aiming illums, orthographic projection, and hiding surfaces

radiance-general
1

Hi all,

I had previously posted on this list requesting help installing Radiance on Mac OS X Jaguar. Thanks for all who helped. I know have it installed, and have started playing around with it. I have walked through the online tutorial, and have ordered Rendering with Radiance, and am anxious to get it in my hands. In the meantime (it is currently out of stock), I need to get an analysis completed for a project I am working on, and would appreciate some help.

However, the following problems have occured under Linux (I haven't had a chance to run this on the Mac).
First, when I use rview to preview my scene, everything seems to pop up normally (I get an image). However, when I use the rpict command, I always get the message
"warning - aiming failure for light source [a]" where a stands for an illum object (the window in the scene). As of now, I am only analyzing daylighting, so I have no other light sources in the scene.

Second, I'm sure this is covered in "Rendering with Radiance," but is there a way to get orthogonal and orthographic projected views in Radiance? Moreover, is there a way to make a surface invisible, yet be considered in the lighting calculation? I would like to look into an interior space in an axonometric view, but look through the ceiling and 1 or 2 walls to see the light distribution throughout a single space.

Thanks.

John An

2

John An wrote:
> Hi all,
>
> I had previously posted on this list requesting help installing
> Radiance on Mac OS X Jaguar. Thanks for all who helped. I know have
> it installed, and have started playing around with it. I have walked
> through the online tutorial, and have ordered Rendering with
> Radiance, and am anxious to get it in my hands. In the meantime (it
> is currently out of stock), I need to get an analysis completed for a
> project I am working on, and would appreciate some help.

Sorry to say, but that book is not out of stock, it's out of print (well, OK it's out of stock too, but permanently). The authors have obtained the copyright from Morgan Kaufman, and are looking for a new publisher, but so far no joy. My co-worker tried ordering direct from MK and they told him the same "out of stock" thing, but no new copies are coming out of MK. A couple of enterprising souls were selling copies on Amazon.com for something like $300USD! =8-( Hopefully, Greg will find a new publisher or some other means of distrubition; it's such a valuable piece of the Radiance toolkit.

> However, the following problems have occured under Linux (I haven't
> had a chance to run this on the Mac). First, when I use rview to
> preview my scene, everything seems to pop up normally (I get an
> image). However, when I use the rpict command, I always get the
> message "warning - aiming failure for light source [a]" where a
> stands for an illum object (the window in the scene). As of now, I
> am only analyzing daylighting, so I have no other light sources in
> the scene.

That happens when you have an irregularly shaped polygon that is defined as an illum. "L-shaped" polys and triangles are the kinds of shapes I'm talking about here. Break it up into more rectillinear forms and you should see that problem go away.

> Second, I'm sure this is covered in "Rendering with Radiance," but is
> there a way to get orthogonal and orthographic projected views in
> Radiance? Moreover, is there a way to make a surface invisible, yet
> be considered in the lighting calculation? I would like to look into
> an interior space in an axonometric view, but look through the
> ceiling and 1 or 2 walls to see the light distribution throughout a
> single space.

Check the manpage for rpict, it explains all the view parameters. The way to see "through" a surface is to use clipping planes in your views. This way, the calculation is correct, but at view time you slice through the building with the use of these clipping planes. These are also explained in the rtrace manual page.

···

----

      Rob Guglielmetti

e. [email protected]
w. www.rumblestrip.org

3

Originally written by Rob Guglielmetti:
<I>Sorry to say, but that book is not out of stock, it's out of print
(well, OK it's out of stock too, but permanently). The authors have
obtained the copyright from Morgan Kaufman, and are looking for a new
publisher, but so far no joy. My co-worker tried ordering direct from
MK and they told him the same "out of stock" thing, but no new copies
are coming out of MK. A couple of enterprising souls were selling copies
on Amazon.com for something like $300USD! =8-( Hopefully, Greg will
find a new publisher or some other means of distrubition; it's such a
valuable piece of the Radiance toolkit.</I>

I had run across some threads discussing trying to find a new publisher, but thought that the situation had been resolved, since Elsevier Science happily took my order. When I called to inquire about my order, they confidenctly said that they would have the book in stock on May 1st. Is there any way that I could temporarily get my hands on the book in the meantime? Is there anyone in the Boston area that would be willing to lend me their book? After years of hearing how great Radiance was, I finally got the courage to jump in and to learn to use this program. The lack of availability of Rendering with Radiance has put a HUGE damper on my plans (though I am still determined to use it for my dissertation).

