Hi!

I'm experimenting with POVRay to simulate a simple physics-lab setup involving
the following:

A spot light
A Mirror Surface (does not need to be visible, just must perfectly reflect the
light.)
Some shadow generating objects before and after the mirror (invisible to the
camera too)
A semi-transparent Screen that catches and displays the reflected photons.

I've accomplished this task already by using "double_illuminate" for the
screen., but I'm struggling with optimizing the render time.
I understand that photons take their time to be calculated. But for a simple
setup like
this I hope I can keep the render time under halve a minute for a 600x800 image.
Can someone give me some hands on tipps about best global settings and about how
useful
commands like "autostop" and other options like "adc_bailout" could be?
Maybe with a little example script, optimized for fast calculation, showing a
sharp and crisp projection of light reflected by a rectangular mirror against a
screen. I don't need fancy blending and anti-alias. Speed is what I seek :-) ...
Thanks in advance!!

PS: I think POVRay is the cooles thing since melted cheese!

Andi

Post a reply to this message

CAD-Andi napsal(a):
> Hi!
> 
> I'm experimenting with POVRay to simulate a simple physics-lab setup involving
> the following:
> 
> A spot light
> A Mirror Surface (does not need to be visible, just must perfectly reflect the
> light.)
> Some shadow generating objects before and after the mirror (invisible to the
> camera too)
> A semi-transparent Screen that catches and displays the reflected photons.
> 
> I've accomplished this task already by using "double_illuminate" for the
> screen., but I'm struggling with optimizing the render time.
> I understand that photons take their time to be calculated. But for a simple
> setup like
> this I hope I can keep the render time under halve a minute for a 600x800 image.
> Can someone give me some hands on tipps about best global settings and about how
> useful
> commands like "autostop" and other options like "adc_bailout" could be?
> Maybe with a little example script, optimized for fast calculation, showing a
> sharp and crisp projection of light reflected by a rectangular mirror against a
> screen. I don't need fancy blending and anti-alias. Speed is what I seek :-) ...
> Thanks in advance!!
> 
> PS: I think POVRay is the cooles thing since melted cheese!
> 
> Andi
> 
> 
the 'autostop' doesn't help much since the mirror is equal to its own 
bounding box
I guess the photons only take one bounce - off the mirror - so the never 
bailout, either
try reducing the photon count or increasing the spacing (whichever you use)
save_file the photons if you can

Post a reply to this message

Jan Dvorak <jan### [at] centrumcz> wrote:
> CAD-Andi napsal(a):
> > Hi!
> >
> > I'm experimenting with POVRay to simulate a simple physics-lab setup involving
> > the following:
> >
> > A spot light
> > A Mirror Surface (does not need to be visible, just must perfectly reflect the
> > light.)
> > Some shadow generating objects before and after the mirror (invisible to the
> > camera too)
> > A semi-transparent Screen that catches and displays the reflected photons.
> >
> > I've accomplished this task already by using "double_illuminate" for the
> > screen., but I'm struggling with optimizing the render time.
> > I understand that photons take their time to be calculated. But for a simple
> > setup like
> > this I hope I can keep the render time under halve a minute for a 600x800 image.
> > Can someone give me some hands on tipps about best global settings and about how
> > useful
> > commands like "autostop" and other options like "adc_bailout" could be?
> > Maybe with a little example script, optimized for fast calculation, showing a
> > sharp and crisp projection of light reflected by a rectangular mirror against a
> > screen. I don't need fancy blending and anti-alias. Speed is what I seek :-) ...
> > Thanks in advance!!
> >
> > PS: I think POVRay is the cooles thing since melted cheese!
> >
> > Andi
> >
> >
> the 'autostop' doesn't help much since the mirror is equal to its own
> bounding box
> I guess the photons only take one bounce - off the mirror - so the never
> bailout, either
> try reducing the photon count or increasing the spacing (whichever you use)
> save_file the photons if you can

Jan,

Thanks for your answer. Does saving the photon map do any good if the
orientation of the mirror changes from frame to frame? I suspect not, because
the photons would be reflected into a diferent direction. Right?
Is there anything else one could think of? Reducing the number of photons
resulted in poor coverage of the screen, expecially when the light hits the
screen under a slim angle.
Thanks again ...

