship explains every symptom below.
• A sensor reporting a plausible but wrong number will never trigger a fault indication. Silence from the control is not evidence that the sensors are good.
• Drift is gradual. Owners nearly always describe a compartment that has been slowly getting worse for months, not one that changed overnight.

The connector lies more often than the sensor does

Before condemning a thermistor, condemn the path to it. Anything that adds resistance in series with the sensor — a corroded pin, a green terminal, a chafed lead making a poor splice — makes the total look higher, and higher means colder to the control. It therefore believes the compartment is colder than it is, runs the system less than it should, and leaves you with a box sitting warm. Anything that adds a parallel path — moisture bridging two pins in a connector, a damp harness in a compartment that has a drain problem — makes the total look lower, and lower means warmer. The control then over-cools, and you get produce freezing in a refrigerator that is reporting a perfectly ordinary number on the front. Coastal kitchens produce the second version more often than most, simply because there is more moisture in the cabinet to begin with.

• A corroded or loose connection reads colder than reality, so the compartment runs warm.
• Moisture or leakage across a connector reads warmer than reality, so the compartment over-cools and freezes food.
• Either fault is cured by cleaning and reseating a connection rather than by a new sensor, which is exactly why the harness gets measured before a part gets ordered.

The two tests that settle it

One of these you can run tonight and one belongs to a technician, and between them they separate a sensor problem from everything else. Yours: stand a jar of water at mid-height in the compartment, drop a thermometer into it, leave it a full day, and go on using the door exactly as you normally would. Water carries enough thermal mass to ignore door openings, so what it reports is the temperature the food has been living at rather than the air at the instant you looked. In the freezer, bury the probe in the middle of the frozen load instead. Then set both against what the panel claims. The technician test is a resistance measurement: the sensor disconnected from the control, an ohmmeter across it at a known temperature, and the reading compared against the manufacturer resistance table for that specific part. An ice-water slurry — ice and water stirred together, not ice on its own — gives a reliable 32 degree Fahrenheit reference in any kitchen.

• Measure at the sensor, then measure again at the control connector. The difference between the two readings is the harness, and that comparison is what finds a corroded pin.
• Compare against the published resistance table for that part, not a general chart. Curves differ between sensors.
• A reading taken while the compartment is still swinging is worthless. Let the temperature settle before the meter comes out.

What drift looks like on each of the sensors

A built-in carries more than one thermistor, and which one has drifted decides what you actually see. An air sensor in a compartment produces the classic complaint: a stable cabinet at the wrong temperature, milk turning early or lettuce freezing, and a display that looks fine throughout. A sensor on or near the evaporator does something far less obvious, because it takes part in the defrost decision. Reading warm — reporting a higher temperature than the coil is actually at — it lets the cycle end before all the ice has cleared; the leftover ice seeds the next round and cooling fades week by week while the number on the front stays confident. Reading cold, it tells the control the coil is colder than it is, so the heater stays on longer than it needs to, which puts melt water where it does not belong and produces a warm spike after every cycle.

• Compartment air sensor: stable temperature, wrong number, no fault indication, and food behaving badly.
• Evaporator sensor reading warm: frost that returns within a couple of weeks of being cleared, and cooling that fades rather than stops.
• Evaporator sensor reading cold: a warm spike after each defrost, sometimes with water in the compartment or on the floor.
• Both compartments wrong at the same time is rarely a sensor. Look at the condenser, the airflow and the heat load first.

What drift is not

This is where money gets wasted, because a thermistor is cheap enough to replace on a hunch and cheap enough for that hunch to survive being wrong. Sort by behavior rather than by number. A compartment that swings — cold, then warm, then cold again — is an airflow, door or condenser problem rather than a sensor problem, because a drifted sensor is wrong consistently rather than intermittently. A compartment that is warm while the coil behind the rear panel is a block of ice is a defrost circuit fault, and the sensor is only one of three suspects in it. A compartment that is warm and silent, with no air movement at the vents, is an evaporator fan. A cabinet where both sides have climbed together starts at the condenser and the grille. Only the stable, confident, consistently wrong number points at a sensor first.

• Swinging temperature: door, gasket, airflow or condenser.
• Warm with an iced coil: the defrost circuit — heater, terminating switch or sensor, in that order of frequency.
• Warm with no air movement inside: evaporator fan motor, or a blade frozen against its shroud.
• Stable at the wrong number, confirmed by a thermometer left in water: the sensor and its connections.

Replacing one, and the part of it that is not a homeowner job

Some sensors clip into an air path and are reachable. Others sit against the evaporator behind the rear panel inside the compartment, which means removing a panel in front of a coil that may be iced and always has fins sharp enough to cut, with the harness routed through foam behind it. That is technician work, and so is anything that involves cutting into a harness rather than unplugging a connector. Two rules matter more than the mechanics of the swap. First, fit the manufacturer part: the resistance curve is the component, and a physically identical thermistor with a different curve leaves you with an appliance that is wrong in a new and less predictable way. Second, never compensate for drift by turning the setting colder. That moves the error rather than removing it: the compartment now runs longer cycles than it was built to, frosts its coil faster, and turns whatever sits nearest the vents to ice, while the number the control is working from is still the wrong one.

• Cut power at the breaker or unplug the appliance before any panel comes off. The defrost heater in that compartment is a live circuit.
• Use the manufacturer sensor. The curve is the part.
• Do not adjust the setpoint to hide a bad reading. The error will reappear somewhere else in the cabinet.
• If a sensor was replaced and the same fault returned within months, look at moisture and the connector rather than at the new part.