Modeling glare from light fixtures

radiance-general
1

Hi-

A question to the radiance community. I've been tasked with modeling a
T5 light fixture in its housing. This is new for me as a project in
Radiance, as I've largely used it to model natural daylight only, with
the use of imported .IES files when needed. I believe that I have to
model the light fixture and its associated housing because, for this
high-end conference room, the client for whom I am working is concerned
about glare emanating from the fixture during video conferencing. The
.IES files just don't have that type of information, as they are really
(at least in this case) files that really just describe the flux as a
mathematical model, and does not physically model the interactions of
the light bulb in its housing. As a result, I can't really show glare
emanating from the fixture at all! (these are all things I'm sure you
are already aware of)

I am wondering a couple of things, and am looking for some advice,
reactions, precautionary tales, etc. based on the information that I
currently have at hand. I'm willing to take harsh criticism so lay it
on me.

It seems my most accurate option is to model a T5 fixture as a glow
object, giving off the appropriate luminous flux inside of a reflective
housing. I would likely have to take some guesses as to the geometry of
the fixture just because, at least at this time, I don't have a physical
sample of the fixture, nor do I have any sort of exploded orthgraphic
projection showing the inner details of the fixture. I'd likely take
just a section of the fixture (found on the cut sheet) and extrude it
the appropriate length. My biggest concern is in regard to modeling
such a high intensity of light in such a small confined space with so
many bounces of light. I am not sure if there are "standards" in terms
of simulation settings, or other helpful pointers, that would normally
be applied to simulate the fixture in this way.

The other option, far simpler though possibly just as effective (and
this is a huge possibly in my mind), is to simply place a square
luminous glow object over the fixture where the light would normally
exit the fixture. I suppose I could build up my own mathematical
representation of how the glow object should distribute its light (in
effect build my own .IES file (of sorts). Though, I truly have never
done this before and I would assume there is a chance of error. Any
guidance for how to go about this would be appreciated, perhaps a
reading supplement to the Ward & Shakespeare book to ground my thinking,
at the very least.

As always, your help is appreciated.

Chien Si Harriman LEED AP

Senior Building Performance Engineer

(P) 415.655.4005

GUTTMANN & BLAEVOET Consulting Engineers

San Francisco | Sacramento | Santa Rosa <http://gb-eng.com/page/Contact>

Walnut Creek <http://gb-eng.com/page/Contact> | Website

<http://www.gb-eng.com/>

Certified Green Business <http://www.greenbiz.ca.gov/AboutUs.html>
committed to the 2030 Challenge <http://www.2030challenge.com/>

HVAC, Plumbing, Fire Protection, Electrical, Lighting,
Telecom/Technology <http://gb-eng.com/page/services>

New and Existing Building Commissioning, Energy & Building Performance
Modeling <http://gb-eng.com/page/services>

2

If the IES file is accurate and of sufficient resolution and you use the Radiance glare tools, the IES file will be sufficient to find sources of glare, even though the lamps and internal elements of the fixture will not be rendered correctly.

Radiance does not model rays emitted from a light source, and will not reliably compute the internal reflections of a fixture, even if the details of its materials and geometry are available.

For presentation, however, you may want to model the lamp and luminaire itself, as well as the IES file. This will not be physically accurate, but is helpful in visualizing the glare. To do this, you wrap the luminaire geometry in an illum (for fluorescents this is usually a long narrow box), to which you attach the IES data. This is called an "impostor surface," because it is an impostor for the actual luminaire. You then use glows within the illum to provide a visual representation of light within the luminaire. Light from a glow does not pass through an illum and hence does not affect the lighting of the scene geometry or the glare calculation.

I wrote a program to do some of this work as part of my thesis. It's not very finished and I haven't open-sourced it, but take a look at the thesis; the issues involved are covered there. See:
  http://dmg.caup.washington.edu/pdfs/Thesis.RandolphFritz.2010.pdf
Be very careful, going through this process, to check your modeled luminaire against the IES data. It is very easy to incorrectly place the IES data or fail to relate it correctly to the luminaire geometry, and this must be carefully checked. I describe my procedures for this on pp. 53-57 of my thesis.

I also have a Python script which reads IES files that might be helpful in working with this; I've been meaning to put it out for a while. If you think it might be helpful, drop me a note.

···

On 2011-05-11 22:27:33 -0700, Chien Si Harriman said:

A question to the radiance community. I've been tasked with modeling a T5 light fixture in its housing. This is new for me as a project in Radiance, as I've largely used it to model natural daylight only, with the use of imported .IES files when needed. I believe that I have to model the light fixture and its associated housing because, for this high-end conference room, the client for whom I am working is concerned about glare emanating from the fixture during video conferencing. The .IES files just don't have that type of information, as they are really (at least in this case) files that really just describe the flux as a mathematical model, and does not physically model the interactions of the light bulb in its housing. As a result, I can't really show glare emanating from the fixture at all! (these are all things I'm sure you are already aware of)

--
Randolph M. Fritz • [email protected]
Environmental Energy Technologies Division • Lawrence Berkeley Labs

3

I believe that I have to model the light fixture and its associated housing
because, for this high-end conference room, the client for whom I am working
is concerned about glare emanating from the fixture during video
conferencing.

