Ambient parameters optimization/tweaking

Hello Everyone!

For my master thesis project, I have conducted an experiment to measure
daylighting/lighting illuminance on the last Equinox days in Vienna. The
room is a small hall (7mx9m), has white walls/ceilings, and only North
facing windows. I have measured reflectivity of all the surface to come
closer to the reality.

Now I have to calibrate my simulation model to match its illuminance output
with the measured illuminance, for both daylight and artificial light
condition.

For daylight, gendaylit sky model is used (listed below), and all
artificials lights are off. For this case, I have tried a number of ambient
parameter combinations (listed below), but since all windows are facing
North, the simulation model only results in 0 illuminance. When I rotated
the North offset, to face windows towards South then I get some
illuminance. From this it seems that windows North orientation is the
reason behind 0 illuminance. In measurements, these values are approx. 1500
lux.

My question is: What ambient parameters should be tweaked so that the model
behaves reasonably (especially for daylight)?

I am following Dr.John Mardaljevic's PhD for ambient parameters
combination. But I don't have much time to do an extensive parametric
analysis of these parameters.

Global, and Diffuse irradiance (for gendaylit) are taken from weather
station in place.

I have tried only rtrace. May be I shall use rsensor now ??

Best regards,
Vaib

*Ambient parameter combinations:*
1. rtrace -w -as 512 -ar 512 -aa 0.15 -ab 2 -I -h foo.oct | rcalc -e
'\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

2. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -as 32 -I -h foo.oct | rcalc
-e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

3. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -ar 2 -aa 0.10 -as 32 -I -h
foo.oct | rcalc -e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

4. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -ar 64 -aa 0.15 -as 512 -I -h
foo.oct | rcalc -e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

*Gendaylit model for (12:00 time):*
# xform -rz 30
# gendaylit 03 22 12:00 -a 48.198724 -o -16.369458 -m -15.00 -W 967.3 187.2
# Ground ambient level: 33.9

void light solar
0
0
3 7791000 7791000 7791000

solar source sun
0
0
4 0.378842567554 -0.642038575073 0.666532
0.533

void brightfunc skyfunc
6 skybright perezlum.cal -i 1 -rz 30
0
10 38.97 29.25 -0.878528
-0.366494 12.997644 -3.44062 0.249106
0.007068

Hi Vaib,

I highly recommend the well-known "Rendering with Radiance" which describes the way the calculations are implemented as well as crucial parameters. For the ambient calculation, all parameters beginning with the letter "a" are of importance. Runnning any Radiance command with the option "-defaults" will show the default values applied unless you override them with setting command-line parameters.

For a Nort-facing room on the Northern hemisphere (Vienna), direct sun will never be observed in the inside. The (diffuse) starts to contribute only if you set the number of bounces for the ambient calculation to at least 1 (which will consider the contribution from the sky to interior surfaces with sky visibility). To enable interreflection of the room-surfaces, the -ab setting must be higher. Try -ab 1 to -ab 6 and see at which setting your illuminance readings stabilize.

Unless you follow a very specific rendering strategy, in most cases "rad" will find rather well-tuned parameters according to the scene and the settings you configure in a "rif"-file. I would start from here, with reasonable values for indirect, detail and quality. If you want to try out the effect of a specific command line parameter, you can specify those in the "rif"-file as well using the "render" keyword. Please see the manpage and above mentioned book for details.

The setting for "-ar" seam to be low in your examples. "-ad" is high at 2048 - by pure guesswork I would have started with -ad 256 -as 128 -aa .1 -ar 64 for a room with simple windows, but this is not a good recommendation without knowing the scene. The settings for "-aa", "-ad", "-as" are interrelated and depend on the detail in the illuminance distribution that would be expected to affect your results. Again, see the book for details.

Cheers, Lars.

Hi Vaib,

I believe you have a more fundamental error in your model. gendaylit and
gensky do not create the sky source, you have to manually add a sky
material and source to your model. If you append the text at the end of
this email to your sky description I believe you will have non zero
illuminance for the north facing window. Though you will still need to

Best,
Andy

skyfunc glow skymat
0
0
4 1 1 1 0

skymat source skyball
0
0
4 0 0 1 360

···

On Wed, Apr 2, 2014 at 6:07 AM, Vaib <vaibhavjain.co@gmail.com> wrote:

Hello Everyone!

For my master thesis project, I have conducted an experiment to measure
daylighting/lighting illuminance on the last Equinox days in Vienna. The
room is a small hall (7mx9m), has white walls/ceilings, and only North
facing windows. I have measured reflectivity of all the surface to come
closer to the reality.

Now I have to calibrate my simulation model to match its illuminance
output with the measured illuminance, for both daylight and artificial
light condition.

