David VE3KCL launched the first flying U3 that I have heard of. He launched two small hydrogen-filled party balloons on 03-Jun-2015 at about 1230Z. The payload consisted of an Arduino Nano board (for weight saving) running the U3 firmware v3.08 with WSPR and JT9 transmissions; with GPS module, Si5351A Synth, solar cells, battery, power management module, and antennas. Projected altitude was 7500m. However at approx 4,233m (according to maximum height in the reported telemetry) it appears that one balloon burst, bringing the balloon into a gradual descent. Flight time was about 2.5 hours and covered 68km.

Summary:

Balloon: 2x mylar party balloons, filled with hydrogen
Payload: LiPo battery & converter, Arduino Nano running U3S v3.08 firmware, Si5351A kit, GPS, 30m half-wave dipole, GPS dipole antenna
Payload weight: 37.5g
Free lift: 3-4g
Projected equilibrium height: 7,500m
Bands: 20m and 30m (14.097050MHz and 10.140150MHz respectively)
Modes: WSPR and JT9
Telemetry: JT9 packets containing latitude, longitude, height, battery, temperature, GPS status & satellites
Flight duration: 2.5 hours
Flight distance: 68km
Max. altitude: 4,233m reported

What an inspiring project! David provides many details of this innovative balloon and the flight. Some extracts are below, with photographs and further description in the following section.

Details

Today we probably set an altitude record with a U3 at 4233 meters ASL. While less than the projected 7500 meters it was a good first try. The Balloon flight however, was cut short by what I think was a party balloon failure... two mylar balloons were tied together to provide enough lift, and the way the U3 reported the decent rate, it appeared as it was coming down with only one balloon inflated.

... It appeared that one of the balloons was not of the greatest quality... one was over inflated and could not be deflated due to the nature of the internal gas seal...I replaced one of the two balloons and used one that had a rather questionable envelope......I won't do that again. I was determined to get the thing off the ground...

...My brother had offered his old motor home as a base station The U3 had run all night in the motor home on axillary battery power and was nice and stable ..pulling in 8 or 9 satellites and the U3 was transmitting telemetry to a dummy load , which was being decoded in all but the most severe JT65 Qrm. The Launch was at about 12:30 gmt and it landed about 15:00 gmt. traveling about 68 kilometers in a straight line

...Maximum height reported was 4233 meters above sea level where the balloon failure probably occurred. The antenna was previously untried and constructed from very light .1mm enameled copper wire crazy glued to dyneema ( kevlar ) fishing line. When the antenna was hooked up and in flight the Si5351 produced a 20mw +27db signal . The full length 30m vertical dipole weighed about 2 grams.

The payload including solar panel U3, antenna,lipo battery, battery control module, boost converter and gps weighed about 37.5 gms. The balloons were filled with hydrogen to provide about 3 to 4 grams of free lift to the payload.

The U3 ran a 12 minute frame, and to save power the gps was turned off for 10 of the 12 minutes. An Aux Output command to D4 (aux 0) which when pulled low turned on a pnp 2n3906 that supplied power to the gps. Your code was loaded onto Atmega 328 Nano for weight saving and the 16mhz crystal was changed out with a 20mhz so the software would work. Unfortunately in the rush, a final picture of the flight model interior was not taken, so the previous version, with changes, is included in the picture files.

The format of the transmission frame was as follows 00 CWID aux 0 (gps turns off after cwid) 01 JT9-1 #T1 aux 8 02 JT9-1 #T2 aux 8 ( #T2 did not decode but did change with position) 03 JT9-1 #LT #A5 aux 8 04 JT9-1 #TN #A3 aux 8 05 JT9-1 #M6 #AT aux 8 06 WSPR 10.140150 13 aux 8 08 WSPR 14.097050 13 aux 8 ( minute 10 to 12 gps turns on and calibrates. This is done by using a aux 0 in the next transmission). The small dipole on the gps was quite adequate and the #T1 telemetry showed up to C3 which would indicate tracking of 12 satellites and a 3d lock. The small gps dipole was much lighter than the amplified ceramic antennas found on most gps units. The 20mw WSPR transmissions were reported as far as Western Canada at about 3300km. 

