I am writing mainly to understand Radiance more on different methods on
irradiance calculation on surface point while maintaining accuracy and
speed.
- I can use rtrace with irradiance interpolation turned on and rely on high
-ad( >1024) for direct light from sky.
running Radiance on Windows, it didn't seem to do run it in multi
processing(-n). I am not sure if this is supported on Windows?
Also breaking the points in chunks and running separate rtrace processes
didn't seem to speed it up much.
I don't go mad with -a numbers but the time tends to increase, sometimes up
to an hour for a room on complex models.
- I tried mkillum to increase the speed/accuracy but as the model is all
triangles I get the "aiming failure" error messages,
and it was mentioned in another post that triangles are poor choice for
light sources(due to the way Radiance samples
direct lighting.) I guess mkillum may not be advised in this case? also the
calc time increases linearly with with more illums in the model.
- rtrace with interpolation switched off, seems to be pretty quick. but as
it is blind monte carlo there is a concern on accuracy on
direct lighting I guess(unless using high -ad). I have seen the window
glazing defined as glow in three phase method(and later to use genbsdf for
T).
I am not quite familiar but if glow material is part of the direct
lighting, would this make it a better approach than pure MC?
Multiprocessing is not supported under Windows using rtrace, unfortunately.
Mkillum is not advisable if you are just computing illuminance points.
Turning interpolation off with rtrace (-aa 0) is a reasonable thing to do for point calculations, especially if you have only a handful of values you are computing. The payoff for the ambient cache is when you have many bounces and/or many calculation points.
Best,
-Greg
···
From: Ali Fatoorechi <[email protected]>
Date: April 11, 2017 12:24:01 PM PDT
Hi Guys,
I am writing mainly to understand Radiance more on different methods on irradiance calculation on surface point while maintaining accuracy and speed.
- I can use rtrace with irradiance interpolation turned on and rely on high -ad( >1024) for direct light from sky.
running Radiance on Windows, it didn't seem to do run it in multi processing(-n). I am not sure if this is supported on Windows?
Also breaking the points in chunks and running separate rtrace processes didn't seem to speed it up much.
I don't go mad with -a numbers but the time tends to increase, sometimes up to an hour for a room on complex models.
- I tried mkillum to increase the speed/accuracy but as the model is all triangles I get the "aiming failure" error messages,
and it was mentioned in another post that triangles are poor choice for light sources(due to the way Radiance samples
direct lighting.) I guess mkillum may not be advised in this case? also the calc time increases linearly with with more illums in the model.
- rtrace with interpolation switched off, seems to be pretty quick. but as it is blind monte carlo there is a concern on accuracy on
direct lighting I guess(unless using high -ad). I have seen the window glazing defined as glow in three phase method(and later to use genbsdf for T).
I am not quite familiar but if glow material is part of the direct lighting, would this make it a better approach than pure MC?
Just a quick note about -aa 0. I've found that illuminance result without
the ambient cache is lower than the result with an ambient cache and the
same simulation settings. I recommend decreasing -lw substantially when
-aa 0 is used. Convergence testing with a few points furthest from the
window is always a good idea, particularly when using -aa 0.
···
On Tue, Apr 11, 2017 at 1:04 PM, Greg Ward <[email protected]> wrote:
Hi Ali,
Multiprocessing is not supported under Windows using rtrace, unfortunately.
Mkillum is not advisable if you are just computing illuminance points.
Turning interpolation off with rtrace (-aa 0) is a reasonable thing to do
for point calculations, especially if you have only a handful of values you
are computing. The payoff for the ambient cache is when you have many
bounces and/or many calculation points.
I am writing mainly to understand Radiance more on different methods on
irradiance calculation on surface point while maintaining accuracy and
speed.
- I can use rtrace with irradiance interpolation turned on and rely on
high -ad( >1024) for direct light from sky.
running Radiance on Windows, it didn't seem to do run it in multi
processing(-n). I am not sure if this is supported on Windows?
Also breaking the points in chunks and running separate rtrace processes
didn't seem to speed it up much.
I don't go mad with -a numbers but the time tends to increase, sometimes
up to an hour for a room on complex models.
- I tried mkillum to increase the speed/accuracy but as the model is all
triangles I get the "aiming failure" error messages,
and it was mentioned in another post that triangles are poor choice for
light sources(due to the way Radiance samples
direct lighting.) I guess mkillum may not be advised in this case? also
the calc time increases linearly with with more illums in the model.
