I believe Mr. Pendergrass' observations may be correct. It appears a software change made to the yaw axis of AHRS that use the digital magnetometer, possibly combined with certain (normal) magnetometer errors, could result in small roll oscillations that are not affected by changing gains. Our website now describes this, a fix has been implemented (software), and is pending a flight test before it is released. I expect to have this improvement available no later than the end of February, 2017.
Mr Pendergrass' observations of not flying as "tightly" or "smoothly" are consistent with this problem.
I apologize for not picking up on this more quickly. Since we had not changed any autopilot software, I had attributed this to other causes. Our autopilot normally performs to such a high level of performance that we routinely are complimented by airline pilots who report it as equal to their Boeing autopilot experience. I personally consider a high-performing autopilot essential to single-pilot flying, especially in RV's and IFR. The fix will restore this level of performance.
During the investigation, attention was also focused on the magnetometer accuracy and calibration. This resulted in a slightly improved factory magnetometer procedure. If you had difficulty achieving accurate magnetic headings after performing fine magnetometer calibration, you are welcome to send us your digital magnetometer and we will re-calibrate it with the new procedure at no charge. This is not likely to affect most installations however.
Greg Toman
President
GRT Avionics