How well can thick glass with six layers and a curve be simulated?

[bildform]

I have a client who would like too see the effect on the view his glass window has.

How physically correct can v-ray simulate this? What would be a good method to approach this?

Thank you for your help?
Jan

[^Lele^]

I'd assume quite accurately, if you input all the relevant physical values (IoR, absorption, and so on) right into your shader.
Did you have any specific issue?

[fixeighted]

I was going to ask; when trying for 'perfect' accuracy in this case, would a nurbs curve based mesh give a more accurate result than a smoothed poly version?

Other than that, one point I would make is be aware of using microfacet/phong brdf with glass, as it is buggy when using 'reflect on back side'. Use blinn.

[^Lele^]

I was going to ask; when trying for 'perfect' accuracy in this case, would a nurbs curve based mesh give a more accurate result than a smoothed poly version?

You're much better off with a subdivided mesh.
The trick is to use the VrayDisplacementMod to do so, and set the edge length to a pixel or less (ofc, be reasonable. ).
Then you'd have (sub-)pixel accuracy as if v-ray rendered nurbs directly.

[fixeighted]

Interesting trick - though I'm unsure how I'd use that displace the curve. How do you mean exactly? I was imagining a window that is curved at the corner.
But without using that trick it doesn't make any difference to the way the rays cast through the glass? Or is it that any difference is imperceptible?

[^Lele^]

The accuracy i speak of is inherent in the method.
Which would be identical to poly modelling plus subdivision, just you can use the vrayDisplacementMod as subdivider, either as dynamic Geo (created at rendertime as needed, discarded from ram if it fills up.), or as static, so prepared before rendering and all loaded at once in ram (quicker, but could be more expensive by virtue of the full preloading, and lack of ability to be unloaded.).

You'd build a window with the interface between glass and frame ever so slightly air-gapped (i.e. slightly removed from being coincidental.), for correct results.

[fixeighted]

LOL excellent....I honestly have never thought to use it that way.
Just found this post from way back in 2011 https://forums.chaosgroup.com/forum/...ay-turbosmooth

I am genuinely surprised this isn't advertised more as a technique, especially noting that Vlado points out it's better than turbosmooth as it's view dependent.
Best thing I've accidentally learned in a long time. Thanks Lele

Another reason why this forum is a bit special

[^Lele^]

You're very welcome!

[bildform]

The vraydisplacement trick is really nice. I did not know this.
My client is using a laser and a real curved glass to find out how much distortion is going on.
I am trying to get the information about the glass and try to simulate it with V-ray. Should I get the data I will post my results.