[Note that the Si5351A module output was connected directly to the antenna, with no PA - generating approx 20mW RF output power].

Interesting conclusions and points to note

Some parts of this project are particularly interesting to me. It validated the U3 for balloon applications, and the use of WSPR and JT9 as different mechanisms to transmit location. Several points are worth identifying, I think.

1) The use of the Si5351A connected directly to the antenna without a PA. This provides 20mW of RF output which proves adequate for balloon tracking purposes (see also Andy VK3YT balloons which use 20mW for worldwide tracking). Most importantly, it reduces power consumption, therefore smaller batteries can be used, smaller solar cells, and the whole thing drastically reduces the payload weight, reducing cost and complexity. 

2) The installation of the U3 firmware on an Arduino Nano is an interesting way to cut the weight of the payload. Obviously QRP Labs would prefer that people abstained from such practices in general ;-)

3) The use of hydrogen as the balloon gas is also atypical, most hobby balloons seem to use Helium. The use of Helium may be somewhat controversial. Although it is the 2nd most abundant element in the universe, it is relatively rare on Earth and it has a limited and fixed, non-sustainable supply. It has many critical medical, scientific and industrial applications. Arguably it should not be wasted, where possible. On the other hand, Helium is inert: very conveniently, it does not explode. Hydrogen is even possible to manufacture at home using electrolysis; but it is explosive and therefore safety issues need to be considered, in respect of storage, handling and use. 

4) The JT9 telemetry was broadly speaking successful, but the #T2 tag did not decode and also the temperature readings are suspect. There may be bugs in the firmware afftecting these data items, which will need further investigation.

5) The WSPR transmissions generated a huge number of reports, at thousands of km distances. Some deficiencies in the WSPRnet database can be identified. David VE3KCL also had a WSPR station transmitting at the home QTH. The reports from the balloon can be identified because of the lower power than the home station transmissions, and also because the balloon transmitted at different (and known) minutes past the hour, compared to the home station. 

6) Interesting use of the Aux parameter functionality for switching the GPS module on/off for power saving! 

7) I was particularly interested to see the use of the wire dipole antenna for the GPS receiver unit, instead of the usual ceramic patch antenna. This provided some weight saving, and was clearly perfectly adequate for very satisfactory GPS reception. 

Path and flight data

This map shows the launch site North-West of Toronto, altitude approx 440m ASL. A total of ten JT9 pairs of JT9 coordinate packets were received containing latitude and longitude. The final position report at 15:03 was at a height of 823m. Extrapolation of the direction and speed (horizontal and vertical speed) allows a fairly precise estimate of the landing site (red circle).  

The height profile of the flight. The initial 440m height reflects the fact that the balloon was operational at the launch site overnight before the launch. Descent was fairly gradual, supporting the theory that one of the balloons had burst.
LiPo battery voltage during the flight. The solar cells were highly effective at quickly taking the batteries to full charge voltage. 

Photographs

Previous model to the flight version, which had the following changes:

1. Connectors routed out the bottom
2. Hole in the middle of the foam for attacment wire
3. Battery and boost converter changed places for balance
4. Connection pins removed for weight
5. Foam thickness changed for weight saving.
6. insulation around Si5351A crystal assy. 

Solar panels constructed from eBay cells, foam insulation wire cut, GPS dipole and tiny coax (from dead laptop), U3 under solar panel in foam covered with food wrap and silver foil. Panel put out 4.5V @ 250mA to charge Lipo battery through a diode. Yellow wire to hang U3 on upper dipole arm and support lower dipole arm. Antenna wire: 0.1mm enameled copper crazy-glued to dyneema (Kevlar) 20lb fishing line. Antenna weight 2g.
Old propane tank converted for hydrogen filling ball point pen end for filling balloon.
Two party balloons were filled with hydrogen and tied together to give about 3 gms of buoyancy to the payload (U3 and antenna).
The U3 is in the centre of the dipole antenna beneath the party balloons
Receiving station in brother's motorhome. You can just see the JT9 transmissions in the left waterfall, very wide due to the 27dB signal from the U3 as it was still in sight.
Vertical antenna constructed from 24' extension painting pole with 4 ground radials and static drain inductor. Party balloons are the black dot near the top of the vertial.