Page 1 of 1

Glass Shadows

Posted: Wed Jan 06, 2016 2:25 am
by gianca
I've been working on a scene that has lots of glass, and I noticed something odd: the shadows of glass objects looks wrong to me when it's seen trough another glass.
With other renderers I solve this by increasing the amount of max ray bounces, but I wasn't expecting this in Maxwell and I would not even know where to look to fix this...
...unless I'm doing something really wrong, which is what I hope and why I'm posting this here.

The scene is a really simple scene I built to test this issue. Standard simple material have been used here, and the glass material is a standard one made using the wizard (not the assistant).
As you can see from the image the shadows behind the front glass pane look wrong, like light is not passing trough at all.
I'm using 3.2.1.0.

Any advice is welcome, even if just to say "yep, that's how it is".

Best,
Gianca

Re: Glass Shadows

Posted: Wed Jan 06, 2016 7:18 am
by seghier
you are right ; i hope they will fix that in future

Re: Glass Shadows

Posted: Wed Jan 06, 2016 12:10 pm
by Half Life
The caustics there will be "late converging" -- somewhere around SL 24-35... though they eventually will appear. This has always been the case with Maxwell AFAIK.

The more diffuse/scattered the scene lighting the less this will be an issue(although this also means the caustics will be less defined)... so the solution is to not use the Maxwell Physical Sun/Sky for lighting these types of tests (basically) and/or let it render longer.

Re: Glass Shadows

Posted: Wed Jan 06, 2016 2:03 pm
by lebulb
A test done a few years ago on a tiny definition :

Image



physical sky (SL16)
Image

default (no other light source) (SL16)
Image

without bulb (SL16)
Image

bulb Hidden from Global Illumination (SL16)
Image

seghier wrote:you are right ; i hope they will fix that in future
No need to fix,
just need more core or super extra algorythm...

Re: Glass Shadows

Posted: Wed Jan 06, 2016 9:34 pm
by gianca
Hey Lebulb,
I'd love to see our tests but they did not come trough.

Thanks for all the responses: I resorted to use an AGS glass.
Bummer as the "real" glass looks way better, but I got tons of glass panes overlapping so that's not really a "real" option.

G
lebulb wrote:A test done a few years ago on a tiny definition :

Image



physical sky (SL16)
Image

default (no other light source) (SL16)
Image

without bulb (SL16)
Image

bulb Hidden from Global Illumination (SL16)
Image

seghier wrote:you are right ; i hope they will fix that in future
No need to fix,
just need more core or super extra algorythm...

Re: Glass Shadows

Posted: Mon Jan 11, 2016 5:29 pm
by F. Tella
Hi Gianca,

As Lebulb is showing in the last image you can hide the glass to GI and it will have the same look as normal glass but won't cast shadows. Then you can add a second geometry hidden to camera and refl/refractions with AGS material to simulate the shadows. Make sure this second geometry is not exactly coplanar with the first one as it will generate some artifacts. This workaround is a bit laborious but renders quite fast and the look is much better than AGS.

Best,
Fernando

Re: Glass Shadows

Posted: Tue Jan 12, 2016 8:55 am
by feynman
I would say Fernando's trick from a few years ago works well also on close-ups; we're using it a lot for rendering glass tabletops.

Re: Glass Shadows

Posted: Tue Jan 12, 2016 2:58 pm
by F. Tella
Another tip on that topic: the bigger the emitter, the easier for caustics to show through the dielectric. Basically it's easier for the rays to reach a big emitter than a tiny one very far away in the sky (the sun). Some time ago I did some experiments changing sun radius; they may come in handy to understand how this works: http://www.maxwellrender.com/forum/view ... sun+radius

Cheers!