Caustics and fog dont render on reflection and refraction.

[ior]

Greetings to all! Especially to the developers of indigo that are doing an amazing job.


This is strange: Caustics and fog (volumetric ss) don´t appear on reflection and refraction.

Iv found a picture from IanT that does exactly what indigo is not doing for me:

http://www.indigorenderer.com/forum/vie ... &start=116

i tried every render configuration possible and material type and always the same. It seems to me that indigo should do that with no much more effort. what am i doing wrong?
Already found the ideal parameters for the perfect render but, no caustics and fog on mirrors or trough glass:

Code: Select all

   <super_sample_factor>5</super_sample_factor>
   <metropolis>true</metropolis>
	<large_mutation_prob>0.400000006</large_mutation_prob>
	<max_change>0.3000000119</max_change>
	<max_depth>32000</max_depth>
	<max_num_consec_rejections>32000</max_num_consec_rejections>
	<bidirectional>true</bidirectional>
	<hybrid>true</hybrid>
	<bih_tri_threshold>512000</bih_tri_threshold>
	<ray_origin_nudge_distance>0.0001</ray_origin_nudge_distance>

I tested a scene with one 100% mirror on the right and one glass sphere on the left were some caustics appear and also volume fog from one light source in the middle, there are some caustics as well on glass teapot on the floor:

The mirror is darker because there is no light from caustics of it in the image of the mirror, as you can see on the ceiling near the end of the mirror.

Don't understand why the light glare on the mirror is bigger in the mirror reflection (using specular with high ior).

Than I have put a sheet of glass in front of the left sphere caustics and fog. They disappeared:

why the sheet of glass turns that sphere shadow so dark? Not realistic (the glass has density 0).


well, for now I would like to know how to have a mirror behaving like a mirror, and a transparent glass that does not hide some effects that I enjoy.

Help appreciated.

[OnoSendai]

Hi ior,
First of all, nice and brutal test scene!
Can you please email it to me (nick at indigorenderer.com), or host it on dropbox and email me a link?

You are right, caustics should show up in the mirror reflections. I'm guessing your problem is the super sample factor 5. Try turning it down to 1 and it should look better.

For MLT, your max change is also way too large (IMO). You should leave it on the defaults.

Don't use hybrid mode

Also make sure you are using the latest indigo build (2.4.10).

Cheers,
nik

[ior]

Thanks.

I can mail you the test scene in maya 2011 format? ill sent you it tomorrow.

If I lower the supersample, the glass get full of fireflies (dont want to change this lower than 3), the caustic and fog result is the same.

according to this post i don´t need MLT so i disabled it:
http://www.indigorenderer.com/forum/vie ... &start=122
Only Bidir and still no caustics in mirror.


No more for today. tomorrow maybe if i have time.

Comments welcome.

[OnoSendai]

Thanks.

I can mail you the test scene in maya 2011 format? ill sent you it tomorrow.

If I lower the supersample, the glass get full of fireflies (dont want to change this lower than 3), the caustic and fog result is the same.

according to this post i don´t need MLT so i disabled it:
http://www.indigorenderer.com/forum/vie ... &start=122
Only Bidir and still no caustics in mirror.


No more for today. tomorrow maybe if i have time.

Comments welcome.

Please send the file in IGS or PIGS format, thanks.

The fireflies are just caustics that haven't converged yet

[ior]

Here is the file in igs format with obj and mtl:

untiled.zip

(281.13 KiB) Downloaded 195 times

Please test it to see if anyone can get caustics and fog on reflection of the mirror. Post the result if yes.

Cheers.

Ps: the file was rendered in a core i7 920 @ 3700 Mh, 12 gig ram, 1 hour.

[pixie]

Here is the file in igs format with obj and mtl:

The attachment untiled.zip is no longer available

Please test it to see if anyone can get caustics and fog on reflection of the mirror. Post the result if yes.

Cheers.

Ps: the file was rendered in a core i7 920 @ 3700 Mh, 12 gig ram, 1 hour.

I render the scene given, tweaked it so that it had more light power so that I could use camera and this was my result. I wonder if someone can spot the oddity.

