The barometric altitude given by the glider still diverges somewhat from the GPS altitude with height, although less than in the first launch. 

This is probably still a calibration error, which has been tweaked (again). (Note the x-axis at right is scaled in seconds)

The G sensor appears to have worked great. This graph is of the max G loading for the most recent 5 seconds, from 4hz AD samples. (Note the x-axis above is scaled in seconds)

There was some wild swinging this time on the way up, as you can see on the left end.  The maximum G load at pullout from the dive (at about 750s) after release was 2.8 - right about what the flight simulation predicted. 

The swings in the cruise portion of the flight mostly correspond to turns, when you look at the second-by-second G loadings, although much is likely turbulence. Most of the return flight time was spent in a banked turn of some kind.

The parachute opening is actually the second-highest spike on the right end, at only 1.93g (very nice!).  The high spike on the right, at 4.5g or so, is the touchdown, while the low spike is next to it probably someone picking the glider up. 

Looking at a small-scale expanded section of the flight data, you can see that the roll rate commanded by the autopilot, and the sensed roll rate, track each other very well.

Also note that the roll command algorithm is always making small control motions to fight turbulence.   (Note the x-axis above is scaled in seconds)

The roll-rate gyro trim was set to be - 4 degrees per second, to zero the sensed roll rate, during the preflight.  Note how much it differed at altitude, when the glider trimmed out after release! 

The internal temperature did rise 10c over that period, which might affect the gyro, but still - that's a fair bit of variation.  It isn't, however, outside of the expected range of gyro neutral drift. (Note the x-axis on the left is scaled in status-telemetry data points, which are 4 seconds apart)

The quickcam also needs attention, something that didn't make it on the work plan for this flight.  Other issues were that the attempted fix for the external temperature probe didn't fully improve it's accuracy.

Overall though, just a fantastic flight. Almost there.

• Use a narrow, fixed, iris in the quickcam's lens assembly instead of "sunglass" type filters, which seem to flatten contrast and colour.

• Put the external temperature sensor out on a small post to separate it better thermally from the aircraft's skin.

• Provide an algorithm to allow the Expert-system level autopilot to adjust the elevator for the ideal trimmed airspeed for the conditions. 

• Provide audio alarms on the ground station, so it's possible to walk away from the table for short periods.