Test flight 4

Late in December we completed the FCS hardware integration, incorporating the CPU board with C6657 DSP and Exynos4412 module, a 3G wireless modem card, an RFD900a, two I/O boards, and a NEO-7N GPS. This enabled the AHRS and communications software running on the DSP to be tidied up, and by the end of December we were ready for an in-flight test.

Photo showing the integrated FCS hardware

The objectives of our fourth test flight were:

Photo showing X8 with top panel removed, being prepared for launch

The weather was perfect: calm and clear, with the sun starting to set behind the trees to the west, so it was out of our eyes.

The reduced power of the motor resulted in a very different take-off approach being required, so it took a few attempts to get it into the air; after we did, it became apparent that the weight distribution was a little nose-heavy, so we decided to land and re-balance.

We then found that the overall reduced weight of the FCS compared with the ODROID-X2 and numerous USB cables made a significant difference to the X8’s already considerable difficulty in reaching the ground. After coasting in ground effect for a while with the motor off, Daniel realised that another landing approach would be required, so he increased throttle for a go-around.

At that point, the small plastic strip preventing the prop blades from folding past the axle snapped, and significant vibration set in as the throttle was increased. To avoid damage to the motor and/or mount, Daniel cut the throttle and completed a nose-first emergency landing into the nearest tree.

Animation showing the X8 hitting a tree after a failed go-around

Airframe damage was minimal, but we thought it best to obtain more experience with the handling characteristics of the new motor configuration so after re-balancing and repairing the plastic strip we removed the FCS board and flew another 15 minutes.

Still from on-board GoPro video showing the sunset

Once in the air, performance was fine, with enough power to loop, roll and climb easily. Power consumption is significantly reduced, but we’ll probably need to increase overall battery capacity to 12Ah or so in order to maintain sufficient reserves in the 1-hour flight required by the Outback Challenge. This time we noticed no ESC overheating, so unless we identify major problems during longer test flights we intend to use this configuration in the Challenge.

Photo of the X8 in flight

Unfortunately, removing the FCS prevented us from logging any sensor data, so we haven’t verified performance in flight. However, in static testing the results are looking promising, with variation in output at rest being under 0.2° and in-run repeatability within 0.5° after 180° rotations on each axis. Between-run repeatability depends on the amount of time spent calibrating the magnetometer (by rotating the system slowly around each axis); on startup the error is less than 3°, and after one rotation per axis the error is less than 1°. The nice thing about using TRICAL for the magnetometer calibration is that there’s no separate calibration process—if you rotate the aircraft sufficiently around each axis prior to launch, you’ll get somewhat better initial accuracy, but even if you don’t it’ll calibrate automatically in flight.

Since the test flight, we’ve verified the telemetry integration between the FCS and our custom GCS software, using an RFD900a connected to the DSP. There’s quite a bit more to do on this software, in particular the control interface, but the main flight data display functions are there and work well.

Screenshot of the GCS interface
github.com/sfwa twitter.com/sfwa_uav youtube.com/user/sfwavideo