mkillum with 400 sources?

Hi!

I am using mkillum on a scene with a lot of faces converted to illums (about 400). This is rather fine-grained and results in longe times spent on mkillum. Now I wonder if I should tweak the rtrace-settings a bit to get quicker and still not too bad results, so far I use defaults. Background: the sources represent glass panes in windows, so each window is allready divided into multiple faces. This is one of the causes for the many illums that are calculated.

What are the settings you use with mkillum, in a daylight-scene, with rather small openings (because of the divided windows)?

TIA+CU Lars.

P.S.: I have been waiting 2 hours for 40 illums, so I am a bit worried about this now :wink:

Hello again.

I have to ask this in a more general way. I have used mkillum to add some illum planes in front of the windows of my scene, and sealed the sides using zero-radiance illums. I am rendering the illums with -s=8 -d=8, and in total there are about 400 calculated illums (and many more zero-radiance illums).

Now, I am experiencing even slower calculation with rpict. I guess it is because of all the illum sources. I think that I can reduce the -ab bounces because of the illum useage, which will give faster rendering times. But still I wonder if there is some kind of a magic limit of the number of illums to use. I mean, is there a number of illums where rendering speed decreases instead of increasing by the useage of the mkillum-tool ?

Are there any other ambient-parameters that I can change now that I have illums? I am thinking of ad, ar, aa, as the light environment is quite different now?

You see, I am still quite a beginner in this mkillum stuff, and a bit worried. Thanks for any help and hints, CU

Lars.

Hi Lars,

Putting illum surfaces inside of your windows rather than using a separate surface for each window pane is one way to reduce calculation times if your mullions are not particularly large or prominent.

All else being equal, you can reduce -ab by 1 and decrease -ad and -as (by a factor of 2 or so) and even increase -aa by 50% and still get similar results. However, you may also have to set -dj .8 and -ps 1, which will undermine the calculation speeds a bit. Overall, illum's pay off most when you have smaller windows or skylights and/or complex goings on outside.

-Greg

···

From: "Lars O. Grobe" <[email protected]>
Date: October 9, 2005 9:22:57 AM PDT

Hello again.

I have to ask this in a more general way. I have used mkillum to add some illum planes in front of the windows of my scene, and sealed the sides using zero-radiance illums. I am rendering the illums with -s=8 -d=8, and in total there are about 400 calculated illums (and many more zero-radiance illums).

Now, I am experiencing even slower calculation with rpict. I guess it is because of all the illum sources. I think that I can reduce the -ab bounces because of the illum useage, which will give faster rendering times. But still I wonder if there is some kind of a magic limit of the number of illums to use. I mean, is there a number of illums where rendering speed decreases instead of increasing by the useage of the mkillum-tool ?

Are there any other ambient-parameters that I can change now that I have illums? I am thinking of ad, ar, aa, as the light environment is quite different now?

You see, I am still quite a beginner in this mkillum stuff, and a bit worried. Thanks for any help and hints, CU

Lars.

Hi Greg,

thanks for your reply. In fact, I am using surfaces not for each pane, but for each opening now, so that is why I have "only" these 400 faces :wink: I tried reducing -ab, -ad and -as, but rendering times are *really* slow now. I have all these zero-radiance illums (sealings), do they have an effect on the rendering speed, even when zero?

You know the scene, it is this church scene I showed you, so it has a lot of these small windows, and as the fenestration is rather complex, I hoped that using illum "coffers" around the windows would improve rendering speed as well as accuracy. However, so far the speed is really disastrous. Outside the building, there is not much detail. I even tried to apply skyfunc to the window panes, but as I have to apply it to *every* small pane than (because I can't use the coffer trick to integrate the whole window than), I get so many hundreds of light sources that calculation is almost impossible.

TIA+CU Lars.

···

Putting illum surfaces inside of your windows rather than using a separate surface for each window pane is one way to reduce calculation times if your mullions are not particularly large or prominent.

All else being equal, you can reduce -ab by 1 and decrease -ad and -as (by a factor of 2 or so) and even increase -aa by 50% and still get similar results. However, you may also have to set -dj .8 and -ps 1, which will undermine the calculation speeds a bit. Overall, illum's pay off most when you have smaller windows or skylights and/or complex goings on outside.

One small trick I use (during the mkillum source creation) to speed up the process is to exclude interior geometry that does not affect the light coming into the space. Also, if there is no exterior geometry, I've created arrays of processed illum planes rather than create many identical illums.

Hope that helps.

Mark

Message: 2

···

Date: Sun, 9 Oct 2005 18:22:57 +0200
From: "Lars O. Grobe" <[email protected]>
Subject: Re: [Radiance-general] mkillum with 400 sources?
To: Radiance general discussion <[email protected]>
Message-ID: <[email protected]>
Content-Type: text/plain; charset=US-ASCII; format=flowed

Hello again.

