The timestamp actually used by ekf2 is the sensor_combined timestamp, so we should be setting it with the actual data sample timestamp, rather than the aggregated sensors topic publication metadata.
Later the sensor_combined message will be replaced with vehicle_imu that carries a separate publication timestamp and raw sample timestamp.
* New NotchFilter methods ahead of RPMFilter implementation.
* Added Direct Form I implementation that support dynamic change of frequencies.
* Added update method to update frequency on an existing filter.
* Added setCoefficients method to easily and efficiently create clones of a filter.
* LowSide, HighSide, Onnotch and array tests testing the applyDF1 method.
The EAS is the calibrated airspeed corrected with a scaling factor to account for sensor error/placement of pitot.
Signed-off-by: Silvan Fuhrer <silvan@auterion.com>
- this is done to minimize real end-to-end latency
- actual pulse width only updated for next period regardless of output module
- add cycle interval perf counter
- skip avionics rail voltage check when USB connected
- skip forced reboot on USB disconnect if circuit breaker set
- avionics voltage preflight check don't silently fail if system_power unavailble
- explicitly set supply check circuit breaker (CBRK_SUPPLY_CHK)
- always check with state machine before reboot/shutdown
- respect BOARD_HAS_POWER_CONTROL (shutdown from command, low battery, power button)
- px4_shutdown_request add optional delay and always execute from HPWORK
- px4_shutdown_request split out px4_reboot_request
This is to prevent a wrong (false positive) failure detection of the airspeed
shortly after system boot due to delays in driver and/or estimator startup (seen in SITL).
Signed-off-by: Silvan Fuhrer <silvan@auterion.com>
- increased REP_XY and REP_Z: improves accuracy a bit, while increasing
measurement time (still allows to go to 50Hz, previous max was 100Hz)
- avoid extra transfer in measure()
- extend regdump output
- general code style cleanup
- 30Hz is the sensors max update rate in continous mode
(though not in forced mode as we are using it)
- 30Hz allows to increase the quality of the measurements
Most often the multiplication in convention descriptions and papers is
described in matrix multiplication style like this:
q · p := Q(q)p
Q(q) :=
[q0 −q1 −q2 −q3]
[q1 q0 −q3 q2]
[q2 q3 q0 −q1]
[q3 −q2 q1 q0]
I'm just rearanging the terms such that it's easily comparable with
these definitions additional to it being clearly described by
documenting we use the hamilton convention.
Daniel Agar