Andi

Post a reply to this message

CAD-Andi napsal(a):
> Jan Dvorak <jan### [at] centrumcz> wrote:
>> CAD-Andi napsal(a):
>>> Hi!
>>>
>>> I'm experimenting with POVRay to simulate a simple physics-lab setup involving
>>> the following:
>>>
>>> A spot light
>>> A Mirror Surface (does not need to be visible, just must perfectly reflect the
>>> light.)
>>> Some shadow generating objects before and after the mirror (invisible to the
>>> camera too)
>>> A semi-transparent Screen that catches and displays the reflected photons.
>>>
>>> I've accomplished this task already by using "double_illuminate" for the
>>> screen., but I'm struggling with optimizing the render time.
>>> I understand that photons take their time to be calculated. But for a simple
>>> setup like
>>> this I hope I can keep the render time under halve a minute for a 600x800 image.
>>> Can someone give me some hands on tipps about best global settings and about how
>>> useful
>>> commands like "autostop" and other options like "adc_bailout" could be?
>>> Maybe with a little example script, optimized for fast calculation, showing a
>>> sharp and crisp projection of light reflected by a rectangular mirror against a
>>> screen. I don't need fancy blending and anti-alias. Speed is what I seek :-) ...
>>> Thanks in advance!!
>>>
>>> PS: I think POVRay is the cooles thing since melted cheese!
>>>
>>> Andi
>>>
>>>
>> the 'autostop' doesn't help much since the mirror is equal to its own
>> bounding box
>> I guess the photons only take one bounce - off the mirror - so the never
>> bailout, either
>> try reducing the photon count or increasing the spacing (whichever you use)
>> save_file the photons if you can
> 
> Jan,
> 
> Thanks for your answer. Does saving the photon map do any good if the
> orientation of the mirror changes from frame to frame? I suspect not, because
> the photons would be reflected into a diferent direction. Right?
Right.
> Is there anything else one could think of? Reducing the number of photons
> resulted in poor coverage of the screen, expecially when the light hits the
> screen under a slim angle.
> Thanks again ...
> 
> Andi
> 
>

Post a reply to this message

CAD-Andi nous apporta ses lumieres en ce 2008/01/08 23:11:
> Hi!
> 
> I'm experimenting with POVRay to simulate a simple physics-lab setup involving
> the following:
> 
> A spot light
> A Mirror Surface (does not need to be visible, just must perfectly reflect the
> light.)
> Some shadow generating objects before and after the mirror (invisible to the
> camera too)
> A semi-transparent Screen that catches and displays the reflected photons.
> 
> I've accomplished this task already by using "double_illuminate" for the
> screen., but I'm struggling with optimizing the render time.
> I understand that photons take their time to be calculated. But for a simple
> setup like
> this I hope I can keep the render time under halve a minute for a 600x800 image.
> Can someone give me some hands on tipps about best global settings and about how
> useful
> commands like "autostop" and other options like "adc_bailout" could be?
> Maybe with a little example script, optimized for fast calculation, showing a
> sharp and crisp projection of light reflected by a rectangular mirror against a
> screen. I don't need fancy blending and anti-alias. Speed is what I seek :-) ...
> Thanks in advance!!
> 
> PS: I think POVRay is the cooles thing since melted cheese!
> 
> Andi
> 
> 
What exactly do you see?
You may be able to go by without using photons. If you only see the screen and 
not the mirror nor the objects that cast the shadows, you can use some tricks.
Use an orthographic camera with a projection plane equivalent to the screen.
Replace the mirror by a white surface.
Have all objects be completely black. pigment{rgb 0}
Keep the spotlight as it is.

The objects before the surface will cast shadows on the plane.
The objects after the plane will hide it and show as black silouetes.
Net result: only use direct view and only test for shadow.

Autostop is usefull if you have objects that don't fill ther bounding box, not 
usefull in your case.
adc_bailout is for multiple reflections and transmission. It limit the number of 
surfaces treated when the surface contribution drop under a given treshold 
value: not your case.

-- 
Alain
-------------------------------------------------
Congregationalist: Shit that happens to one person is just as good as shit that 
happens to another.

Post a reply to this message

> A Mirror Surface (does not need to be visible, just must perfectly reflect
the
> light.)

If it's a planar mirror, you don't need photons at all. Just make a flat box
that represents the surface of the mirror and use this with a
projected_through light source. If this light source is on the opposite side
of the mirror, so that it's in the "mirror image" position of the real
light, then it will send light in exactly the right direction as though it
bounced off the mirror.