For daylight, gendaylit sky model is used (listed below), and all
artificials lights are off. For this case, I have tried a number of ambient
parameter combinations (listed below), but since all windows are facing
North, the simulation model only results in 0 illuminance. When I rotated
the North offset, to face windows towards South then I get some
illuminance. From this it seems that windows North orientation is the
reason behind 0 illuminance. In measurements, these values are approx. 1500
lux.

My question is: What ambient parameters should be tweaked so that the
model behaves reasonably (especially for daylight)?

I am following Dr.John Mardaljevic's PhD for ambient parameters
combination. But I don't have much time to do an extensive parametric
analysis of these parameters.

Global, and Diffuse irradiance (for gendaylit) are taken from weather
station in place.

I have tried only rtrace. May be I shall use rsensor now ??

Best regards,
Vaib

*Ambient parameter combinations:*
1. rtrace -w -as 512 -ar 512 -aa 0.15 -ab 2 -I -h foo.oct | rcalc -e
'\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

2. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -as 32 -I -h foo.oct | rcalc
-e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

3. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -ar 2 -aa 0.10 -as 32 -I -h
foo.oct | rcalc -e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

4. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -ar 64 -aa 0.15 -as 512 -I
-h foo.oct | rcalc -e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

*Gendaylit model for (12:00 time):*
# xform -rz 30
# gendaylit 03 22 12:00 -a 48.198724 -o -16.369458 -m -15.00 -W 967.3 187.2
# Ground ambient level: 33.9

void light solar
0
0
3 7791000 7791000 7791000

solar source sun
0
0
4 0.378842567554 -0.642038575073 0.666532
0.533

void brightfunc skyfunc
6 skybright perezlum.cal -i 1 -rz 30
0
10 38.97 29.25 -0.878528
-0.366494 12.997644 -3.44062 0.249106
0.007068

_______________________________________________

A short add to Andy's post, you will notice Andy is defining the sky and ground hemispheres as a single 360-degree source, called "skyball". This works because skyfunc will automatically use the ground ambient value (adjusted with the –g option to gensky) when the ray direction vector has a negative zed value.

This is, in a word, cool.

Some more words: this means Radiance will automatically derate the "sky" luminance when a traced ray hits below the horizon, knowing that it's actually hitting the "ground", and not the sky. The default for –g is 0.20, for a 20% reflectance ground "plane". Reasonable enough, but be advised you should really also have actual geometry to define the local exterior environment — this includes the ground as well as neighboring buildings, so that your local ground is shaded by your building and its neighbors.

- Rob

···

On 4/2/14, 11:24 AM, "Andrew McNeil" <amcneil@lbl.gov<mailto:amcneil@lbl.gov>> wrote:

Hi Vaib,

I believe you have a more fundamental error in your model. gendaylit and gensky do not create the sky source, you have to manually add a sky material and source to your model. If you append the text at the end of this email to your sky description I believe you will have non zero illuminance for the north facing window. Though you will still need to adjust ambient parameters to suit your model as Lars describes.

Best,
Andy

skyfunc glow skymat
0
0
4 1 1 1 0

skymat source skyball
0
0
4 0 0 1 360

On Wed, Apr 2, 2014 at 6:07 AM, Vaib <vaibhavjain.co@gmail.com<mailto:vaibhavjain.co@gmail.com>> wrote:
Hello Everyone!

For my master thesis project, I have conducted an experiment to measure daylighting/lighting illuminance on the last Equinox days in Vienna. The room is a small hall (7mx9m), has white walls/ceilings, and only North facing windows. I have measured reflectivity of all the surface to come closer to the reality.

Now I have to calibrate my simulation model to match its illuminance output with the measured illuminance, for both daylight and artificial light condition.

For daylight, gendaylit sky model is used (listed below), and all artificials lights are off. For this case, I have tried a number of ambient parameter combinations (listed below), but since all windows are facing North, the simulation model only results in 0 illuminance. When I rotated the North offset, to face windows towards South then I get some illuminance. From this it seems that windows North orientation is the reason behind 0 illuminance. In measurements, these values are approx. 1500 lux.

My question is: What ambient parameters should be tweaked so that the model behaves reasonably (especially for daylight)?

I am following Dr.John Mardaljevic's PhD for ambient parameters combination. But I don't have much time to do an extensive parametric analysis of these parameters.

Global, and Diffuse irradiance (for gendaylit) are taken from weather station in place.

I have tried only rtrace. May be I shall use rsensor now ??

