Spots when recreating Turrell installation

Folks,

If you have the chance to go to the Guggenheim in NYC, go and see the
James Turrell exhibit. You've never seen the Guggenheim quite like
this before, and as appreciators of light and its behavior, you will
find yourself both satisfied and intrigued by his installations.

I've been working to recreate a few of them in Radiance, and am
running into some difficulties. One piece, "Afrum I (White)", should
look like this:

It's a point light in a small box above the entryway to the room, a
special 6-sided hole cut in the wall, and the light beams this
cube-like shape onto the corner. The light patch illuminates the room.

Instead, what I get looks like this:

Here are the relevant rendering parameters:
-ds 0.02 -lw 1e-5 -ab 5 -aa 0.03 -ar 256 -ad 1024 -as 512

I suspect that I simply need more ambient subsamples, as a few ambient
points are catching the patch and some are not. But increasing -ad and
-as do nothing to improve quality.

Or, WAIT, are those bright spots seeing the little hole (it's 6" wide)
through which the light is originally emerging? How can I prevent
that?

Mark

Hi Mark,

Funny you should be trying to simulate this -- my wife (who is an artist herself) was just down in L.A. and Turrell has what must be a very similar exhibit there, as she sent me a nearly identical photo of it.

My first guess would be the same as yours, which would be to crank the -ad parameter much higher. You may not need to increase the -as all that much. Finding the bright spots is the difficult part, since it is so near the horizon as seen from points on the adjacent walls. (Cosine-weighted sampling means that the horizon gets fewer samples than the zenith.)

To avoid any contributions from the source box, make the surrounding out of a special material and name it with an -ae option.

Failing that, I would go full Monte Carlo and see if that works:

  -aa 0 -ad 1024 -as 128 -lw .0002 -ab 2

Your scene breaks one of the basic assumptions in the Radiance indirect calculation, which says that indirect lighting is fairly well-distributed in the space. Here, you have a patch acting as a light source.

Once I implement the irradiance Hessian described in the recent Siggraph Asia paper by Schwarzhaupt et al., the situation might improve:

  http://zurich.disneyresearch.com/~wjarosz/publications/schwarzhaupt12practical.html

Best of luck!
-Greg

···

From: Mark Stock <[email protected]>
Date: September 1, 2013 8:08:28 PM PDT

Folks,

If you have the chance to go to the Guggenheim in NYC, go and see the
James Turrell exhibit. You've never seen the Guggenheim quite like
this before, and as appreciators of light and its behavior, you will
find yourself both satisfied and intrigued by his installations.

I've been working to recreate a few of them in Radiance, and am
running into some difficulties. One piece, "Afrum I (White)", should
look like this:
http://nyogalleristny.files.wordpress.com/2013/08/afrum1.jpg

It's a point light in a small box above the entryway to the room, a
special 6-sided hole cut in the wall, and the light beams this
cube-like shape onto the corner. The light patch illuminates the room.

Instead, what I get looks like this:
http://markjstock.org/transfer/img06.png

Here are the relevant rendering parameters:
-ds 0.02 -lw 1e-5 -ab 5 -aa 0.03 -ar 256 -ad 1024 -as 512

I suspect that I simply need more ambient subsamples, as a few ambient
points are catching the patch and some are not. But increasing -ad and
-as do nothing to improve quality.

Or, WAIT, are those bright spots seeing the little hole (it's 6" wide)
through which the light is originally emerging? How can I prevent
that?

Mark

Greg, interested parties,

Looking back at the hole in the wall (made with antimatter) I noticed
that even putting the "light bulb room" wall material and the
antimatter material in the ambient exclude list, there was still
ambient light pouring from the little room into the main viewing room.
That's what was causing the errant ambient spots. Possibly because
antimatter messes up the ambient exclude logic?

As soon as I changed the color of the little "light bulb room" to pure
black, the situation improved:

Old: http://markjstock.org/transfer/img06.png
New: http://markjstock.org/transfer/img11.png

Ambient parameters might actually be too high. But the rendering looks
nice! Turrell must have used a very black little room for his light
source!

Files: http://markjstock.org/transfer/afrum1.tar

Improvements/comments welcome. More Turrell pieces are in the works,
as I figure them out.

Mark

···

On 9/2/13, Greg Ward <[email protected]> wrote:

Hi Mark,

Funny you should be trying to simulate this -- my wife (who is an artist
herself) was just down in L.A. and Turrell has what must be a very similar
exhibit there, as she sent me a nearly identical photo of it.

My first guess would be the same as yours, which would be to crank the -ad
parameter much higher. You may not need to increase the -as all that much.
Finding the bright spots is the difficult part, since it is so near the
horizon as seen from points on the adjacent walls. (Cosine-weighted
sampling means that the horizon gets fewer samples than the zenith.)

To avoid any contributions from the source box, make the surrounding out of
a special material and name it with an -ae option.

Failing that, I would go full Monte Carlo and see if that works:

  -aa 0 -ad 1024 -as 128 -lw .0002 -ab 2

Your scene breaks one of the basic assumptions in the Radiance indirect
calculation, which says that indirect lighting is fairly well-distributed in
the space. Here, you have a patch acting as a light source.

Once I implement the irradiance Hessian described in the recent Siggraph
Asia paper by Schwarzhaupt et al., the situation might improve:

  http://zurich.disneyresearch.com/~wjarosz/publications/schwarzhaupt12practical.html

Best of luck!
-Greg

From: Mark Stock <[email protected]>
Date: September 1, 2013 8:08:28 PM PDT

Folks,

If you have the chance to go to the Guggenheim in NYC, go and see the
James Turrell exhibit. You've never seen the Guggenheim quite like
this before, and as appreciators of light and its behavior, you will
find yourself both satisfied and intrigued by his installations.

I've been working to recreate a few of them in Radiance, and am
running into some difficulties. One piece, "Afrum I (White)", should
look like this:
http://nyogalleristny.files.wordpress.com/2013/08/afrum1.jpg

It's a point light in a small box above the entryway to the room, a
special 6-sided hole cut in the wall, and the light beams this
cube-like shape onto the corner. The light patch illuminates the room.

Instead, what I get looks like this:
http://markjstock.org/transfer/img06.png

Here are the relevant rendering parameters:
-ds 0.02 -lw 1e-5 -ab 5 -aa 0.03 -ar 256 -ad 1024 -as 512

I suspect that I simply need more ambient subsamples, as a few ambient
points are catching the patch and some are not. But increasing -ad and
-as do nothing to improve quality.

Or, WAIT, are those bright spots seeing the little hole (it's 6" wide)
through which the light is originally emerging? How can I prevent
that?

Mark

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general