Note from the editor: This story highlights a team of IGS employees that put a SQL server in space*. Why did they do this? They just wanted to see if they could do it.
Without giving away any spoilers, the result was pretty awesome.
This piece was written by Drew Furgiuele and originally appeared on his blog, port1433.com. Check it out to stay up-to-date on the latest and greatest from the HAASP project.
Julie and I just had pulled into the Birch Bark Canoe Livery to stop and get our bearings. Without much of a place to pull over on the narrow country roads of western Urbana, this seemed like a nice wide open area to stop. The rest of the team had fanned out to different locations to the north, west, and south of the city. According to the APRS map interface, the balloon was still climbing and was passing almost directly over current location. We hopped out of her minivan to both stretch our legs a bit and to watch the skies. Our predictions had the balloon bursting pretty much any minute. The Kenwood TH-D74 in my hand blurted out modem tones every 30 seconds or so with data packets coming back from our transmitter on board the payload. On the online map we watched the balloon continue it’s track westbound, over our location. The flight was just coming up on two hours, and based on our calculations we had an expected burst time around 130 minutes.
And then… five minutes of silence. The APRS packets stopped; I wasn’t receiving any data on my handheld transceiver, and the online overlay wasn’t getting any updates. There wouldn’t be many reasons why the transmitter would go off the air, but they were all bad. We held our breath.
And then my phone lit up. Rick was the first to realize it, and he alerted the group via text. “It popped,” he wrote. That meant that somewhere, almost 100,000 feet up, our 7 foot diameter balloon had finished expanding to 28 feet, and exploded. Free from lift, it was now returning to earth. Other team members started confirming it by looking at the data coming back from APRS. The overlay started showing altitude numbers descending, but they were a far cry from the gradual numbers we saw when the balloon was on the way up. HASSP-2 was coming back to earth, and it was coming back fast.
Flash back to earlier that day, when we arrived at Neil Armstrong Airport, we were eager to get to work. The last time we were here, we were amateurs with no real experience inflating a weather balloon. This time, well… we still were a bunch of amateurs with no real experience inflating a weather balloon. Our experience last time, though, did teach us a lot. After we licked our wounds from the last launch we all sat down and conducted what amounted to a team retro of the launch. We identified a lot of areas we could have improved, and this time we made sure to change what we could.
The airport manager, Ted, met us there. He had already cleared us out some space in one of the hangers to work in.
Inside here, free from the outside wind (which even outside was much calmer than last time) we got to work. To save some time, we pre-assembled a lot of payload train for the balloon, and also used a pre-fabricated harness and carabiner that we could quickly slide onto the capsule once we were ready for launch and then attach it to the train that contained the radar reflector, parachute, and balloon.
The rest of the team, myself included, started inflating the balloon. Our biggest failing from our last launch was an inadequate amount of lift. As you can imagine, I was overly concerned about making sure we got it right this time. Working indoors, along with a better way to measure lift, were key to making sure we got it right.
We took quite a few readings to try and account for any variations, and when we were done we ended up with a little more lift than we calculated, but not by much. We then proceeded to tie off the balloon. This is by far the most stressful part of the entire operation. Weather balloons have what amounts to a very long, thick neck where you inflate them from. Unlike, say, a party balloon, you don’t just knot the end and call it a day. Instead, you need to tie one end of the neck off to prevent the escape of the helium, then bend the neck back. Then, you tie that off. You’re left with a “U” shaped bend at the bottom of the balloon that you then attach your payload train to: the parachute, then the reflector, and ultimately the payload. Oh, and you have to do this while, you know, trying to wrangle a 7 foot diameter balloon.
The final step was turning on all the on-board electronics: the APRS transmitter, cameras, and of course the on board SQL Server. With everything up and running, the team walked the balloon and train out to the runway (or: about as far away from power lines as possible). We counted down, and then let go…
The entire collection of things shot straight up into the sky. With no clouds to serve as a reference point, I had initial moment of panic; it didn’t look like it was rising. My teammates, though, were convinced. And the APRS data started coming back too, and the altitude number was going up.
We retired to the hanger and to clean up our stuff and run some final predictions based on our final lift and weight numbers. Our predicted track didn’t change much, so after we loaded up our vehicles, we dispersed to separate locations around the city of Urbana, OH. The plan was that with us spread out, someone would probably be near where the payload would land.
“It’s slowing” Jacob texted us. Around ~40,000 feet the air thickened up enough and the chute started to do it’s job. The tracker still had it moving East and coming down fast, but at a much safer rate. I finally exhaled.
Rick chimed in next “We’re are in downtown Urbana, headed south.” The track was moving southeast now, and with the capsule coming down rather quickly (but controlled), we had to move to where we could see it quickly. If the payload landed without us seeing close to where, it would make recovery difficult. This mostly has to do with how the APRS network functions. Below a certain altitude, the signal wouldn’t be heard by any repeaters and thus not forwarded to our online overlay. If I got within a reasonable distance with my handheld, I would hear the signal and it could guide us in. That initial landing, however was key. We needed to get as close as we could.
The search teams scrambled towards the south, with eyes on the sky. Bob, who was riding along with Rick, had a brief sighting, but lost it in the sun. We officially had our last repeater hear the signal at 11:18 local time, just as the track started to snake back to the west. We drove to where we thought the final resting place was based on the track, which was Middle Urbana Rd.
