test file

Hello all:

I use radiance to simulate *the same model twice*. But the result (sensor
point illuminance) is different. Could anyone know the reason?

Thank you in advance,

Jia

Jia,

There will always be some difference between two identical simulations in
Radiance. The monte carlo sampling method used by radiance is like rolling
the dice several times. If your sample density is high enough your results
for identical simulations will exhibit minor differences. However with a
low sampling density a few lucky hits or misses can skew the simulation
results significantly.

The radiance simulation parameters allow us to adjust the sampling density
and other parameters to improve the correlation of the results. Trouble is
there is no single set of parameter settings that works for every model.

To allow the group to best help you refine your simulation parameters you
should send the following:
1 ­ The radiance commands you used to run your simulation.
2 ­ The results of your simulations.
3 ­ A description of your model
4 ­ A rendering if you have one (proper etiquette is to upload the image to
a website and include a link in your email to the list).

Best,
Andy

···

On 1/14/10 11:01 AM, "Hu,Jia" <[email protected]> wrote:

Hello all:

I use radiance to simulate the same model twice. But the result (sensor point
illuminance) is different. Could anyone know the reason?

Thank you in advance,

Jia

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general

____________________________________________________________
Electronic mail messages entering and leaving Arup business
systems are scanned for acceptability of content and viruses

Hello Andy:

Sorry for my late reply. I am new for Radiance and have to learn more to
reply you. Before discussing the previous question, could I ask how I
can open the text editor and save as file with extension like .rad after I
install Ubuntu 9.10 because I am unfamiliar with Ubuntu (Linux)?

For the question in my last email, I asked that because someone did some
simulations before but the results seems weird. In those simulations, two
identical small offices are developed with only one difference. In one
office, the angles of the blinds are 0 degree. For another, the windows
are separated into three sections (the height for each section is the
same). The blinds with 0 degree are also the same for each section. The
illuminance of the sensor points are calculated for the two
offices. Theoretically, the illuminance should be identical or quite
similar. But as I said in my last email, the illuminance shows a big
difference.

In addition to your explanation, is that right another possible reason is
the computer hardware difference. For example, one computer may calculate
5+3=8, while another calculates 5+3=7.8888. If possible, Is there any test
file to test the accuracy of computer before running Radiance?

Thanks.

Jia

···

On Thu, Jan 14, 2010 at 2:36 PM, Andrew McNeil <[email protected]>wrote:

Jia,

There will always be some difference between two identical simulations in
Radiance. The monte carlo sampling method used by radiance is like rolling
the dice several times. If your sample density is high enough your results
for identical simulations will exhibit minor differences. However with a
low sampling density a few lucky hits or misses can skew the simulation
results significantly.

The radiance simulation parameters allow us to adjust the sampling density
and other parameters to improve the correlation of the results. Trouble is
there is no single set of parameter settings that works for every model.

To allow the group to best help you refine your simulation parameters you
should send the following:
1 – The radiance commands you used to run your simulation.
2 – The results of your simulations.
3 – A description of your model
4 – A rendering if you have one (proper etiquette is to upload the image to
a website and include a link in your email to the list).

Best,
Andy

On 1/14/10 11:01 AM, "Hu,Jia" <[email protected]> wrote:

  Hello all:

I use radiance to simulate *the same model twice*. But the result (sensor
point illuminance) is different. Could anyone know the reason?

Thank you in advance,

Jia

------------------------------
_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general

____________________________________________________________
Electronic mail messages entering and leaving Arup business
systems are scanned for acceptability of content and viruses

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general

Jia.

See further answers within ...

Hello Andy:

Sorry for my late reply. I am new for Radiance and have to learn more to
reply you. Before discussing the previous question, could I ask how I
can open the text editor and save as file with extension like .rad after I
install Ubuntu 9.10 because I am unfamiliar with Ubuntu (Linux)?

I currently don't have an Ubuntu system installed but I think this still
applies:

Gedit is the default Ubuntu text editor, which you can launch by clicking
Applications->Accessories->Text Editor on the desktop menu system.

When you save files always make sure that you save in plain
text (*.txt).

For the question in my last email, I asked that because someone
did some simulations before but the results seems weird. In those
simulations, two identical small offices are developed with only one
difference. In one office, the angles of the blinds are 0 degree. For
another, the windows are separated into three sections (the height
for each section is the same). The blinds with 0 degree are also the
same for each section. The illuminance of the sensor points are
calculated for the two offices. Theoretically, the illuminance should
be identical or quite similar. But as I said in my last email, the
illuminance shows a big difference.

As Andy said, without details about the rendering parameters and
the scene/window geometry it's hard to tell why the values are so
different. Scenes with blinds are sensitive to the right rendering
parameters because you have to account for multiple reflections
to achieve satisfactory results. If the scenes were rendered with
settings that are not accurate enough you will get largely differing
results.

Another option is that the two scenes had different sky settings
which will also result in different (absolute) illuminance values.

In addition to your explanation, is that right another possible reason is
the computer hardware difference. For example, one computer may calculate
5+3=8, while another calculates 5+3=7.8888. If possible, Is there any test
file to test the accuracy of computer before running Radiance?

Radiance was written to be usable on multiple platforms. There
is a built in tolerance for rounding errors which takes care of the
above situations. However, you can get different results (to the
point where the calculations are untrustworthy) if you apply very
strong optimisation options when compiling the binaries. I assume
that you have used the binaries from the Debian/Ubuntu distribution
provided by Bernd Zeimetz so this should not be the cause of
your problem.

