Module 5 of 16
Sensors & Feedback Devices
Thermistors, thermocouples, pressure, flow, and position feedback
Learning objectives
- Identify sensor type from circuit, connector, and expected output
- Convert a thermistor resistance to a temperature and compare it against reality
- Test thermocouples, pressure transducers, flow meters, and hall-effect sensors
- Recognise when a fault code is describing a sensor problem rather than the condition it names
- Predict how a control board will behave when a given sensor drifts rather than fails outright
- Test hall-effect, optical, pressure and position sensors with the correct method for each type
- Distinguish a bad sensor from a bad reference voltage, a bad harness, or a correct sensor reporting a real fault
A machine only knows what its sensors tell it
Every modern appliance decision — run the compressor, start defrost, stop filling, unlock the door — comes from a sensed value compared against a setpoint. When the sensed value is wrong, the machine behaves perfectly logically and completely incorrectly. That is why 'the board is crazy' calls are so often a two-dollar thermistor with a corroded pin.
The technician's job with any sensor is the same three-step check: is the sensor reporting a plausible value, does that value match an independent measurement, and does the control see the same value the sensor is producing?
- Plausible: within the sensor's physical range and not pinned at either extreme
- Accurate: matches your own thermometer, gauge, or observation
- Delivered: the control reads the same value the sensor outputs — the harness in between is a suspect
Thermistors and RTDs
Most appliance temperature sensing uses an NTC thermistor: resistance falls as temperature rises. Common curves are 10k at 25 °C in refrigeration and 50k or 100k in some laundry and cooking applications, but the curve is model specific — always use the manufacturer's table rather than a remembered number.
Diagnose by comparing three things: the measured resistance, the temperature that resistance implies, and the actual temperature at the sensor bulb. A thermistor reading 10k in a 2 °C compartment is telling the board the space is at 25 °C, and the board will happily run the compressor forever or not at all.
- NTC: resistance down as temperature up. PTC and RTD behave the opposite way
- Open thermistor reads OL and usually drives the machine to a default or fault mode
- Shorted or water-intruded thermistor reads far low, implying a high temperature
Test at two temperatures where possible — ice water and body heat — so you prove the curve, not one point.
Thermocouples and flame sensing
A thermocouple generates a small DC voltage from a temperature difference between its hot junction and the cold end — typically 25 to 30 millivolts open circuit on a gas pilot assembly, dropping to perhaps 12 to 18 mV under load in the gas valve's magnet. Measure it in millivolts DC; an ohm reading tells you nothing useful.
Flame rectification on modern gas equipment is different again: the control sends AC through the flame to ground and reads the DC microamp current the flame rectifies. A few microamps below the control's threshold and the board shuts down for safety, exactly as it should. Dirty flame rods, poor ground, and a lazy flame are the usual causes — not the board.
- Thermocouple: 25–30 mV open, 12–18 mV under load; below that, the valve drops out
- Flame rectification: typically 1–10 µA DC; below the control threshold means a cleaning or ground problem
- The burner ground path is part of the sensing circuit — a rusty burner bracket kills flame signal
Pressure, flow, and position feedback
Pressure transducers output a voltage or frequency proportional to pressure and have replaced simple pressure switches in many machines. They fail slowly and quietly, drifting rather than dying, which produces intermittent complaints. Compare their reported value against a gauge before assuming the system is at fault.
Flow meters are usually a small turbine with a hall-effect pickup producing pulses; the control counts pulses to measure water volume. A sticky turbine underreports flow, so the machine keeps filling. Position feedback — lid switches, door latches, hall sensors on a drum or motor — is mostly digital, and the useful test is watching the state change as you actuate the device.
- Transducer: verify supply voltage (often 5 V) before condemning the output
- Flow meter: pulses should appear on a scope or a frequency-reading meter during fill
- Hall sensors: expect a clean high/low transition, not a slow ramp
Four signal families, four test methods
Almost every sensor in modern appliances and light commercial equipment falls into one of four signal families, and the family — not the part number — decides how you test it. Resistive sensors (thermistors, RTDs, some level floats) change resistance with the measured quantity, and the board reads them as a voltage divider against an internal pull-up. Ratiometric analogue sensors (most pressure transducers, some position sensors) are powered by a regulated 5 V reference and return a voltage that is a fixed fraction of that reference. Digital pulse sensors (hall-effect tachs, flow meters, optical encoders) return a switching signal whose frequency or count carries the information. Switched sensors (float switches, reed switches, lid and door interlocks) simply open or close.
The consequence is practical: a resistance reading proves a thermistor and proves nothing about a transducer. A transducer that reads open on ohms may be perfectly good. Identify the family from the wiring — two wires and no supply means resistive or switched; three wires with a 5 V rail means ratiometric or hall-effect — before you pick up a meter.
