NSL-65

NSL-65 was a test flight flown on the morning of Sunday July, 8th 2018. Paul L. had been working on some engineering issues concerning radio interference of GPS signals and needed more than just a ground integration test. The jet stream looked like it would behave for a few hours. But there was a tropical depression off the coast that was increasing winds. So it was a spur of the moment flight on an unusually cool summer morning. Brian B. and Matthew L. joined in to launch/chase/recover.

The payload consisted of the NSL-58 box with the following:

• Raspberry Pi Zero with GPS, camera, LoRa and Pi-In-The-Sky software (like NSL-63)

• PowerAdd 360 degree camera (dual fisheye) on its side, sticking out the side of the box

• Mobius Mini Action Camera, looking horizontally

• AP510 APRS tracker (NSL-63's damaged unit with new antenna)

• Outside Hung Compact Rain-Activated Pull-down

• Directed Arboreal Recovery Node with Integrated Tether

The payload had a 92cm chute and was hauled-up by a 600g cell with $10 worth of H_2.

In earlier tests, all on board GPS units were drowned out by interference from the Pi camera. Cameras often cause RFI problems, but can usually be overcome by covering them in aluminum foil. On this flight, the Pi and its camera had to be surrounded by layers of grounded shielding. Also the Pi's GPS module had to be relocated to its own shielded compartment; and its simple chip antenna was replaced with a more robust unit a dozen centimeters away. The RF-noisy buck converter for the Pi battery had to be replaced with a different, shielded unit. After all of this, the payload weighed in at 900g !

The flight was also a test of some Pi-In-The-Sky tracking gear. Paul had a 434MHz LoRa gateway sticking out a window at his house. Another was in his car with a mag mount antenna as well as a large Yagi antenna. Brian had his gateway on a breadboard in his car. All of these were connected via WiFi of some sort in order to transmit live telemetry and images to Habhub.org. During this flight, the small telemetry packets typically came in successfully, but image packets often had CRC errors.

360 Images

The inexpensive Poweradd 360 Degree VR Camera (caught on sale at $68) had not been flown before. It was unknown how long it would run in the cold or how much it may go out of focus. It consists of two fish-eye cameras that produce a single image like this.

These images then must be converted into other formats for the various 360 viewers to understand. For example, here is an image captured at burst in dual fish-eye and then converted into equirectangular:

The camera can take stills or video. The short clip below was made by setting the camera to take low quality images every 5 seconds and put them together as a video (time lapse).

In this mode, the camera ran for about 2 hours and made a very small video file. This has potential, but the resolution needs to be increased as the pixels are smeared during the conversion process. Taking individual photos instead of time lapse video should improve this.

In full video mode, the camera would need to be mounted vertically somehow to place the center of rotation along the payload access of rotation. Then with some computing power the video could be un-spun to produce a stabilized 360 movie.

The Pi's on board landing prediction software down-linked suggested coordinates in the telemetry, but the act of placing those on a chase map was cumbersome. A python script to load this information to the Habhub map bombed out, and even after being edited during the chase, would still not place an icon on the Habhub map. Also, the landing area was on the edge of 3G coverage, so as the cars drove through the countryside, the Internet would cut in and out.

Even with these issues, the Pi's on-board prediction software was fantastic. It guessed within 1km of the actual landing area a full 30 minutes before landing!

The chase crew was in the general landing area minutes before the landing. The two chase cars took up positions on intersecting roads and each ended up a km or so from the actual landing site. Tall trees obstructed the view of the landing.

The chute caught in the top of a tall pine tree, and the payload plowed down through the branches and finally hung 10-15m off the ground. The chase crew spoke to a native and gained access to the area and then loaded up on bug spray and gear. Once found, Paul activated the nichrome cut-down system and the payload fell -- less than a meter. While the "Directed Arborial Recovery Node" part worked, the "Integrated Tether" snagged. And worse, the spool of string in the "Outside Hung Compact Rain-Activated Pull-down" popped out on a branch and also tangled up. It would be a week before the next rain would drop the spool to the ground. And the payload was too high for the crew's pruning pole.

The team ended up throwing sticks at the payload! Brian and his 'lucky stick' hit the payload but it didn't budge much. Eventually Brian managed to get a water-bottle-tied-to-a-string up and over the payload and it was pulled down. There was much rejoicing.

Sadly, the Mobius Mini camera was accidentally shutdown during payload packing so we don't have any video from it. But the Pi took 140 flight pics (and plenty more close-up images of a pine tree).

Recorded temperatures during the flight. The outside temp sensor (wired to the Pi) was covered in white heat-shrink tubing to limit solar heating. The Pi had a second temp sensor in the small cardboard box that held the Pi, LoRa, and camera -- this was a relatively empty box, within the payload, but with no foam insulation. The AP510 temperature sensor is on the AP510 board and nestled in black foam within the payload.