PDA

View Full Version : Dielectric Black Areas



Quantenschaum
09-25-2007, 05:32 PM
hi guys.
this is actually my 1st post here although i keep wavin' since a looong time.

since i'm going to do a job soon involving gold jewellery and different gemstones, it was only recently that i really fiddled
around with the dielectric node. the forum here seems full of questions about the proper settings and believe me, i studied them a lot.

so here's my suspicion:
1: it's definetly BUGGY or missing something like higher Ray Recursion Limit.
2: i'm totally missing something and should wear a clown costume for life.

here's my setup:

Render Globals: Ray Recursion Limit = 24 (it's the max!), Ray Trace >Shad >Trans >Reflect >Refract = ON
Object: SubD with 1 surface. Surface: Double Sided with classic "Switch setup" consisting of the Dielectric (glass),
Standard (air), Spot Info (poly side) and Switch node. the glass material uses IOR = 1.5, Absorb = 0, Rough = 0
for faster rendering i did not use Reflection/Refraction Blurring or Dispersion.

http://www.sheep.ch/dump/FistGrey.jpg http://www.sheep.ch/dump/FistEnv.jpg

for the 1st image i used a plain 128 grey BG color, so that we won't see any distracting reflection or refraction patterns.
sounds weird but what i really want to show are those damn black areas! they just make no sense! it doesn't matter
what background , what IOR i used... they're here to stay! :compbeati

so... anybody with nothing better to do, download my little setup and try for yourself. 50743
any help welcome but please read my settings carefully, so you can avoid the obvious basic questions... :tongue:

jameswillmott
09-25-2007, 06:04 PM
Are you using 9.3?

Turn double sided off and just plug the Dielectric into the Material input.

9.2?

You'll have to set up air polygons or an air surface switch arrangement in nodal to switch the backsides of your polygons to an Air material. This arrangement probably won't work correctly with 9.3 though.

Quantenschaum
09-25-2007, 06:15 PM
jameswillmott,

i'm using 9.2

as i described in my post, i totally did the "switch-node-thing" for air polys.
believe me, i checked and checked and checked already the basics.

in my desperation, i repeatedly tried to just connect it directly to material and toggling double sided on and off. but nooooooo...

fact: those black areas only change shape when modifying the IOR (naturally) but they stay black, no matter what goes on around the object.

just try out my scene and see if you get anywhere

Quantenschaum
09-25-2007, 06:23 PM
oh... and i even tried the olden ways of actually doing Air Polys in Modeler...
anyways... air polys eitherways are necessary.

would it work in 9.3 ???

SHOW MEEEEE!!!!

jameswillmott
09-25-2007, 07:41 PM
I tried your scene in 9.3, after rearranging your node graph and turning off Double Sided, it seemed to work fine.

Quantenschaum
09-25-2007, 07:52 PM
woah! what did you exactly rearrange? Dielectric Node only plugged directly into Material? if so, it seems to be a 9.2 buggy thingy...

jameswillmott
09-25-2007, 08:08 PM
woah! what did you exactly rearrange? Dielectric Node only plugged directly into Material? if so, it seems to be a 9.2 buggy thingy...

Yes, Dielectric directly into Material.

Quantenschaum
09-26-2007, 04:36 AM
ok james...
i updated to 9.3 and it seems to work correctly now!

i also tried to plug the Dielectric node directly into Material and uncheck Double Sided, but it still seems that the classic "Switch" setup produces a slightly different result and i'm having a hunch that this is still the correct way.
the differences are VERY subtle, but you can spot them just around the thumbnail of the fist object.

still: thx james for the tip and for kicking me to the 9.3 update! :thumbsup:

direct to material: 50754 switch node: 50755

jameswillmott
09-26-2007, 06:20 AM
ok james...
i updated to 9.3 and it seems to work correctly now!


Good decision! Glass effects are so much easier in 9.3...




i also tried to plug the Dielectric node directly into Material and uncheck Double Sided, but it still seems that the classic "Switch" setup produces a slightly different result and i'm having a hunch that this is still the correct way.
the differences are VERY subtle, but you can spot them just around the thumbnail of the fist object.

still: thx james for the tip and for kicking me to the 9.3 update! :thumbsup:

direct to material: 50754 switch node: 50755

You have self intersecting geometry around that area, which is confusing the Switch setup and that is why it produces a different result to the volume stack. But I believe the volume stack refractions are the correct ones.


