I have a servo gimbal which is 3 axis, which has the roll and pitch stabilized by the pixhawk.
It does react and the system works but since it is handmade and also not at the center of rotation of the frame, it shows some sort of undercorrection or overcorrection.
This is easily adjustable by adjusting the min/max angle ranges or the PWM endpoints. but, adjusting it just right is pretty impossible, partly because it is pretty difficult to judge difference by eye
So is there some way to make this process easier? maybe using another imu on the final stabilized platform to make it similar to how a brushless gimbal (STorM32) works?
For the time being, an esp with a mpu6050 with a filter and some decently heavy averaging, I calibrated it upto sub degree level inaccuracy, after which it is impossible because of servo resolution
You could google around for fpv headtracker code. There are various of these around. You just need to get it to do the opposite sense. Here is one based on an old one by Denis Frie which I used some years ago.
Since servos in the gimbal may react slower to position/angle changes in the vehicle’s roll and pitch as the vehicle moves about a target, the camera shot may have some visible lag in it. This can be reduced by using these parameters to have the gimbal outputs move a bit ahead of the movements of the vehicle.