[ior]

what is wrong with this picture?:

Just to demonstrate the point ... 30 seconds on a laptop (no MLT... all important caustics already visible):

this is from a post from 2007:
http://www.indigorenderer.com/forum/vie ... &start=122

Remember! this was done in 2007 with first versions of indigo, that allowed caustics in reflection only with bidir path tracing, no MLT. Now it is impossible with indigo render!
With the latest versions of indigo only with MLT you can get caustics on mirrors and it takes too long to render decent results. Even longer with only bidir (till fireflies converge ).

I render the scene given, tweaked it so that it had more light power so that I could use camera and this was my result. I wonder if someone can spot the oddity.

I spot the oddity! you used MLT and rendered for long time, till most of the sparkles disappear (you can still see some imperfections on the fog reflection from the sphere and still no caustics on the reflection caustics of the pot!).


So what i think is: or indigo developers removed reflection caustics (etc.) to get some improvement in general rendering (do not understand that if you can have an option to override that) or something in this post is not right:
http://www.indigorenderer.com/forum/vie ... &start=122

Fryrender can do that. indigo does not now, why?:
http://www.indigorenderer.com/forum/vie ... &start=140
http://www.indigorenderer.com/forum/vie ... &start=122

Wait to see some response from developers.

[pixie]

I render the scene given, tweaked it so that it had more light power so that I could use camera and this was my result. I wonder if someone can spot the oddity.

I spot the oddity! you used MLT and rendered for long time, till most of the sparkles disappear (you can still see some imperfections on the fog reflection from the sphere and still no caustics on the reflection caustics of the pot!).

I try to use the right tool for the job, and for caustics interaction with the medium the best tool is bidir MLT. Still the oddity I was talking about was that, despite having my indigo registered (as you can see from my badge) having the image watermarked

[CTZn]

Hi ior,

Despite the fact that you are using an uniform scattering, the scene is still offering too complex light paths to afford discarding MLT, as pixie said.

The mirror's medium is using absorbtion layer, though as it is not a transparent material this may have no impact at all.

About emission power, I suggest that you use the spectrumNode wich you will find under the HyperShade color utilities. Plug it into the base emission, and experiment with the blackbodies gain and camera/linear exposure. Edit: Else what, do not hesitate to use super colors (>1) as base emission.

A revised version of mti will be out the next weeks.

[pixie]

A revised version of mti will be out the next weeks.

Yet another?

[CTZn]

ah come on, don't push me this often pixie or else !

;P

[ior]

Hi.

Despite the fact that you are using an uniform scattering, the scene is still offering too complex light paths to afford discarding MLT, as pixie said.

Even without ss fog, the results are the same.
MLT take too long to have decent mirror caustics on reflection and does not do the job properly (as the caustics of the pot do not appear on the mirror reflection), don´t think this has anything to do with light power.


What about the image from lwan done with simple bidir in 2007 on a laptop in 30 seconds?:
http://www.indigorenderer.com/forum/vie ... &start=122

Now with latest versions of indigo it is impossible to do that, as I said before. Please comment about Iwan image.

[pixie]

Remember! this was done in 2007 with first versions of indigo, that allowed caustics in reflection only with bidir path tracing, no MLT. Now it is impossible with indigo render!

From what I could gather from the thread this wasn't rendered with Indigo, rather with fryrender v1.8.

BiDirPath

2.4.10

I see Caustics, but I also see a dark glass, Whereas on single Path MLT It's not as dark.

PT+MLT

I did some more tests, and although I don't have that particular build (one can't seem to download past 1.1.18) nor that particular laptop, I've downloaded the 1.1.18 build and did some comparisons against the current build. 1Minute each, SS set to 3 just for the kick.

[pixie]

I did some more tests and I fin this odd behaviour:

The very same glass ball from this scene:

PT+MLT

The time went further on and made some improvements:

MLT Bidir

Wen't on dreaming the impossible:

And fixing internal Indigo bugs all by myself:

Hey Ono, I want my paycheck!

[ior]

Hei comequi é? With no license the .pigs file is useless.

Lots of processing power plus Mlt + BiDir is the key?

Try to render my orignal scene and have some caustics on the reflection of the pot (no way!).


Tomorow ill be here (maybe), for now using a portable internet cause my isp is down for a while.