Missions‎ > ‎


Flew the evening of Oct 2, 2020 in an attempt to video two rocket launches.

Predicted flight paths of Antares NG-14 and SpaceX GPS-3 SV04 as modeled on FlightClub.io

Paul L's box from NSL-86 flew with the following changes:
1) The Mobius Mini camera was replaced with a Caddx Dolphin HD Camera (inexpensive low-light camera)
2) The carbon fiber rod was removed and replaced with 10m of string.
3) A thin sheet of styrofoam was attached to the side of the box, to act as a "wing".

The addition of the styrofoam should change the angular momentum of the payload.  But how would it dampen the swinging on a particular axis?  If the balloon is dragged quickly to the northwest, will the wing act as a sea anchor and cause the payload to point itself in that same direction?  Or will it all just move along in the same air mass?

With dry weather forecast for the evening, the parachute line attached to the payload with PVA string.  This could allow the payload to fall from a tree after a rain.

The two rocket launches were scheduled within 30 minutes of each other that evening -- The Antares would launch from Wallops Island, Virginia; and Falcon 9 from Florida (with a drone ship landing off the coast of NC).  With a somewhat strong jet steam in the area, Paul chose a long flight path from Aberdeen to Rocky Mount, NC.  An old Kaymont 600g cell, loaded with whatever H2 remained in the tank (expected 1400L or 50cu ft); should work to image both flights.  Chris G. volunteered to be Public Address Officer.

Prep and launch from a public park in Aberdeen proceeded without issue.   Although, the tank only gave up about 1300L of gas, yielding a longer flight than originally planned.
The above simultaneous images of balloon launch, from the Caddx and RunCam cameras, show their ability to see in the dark.  The Caddx slipped somewhat while packing which explains the plastic visible along the edges of the images.

     Click on images to expand

     Equivalent images taken shortly after launch.  

The Caddx color HD camera shows much more detail -- Mars can be seen next to the moon.  But its images tended to blur more when in motion due to a slower shutter rate.

 Northeast NC from 13km up

 View southeast with Jupiter

 View west and the glare of RTP lights

The Caddx runtime was 2 hours, 40 minutes at room temperature.  But lasted only 1 hour 20 minutes at -40C.
At 2113EDT, the built-in battery for the Caddx ran out and the camera shut down.  Three minutes later, the Antares NG-14 rocket lifted-off from the coast of Virginia.  This was caught on the remaining RunCam camera.  Paul noted that there was at least 10 minutes of "wasted" video taken prior to launch, if he had "just turned on the camera at launch" he would "have gotten HD video of Antares".

  <--  Click to view GIF of Antares launch

For those interested in the raw RunCam video and associated Google Earth track, download here (175MB).

Sadly, the SpaceX launch aborted at T-2 seconds.

As the flight slowly ascended, telemetry showed that the old AP510 battery was running out of power.  The LiPo had seen several dozen eventful flights and could no longer handle the cold.  As the voltage dropped, the unit shut down.  This left only the LoRa T-Beam tracker, which recorded burst at 31,281m.

  <--  Click to view GIF of burst

  Habhub map showing descending payload, chase car, and landing predictions

  LoRa receiver out the chase car window

The value of landing prediction algorithms
The T-Beam LoRa 915MHz tracker, with its stock antenna, is limited to under 40km reception range at altitude; and only about 1-2km when on the ground.  So the receiving antenna in the chase car needs to keep up with the flight.  LoRa is also subject to frequency drift due to the extreme temperature changes.  The simple Arduino receiver could not compensate automatically (unlike the Raspberry Pi version) and must to be manually adjusted to keep in tune.

For this flight, the chase car could not keep up with the strong winds.  The last packet clearly received by the chase car was 22 minutes prior to landing, while the payload was still over 9km up.  The chase car was just too far away on slow country roads for clear LoRa reception.  Thankfully, each LoRa packet also contained an estimated projected landing coordinate; in this case a further 20km to the northeast.  In previous flights, these on-board estimations were shown to be pretty accurate.  As the payload was out of contact, the only thing to do was to head to the predicted landing area 45 minutes away.

Paul eventually reached the projected landing area. With fingers crossed, he slowly drove the country roads with the receiver's Yagi antenna hanging out the car window.  After a few moments, the software's received signal strength meter showed something on a nearby frequency.  Paul pulled over beside an open field and adjusted receive frequency -- a clean packet came through.  That was enough to identify the true landing location -- a mere 1.6km from the guess made much earlier in the flight.
  Accuracy of predicted landing sites during descent

It landed in a tree farm.   Being almost midnight, Paul headed home to sleep.

The following week, he headed back out to the site and met with manager of the farm.  After some bushwhacking through briars, the payload was located at the top of a young (7m) pine tree.

        AP510 on-board data

  T-Beam data logged at Habhub

The piece of styrofoam worked surprisingly well.  The video shows that spinning was dampened and that the cameras were typically forced to point in the direction of travel.  Occasionally the payload would spin 180 degrees, but would soon be righted back so that the wing trailed the direction of travel.  This gave the impression of a very calm flight.  To see an example, download the 175MB zip (link halfway down this page).   Further tests are needed.  Perhaps an 'X' of foam-core could be constructed to dampen in multiple directions.