Irradiance Particles

[bart]

In preparation for an IP Tips page, here we'll explain the new Irradiance Particles (IP) feature of mental ray 3.6+ and 3.7, comparing it with Final Gather (FG) and Global Illumination (GI), ie, photon mapping. One selects these techniques to calculate indirect diffuse illumination at a given sample point hit by the eye ray at (tile) rendering time. Although FG and GI can be used either together or separately, IP remains exclusive. However, do note that Caustics can be used with IP.

Any indirect illumination technique (FG, GI/Caustics, IP) requires that your material shader include an indirect diffuse illumination component. (Detail for shader writers: this is the simple inclusion of the function mi_avg_radiance().)

In short, for IP mental ray performs two stages of pre-processing before tile rendering. In the first stage before rendering, mental ray shoots importons into the scene from the camera to determine the positions of these new irradiance particles. In a second stage before rendering, at each of these positions, mental ray determines the amount of incoming direct illumination, and indirect illumination if indirect passes are allowed.

After the pre-processing, during tile rendering, mental ray looks above the shading sample point (where the eye ray hits) for light contribution from irradiance particles.

If the IPs had only collected direct illumination, only one bounce of indirect diffuse light will be contributed to that shading sample. This is similar to FG with no secondary FG rays, or FG trace depth set to 0.

Because the IP contribution is calculated at each shading sample point, this is more like the new FG force mode, where each sample point sends out FG rays rather than interpolating pre-processed FG points. However, one can set an IP interpolation mode similar to the normal FG interpolation modes. We often call that smoothing as well. In the interpolated mode, nearby IPs contain pre-computed irradiance which can then be interpolated for the given shading sample point. In practice, the noise from undersampling is a tradeoff between graininess and splotchiness. The smoothing modes are faster because mental ray performs less calculation. The interpolation/smoothing modes also work faster with motion blur because of re-use.

In addition to calculating the indirect diffuse contribution from other objects in the scene, this technique can also add indirect diffuse illumination from the environment. One might want to think of these as special environment IPs, which have only an environment direction associated with them rather than a position.

Note that the irradiance particles technique incorporates a key concept called "importance sampling". This helps keep sampling more efficient, focusing render time where it is needed. The sampling of IPs above any given point are not uniform, but rather take into account how important any given direction above that point is. For example, some directions may include more light, or more changing of illumination conditions than others. So before shooting the (IP) rays, there's an estimate of the importance of some directions with respect to others, and more rays are shot towards brighter parts of the particle map, while less rays are shot towards the darker parts.

Since the IP technique replaces other indirect illumination techniques, such as FG and GI, we must specify what mental ray does when several techniques have been accidentally specfied together. Simply put, FG precedes, IP which precedes GI. If IP and GI are enabled, then GI is switched off. If FG and IP are enabled, then IP is switched off. As previously mentioned, Caustics can be used with IP. This is because they account for different types of indirect illumination (IP currently accounts for indirect diffuse illumination from diffuse surfaces, whereas Caustics accounts for indirect diffuse illumination from specular surfaces.).

Currently, the IP technique depends on photon shaders, as they are used in the importon stage in the reverse light path direction, ie eye-to-light rather than light-to-eye. If the material of an object has no photon shader, then mental ray cannot scatter the importons.

In general, the IP technique should be faster than FG for non-trivial scenes, and especially in scenes with very uneven lighting distribution, because of it's importance driven algorithm. We might classify non-trivial scenes as those which take more than a few minutes to render.

Currently, the IP techique is best suited for scenes with more diffuse-to-diffuse surface interactions. FG will handle scenes with lots of mirrors and transparency better. Also FG will better handle light cards, or materials which have only self-illumination/incandescence.

The smoothing artifacts of IP (when used with interpolation) should be less evident than the smoothing artifacts produced by photon mapping for similar speed/quality. As with FG, GI will handle non-diffuse surface interactions better.

Unlike FG or GI, currently we have no per-object settings for the IP technique. We set up IP with 'global' string options.

[bart]

Now, we'll proceed first with some simple scene examples, and we'll move on to more complex ones as we go. Thanks to dagon, we have many scene examples to show. But we should start slowly. Note that I will use a scene which works in Maya 2008 SP1/Ext. 2 to begin with, but we will also eventually supply similar scenes using XSI 7, and 3ds max 2009. You might need special shaders to set the string options for 3ds max, in that version.

