On 9/10/25 20:08, Cousin Ricky wrote:
> I tried samples 10, 10, and that improved the reflection; but now the
> media photons are considerably dimmer than before.

Dimming is a side effect of media sampling - any sort of adaptive 
sampling really - including AA with small details.

Where the media density isn't constant throughout a container, or, here, 
  where the distribution of deposited media photons isn't of constant 
density for all ray paths through the media - how the samples you do 
take, land, matters to the intensity of the overall result.

A simple example. Suppose looking down on the media box we have, 
vertically, a very thin layer of deposited photons. In our set up we 
take 3 primary samples and no adaptive samples. Further, camera ray 
paths traveling down into the media box are such that only the middle 
sample of the 3 lands in the middle of our thin layer of deposited media 
photons. The resultant media intensity is calculated as 1/3 for the 
three samples.

Now we up the samples to 10, but still only one of those 10 sample lands 
in our thin layer of deposited photons. The calculated media intensity 
is now 1/10 for the samples taken.

Our adaptive sampling realities are more much complicated - not the 
least of which is the adaptive bit which tends to pile on samples 
between changing initial samples. This often both dims and brightens - 
depending.

With media photons we have other factors like the photon gathering 
radius and light intensity and density (radial pattern spacing) used 
when depositing photons. Users might use an 'absorption' color to 
control sampling, over brightness/intensity (color channel clipping) 
where ray paths are relatively long.

With that mirror's reflected rays, and our simple example, we could have 
reflected ray paths which run much 'longer' within that thin layer of 
deposited photons than they do more generally in the scene - resulting 
in unrealistically bright reflections.

The rule is more samples / rays for any given render is more accurate. 
Of course, that gets expensive(*).

There may be additional causes the dimming you see. How the media 
adaptive sampling gets done is itself complicated (**) and there are 
code comments for suspected shortcomings and open questions as to 
particular bits of the code.

(**) - The intensity / grayscale values used while controlling the 
adaptive process for different colors, changes with color, for example.

Bill P.

---

(*) - In yuqk, I co-opted a number of existing features under one 
keyword called 'amplify'. I've also added the 'amplify', rgb, color 
vector multiplier feature to additional {} blocks too (not yet media{}). 
The thought is to let users adjust the color contribution intensities 
from various features in some standardized, SDL defined, way.

If media gets a little dim to tastes, use perhaps

media { Media00 ... amplify <1.3,1.2,1.5> }

If after, the reflection from that mirror looks too bright,

reflection { ... amplify <1/1.3,1/1.2,1/1.5> }

Post a reply to this message