IMU and gyroscope

Summary¶

The gyroscope is the sensor that closes the innermost flight control loop. It measures the rotation rate of the airframe on three axes — roll, pitch, yaw — delivering new readings at up to 32,000 times per second. The MEMS gyroscope works using the Coriolis effect: a microscopic proof mass driven to oscillate at a known frequency deflects perpendicular to its oscillation when the chip rotates, and that deflection is proportional to rotation rate. The gyro cannot distinguish genuine rotation from vibration — both look identical in the sensor output — which is why motor vibration filtering is not optional. The accelerometer in the same IMU package measures the specific force (gravity plus linear acceleration) needed for level sensing and GPS-assisted modes.

Concept¶

MEMS gyroscope — how it works¶

Inside the ICM-42688-P, a microscopic proof mass (a few micrograms of etched polysilicon) is driven to oscillate at a precisely controlled resonant frequency — typically 20–40 kHz — by electrostatic actuation. When the chip rotates, the Coriolis force deflects the oscillating mass perpendicular to both its oscillation axis and the rotation axis. A capacitive sense element measures this deflection. The deflection magnitude is proportional to rotation rate.

The Coriolis effect is the physical basis of all MEMS gyroscopes. It is the same effect that causes hurricanes to spiral as air moves toward the equator in the Earth's rotating reference frame. In the gyroscope, the rotating reference frame is the chip itself.

The genius of using the Coriolis effect is that measurement is self-referencing. The drive frequency is precisely controlled; the sense measurement is synchronously demodulated at that same frequency. Environmental vibration at different frequencies is rejected by this demodulation. This is why MEMS gyroscopes can measure rotation in the presence of motor vibration — the motor vibration is at different frequencies than the drive frequency and is naturally rejected.

However, vibration at or near the drive frequency does contaminate the output. And more practically: vibration at flight-relevant frequencies (100–2000 Hz) generates gyro output that is indistinguishable from real rotation at those frequencies. The gyro samples rotation rate — it cannot know whether the rate change came from the aircraft moving or from the frame vibrating.

Gyro output data rate¶

The ICM-42688-P outputs data at up to 32,000 Hz (ODR 32 kHz). Betaflight samples this at 8 kHz — one reading every 125 µs. The gyro data arrives via SPI at ~10 MHz; a DMA transfer reads 16 bytes in ~13 µs without CPU involvement.

The accelerometer¶

The accelerometer in the same IMU chip measures specific force — the sum of all non-gravitational forces per unit mass. In steady hover, specific force is approximately 1g pointing downward (opposing gravity). The accelerometer tells the FC which direction is "down" — essential for self-leveling (angle mode) and for the GPS-assisted modes that need an absolute attitude reference.

In rate (manual) mode, the accelerometer is largely irrelevant — the FC tracks attitude by integrating gyro rate. Over time, gyro drift accumulates, but in the seconds-to-minutes typical flight duration this error is small enough to be managed by I-term correction.

In angle mode or GPS hold, the accelerometer provides the gravity reference that prevents accumulated gyro drift from giving a wrong "down" direction.

Noise sources that contaminate gyro output¶

Real attitude signal occupies 0–100 Hz for any realistic flight manoeuvre. Everything above ~100 Hz is noise. Sources:

The D term amplifies high-frequency noise. Every noise source above therefore contributes to motor command noise via the D term unless filtered.

Reference¶

ICM-42688-P key parameters¶

Filter chain (Betaflight)¶

Raw gyro (8 kHz) → RPM filter (36 notch positions) → Dynamic notch filter (3 notches per axis) → Gyro lowpass filter 1 (static PT1/Biquad) → Gyro lowpass filter 2 (static PT1/Biquad) → D-term lowpass filter → PID calculations

Each filter stage removes a class of noise at the cost of some phase lag. The RPM filter contributes least phase lag per noise rejection because its notches are narrow. Static lowpass filters contribute the most. Minimising the number and aggression of static filters while relying on RPM + dynamic notch filtering is current best practice. → See imu-filter-tuning.

Procedure¶

Rationale¶

Why the gyro is on a PCB moat¶

Frame vibration enters the FC through the mounting screws. A rigid PCB transmits this vibration directly to the IMU. The moat (a slot cut around the IMU island) mechanically decouples the IMU mounting point from the screw mounting points. Vibration must travel around the moat, attenuating significantly. Combined with the nylon standoffs (which further attenuate transmission from the ESC to the FC), two stages of mechanical isolation precede the software filtering chain.

Connections¶

requires: - closed-loop-control related: - flight-controller-hardware - vibration-isolation-theory - rpm-filter leads_to: - imu-filter-tuning - rpm-filter