NSL-61

NSL-61 was a short test flight flown on April 6, 2018. The main purpose was to test two 434MHz LoRa ground stations and antennas. Paul, Michael, Robert, and Matthew L. participated. Michael ran the home "ground station" in Apex, while the others launched and chased.

Payloads

The flight train consisted of two separate payloads strung along a 9m line. For this flight, a 1.5cm-wide orange ribbon was used as the line. This was to see if ribbon would, in any way, reduce the constant spinning/unspinning present on flights that normally just use string (mason line). The ribbon began at a small, 300g balloon, passed through a 68cm round parachute, and then down to two payloads. Each were spaced about 3m apart. The entire ribbon, chute, and payloads came to 350g.

The upper payload was a 434MHz LoRa tracker similar to those designed and flown by Dave Akerman (http://www.daveakerman.com/?p=2101). It was basically a Pi-in-the-Sky LoRa tracker with the ability to send two way data and down-link camera images.

It consisted of a Raspberry Pi Zero with a RFM96W (or 98W) 434MHz LoRa module and a Ublox 6 GPS module. A cheap AA battery powered USB phone charger was attached for +5 volts (http://www.daveakerman.com/?p=2182). A piece of paper was jammed between battery contacts to act as a power switch -- it was pulled out to activate the unit before flight. The venerable Energizer Ultimate Lithium batteries were used as they can handle the cold.

Paul has had issues with GPS interference in the past with this rig. Apparently, this GPS module generates noise and interferes with other GPS trackers, so this payload was shielded and placed 3m away from the second payload box. Also, the cheap phone charger boost circuit causes RF noise as well. So it had to be wrapped in aluminum foil so that its own GPS would work. The entire package, except for the camera lens, GPS antenna, and LoRa antenna were then wrapped in foil and grounded. To take advantage of solar heating, the unit was then wrapped in black foam and then clear box tape. The whole tracker unit came to about $41.

The lower payload was connected to the end of the ribbon and held just an APRS tracker and video camera.

It was simply a small plastic box with an AP510 APRS tracker and Mate808 video camera. These devices shared a single 4200mAh Lipo (re-flight of the hardware from NSL-60). They were nestled in black foam within the translucent box to create a warm greenhouse. An Outside Hung Compact Rain-Activated Pull-down (white spool of string held with a PVA thread) was attached outside to assist in tree recovery.

LoRa Gateways

Two, practically identical, LoRa receivers were built -- one for a chase car, and one for home. These would receive the 434MHz signals and gate the telemetry and images over to Habhub.org (http://www.daveakerman.com/?p=1905). They also had the capability to transmit commands up to the flight. Both gateways were slapped together in a day and totaled less than $50.

Each gateway consisted of a $10 Raspberry Pi Zero W (has WiFi) and an attached 434MHz LoRa module. A GPS module could have been used for the car, but wasn't in this test flight.

The car unit received its primary power from an attached phone charger bank. This battery acted as an uninterruptible power supply and was topped-off periodically from car power. The Pi connected via WiFi to a cellular MiFi hotspot in the car, allowing the Pi to go directly out to the Internet while traveling down the road. A separate chase car computer, typically used for maps and telemetry, could remote into the Pi as needed.

The home unit was simply stuck out a house window. It received power through a USB umbilical that just plugged into a USB phone charger in the house. This Pi connected to the home WiFi. A home PC could remote into the Pi as needed.

The Flight

Due to a strong easterly jet steam still affecting NC, the team chose to fly from a park to the west, in Siler City. This was to be a short/quick test, so a small 300g cell was used. It was filled with Hydrogen to provide 1500g of neck lift for rapid ascent. The projected flight path would take it just south of the Apex ground station and land it just east of Wilson. This track was faster than a car could travel the same distance, so this was a great test of distributed ground receivers.

Friday was to be a breezy day, so the team raced to get it launched before the ground winds got too high. They had to struggle a few times as gusts tried to yank away the filling balloon. Ground tests completed, NSL-61 was launched just before 0930EDT.

The payload shot up through the pollen-laden haze while the chase car gateway received telemetry and images. At 2km in altitude, the Pi automatically switched from sending tiny thumbnail images to slightly larger 320x240 pixel images. Below are samples of Pi images down-linked through the chase car gateway during the first 15 minutes of flight.

[Note from Paul: The haze obscured the images; perhaps the Pi should be configured to tweak contrast/gamma in future flights. Interestingly, the Pi-in-the-sky software takes multiple images during each transmission cycle. It chooses which image in the group to send by picking the largest file size -- as black images of space or washed out images of the sun may result in smaller JPEG files.]

As the flight progressed, the payload started to pull-away from the chase car (a fast food stop may have contributed). As it gained distance, the received image data started to break-up with CRC errors. The chase car antenna was swapped between 1/4-wave and 5/8-wave but no difference was noticed. The much shorter telemetry packets often survived, but the large image packets suffered more and more errors.

The winds remained turbulent the entire flight, tossing the payloads violently. This may have contributed some to packet loss as the LoRa antenna was thrown horizontal. Video shows this as one of the most violent flights in years. The ribbon stood no chance in dampening movement as the payload was yanked around. The ribbon idea will have to be tried again on a calmer flight.

As the payload quickly approached Apex, the home gateway started picking up data and handing it off to Habhub. This required a bit of coordination though, as the transmitter frequency shifted slowly as the payload ascended. This shifting is normally handled by a gateway's automatic frequency control feature. But the home station needed an idea on where to start looking for the signal. Once in the ballpark, the home gateway's AFC took over.

Both gateways were initially configured to gather a list of missing packets and then relay this info back up to the payload for re-transmission on a specific frequency. But this became problematic as the frequencies drifted. The downlink and uplink frequencies were initially hard coded to 434.450MHz. But during ascent, the payload drifted up 6kHz. The AFC corrected this, but whenever the ground gateways decided to uplink, they would re-tune back down to the lower frequency and then lose lock. It became annoying to have to manually tweak the frequency every minute -- losing packets in the process. The team ended up simply turning off the uplink feature on both gateways. In the future, the gateways can be configured to simply uplink on whatever the downlink frequency happens to be.

Landing Prediction Telemetry

The on-board landing prediction software was nothing short of amazing! Within moments of burst, the Pi was transmitting its predicted landing area -- from several counties away. At 60,000 feet it guessed its own landing area to within 2.3 miles! Looking through the data afterwards, it is interesting to see the Pi recalculating the parachute's effectiveness ( See http://www.daveakerman.com/?p=2059 to learn more).

The chase crew was behind the payload due to the high wind speed, but they received some of this telemetry. As a result, they set their driving way-point to the highway interchange in the image above. It landed not far away.

Recovery was a breeze. It landed just off the road in a bare field.

One of the concerns with using spring-type battery holders, is that the shock of burst or landing can disconnect the batteries for a few milliseconds. This has happened before. Taping the batteries down in the holder can exasperate the issue as the mass of all of the batteries combine. To limit this, the batteries were first covered by a slip of paper and then taped down -- allowing each battery to independently slide slightly on its springs (Soldering the batteries helps, but it is a pain to do properly). In addition, a capacitor was placed in the boost circuit to limit brownouts. This cap apparently wasn't large enough as the Pi took a power hit and rebooted upon impact.

Ascent rate and temperature data from the AP510. Sadly the temperature data from the Pi just read 26.1C the entire flight.