[Nova66]

I have a case where the shadow of a pane of glass is too dark when seen through that same pane of glass.

On the left side of this rendering you can see how Luxology correctly renders the shadows. On the right side are the Maxwell renders where I've circled the problem areas in red:



I've added a ZIP file to this post that includes the original MicroStation file and the exported MXS file.


Cheers,
Andrew Novinc.

[polynurb]

there are a few problems with shadows and sunlight right now..
i hope it will be fixed soon.

*edit

but i am wondering if the results in your test a really wrong too.. wouldn't it depend a lot on the reflective properties of the glass material? as the piece of the shadow that is not lying between the walls recieves less indirect illumination bouncing of them...

[Half Life]

SL also plays a part here -- in my experience glass does not reveal its full character in Maxwell until about SL 35 currently.

Best,
Jason.

[RobMitchell]

SL also plays a part here -- in my experience glass does not reveal its full character in Maxwell until about SL 35 currently.

Best,
Jason.

SL 35? That's pretty discouraging to hear... I don't think I've ever had a render reach the default 25 on a typical job.

[Nova66]

Thanks for taking a look at this guys but to me it look like an actual problem with the rendering rather than an issue with SL or Reflection.

In this rendering the Green circle shows that SL or Reflection isn't the problem because the same geometric configuration fails under the Red circles.



The ZIP file includes the MXS file and you can see from all sorts of angles that things aren't quite right. Note that the Fire preview doesn't go to a high enough SL so you'll need to use the actual renderer.

Andrew.

[Half Life]

If you render high enough you will see that change.

I've done it only to prove to myself that it's out there -- but yes, not terribly practical... this is certainly an element (dielectrics) in which there is still much room for improvement/optimization/prioritization in future versions.

As such AGS unfortunately still has much relevance -- I do very much look forward to the day when nobody will use AGS ever again.

Best,
Jason.

[Half Life]

BTW if you try this with SSS planes of similar thickness and low scattering you will see exactly the same effect... I document this in my Sketchup to Maxwell video series.

Best,
Jason.

[polynurb]

Thanks for taking a look at this guys but to me it look like an actual problem with the rendering rather than an issue with SL or Reflection.

yep.. you are right.. i missed the effect in the top row.

i am still using 2.5.. so i can't open your mxs correctly; but if you are using the same "real glass" material on the top row too, there must be something wrong.

"sun shadow through glass".. this reminds me of something..

[Half Life]

It's the effect of the dark portions of the glass shadow only being lit by caustic light -- caustic light is optimized to reflect back through the surface of the glass very late in the render process... typically not making an appearance at all until SL 28 or so, and taking much longer to clear.

It's there -- just let it render longer.

All the non-caustic light is accounted for much earlier which is why things look the way they do... try it and see.

This is one of the dozens of examples I could point to as to why it is important to let your renders in Maxwell cook for as long as you can -- being unbiased means there is almost always something more you can get as far as realism is concerned if you let it render long enough (or use a render farm).

BTW the Modo render "glass" isn't really much better than AGS -- if you let the Maxwell glass render long enough you will see it is orders of magnitudes superior.

Best,
Jason.

[Bubbaloo]

I have done a test with pool caustics before, where it started to look good at S.L. 33. It's the same thing happening with the shadow behind glass. It's considered a limitation of the software, but not a bug. As a side note, caustics will clear up even slower with Fire due to the low S.L. optimizations.

[Nova66]

It's there -- just let it render longer.
All the non-caustic light is accounted for much earlier which is why things look the way they do... try it and see.

Thanks Jason & Brian you were both spot on. I did a postage stamp render (200x150 pixels) and after about 12 hours it got to SL 30:



You can see the shadow starting to clear up but SL 35 is another 58 hours away

We're all in this because Maxwell is such a fine unbiased renderer but it's shocking and a little scary that such a simple model can take so long to clear up at such a low resolution. I know there's a lot of load balancing going on and if it was a simple task it would have been addressed long ago but I'm wondering if this issue is under continuing development or as Brian says "It's considered a limitation of the software".

Thanks for your insight
Andrew.

[Bubbaloo]

It's a known limitation but those guys are always exploring ways to improve.

[Half Life]

I've let some of my tests render for over 300 hours Definitely not practical but certainly something I expect will be improved upon in future releases.

Best,
Jason.

[polynurb]

still i am a bit puzzled .. i think the effect is not as strong when using physical sky + sunlight..

i tend to use real glass as much as possible, due to the better quality of reflection.

this is a crop from a recent work using sun+physical where the above problem seems to be less apparent:

[brodie_geers]

True. I got into this issue back here... http://www.maxwellrender.com/forum/view ... 97&t=36001

-Brodie