- rtrace with interpolation switched off, seems to be pretty quick. but as
it is blind monte carlo there is a concern on accuracy on
direct lighting I guess(unless using high -ad). I have seen the window
glazing defined as glow in three phase method(and later to use genbsdf for
T).
I am not quite familiar but if glow material is part of the direct
lighting, would this make it a better approach than pure MC?
On Wed, Apr 12, 2017 at 12:31 AM, Andy McNeil <[email protected]> wrote:
Just a quick note about -aa 0. I've found that illuminance result without
the ambient cache is lower than the result with an ambient cache and the
same simulation settings. I recommend decreasing -lw substantially when
-aa 0 is used. Convergence testing with a few points furthest from the
window is always a good idea, particularly when using -aa 0.
On Tue, Apr 11, 2017 at 1:04 PM, Greg Ward <[email protected]> wrote:
Hi Ali,
Multiprocessing is not supported under Windows using rtrace,
unfortunately.
Mkillum is not advisable if you are just computing illuminance points.
Turning interpolation off with rtrace (-aa 0) is a reasonable thing to do
for point calculations, especially if you have only a handful of values you
are computing. The payoff for the ambient cache is when you have many
bounces and/or many calculation points.
I am writing mainly to understand Radiance more on different methods on
irradiance calculation on surface point while maintaining accuracy and
speed.
- I can use rtrace with irradiance interpolation turned on and rely on
high -ad( >1024) for direct light from sky.
running Radiance on Windows, it didn't seem to do run it in multi
processing(-n). I am not sure if this is supported on Windows?
Also breaking the points in chunks and running separate rtrace processes
didn't seem to speed it up much.
I don't go mad with -a numbers but the time tends to increase, sometimes
up to an hour for a room on complex models.
- I tried mkillum to increase the speed/accuracy but as the model is all
triangles I get the "aiming failure" error messages,
and it was mentioned in another post that triangles are poor choice for
light sources(due to the way Radiance samples
direct lighting.) I guess mkillum may not be advised in this case? also
the calc time increases linearly with with more illums in the model.
- rtrace with interpolation switched off, seems to be pretty quick. but
as it is blind monte carlo there is a concern on accuracy on
direct lighting I guess(unless using high -ad). I have seen the window
glazing defined as glow in three phase method(and later to use genbsdf for
T).
I am not quite familiar but if glow material is part of the direct
lighting, would this make it a better approach than pure MC?
And a quick trick that used to work is to run several iterations and average them, if you are only testing a single point.
It is more precise than a single shot.
G
62 High Street
Steyning
West Sussex
BN44 3RD
T +44 (0)1903 879323 <>
M +44(0)7830751409 [email protected] <mailto:[email protected]> www.surveymbs.com <http://www.surveymbs.com/>
On Wed, Apr 12, 2017 at 12:31 AM, Andy McNeil <[email protected] <mailto:[email protected]>> wrote:
Just a quick note about -aa 0. I've found that illuminance result without the ambient cache is lower than the result with an ambient cache and the same simulation settings. I recommend decreasing -lw substantially when -aa 0 is used. Convergence testing with a few points furthest from the window is always a good idea, particularly when using -aa 0.
Multiprocessing is not supported under Windows using rtrace, unfortunately.
Mkillum is not advisable if you are just computing illuminance points.
Turning interpolation off with rtrace (-aa 0) is a reasonable thing to do for point calculations, especially if you have only a handful of values you are computing. The payoff for the ambient cache is when you have many bounces and/or many calculation points.
I am writing mainly to understand Radiance more on different methods on irradiance calculation on surface point while maintaining accuracy and speed.
- I can use rtrace with irradiance interpolation turned on and rely on high -ad( >1024) for direct light from sky.
running Radiance on Windows, it didn't seem to do run it in multi processing(-n). I am not sure if this is supported on Windows?
Also breaking the points in chunks and running separate rtrace processes didn't seem to speed it up much.
I don't go mad with -a numbers but the time tends to increase, sometimes up to an hour for a room on complex models.
- I tried mkillum to increase the speed/accuracy but as the model is all triangles I get the "aiming failure" error messages,
and it was mentioned in another post that triangles are poor choice for light sources(due to the way Radiance samples
direct lighting.) I guess mkillum may not be advised in this case? also the calc time increases linearly with with more illums in the model.
- rtrace with interpolation switched off, seems to be pretty quick. but as it is blind monte carlo there is a concern on accuracy on
direct lighting I guess(unless using high -ad). I have seen the window glazing defined as glow in three phase method(and later to use genbsdf for T).
I am not quite familiar but if glow material is part of the direct lighting, would this make it a better approach than pure MC?