Pitch Bounce back

[ResinSmoker]

When tuning for 6S LiPo's, seeing a rather large amount of Pitch bounce-back regardless of Terms used or filter settings applied. (See red circled image below)
https://farm2.staticflickr.com/1744/41849129875_de83dfa757_h.jpg

In the above Plasmatree image, the trace should quickly rise to "1", with minimum overshoot and then flatten out. This graph shows the relation between the PID inputs and the what the Gyro experiences in the real world. From what I can tell, the Pitch and Yaw are overshooting causing oscillations on the other Axis. After about a week of testing various PID combinations, I'm thinking that maybe the Yaw compensation while using a 6S-LiPo is off somehow.

The video example is shown with increasing values of D by "2" for each frame. D-Term 17 shows twice as one is with stock filtering and the other is using Dynamic filtering. (Pitch "baseline" starts at 9/10/0) Yaw smooths out as Pitch improves, which is normal for a Tricopter.
https://youtu.be/YgQUPb1tRBE

Quick test using 4S-LiPo, please notice the same bounce-back is present just not as strong as before.
https://farm1.staticflickr.com/892/42032384814_2828c2f135_h.jpg

Here is a video with I & D at ZERO and P scaling upwards.
https://youtu.be/WMkpUDe8STY

Here is an animation of a recent test to slowly increase P & D proportionally by a value of "2".
https://youtu.be/m1zagikCYNE

Thus far I've completed in excess of 60 - 80 test flights for Plasmatree.
Most of which can be viewed here with log files:
https://drive.google.com/file/d/1J7MG7rEGm9G5l-QOjyDGtQk5mqs7V0I_/view?usp=sharing