Originally written by Rob Guglielmetti:
<I>That happens when you have an irregularly shaped polygon that is defined
as an illum. "L-shaped" polys and triangles are the kinds of shapes I'm
talking about here. Break it up into more rectillinear forms and you
should see that problem go away.</I>

I see. Thankfully, I only have 4 or 5 windows in my scene. I dxfed my model from a wondering modeling program called Sketchup (the most intuitive modeling program I've come across), and had been meaning to optimize the .rad file, but didn't realize that the excessive polygons would impact the calculations. Does this mean that I should optimize (de-triangulate the rectangles) for all surfaces if I want accurate calculations?

Thanks for your response.

Please, if anyone can help me out with getting my hands on Rendering with Radiance, I would truly appreciate it.

John An

4

I had run across some threads discussing trying to find a new publisher, but thought that the situation had been resolved, since Elsevier Science happily took my order. When I called to inquire about my order, they confidenctly said that they would have the book in stock on May 1st. Is there any way that I could temporarily get my hands on the book in the meantime? Is there anyone in the Boston area that would be willing to lend me their book? After years of hearing how great Radiance was, I finally got the courage to jump in and to learn to use this program. The lack of availability of Rendering with Radiance has put a HUGE damper on my plans (though I am still determined to use it for my dissertation).

Well, I guess we'll find out tomorrow then. =8-) I don't recall any announcements from Greg about this situation getting resolved, but if it's true that's GREAT! Do let us all know how you make out. I'd still like to get another copy for the office.

Congrats on taking the plunge with Radiance. It's a great program (that I am only beginning to understand), and the massive effort you put into learning it *will* be rewarded. And this list is frequented by many of the gurus, and they are quite open and generous with their advice. It's a great bunch of people on here.

I see. Thankfully, I only have 4 or 5 windows in my scene. I dxfed my model from a wondering modeling program called Sketchup (the most intuitive modeling program I've come across), and had been meaning to optimize the .rad file, but didn't realize that the excessive polygons would impact the calculations. Does this mean that I should optimize (de-triangulate the rectangles) for all surfaces if I want accurate calculations?

It's not the number of polys that's causing that aiming failure, it's the shape. And it's only with illum sources. So no, you don't need to de-triangulate all your surfaces for accurate calcs. Just fix those illum polygons and that error will go away. That said, long thin polygons are not a "good thing", and others on here could give you a better explanation of why, but it's not the end of world if you have some in there (I don't think).

I played with Sketchup a little bit, among other modelers, but I always crawl back to AutoCAD, for better or worse. I guess once you get comfy with something, it's hard to leave it. If Skecthup works for you, use it. But I find AutoCAD to be a very precise tool, and Gerog Mischler's radout utility is a great .DWG > Radiance exporter. It's a good combo!

Thanks for your response.

Ha, it's my pleasure to be able to *answer* a question here, opposed to always being the "asker". (Hopefully I did not steer you wrong, but if I did I will surely be corrected here...) Good luck in your pursuits with Radiance!

Rob Guglielmetti
[email protected]

···

On Wednesday, April 30, 2003, at 09:02 PM, John An wrote:

5

John, have you tried interlibrary loan? Since you're in Boston I'd expect your local library could easily find you a copy.

Randolph

···

On Wednesday, April 30, 2003, at 06:02 PM, John An wrote:

I had run across some threads discussing trying to find a new publisher, but thought that the situation had been resolved, since Elsevier Science happily took my order. When I called to inquire about my order, they confidenctly said that they would have the book in stock on May 1st. Is there any way that I could temporarily get my hands on the book in the meantime? Is there anyone in the Boston area that would be willing to lend me their book? After years of hearing how great Radiance was, I finally got the courage to jump in and to learn to use this program. The lack of availability of Rendering with Radiance has put a HUGE damper on my plans (though I am still determined to use it for my dissertation).

6

A bit more clarification.

