Incorrect roll value in 'real life'

[fralbo]

Hello, First, I don't want to say that your calculations are wrong. In fact, there are perfectly right on table. Maybe it's not really the right place to talk about the particular problem I encountered but I didn't find a place where to address the right persons. Tell me if I should write it somewhere else.

I tried to use MPU6050 to measure roll angle on a motorcycle and I was surprised by the results shown on the image below: In blue are the measured angle, in red the theoritical one.

On this image we can see that the roll increases (in absolute value) but, as we maintain the angle during the turn, we can see that the measured angle decreases to an intermediate value. The peak we can see at 409 and 834 certainly occurs at the end of the turn when we straighten the motorcycle. In fact, just after engaging the turn, the acceleration oriented through z axis increases whereas y and x, if the speed is constant, stay at 0.

If think the fact that the z increases, causes the problem. This issue is certainly present on plane and drone as they turn in the same way as a motorcycle.

Did you already encounter this issue and how did you address it ?