Blotching caused by translucent material

radiance-general
1

Hi,

Can someone help me to set the Radiance parameters to eliminate blotching in my images? I am running renderings, DGP and glare simulations for a simple geometry. I have had no issues with the images when I use clear glazing. One of the tests that I am conducting, however, uses translucent glazing with the following parameters:

void trans testTrans007 0.478 0.478 0.478 0.000 0.010 0.178 0.635
I've run radiance with two different parameter sets, the first with
_av_ is set to 0xScale is set to 2_ab_ is set to 6_dc_ is set to 0.5_aa_ is set to 0.2_ad_ is set to 2048_st_ is set to 0.5yScale is set to 2_ps_ is set to 4_ar_ is set to 64_as_ is set to 2048_ds_ is set to 0.25_pt_ is set to 0.1_dr_ is set to 1_pj_ is set to 0.9_dp_ is set to 256_dt_ is set to 0.25_lr_ is set to 6_dj_ is set to 0.5_lw_ is set to 0.01
resulting in the image named Blotching001
The second case has the following parameters:
_av_ is set to 0xScale is set to 6_ab_ is set to 6_dc_ is set to 0.75_aa_ is set to 0.1_ad_ is set to 4096_st_ is set to 0.15yScale is set to 6_ps_ is set to 2_ar_ is set to 128_as_ is set to 4096_ds_ is set to 0.05_pt_ is set to 0.05_dr_ is set to 3_pj_ is set to 0.9_dp_ is set to 512_dt_ is set to 0.15_lr_ is set to 8_dj_ is set to 0.7_lw_ is set to 0.005
resulting in the image named Blotching002. I have also attached an image with no blotching. This case uses glazing with 65% light transmittance.
Although the second case is much better, it is still very bad for hours when the sun is lower in the sky. The two above images are rendered for a clear sky at 18:00 in Germany.

Can someone help?

Kind regards,
Orn Erlendsson

2

Hi Orn,

It looks like the specular transmission component is sampled
stochastically, this happens because you have non-zero roughness in your
trans definition. To reduce the blotches in your sun patches you'll need to
crank ad, reduce aa and set st to zero. I'd keep lw at roughly 1/ad because
you want to heavily sample the window but with so many rays you don't need
them to split much further. You could also consider turning off the ambient
cache by setting -aa 0. This changes the effect from blotches to noise,
which sometimes has a better appearance.

Or you could change roughness from 0.010 to 0. I'm not sure how important
the roughness is to you simulation, but I don't expect it'll have much
affect on your illuminance result. But if you're interested in simulating
appearance accurately then the roughness is important.

I'd also suggest double checking your trans definition. It has a very low
total transmission (~9%) and is mostly specular.

Best,
Andy

···

On Fri, Jun 12, 2015 at 5:47 AM, Örn Erlendsson <[email protected]> wrote:

Hi,

Can someone help me to set the Radiance parameters to eliminate blotching
in my images? I am running renderings, DGP and glare simulations for a
simple geometry. I have had no issues with the images when I use clear
glazing. One of the tests that I am conducting, however, uses translucent
glazing with the following parameters:

void trans testTrans
0
0
7 0.478 0.478 0.478 0.000 0.010 0.178 0.635

I've run radiance with two different parameter sets, the first with

_av_ is set to 0
xScale is set to 2
_ab_ is set to 6
_dc_ is set to 0.5
_aa_ is set to 0.2
_ad_ is set to 2048
_st_ is set to 0.5
yScale is set to 2
_ps_ is set to 4
_ar_ is set to 64
_as_ is set to 2048
_ds_ is set to 0.25
_pt_ is set to 0.1
_dr_ is set to 1
_pj_ is set to 0.9
_dp_ is set to 256
_dt_ is set to 0.25
_lr_ is set to 6
_dj_ is set to 0.5
_lw_ is set to 0.01

resulting in the image named Blotching001

The second case has the following parameters:

