Wheat Field flight Gone Bad

Please let me know if you have any suggestions/comments/etc.

I had an interesting flight occur over a wheat field. We were flying with an NIR sensor weighing a little more, and off CG (forward) than I was comfortable with, but tried it anyway.

The results were unfavorable. After put into “auto” it appeared to be tracking the planned mission pretty well until it seemed to get lost. Similar to compass “toilet bowl” effect. it was not pretty, the pilot was able to get the aircraft back home safely in “loiter”. Once in control, “loiter” seemed to be working okay.

The mission was planned last year from a much older Mission Planner version. I believe .17.

We then removed the payload and calibrated the compass and accel. The results were similar. It seemed to be lost.

Here is a link to the log:
drive.google.com/open?id=0BzAok … authuser=0

Auto Analysis

Duration 0:08:05
Vehicletype ArduCopter
Firmware Version V3.2.1
Firmware Hash 36b405fb
Hardware Type
Free Mem 0
Skipped Lines 0

Test: Autotune = UNKNOWN - No ATUN log data
Test: Balance/Twist = GOOD -
Test: Brownout = GOOD -
Test: Compass = FAIL - Large change in mag_field (38.59%)

Test: Dupe Log Data = GOOD -
Test: Empty = GOOD -
Test: Event/Failsafe = GOOD -
Test: GPS = GOOD -
Test: IMU Mismatch = FAIL - Check vibration or accelerometer calibration. (Mismatch: 2.74, WARN: 0.75, FAIL: 1.50)
Test: Parameters = GOOD -
Test: PM = FAIL - 48 slow loop lines found, max 13.60% on line 953
Test: Pitch/Roll = GOOD -
Test: Thrust = GOOD -
Test: VCC = GOOD -