An accelerometer is a sensor that measures changes in motion, and it shows up in an enormous range of devices you interact with daily. Your smartphone uses one to rotate the screen when you turn it sideways. Your car uses one to trigger airbags in a crash. Your fitness tracker uses one to count your steps. Beyond consumer electronics, accelerometers monitor earthquake activity, detect when elderly people fall, keep factory equipment running, and help engineers determine whether bridges are structurally sound.
How an Accelerometer Works
Most modern accelerometers are tiny chips smaller than a fingernail, built using a technology called MEMS (micro-electromechanical systems). Inside the chip, a small weighted structure called a proof mass is suspended between two fixed plates by microscopic springs. When the device accelerates, the proof mass shifts toward one plate and away from the other. That physical shift changes the electrical charge (capacitance) between the plates, and a circuit converts that change into a voltage signal proportional to the acceleration.
Because the sensor measures movement along a line, most devices use three accelerometers oriented at right angles to capture motion in all three dimensions. When paired with a gyroscope (which measures rotation) and sometimes a magnetometer (which senses compass direction), the combination is called an inertial measurement unit, or IMU. This package can track an object’s full orientation, including pitch, roll, and heading, without relying on GPS or any external signal.
Smartphones and Tablets
The accelerometer became a mainstream consumer sensor with the original iPhone in 2007, where its sole job was rotating the screen between portrait and landscape mode. Since then, its role has expanded dramatically. Smartphones use accelerometer data to detect when you raise the phone to your ear during a call, to track your steps throughout the day, and to enable tilt-based controls in games. Combined with a gyroscope, the accelerometer also powers augmented reality apps and image stabilization in your camera.
Fitness Trackers and Step Counting
When you walk or run, your body produces a rhythmic, wave-like pattern of acceleration with distinct peaks during each step. A waist-worn or pocket device counts steps by detecting either the peaks or the zero-crossings in that vertical acceleration curve over time. Wrist-worn trackers work on the same principle but face a unique challenge: they can miss steps when your wrist is stationary (like when you’re pushing a stroller) and log false steps when you’re gesturing or folding laundry.
To filter out noise, many step counters require at least four continuous seconds of a regular, rhythmic stepping pattern before they start recording. This prevents brief arm movements or single jolts from inflating your count. Accuracy varies by placement. Waist-mounted sensors tend to be the most reliable because they sit close to your center of mass, where the stepping signal is strongest and cleanest.
Fall Detection in Smartwatches
Fall detection systems rely on a two-step process. First, the accelerometer watches for a sudden spike in force. A typical threshold is around 2 g, meaning twice the force of gravity, which is well above what normal daily activities produce. Second, the system checks what happens in the two seconds after the spike. In a real fall, the person ends up lying on the ground, which produces a rotation of roughly 90 degrees from their previous standing position. The algorithm calculates this rotation (allowing a margin of plus or minus 30 degrees to account for uneven ground or awkward landing positions) and, if both conditions are met, triggers an alert.
This layered approach matters because plenty of everyday movements create brief force spikes. Sitting down hard, clapping, or dropping the device could all exceed 2 g momentarily. Requiring both the impact spike and the subsequent body rotation dramatically reduces false alarms.
Car Crash Sensors and Airbag Deployment
In a vehicle collision, airbags need to inflate within about 30 milliseconds, which is roughly the time it takes to blink. The accelerometer in a crash sensor is calibrated to measure forces from negative 35 g to positive 35 g. When the sensor detects a rapid deceleration pattern consistent with a crash, the car’s processor compares the accumulated velocity change against a preset threshold. If it exceeds that threshold, the system fires the airbag inflator. The entire decision, from impact detection to deployment signal, takes less than 20 milliseconds. Different thresholds are set for different crash types: frontal, oblique, offset barrier, and pole impacts all have distinct deployment criteria.
Predictive Maintenance in Factories
Every motor, pump, fan, and gearbox vibrates while it runs. When bearings start to wear or a rotor becomes unbalanced, the vibration pattern changes in subtle but measurable ways long before the machine actually fails. Industrial accelerometers pick up these vibration signatures continuously, allowing maintenance teams to replace parts on a predictable schedule rather than waiting for a breakdown.
The key specification for industrial sensors is frequency response. To catch bearing defects, an accelerometer needs to measure vibrations at frequencies 40 to 50 times the shaft’s rotational speed. For fans and gearboxes, the sensor needs to cover at least 4 to 5 times the blade-passing frequency. Higher-end piezoelectric sensors can resolve vibrations as fine as 1.4 millig, roughly ten times more sensitive than typical MEMS sensors. That extra resolution lets engineers spot developing faults weeks or months earlier.
Earthquake Monitoring
Seismic networks use accelerometers to detect ground motion during earthquakes. The Advanced National Seismic System classifies these sensors into three tiers. Class A sensors offer the highest resolution (20 to 26 bits of useful data), suitable for research-grade seismology. Class C sensors, built with MEMS technology, deliver about 14 to 15 bits of resolution over a range of plus or minus 2 g. While less precise, Class C sensors are inexpensive enough to deploy in dense networks across earthquake-prone regions.
Field testing of low-cost MEMS seismic sensors in southwest China showed they achieved a dynamic range of roughly 88 to 91 decibels across frequencies from 0.005 to 25 Hz. That bandwidth captures everything from the slow rolling waves of a distant earthquake to the sharp jolts of a nearby one. Dense networks of these affordable sensors are particularly valuable for earthquake early warning systems, where having many observation points close together matters more than having a few ultra-precise ones far apart.
Bridge and Building Monitoring
Engineers attach accelerometers to bridges, buildings, and other structures to continuously measure how they vibrate under everyday loads like traffic, wind, and temperature changes. The goal is to identify the structure’s natural frequencies, the specific vibration rates at which it resonates most strongly. A healthy bridge has predictable natural frequencies; shifts in those frequencies over time can signal developing cracks, foundation settling, or weakening connections.
In a real-world example, dynamic load testing of a bridge identified four distinct natural frequencies. The lowest was 0.42 Hz, noticeably higher than the 0.29 Hz predicted by theoretical calculations. The most active vibrations clustered around 2.5 Hz. Engineers compare these measured values against computer models to assess whether the structure is behaving as expected or showing early signs of degradation. Accelerometers record both vertical and lateral (side-to-side) accelerations, since problems like deck flutter or foundation scour tend to show up as unusual lateral vibration before they become visible to the eye.
Navigation Without GPS
In tunnels, underwater, indoors, or in dense urban canyons where GPS signals can’t reach, accelerometers provide a backup navigation method called dead reckoning. By continuously measuring acceleration in all three axes and integrating that data over time, a navigation system can estimate how far and in what direction it has moved from a known starting point. Submarines, mining vehicles, and indoor robots all rely on this approach.
The limitation is drift. Small measurement errors accumulate with every passing second, so a purely accelerometer-based position estimate becomes increasingly inaccurate over time. Practical systems address this by fusing accelerometer data with gyroscope readings, magnetometer input, and intermittent GPS fixes whenever a satellite signal becomes available again. Algorithms developed specifically for this sensor fusion, including one published by Sebastian Madgwick in 2010 that became widely adopted, efficiently combine these inputs to maintain a reliable estimate of position and orientation even during extended GPS outages.