[REQ] ray bending medium (ideal lens material)

[CTZn]

Something that would allow one to build lenses out of sphere shapes using precedence... without "caring" about reflected light... so yes, the request being about an ideal object it's a biased one I suppose.

It might be fun with camera or punctual (ceiling) light sources, though I understand that the resulting mediums would not be designed to be seen but their effect.

I'm assuming here, that discarding reflection rays may produce a material more efficient than <specular> for the matter.

[galinette]

It would not be unbiased. Theoritically, it's possible to make the reflection coefficient on the glass-air interface as low as possible with thin film stacks (a succession of many coatings). It's just technically impossible as this is a theoritical limit.

For instance, with a few layers, you can lower the reflection coefficient down to 0.5% in normal incidence. The remaining 0.5% are colored (generally dark blue as on expensive watch glasses) and increase with angle, but it can be improved by complexifying the stack.

There's nothing unphysical I think.

Etienne

[CTZn]

Yes, an ideal lens is beyond the theorical limits therefore it's biased

I think I came with that request after a fresnel-like lens test under sun, with bidir. Performances were so poor that I made the lens thinner and with an higher IOR to compenssate, with little help.

[galinette]

There is a big difference between:
- Beyond the theoritical limit -> biased
- Just at the theoritical limit -> theoritical but unbiased

This request is the second case and does not break any physical or rendering rule.

Etienne

[CTZn]

Ah sorry didn't mean to edit the message above... nvm.

I understand what you say well Etienne ! But are you suggesting that you would rather see the implementation of a limitrophic, physically plausible coated glass material instead of an ideal lens ?

I understand that the theorical model is very close to the ideal (biased) one, but how would performances compare then (say, with bidir) ?

[galinette]

Actually, I'd like to see the possibility to override the fresnel equations of the specular material, to replace them by a custom user-defined R(wavelength, theta) and T(wavelength, theta). I know that Nick is already planning to give it a try.

This would allow making any coated transparent material, as well as fulfilling your request. You would just have to set R=0 and T=1. And this would allow me to make the real-life antireflection case!

Etienne

[CTZn]

You would just have to set R=0 and T=1.

Fair enough