_av_ is set to 0
xScale is set to 6
_ab_ is set to 6
_dc_ is set to 0.75
_aa_ is set to 0.1
_ad_ is set to 4096
_st_ is set to 0.15
yScale is set to 6
_ps_ is set to 2
_ar_ is set to 128
_as_ is set to 4096
_ds_ is set to 0.05
_pt_ is set to 0.05
_dr_ is set to 3
_pj_ is set to 0.9
_dp_ is set to 512
_dt_ is set to 0.15
_lr_ is set to 8
_dj_ is set to 0.7
_lw_ is set to 0.005

resulting in the image named Blotching002. I have also attached an image
with no blotching. This case uses glazing with 65% light transmittance.

Although the second case is much better, it is still very bad for hours
when the sun is lower in the sky. The two above images are rendered for a
clear sky at 18:00 in Germany.

Can someone help?

Kind regards,
Orn Erlendsson

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

3

Orn,
With clear glass, a very significant contribution is made by the direct sun, so all the ambient bounces (and their calculation parameters) contribute less to the scene, and you see less blotches.

For the translucent glass, try again with all of your first run parameters, but instead of -ar 64 try -ar 500 or -ar 1000. This probably has the biggest impact on visibility of blotches. This is particularly important if your model includes a room that's part of a much larger building, or a full building floor with a very large site context modeled around it.

In the parameters you tested in the second run, when you increased -ad and lowered -aa those should also improve accuracy, but might not be necessary depending on how much importance you put on accuracy versus calculation time.

···

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

4

I retract my answer. Andy is right about it being splotches in the direct transmission. I was too quick to jump to conclusions when I read your subject and barely looked at the images.

···

From: Christopher Rush [mailto:[email protected]]
Sent: Friday, June 12, 2015 1:52 PM
To: Radiance general discussion
Subject: Re: [Radiance-general] Blotching caused by translucent material

Orn,
With clear glass, a very significant contribution is made by the direct sun, so all the ambient bounces (and their calculation parameters) contribute less to the scene, and you see less blotches.

For the translucent glass, try again with all of your first run parameters, but instead of -ar 64 try -ar 500 or -ar 1000. This probably has the biggest impact on visibility of blotches. This is particularly important if your model includes a room that's part of a much larger building, or a full building floor with a very large site context modeled around it.

In the parameters you tested in the second run, when you increased -ad and lowered -aa those should also improve accuracy, but might not be necessary depending on how much importance you put on accuracy versus calculation time.

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

5

Chris, I think you're suggestion is valid, particularly for reducing the
size of the blotches.
Andy

···

On Fri, Jun 12, 2015 at 11:03 AM, Christopher Rush < [email protected]> wrote:

I retract my answer. Andy is right about it being splotches in the
direct transmission. I was too quick to jump to conclusions when I read
your subject and barely looked at the images.

*From:* Christopher Rush [mailto:[email protected]]
*Sent:* Friday, June 12, 2015 1:52 PM
*To:* Radiance general discussion
*Subject:* Re: [Radiance-general] Blotching caused by translucent material

Orn,

With clear glass, a very significant contribution is made by the direct
sun, so all the ambient bounces (and their calculation parameters)
contribute less to the scene, and you see less blotches.

For the translucent glass, try again with all of your first run
parameters, but instead of -ar 64 try -ar 500 or -ar 1000. This probably
has the biggest impact on visibility of blotches. This is particularly
important if your model includes a room that’s part of a much larger
building, or a full building floor with a very large site context modeled
around it.

In the parameters you tested in the second run, when you increased –ad and
lowered –aa those should also improve accuracy, but might not be necessary
depending on how much importance you put on accuracy versus calculation
time.

____________________________________________________________
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

6

Thank you all for your reply. I have now tried several variations with your suggestions and although the blotching issue is not resolved, the images do get slightly better. Could the issue in fact be that the diffuse and specular transmissive properties are so low that in reality very little light would pass into the scene, thus causing a very low light condition closely represented by the blotching case we see from the simulations?

