Bug - Meshing don't preview any quad/tri's All  1  2-9

Previous
Next
 From:  tyglik
1563.2 In reply to 1563.1 
I don't know why, however after separating that solid and joining all surfaces again, it works.
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
1563.3 In reply to 1563.1 
Hi PaQ, thanks for posting that - the edge preview not showing is related to the inverted normals.

In the mesh preview, if the object is closed (every edge joined to another edge), mesh edges are not shown for backfacing polygons to help clean up the display and reduce edges that might bleed through thin areas.

However, in this case the orientation of the object is inverted, also inverting that backface detection.

The bug is in some earlier process that did not calculate the orientation of the surfaces towards the outside of the object correctly... Unfortuantely it will be difficult at this point to figure out which operation didn't do that properly.

Doing a Separate and then re- Join like Petr mentions does do a proper job of determining the "outside", that's why it fixes this up.

Please let me know if you run across this again or if you notice any pattern that you notice the inversion happening after a particular operation.

- Michael
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  manz
1563.4 In reply to 1563.3 
Hi Michael,

I have noticed this inversion of normals, but not on such a full scale, and that was from a view point within MoI

I have seen this where I have made boolean, then I notice I can see through the model on some areas and I thought the model was broken, but on checking (using the select solids) did show the model as solid.

I have been trying to find a good ( simple) example of that.

I dont know if this is related?

- Steve
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  PaQ
1563.5 In reply to 1563.4 
Ok thanks for the explanation, and for the tip (split/rejoin).
I think the miss orientation appends after a boolean (merge), but I'm not sure. I'll try to investigate a little bit.
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
1563.6 In reply to 1563.4 
Hi Steve, was the one you noticed also on meshing or on just regular working object display?

That particular thing I was talking about here is just specific to the mesh display.

On the regular object display, it is possible for edges to bleed through somewhat more after a boolean - that is a display artifact that is difficult to correct, not an indication of a broken model.

What happens is for all edges and curves, there is a zbias factor that pulls the z values of the edge forward towards the eye point when the edges or curves are rendered. This is so that the edges don't appear to be partially submerged into the surfaces of the object, which would otherwise cause a kind of broken and stippled edge drawing appearance.

If you have areas of your model that have thin walls, it is possible for the zbias value to be greater than that wall thickness, so that can pull those edges forward enough to be visible even though they really shouldn't be.

It is very difficult to balance this because lowering the bias factor too much results in a bad display of all edges, not just those thin wall ones. So right now this is designed to err more on the side of pulling too much rather than pulling too little.

The z bias factor for a brep is calculated relative to the bounding box size of the brep, so that is why it can change after you do a boolean operation, since that can alter the bounding box of the resulting object.

In the future I am hoping to be able to tune this up somewhat and try to avoid drawing "backfacing edges". The thing is, a backfacing edge of a curved object is not as easy to detect or generally deal with on a curved object as it is on a flat object like a polygon. Unlike a polygon a curved edge can have some portion of it backfacing and some part of it front facing, that doesn't happen on the line edge of a polygon.

- Michael
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  PaQ
1563.7 In reply to 1563.5 
I still have a question about the normal direction.
It's not that I'm completely stupid, but my english level don't help me sometimes to really undertand some technical thinks.

Let's say that an object looks right when I do the meshing, I mean I can clearly see the polygons, the backface curling work well ... however when I open
the .obj in my rendering software (actually modo), the normals are inverted. Is it something that happends sometimes and I have to deal with ? or it's also something
I can call a bug :) ... the problem actually is that I can't really flip the normals in modo because the vertex normal map is not inverted, so I still have something wrong.
I can only force to render everything in 'double side' mode, and that can be a problem later with some transparency surface.
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
1563.8 In reply to 1563.7 
> Is it something that happends sometimes and I have to deal with ?
> or it's also something I can call a bug :)

Hi PaQ - if your object is a fully closed solid like your example here, then that is a bug.

But if your object is not a fully closed solid (if it has some edges that are open and not attached to another surface), then that is not a bug - it isn't generally possible for MoI to automatically determine an inside and outside type region from an open surface.


> the problem actually is that I can't really flip the normals
> in modo because the vertex normal map is not inverted,
> so I still have something wrong.

Can you please report this to the modo developers? When you perform a flip in Modo they should also just flip the vertex normals as well instead of leaving them the same.

I had previously assumed that MoI did not need flipping functionality inside of it because it seemed to me that programs that specialized in polygon editing should be able to handle a pretty basic operation like flipping orientations without messing stuff up....

But that has turned out to be a problem in Modo and Cinema4D as well.

So I would like to add in something to let you see the orientations of open objects and flip them around, I have not quite figured out where to put that yet.

- Michael
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
 From:  PaQ
1563.9 In reply to 1563.8 
Ok now I get it about the fact the object has to be closed.

> Can you please report this to the modo developers? When you perform a flip in Modo they should also just flip the vertex normals as well instead of leaving them the same.

Sure I will !
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged
 

Reply to All Reply to All

 

 
 
Show messages: All  1  2-9