advanced rendering for AutoCAD
Any idea why the frosted glass panels running parallel to the windows are so flat and white, whereas those running perpendicular to the windows look correct?
The panels are drawn as a solid block 6mm thick and I've attached the material settings.
I used the Path Tracer engine for this and am currently running a Hybrid render to see if it makes any difference.
Send the model by email when you get a chance-- if nothing else I can at least check for a problem with the model and make sure the new stuff doesn't have the issues. If it's a bug in the current stuff I'm not sure there's a whole lot I can do.
I've sent you the model by WeTransfer.
Haven't had much chance to look at this during the weekend-- however-- I did run the simulation with the new stuff. This is what I got (with and without the ceiling lights):
1. I'm guessing a lot of the problem comes from the "Thin" switch and the IOR. I did use an IOR of 1.5 and got rid of the Thin switch for the acrylic material. (In the new product the Thin switch is on the material.) The use of these settings is likely causing weird refraction-- the brighter panels-- the ones which refract the daylight-- should be the ones closer to the window. The Path Tracer has a second problem, the use of automatic daylight sources, which is very likely causing some inaccuracies. The Hybrid engine looks like it's doing a more credible job-- but the refraction error is causing the back panels to strike the windows rather than the front ones.
2. This simulation ran forever, but notice how noisy the daylight image still remains. This is due to the large size of the space (ceiling height too) and the fact that it's lit from only one side. The reflected light in the foreground is taking a long time to resolve. Not something you necessarily need to worry about-- this is more for me-- but very interesting.
3. The reflected caustics you see on the back wall using the new product are real-- sunlight off of the chair legs, I believe-- but may be a nuisance and undesirable. Not sure what to do about this.
I'll look into this a little more using the old product-- see if I can learn anything else.
Thanks for looking into this Roy,
I'll do a few tests of my own.
Nice! Many path renderers are introducing denoising features with good results.
I have a question on those images Roy: the bright elliptical spots on the wall between the two windows are reflection caustics from highly reflective surface on the chairs? Looking really bright, but I kinda like it. Those spots are not on Peter's images though.
Most are using nVidia's stuff for denoising-- it's pretty impressive.
They are reflective caustics which you get automatically in the new engine-- yes, I think they come from sunlight striking the chair bases. They should be pretty accurate but some might consider them a nuisance. Not sure what to do about this yet.
Thank you Roy. The more physically based effects nXt has, the more realistic pictures it will compute. All of those need a lot of computing power under (or on top of) our desks though. Are you doing any move to engage GPUs on rendering process in nXt? Most of commercial or non-commercial renderers looks like they are moving that direction (Real Time - RT technology, GPU rendering etc.) Rendering times are abyssally shorter on even mid range cards like 1060 or 1070 (for a few hundreds of €) compared to very expensive CPUs with many cores (and costing thousands of € or $).
Not yet. It's not quite like that-- where you can offload certain portions. It's more like an all or nothing proposition.