Question : light material sample distribution

[v-cube]

Hello Paul,

I am working on a fairly complex interior scene with lots of emissive light materials.
The biggest challenge is actually to balance the different light groups against each other to achieve a nice overall appearance.
I understand that the "surface brightness" node ,if enabled, handles the light power relative to the surface area covered by the material ( increasing the surfaces area would result in more distributed light in the scene...)

Since I get a lot of emissive material related noise, I try to identify the noise sources (certain materials... light ID is a really great help here!) and try to adjust the samples of the materials accordingly.

Here is my first question : is the value for the material sample absolute or relative (would increasing the surface area automatically result in 1.more samples in the scene or 2.would the sample amount be thinner distributed onto the surface leaving the total amount of samples in the scene the same?)

my second question: is the amount of samples being emitted dependant from or connected to the amount of light a material is distributing to the scene? I mean, do brighter lights , which have a higher optical impact on the scene, automatically get more samples may be due to some sort of global noise threshold? If that is not the case, would this make sense?

what would be the best strategy to have a "even" noise distribution in the scene with several lights in general?

thanks for any feedback

Andreas

[formatio]

Hi

There is a setting for emmisive node called sample rate:

From my experience what it does is allocate more samples to that particular emmisive (light source). It doesn't multiply samples, it simply make chosen emmisive to be more important in calculation.

This should help evening up the noise. Try fiddling with it and you will instantly get the idea.

Regards,
Mac

[v-cube]

Hello Mac,

thank you for your response

There is a setting for emmisive node called sample rate:

yes I am very well aware that the light samples can be modified to increase the "importance/ impact" of the light material during the rendering phase, I know that you can try to "balance" the sample amounts from different light materials against each other (actually I tried to describe that in my initial post...)

The questions remain:

1. is the value for the material sample absolute or relative (would increasing the surface area automatically result in 1.more samples in the scene or 2.would the sample amount be thinner distributed onto the surface leaving the total amount of samples in the scene the same?)
2. is the amount of samples being emitted dependant from or connected to the amount of light a material is distributing to the scene? I mean, do brighter lights , which have a higher optical impact on the scene, automatically get more samples may be due to some sort of global noise threshold? If that is not the case, would this make sense?
3.what would be the best strategy to have a "even" noise distribution in the scene with several lights in general?

best

Andreas

[formatio]

I think its option 1 - and it is relative - sampling rate works like importance in my opinion.

I usually tweak this setting until I get nice results - there is no perfect formula i think

[face_off]

Since I get a lot of emissive material related noise

Is it bright noise, or dark noise? If dark, reduce the kernel->path termination power (I've been using 0.1). For bright noise, you can cheat and increase Caustic Blur (however that may impact the end result quality beyond what you are prepared to accept).

Paul