You might have to move the flat box very slightly off the surface of the
mirror to avoid coincident surface issues.

 - Slime
 [ http://www.slimeland.com/ ]

Post a reply to this message

Alain <ele### [at] netscapenet> wrote:
> What exactly do you see?
> You may be able to go by without using photons. If you only see the screen and
> not the mirror nor the objects that cast the shadows, you can use some tricks.
> Use an orthographic camera with a projection plane equivalent to the screen.
> Replace the mirror by a white surface.
> Have all objects be completely black. pigment{rgb 0}
> Keep the spotlight as it is.
>
> The objects before the surface will cast shadows on the plane.
> The objects after the plane will hide it and show as black silouetes.
> Net result: only use direct view and only test for shadow.
>
> Autostop is usefull if you have objects that don't fill ther bounding box, not
> usefull in your case.
> adc_bailout is for multiple reflections and transmission. It limit the number of
> surfaces treated when the surface contribution drop under a given treshold
> value: not your case.
>
> --
> Alain
> -------------------------------------------------
> Congregationalist: Shit that happens to one person is just as good as shit that
> happens to another.

Alain,

Interesting idea. Thinking through it though, for my application, i think it
would not work. Shadows grow with the distance of the object to the screen.
Using parallel projection is not going to give me that effect. Also my screen
is not paralell to the mirror glass. (I think I did not mention that in my
scene description, sorry.) Thanks, anyway!

Post a reply to this message

"Slime" <fak### [at] emailaddress> wrote:
> > A Mirror Surface (does not need to be visible, just must perfectly reflect
> the
> > light.)
>
> If it's a planar mirror, you don't need photons at all. Just make a flat box
> that represents the surface of the mirror and use this with a
> projected_through light source. If this light source is on the opposite side
> of the mirror, so that it's in the "mirror image" position of the real
> light, then it will send light in exactly the right direction as though it
> bounced off the mirror.
>
> You might have to move the flat box very slightly off the surface of the
> mirror to avoid coincident surface issues.
>
>  - Slime
>  [ http://www.slimeland.com/ ]

Hi Slime!

Absolutely right! This works perfect for a planar mirror. Unfortunately I have
to deal with convex and free-form shaped mirror surfaces. Great idea though and
I may end up implementing it for the case of planar mirrors. Thanks!
Anything else how I really could give those photons a push?

Post a reply to this message

Another thought:

Would a smaller ammount of mirroring surface speed up rendering? Let's imagine a
rectangular mirror 100x200 that bounces photons onto a screen (screen not
parallel to the mirror surface). The lit up area on the screen would look like
a certain polygonal shape and would be filled completely. Would it take less
time rendering if the rectangle was only reflecting photons from a 5mm rim
around the perimeter of the mirror creating somewhat of an outline polygonal
shape on the screen?

Post a reply to this message

CAD-Andi nous apporta ses lumieres en ce 2008/01/15 14:07:
> Another thought:
> 
> Would a smaller ammount of mirroring surface speed up rendering? Let's imagine a
> rectangular mirror 100x200 that bounces photons onto a screen (screen not
> parallel to the mirror surface). The lit up area on the screen would look like
> a certain polygonal shape and would be filled completely. Would it take less
> time rendering if the rectangle was only reflecting photons from a 5mm rim
> around the perimeter of the mirror creating somewhat of an outline polygonal
> shape on the screen?
> 
> 
If the spotlight don't cober the whole surface, no.
If the surface is not flat, not realy.
About shooting only at the rim, no gain there, as you still start shooting 
photons at the center of the bounding box. The only effect will be that those 
photons will be blocked before they reatch the target.

A question: do you use more than one light_source? Are they at the same location?

If you answered yes to both, replace them with a single light with an intensity 
egual to the sum of all current sources. This will reduce the number of protons 
to shoot and improve the quality. When several coincident light_source are 
shooting photons at the same target, they all shoot photons in the same pattern. 
It will also cut down on the shadow testing, resulting in another speed increase.

-- 
Alain
-------------------------------------------------
EVERYTHING HAS A GENDER

You may not know this but many nonliving things have a gender...

A Web Page is Female, because it's always getting hit on.

Post a reply to this message