We describe a hydrogen-filled weather balloon launch in central North Carolina, and present a video, still images, and data from our GPS data logger.
Launch and Flight Video:
This flight was designated “Jake 7”, as it is our seventh tracked balloon launch attempt (5 successes, 2 failures so far).
Our goals: get video, test using hydrogen instead of helium for lift gas, try launching in close proximity to the ocean without losing the balloon, use a plastic rather than nylon parachute, and have a faster descent rate than on previous launches to cut down on flight distance. We were also excited to get latitude, longitude, and altitude from launch to landing (our most recent flight before this one was deliberately cut down at 20,000 m, and that time the GPS only started tracking at 10,000 m). We were not going for a spectacularly high altitude during this launch because the jet stream was blowing towards the Atlantic at 100 miles an hour.
The GPS flight data can be downloaded here: jake7_flight_data.txt
Results at a glance:
-Hydrogen is great. It’s cheap-currently 1/3 the price of helium and weighs less, meaning more lift. I got a 200 cubic foot hydrogen tank for about $70, including 10 days of tank rental. Hydrogen is also easy to find. You can pick up a tank at a welding supply shop such as this one in North Carolina.
-Hydrogen is also very flammable and much more dangerous than helium. Driving with a 200 cubic foot tank in the back seat is nerve wracking to say the least. This may be a downside for the faint of heart.
-By overfilling our balloon we got a very fast ascent rate. This cut down on our altitude (we made it to 79,000 ft, in contrast to 88,000 on Jake 2), but the upside is we didn’t lose our payload in the ocean.
-Got video and still images throughout the entire flight, but the payload was spinning very quickly. The raw video makes one seasick and a lot of the stills are out of focus. The edited video is the best I can do. The spinning issue needs to be fixed in the future.
-The plastic parachute was too flimsy and tore apart during the descent, resulting in a pretty hard landing. More on this later.
-The Arduino Uno and GPS shield recorded data through the entire flight. Here’s an altitude versus time plot (see the raw data link above):
a 600 gram weather balloon from Kaymont
a 200 cubic foot tank of hydrogen
one SPOT satellite tracker, so you know where the balloon went
two lunch boxes
tubing to move hydrogen from the tank to the balloon (1 inch inner diameter, if I recall right – measure the tank outlet)
plastic parachute (I will describe how I made it, but it failed! this is a “what not to do”)
Canon still camera, configured using CHDK to take pictures every 10 seconds
Kodak PlaySport video camera
I used a Python script I wrote to predict where the balloon would go based on the weather forecast. Since we had a strong (~100 mph) jet stream, we found that there was a good chance of a water landing if the balloon ascended or descended slowly. So we filled the balloon with a lot more gas than usual so it would rise quickly.
After release, the payload swung back and forth violently. Sometimes it was almost parallel to the ground. This swinging motion was probably because the balloon was rising quickly through a fair amount of wind shear. We watched the balloon disappear into a partly cloudy sky, and tracked it for about 20 minutes with the SPOT. After that, it was above the SPOT maximum altitude, so we had to wait and hope it talked to us on the way down.
At high altitudes, the payload box spun rapidly. This made for some nausea-inducing video! The burst is audible, however, and the camera swings up briefly just after the pop. You can see the expanding ball of plastic shreds in the YouTube video. Pretty neat!
As the payload descended, the video camera swung upwards and pointed at the sky. The still camera swung down and looked straight at the ground. This is probably because the still camera was heavier, so the payload was off balance.
The still camera took excellent pictures until it went lens-first through a cloud. All the pictures are foggy from then on.
The last photo before the cloud:
The payload started falling at about 150 miles per hour. As it descended, it encountered denser air and slowed to a fifth of its initial velocity. I designed the parachute to slow everything down to 20 mph – but the thin plastic I used ripped during the descent, so the payload fell 10 mph faster than expected. Everything survived just fine except for a crack in the Arduino case. In addition, the lens cover no longer closes on the Canon camera.
We were very fortunate during the landing. Had the balloon burst any later, the payload would have landed in a forest full of gigantic trees. Instead, we found our lucky pink lunch boxes lying in a fallow cotton field in Selma, North Carolina – about 50 feet from the forest margin.
The moral of the story is: make a strong parachute. I used the thinnest paint dropcloth I could find at Lowes (this material is what we use to make our giant solar balloons) to save on weight and also because I had it lying around already. However, this plastic rips easily in the best of conditions. Falling at 150 mph through thin, bitter cold air is certainly not a good place for it to be. Next time, we’re going for the rip stop nylon parachute we used in our first weather balloon launch.
The take home message:
-Hydrogen is cheaper and lifts better. We will be using it from now on, and thinking non-flammable thoughts as we do.
-Parachutes should not be made of plastic, and should be tested before deployment. I think I will hold one out the window of the car at 20 or 30 mph. If it can survive that, it will work (note that it falls 150 mph at 80,000 ft, but the air is thinner so the force is the same).
-A fast ascent seems to lead to bad image and video quality. In any case, though, we need to think hard about how to make a stable camera platform.
We’ll leave you with a couple more cool plots – one showing ascent and descent rate, another showing the windspeed vs elevation, and a final one showing the flight track. All of these plots were made using the data set included in this blog post.