First here is an example of the string options on which IP depends (grabbed from an export):

Code: Select all

    "importon emitted" 1000
    "importon merge" 0.
    "importon trace depth" 2
    "importon traverse" on
    "irradiance particles" on
    "irradiance particles env" on
    "irradiance particles env rays" 64
    "irradiance particles env scale" 1.
    "irradiance particles file" "ipsimple"
    "irradiance particles indirect passes" 0
    "irradiance particles interpolate" "always"  # "never" is like FG force
    "irradiance particles interppoints" 64
    "irradiance particles rays" 64
    "irradiance particles rebuild" on
    "irradiance particles scale" 1.

You see the options for the importon stage and then the IP irradiance calculation stage. Note that we have 0 indirect passes, which means it is similar to FG without secondary trace depth enabled, ie no secondary FG rays.

Also, note that importon trace depth is 2, not much bouncing around.

[bart]

Here's a render from a simple scene using those settings. It uses only mia_physical_sky/sun and mia_exposure_photographic on the camera lens.
The Maya 2008 SP1/Ext2 version of the scene is here: http://www.lamrug.org/resources/scenes/ipsimple.mb
That scene is not without issues (odd crashes at unexplainable times for me) in that version of Maya.
An application-clean version of the exported mi file for standalone is here: http://www.lamrug.org/resources/scenes/ipsimple.mi



Now, if we increase importon trace depth to 4, we'll get a bit better importon bounce around from the camera's perspective. Remember, that these are importons which determine the locations for Irradiance Particles. They also help determine importance. Depth of 4 in image below:



Note, that we could have also increased importons emitted from 1000 to something greater, like 2000 as it is below, with depth kept at 2.



Note,if our importons don't bounce enough with low numbers, it could lead to unwanted effects. Here is an importon depth of 1 with only 700 importons. Note the anomalous lighting under the main table.



Now, if we want more actual light bounce, because after all this is a highly indirectly lit scene, lets increase the IP indirect passes, below we go next to 1, then 2, 3 and 4, all with an importon depth of 2 and 1000 importons. Like increasing FG secondary diffuse reflection trace depth, you will notice increased light.

[bart]

Now, similar to what we do in training class, it can be instructive to reduce interpolation in order to see how well the IPs are distributed. It also gives you a feel for how much smoothing happens as the interpolation number is increased.

Below, we show the number of interpolation points at 1,2, then 10, and 20. Above they are set at 64.

[bart]

Now, let's keep the interpolation down to 1, and get a good distribution of IPs. Above, there were 1000 importons emitted, and in these nexte images, we'll used 2000, 3000, and 4000.

[dagon]

Bart has already explained how the IP works, now i'll show you an analysis over the IP options, with some hint on the IP setup.
We will see two different approachs:
1- non-interpolated IP (Interpolate = never)
2- interpolated IP (Interpolate = always)
The big difference over the two methods is the "final irradiance" computation, that is pre-computed in the interpolated IP (the final IP pass) and computed at-rendertime (during the raytracing) in the non-interpolated IP.


1- non-interpolated IP
This method is very qualitative, as you can get something similar to a brute force render, but it's quite expensive in therms of rendertime. We'll see how to speed it up.

Depth and Passes
Here's a chart where you can see the differences in therms of quality and rendertimes



The Depth option control the "bounces" of importons over the scene, the higher the depth the qualitative the results.
As you can see rendertimes are better with higher Depths.
The Passes option control the diffuse bounces on the scene.
As you can see in the non-interpolated IP increasing the Passes value hasn't a big effect in therms of rendertimes, the time spent in the IP generation is compensated in the raytracing phase (this is because better illumination means less contrast and faster AA in the non-interpolated IP).
Hint: use always a good Depth (2/4) and change Passes depending on your scene, don't increase too much the Passes value, as it increase rendertimes (exteriors:2/3, interiors:3/5)


Importon Emitted (Importon Density)
This define how much importons are traced from the camera to the scene.
You can use two different options:
- Importons Emitted, define the amount of importons emitted, this number is unleashed to the image resolution.
- Importons Density, define the amount of importons related to the image resolutions, where Density = 1 means 1 importon per pixel, Density = 0,5 means 1 importons per 2 pixels, etc.