- Two wires, no supply: resistive or switched — measure resistance, compare against the chart at a known temperature
- Three wires with 5 V and ground: ratiometric or hall — measure signal voltage relative to the actual reference, not to a fixed number
- Frequency-based: read Hz with a meter that supports it, or watch the value the control reports while you drive the sensor
Never inject supply voltage into a sensor circuit to 'wake it up'. A 5 V input on a board rail tolerates almost no overvoltage, and the board, not the sensor, is what fails.
Drift is harder than failure, and more common
An open or shorted sensor is easy: the board sees an impossible value, throws a code, and defaults to a safe mode. Drift produces no code at all. A thermistor that reads four degrees warm keeps the compressor running longer than it should, produces frost patterns nobody can explain, and looks fine on a resistance check unless you compare it against a real reference at a real temperature.
That is why the professional method is comparison, not inspection. Read the sensor with the machine at a known condition, compare against a calibrated thermometer or gauge, and compare sensors against each other. Two identical thermistors in the same compartment should agree within a degree or two; when one disagrees, you have your answer without any chart at all.
Ratiometric sensors drift differently. A pressure transducer with a contaminated or slowly leaking port reports steady values that are simply wrong, and the control obeys them exactly. Confirm the pressure independently before condemning anything downstream of it.
- Compare against a calibrated reference at a known condition, not against memory
- Compare like sensors against each other in the same compartment or circuit
- A drifting sensor produces symptoms with no fault code — treat unexplained behaviour as a calibration question
When the sensor is innocent
A large share of sensors replaced in the field were good. Three things impersonate a bad sensor. First, a sagging reference rail: if the board's 5 V reference is at 4.3 V, every ratiometric sensor reads low simultaneously — the tell is that several unrelated sensors are wrong at once. Second, harness resistance: a corroded pin adds ohms in series with a resistive sensor, which the board reads as a colder temperature; wiggle-testing while watching the reported value exposes it. Third, a correct sensor reporting a real fault, which is the outcome nobody wants to accept when the customer wants a part changed.
The discipline is to prove the signal path before condemning the transducer at the end of it. Measure the reference, measure at the board connector as well as at the sensor, and only then decide.
If several sensors read wrong in the same direction at the same time, suspect the board's reference supply or a shared ground, not the sensors.
Failure modes and what confirms them
| Symptom | Mechanism | The tell |
|---|---|---|
| Compartment far colder or warmer than setpoint | Drifted or shorted temperature sensor | Measured resistance implies a temperature that does not match a thermometer at the bulb |
| Sensor fault code that clears when the harness is wiggled | Corroded pin or chafed lead, not the sensor | Resistance measured at the board differs from resistance at the sensor |
| Gas burner lights then drops out after a few seconds | Weak flame signal or failed thermocouple | Microamps below control threshold, or thermocouple under 12 mV loaded |
| Machine overfills or underfills with no valve fault | Flow meter turbine sticking or scaled | Pulse count far below the flow measured by a bucket test |
| Intermittent cycling for no apparent reason | Pressure transducer drifting out of calibration | Reported pressure disagrees with a gauge on the same port |
| Several unrelated sensors read wrong in the same direction | Sagging 5 V reference or a degraded shared ground on the board | Measured reference below specification at the sensor connector while sensors themselves check out |
| Temperature complaint with no stored fault code | Thermistor drift within the range the board considers valid | Sensor resistance disagrees with a calibrated reference at the same measured condition |
| Analogue level or pressure value plateaus mid-range | Partially blocked air dome, kinked sensing hose, or a fouled transducer port | Signal voltage stops changing while the physical quantity keeps rising |
| Intermittent speed or position fault under load only | Hall-effect dropout from gap, contamination, or a loose connector pin | Missing pulses during slow rotation with balanced motor windings |
| Optical sensor reports a permanent blocked or clear state | Contaminated lens or emitter degradation rather than circuit failure | State changes correctly only after cleaning, or never changes when the path is interrupted |
Test procedures
Manufacturer differences
Samsung
Heavy use of NTC thermistors on a common harness plug, with the main board reporting per-sensor values through service mode rather than through individual fault codes.
What it changes: Enter service mode and read the board's own reported temperatures first; a value that is plausible but wrong points to drift, while a rail value points to open or short.
LG
Inverter and sealed-system control depends on multiple thermistors plus a hall-effect rotor position feedback on the compressor and on direct-drive washer motors.
What it changes: Rotor position faults present as start failures or rough operation, not as temperature complaints. Test the hall sensor by frequency or by service-mode feedback, never by resistance across the signal pin.
