[Radiance-general] Re: Radiance-general Digest, Vol 38, Issue 14
Rob Guglielmetti
rpg at rumblestrip.org
Fri Apr 27 02:21:11 CEST 2007
steve michel wrote:
> A linear fixture is exactly what I had in mind. For a while I even
> searched for 'bare bulb' photometry files from big lamp manufacturers
> (GE or Philips) but none exist for fluor tubes (only par and mr16
> spots). But is the illum distribution method another one or a subset
> of using flatcorr and lboxcorr techniques you describe??
No photometric files exist for bare fluorescent tubes probably because
you can't really test that condition, nor would anyone want to spend the
money to do so. A fluorescent tube will always be in a fixture, even if
it's a simple striplight -- a simple ballast housing and two
lampholders. And that striplight housing will block part of the lamp,
thereby altering the distribution, and THAT is what you would want to
photometer. There are photometric files for simple fluorescent
striplights, I believe people like H.E. Williams have them. The reason
there are photometric files for PARs -- a seemingly plain bare lamp
similar to the fluorescent lamp example -- is that those lamps have a
reflector and that reflector's effect IS worth evaluating. PARs are in
essence a "fixture", since they contain a light emitting filament as
well as an optical device to take that light and do something with it.
Photometric files are all about describing what a fixture does with
those lumens the lamp starts out with.
Anyway, the flatcorr and lboxcorr techniques are really enhancements to
the "illum distribution method" we've been discussing, and all of this
is a more complex but elegant way of describing the luminous output of a
given photometric file. Think of a photometric file as a description of
the fixture performance at a single point, because that's what it is.
This works great for a 5" round aperture recessed downlight; stick that
description on a disc-shaped piece of geometry, slap it on your ceiling
and render away. But when your photometric file is describing the
luminous performance of a linear fixture, you need to spread this out
along the lamp length. This is what the -ds option to rpict does, as we
discussed earlier. Yes, it's a hack, yes, it's less than perfect, and
yes, many smart people have offered alternate ways to photometer
fixtures to improve on this, but implementation and adoption of a
standard for that is not on the horizon, as far as I know.
So, we take our point "distribution description" and array it along a
polygon, at a frequency determined by the -ds option. This works, until
you have a three dimensional fixture model, or even a large flat fixture
that is close to other geometry. Then you must stretch and pull that
point description into a box shape, and apply it to an illum material.
Illum is a magical material (I know Greg is cringing hearing anything in
Radiance described as 'magic'; trust me, it's still physically-based!)
that can emit light but becomes invisible when you look directly at it.
So the idea is to create this "imposter geometry" that acts as the
light-emitting boundary and origin for the fixture. Within the box you
have 3D geometry that represents the physical appearance of the fixture,
but the impostor box is doing all the lighting. So flatcorr and
lboxcorr are just little helper files that take that point "distribution
description" and do all the stretching and pulling. If the ies file was
for a table lamp, you might be able to use a simple sphere to do the
same thing (you'd describe a sphere whose radius is larger than the
lampshade, which in this case is the thing doing the emitting, in the
physical world).
Are we having fun yet? Well, that's the general idea of what these do.
This is not an alternate description of the distribution! It's a way of
taking the distribution (again, described as a point) and mapping it to
a 3D form. For raw quantitative analysis, it's unnecessary. But if you
want 3D fixtures in your renderings, it's a process you'll need to
understand. The juicy bits of the technique start on page 320 in
Rendering with Radiance.
- Rob Guglielmetti
More information about the Radiance-general
mailing list