N.B.
In this analysis i'm using the Importons Emitted option, but you can easily relate it to the Density option (i'm using a resolution of 640x480 (307.200 pixels) so 30.000 is Density ~0,1, 3.000 is ~0,01)
In Maya 2009, if you use the standard Render Settings, you can't use the Emitted option, but If you want you can enable it via Strings:

Code: Select all

setAttr -type "string" miDefaultOptions.stringOptions[7].name "importon emitted";
setAttr -type "string" miDefaultOptions.stringOptions[7].type "integer";
setAttr -type "string" miDefaultOptions.stringOptions[7].value "10000";

As you can see the number of importons (emitted/density) affect the quality of details founded in the scene, the higher the number the higher the detail level.
Hint: start with a low value, using low AA settings, and increase it only if you need more details (density 0,02/0,3; emitted 10.000/100.000)


Irradiance Particles Rays
This define how much QMC rays are shooten for the irradiance sampling, and determine the quality of the visual results.







As you can see lower values involve more noise (and lower rendertimes), higher values involve less noise (but higher renderimes)
It's very important to understand that the number of rays is influenced both by the number of importons and the number of passes. For each bounce of light (pass) mental ray use the IP Rays value to sample the irradiance.


Irradiance Particles Env. Rays
This control the number of rays used for the environment sampling.
Irradiance Particles can do importance sampling on the scene environment, this is one of the most important advantage over the other GI methods (GI/FG)







As you can see the Env. Rays define the quality of light coming from the environment.
Rendertimes are not affected too much if you use reasonable values.


IP Setup

Let's start a short tutorial to show you how get a good non-interpolated IP in few steps and with fast rendertimes (remember, this is a very qualitative method, so fast is a relative therm )
N.B.
I'm using Maya 2009 for this tutorial, if you use a previous version of mental ray (Maya 2008, Max 2009) you may expect worse results.

In first we need to find a good level of details for our scene.
For fast feedback set a low value for Rays (Importons and Env.) and AA:
Importons Rays = 64
Env.Rays = 64
AA: -2 -1
Specify a location and a name for you IP map.





We can get a good level of details with 30.000 importons, this is my choice.

Now we need to "smooth" the noise, but we need also to keep low rendertimes.
The "official" way to smooth the noise is to higher the Irradiance Particles Rays (and the Env. Rays), but this is quite expensive in therms of times, you can see it in the next render:
Importons Emitted = 30.000
Importons Rays = 1000*
Env.Rays = 500
AA: -2 -1


We can use a little "trick" to optimize our renderimes.
We need to use a low level of Importons and Env. Rays (50/100) to keep the importons generation fast.
Then we can re-use the map unchecking the "rebuild" option and increasing the Rays values, this changes the "final

irradiance" computation (at-rendertime) without increasing the times of the IP generation.
We can stop the firt render (the one with low level of Rays) before the raytracing phase.
First render:
Importons Emitted = 30.000
Importons Rays = 64
Env.Rays = 64
AA: -2 -2 (if you don't remember to stop it )
Second render (rebuild off):
Importons Emitted = 30.000
Importons Rays = 1000*
Env.Rays = 500
AA: -2 -1



Ok, now you if you compare it to the previous render you can see that the quality is quite the same, but the rendertimes are much better now.



*in Maya 2009 you can't set more the 512 Rays, you can use the string options to set an higher value:

Code: Select all

setAttr -type "string" miDefaultOptions.stringOptions[12].name "irradiance particles rays";
setAttr -type "string" miDefaultOptions.stringOptions[12].type "integer";
setAttr -type "string" miDefaultOptions.stringOptions[12].value "1000";

coming next: interpolated IP

[dagon]

2- interpolated IP
This method is very similar to the regular FG, now we will see the differences in setup and quality.

Depth and Passes
Here's a chart where you can see the differences in terms of quality and render times



As you can see, there are some differences from the non-interpolated IP example; with interpolation, most of the render time (in a simple scene like this) is spent on IP generation, so, in this case, more bounces affect even more.

rendertimes
Hint: use always a good Depth (2/4) and change Passes depending on your scene, don't increase too much the Passes value, as it increases render times (exteriors:2/3, interiors:3/5)


Importon Emitted (Importon Density)
This defines how many importons are traced from the camera to the scene.
You can use two different options:
- Importons Emitted, defines the amount of importons emitted, this number is unleashed to the image resolution.
- Importons Density, defines the amount of importons related to the image resolutions







Once again, the number of importons (emitted/density) affects the quality of details found in the scene; the higher the number, the higher the detail level.

Hint: with the interpolated IP you can try two different ways to get the details; we will see it on the last part of the tutorial.


Irradiance Particles Rays
This defines how many QMC rays are shot for the irradiance particles sampling, and determines the quality of the visual results.







Increasing the number of Rays you can reduce the noise generated, but this is very expensive in therms of render times, especially if you use many bounces.


