So it looks good when “far away” but horrible up close when I want to model.
I extracted some objects since I can’t share the model, and then meshed them to see if that would make a difference, but no. So here are some of those meshes:
Obviously I can move it closer to origin, but then I can’t sync it with the architect. So sometimes we need to be able to work “here”.
No objects far from origin.
No bad meshes.
No blocks hiding in the bushes.
(There are some bad meshes in the original file, but nothing culldegeneratemeshfaces can’t cure, but none in the file I included)
Hi Jorgen it looks like there are coincident, or near coincident faces - the larger white mesh and the tan ones - these are .01 apart - I’m just guessing these are fighting in one level of depth in the display.
Thanks, but that is still 10x away from the file tolerance, and the mesh display shouldn’t struggle with these distances since the view bounding box isn’t large (IMO) Thanks for poking!
Hi Jorgen - yeah, my guess of the moment was that it is not tolerance related so much as depth related but I see it does look better nearer the origin. I’ll poke some more but I suspect it will need a bigger brain than mine. (These all look better if they are unwelded too, at about 30 degrees, but, different thing…)
Also noticed hat if I copy the obiects to the origin and stick them on a different layer, and turn the layer on and off, my current view of the origionals changes - which brings me back to my first guess - this has to do more with the extents of the scene, possibly, than with messy meshes.
Here is an exploration of the problem. I do not know if there is a problem here or not, but this is a look into the mechanism that is causing the display inaccuracy.
It is not file tolerance, but video card Z-buffer resolution. When faces are really close to each other video cards can have trouble with near co-planar meshes. I think most Z-buffers only have 125 layers to sort objects int across the depth of the camera frustrum.
The images below show the problem with the two meshes not hiding correctly. The viewport below shows the orientation of the camera and the deep frustrum the perspective view contains. So, trying to sort the hidden with the long joined meshes is causing trouble
As a test then unweld and explode all meshes, which allows better sorting of objects. Then delete some of the meshes that are growing the frustum. The image below shows the same view, but the camera frustrum is much smaller and the hidden view is much better:
There very well could be a problem here. It is the camera frustrum that is changing.
I added more information to the report RH-77172. The fact that moving it makes a difference.
Here is the model in both locations. This is the model at the original coordinates, see how large the camera frustrum is compared to the model in the upper left of the viewport:
Thanks for poking!
