- new static notch filter configured via IMU_GYRO_NF1_FRQ and IMU_GYRO_NF1_BW
- existing notch parameters IMU_GYRO_NF_FREQ and IMU_GYRO_NF_BW become
IMU_GYRO_NF0_FRQ and IMU_GYRO_NF0_BW
- this simplifies the reset by allowing a notch filter to reset as
needed
- improves cascade initialization, on reset each filter will reset
properly from the previous
- non-trivial perf counters (elapsed & interval) are relatively expensive
- if ESC and FFT notch filtering are enabled this reduces 6 updates (2 per axis) to 1
- avoid unnecessary filter reset on parameter update
- additional minor optimizations (precomputing dt inverse, etc)
- moving filter reset check and dynamic notch filter update checks out of the update loops
- this were necessary previously when the scale factor wasn't applied prior to filtering the otherwise raw data
Having a 2nd order low-pass filter in the derivative loop reduces
stability at low cutoff values as too much phase is lost through
the filter. Using a 1st order filter avoids this issue because its
maximum phase loss is 90 degrees instead of 180 degrees for a 2nd order
lpf.
- initial frequency peak tracking SNR increased from 10->15 db
- after initial detection the threshold decreases to SNR 5db
- gyro_fft large method refactored into smaller pieces
- sensors/vehicle_angular_velocity: dynamic notch FFT make sample rate
check a percentage and relax lower bound safety threshold
- apply sensor scaling immediately to keep things simple (FIFO vs regular)
- inline filter helpers (minor performance improvement)
- dynamic notch filtering
- reorder by axis (applied per axis)
- don't remove notch filters immediately if ESC or FFT data times out
- constrain notch filter frequency and bandwidth to safe range (minimum bandwidth for flaot precision, Nyquist, etc)
- add safe constraint on dt
- merge NotchFilterArray into regular NotchFilter (apply vs applyArray)
- only use direct form 1 to prevent reset confusion
- safe default field initialization
- update VehicleAngularVelocity usage
For the sake of efficiency (at 8 kHz) all filtering is performed on the raw data before the calibration and rotation is applied to only the final output. As a result we have to be a bit more careful when switching between sensors or in cases where the gyro scale factor changes (eg icm42688p 20 bit data rescaled to fit in int16 output).
- rotates accel & gyro FIFO data before publication both to simplify downstream usage (including log review) and fix other issues
- to best handle int16_t data rotations are now either performed with swaps if possible, otherwise promoted to float, rotated using the full rotation matrix, then rounded back to int16_t
- fix sensors/vehicle_angular_velocity filter reset both with proper rotation and new calibration uncorrect helper
- in FIFO case filtering is done before calibration is applied, but we need to handle a possible reset from a completely different sensor (vehicle body angular velocity -> sensor frame uncorrected data)
- parameter updates can be quite expensive because they trigger nearly all modules to reload all of their parameters immediately
- limit modules from updating faster than once per second
Daniel Agar