Best regards,
Vaib

Ambient parameter combinations:
1. rtrace -w -as 512 -ar 512 -aa 0.15 -ab 2 -I -h foo.oct | rcalc -e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

2. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -as 32 -I -h foo.oct | rcalc -e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

3. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -ar 2 -aa 0.10 -as 32 -I -h foo.oct | rcalc -e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

4. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -ar 64 -aa 0.15 -as 512 -I -h foo.oct | rcalc -e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

Gendaylit model for (12:00 time):
# xform -rz 30
# gendaylit 03 22 12:00 -a 48.198724 -o -16.369458 -m -15.00 -W 967.3 187.2
# Ground ambient level: 33.9

void light solar
0
0
3 7791000 7791000 7791000

solar source sun
0
0
4 0.378842567554 -0.642038575073 0.666532 0.533

void brightfunc skyfunc
6 skybright perezlum.cal -i 1 -rz 30
0
10 38.97 29.25 -0.878528 -0.366494 12.997644 -3.44062 0.249106 0.007068

_______________________________________________

Thank you Lars, Andrew, Rob..

After including the sky glow and source, the issue of 'no illuminance' is
solved.. That make sense..
I am now trying different ambient options based on Lars suggestion, and the
results are getting closer.. Thank you all again..

Best regards,
Vaib

···

On 2 April 2014 19:47, Guglielmetti, Robert <Robert.Guglielmetti@nrel.gov>wrote:

A short add to Andy's post, you will notice Andy is defining the sky and
ground hemispheres as a single 360-degree source, called "skyball". This
works because skyfunc will automatically use the ground ambient value
(adjusted with the -g option to gensky) when the ray direction vector has a
negative zed value.

This is, in a word, cool.

Some more words: this means Radiance will automatically derate the "sky"
luminance when a traced ray hits below the horizon, knowing that it's
actually hitting the "ground", and not the sky. The default for -g is 0.20,
for a 20% reflectance ground "plane". Reasonable enough, but be advised you
should really also have actual geometry to define the local exterior
environment -- this includes the ground as well as neighboring buildings, so

- Rob

On 4/2/14, 11:24 AM, "Andrew McNeil" <amcneil@lbl.gov<mailto: > amcneil@lbl.gov>> wrote:

Hi Vaib,

I believe you have a more fundamental error in your model. gendaylit and
gensky do not create the sky source, you have to manually add a sky
material and source to your model. If you append the text at the end of
this email to your sky description I believe you will have non zero
illuminance for the north facing window. Though you will still need to

Best,
Andy

skyfunc glow skymat
0
0
4 1 1 1 0

skymat source skyball
0
0
4 0 0 1 360

On Wed, Apr 2, 2014 at 6:07 AM, Vaib <vaibhavjain.co@gmail.com<mailto: > vaibhavjain.co@gmail.com>> wrote:
Hello Everyone!

For my master thesis project, I have conducted an experiment to measure
daylighting/lighting illuminance on the last Equinox days in Vienna. The
room is a small hall (7mx9m), has white walls/ceilings, and only North
facing windows. I have measured reflectivity of all the surface to come
closer to the reality.

Now I have to calibrate my simulation model to match its illuminance
output with the measured illuminance, for both daylight and artificial
light condition.

For daylight, gendaylit sky model is used (listed below), and all
artificials lights are off. For this case, I have tried a number of ambient
parameter combinations (listed below), but since all windows are facing
North, the simulation model only results in 0 illuminance. When I rotated
the North offset, to face windows towards South then I get some
illuminance. From this it seems that windows North orientation is the
reason behind 0 illuminance. In measurements, these values are approx. 1500
lux.

My question is: What ambient parameters should be tweaked so that the
model behaves reasonably (especially for daylight)?

I am following Dr.John Mardaljevic's PhD for ambient parameters
combination. But I don't have much time to do an extensive parametric
analysis of these parameters.

Global, and Diffuse irradiance (for gendaylit) are taken from weather
station in place.

I have tried only rtrace. May be I shall use rsensor now ??

Best regards,
Vaib

Ambient parameter combinations:
1. rtrace -w -as 512 -ar 512 -aa 0.15 -ab 2 -I -h foo.oct | rcalc -e
'\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

2. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -as 32 -I -h foo.oct | rcalc
-e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

3. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -ar 2 -aa 0.10 -as 32 -I -h
foo.oct | rcalc -e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

4. cat allSens.txt | rtrace -w -ab 7 -ad 2048 -ar 64 -aa 0.15 -as 512 -I
-h foo.oct | rcalc -e '\$1=(\$1*.265+\$2*.67+\$3*.065)*179'

Gendaylit model for (12:00 time):
# xform -rz 30
# gendaylit 03 22 12:00 -a 48.198724 -o -16.369458 -m -15.00 -W 967.3 187.2
# Ground ambient level: 33.9

void light solar
0
0
3 7791000 7791000 7791000

solar source sun
0
0
4 0.378842567554 -0.642038575073 0.666532
0.533

void brightfunc skyfunc
6 skybright perezlum.cal -i 1 -rz 30
0
10 38.97 29.25 -0.878528
-0.366494 12.997644 -3.44062 0.249106
0.007068

_______________________________________________