Just completed another couple of test flights. Different location, just in case the golf course that I used previously had some undetected magnetic anomalies or other issues. To summarize the last test - large GPS-mode track errors with J-hook that mostly improved during flight but randomly reappeared, CL off by at least 2 x declination, substantial altitude changes and lots of uncommanded yaw.
EQUIPMENT
HARDWARE: P2, H3-3D v1.0, GoPro H3+, Lightbridge, Futaba 14SG.
FIRMWARE:
Main Controller: 3.05
GPS: 6.0.1.4
P330CB: 1.0.1.30
Gimbal CMU: 1.0
Battery: 2.0.0.0
H3-3D IMU: 1.0.0.6
BASIC GAINS: (P, R, Y, T) 125, 125, 120, 140.
ATTI GAINS: (P, R) 260, 260.
I have never altered these gains - just the gimbal tilt gain.
TEST CONDITIONS
Winds calm. This time startup heading was 300˚T (as opposed to 127˚T). I would have explored an orthogonal startup, but the test was cut short by a thunderstorm. Another difference, outside my control, is that I had 10 or 11 satellites at all times - more than I have seen before, and in contrast to just 7 or 8 last time. I used elevator trim, rather than sticks, to perform the runs in order to ensure that I had no inadvertent roll input.
RESULTS
Good news and bad news. The bad news is that the results were completely different. The good news is that I saw no discernible TBE, GPS runs were fairly straight right from the start (± < 5˚) with virtually no trace of J-hook and, remarkably, CL was approximately straight and within a couple of degrees of startup heading. Altitude control appeared to be good ± 1 m. Yaw on takeoff was minimal, and I saw only slight uncommanded yaw throughout the flights. HL brought the P2 back on a straight line to the takeoff point and was orientation independent, as it should be.
I'll post some video once I have processed it, but it really doesn't show much except mostly straight flight. It does seem to demonstrate that the H3-3D does not always look completely straight ahead.
I have no idea what is going on here, but possible variables accounting for the very different results: different location, different startup heading, lots of satellites, complete operator incompetence.
NEXT
I also tested out a combo trim setting (yaw + elevator) to drive the P2 in well-defined circles (primarily for descending), and it performed very well. I'll test that on my other P2 with GPS tracking to check the geometry, which may be an interesting indication of how isotropic the FC is, and also generate some precise linear tracks. Next tests will be the same location, but orthogonal startup heading, and then back to the golf course to compare.