In the end, I found the payload after two days of searching in the Pawnee Grasslands of Colorado. I wound up with a 25 minute video which shows only the beginning of the ascent, a bunch of flight data to parse through, and a lone logged position waypoint from the GPS at 19,000 meters (~62,000 feet). 

Early on July 30th we set out to Ault, CO to find a launch pad. After some searching, we found 'East Ault City Park' which is basically a dirt pile next to some warehouses.

Natalie filled the balloon while I prepared the payload. This is a shot from inside the payload box with two 125 ft^3 helium tanks in the foreground. I used about 175 ft^3 of helium. 1 ft^3 of He has about 28.2 grams of lift, so 175 ft^3 has about 4900 grams of lift at sea level, actually slightly less at higher elevations. The balloon weighted about 2280 grams and my payload weighed about 2500 grams. Combined this is less than the 4900 grams, which is what you want for lift. 

You don't want to fill the balloon too much since the balloon will not reach it's full height and pop too soon. If you fill the balloon too little, you might not have enough lift. Afterwards, I realized I filled the balloon too much. The actual weight of the balloon was much less than 2280 grams, since I cut the skirt off of the balloon. This is why there is shredded rubber on the underside.

Another picture from the box right before launch. 

Here is the video of the flight, nothing spectacular. I have no idea why the camera cut out after about 25 minutes, which means I only have video up to about 10,000 meters (33,000 feet). :( 

The repetitive clicking you hear is the 1 Watt Xbee Xtend radio interfering with the microphone on the camera, the same thing would happen to my computer speakers during testing.

This is the Spot locator track, which shows the entire flight. Notice the bubble highlighted. That is approximately where the camera cut out. As you can see the flight went much farther and higher than what was filmed; about 1.5 hours longer out of 2 hours. The ascent probably spanned the lower half of the 'Z' track, with the upper half of the 'Z' track corresponding to the decent. The green shaded region is the Pawnee National Grasslands.

My flight computer's GPS cut out at about the same time as the camera. Although amazingly, I did record a lone waypoint after an hour at about 19,000 meters! This corresponds to a position close to waypoint 12 in the Spot track above. So I assume the balloon went to at least 20,000 meters. I only wish my camera was still running. 

Here I am tracking the balloon with my radio about 10 miles outside of Ault. I tracked the balloon for 45 minutes in one spot, well beyond the the failure of the camera and GPS. 

Here are the last three tracks I received from the Spot locator. This was really weird because the last waypoint is a 'Check-In' message, which means that one of the buttons on the unit was hit! I assumed this meant there was a crash landing, but why wasn't the Spot locator still reporting. I wasn't completely sure if this was the true landing point or not, since the Spot locator does not give elevation for some reason. 

We immediately went to the last waypoint. It took us about 30 minutes to hike to what we thought was the waypoint. Our searching was not successful. We returned to the car completely tired from driving and hiking all day in the heat. We decided to go home and analyze the positions I had recorded from the Spot locations, where we searched, and the data I received from the radio. 

Once I got home, I calculated the time of the flight and it seemed on par for a full flight, meaning the balloon should of been coming down close to the last waypoint. The flight also closely resembled the modeled flight (CU Flight Predictor). I also made sure the 'Check-In' notification was in fact due to a button hit. After I figured all of  this out, I realized I had to go back the next day and do a better search.

The next day, my Dad and I drove out to the grasslands to search again. I realized that the day before we were off by about 500 meters and were not looking in the correct spot. I headed directly for the spot where I now thought it was located and after about 30 minutes of walking I saw the payload in my binoculars, success! The picture above shows you the conditions I was searching in, dry, hot, barren grasslands. (Actually, the grasslands are anything but barren. To the naked eye, you see nothing, but look through some binoculars and there is wildlife EVERYWHERE. I kept thinking the white asses of the pronghorns and antelope were the payload box, damn nature. 

When I found the payload, I was so excited I forgot to take pictures and I immediately ripped open the box to see what was on the camera only to be completely disappointed that I didn't capture the entire journey. I did notice that when I approached the payload, the flight computer, GPS, and radio were all ON and working! The parachute wasn't tangled and the payload box had a small depression on the corner and that was it, it didn't seem like a hard landing at all. 

I reviewed the logged data and plotted the final descent of the payload. It looks like the GPS cut back in at about 2000 meters (about 600 meters above ground). This was very interesting, since now I could see how fast the payload hit the ground. I calculated about a 3-4 m/s descent rate during that time, which is about what is expected with a working parachute.

Top left is the parachute, top right is part of the shredded balloon, far right is the payload, bottom middle is the flight computer, batteries, camera, radar reflector, and Spot locator. The object on the bottom left is the receiver antenna. 

For the radios I used two Xbee Xtend 1W 900MHz transcievers. I choose a directional Yagi antenna for the receiver, since knew the general direction of the balloon. I choose a clover leaf antenna for the transmitter (see photo below), so that the radiation pattern would only be focused in a downward 35 degree cone. This proved to be a problem because the payload seemed to swing a bunch (it also spun a lot but not as much as I expected) and the signal would cut out periodically, due to the payload swinging away from me. Otherwise the radio worked well, considering I didn't actively track the balloon. 

The flight comptuer consisted of an ATmega328 powered at 3.3V, an ADXL345 accelerometer, an HMC5843 magnetometer, a BMP085 temperature and pressure sensor (that was buggy throughout the entire flight), a Copernicus GPS module, OpenLog, and status LED. The OpenLog recorded everything that was put out on the radio onto an SD card. In retrospect, I should of additionally included the raw GPS NMEA sentences.

I also had a one second timer running that time stamped each message, which was very helpful.

The radio power supply was also on board and was electically isolated.

The fact that when I found the balloon, the flight computer and radio were working fine makes me think the issue with the GPS was bad view of the sky with the embedded ceramic patch antenna I used. Next time I will use a helical GPS antenna, so direction of the antenna won't matter as much. 

There was a bug in the code for the temperature and pressure sensor that I still can't figure out, even after verifying the I2C bytes and ACKs from the sensor and replacing hardware. Most likely, there is a variable overflowing, since there is so much typecasing going on in the code. However, the night before, I thought the problem could be that I was missing a makefile flag somewhere for floats or a math library. I was doing my development in Linux and the makefile I was using was minimal, so I compiled in Windows using a different makefile and everything seemed to work great. Obviously, when I went to launch the balloon, the sensor crapped out again.

Here is all of my code, PCB layout files for the flight computer, log file from the openlog and the receiver, and Spot locator track.