What is happening with 9.3 is this.

1) The camera ray hits the outer glass face of the thumb.
2) It gets refracted and continues.
3) It hits another front face of glass, it is not bent by refraction and thus continues on it's original path.
4) It hits a backface of glass, and again is not refracted and continues on the same path.
5) It hits a backface to air, and is refracted correctly.

Now, with the Switch setup, it does this.
1) The camera ray hits the outer front glass face of the thumb.
2) It gets refracted and continues.
3) It hits another front face of glass, it is not bent by refraction and thus continues on it's original path.
4) It hits a backface, but here your switch setup tells the renderer that all backfaces are to air, and so it is refracted. This glass->air refraction happens inside the model, and is thus incorrect.
5) It hits a backface to air, and is refracted correctly.

Quantenschaum
09-26-2007, 08:33 AM
hmmmm...
i think your description of the SWITCH setup is not correct. actually it describes the effects of NOT using it... hehe...

1. the ray hits the Dielectric surface and gets refracted
2. before the ray can hit another Dielectric surface, it HAS to and WILL hit the Air surface that the SWITCH setup provides!

i'm already into rendering diamonds and i will post an example soon...

thx for your Brain-Time anyways!

Quantenschaum
09-26-2007, 08:58 AM
so... here it is. all settings are as always (in this thread)
Direct Setup has Double Sided Unchecked.
which ring would you buy? :hey:

believe me, the right one looks like the real thing,
which i held in my hands (for a short while...)

50760

jameswillmott
09-26-2007, 09:03 AM
hmmmm...
i think your description of the SWITCH setup is not correct. actually it describes the effects of NOT using it... hehe...


Wrong? I don't think so... :)

Not using a Switch setup for this model makes the ray refract only once, which is when it first enters the glass. Every other time it hits a surface it will have the same IOR as it's current medium and no refraction will occur.



1. the ray hits the Dielectric surface and gets refracted
2. before the ray can hit another Dielectric surface, it HAS to and WILL hit the Air surface that the SWITCH setup provides!


:)

I don't understand where you get point #2 from, a ray can only hit an Air surface if it hits the back side of a polygon using a Switch setup. The Switch setup will not create new geometry for the ray to hit...

The way the fist model is built, the camera ray hits two front sides in a row through the thumb area, followed by two back sides as it passes through the fist and exits the model.

Using the Switch setup, the ray is told it enters Glass twice and then exits to Air twice, refracting the first time it enters Glass, and the first time it exits to Air.

Using the volume stack, the ray is told it enters Glass twice, exits to Glass once, and then exits to Air once, refracting the first time it enters Glass, and refracting the only time it exits to Air.

Because the model is built wrong, the Switch setup works incorrectly.

jameswillmott
09-26-2007, 09:05 AM
so... here it is. all settings are as always (in this thread)
Direct Setup has Double Sided Unchecked.
which ring would you buy? :hey:

believe me, the right one looks like the real thing,
which i held in my hands (for a short while...)

50760

Did you turn on Reflections?

Quantenschaum
09-26-2007, 09:43 AM
that is what the SWITCH node is for!
you feed it with the Spot Info node's POLYGON SIDE parameter.
then you plug the Dielectric Node (Glass) into the TRUE input and
the Standard Material Node (Air) into the FALSE input. so now the backside of the Glass polygons become Air polygons when hit from the backside.

maybe the "Fist" object wasn't a good choice to demonstrate, cuz it's basically a ball. diamonds on the other hand have very strong internal refraction/reflection effects due to their strict geometry.

the "what's new in 9.3" manual states that you won't need air-polygons anymore. but what they mean is: you dont have to model them anymore.

Quantenschaum
09-26-2007, 09:52 AM
oh... of course i meant the LEFT ring to be the real thing! my bad, sorry.

jameswillmott
09-26-2007, 05:13 PM
Are you able to post the model for dissection? :)

Quantenschaum
09-30-2007, 08:15 PM
i did some further dielectric-node testing and definetly came to a conclusion.
check out my tests and see for yourself... images and all:

http://www.sheep.ch/gmd/dielectric.html

Quantenschaum
09-30-2007, 08:31 PM
ooops... forgot...
here's the Lens Scene to fiddle a little...

http://www.sheep.ch/gmd/Lens_Test1.zip