mjb/altos/src, branch 1.8.1

altos: Don't compute filtered average of height error when HAS_ACCEL

2017-08-28T07:22:43+00:00

We only use this for baro-only devices to avoid firing drogue charges at mach transitions; we trust the combination of accel+baro to do the right thing when available. Signed-off-by: Keith Packard

altos/test: Fake baro data for flight 12 serial 2093

2017-08-28T00:40:53+00:00

This flight had a baro spike due to an accidental drogue charge firing but is otherwise quite useful when testing for various mach delay effects, so fake out the data during that spike. Signed-off-by: Keith Packard

altos/test: Integrate raw accel to provide speed for comparison

2017-08-28T00:40:53+00:00

This can provide a useful visualization of the 'true' vs 'kalman' speed value, as the kalman is necessarily delayed due to the model assuming constant acceleration. Signed-off-by: Keith Packard

altos/test: telemetrum baro data is inverted before being written to eeprom

2017-08-28T00:40:53+00:00

No need for the test code to invert it during replay Signed-off-by: Keith Packard

altos: Don't eliminate baro above mach speed, just trust it less

2017-08-28T00:40:40+00:00

Instead of completely eliminating the baro sensor above mach speed, just derate it a bit so that the accel will dominate for speed computation and keep the device from false-triggering across mach transitions. When we completely ignored the baro sensor above mach, and the flight spent considerable time in that speed range, then the estimated height could be far from the real value. When the estimated speed dropped back down and the baro values were brought back into the computation, then the resulting rapid shift in estimated speed could trigger accidental apogee detection. By mixing in a bit of baro data even above mach, we keep the estimated height closer to the baro value and prevent this error, at least in flights measured so far. The flight known to have this problem is: 2015-09-26-serial-2093-flight-0012.eeprom Signed-off-by: Keith Packard

altos: Eliminate separate height error filter for accelerometer devices

2017-08-28T00:40:40+00:00

We don't use the error value in flight for those models anyways; it's only useful on baro-only hardware. Signed-off-by: Keith Packard

altos: Perform time comparisons using 16-bit arithmetic to handle wrap

2017-08-28T00:38:58+00:00

Subtracting two 16-bit unsigned values to perform time comparisons yields mystic results unless we carefully cast that to int16_t. Signed-off-by: Keith Packard

Merge remote-tracking branch 'origin/master'

2017-08-12T03:45:25+00:00

altos/test: Fix AES key format in test code

2017-07-23T23:47:40+00:00

Use temp variable instead of stepping on the AES name. Signed-off-by: Keith Packard

altos/test: Fix printf format for int64 arithmetic test

2017-07-23T23:46:24+00:00

Use 'long long' and %lld for 64-bit values when printing. Signed-off-by: Keith Packard