Aurora Borealis: Visions from the Stratosphere

In October of 2015, we formed Night Crew Labs and set out with the ambition to take video of the northern lights from a high altitude balloon. After many local test flights in California, several design iterations, and much planning we went to Alaska in March of 2017 to do just that.

The Night Crew, consisting of Bryan Chan, Ashish Goel, Tyler Reid, Corey Snyder, and Paul Tarantino (with special guest Adrien Perkins), arrived in Fairbanks, AK on March 17, 2017 and began our balloon flight campaign. We planned for up to four separate balloon flights, allowing for several day and night launches over Fairbanks. We chose March, near the vernal equinox, as it is the optimal time to see the Aurora Borealis due to the right combination of clear skies and space weather near Fairbanks. When 8 tanks of helium arrived at our location, we knew this was for real.

This slideshow requires JavaScript.

We spent a total of two and half weeks in Fairbanks. The first week was spent doing reconnaissance and testing equipment. Neither our flight cameras or us had ever seen the northern lights, so we had to make sure we had our settings right. We were fortunate to have high auroral activity (kp’s of 4 to 6) and clear skies, giving rise to some of the most stunning views of our lives.

Since we were up all night chasing aurorae, we used what little daylight hours we had while awake to scout out and plan launch and landing locations that were accessible by road, at least as best we could. It gets rural fast outside of Fairbanks itself, but the winds were (at least somewhat) in our favor.

We flew a total of 3 balloon flights over Fairbanks, two during the day, and one at night to capture video of the aurora.

This slideshow requires JavaScript.


Launch & Recovery

To film the aurora, we had to launch the balloon at night. The earliest we could launch was around midnight, as it takes hours for the lingering sun to finally set. Even the slightest amount of light could have overpowered our sensitive camera and ruined the shot. As with our two prior day launches in the area in the days leading up to this launch, the prevailing winds meant all flights had to be launched approximately 10 miles west of Fairbanks on turn-outs off the side of the mountain ridge highway. This was the only way to stand a chance of landing within a few miles of one of the few roads in the area.

Our day flights were met with great success where our two GPS tracking systems worked flawlessly, giving us confidence to fly our more expensive night rig. Both day flights landed within a half mile from a road, making recovery straight forward, even while breaking trail with snowshoes. The night flight, however, was met with many challenges. First, we lost radio link. This was not unexpected, as it was overcast that day and clouds have caused signal loss on past flights. Then, we were dismayed to see that our satellite link wasn’t working either. It turned out that the larger and more metallic night equipment blocked the signal to the satellites due to our antenna placement. At such a high latitude, the Globalstar satellites we rely upon can be low on the horizon, making signal acquisition even more of a challenge. We sped to the estimated recovery site, hoping to get lock from our radio but we weren’t so lucky.

This slideshow requires JavaScript.

We went back to basecamp, slept a few hours, and came up with a plan. Our only hope of recovery was to fly over the area and search. We hired a small local plane to fly a search pattern over the possible landing area. Luckily, our pilot had experience searching for balloons and sounding rockets at the nearby Poker Flat Research Range. From experience, he figured we’d have a 50/50 chance of spotting it. We took off from Fairbanks Airport with binoculars in hand and hoped for the best. Our saving grace, however, was our radio. It was still beaconing its latitude and longitude and we picked up the signal from the plane. We were lucky. We had landed in a dense stand of trees and never would have spotted our bright orange parachute from the sky.

With the coordinates in hand, we planned a route to hike in and recover our payload. There was a problem: it was 6 miles one way, 2 miles of which were off trail through snow more than knee deep. Determined, we drove to the Wickersham Trailhead and stated on our way. We made good time for the first four miles, but slowed to a crawl through the deep snow in the last two. As the sun was setting, we found our payload, still beaconing after more than 21 hours. With darkness setting in, we started the 6 mile hike back at a brisk pace. Just after midnight, we made it back to our truck. In total, it took more than 24 hours from launch to recovery. We were cold, but we had found it and were ecstatic. For more about this adventure, check out our recovery video below!


Camera System

To capture the relatively faint aurora borealis at night on video, we needed a low-light capable camera system. We decided on the Sony a7S: a low weight, full-frame mirrorless camera with incredible low light sensitivity. Before the Alaska mission, we decided to flight test the Sony a7S performance in the stratosphere. In 2015 and 2016, we launched separate high-altitude balloon missions, testing the camera’s capability. After stellar results, we decided to continue baseline it as the flight camera to capture the aurora. Along with the camera, we used the following supporting equipment to make up the camera payload:

  • Sony a7S (Mark I)
  • Rokinon 24mm, f/1.4 lens
  • Atomos NInja Flame 4K Recorder
  • Varavon External Battery
  • Neat Video (software plugin)

This slideshow requires JavaScript.

