this seems to have been the cause of the 'flips' seen by Marco and
others. Testing by Craig and Alan shows that the flips are gone when
the barometric acceleration is removed.
It looks like a 5 point average filter is not enough to keep the
vertical acceleteration noise low. With high noise in the z axes, the
x and y axes are scaled back when the ge vector is normalised.
this follows the method that Bill developed in his fastRotations
paper. We've demonstrated that this is indeed needed in APM, as we
were able to produce the 'dizzy' effects in both the ArduPlane and
ArduCopter simulator
the new AHRS code doesn't use calculate() and the compass.heading
attribute. Instead it works on the raw magnetometer vector. This
change removes the internal calculate state from the compass object
and instead adds calculate_heading() for use by older code that
doesn't go via AHRS.
This significantly reduces the calculation involved in compass updates
The null offsets enable/disable code is also removed, as it is not
needed now that compass offsets are not linked to the AHRS state.
this allows the barometer driver to calibrate and return altitude and
climb rate values. This will be used by the AHRS drift correction code
for vertical velocity
The climb rate uses a 5 point average filter
omega_I applied continuously. _ki larger. Stop integrating when _omega.length()>20
The key change was the scaling of ge to ensure the error is not
quadratic
This makes 3 major changes:
1) fixes the scaling of the yaw drift correction term to fix the time
constant
2) don't integrate the mag vector over multiple readings
3) accumulate omega_I changes over 15 seconds before applying, to try
to prevent omega_I picking up short term responses
if a UBlox is configured for NMEA only, with no UBX messages at all
then it would never trigger the GPS_AUTO detection. This adds a UBX
config message to the init strings that enables the NAV_SOL message
the parser was broken in several ways:
1) when it received an unknown message it didn't update the ck_a and ck_b fields, so it
thought the message had a bad checksum, which meant it got out of protocol sync
2) the read() method would return false if the last message from the GPS was of an unknown type.
So we relied on the last msg always being one that we understand and want
3) the parser considered any valid UBLOX message to be 'new data', whereas we only actually get
a new fix when we get box a new position and velned message
4) the total message size per update is more than 128 bytes, but the serial port was opened
with only a 128 byte buffer, so we got corruption regularly
Andrew Tridgell
Jonathan Challinger
rmackay9
Jason Short
Amilcar Lucas
Wingspinner