Fixes common MPU-9250 accelerometer issues for RPi Linux MCU and
improves reliability on all other architectures by adjusting the MPU-*
reading algorithm to only read whole Klipper messages' worth of
data from MPU-* and eliminating many unnecessary checks of the MPU FIFO
fill-level that consumed bus bandwidth needed for data transfer. Fixes
intermittent "Lost communication with MCU 'rpi'" due to "Timer too
close" and transposed / corrupted data due to FIFO overrun/data loss
when using MPU-* accelerometers. In addition FIFO overrun checks are
performed by testing the MPU interrupt flag vs. inferring from the FIFO
fill level.
Stress tested for 13hrs with two MPU-6500 attached to one I2C bus on RPi
and one on a PR2040
Stress tested for 23hrs with two MPU-6500 attached to one I2C bus on RPi
and one on a ATmega328P (Seeduino Nano)
Signed-off-by: Matthew Swabey <matthew@swabey.org>
Calling python functions can have high overhead. Inline the
twos_complement code in the _extract_samples() inner loop.
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
The MPU9250CommandHelper() class is nearly a duplicate of
ADXL345CommandHelper(). Rename ADXL345QueryHelper() to
AccelCommandHelper and remove user facing references to "adxl345". Use
it directly from mpu9250.py .
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
The mpu9250.ClockSyncRegression() class is a duplicate of
adxl345.ClockSyncRegression(). Remove the duplicate copy and use the
code from adxl345.py .
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
The MPU9250QueryHelper() class is a duplicate of ADXL345QueryHelper().
Rename ADXL345QueryHelper() to AccelQueryHelper() and use it directly
from mpu9250.py .
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>
Dr. Matthew Swabey