Whirlpool
Mixed generations: older resistive sensing and mechanical pressure switches alongside newer analogue pressure transducers on washers and dishwashers.
What it changes: Confirm which generation you are on before testing. A pressure switch is a continuity test; a transducer is a supply-plus-signal test and will read open on ohms while being perfectly serviceable.
GE
Board-centred architectures with sensor values available over the internal service bus and, on many platforms, through a diagnostic display sequence.
What it changes: Use the reported values as your first instrument. Comparing the board's value against your own reference immediately separates a sensor problem from a control interpretation problem.
Bosch
Precision-oriented sensing with tight tolerance bands, including NTC sensing in dishwashers and heat-pump dryers where the control acts on small differences.
What it changes: Small drift causes real symptoms here. Two or three degrees of error is enough to change cycle behaviour, so hold these sensors to a tighter standard than a domestic refrigerator thermistor.
Commercial refrigeration controllers
Programmable electronic controllers with configurable probe types, offsets, and alarm bands; probes are field-replaceable and often extended with added cable.
What it changes: Check the configured probe type and any programmed offset before trusting a displayed value, and account for added cable resistance on long probe runs.
Test method by sensor family
Pick the method from the signal family. The wrong method produces a confident wrong answer.
| Sensor type | Signal | Correct test | Good result | Suspect result |
|---|---|---|---|---|
| NTC thermistor | Resistance falls as temperature rises | Resistance at a known temperature, compared to the chart | Within chart tolerance and tracks as you warm it in your hand | Open, shorted, or plausible but several degrees off a calibrated reference |
| RTD / PTC probe | Resistance rises with temperature | Resistance at a known temperature | Tracks smoothly and repeatably | Jumps or drifts while undisturbed |
| Pressure transducer | Ratiometric voltage on a 5 V reference | Measure reference, then signal, with power applied | Signal sits in the documented fraction of the measured reference | Signal pinned near 0 V or near the rail, or steady but contradicted by gauges |
| Hall-effect tach or rotor sensor | Digital pulse train | Frequency while rotating, or the control's reported speed | Frequency rises and falls smoothly with speed | No pulses, intermittent pulses, or output stuck high or low |
| Optical sensor | Switched or pulsed light path | Block and unblock the path while watching the signal | Clean state change every time | Sluggish change, or no change until the lens is cleaned |
| Level sensor (analogue) | Frequency or voltage versus water level | Compare reported level against actual with a controlled fill | Reported level tracks actual through the fill | Plateaus mid-fill, or offset consistently — check the air trap and hose first |
| Position or door switch | Open or closed contact | Continuity through the actual travel of the mechanism | Clean make and break at the correct point of travel | Makes only when pressed harder than the mechanism can press it |
Safety and professional boundaries
- Sensor circuits are low voltage, but they sit inches from line voltage — verify what you are probing before you probe it.
- Do not backfeed voltage into a sensor input to test it; board inputs are not protected against it.
- Sealed-system sensors on commercial equipment may require refrigerant handling certification to access — stop where your certification stops.
Two-point thermistor test
- 1.Isolate the machine and disconnect the sensor at its connector.
- 2.Measure resistance and note the ambient temperature at the sensor.
- 3.Place the bulb in ice water (0 °C) and record the reading once stable.
- 4.Warm the bulb in your hand (about 32–35 °C) and record again.
- 5.Compare both readings against the manufacturer's resistance table; a curve that is out at either end condemns the sensor.
Prove the harness, not just the sensor
- 1.Measure the sensor's resistance at the sensor connector.
- 2.Measure the same circuit at the control board connector with the sensor still attached.
- 3.A meaningful difference is harness resistance — a corroded pin, a chafed lead, or a poor crimp.
- 4.Flex and wiggle the harness while watching the reading to catch intermittents.
Measure flame rectification current
- 1.Set the meter to DC microamps and place it in series with the flame rod lead.
- 2.Start the burner and read the current once the flame is established.
- 3.Compare against the control's minimum — typically 1 µA or higher, with healthy systems at 3–10 µA.
- 4.If low, clean the flame rod with a non-abrasive pad, confirm the burner ground, and re-measure before condemning the control.
This is a live gas test. Confirm ventilation, keep your face clear of the burner, and be ready to shut the gas off.
Test a ratiometric sensor correctly
- 1.Identify the supply, ground and signal pins from the wiring diagram before probing.
- 2.With power applied, measure the reference voltage between supply and ground at the sensor connector, not at the board.
- 3.Measure signal to ground and express it as a fraction of the measured reference.
- 4.Change the physical quantity — fill, pressurise, or move the mechanism — and confirm the signal tracks smoothly.
