Accelerometer Tester

Accelerometers are crucial for many features users take for granted: screen auto-rotation, step counting, sleep tracking, fall detection, augmented reality stabilization, image stabilization in camera, and game motion controls. Testing helps in several scenarios. Developers building motion-based apps need to verify sensor availability, frequency, and noise floor before designing algorithms. Users with suspected hardware faults can confirm whether sensor drift, dropped readings, or stuck values indicate a defect worth a warranty claim. Second-hand buyers can verify a phone's sensors before purchase. Curious users learning about physics can observe gravity directly, see how rotation and translation appear differently to a fixed-frame sensor, and develop intuition about MEMS (microelectromechanical) device behavior.

A stationary phone lying flat face-up should show: Z near +9.81 m/s² (gravity pointing into the screen), X and Y near zero with a small noise floor of ±0.05 to ±0.2 m/s² depending on the sensor grade. Rotating the phone to landscape switches the gravity vector to X. Standing the phone vertical puts gravity on Y. Magnitudes should always sum to approximately 9.81 m/s² when stationary: sqrt(X² + Y² + Z²) ≈ 9.81. Walking generates oscillating peaks of ±2 to ±5 m/s² at roughly 2 Hz. A small drop or shake generates ±15 to ±30 m/s² spikes. Free fall would show all axes at 0 m/s² for the duration of the fall. The shipped sensor in flagship phones runs at 50–400 Hz with about 12-bit resolution.

Even at perfect rest, MEMS accelerometers show three kinds of noise. White noise from electrical fluctuations sets the noise floor at about 0.01–0.1 m/s²/√Hz for consumer-grade sensors. Bias drift slowly shifts the zero-g reading by ±0.01 m/s² per minute as temperature changes. Vibration from your hand (1–10 Hz tremor and breathing motion) couples into the sensor at ±0.05 to ±0.3 m/s². Together these produce visible jitter. Better sensors use lower-noise capacitive readout and digital filtering; worst-case sensors in budget phones can show ±0.5 m/s² noise even when truly still. To reduce noise, average over 100–200 samples (~1 second) or apply a low-pass filter. Be aware that filtering also delays your response to real motion.

Noise has three characteristics: it is roughly Gaussian-distributed around the current bias, its amplitude is small (under 0.5 m/s² for a healthy sensor), and it shows no temporal pattern. Real motion has signature shapes — a footstep produces a sharp upward peak followed by a downward rebound lasting 100–300 ms, a phone tilt produces a smooth multi-second transition as gravity redistributes across axes, a tap produces a single-sample impulse with high frequency content, and free fall produces simultaneous near-zero values on all three axes. By examining the magnitude (vector norm) you can filter out tilt (which preserves total magnitude) and isolate translation. Modern fitness apps and step counters combine accelerometer with gyroscope to disambiguate between rotation and translation, which is why they work better than naive single-axis thresholds.

These three together form the inertial measurement unit (IMU) inside your phone. The accelerometer measures linear acceleration including gravity, in m/s². The gyroscope measures angular velocity around three axes in radians per second (or degrees per second), independent of orientation. The magnetometer measures the local magnetic field vector in microtesla (μT), giving an absolute heading reference for compass functions. Each sensor alone has limitations: accelerometer cannot distinguish rotation from translation, gyroscope drifts over seconds-to-minutes, magnetometer is distorted by nearby metal and electronics. Sensor fusion algorithms like Madgwick or Kalman filter combine all three to produce a robust orientation estimate that is used for image stabilization, AR tracking, and navigation. This tool tests only the accelerometer in isolation.

A MEMS (microelectromechanical systems) accelerometer is a tiny silicon chip about 2×2 mm containing a microscopic proof mass suspended by springs. When the phone accelerates, inertia keeps the proof mass momentarily behind, deflecting the springs in proportion to acceleration. The deflection is measured capacitively: pairs of comb-like electrodes on the proof mass and fixed structure form variable capacitors whose values shift with displacement. ASIC electronics convert the capacitance change to a voltage, then to a digital count via an integrated analog-to-digital converter. Typical sensitivity is in the 0.0001 g range per LSB (least significant bit) at the smallest scale setting. Bosch BMI160, ST LSM6DSO, and InvenSense ICM-42688 are the chips inside most flagship phones and follow the IEEE 1057 specification for measurement accuracy.

Yes. Acceleration is reported in the International System (SI) base unit m/s² as defined in BIPM CGPM resolutions. The gravitational reference of 9.80665 m/s² is the standard gravity defined by the third CGPM (1901) and rounded to 9.81 in most engineering contexts. For sensor specifications, IEEE 1293 covers single-axis accelerometers and IEEE 1554-2014 covers multi-axis IMUs, defining test methodology for bias instability, scale factor, noise density, and cross-axis sensitivity. Consumer phone accelerometers typically operate at ±2g, ±4g, ±8g, or ±16g full-scale ranges; the range is software-selectable to trade resolution against maximum measurable shock. Sports and automotive accelerometers go to ±100g; impact-survival aerospace sensors reach ±500g or more. The IPC J-STD-001 and JEDEC JESD22 standards also govern shock testing for the chips themselves.

DeviceMotion works on devices that actually have a motion sensor, served over a secure context. Android Chrome and most mobile browsers work immediately with no permission prompt. Safari on iOS 13+ works but requires you to tap a button so Safari can show its motion-permission dialog — that is exactly what the Start button does here. Desktop and laptop computers usually have no accelerometer at all, so even though the DeviceMotion API exists in Chrome and Firefox, no events ever fire; this tool runs a short watchdog and shows an honest "no data" message instead of pretending the sensor is live. A few convertibles and tablets with rotation sensors are the exception. Critically, the page must be loaded over HTTPS: browsers block motion sensors on insecure (http) origins, so on an http page the tool will report that a secure connection is required rather than spinning forever on empty readings.

On Android Chrome no permission prompt is needed; on iOS 13+ a single tap grants motion access. Either way every sample is read by the W3C DeviceMotion API and stays inside your browser — readings never leave the page, even when you export them. While a test runs the tool records each sample into an in-memory session and shows a live sample count plus the average and minimum magnitude alongside the existing peak (max), which is what a noise-floor or drop check needs. Press Export CSV or Export JSON to download the recorded session. The CSV has columns time_ms, x_ms2, y_ms2, z_ms2, magnitude_ms2 (one row per sample); the JSON includes the tool name, an ISO timestamp, the unit, the total sample count, the session duration in milliseconds, and the full samples array. Attach the file to a warranty claim, diff a suspect unit against a known-good baseline, or analyse sample-rate jitter offline.