Displacement problem

[corestudio]

Hi. We're having a problem with displacements. See attached. We didn't initially notice the problem when we were rendering half res, but as you can see when we switch to full 1080 renders the problem is quite obvious. Seems like the displacement is getting clipped in some strange way when we render full res. Any ideas? running up against a deadline on our first vray job and getting nervous!

displacement sources are high res 32bit tif files.

(not sure why attachments aren't working so I'll use drop box)

https://dl.dropbox.com/u/4327868/vray/960.jpg
https://dl.dropbox.com/u/4327868/vray/1080.jpg

[bazuka]

there is an option on file node->att->vray-> allow negative colors

maybe this can help?

[corestudio]

Thanks. But hmm. Why different results at different resolutions? I'll give it a shot, but there are no negative values in the displacement image.

[Morbid Angel]

It appears to be not so much the negative color as much as the float value is getting clipped. If you add displacement control attributes to the geometry, there is clipping bounds there, which specify at which range the float value is terminated. You need to sample the highest float value in your displacement image for example it will be 3.4 and set your far clipping to be higher then that (but not too high) like 3.5 Also as bazuka mentioned, the negative value is important, though you may not see it, all float images contain positive and negative value, which must be accounted for.

[corestudio]

Thanks for the reply Dmitry. So you're talking about setting the Displacement Bounds to Explicit, and then setting the values in the swatches, correct? Would this result in the renders looking different at different resolutions? That's the big mystery to me. Same floating image for displacement, but different results depending on the *output* resolution.

I've been using Normal Displacement as I wasn't able to get nearly the same detail with 2D Displacement. These maps are 8K

re float images containing positive and negative values. you mean they *can* contain both positive and negative, right? there is no requirement that they have negative values. These maps have been normalized to 0-1

Thanks for the help. I really need to get this sorted out soon

[Morbid Angel]

well if your map has been normalized to be 0-1 then it may have clipped the highest point...check that its normalized properly.

[corestudio]

yeah I think what's happening is that the normalization is so extreme that it's effectively killing data even at 32bit. Or at least vray sees it that way. Before I left for the night I started experimenting expanding the map from 0-1 out to 0-4 or 5, and compensating with the Amount (? I think) value. The detail started to come back. The resolution thing was throwing me for a loop. Thanks for pointing me in the right direction.

[MoonDoggie]

I think that's a bit of a stretch? Killing data at 32bit? You mean there's more than 4 bytes of data per pixel?

1111 1111 1111 1111 1111 1111 1111 1111 would effectively be 16,777,215 (dec) per channel. That's crazy normalization. Maybe your displacement generation is odd. I would try a linear format for your displacement map (.exr should be fine).

[corestudio]

That's why I said "at least vray sees it that way". The normalization was in nuke taking 0-~800 data values down to 0-1. I just changed the math to get it to work, so normalization is now 0-100 and adjusted the Displacement Amount and equivalent Min/Max values in Displacement control. Works now. Agreed about the unlikeliness of data being lost, but there you go. From what we now understand about vray's displacement, it should have worked with the values we had in there (i.e. maps normalized to 0-1, appropriate Displacement Amount and Min and Max values set 0-1) but it didn't. And it didn't ONLY when we rendered at full res. I can't stress this enough. The R&D done at half res all looked right. ONLY when we rendered full did the clipping show up. Seems like a bug no matter how you slice it, but we're new to vray so maybe we're just a bunch of tools.

Some of the maps are .exr already. We did those experiments too wondering if tif vs exr would make a difference. There is no difference between the .tif and .exr files when they're interpreted correctly.

[vlado]

You might also want to try a newer nightly build; we had a bug where displacement could disappear close to the camera on dense meshes.

Best regards,
Vlado

[MoonDoggie]

Aye. Just checking all possible problems. Welcome to VRay btw.

[Morbid Angel]

you might also want to try a newer nightly build; we had a bug where displacement could disappear close to the camera on dense meshes.

Best regards,
vlado

yes we did!

[corestudio]

Sorry to be dense, but where are the nightly builds?

[vlado]

Please email us to vraymaya@chaosgroup.com and our support guys will explain.

Best regards,
Vlado