fully transparent specular material

[zuegs]

Hello indigos...
can someone explain me why following material isn't completely transparent for light? When i have a sunlight or meshlight shineing through this material only a part of the light energie arrives on floor.

I thought, that a ior=1 or ior=1.001 should not reflect light and all light should enter inside the material and as material doesn't absorb it should also leave the material at the otherside without any loss???

Code: Select all

	<medium>
		<name>glass</name>
		<precedence>10</precedence>
		<basic>
			<ior>1</ior>
			<cauchy_b_coeff>0</cauchy_b_coeff>
			<absorption_coefficient_spectrum>
				<uniform><value>0</value></uniform>
			</absorption_coefficient_spectrum>
		</basic>
	</medium>
	<material>
		<name>glass</name>
		<specular>
			<transparent>true</transparent>
			<internal_medium_name>glass</internal_medium_name>
		</specular>
	</material>

[OnoSendai]

Hi Zuegs.
It should be totally transparent.
Try latest version of Indigo (0.9 test 3).
And post some pics as well perhaps

[SmartDen]

Hmm... what is totally transparent object good for?

[OnoSendai]

It's good for testing.. that's about it

[zuegs]

Hi Nick, hi SmartDen,

it's for testing, right and so I tested a bit arround and found that the effect happens by "MLT+BiDir" when "Max_num_consec_rejects" is low.

Funny is that when using "MLT only" the specular material renders fully transparent already with a very low "max_num_consec_rejects" like 5. But "MLT with BiDir" need a minimum "max_num_consec_rejects" of about 50 to get mostly fully transparent.

Why "MLT" with "BiDir" needs such high "max_num_consec_rejects" to cross one simple specular transparent cube

[OnoSendai]

Hi Zeugs.
when you set Max_num_consec_rejects low, the resulting calculations are actually biased. (Technically any finite Max_num_consec_rejects leads to biasing)

Bidir can need Max_num_consec_rejects quite high to render a transparent cube, because some sampling paths are quite difficult to sample (low sampling density).

[zuegs]

Hi Nick,
Thx for explanation, so it is in the nature of bi-dir to need high Max_num_consec_rejects
because i like to set "Max_num_consec_rejects" quite low to get a faster noise-free render

Is it possible to weight less the rejects inside transparent cube mesh? I mean a reject inside a transparent mesh only counts 0.1 instead of 1 and will so have the possibility to reject more until reaching the "Max_num_consec_rejects" limit. So the direct rejects (through air) could held low for faster noise-freeing

[Kram1032]

...and is it possible, to make that value "infinite"?

[CoolColJ]

I just set mine to 3000 and leave it there

The shadow areas get much darker with it set high in some tests I did, looks nicer. Put it at 1 and everything is very washed out looking

I find not to much speed difference at all with it set high, Max bounces has more speed differences. Like 3 is faster than 1000, but I can't see much visual difference bewteen 5 and 1000....

[Kram1032]

I usually use 10000 xD