Radiance is not the right tool to simulate the internal reflections of the
fitting. You could do it with a lot of rendering time and high accuracy
values. However due to inaccuracies in your material descriptions etc I
expect that the final result would not be more accurate than the IES data
you already have.

Since you're focused on video conferencing you also have to model the video
screens accurately. You can look at the various workshop presentations on
this topic.

Also keep in mind that you may be able to "prove" that there is no glare
caused by the light fittings but your client will still be disappointed by
the reflections of the fitting. Glare is a subjective experience and all you
can do with Radiance is to show that you meet some existing standard.

As a lighting designer I would recommend to reduce the potential for glare
as far a possible (different fitting, louvre accessories, different layout
and fitting type etc). If your clients don't want to spend money on good
lighting they have to live with the consequences.

The .IES files just don't have that type of information, as they are really
(at least in this case) files that really just describe the flux as a
mathematical model, and does not physically model the interactions of the
light bulb in its housing. As a result, I can't really show glare emanating
from the fixture at all! (these are all things I'm sure you are already
aware of)

IES data (and LDT) does indeed not reflect the physical characteristics of
certain light fittings very well. If the light source is small and directly
visible (think LEDs, halogen or discharge lamps) you will almost certainly
experience glare when you look directly into the lamp. The intensity of a T5
lamp can also be perceived as disturbing. However, if that is an issue here
you should look into diffusers etc. to cover the light source from direct
views.

Glare that results from the average intensity of the fitting is accurately
represented in the IES file and its geometry (disc, rectangle or box). Some
lighting design apps can generate a UGR diagram from the IES file. If you
use the IES file in Radiance I think you will also loose some angular
resolution because the conversion to a Radiance *.dat file has a fixed
number of vertical angles (although I'm not sure about this).

It seems my most accurate option is to model a T5 fixture as a glow object,
giving off the appropriate luminous flux inside of a reflective housing. I
would likely have to take some guesses as to the geometry of the fixture
just because, at least at this time, I don't have a physical sample of the
fixture, nor do I have any sort of exploded orthgraphic projection showing
the inner details of the fixture. I'd likely take just a section of the
fixture (found on the cut sheet) and extrude it the appropriate length.

If you are only interested in the source visibility you can use an extruded
cross section to simulate the fitting. If you want to use a mirror material
for the extrusion you have to change the "glow" to a "light" or you won't
get any secondary light sources. You can also reduce the ambient bounces a
lot because you only want to simulate the visibility of the light source.
This should speed up your calculation but it requires that you have a good
representation of the reflector geometry.

My biggest concern is in regard to modeling such a high intensity of light
in such a small confined space with so many bounces of light. I am not sure
if there are "standards" in terms of simulation settings, or other helpful
pointers, that would normally be applied to simulate the fixture in this
way.

There is no limit for the intensity you can define. Keep in mind that a
glare analysis is done without depreciation of the lamp and fitting (new
lamp and no dirt).

You are also less interested in ambient bounces than source reflections
("direct relays"). The bounces are only necessary if you want to show the
fitting within the conference room ambience. For your video conferencing
situation it should be enough to produce a reflection of the light source on
the video screen.

The other option, far simpler though possibly just as effective (and this
is a huge possibly in my mind), is to simply place a square luminous glow
object over the fixture where the light would normally exit the fixture.

This is exactly what ies2rad does.

I suppose I could build up my own mathematical representation of how the
glow object should distribute its light (in effect build my own .IES file
(of sorts). Though, I truly have never done this before and I would assume
there is a chance of error.

Exactly. You will not only introduce some errors due to inaccurate
information but you also have hardly a way to verify your model. If
something is wrong with your calculation results you will have a hard time
to prove it wasn't your hand made distribution. That's hardly worth it.

Regards,
Thomas

···

On Thu, May 12, 2011 at 12:27 AM, Chien Si Harriman <[email protected]>wrote:

4

Hi!

I suppose I could build up my own mathematical representation of how the glow object should distribute its light (in effect build my own .IES file (of sorts). Though, I truly have never done this before and I would assume there is a chance of error.

Exactly. You will not only introduce some errors due to inaccurate information but you also have hardly a way to verify your model. If something is wrong with your calculation results you will have a hard time to prove it wasn't your hand made distribution. That's hardly worth it.

Maybe my view is a bit too simplifying, but, imho: I think basically, for glare, there are two cases:

1) luminaire designed without thinking about glare. Lamp directly visible. IES data does not reflect well the spatially non-uniform distribution, which is averaged over aperture area. Will probably make your client unhappy anyways in your usecase.

2) luminaire optimized to reduce glare. No direct visibility of the lamp, thus little nonuniformity over the aperture. Should be matched not all that bad by IES data.

If you really model the reflector and source, you may actually try to validate it by sampling at the directions of you IES file. Unless the data is not wrong at all, you would get the same averaged values if your model is ok. By the way, is your IES data based on measurements, or is it also the result of a simulation?

