I've been noticing that bump mapping in Octane can sometimes act a little but unpredictably when UVW transforms are applied to scale the bump map. I haven't quite tested it out before but I did take 15 minutes off today to explore what is happening.
From the looks of it, UVW transforms can mess up bump maps really badly.
I'm attaching two images. One has the material scaled at 100% with default UVW transforms (scale = 1). The other image has the material scaled at 400% with a UVW transform at a value less than 1 - I wanted to match the scale on both of these.
As you can see on the second image, something really wonky happens - is it texture filtering or incorrect scaling by the UVW transform node I have no idea. Bottom line it looks borderline corrupted.
Not sure whether I'm doing something wrong here or I'm just being stupid but giving other renderers a spin and doing the same exact approach (with their implementations of UVW transforms) yields expected results where both of the images match each other visually.
@nejck, please submit a scene that exhibits your concern and we will take a look at it. It is the best way to insure that we can see your issue the way you see your issue.
jayroth2020 wrote:@nejck, please submit a scene that exhibits your concern and we will take a look at it. It is the best way to insure that we can see your issue the way you see your issue.
Yep, would be happy to do that! I might be out of the loop here but is there a private uploader or a private email I can use to transfer the files to?
I am writing my reply back in this forum, as opposed to a PM, since it was asked here first and others have expressed similar questions as well. Unlike the images in your post, the file that I received had two spheres next to each other. I am referencing your file, not the posted images. Apologies for that, but I am only dealing with what I was given. That said, for the others reading this, the information I present should be easy to digest, even if you have not seen nejck's scene.
I have taken a look at your file and I have some initial comments with the file that I received (to insure that we on the same page):
a) both spheres had their projections scaled, not just the one on the left.
b) the sphere on the right was scaled by a factor of 10 in the material tag.
c) the sphere on the right was using UV projection, whereas the sphere on the left was using cubic projection. That difference is important when doing an A|B comparison.
d) in addition to the differences in projection, there were also differences in each of the materials transform nodes. The sphere on the left was scaled at .2 -- the sphere on the right was scaled at .005
Some comments:
1. Regarding scaling using tiles in the material tag vs. a transform node in the material itself. Of course, you are free to do this, but please keep in mind what you are asking, mathematically. You are doing some MASSIVE scaling by first setting your material tiles for 4000 on U and V, and then further scaling the map itself down to .005 -- frankly, I see no way how to avoid artifacts when doing this. Even though your texture is very large, it is still pixels, and at some point, you get to diminishing returns where effectively you are reaching the end of any precision that you can expect. That leads to artifacts, such as those apparently exhibited in the sphere on the right in your image.
2. From a "best practices" standpoint, think about using one method or the other to size your image maps. I realize that you can run into limits from time to time, but it is easy to get confused about where you may have scaled your map, or, in this case, just how much you have actually reduced your map. In this case, the ultimate scale would appear to be ridiculously sub atomic...
3. It is hard to discern the validity of this particular example, and I think you were trying to illustrate what you perceived as a problem, however, if this were some sort of actual use, then an image map would be the worst way to go here, especially since the provided image texture was basically noise. Noise nodes are much better for this, and resolve infinitely, for all intents and purposes. As others have said that they have experienced similar issues, please consider using noise nodes if you find yourselves scaling your image textures so such an atomic (or sub-atomic) size.
Now, as to the question of whether or not this would be considered a bug, I am not an Otoy developer, so I can only speak for myself. Given the amount of scaling that is going on, and given that modern processors still have limits in precision, which I suspect this example far exceeds, I would consider these results to be expected. More of a practical limitation than a bug.
Does this make sense? Let me know if there is anything else you would like me to look at.
Yeah, sorry about the different scene example I've sent in, I totally misplaced the original somewhere and to make matters worse I've really dropped the ball on naming it sensibly
What you've noted makes sense, it does and I really appreciate the in depth reply! Still, I think we might be missing the point here a bit. The example scene could be tweaked better to showcase what I mean but essentially if you scale using C4Ds material scaling options (in material tag etc...) I can't get this same behavior going whereas if you do similar scaling to that with the UVW transform node it goes haywire.
Maps that I typically use for noise are at least 4k, typically 8k in size for closeups. Its important to note that in other renderers (albeit I only tested with CPU renderers) I can't quite reproduce this behavior - sure, they don't have UVW transform nodes but if I use a Layer -> Transform to scale the map I don't seem to be getting these types of "issues". And you can scale it to some quite extreme values even though the renderer is doing texture filtering on top of everything.
So I suspect it could be some sort of a limitation of the UVW node or something akin to that. I could be wrong though