There were a couple factors in deciding to use the Atomos Ninja Flame 4K external recorder. Using an external recorder with its own batteries greatly increased the camera’s own battery life. The ability to record over one hour of footage (the minimum duration of a balloon mission) without worry of battery drain was very important to mission success. Lithium polymer batteries drain rapidly and die in cold temperatures, and we were unsure of how cold the entire rig would get (outside air temperature in the stratosphere can reach -50 deg C). Additionally, since we wanted to record the entire flight, the camera internal recording limit of 29 min 50 sec was bypassed by using an external recorder. The performance of the Atoms Ninja Flame was impressive; it performed flawlessly during the flight and actually kept everything else on the rig fairly warm due to its heat output during operation. It ended up filling up its 1 TB hard drive and then idled for the next 10 hours until it finally drained its two batteries.

In order to extend the camera’s battery life, we used a Varavon 10400 mAh external battery. Based on our previous missions, we expected the camera to last over 2.5 hours while video recording to an external source.

A wide and very fast lens was needed to capture sweeping views of the dim aurora. Since we couldn’t predict which angle the aurora would present itself, a wide field-of-view lens was desired to capture as much of the sky as possible. Based on some ground-based, real-time aurora videos (such as Philip Bloom’s excellent aurora/whales vlogumentary), we were confident that the combination of a f/1.4 lens, and ISO set to 51,000 with a Sony a7S would capture the aurora in real-time video. Thus, the Rokinon 24mm, f/1.4 lens fit our needs for the aurora flight. While in Alaska, we performed some ground-based video recording tests using these settings, and came up with fantastic results. For the flight, to reduce the chances of water condensation and/or freezing on the lens, we wrapped flexible electric heaters around the end of the lens barrel. This would raise the lens’ glass temperature a few degrees to reduce local air moisture and prevent any icing on the lens.

Neat Video, a de-noise plugin for Adobe Premiere Pro, was used to get as clean a video as possible. Neat Video seriously does wonders for denoising. Compared to the noise performance of native Adobe creative cloud noise reduction tools, there was no comparison. Neat Video blew everything else out of the water. The one drawback to the software was that it took a very long time to render. One frame of de-noised video took about 20 seconds for my Macbook Pro to render.



The complexity of this mission in the harsh Alaska conditions called for some custom hardware. This began with the NCL Balloon Integrated Re-programmable Computer (BRIC) – Mark II. The BRIC consists of an Arduino MEGA which runs our custom flight management software. We had a custom Printed Circuit Board (PCB) fabricated to allow for ease of integration with other components including a GPS unit, radio telemetry link, barometric altitude sensor, and 8 thermistors (temperature sensors). The GPS helps us determine position, the radio allows us to track the payload in real time, and the thermistors give feedback for our electric heating system needed to keep everything from freezing in the harsh -50 deg C (-58 deg F) high altitude environment.

This slideshow requires JavaScript.

The PCB design also allowed us to easily assemble multiple units, allowing us to bring two complete builds with us to Alaska. This was a necessity, as there was a very real possibility we would lose one in the remote Alaskan wilderness. We also had a day and night configuration, the difference between them being cameras. The day flight configuration flew a GoPro Hero 4 Black, a 360Fly virtual reality camera, and a Google Pixel phone. The GoPro Hero 4 and 360Fly recorded in 4K during the entire flight, while the Google Pixel was set to take pictures every two seconds. The night flight was designed around the low-light capable Sony a7S with an external 4K recorder. All electronics and cameras were housed in a custom 3D printed structure which was fabricated at a discount by RYT Manufacturing in Santa Clara, California. For more details on the build, please check out the video below!



In addition to capturing video of the northern lights, we collected radiation measurements during the flight. For more on the science of the aurora and these radiation measurements, check out this in-depth article by NCL’s Ashish Goel.


Thank You’s

Thank you to Alana Vilagi, Eli Barry-Garland, and Mary-Cathrine Leewis for their hospitality and expert knowledge of both cold weather operations and of the Fairbanks area.

Thank you to Lucas Hulien for his help with logistics before, during, and after our stay in Fairbanks.

Thank you to RYT Manufacturing in Santa Clara, California for their expertise, quick turnaround, and discounted pricing for our project.

Thank you to the folks at the Federal Aviation Administration (FAA) center in Fairbanks and to those at Air Traffic Control (ATC) who worked with us to make these flights possible.