Cheers, Lars.

5

If you model estimated geometry surrounded by the "impostor surfaces," would this impact your results depending on how you are estimating glare? If you use luminance for glare evaluation, do you get the luminance value of the faux luminaire geometry behind the illum "impostor surface" or do you get the luminance value of the IES illum?

If you get the value of the IES file then the detailed geometry is just for show in any renderings. If you get the value of the detailed geometry then you're back to the question of if it's accurate.

···

-----Original Message-----
From: Randolph M. Fritz [mailto:[email protected]]

For presentation, however, you may want to model the lamp and luminaire itself, as well as the IES file. This will not be physically accurate, but is helpful in visualizing the glare. To do this, you wrap the luminaire geometry in an illum (for fluorescents this is usually a long narrow box), to which you attach the IES data. This is called an "impostor surface," because it is an impostor for the actual luminaire.
You then use glows within the illum to provide a visual representation of light within the luminaire. Light from a glow does not pass through an illum and hence does not affect the lighting of the scene geometry or the glare calculation.
____________________________________________________________
Electronic mail messages entering and leaving Arup business
systems are scanned for acceptability of content and viruses

6

Right. Since what the rendering shows is the false luminaire geometry, that's what a direct view of the luminaire will show. So you would have to use a straight "light," not an "illum," to get a calculation of glare looking into the luminaire. This will not produce a realistic image, but I expect that, for a T5 fluorescent, it would produce a useful glare calcuation. As Lars points out, it probably won't be as useful for LEDs fixture that allows for a direct view of many small points of light.

I wonder if HDR images of an LED fixture would provide a useful basis for glare calculation. I suppose one could, if one had the resources, move the camera around the luminaire, the way a photometer does, and create multiple images which, if we had a way of getting those images into Radiance, might provide a fairly accurate basis for glare computation.

Randolph

···

On 2011-05-16 06:32:41 -0700, Christopher Rush said:

If you model estimated geometry surrounded by the "impostor surfaces," would this impact your results depending on how you are estimating glare? If you use luminance for glare evaluation, do you get the luminance value of the faux luminaire geometry behind the illum "impostor surface" or do you get the luminance value of the IES illum?

7

They do, in the crude sense of just looking at contrast ratios. Never tried with a fisheye lens and running evalglare, but I actually have a slide showing an LDR image alongside a HDR image that was tonemapped with pcond -h that suggests the discomfort of looking at an LED striplight, that I am presenting at Lightfair in (yikes) less than two hours. Gotta go!

- Rob

···

On May 16, 2011, at 9:24 AM, Randolph M. Fritz wrote:

I wonder if HDR images of an LED fixture would provide a useful basis for glare calculation. I suppose one could, if one had the resources, move the camera around the luminaire, the way a photometer does, and create multiple images which, if we had a way of getting those images into Radiance, might provide a fairly accurate basis for glare computation.

8

If you have a simple fitting where the LEDs have individual lenses and no
common reflector you can approximate the appearance with the data from the
IES file: Instead of one (box-shaped) luminaire you model each LED as a disc
(size of lens) and apply the IES file to each LED. Then you have to increase
the output of the LEDs to compensate for the loss in area of the luminous
opening (large rectangle vs small discs). This can be a fairly good
approximation for the high intensity of LEDs.

Regards,
Thomas

···

On Mon, May 16, 2011 at 10:24 AM, Randolph M. Fritz <[email protected]> wrote:

I wonder if HDR images of an LED fixture would provide a useful basis for
glare calculation.

9

Hi!

···

On May 16, 2011, at 18:24, "Randolph M. Fritz" <[email protected]> wrote:

I wonder if HDR images of an LED fixture would provide a useful basis for glare calculation. I suppose one could, if one had the resources, move the camera around the luminaire, the way a photometer does, and create multiple images which, if we had a way of getting those images into Radiance, might provide a fairly accurate basis for glare computation.

I think we hit near field photometry here. HDR images may be a useful input. Anyone who did this in Radiance so far?

Basically you need to end up with Radiance(position,direction) as a function. If you have enough HDR images, where you can relate pixel positions to location at luminous surfaces under known directions, this should be doable using Radiance tools.

Cheers, Lars.

10

Sigh. Everyone except me seems to be at Lightfair.

They do, in the crude sense of just looking at contrast ratios. Never tried with a fisheye lens and running evalglare, but I actually have a slide showing an LDR image alongside a HDR image that was tonemapped with pcond -h that suggests the discomfort of looking at an LED striplight, that I am presenting at Lightfair in (yikes) less than two hours. Gotta go!

Seattle put in an LED street light head just down the block. The point sources are so bright it's hard to look at. And yet the old luminaires (mercury vapor?), which probably have similar photometry, are perfectly acceptable.

PS: Check out the Lighting Sciences Group/Google home automation demo.

···

--
Randolph M. Fritz • [email protected]
Environmental Energy Technologies Division • Lawrence Berkeley Labs