Can POVRay be used to make nearly-accurate-to-physics models for optics beyond
refraction and reflection... to diffraction, interference, etc.?

If so, does anyone have examples of setting up objects with whatever settings
are As Realistic As Possible, minimizing "cheating", understanding that the
photon model has some limits, accepting that this endeavor will take a Huge
number of compute cycles, and so on.

Ben

Post a reply to this message

Le 18/11/2011 05:58, Ben a écrit :
> Can POVRay be used to make nearly-accurate-to-physics models for optics beyond
> refraction and reflection... to diffraction, interference, etc.?
> 

Diffraction & interference need notions that are out of the actual
model: accurate set of length-waves (including the spectral
distribution) and phase. At least.

Dispatching a set of length-wave instead of a single rgb would be
possible (it's just a bigger data set, and there are only a few special
case which would turn a single ray in many (such as scene with ior).

Taking into account the phase is a totally different story.

what is in your etc ? Birefringence ? Ray absorption ? Cristal with
effect on the length-waves (such as a blue ray exits as red..) ? Using
Black-light (UV) light sources ?

How deep in non linear optic do you want to dive ?

What is a realistic picture of an Ageratum ? the one identical on rgb
display to the mental picture of your humain brain that your human eye
perceived ? or the one identical to the photo taken by a camera ?

(Ageratum has "blue" flower, but they appears "hot pink" on film: this
happens with many flowers and is known as Ageratum effect in the photo
world. it depends on lighting conditions, film or CDD models...)

> If so, does anyone have examples of setting up objects with whatever settings
> are As Realistic As Possible, minimizing "cheating", understanding that the
> photon model has some limits, accepting that this endeavor will take a Huge
> number of compute cycles, and so on.
> 
The main model of povray is not photon model, it's ray model.
In fact, the model is mainly backward ray tracing. only optional photons
follow the forward ray tracing approach.

-- 
A good Manager will take you
through the forest, no mater what.
A Leader will take time to climb on a
Tree and say 'This is the wrong forest'.

Post a reply to this message

Ben <nomail@nomail> wrote:
> Can POVRay be used to make nearly-accurate-to-physics models for optics beyond
> refraction and reflection... to diffraction, interference, etc.?

  Short answer: No.

  The so-called unbiased renderers, which support BRDF definitions, are
designed for this exact purpose. In other words, if you are able to
describe the physical property of an object with a BRDF, the renderer
will then be able to simulate that property with a high degree of accuracy.
(Of course in many cases it will be extremely slow, and getting a good,
non-grainy result can take humongous amounts of time.)

  Unfortunately POV-Ray is not such a renderer. (On the plus side you
get non-grainy results which are often quite good pretty fast in most
cases.)

-- 
                                                          - Warp

Post a reply to this message

Thanks!  Can you recommend any unbiased renderers, then?  My goal is to
eventually produce real optical devices to test the predicted results (series of
optics).


Warp <war### [at] tagpovrayorg> wrote:
> Ben <nomail@nomail> wrote:
> > Can POVRay be used to make nearly-accurate-to-physics models for optics beyond
> > refraction and reflection... to diffraction, interference, etc.?
>
>   Short answer: No.
>
>   The so-called unbiased renderers, which support BRDF definitions, are
> designed for this exact purpose. In other words, if you are able to
> describe the physical property of an object with a BRDF, the renderer
> will then be able to simulate that property with a high degree of accuracy.
> (Of course in many cases it will be extremely slow, and getting a good,
> non-grainy result can take humongous amounts of time.)
>
>   Unfortunately POV-Ray is not such a renderer. (On the plus side you
> get non-grainy results which are often quite good pretty fast in most
> cases.)
>
> --
>                                                           - Warp

Post a reply to this message

"Ben" <nomail@nomail> wrote:
> Thanks!  Can you recommend any unbiased renderers, then?  My goal is to
> eventually produce real optical devices to test the predicted results (series of
> optics).

Check out Luxrender (free and open-source), Octane, Indigo or Maxwell
(commercial).

Post a reply to this message

Am 19.11.2011 03:02, schrieb Ben:
> Thanks!  Can you recommend any unbiased renderers, then?  My goal is to
> eventually produce real optical devices to test the predicted results (series of
> optics).
>
>
> Warp<war### [at] tagpovrayorg>  wrote:
>> Ben<nomail@nomail>  wrote:
>>> Can POVRay be used to make nearly-accurate-to-physics models for optics beyond
>>> refraction and reflection... to diffraction, interference, etc.?
>>
>>    Short answer: No.
>>
>>    The so-called unbiased renderers, which support BRDF definitions, are
>> designed for this exact purpose. In other words, if you are able to
>> describe the physical property of an object with a BRDF, the renderer
>> will then be able to simulate that property with a high degree of accuracy.
>> (Of course in many cases it will be extremely slow, and getting a good,
>> non-grainy result can take humongous amounts of time.)

