An accelerometer measures linear acceleration along its axes, in metres per second squared (m/s²) or in multiples of , the acceleration of gravity at the Earth’s surface ( m/s²). A 3-axis accelerometer reports acceleration along three perpendicular directions, conventionally , , .
A modern MEMS (micro-electro-mechanical systems) accelerometer works on a spring-and-mass principle implemented in silicon. A tiny mass is suspended by tiny springs, and as the device accelerates, the mass’s position relative to the housing changes. The displacement is read out capacitively — the mass and the housing form a variable capacitor whose capacitance depends on the separation.
A stationary accelerometer doesn’t read zero. It reads the acceleration that opposes gravity — typically about m/s² straight up if the device is held flat — because the housing has to push the mass upward against gravity to keep it from falling. This is why an accelerometer at rest can sense which way is down. The caveat: it can only do so while it isn’t otherwise accelerating. A phone sitting on a table reads pure gravity; a phone in a car making a turn reads gravity plus the lateral acceleration of the turn, and the two are indistinguishable from a single accelerometer reading. This is exactly why the Gyroscope and Magnetometer are needed for robust orientation tracking.
Accelerometers are sensitive to vibration, which is sometimes the measurement of interest (machine condition monitoring) and sometimes a noise source (a human walking with a phone in their pocket). The accelerometer is one of the three sensors inside an IMU, alongside the Gyroscope and Magnetometer, and the three are typically combined through Sensor fusion to estimate orientation.