I have to ask this in a more general way. I have used mkillum to add
some illum planes in front of the windows of my scene, and sealed the
sides using zero-radiance illums. I am rendering the illums with -s=8
-d=8, and in total there are about 400 calculated illums (and many more
zero-radiance illums).

Now, I am experiencing even slower calculation with rpict. I guess it
is because of all the illum sources. I think that I can reduce the -ab
bounces because of the illum useage, which will give faster rendering
times. But still I wonder if there is some kind of a magic limit of the
number of illums to use. I mean, is there a number of illums where
rendering speed decreases instead of increasing by the useage of the
mkillum-tool ?

Are there any other ambient-parameters that I can change now that I
have illums? I am thinking of ad, ar, aa, as the light environment is
quite different now?

You see, I am still quite a beginner in this mkillum stuff, and a bit
worried. Thanks for any help and hints, CU

Lars.

Hi,

ok, I found the problem, not the solution. I have a cad model for the faces, and they are all polygons. Now, if I use obj and obj2rad, they are all triangulated. And if I use dxf and dxf2rad, only 3- or 4-sided polygons are possible. One way would be to use polylines (closed) with dxf2rad, but somehow, this gives me just an empty rad file. At the moment I have really no idea how to get my illum surfaces (=polygons) from cad to radiance...

CU Lars.

Lars O. Grobe wrote:

Hi,

ok, I found the problem, not the solution. I have a cad model for the
faces, and they are all polygons. Now, if I use obj and obj2rad, they
are all triangulated. And if I use dxf and dxf2rad, only 3- or 4-sided
polygons are possible. One way would be to use polylines (closed) with
dxf2rad, but somehow, this gives me just an empty rad file.

Make sure that you're using 2D-Polylines. 3DPolys will be ignored,
even if they are actually planar (it would be too much work for
dxf2rad to check that). Also make sure that they are really closed
with the "closed flag". It is not enough to just place the last
vertex on top of the first one.

-schorsch

···

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

Make sure that you're using 2D-Polylines. 3DPolys will be ignored,
even if they are actually planar (it would be too much work for
dxf2rad to check that). Also make sure that they are really closed
with the "closed flag". It is not enough to just place the last
vertex on top of the first one.

Hi Schorsch,

thank you for the hint, for someone not that familiar in Autocadland this is really hard... especially as I am currently working in formZ and have to get the geometry (dxf) from there.

I did it a bit different now. I read again the obj2rad manpage, and found out that the obj geometry is triangulated ONLY if not planar OR if there is surface normal information included in the file. So I switched off the normals - et voilà, I got the polygons!

I have just started the next run of mkillum now, which is much more promising, cause there is really nothing worse to feed to mkillum than triangulated faces. I hope the normals will still be right, as the faces were editet correctly, but you never know what cad / modelling software is doing to your scene...

So thank You, and sorry for some of my questions which were a bit misleading cause the problem was completely different than I had guessed, but I am just getting crazy with that mkillum step in my scene at the moment :wink:

CU, greetings to Munich, Lars.

If you are using AutoCAD, then you might try Georg Mischler's radout. It's much better than torad. Never used dxf2rad, but I always make sure and use 3d faces to draw illum sources (and most others for that matter). That way you can shade and/or turn the back face off within cad and check to make sure the faces are drawn in the right orientation. What I have not been able to figure out (and perhaps someone knows) is how to check to make sure all 3D face points are coplanar. I've run into problems before where objects are not coplanar, but with hundreds of planes, who knows which one is drawn incorrectly! The only quick fix to that is to triangulate.....

Mark

···

Date: Mon, 10 Oct 2005 17:57:44 +0200
From: "Lars O. Grobe" <[email protected]>
Subject: Re: [Radiance-general] mkillum with 400 sources?
To: Radiance general discussion <[email protected]>
Message-ID: <[email protected]>
Content-Type: text/plain; charset=US-ASCII; format=flowed

Hi,

ok, I found the problem, not the solution. I have a cad model for the
faces, and they are all polygons. Now, if I use obj and obj2rad, they
are all triangulated. And if I use dxf and dxf2rad, only 3- or 4-sided
polygons are possible. One way would be to use polylines (closed) with
dxf2rad, but somehow, this gives me just an empty rad file. At the
moment I have really no idea how to get my illum surfaces (=polygons)
from cad to radiance...

CU Lars.

Never used dxf2rad, but I always make sure and use 3d faces to draw illum sources (and most others for that matter).

Hi,

yes, I guess for most this is ok. But if you have polygonal surfaces (which means that there may be more than 4 vertices), 3dfaces cannot represent these any more without first splitting them to triangular or quadrangular faces, if I understood the dxf-faces right. Am I wrong here? And triangular faces are the worst to use for mkillum.

Closed polylines might work, but as you can see from Schorsch's mail, there is again a difference between 2d and 3d polylines. And here, if you are not working with Autocad, it is really difficult to find out what kind of polylines are written to your dxf...

CU Lars.