3d tiles - coordinate system of vertices and bounding sphere?


I am converting 3d models from a CAD application into 3d tiles and viewing them in the Cesium viewer widget.
I am using the 3d-tiles-transform branch, and defining my bounding volumes as spheres.

I am defining the bounding sphere and my vertices in the same (world) coordinate system. Is that correct?

I find that if I have e.g. a simple triangle centered around the origin (and bounding sphere centered at the same place), it displays correctly in the viewer.

If I have the same simple triangle offset some distance from the origin (and the bounding sphere offset the same distance), in the viewer the geometry is rendered offset from and outside of the bounding sphere. (I have turned debugShowBoundingVolume on).

I am certain I am missing something obvious, but not at all certain what that obvious thing is...

I will follow up with sample data sets.


Sorry, but I don't see a way to attach files to a topic.
Here are two images demonstrating the issue: http://imgur.com/a/6nOSH

Hello,I am also converting 3d models into 3d tiles(b3dm),but I met some trouble.
Can you explain how you transform the 3d models in detail?

In case anyone else runs into the same issue...
Sean Lilley and Patrick Cozzie graciously followed up via email, and I independently discovered that Cesium assumes incoming models are expressed in a y-up coordinate system, and therefore unconditionally applies a y-up to z-up transform. My models are already in z-up, so I must transform them to y-up during publishing.
The assumption of y-up seems somewhat arbitrary to me and appears to contradict the assertion that bounding spheres and geometry are defined within the same coordinate system.

Xingwang, if you can be specific about the trouble you are having I can try to help.

Regards (and thanks, Cesium team!),