Well, first off, the new external thermometer mounting works much better.  The temperatures are about what you would expect for the altitudes.

So what happened here? 

1)  The glider cut away from the balloon, recovered into level flight, and attempted to trim out any roll rate gyro error.

2) A period of spiral-diving occurred, which we on the ground mistook for problems with navigating in the high winds.  In fact, it was due to inaccurate roll trim.

3) During that time, the auto-pitch trim feature attempted to pick up speed to fight the high winds.  However, the elevator response was 2-3 times stronger than anticipated, so it over controlled.  G's built up quickly, and the bank was probably over 70 degrees at times.

4)  After we on the ground guessed what might really be wrong, the autopilot was commanded to try trimming itself out again in roll.

5)  After the wings were leveled, the great amount of speed built up (200 kts or so at 24,000 feet ASL) resulted in a series of loop-de-loops.  That not being enough excitement for it, it mistook the lack of progress into the wind to mean it should pitch down yet again, causing a drastic nose-under bunt of over -3 G's, and possibly even an inverted loop or two.

6) Finally sorting itself out, it began to correctly fly straight and level to the commanded landing site.

7) While on its way to Saltspring Island to land, we told it (not realizing the pitch issue at the time) to try pitching down just one notch, to be sure it would have the airspeed to make it to the landing site.  From the ground station, the result seemed to be that it squirmed around a bit, then sorted itself out quickly at the new trimmed speed.  In fact, the elevator response was so sensitive that the one extra notch made the wings lift negative - so it did a quick little bunt, came out flying level in the wrong direction, and turned itself around to go where it needed to so quickly we didn't really notice from the ground.  It then flew the last 10km or so happy as can be, upside down.  From the ground all we saw, was that it was going where it was supposed to, and the G meter read "1.0".  Nobody noticed the negative sign in front of the one, although we did wonder why the glide wasn't quite as good as it had been earlier.

• The airframe flies well throughout its design envelope in positive and negative G's, at speeds of up to 200 kts, and altitudes of up to 35,000 feet.
   
• As long as the roll-trim is moderately good, the autopilot and roll-control system can fly the glider well at high speeds and moderately high altitudes.  In fact, the speeds inadvertently demonstrated are as high as those it would experience during a normal flight from 60,000 feet.
   
• The roll-trim algorithm needs some adjustment to make it more accurate in conditions of strong winds and low groundspeeds.
   
• The elevator response is much more sensitive than notes from manual flight testing had indicated.
   
• The CL and airspeed versus elevator response, and glide performance, are now known for a large range of trim points, including inverted.  The data is better and covers a broader range than we would ever have obtained on purpose.
   
• Always make negative signs very visible.  Those unassuming little dashes can be important.

• Open up the quickcam's iris a bit more.
   
• Reduce the elevator trim volume.
   
• Adjust the roll-trim out algorithm..
   
• Use a different algorithm to keep the roll gyro from drifting.  There's too much hysterisis involved in trying to compensate for temperature changes at present.
   
• Negative signs on the ground display are to be bright flashing orange.