Tangent Basis of Blender does not match xNormal. (Both use mikktspace) Which is 100% correct? #27908

New Issue

Andy Davies · 2011-07-08T18:57:16+02:00

Andy Davies commented

2011-07-08 18:57:16 +02:00

%%%Hey,

I have been doing some tests and noticed that Normal maps that are produced in Blender do not match ones that are created with the same models in xNormal 3.17.5.
As im sure you are aware, xNormal now uses the mikktspace by Morten S. Mikkelsen, so it should match Blender 100%
Im wondering if this is due to an error in xNormal, or in Blender, or due to different ray casting techniques for both of the programs

I have attached ta difference map, along with a zip file that includes the rendered maps from each application, along with the .blend, the LP and HP meshes used for baking.

Any ideas?

Blender 2.58a
Windows 7 64bit
8GB RAM
Nvidia 460GTX%%%

%%%Hey, I have been doing some tests and noticed that Normal maps that are produced in Blender do not match ones that are created with the same models in xNormal 3.17.5. As im sure you are aware, xNormal now uses the mikktspace by Morten S. Mikkelsen, so it should match Blender 100% Im wondering if this is due to an error in xNormal, or in Blender, or due to different ray casting techniques for both of the programs I have attached ta difference map, along with a zip file that includes the rendered maps from each application, along with the .blend, the LP and HP meshes used for baking. Any ideas? Blender 2.58a Windows 7 64bit 8GB RAM Nvidia 460GTX%%%

Andy Davies commented

2011-07-08 18:57:16 +02:00

Changed status to: 'Open'

Brecht Van Lommel commented

2011-07-08 19:26:17 +02:00

%%%There are some differences on the edges, but that has nothing to do with the tangent space, just difference in baker implementations. Also some pixels +1 -1 but again that's just because they are two different applications, you can't expect them to give 100% the same output.%%%

Andy Davies commented

2011-07-08 19:50:03 +02:00

%%%Thanks for the reply Brecht.

When you say edges, do you mean the edges of the islands? I expected those to be different :)
I was concerned more about the centre section of the island that it top left, and the centre vertical lines of the Bottom island, as they show quite a difference.
Would you put this down to the difference between the applications?

Thanks for looking%%%

%%%Thanks for the reply Brecht. When you say edges, do you mean the edges of the islands? I expected those to be different :) I was concerned more about the centre section of the island that it top left, and the centre vertical lines of the Bottom island, as they show quite a difference. Would you put this down to the difference between the applications? Thanks for looking%%%

Brecht Van Lommel commented

2011-07-08 20:05:46 +02:00

%%%I think that too is just difference between bakers, the tangent space is just one of many factors influencing the result. The main reason to have matching tangent spaces is so that you can render the resulting normal map the same way. For baking there's more things that can be different, which point on the low poly mesh corresponds to which point on the high poly mesh is not well defined, it's something that I would expect to vary a bit.%%%

Brecht Van Lommel commented

2011-07-08 20:05:46 +02:00

Changed status from 'Open' to: 'Archived'

Brecht Van Lommel closed this issue

2011-07-08 20:05:46 +02:00

Andy Davies commented

2011-07-09 01:10:37 +02:00

%%%ok. Thanks for taking the time to look at this :)%%%

Morten Mikkelsen commented

2011-07-10 06:48:09 +02:00

%%%Hey Andrew,

What Brecht is saying is absolutely correct. I just thought I'd add a few extra details. It is true that there's no way you can expect two different bakers to
pick up/sample the same field of normals from the hi-res. The advantage to a shared tangent space is that you're able to import and correctly "decode"
the normal map (ie. use it for rendering) such that your viewer can reconstruct the original captured normals. So for instance you can bake the map in Blender and use the result in the viewer in xNormal and vice versa works too.

Other things you need to be aware of is that if xNormal splits quads along a different diagonal then Blender then you get different interpolated tangent spaces.
So to be on the safe side, especially when scrutinizing results like this, it can be best to triangulate in Blender before exporting and baking .

Another thing you should know is by default xNormal calculates its own vertex normals which will affect tangent space too.
So you should choose "use exported normals" in xNormal where your lo-res is selected.

%%%

%%%Hey Andrew, What Brecht is saying is absolutely correct. I just thought I'd add a few extra details. It is true that there's no way you can expect two different bakers to pick up/sample the same field of normals from the hi-res. The advantage to a shared tangent space is that you're able to import and correctly "decode" the normal map (ie. use it for rendering) such that your viewer can reconstruct the original captured normals. So for instance you can bake the map in Blender and use the result in the viewer in xNormal and vice versa works too. Other things you need to be aware of is that if xNormal splits quads along a different diagonal then Blender then you get different interpolated tangent spaces. So to be on the safe side, especially when scrutinizing results like this, it can be best to triangulate in Blender before exporting and baking . Another thing you should know is by default xNormal calculates its own vertex normals which will affect tangent space too. So you should choose "use exported normals" in xNormal where your lo-res is selected. %%%

Andy Davies commented

2011-10-07 16:20:48 +02:00

%%%Thanks Morten :)
Sorry for the late reply...I missed the notification email for this.
I love digging deeper into this and it is great to actually understand why there are difference between bakers that go further than just the normal tspace/swizzle. I think its good for artists to have a solid understanding of these "behind the scenes" factors that directly effect their work. :)
Cheers guys!%%%

%%%Thanks Morten :) Sorry for the late reply...I missed the notification email for this. I love digging deeper into this and it is great to actually understand why there are difference between bakers that go further than just the normal tspace/swizzle. I think its good for artists to have a solid understanding of these "behind the scenes" factors that directly effect their work. :) Cheers guys!%%%

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Tangent Basis of Blender does not match xNormal. (Both use mikktspace) Which is 100% correct? #27908