Sensors

Water leak sensors are common in smart homes: near washing machines, under water purifiers, inside sink cabinets, near floor-heating manifolds, balconies, and basements.

They are simple and useful, but they are not leak-location instruments.

Most of them only know whether conductive liquid has reached the probe area.

The first model is: a water leak sensor detects resistance or conductance change between electrodes. The leak alarm is an interpretation that conductive liquid appeared at the probe location.

An electronic scale looks direct: place an object on it and the display shows grams or kilograms.

The sensor does not directly see mass. It senses tiny deformation caused by force, then calibration converts that signal into a weight or mass reading.

The first model is: gravity creates load, load deforms an elastic element, strain gauges turn deformation into resistance change, a Wheatstone bridge converts that into tiny voltage, and calibration maps the voltage to weight.

Current sensing is easy to underestimate. An ADC reads voltage, not current. To measure current, the system must first convert current into something measurable.

The first model is: current must be transformed into voltage, magnetic field, or induced current before electronics can measure it.

Shunt resistor: current -> voltage drop
Hall sensor: current -> magnetic field -> Hall voltage
Current transformer: AC current -> magnetic flux -> induced current

These are not three interchangeable implementations of the same idea. They have different physics and tradeoffs.

Gas sensors can look very certain. A screen may show CO2 800 ppm, TVOC 0.5 mg/m3, formaldehyde 0.08 mg/m3, or “air quality good”.

Many low-cost sensors are not that certain.

The first model is: a gas sensor usually does not count molecules directly. It lets gases affect a sensitive material, electrode reaction, optical path, or ionization process, then interprets the response as concentration, index, or alarm state.

Gas enters sensor
-> Sensitive material, electrode, or optical path responds
-> Resistance, current, or light intensity changes
-> Compensation and calibration
-> Concentration, index, or alarm output

Two distinctions matter:

Hall sensors appear in door contacts, limit switches, speed sensing, brushless motor commutation, magnetic encoders, and current sensors.

They are often thought of as “small switches that trigger when a magnet gets close”. That is usable, but incomplete.

The first model is: a Hall sensor detects magnetic field, not the object itself. Position, rotation, or current can be detected only if they reliably change the magnetic field at the sensor.

The phrase “mmWave radar can detect still people” is easy to misunderstand. It can sound as if a radar naturally knows there is a person sitting still in the room.

That is not the right model.

Radar receives electromagnetic echoes. It observes distance, angle, reflection strength, velocity, phase, and how those values change across frames. Whether those changes mean “a human is present” is an algorithmic interpretation.

The first model is: mmWave presence radar does not detect the concept of a human directly. It observes echo energy, velocity, phase, range bins, angle bins, and multi-frame changes, then classifies some patterns as human presence.

Optical distance sensors are often described as “using light to measure distance”. That is true, but too broad. Different optical sensors can mean very different output semantics.

A simple reflective proximity sensor may only know that reflected light became stronger or weaker. A ToF sensor estimates distance from light travel time or phase. The reported number is not simply “what the sensor sees”; it depends on reflectance, geometry, ambient light, optics, and algorithms.

PIR sensors are everywhere in smart homes: hallway lights, night lights, security triggers, bathrooms, cabinets, and battery-powered motion sensors.

They are cheap, mature, and low power. They are also often misunderstood.

A PIR sensor is not a camera. It does not know whether an object is a person. It is also not a mmWave radar that can observe tiny body motion over time.

The first model is: a PIR sensor detects changes in infrared heat radiation across its field of view. It is good at detecting human motion, not at proving continuous still presence.

A pressure sensor seems to output only one thing: pressure. In engineering, that pressure is often used to infer liquid level, altitude, filter blockage, airflow, or tank state.

The first model is: a pressure sensor measures force per unit area. Liquid level, altitude, and flow-related states are model-based interpretations of that pressure.

External pressure acts on diaphragm
-> Diaphragm deforms
-> Piezoresistive or capacitive structure changes
-> Analog front end and ADC
-> Pressure output
-> Application model interprets level, altitude, or differential pressure

What Pressure Is

Pressure is force distributed over area:

A temperature and humidity sensor looks like it should directly describe the room: 25°C, 55% RH, comfortable or not.

That is too simple.

The sensor measures the local air around itself. The reading may not represent the whole room, the air outside the enclosure, or how a person feels.

The first model is: a temperature and humidity sensor measures the temperature and water-vapor state of air near the sensing element. Enclosure, airflow, self-heating, response time, and contamination decide whether that local air represents the environment you care about.