Just to clarify the light transmission of translucent glazing is equal to the sum of the diffuse and specular transmission.... right?

Örn

···

From: [email protected]
Date: Fri, 12 Jun 2015 12:17:55 -0700
To: [email protected]
Subject: Re: [Radiance-general] Blotching caused by translucent material

Chris, I think you're suggestion is valid, particularly for reducing the size of the blotches.Andy

On Fri, Jun 12, 2015 at 11:03 AM, Christopher Rush <[email protected]> wrote:

I retract my answer. Andy is right about it being splotches in the direct transmission. I was too quick to jump to conclusions when I read your subject and barely looked at the images.

From: Christopher Rush [mailto:[email protected]]

Sent: Friday, June 12, 2015 1:52 PM

To: Radiance general discussion

Subject: Re: [Radiance-general] Blotching caused by translucent material

Orn,
With clear glass, a very significant contribution is made by the direct sun, so all the ambient bounces (and their calculation parameters) contribute less to the scene, and you see less blotches.

For the translucent glass, try again with all of your first run parameters, but instead of -ar 64 try -ar 500 or -ar 1000. This probably has the biggest impact on visibility of blotches. This is particularly
important if your model includes a room that’s part of a much larger building, or a full building floor with a very large site context modeled around it.

In the parameters you tested in the second run, when you increased –ad and lowered –aa those should also improve accuracy, but might not be necessary depending on how much importance you put on accuracy versus
calculation time.

____________________________________________________________

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

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

7

This seems like a poster case for the new photon-mapping routines. If you are working with the latest alpha version, you should definitely give that a try. The irradiance cache isn't built to handle specular scattering like this. You would have to turn it off with -aa 0 and put up with the noise. Using a caustic photon map avoids the issue. It's even built into the latest rad program, so you just need to set up your maps using:

  PGMAP= scene_g.pmp 200K 50
  PCMAP= scene_c.pmp 1M 50

Rad will run mkpmap and set the rvu or rpict options appropriately, and you can watch what it does. It typically spits out things like:

  oconv sky.rad window3b.rad office.rad window_port.rad > test3f.oct
  mkpmap -apo window_port -fo+ -apg test3f_g.pmp 200K -apc test3f_c.pmp 1M test3f.oct
  rpict -vu 0 1 0 -vf fish.vf -dp 512 -ar 185 -ms 1.3 -ds .3 -dt .1 -dc .5 -dr 1 -ss 1 -st .1 -ab 1 -af test3f.amb -aa .1 -ad 1536 -as 392 -av 10 10 10 -lr 8 -lw 1e-4 -ap test3f_g.pmp 50 -ap test3f_c.pmp 50 -av 0.8 0.8 0.8 -x 64 -y 64 -ps 1 test3f.oct > /dev/null
  rpict -vu 0 1 0 -vf fish.vf -x 1024 -y 1024 -dp 512 -ar 185 -ms 1.3 -ds .3 -dt .1 -dc .5 -dr 1 -ss 1 -st .1 -ab 1 -af test3f.amb -aa .1 -ad 1536 -as 392 -av 10 10 10 -lr 8 -lw 1e-4 -ap test3f_g.pmp 50 -ap test3f_c.pmp 50 -av 0.8 0.8 0.8 -ps 6 -pt .08 test3f.oct > test3f_fish.unf
  pfilt -1 -e .05 -r .6 -x /2 -y /2 test3f_fish.unf > test3f_fish.hdr
  rm -f test3f_fish.unf

You can dig through these for hints of how to run everything if you prefer doing it by hand.