It should be noted that unbiased renderers typically don't take you much 
further than POV-Ray with radiosity and photons enabled, as they don't 
model light as waves either, and therefore can't support interference 
(e.g. diffraction or iridescence) nor polarization-related effects (e.g. 
birefringence).

The main difference between biased (e.g. POV-Ray) and unbiased renderers 
is not primarily what types of illumination effects they support, but 
that (by definition) the latter use algorithms which are guaranteed to 
not introduce any artifacts except per-pixel statistical noise (which 
can be reduced by investing more computing time).

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Am 19.11.2011 03:02, schrieb Ben:
> > Thanks!  Can you recommend any unbiased renderers, then?  My goal is to
> > eventually produce real optical devices to test the predicted results (series of
> > optics).
> >
> >
> > Warp<war### [at] tagpovrayorg>  wrote:
> >> Ben<nomail@nomail>  wrote:
> >>> Can POVRay be used to make nearly-accurate-to-physics models for optics beyond
> >>> refraction and reflection... to diffraction, interference, etc.?
> >>
> >>    Short answer: No.
> >>
> >>    The so-called unbiased renderers, which support BRDF definitions, are
> >> designed for this exact purpose. In other words, if you are able to
> >> describe the physical property of an object with a BRDF, the renderer
> >> will then be able to simulate that property with a high degree of accuracy.
> >> (Of course in many cases it will be extremely slow, and getting a good,
> >> non-grainy result can take humongous amounts of time.)
>
> It should be noted that unbiased renderers typically don't take you much
> further than POV-Ray with radiosity and photons enabled, as they don't
> model light as waves either, and therefore can't support interference
> (e.g. diffraction or iridescence) nor polarization-related effects (e.g.
> birefringence).
>
> The main difference between biased (e.g. POV-Ray) and unbiased renderers
> is not primarily what types of illumination effects they support, but
> that (by definition) the latter use algorithms which are guaranteed to
> not introduce any artifacts except per-pixel statistical noise (which
> can be reduced by investing more computing time).

And phisically-based materials.

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> It should be noted that unbiased renderers typically don't take you much 
> further than POV-Ray with radiosity and photons enabled

  I wouldn't say that. As said, if the unbiased renderer supports defining
BRDF functions (and possibly other similar functions, such as BSDF's for
scattering, BTDF's for transmittance, and so on), it allows for a much
more accurate representation of physical surfaces, far beyond the simplistic
phong lighting model supported by POV-Ray (even if you enhance it with
diffuse inter-reflection of light between surfaces, photon mapping and
"faked" dispersion).

  Naturally there necessarily are limits to what even an unbiased rendered
with all those features can do (for example, it may be very difficult to
simulate interference in a physically accurate way, especially since
interference is affected by quantum effects that go well beyond what a
simple renderer supports and can simulate). However, I think that saying
"they don't take you much further than POV-Ray" is an exaggeration.

-- 
                                                          - Warp

Post a reply to this message

Diffraction and iridescence would be on my list of need-to-have, as there would
be edges and thin materials involved... but I'm thinking "calculus" rather than
"quantum" would be good enough, so to speak, in the ugrad physics sense.

Which tool would be good if I need to write a program to vary the parameters of
the model?  I had been looking at Pov-Ray since I could just generate the scene
files as text, but whether pov-like files or an API, it'd be fine as long as I
don't have to use an editor for every single variation.

Thanks for these great suggestions!

The only other way of proceeding I could think of is if there is some optics
simulator designed specifically for scientific estimate, rather than rendering.
(I Googled it, of course, but nothing particularly useful turned up, or else I
didn't have the right search phrase.)

Post a reply to this message

Am 19.11.2011 15:24, schrieb Ben:
> Diffraction and iridescence would be on my list of need-to-have, as there would
> be edges and thin materials involved... but I'm thinking "calculus" rather than
> "quantum" would be good enough, so to speak, in the ugrad physics sense.
>
> Which tool would be good if I need to write a program to vary the parameters of
> the model?  I had been looking at Pov-Ray since I could just generate the scene
> files as text, but whether pov-like files or an API, it'd be fine as long as I
> don't have to use an editor for every single variation.
>
> Thanks for these great suggestions!
>
> The only other way of proceeding I could think of is if there is some optics
> simulator designed specifically for scientific estimate, rather than rendering.
> (I Googled it, of course, but nothing particularly useful turned up, or else I
> didn't have the right search phrase.)

I recall a former work colleague of mine talking of how he had been 
involved in the writing of a piece of software to speed up the 
simulation of the distribution of electromagnetic waves (he was a telco 
engineer); said when he finally presented the piece of software in 
(IIRC) the 90's, some members of his audience were wearing military 
uniforms with lots of "lametta", and seemed to be pretty interested in 
the technology.

Morale of the story? I guess when military authorities are interested in 
ways to speed up physical simulations, chances are they're pretty 
time-consuming even on top-notch supercomputers. And I doubt whether 
even two decades of Moore's law have been enough to put that much 
computing power (plus the patience to wait for the results) at the 
fingertips of ordinary mortals by now.

Post a reply to this message