How does a phone sense a turn?
A phone’s gyro usually uses a tiny vibrating structure on a chip, rather than a little spinning wheel. Turning the chip adds a sideways response. Electronics read that change to detect which way it turns and how fast.
Adding up these small turns estimates the angle already turned. Small errors can add up too, so phones also use information from other sensors.
Try two small comparisons
- Keep the chip still and watch the tiny structure vibrate.
- Reverse the chip’s rotation and compare the sideways response.
More to explore
Look closer: observations and controls
Is a tiny spinning top hiding inside your phone?
Why does a game view turn when you turn the phone? Usually there is no tiny spinning wheel inside. A small structure vibrates back and forth. Watch how rotation adds sideways motion, which a circuit uses to measure the turn.
Turn your phone and a game’s view can turn with it. A gyroscope commonly supplies the angular rate; software then estimates orientation. Rate answers “How fast is it turning now?” Orientation answers “Which way is it facing?” Android’s sensor overview lists the gyroscope and rotation-vector sensor separately.
Start with a traditional free rotor. A spinning wheel carries angular momentum. With very little external torque, its axis approximately preserves its direction in inertial space. Gimbals let the support change direction without forcing the rotor axis to follow, providing an orientation reference. Our rotor is an ideal example without friction or external torque. A real gyro under gravity torque can precess; its axis is not unconditionally fixed. NASA’s gyro activity uses a bicycle wheel to explore angular momentum.
A common phone MEMS gyroscope uses a different mechanism. MEMS means microelectromechanical systems: masses, springs and electrodes fabricated on a chip. Electronics drive a silicon mass back and forth. When the chip also rotates around an axis perpendicular to its plane, the moving mass experiences a transverse Coriolis force in the chip’s rotating frame. Reversing rotation reverses that force; reversing the mass velocity does too.
The resulting transverse vibration changes sensing capacitances. Electronics read the tiny signal and synchronously demodulate it against the known drive oscillation to recover rotation direction and rate. The two opposing plates shown here are a simplified differential arrangement. Real chip structures vary. ADI’s MEMS rate-sensor explanation describes this vibration-to-Coriolis-to-capacitance chain.
Follow a turn, then watch what happens when it stops
- In the free-rotor chapter, track the pale blue shaft line. The support changes direction while the ideal rotor axis keeps its inertial direction.
- First observe the MEMS drive motion with the chip stationary. It still vibrates along the drive direction, but has no transverse response caused by rotation.
- In exploration, change “Chip angular rate” from positive to negative, then drag “Vibration phase.” The green arrow shows drive velocity; the gold arrow shows instantaneous transverse force. Displacement does not instantly copy the force: springs and damping introduce a phase difference.
- In the integration chapter, follow one reproducible turn. The area under the rate curve grows together with the accumulated angle. After the phone stops, rate returns to zero while the accumulated 60° remains.
- Both phones in the drift comparison are physically stationary. Integrating a biased gyro reading alone slowly tilts the estimated pose. A trusted gravity reference constrains roll error. Disable gravity correction and both estimates drift under the same bias.
How does Coriolis force become an angular-rate reading?
Let x be the drive direction in the chip plane, y the transverse sensing direction, and positive Ω point out of the chip. In the chip’s rotating frame, a mass moving with x-velocity vₓ experiences:
The drawing has x to the right and y upward. Positive rotation with rightward velocity produces downward force. Reversing either Ω or vₓ reverses Fᵧ. With transverse spring stiffness k and damping c, the mass obeys:
The model solves the steady forced-oscillator response, then calculates the two ideal plate capacitances using C = εA/d. Electrode areas are equal and displacement changes their opposing gaps. The rate readout recovers angular rate from the steady response’s two quadrature components and a known transfer function, representing synchronous demodulation across cycles. It does not equate an instantaneous capacitance with rate or implement a real sensor’s complete demodulation circuit. The 4 kHz drive is a teaching parameter, displayed in slow motion. Drive and sense displacements use separate magnification factors; the picture is not a scale drawing and cannot establish actual frequencies, dimensions or displacement ratios. It does not simulate a specific phone chip, full electronics, feedback or noise.
Why can the estimated angle drift while the device is still?
For rotation around one fixed axis, multiplying angular rate by each short time interval and adding the increments gives the change from an initial angle:
A real gyro can retain a small nonzero reading at rest: residual bias. With constant residual bias b, integration error grows as b × t. This demonstration uses 0.35°/s, accumulating 21° in 60 seconds. It is an example, not a phone accuracy specification. Real bias can vary with time and temperature, and readings also contain noise. ADI’s bias-drift discussion explains the accumulating error.
When stationary, or when linear acceleration is negligible, an accelerometer provides a gravity-based tilt reference that can constrain roll and pitch. Turning around the vertical axis does not change the gravity direction, so gravity alone cannot correct yaw drift. Under sustained acceleration, its reading cannot simply be treated as gravity. ADI’s inclination-sensing application note states this assumption.
Our stationary roll comparison uses a one-dimensional complementary correction with e′ = b − e/τ and τ = 2 seconds. General three-dimensional orientation updates normally use rotation matrices or quaternions. Integrating the three body-axis rates separately does not directly produce three Euler angles at arbitrary orientations.
Do phones and missiles use the same kind of gyro?
Both may need to sense their own rotation, but their sensors and downstream systems can differ substantially. Phones commonly use small, low-power MEMS gyros together with accelerometers and other data for orientation interaction, tracking and stabilization. Simple screen autorotation can mainly use an accelerometer; it is not always a gyro function. Bosch’s motion-sensor overview describes MEMS IMUs for phones and other devices.
Aircraft and other vehicles, including missiles with inertial navigation, use gyros to track body rotation. Accelerometers measure specific force. A computer combines that with orientation, a gravity model, and an initial position and velocity to estimate changes in velocity and position. A gyro itself neither measures position nor supplies a destination. Navigation errors accumulate and can be corrected using suitable external information.
Higher-precision systems may use fibre-optic or ring-laser gyroscopes, which compare counter-propagating light affected by rotation. Their mechanism does not depend on a MEMS vibrating mass. The type used in a particular vehicle depends on its design. ESA’s IMU project illustrates three gyro axes combined with three accelerometer axes, including a fibre-optic gyro and MEMS accelerometer design.