Strange pixel values in EXR

[re:FORM]

I'm not sure what the cause of this is, but I'm getting random pixels being rendered with a value of "1.#inf" and "-1.#ind" when checking using the AE info panel.
Is this due to a problem with my EXR file, my sampling settings, or some other issue?
Incidentally, is it normal for the EXR to be so bright as default? I have to heavily under-expose the image to get it looking the same as the VFB.

http://www.reformstudios.com/03-reso...Climb_0008.exr

[re:FORM]

In addition to this, I'm getting other parts of the image suffering from transparency where I know there's no way that you can see through to the background. Does a low Max reflection/refraction rate lead to pixels being flagged up as transparent? The problem appears to be limited to my tree leaves, and my max refraction is set to 4, so I know the max refraction limit is being hit... I just didn't expect the pixel to default to transparent.

Funny scene this... very heavy in proxies and glossies... GI is all brute-force, and sampling is DMC 2-16 nThres = 0.01.

Any ideas?

[vlado]

You probably have doubled faces on the leaves and the rays get trapped intersecting the same geometry over and over again. If you increase the secondary rays offset a bit (e.g. to 0.001) it may help. Best would be to remove the double faces.

Best regards,
Vlado

[re:FORM]

Thanks for the suggestion Vlado... unfortunately its an old scene and I don't think I have the original mesh's at hand as they were all converted to proxies at the time. I've not had coplaner faces before using onyx but if that's a known problem, then its good to know.
I'll try upping the secondary ray offset and let you know if it helps..

Do you have any idea whats causing the strange pixel values that I mentioned before?

[vlado]

Missing or incorrect UVW coordinates can sometimes cause this. Have you tried to turn on Clamp output in the Color mapping?

Best regards,
Vlado

[re:FORM]

Well, its not coplanar faces as the strange pixels have appeared on a surface that is definately a single surface.
Increasing the secondary ray offset also hasn't removed the problem.

Here's some screen grabs of the offending pixels...



Just testing subpixel and clamp output... Would prefer not to clamp as I like to have the extra headroom in post.

[vlado]

Did you check the UVW coordinates?

Best regards,
Vlado

[re:FORM]

When you say "check the uvw coords", how should I do this to identify problems? The mesh in question was collapsed (I dont have the base anymore)... so I applied a UVWunwrap and it looks OK. No sign of problems.

Initial testing suggests it might be the old clamp-output issue again.... I have a feeling I've been ignoring the most obvious likely cause! I guess I've not experienced this symptom before... normally its been super-bright pixels, not black ones...

It will take a while to render out a few frames to see if sub-pixel mapping fixes the problem, but I'll keep you updated.

Thanks again for your patience!
p.

[re:FORM]

Ok... some progress.
Pixel glitches seem to be fixed by enabling sub-pixel-mapping. Doh!

Strange transparency is solved with the secondary ray offset suggestion, but I need to increase the value to 0.1 (0.09 leaves some alpha artifacts) to get the correct alpha rendering.
I checked the helpfile but there's no mention of any possible side-effects of increasing the offset by so much. Are there any issues I should be aware of when using a secondary ray offset of 0.1?

Thanks for helping me sort this out.

[vlado]

There might be weird light leaks or shadow offsets, but it really depends on the scale of your scene.

Best regards,
Vlado

[re:FORM]

Ok, I'll keep my eye out for them. The scene is pretty small, just 50x50m. I'm curious as to why my offset needs to be so large compared to what you suggested. Ach well, if it works then I'm happy.

That'll be yet another beer I owe you

thanks again.
p.