illum rules of thumb

I'm after both. (Renderings and falsecolor images mostly, but rtrace
calcs as well). Not sure I understand why it's such a bad idea to place
an illum plane on the outside of a window. (I am not using my glass as
the illum planes). Guess I'm going to have to try out a few test case
scenarios and see what seems to work better before running the real
thing.

Mark

Mark de la Fuente wrote:

Lars,

I wish I could attach images, as perhaps there is another way to do
this. But the reason for the exterior illums is that the interior
architecture follows the windows, forcing me to use triangular illums

at

some point if I am to seal each opening. (window panes are not
rectangular either) Or each opening can be sealed from the outside.

Hi Mark,

Yeah, triangular polygons are not a good fit with mkillum, because of
the way source sampling works. The triangular polygons are usually the

source of the "aiming failure" errors seen when running mkillum. The

best explanation is of course in the Rendering with Radiance text. Do

you have a copy? If so, this problem is addressed on pg 577.

Your building seems a challenge for mkillum. Are you trying to create

pretty pictures, or get some illuminance data at points? If the
former,
it may still be worth the effort to try and seal your opening -- from
the inside -- with a combination of polygons less likely to generate
failures. If the latter, it's probably best to simply crank up the
ambient calculation parameters, and wait it out. (This method could
also be used for pretty picture generation, but now you're talking
about
*really* waiting...

···

----

      Rob Guglielmetti

e. rpg at rumblestrip.org
w. www.rumblestrip.org

Hi Mark,

in short, the illums help you to reduce the errors which are caused when the ambient calculation sends out its random samples. If the lightsource (or illum) is outside, than the geometry between the source (or illum) will still cause errors. You can answer this by either setting the ambient paramters higher (so that every detail of your geometry is hit, which is very slow) or place the illum on the inside.

You might try to break down your illum panes into smaller surfaces. Don't use too much, as this will also slow down calculation, but maybe if you divide each face into some smaller faces, you can use rectangles?

As Rob said, if you can take a look at the chapter in "Rendering with Radiance", that might explain a lot.

Good luck, Lars.

···

Am Freitag, 30.07.04 um 23:25 Uhr schrieb Mark de la Fuente:

I'm after both. (Renderings and falsecolor images mostly, but rtrace calcs as well). Not sure I understand why it's such a bad idea to place an illum plane on the outside of a window. (I am not using my glass as the illum planes). Guess I'm going to have to try out a few test case scenarios and see what seems to work better before running the real thing.

Mark

--
Lars O. Grobe
[email protected]

--
Lars O. Grobe, [email protected], ++90-212-2458330

Kardesler Apartment, Turnacibasi Sokak 28
Galatasaray - Beyoglu, 34433 Istanbul

Mark de la Fuente wrote:

I'm after both. (Renderings and falsecolor images mostly, but rtrace
calcs as well). Not sure I understand why it's such a bad idea to place
an illum plane on the outside of a window.

Neither do I... :wink:

If you have windows with lots of subdividing geometry, then in
most cases the optimal solution is to turn each subpane into an
illum. In cases where this would create too many illums, or
illums with very non-rectangular shapes, it may be better to
place a big illum on the inside *or* the outside of the windows.
None of the two positions is inherently better or worse than the
other.

Advantages inside:
- Internal rays will hit the illum directly, which improves speed
  and smoothness of the rendering.
- You can make one or several illums per opening, depending on
  how close it is to any other geometry (the farther away the
  other geometry is, the larger the illum can be)
Disadvantages inside:
- Some of the rays cast by mkillum will hit your window geometry.
  Depending on the number of sampling positions (s=) and the type
  of geometry, this may negatively affect the accuracy (more or
  less rays may hit window geometry than statistically justified).
  Since the sampling positions are determined at random, this
  effect is hard to predict, let alone to correct. The need to go
  through the window geometry will also slow down mkillum to some
  degree.

Advantages outside:
- Since there is no obstruction by the window geometry, the
  values generated by mkillum will be very accurate, and the
  speed of the mkillum runs will only be determined by the
  complexity of your exterior scene.
- You can make one or several illums per opening, depending on
  external obstruction. If one part of an opening is shaded eg.
  by a tree, you can make that part an extra illum, which avoids
  averaging between areas of different illuminance.
Disadvantages outside:
- Direct sampling from the inside will have to get past the
  window geometry to hit the illums. This can affect both
  rendering speed and accuracy. You can improve the accuracy by
  suitable values for -ds and -dj (you should set those anyway
  when using illums). You can also subdivide your illums in a way
  that makes it more likely that their centers will be close to a
  clear part of your window geometry.

Considering all that, I wouldn't hesitate at all to place your
illum surfaces outside in your situation.

-schorsch

···

--
Georg Mischler -- simulations developer -- schorsch at schorsch com
+schorsch.com+ -- lighting design tools -- http://www.schorsch.com/

Hi!

Disadvantages outside:
- Direct sampling from the inside will have to get past the
  window geometry to hit the illums. This can affect both
  rendering speed and accuracy. You can improve the accuracy by
  suitable values for -ds and -dj (you should set those anyway
  when using illums). You can also subdivide your illums in a way
  that makes it more likely that their centers will be close to a
  clear part of your window geometry.

But what about the indirect calculation? As far as I understood, it is one of the main reasons to use illum surfaces at the windows to reduce the errors caused by the ambient calculation, which can cause visible artifacts. And this problem could still arise if the illum is placed outside. Am I wrong here?

CU Lars.

···

--
Lars O. Grobe
[email protected]

Lars O. Grobe wrote:

Hi!

> Disadvantages outside:
> - Direct sampling from the inside will have to get past the
> window geometry to hit the illums. This can affect both
> rendering speed and accuracy. You can improve the accuracy by
> suitable values for -ds and -dj (you should set those anyway
> when using illums). You can also subdivide your illums in a way
> that makes it more likely that their centers will be close to a
> clear part of your window geometry.

But what about the indirect calculation? As far as I understood, it is
one of the main reasons to use illum surfaces at the windows to reduce
the errors caused by the ambient calculation, which can cause visible
artifacts. And this problem could still arise if the illum is placed
outside. Am I wrong here?

The purpose of mkillum is to reduce the calculation required
outside your space. Reducing artifacts is a secondary effect.
It happens because the illum surface averages the large variety
of luminances found outside. Window geometry between the space
and the illum may have some adverse effect, but it should be
minor in comparison to the improvements mkillum gives you in
either case.

As always, need to weight the pros and cons in the context of
your actual scene.

-schorsch

···

--
Georg Mischler -- simulations developer -- schorsch at schorsch com
+schorsch.com+ -- lighting design tools -- http://www.schorsch.com/