There is no standard test scenario for Radiance. I think you can
install example scenes in Ubuntu as a separate package. Some
of these are also discussed the book "Rendering with Radiance".

Once you have Radiance running (i.e. it produces images that look
about right) it's hard to verify the accuracy because each rendering
depends on the scene complexity and selected rendering parameters.
Basically you have to do a set of verification renderings with changing
settings until you have shown that your error is within a certain limit.

I also used other lighting simulation apps to check that my
results are at least in the same order of magnitude (especially
when artificial lighting is involved).

Mark Stock has written a benchmark scene which you can use
to test your installation. It will run for a few hours, though. It renders
a scene at high quality settings (for a picture output) and if your
result looks like the (tiny) image on his page Radiance works fine.

http://markjstock.org/pages/rad_bench.html

Regards,
Thomas

···

On Sat, Jan 16, 2010 at 4:37 PM, Hu,Jia <[email protected]> wrote:

Thank you for clarification. Is there any material for novice
besides reference listed in this website: http://radsite.lbl.gov/radiance/?

Jia

···

On Sun, Jan 17, 2010 at 5:58 AM, Thomas Bleicher <[email protected]>wrote:

Jia.

See further answers within ...

On Sat, Jan 16, 2010 at 4:37 PM, Hu,Jia <[email protected]> wrote:
> Hello Andy:
>
> Sorry for my late reply. I am new for Radiance and have to learn more to
> reply you. Before discussing the previous question, could I ask how I
> can open the text editor and save as file with extension like .rad after
I
> install Ubuntu 9.10 because I am unfamiliar with Ubuntu (Linux)?

I currently don't have an Ubuntu system installed but I think this still
applies:

> Gedit is the default Ubuntu text editor, which you can launch by clicking
> Applications->Accessories->Text Editor on the desktop menu system.

When you save files always make sure that you save in plain
text (*.txt).

> For the question in my last email, I asked that because someone
> did some simulations before but the results seems weird. In those
> simulations, two identical small offices are developed with only one
> difference. In one office, the angles of the blinds are 0 degree. For
> another, the windows are separated into three sections (the height
> for each section is the same). The blinds with 0 degree are also the
> same for each section. The illuminance of the sensor points are
> calculated for the two offices. Theoretically, the illuminance should
> be identical or quite similar. But as I said in my last email, the
> illuminance shows a big difference.

As Andy said, without details about the rendering parameters and
the scene/window geometry it's hard to tell why the values are so
different. Scenes with blinds are sensitive to the right rendering
parameters because you have to account for multiple reflections
to achieve satisfactory results. If the scenes were rendered with
settings that are not accurate enough you will get largely differing
results.

Another option is that the two scenes had different sky settings
which will also result in different (absolute) illuminance values.

> In addition to your explanation, is that right another possible reason is
> the computer hardware difference. For example, one computer may calculate
> 5+3=8, while another calculates 5+3=7.8888. If possible, Is there any
test
> file to test the accuracy of computer before running Radiance?

Radiance was written to be usable on multiple platforms. There
is a built in tolerance for rounding errors which takes care of the
above situations. However, you can get different results (to the
point where the calculations are untrustworthy) if you apply very
strong optimisation options when compiling the binaries. I assume
that you have used the binaries from the Debian/Ubuntu distribution
provided by Bernd Zeimetz so this should not be the cause of
your problem.

There is no standard test scenario for Radiance. I think you can
install example scenes in Ubuntu as a separate package. Some
of these are also discussed the book "Rendering with Radiance".

Once you have Radiance running (i.e. it produces images that look
about right) it's hard to verify the accuracy because each rendering
depends on the scene complexity and selected rendering parameters.
Basically you have to do a set of verification renderings with changing
settings until you have shown that your error is within a certain limit.

I also used other lighting simulation apps to check that my
results are at least in the same order of magnitude (especially
when artificial lighting is involved).

Mark Stock has written a benchmark scene which you can use
to test your installation. It will run for a few hours, though. It renders
a scene at high quality settings (for a picture output) and if your
result looks like the (tiny) image on his page Radiance works fine.

http://markjstock.org/pages/rad_bench.html

Regards,
Thomas

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general

Axel Jacobs has written some excellent tutorials (and is still adding more).

http://luminance.londonmet.ac.uk/learnix/docs.shtml

Regards,
Thomas

···

On Sun, Jan 17, 2010 at 4:30 PM, Hu,Jia <[email protected]> wrote:

Thank you for clarification. Is there any material for novice
besides reference listed in this website: http://radsite.lbl.gov/radiance/?

Thanks, that is very helpful

···

On Sun, Jan 17, 2010 at 6:07 PM, Thomas Bleicher <[email protected]>wrote:

On Sun, Jan 17, 2010 at 4:30 PM, Hu,Jia <[email protected]> wrote:
> Thank you for clarification. Is there any material for novice
> besides reference listed in this website:
http://radsite.lbl.gov/radiance/?

Axel Jacobs has written some excellent tutorials (and is still adding
more).

http://luminance.londonmet.ac.uk/learnix/docs.shtml

Regards,
Thomas

_______________________________________________
Radiance-general mailing list
[email protected]
http://www.radiance-online.org/mailman/listinfo/radiance-general