Autorotation results aren't ideal when looking at a single half-period (rather than the full model), since the rotation is around the absolute Z axis and the models tend to live in one quadrant. So it's spinning on an external axis, which isn't very useful.
It's possible this can be corrected through some grouping method in THREE.js and rotating on a relative z axis, but more likely we'll need to compute the center of mass of the x,y values of all the surface and coil points and rotate around that. Or, to be less computationally expensive, we could just rotate around the x, y center of the bounding box.
Autorotation results aren't ideal when looking at a single half-period (rather than the full model), since the rotation is around the absolute Z axis and the models tend to live in one quadrant. So it's spinning on an external axis, which isn't very useful.
It's possible this can be corrected through some grouping method in THREE.js and rotating on a relative z axis, but more likely we'll need to compute the center of mass of the x,y values of all the surface and coil points and rotate around that. Or, to be less computationally expensive, we could just rotate around the x, y center of the bounding box.