PDA

View Full Version : Self shadow clearly not working



Dan Ritchie
04-17-2015, 02:54 PM
Self shadow clearly not working with translucency.

127892

Can anyone explain why self shadow is so wonky in general?

jboudreau
04-17-2015, 04:47 PM
Self shadow clearly not working with translucency.

127892

Can anyone explain why self shadow is so wonky in general?

Hi

Can you provide your scene so we can help you better

Thanks,
Jason

Nicolas Jordan
04-17-2015, 04:57 PM
I did some quick tests in LW 2015 SP2 and I can confirm that "self shadow" and "receive shadow" do not seem to work as expected in my tests.

bobakabob
04-17-2015, 05:28 PM
Phew, that looks like a shocker... Have you reported it? Would be helpful to see the scene file.

Sensei
04-17-2015, 05:30 PM
Self shadow clearly not working with translucency.

Dan, unfortunately I have to say, you don't understand how renderer is integrating data from various sources/channels.

Translucency has nothing to do with self shadow option.

Take a look at these screen-shots:

127894

127893

127896

127895

When we have 2 balls in single layer (single Layout object), when renderer is checking whether spot is covered by anything it is sending ray to every light source.
It can ignore currently processed object,
or
not ignore currently processed object (to which world spot belongs).
That's what "Self Shadow" trigger option does.

If I would separate these two balls to two layers, and make from them two layers, self shadow would do nothing..

Diffuse = dot product between surface smoothed normal vector (and modified by node editors and bump maps) and each light sample direction vector.
If it's negative it's normal diffuse effect
diffuse = -dot( normal, light_sample_direction)
if it's positive it's normally ignored (vectors quasi-parallel).
But when translucency is > 0 then positive part is used and multiplied:
translucency = dot( normal, light_sample_direction ) * translucency;

ps. Eventual issue in 2015.2 is another story.
In 2015.1 my example is working the same as in 11.6.x

ps2. object = single layer = single Layout item.

jboudreau
04-17-2015, 05:51 PM
I did some quick tests in LW 2015 SP2 and I can confirm that "self shadow" and "receive shadow" do not seem to work as expected in my tests.

Hi Guys

It doesn't work here in 2015.2 either unless you do a F9 render which does work and then all of a sudden the VPR works too. So it looks like a refresh problem

So to fix it for now just do a F9 Render and then it works

Another way of getting it to work in 2015.2 is to turn vpr off and back on again.

Thanks,
Jason

squarewulf
04-17-2015, 06:18 PM
Hi Guys

It doesn't work here in 2015.2 either unless you do a F9 render which does work and then all of a sudden the VPR works too. So it looks like a refresh problem

So to fix it for now just do a F9 Render and then it works

Thanks,
Jason

Have you tried switching VPR out of draft mode? Sometimes that fixes stuff. Just a thought.

jboudreau
04-17-2015, 06:22 PM
Have you tried switching VPR out of draft mode? Sometimes that fixes stuff. Just a thought.

Just tried switching to draft mode and it doesn't work but changing from vpr to another shaded mode then back to VRP works

Thanks
Jason

Nicolas Jordan
04-17-2015, 07:07 PM
Hi Guys

It doesn't work here in 2015.2 either unless you do a F9 render which does work and then all of a sudden the VPR works too. So it looks like a refresh problem

So to fix it for now just do a F9 Render and then it works

Another way of getting it to work in 2015.2 is to turn vpr off and back on again.

Thanks,
Jason

That makes sense. I was using VPR also and forgot to mention that. I didn't actually try an F9.

Dan Ritchie
04-18-2015, 02:48 PM
You know, I've written a few raytracers, so I know all that.

What I am trying to understand, is why the shadow looks the way it does? I was under the impression that shadows were cast by a surface even if it wasn't double sided, but I must be wrong.

Dan Ritchie
04-20-2015, 11:12 AM
Ok, on closer inspection, I appear to be wrong. This is the intended behavior. When I stack two single sided planes on top of each other, the first one does not cast a shadow. Therefore, the backside of the glass is only receiving shadows from the faces that are facing toward it. This is unfortunate, but the intended behavior. Seems like that behavior has changed some time over the last 20 years, but it makes sense, because translucency wouldn't work otherwise.

Sorry for crying wolf. Bare with me, though, because I got a little more self shadow crying to do yet.

spherical
04-20-2015, 02:39 PM
I tried by reproducing your scene from scratch. Got forked off on other tasks for the moment but did come to some conclusions.


That which you found, above, is correct. The situation created by the bowl shape is one that Translucent doesn't cover. Translucent is essentially a cheat in the first place, so cannot be expected to work in all situations. Only volumes that have no interfering inner surfaces that fold over upon themselves would look correct.
There might be a way to fake it by selecting the bowl inner surface and making only those polys double-sided, so they would cast a shadow, but I didn't have time to try that.
For that object, I'd use an SSS material anyway. Far more realism, as expected, with way more control.