If one has a model which is 100 feet in the x (width) direction and 50 in the y direction, and one's floor is at z=0, and one is using feet (which I prefer for AS units), then the options:
   -vl -vd 0 0 -1 -vu 0 1 0 -vp 50 25 5 -vo 1
will give a plan cut at the conventional four feet off the floor. Also, this:
   -vl -vd 0 0 -1 -vu 0 1 0 -vp 50 25 7 -vo -1
I believe will give the reflected ceiling plan so popular with lighting designers, cut at eight feet off the floor, which I expect will include most overhead luminaires.

Explanatory notes:

"-vl" sets a "parallel" (orthographic or axonometric) projection.

"-vd" (view Direction) sets view direction; the option takes x y and z values but these are used to define a vector and (so far as I can tell) only the direction of the vector is used. "-vd 0 0 -1" is a plan, since it is a vector pointing straight down.

"-vo" (fOre) sets a clipping plane near the viewpoint and "-va" (Aft) sets one behind the space. "-vo 1" sets a cutting plane 1 unit from the viewpoint. "-vo -1" will--I think--show a reflected image one unit behind the viewpoint

"-vp" (viewPoint) sets the viewpoint in model space (this will be the center of an orthographic projection.) "-vp 50 25 5" is a viewpoint five feet up in the center of my example model.

"-vu" (up) gives a vector in model space which indicates the "top" (maximum y coordinate) of the 2-D projection. "-vu 0 1 0" (I believe this is default) will make the y direction in model space the y direction in the projection.

I believe that's a summary of the basics of Radiance parallel projections. Hope it helps!

Randolph

···

On Wednesday, April 30, 2003, at 03:08 PM, Rob Guglielmetti wrote:

> Second, I'm sure this is covered in "Rendering with Radiance," but is
> there a way to get orthogonal and orthographic projected views in
> Radiance? Moreover, is there a way to make a surface invisible, yet
> be considered in the lighting calculation? I would like to look into
> an interior space in an axonometric view, but look through the
> ceiling and 1 or 2 walls to see the light distribution throughout a
> single space.

Check the manpage for rpict, it explains all the view parameters. The way to see "through" a surface is to use clipping planes in your views. This way, the calculation is correct, but at view time you slice through the building with the use of these clipping planes. These are also explained in the rtrace manual page.

7

Thanks Rob and Randolph for your help.

First off, de-triangulating the illums did the trick; no more aiming problems. I'm curious to know how inaccurate the calculations are when there are aiming errors.

Secondly, while I had searched the Harvard libraries for Rendering with Radiance, I hadn't searched the MIT libraries. Low and behold, MIT has a copy, which I quickly snatched up. And unfortunately, it seems as if Rendering with Radiance is no longer in print. I found a used copy of it on sale at Amazon for about $150 (I haven't ordered it). I really do want a copy of my own for future reference, so if anyone finds out more about the availability of Rendering with Radiance, please keep me (and us, I assume) informed.

Now, to my next question. I've flipped through Rendering with Randiance the last 30 minutes, and haven't found what I needed. I will be reading through it more carefully, but I haven't seen anything regarding setting up a reference grid. In Desktop Radiance, there is an option to create a reference grid.

So, how does one go about setting up a reference grid or reference point in the "regular" Radiance?

Thanks again.

John An

8

John An wrote:

> First off, de-triangulating the illums did the trick; no more aiming
> problems. I'm curious to know how inaccurate the calculations are
> when there are aiming errors.

AFAIK, error could potentially be quite large. Since you now have RwR in your posession, head over to page 511 for the juicy bits about this dilemma.

> Secondly, while I had searched the Harvard libraries for Rendering
> with Radiance, I hadn't searched the MIT libraries. Low and behold,
> MIT has a copy, which I quickly snatched up. And unfortunately, it
> seems as if Rendering with Radiance is no longer in print. I found a
> used copy of it on sale at Amazon for about $150 ...

That's only about 50% over list price, certainly the cheapest I've seen it in a while. :\

> Now, to my next question. I've flipped through Rendering with
> Randiance the last 30 minutes, and haven't found what I needed. I
> will be reading through it more carefully, but I haven't seen
> anything regarding setting up a reference grid. In Desktop Radiance,
> there is an option to create a reference grid.
>
> So, how does one go about setting up a reference grid or reference
> point in the "regular" Radiance?

