calculation weirdness

Hi all, I'm running a radiance calculation, and as always, I compare my radiance calcs with AGI to make sure I didn't mess up. I know enough to be dangerous! :slight_smile:

My comparison consists of placing a calculation point 10' below an iesfile based light fixture.

AGI calc = 64810 lux

Radiance = 40,822 lux.

ies2rad -di -t white iesfile.rad
rtrace -w -h -I -ab 0 -aa .1 -ad 2048 -as 512 -ar 512 -ds .1 -aw 0 test.oct < input_pt.txt | rcalc -e '$1=47.4*$1+120*$2+11.6*$3'

OK, red flag, I've messed up somehow. The numbers are off by 30%.

After almost giving up, I ran the same command but deleted the "-ds .1" and got 64844 lux!

What's going on here? I would think source substructuring "-ds .1" would be better than the default (which I think is .2) I appreciate your thoughts.

Mark

Hi Mark,

Thanks for sharing your files with me. It seems that your light fixture has a very tight output distribution, which falls off dramatically to the left and right. The more you subdivide the light source, the more accurate your results. Your second Radiance result is the more accurate. Apparently, AGI32 only sends a single sample to the center of the source, which is what you also get from Radiance with the default -ds 0.2.

I discovered this by looking up from your sample point 120 inches below the light source, which is itself 20 inches in diameter. A perspective view shows that even at this distance, the fall-off is quite noticeable.

Bottom line, the rtrace -ds 0.1 result is more accurate. Whew!

-Greg

路路路

From: "Mark de la Fuente" <[email protected]>
Date: July 31, 2006 5:56:01 PM EDT

Hi all, I'm running a radiance calculation, and as always, I compare my radiance calcs with AGI to make sure I didn't mess up. I know enough to be dangerous! :slight_smile:

My comparison consists of placing a calculation point 10' below an iesfile based light fixture.

AGI calc = 64810 lux

Radiance = 40,822 lux.

ies2rad -di -t white iesfile.rad
rtrace -w -h -I -ab 0 -aa .1 -ad 2048 -as 512 -ar 512 -ds .1 -aw 0 test.oct < input_pt.txt | rcalc -e '$1=47.4*$1+120*$2+11.6*$3'

OK, red flag, I've messed up somehow. The numbers are off by 30%.

After almost giving up, I ran the same command but deleted the "-ds .1" and got 64844 lux!

What's going on here? I would think source substructuring "-ds .1" would be better than the default (which I think is .2) I appreciate your thoughts.

Mark

Greg,

I really appreciate you looking into this. I was going to drop the -ds
parameter all together, confident with my AGI/Radiance check!

The light fixture in question is a sportslighting fixture for a hockey
rink. There will be 200+ fixtures all aimed at slightly different
spots. The fixtures as you mentioned have a very strong but also very
narrow distribution in one axis. It sounds like I need to run a -ds
convergence checker at 100' to see what the largest value possible
should be to get accurate results. (I typically do this for the ab
parameter only and just settle on some of the others based on rad's
output)

Does the size of the scene affect the -ds parameter? My final scene
will be a large stadium.

Given this distribution, how will having an illum at each fixture
help/hurt renderings or point-by-point calculations?

Thank you for your guidance.

Mark

Hi Mark,

You can always experiment with smaller and smaller values of the -ds parameter, but I found that the convergence doesn't change much after -ds 0.1. The size of the scene doesn't affect what -ds value will work.

Using an illum would not help in this case, as far as I can figure.

-Greg

路路路

From: "Mark de la Fuente" <[email protected]>
Date: August 2, 2006 7:26:20 PM EDT

Greg,

I really appreciate you looking into this. I was going to drop the -ds parameter all together, confident with my AGI/Radiance check!

The light fixture in question is a sportslighting fixture for a hockey rink. There will be 200+ fixtures all aimed at slightly different spots. The fixtures as you mentioned have a very strong but also very narrow distribution in one axis. It sounds like I need to run a -ds convergence checker at 100' to see what the largest value possible should be to get accurate results. (I typically do this for the ab parameter only and just settle on some of the others based on rad's output)

Does the size of the scene affect the -ds parameter? My final scene will be a large stadium.

Given this distribution, how will having an illum at each fixture help/hurt renderings or point-by-point calculations?

Thank you for your guidance.

Mark

Hi Mark.

I don't know how your going to evaluate your model but if you're
calculating
illumination over an area or even falsecolor image I wouldn't worry much
about
the accurate contribution of one individual luminiare. You'll have
several luminiares
contributing at any given spot which should even out the computational
differences
of each luminaire.

Based on your description the -ds option will probably have no effect
anyway
because all your light sources are small in relation to their distance.
If you want
to compare results between different applications you'd have to setup
the same
model in each and run a standard (for each application) calculation.
This will
account for the different types of calculation methods (tracing vs.
radiosity)
and details (as the -ds effects).

Please note that I have seen a noticeable slowdown in my recent
calculations
of scenes with large numbers of light sources. If anyone has an idea how
to
optimise that it be great.

Thomas

路路路

_____

From: [email protected]
[mailto:[email protected]] On Behalf Of Mark
de la Fuente
Sent: 03 August 2006 00:26
To: Radiance
Subject: Re: [Radiance-general] calculation weirdness

Greg,

I really appreciate you looking into this. I was going to drop the -ds
parameter all together, confident with my AGI/Radiance check!

The light fixture in question is a sportslighting fixture for a hockey
rink. There will be 200+ fixtures all aimed at slightly different
spots. The fixtures as you mentioned have a very strong but also very
narrow distribution in one axis. It sounds like I need to run a -ds
convergence checker at 100' to see what the largest value possible
should be to get accurate results. (I typically do this for the ab
parameter only and just settle on some of the others based on rad's
output)

Does the size of the scene affect the -ds parameter? My final scene
will be a large stadium.

Given this distribution, how will having an illum at each fixture
help/hurt renderings or point-by-point calculations?

Thank you for your guidance.

Mark

***********************************************************************************
This e-mail, (and any attachments) is confidential and may be privileged. It may be read, copied and used by the intended addressee only. If you have received this in error please contact BDP immediately.

If you have any queries, please contact the sender.
***********************************************************************************
Building Design Partnership
Registered in England No 2207415:
Registered Office: Building Design Partnership Ltd, Sunlight House, PO Box 85, Quay Street, Manchester, M60 3JA, http://www.bdp.co.uk
***********************************************************************************