[Radiance-general] Sampling thorugh TRANS material

[Greg Ward]

Hi Germán,

The default -lw setting for rcontrib is actually just 0.002, which means that at most 500 rays will be sampled in the initial scattering, no matter what -ad is set to.  You really need to adjust these settings in tandem when the irradiance cache is turned off, as it must be for rcontrib.

Illuminating the interior via a BSDF material means that you are randomly sampling something that is also randomly sampling, so you have two major contributors to variance.  You can see some of these effects in my 2011 Radiance Workshop talk, "The Bidirectional Scattering Distribution Function as a First-class Citizen in Radiance":

	http://www.radiance-online.org/community/workshops/2011-berkeley-ca/presentations/day2/GW5_BSDFFirstClass.pdf

Best,
-Greg

> From: Germán Molina Larrain <germolinal at gmail.com>
> Subject: Re: [Radiance-general] Sampling thorugh TRANS material
> Date: March 9, 2016 10:32:22 AM PST
> 
> Thanks, Greg... It worked. Is there any special thing about TRANS that requres so many ADs? This is the first time I have required a value that high to converge.
> 
> Best,
> 
> 2016-03-04 22:35 GMT-03:00 Greg Ward <gregoryjward at gmail.com>:
> Hi Germán,
> 
> I'm really only guessing, but I would say you need to increase your -ad value (a lot) to get a good result for your workplane in this situation.  You have the noise of the BSDF on the window to filter, which requires a great many samples.  Setting -ab 12 is probably higher than needed.  What is your -lw setting?  This is important for determining the number of rays actually traced.  Also, -aa gets set forcibly to zero by rcontrib, since it doesn't support the ambient cache.  Try:
> 
> 	-ad 20000 -lw 2e-5 -ab 5
> 
> to see if this produces smoother results.
> 
> Cheers,
> -Greg
> 
>> From: Germán Molina Larrain <germolinal at gmail.com>
>> Subject: [Radiance-general] Sampling thorugh TRANS material
>> Date: March 4, 2016 12:33:12 PM PST
>> 
>> Hello everyone, 
>> 
>> I have a question about trans material... The story is the following:
>> I already read what Axel Jacobs has to say about Trans, read the Radiance Reference Manual and also Googled a bit. 
>> After all that, I found out how to model Fabric materials using Trans (I need some uncolored specular transmission and a bit colored Diffuse transmission). After the research, I noticed that Trans, as I wanted to model Fabrics, would only work for gray textiles... but that is good enough for me.
>> I managed to do all that, and actually run genBSDF to check if the diffuse and direct parts were OK.... they came very close, but not quite perfect, even if my parameters were pretty high. 
>> Assuming it was as good as it would get, I continued with my simulation... I am using rcontrib's 2-phase method, with "-ab 12 -ad 4096 -aa 0.1" parameters... but I get some NOT SMOOTH results. LINK TO PICTURE
>> Any hints on the parameters I have to use in order to model this product well? Is this caused by the fact that there is no direct light in the calculation of the DC matrix? Will my approach work?
>> 
>> Regards, 
>> 
>> Germán
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.radiance-online.org/pipermail/radiance-general/attachments/20160309/bdedfbb9/attachment.html>