This thread might be useful for ya. Be sure to read the entire thread because (surprise!) I made some errors along the way...

http://www.radiance-online.org/pipermail/radiance-general/2002-May/000258.html

Also, Georg Mischler's Rayfront will generate these grids for you automatically.

···

----

      Rob Guglielmetti

e. [email protected]
w. www.rumblestrip.org

9

John An wrote:

Thanks Rob and Randolph for your help.

First off, de-triangulating the illums did the trick; no more aiming problems.

Actually, this brings up a question I have for the list. What to do if you have a warped plane that is a skylight? I mean, I can draw it as a warped parallelogram, but Radout is gonna split it into two long thin triangles, and I'll get the aiming failure. Do you use LOTS of triangles to approximate the warp, or is there some other way?

···

----

      Rob Guglielmetti

e. [email protected]
w. www.rumblestrip.org

10

Rob Guglielmetti wrote:

John An wrote:
> Thanks Rob and Randolph for your help.
>
> First off, de-triangulating the illums did the trick; no more aiming
> problems.

Actually, this brings up a question I have for the list. What to do if
you have a warped plane that is a skylight? I mean, I can draw it as a
warped parallelogram, but Radout is gonna split it into two long thin
triangles, and I'll get the aiming failure. Do you use LOTS of
triangles to approximate the warp, or is there some other way?

Do your windowing contractor a favour, and avoid constructs like
that to begin with? :wink:

Note that the illum doesn't have to be the same object as the
window pane. In such a case, you probably want to add an extra
polygon below the window, eg. around the lower outline of the
frame of the skylight. This surface then gets treated by mkillum
instead of the actual window.

If the resulting shape is still very long and thin, then consider
dividing it into two or more sections that are closer to a
square. Otherwise, the source subdivision may result in a very
suboptimal set of virtual light sources (their location is
determined on a regular grid within the bounding rectangle of the
polygon).

-schorsch

···

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

11

Georg Mischler wrote:

Do your windowing contractor a favour, and avoid constructs like
that to begin with? :wink:

Heh heh, I didn't design the darned thing, I'm just tasked with studying it! =8-/

Note that the illum doesn't have to be the same object as the
window pane. In such a case, you probably want to add an extra
polygon below the window, eg. around the lower outline of the
frame of the skylight. This surface then gets treated by mkillum
instead of the actual window.

True, I didn't think of that. I will try that. Thanks Georg!

···

----

      Rob Guglielmetti

e. [email protected]
w. www.rumblestrip.org

12

Clearly, your name isn't Aalto. :slight_smile: Though I think he may have hired
boat carpenters for some of his reflectors--that's how they're put
together; ribs and slats. Rob, what does your skylight look like? Is
the glass actually curved?

Randolph

···

On Thu, May 01, 2003 at 03:51:21PM -0400, Georg Mischler wrote:

Rob Guglielmetti wrote:

> John An wrote:
> > Thanks Rob and Randolph for your help.
> >
> > First off, de-triangulating the illums did the trick; no more aiming
> > problems.
>
> Actually, this brings up a question I have for the list. What to do if
> you have a warped plane that is a skylight? I mean, I can draw it as a
> warped parallelogram, but Radout is gonna split it into two long thin
> triangles, and I'll get the aiming failure. Do you use LOTS of
> triangles to approximate the warp, or is there some other way?

Do your windowing contractor a favour, and avoid constructs like
that to begin with? :wink:

13

Clearly, your name isn't Aalto. :slight_smile: Though I think he may have hired
boat carpenters for some of his reflectors--that's how they're put
together; ribs and slats. Rob, what does your skylight look like? Is
the glass actually curved?

Well, the schematic drawings show it that way, but I'm sure it will be more faceted in reality. At first I didn't care and just modeled it as a warped plane. But now I wanna generate some prettier renderings and I can see that the mkillum process really helps in that regard. I guess I'll "straighten it out".

About half the planes in my model need to get fixed, but even with all the aiming failures, the renderings generated with illums on all the windows are much nicer. But I can see some artifacts that indicate to me that the accuracy is being compromised here.

Rob Guglielmetti
[email protected]