Would non-linear absorption make sense?

[danw]

I'm trying to eyeball a decent red-wine look using mainly refraction attenuation.

 

I'm finding no matter what I use, I either get a decent fade-off from strong red to black, but the thinnest parts still look far too transparent... or if I crank up the density, I get a decent red tint in the thin areas, but it feels as though the fluid fades off to black far too rapidly, leaving very little redness anywhere except the very thinnest areas.

 

I'm not really much for decoding scientific papers, but I'm wondering, in a general sense, whether the notion of non-linear light absorbtion makes any scientific sense - ie, wire a "power" node into the "I" value being piped into the attenuation calculation inside the Surface Model shader.  I've tried it, and with a slightly-below-1.0 exponent, it does succeed in reddening the thin areas while slowing the fade to black... but it niggles me that I might just be creating imaginary light-physics.  Does anyone know if there's any actual scientific precident for doing this?

[Atom]

I found this article interesting. More about fluid in a glass, not necessarily wine related. But part of the look of wine is the glass it is in.

 

http://adaptivesamples.com/2013/10/19/fluid-in-a-glass/

[danw]

Ah yes, I'd given the idea of modelling fluids and glass as discrete medium-to-medium interfaces rather than enclosed objects quite a bit of thought over the years, and set up something very much like that a few months back... it certainly does make for a far closer fit to reality when rendering.

 

Actually, I think I may have been overthinking the red wine look... after putting together some accurately controlled reference video while drinking at a restaurant the other day, I realise red wine really does absorb light like crazy - I think I just needed to embrace incredibly high attenuation density and trust the renderer to know what it's doing :-)

 

It'd still be interesting to know if there are any digestible reference on the topic of absorption non-linearity though.

 

I'd also love to know if anyone knows of an artist-friendly resource that suggests whether and to what extent various materials and fluids are back- or forward-scattering.  All I ever turn up on that topic is something to the effect of "it depends on the average size of the particles relative to the wavelength of light", and *never* any indication of what materials actually contain what sized particles.

Edited March 19, 2016 by danw

[fathom]

not sure you can really limit the effect to simple absorption.  i would think there's gotta be some scattering going on as well.  maybe not a ton, but maybe enough to kick things up from black.

[danw]

Yeah, I actually set that up too using a subtle uniform volume with multiple pbr scatter bounces... although red wine is an odd one - it's so densely absorbing that you either get the barest hint of scattered light out of the volume, or it makes it far too cloudy to be red wine.

 

That's another thing it'd be interesting to get a pseudo-scientific idea of - "turbidity" appears to be a quantifiable measurement - it'd be cool to know how "Turbidity Units" relate to uniform volume density values inside of Houdini.

[fathom]

http://www.synergytaste.com/sites/synergytaste.com/files/SEN-TN-0010-Estimating-Turbidity-NTU-from-Absorption-Data.pdf

 

that's a paper that describes estimating turbidity from absorption at a specific wavelength.  it implies a linear relationship, but the question is then one of units and what the paper's absorption values measure and how to convert that to houdini's attenuation values.

[danw]

Interesting read, thanks.  Seems odd that they're suggesting a direct linear correlation between turbidity and absorption... my instinct tells me that's a relationship that only works one way - I suppose increased cloudiness would invariably increase absorption, but I can't see that increased absorption necessarily implies cloudiness.  Fairly sure red wine would be the fluid to disprove that if anyone asserted it - I came across a resource somewhere (which annoyingly I can't re-find) that suggested the best quality red wines have as low as 4-5 NTU turbidity - which seems to be not far from crystal-clear.

[fathom]

thinking about it more, turbidity is really a measure of scattering.  you measure turbidity with a light 90 degrees to the sensor.  absorption is similar, but different.  they're both forms of extinction, but they're two separate parts of it.  that said, in clear water, there's maybe not enough absorption (particularly at the wavelength they selected) to matter much, so you're really measuring overall extinction which is mostly scatter/turbidity.

 

or something.

 

edit: they state the wavelength they selected is mostly affected by particulates and not he tea itself, which implies turbidity is measuring stuff on a different scale that whatever is pigmenting the liquid, be it wine or tea...

Edited March 24, 2016 by fathom