Originally Posted by bjhoogs
Not sure if this will help at all, but...
I think what is meant is that it takes some time for the accelerometer to stop registering the horizontal acceleration.
When a multicopter tilts forward it will accelerate forward. We know this because it will transition from stationary to moving forward. Eventually it will reach it's maximum speed for the given motor speed, this occurs when the drag is equal to the force propelling it forward. At this point it will continue on at a constant velocity and will not be accelerating.
An accelerometer measures Acceleration. When the multicopter is not accelerating the force of gravity is the only force that registers on the accelerometers (and it can tell which way is up). When it is accelerating forward there are components due to gravity and due to the acceleration of the craft registering in the accelerometers. Using accelerometers only there is no way to tell which components of that acceleration are due to gravity and which are due the craft accelerating, you have to wait for it to stop accelerating.
A 'draggy craft' is one that can change atitude (tilt) without accelerating.
Yes, you're correct and I understand the physics of what you described.
However, I need to go back to the original quote:
It needs some speed to get air resistance
Therefore the autolevel works best on draggy crafts.
It implies that some type of air resistance, perhaps from steady state airspeed, turbulent air flow from "draggy" frames, a laminar flow, prop wash, or any type of air flow etc... would have a determination in the speed of correction by the auto-level feature. To imply that, isn't correct.
The KK2 board doesn't receive any input from a pitot tube or other source, The initial explanation could lead someone to believe air pressure or air flow has something to do with it. It simply does not.