Next, we pulled out the equipment. We had the drone available to help scout the location and look for the chute. The terrain wasn’t the best; it was a little hilly, with lots of trees and some dense corn fields. My radio was also ominously silent. That could mean a few things: one, did the payload land in such a way that the antenna was impaired and not broadcasting at full strength? Had the crash been rough, and incapacitated the payload? Or, was our prediction off and we were just too far away?
I took my handheld transceiver and waded into a cornfield, just to see if I could hear it. Bob, meanwhile, spotted something “suspicious” in a nearby tree (but it turned out to be a swing). It was then that we found a hand warmer on the road. My heart sank; this was the last thing we wanted to see; after all, what are the odds of finding one on the side of the road in the middle of August, near the track our payload was following? “Is it still warm?” I asked Paul. In response, he put it to his cheek. “Sort of.” He replied. Matt took a look, and realized it wasn’t same brand as the ones we used. We were relieved, except for Paul, who put some random hand warmer on his face. Who knows where that thing could have been.
Sadly, we walked back to the parked cars. We were starting to lose a little confidence, but we stopped and took a second look at the data we had. We retraced the route, and did some quick calculations. We determined that based on how fast the payload was coming down and moving (ground speed), that we were slightly off; the payload probably landed one road over, on state road 92. We loaded up the cards and headed about a half mile to the east of where we already were. Back in Julie’s minivan, I rolled down the window… and I was immediately greeted by the chirp of the APRS beacon. We again pulled over in a long driveway and determined that the signal was coming from a cornfield right on the side of the road, and it was approximately .2 miles away from our location.
After getting permission from the local farmer, Matt, Ryan and I set out for another nearby cornfield. Late August in Ohio means high corn; so high that it was even over Matt’s head. Stalk after stalk battered us as we moved through. Meanwhile, Bob was back on the drone controls looking to see if he could help steer us in. As we got deeper in the corn, my radio kept giving us constant bearing and distance. We counted off the distance: .2 miles, .1, .05, .02… until…
Jackpot! The chute was caught in the stalks, but the payload was on the ground safely. Mission accomplished!
We hurried back to the cars. The team already had the streaming station set back up, and we quickly opened up the capsule. Everything survived, but the only thing still “running” was the beacon. Our cameras were out of juice, and worse yet, the SQL Server was powered off. We also observed a lot of condensation on the battery pack, the APRS transmitter, and even on the heat sink of the Joule. We were anxious to take look at the data and camera footage, and the fact that we just, you know, put a balloon in heckin’ space, we felt the need to celebrate. We hauled our asses back to Dublin to a local watering hole not far from work, and started ordering rounds. After the day I had, I needed it.
It was about after my second drink that the gravity of both what we accomplished and didn’t accomplish started to set in. The camera footage is breathtaking; as of this writing we’re still working on processing it, that’s how much footage we have. The APRS data also was a big win; we were not only able to track the balloon real time, but we also were able to export the APRS data and work with it too. This includes plotting our actual flight path and altitude profile.
The team was thrilled, and I was too…
… kind of.
Even though we launched a balloon, tracked it, and have video footage of damn near all of it… I wasn’t satisfied. And part of me still isn’t; after powering on the SQL Server and looking at what data we did capture, it looks like we only have about 30 minutes of flight data. We aren’t sure why.
Still, what we did, and how we did it, I wouldn’t have changed a thing. When I got home later that night, I cried. Not because I was sad, it was because I was so thankful. A lot of people helped make this project what it was. And every single person means something special to me. They support me in so many ways, and I’m proud to be their coworkers and friends. And maybe that’s the real goal here: camaraderie.
They are also just as excited as I am to try again, and I wouldn’t want to without them.
The big list of “thank yous”
First and foremost, I am so thankful to both the entire team and their families for all the time, effort, and passion they put into this project with me. When I first said I wanted to do this, I don’t think I could have gotten the response I did anywhere but IGS. These people are the best and brightest I have ever worked with. The picture above is missing a few team members, but they were all equally important in helping me plan and execute on the project.
Special thanks also goes to Ted, the airport manager at the Neil Armstrong Airport. Ted was just as critical to our success as the core team, but being such a gracious host for our launches and even helping with FAA notices as well. He even let some of the kids play around with the flight simulator they have there (which, I admit, I wish I would have time to play with, too)
I’d also like to thank Brent Ozar for helping out by donating two GoPro cameras to the project as well. I didn’t solicit donations from outside for the second launch, but Brent just up and did it. The footage they captured was nothing short of amazing, and I can’t wait to finish processing it to show it off.
Finally, I need to thank everyone who followed along, gave encouragement, and cheered us on. This mostly took place on twitter and in the #hassp channel on the SQL Server Community Slack channel. Seriously you guys, it meant a lot to have the people I did there to watch this with us.
Until next time…
So what happens now? Well, since we didn’t quite get all the data collection we wanted, the team and I are sure we’re going to launch again. We’re not quite sure when, yet, but I’m sure I’ll get the word out when we know. SQL Server got closer to space on that Saturday in northwest Ohio, but it didn’t quite get where it needed to go. Next time though, it will. I also really enjoyed learning what I did, and what it took to make this all work. I’d love to share my thoughts about it, so if you ever see me anywhere and want to learn more, I’d love to talk about it.
If you missed the launch, our live stream is still available to watch on YouTube. It features our custom overlay of the flight, and even our drone cameras of the launch and recovery. It’s really cool to see how this all came together so you should give it a watch. When we finish processing the videos from space, we’ll put them out there too. And finally, I’ll also make what data we did collect available on the site as well, along with the APRS data. That’ll be on my Github.