A MEMS gyroscope measures angular velocity by exploiting the Coriolis effect: when a vibrating mass rotates, it experiences a sideways force perpendicular to both its vibration direction and the rotation axis. Etch a tiny proof mass into silicon, drive it into oscillation, and measure the perpendicular deflection — that deflection is proportional to the rotation rate around the third axis.

The physics: Coriolis acceleration is a_c = 2ω × v, where v is the drive velocity and ω is the rotation rate you're trying to measure. Because the drive velocity is known and controlled, the sideways Coriolis force becomes a direct proxy for ω. That sideways motion is picked up capacitively by fixed comb-finger electrodes flanking the moving mass.

Why the tuning fork shape? Two counter-vibrating masses cancel the linear momentum of the drive motion but produce opposite Coriolis deflections when rotated. Differential sensing rejects linear acceleration (which shakes both masses the same way) while amplifying rotation signals. Without this trick, every bump on the road would look like rotation.

Real-world example: The InvenSense MPU-6050 combines a 3-axis MEMS gyro and 3-axis accelerometer on a 4×4 mm chip for about $3. It's inside virtually every quadcopter, VR headset, and smartphone. When your phone rotates to landscape, the accelerometer detects the final tilt — but the gyro is what tracks the motion of rotation smoothly during the swipe or game input, because accelerometers alone are noisy and can't distinguish tilt from horizontal acceleration.

Sensor fusion rule of thumb: Gyros are accurate short-term but drift (integration of small bias errors accumulates as angle error over time — typically 1–10°/hour for consumer MEMS, ~0.01°/hour for tactical grade). Accelerometers give absolute tilt reference but are noisy. A complementary filter blends them:

• angle = 0.98 × (angle + gyro_rate × dt) + 0.02 × accel_angle

The 98% weight on the integrated gyro captures fast, smooth motion; the 2% pull toward the accelerometer washes out gyro drift over seconds. Kalman filters do the same job more rigorously with statistical weighting.

Failure modes: MEMS gyros are sensitive to vibration at their drive frequency (typically 10–30 kHz) — mounting one near an ultrasonic transducer or high-RPM motor can inject huge false rotation signals. They also have temperature-dependent bias, which is why premium IMUs include a temperature sensor and factory calibration table.