Cheers,
-Greg

···

From: Örn Erlendsson <[email protected]>
Subject: Re: [Radiance-general] Blotching caused by translucent material
Date: June 24, 2015 12:54:38 AM PDT

Thank you all for your reply. I have now tried several variations with your suggestions and although the blotching issue is not resolved, the images do get slightly better. Could the issue in fact be that the diffuse and specular transmissive properties are so low that in reality very little light would pass into the scene, thus causing a very low light condition closely represented by the blotching case we see from the simulations?

Just to clarify the light transmission of translucent glazing is equal to the sum of the diffuse and specular transmission.... right?

Örn

From: [email protected]
Date: Fri, 12 Jun 2015 12:17:55 -0700
To: [email protected]
Subject: Re: [Radiance-general] Blotching caused by translucent material

Chris, I think you're suggestion is valid, particularly for reducing the size of the blotches.
Andy

On Fri, Jun 12, 2015 at 11:03 AM, Christopher Rush <[email protected]> wrote:
I retract my answer. Andy is right about it being splotches in the direct transmission. I was too quick to jump to conclusions when I read your subject and barely looked at the images.

From: Christopher Rush [mailto:[email protected]]
Sent: Friday, June 12, 2015 1:52 PM
To: Radiance general discussion
Subject: Re: [Radiance-general] Blotching caused by translucent material

Orn,

With clear glass, a very significant contribution is made by the direct sun, so all the ambient bounces (and their calculation parameters) contribute less to the scene, and you see less blotches.

For the translucent glass, try again with all of your first run parameters, but instead of -ar 64 try -ar 500 or -ar 1000. This probably has the biggest impact on visibility of blotches. This is particularly important if your model includes a room that’s part of a much larger building, or a full building floor with a very large site context modeled around it.

In the parameters you tested in the second run, when you increased –ad and lowered –aa those should also improve accuracy, but might not be necessary depending on how much importance you put on accuracy versus calculation time.

8

Did you try setting roughness to zero in your trans definition? That solves
it for me.
Though if the small extra scatter you get from the roughness is important
to your simulation, than Greg has a good point. Photonmap is best.

Andy

···

On Wed, Jun 24, 2015 at 12:54 AM, Örn Erlendsson <[email protected]> wrote:

Thank you all for your reply. I have now tried several variations with
your suggestions and although the blotching issue is not resolved, the
images do get slightly better. Could the issue in fact be that the diffuse
and specular transmissive properties are so low that in reality very little
light would pass into the scene, thus causing a very low light condition
closely represented by the blotching case we see from the simulations?

Just to clarify the light transmission of translucent glazing is equal to
the sum of the diffuse and specular transmission.... right?

Örn

------------------------------
From: [email protected]
Date: Fri, 12 Jun 2015 12:17:55 -0700
To: [email protected]
Subject: Re: [Radiance-general] Blotching caused by translucent material

Chris, I think you're suggestion is valid, particularly for reducing the
size of the blotches.
Andy

On Fri, Jun 12, 2015 at 11:03 AM, Christopher Rush < > [email protected]> wrote:

I retract my answer. Andy is right about it being splotches in the
direct transmission. I was too quick to jump to conclusions when I read
your subject and barely looked at the images.

*From:* Christopher Rush [mailto:[email protected]]
*Sent:* Friday, June 12, 2015 1:52 PM
*To:* Radiance general discussion
*Subject:* Re: [Radiance-general] Blotching caused by translucent material

Orn,

With clear glass, a very significant contribution is made by the direct
sun, so all the ambient bounces (and their calculation parameters)
contribute less to the scene, and you see less blotches.

For the translucent glass, try again with all of your first run
parameters, but instead of -ar 64 try -ar 500 or -ar 1000. This probably
has the biggest impact on visibility of blotches. This is particularly
important if your model includes a room that’s part of a much larger
building, or a full building floor with a very large site context modeled
around it.

In the parameters you tested in the second run, when you increased –ad and
lowered –aa those should also improve accuracy, but might not be necessary
depending on how much importance you put on accuracy versus calculation
time.

____________________________________________________________
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

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

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