- 5.If the signal is flat, remove the sensing hose or port and apply the quantity directly to separate a dead sensor from a blocked path.
Live low-voltage work: keep probes off adjacent pins, use insulated back-probes where available, and confirm the circuit is low voltage before you touch it.
Prove a temperature sensor by comparison
- 1.Read the value the control reports through service mode and record it.
- 2.Measure the same location with a calibrated thermometer placed as close to the sensor as possible.
- 3.Measure the sensor's resistance and compare against the chart at your measured temperature.
- 4.Compare against a like sensor in the same circuit or compartment where one exists.
- 5.Condemn the sensor only when the reported value, the resistance, and the reference disagree consistently.
Expected readings and what they mean
| Measurement | Expected | Meaning |
|---|---|---|
| 10k NTC at 25 °C | ≈10,000 ohms | Baseline; roughly 16k at 10 °C and 28k at 0 °C on a common curve |
| Thermocouple, open circuit | 25–30 mVDC | Below 20 mV, the gas valve magnet will not hold |
| Flame rectification current | 3–10 µA DC | Below the control threshold the board locks out — usually cleanliness or ground |
| Pressure transducer supply | 5.0 VDC ±0.25 | Bad supply makes a good sensor report nonsense |
| Flow meter output | Steady pulse train during fill | No pulses with water flowing means a stuck turbine or dead pickup |
| Board 5 V sensor reference at the connector | 4.9 to 5.1 V, steady | Low or unstable reference makes every ratiometric sensor read wrong at once — fix the rail first |
| Ratiometric signal at rest | Documented fraction of the measured reference, typically well clear of both rails | Pinned near 0 V or near the reference indicates sensor or wiring failure, not a real extreme value |
| Hall-effect output while rotating | Clean pulse train, frequency proportional to speed | Dropouts under slow rotation explain faults that only appear under load |
| Series resistance added by harness and connectors | Negligible compared with sensor resistance | Corroded pins add ohms the board reads as a colder temperature |
Field scenarios
Mini-scenario: refrigerator with no code and wrong temperatures
- French-door refrigerator, fresh food section running 8 degrees warmer than setpoint
- No fault codes stored, compressor and condenser fan both running
- Evaporator has a full, even frost pattern; defrost cycles appear normal
- Service mode reports the fresh food sensor at setpoint
The board says the compartment is at setpoint and your thermometer says it is not. What do you do next?
Mini-scenario: washer overfilling, transducer reads open
- Front-load washer overfills and the cycle eventually faults on level
- Three-wire analogue pressure sensor on the tub air dome
- Resistance across the signal and ground pins reads open
- Air dome hose is connected and looks intact
Ohms across the sensor read open. What does that tell you, and what is the next test?
Mini-scenario: direct-drive washer won't start under load
- Direct-drive washer starts and spins an empty tub, stalls with a full load
- Motor windings measure balanced and within spec
- Intermittent motor or rotor position fault stored
- Rotor magnets and stator look clean, bolt torque is correct
Windings are good and the fault names rotor position. What is the next correct test?
The freezer that says it is warm
- A reach-in freezer displays a high-temperature alarm and runs the compressor continuously, but the product is frozen solid and your own thermometer reads -20 °C in the box.
- The evaporator sensor measures 9.8k ohms at the connector.
What is the fault?
The sensor is reporting roughly 25 °C to the control while the box is at -20 °C, so the control never stops calling. A 10k NTC should read tens of thousands of ohms at freezer temperature — this one is shorted or water-intruded. Replace the sensor and verify the new reading against your thermometer before you leave.
Takeaway: When the machine's story and the physical reality disagree, suspect the sensor that tells the story.
Knowledge check
An NTC thermistor in a 0 °C compartment measures 10k ohms on a 10k-at-25 °C curve. What does that tell you?
A gas burner lights and then drops out after four seconds. Flame current measures 0.4 µA. What do you do first?
Sensor resistance measures correctly at the sensor but wrong at the board connector. What is the fault?
Three analogue sensors on the same board all read about 12 percent low. Nothing else has changed. What is the most defensible next step?
A customer insists a previous technician replaced the same thermistor twice. What should shape your approach on arrival?
Key takeaways
- A machine's behaviour is only as sane as its sensed values.
- Test thermistors at two temperatures to prove the curve, not one point.
- Measure at the sensor and at the board — the difference is the harness.
- Flame signal problems are cleanliness and ground problems far more often than board problems.
- The signal family decides the test method; the part number does not.
- Drift causes symptoms without codes, and comparison against a reference is the only way to see it.
- Prove the reference and the harness before condemning the sensor at the end of them.
Finished this module?
Mark it complete to track your progress toward the Technical Systems Certificate.