Irradiance Particles Env. Rays
This controls the number of rays used for the environment sampling.







Render times are not affected too much if you use reasonable values.


Interpoint
This controls how many IPs are "combined" for the interpolation. Increasing this value you get a smoother render, but with less details.






IP Setup

Now we will see how to setup a good interpolated IP in few steps, using two different approach:
2.1- Increase the number of importons (we will see that the more importons we use, the more Rays we need to smooth the solution)
2.2- Use a low number of importons and add an AO pass (the AO+color bleed mode, or ao_do_details = 2, is recomended)

N.B.
I'm using Maya 2009 for this tutorial, if you use a previous version of mental ray (Maya 2008, Max 2009) you may expect worse results.


2.1
The first step is always to find the right level of details.
For fast feedback set a low value for Rays (Importons and Env.) and AA:
Importons Rays = 64
Env.Rays = 64
AA: -1 1
Specify a location and a name for you IP map.





The problem in this kind of setup is to find the right balance between quality and render times.
The number of importons used is the first step to find this balance. As you can see, if we increase the importons too much, we get very high render times (and, for now, we are using only few Rays, this problem is much worse when we increase it), my choice is 30.000.

Now to "smooth" the noise let's increase the Rays

Importons Emitted = 30.000
Importons Rays = 500*
Env.Rays = 500
Interpoint = 32
AA: -1 1


Unchecking "rebuild" we can increase the Interpoint (=128) to reduce the restante noise a little bit


With interpolation the "trick" we used to optimize render time in the non-interpolated IP isn't as effective (there are too many artifacts generated); so if we want speed up render times a bit, we need another solution.

2.2
This is a method I've used many times with FG.
FG can be very fast for clean solution, but it tends to loose details.
AO+color bleed (or ao_do_detail=2) can help to enhance the detail level without increasing too much rendertimes.
We can use this approach for the IP too, what we need is a smooth and fast render.
We can get it reducing the number of importons, let's try with:
Importons Emitted = 10.000
Importons Rays = 500
Env.Rays = 500
Interpoint = 32
AA: -1 1



Now we can add the AO+color bleed to the mia_material_x and, eventually, we can increase the Interpoint to 128
Samples = 32
Distance = 50
Ambient Shadow Color = black
Ambient Light Color = black


Ok, that's it, very fast setup, good quality and fast render times, we will see that the IP can be much more effective with complex scene.
Obviously you can mix the first method (2.1) to this one to get different balances between quality and render times.




*in Maya 2009 you can't set more the 512 Rays, you can use the string options to set an higher value


coming next: Importance Sampling with HDRI example

[brogh]

Excellent Post guys !,

I will also test some things tomorrow this new release of MR seems awesome, actually i'm more un passes

Thanks For this !

Cheers

[popbot]

irradiance particles looks very nice - what about animation i read it works better than fg some one test it

[dagon]

Importance Sampling with HDRI


- Example 1: Simple interior

Now we will see probably the most interesting feature in the IP, the environment sampling.
As you probably know, FG and GI aren't very good for HDRI lighting (IBL).
GI can't shoot photons from environment, and FG can't do a good sampling on the high dinamic range of the environments (there's why there is a filter on the FG options).
We are used to use some tricks to surpass this problem, the most common is to use some sort of direct light (a light for direct illumination, direct/spot/arealight or, in the last mray releases, skyportals) to get shadows with a good contrast.
What we'll see now is how to use the IP to surpass this problem, we don't need direct lights anymore (exept for artificial lighting... obviously)

Let's start the test with our scene, this is a very simple scene, but it can show very well the prerogatives of IP.
you can download the scenre from here: http://www.lamrug.org/resources/scenes/ipsimple09hdri.mb
I'll use a simple HDRI file, probe_16-20_anglemap.hdr, that you can download from http://gl.ict.usc.edu/skyprobes/
No direct lights in the scene, only an IBL* and and the IP.

The setup is very simple, so i'm not annoying you anymore, let's see my steps:
- Finding a good level of details:





My choice is 30.000 importons, now i want a good sampling, keeping good rendertimes, we'll use our method for the non-unterpolated IP:

- Defining the sampling quality
First render with the IP only (render stopped after IP generation)
importon emitted = 30.000
IP rays = 64
env. rays = 500
passes = 4
depth = 4
save the IP with a name
Second render with the final image
importon emitted = 30.000
IP rays = 500
env. rays = 1000
passes = 4
depth = 4


As you can see here i'm using an higher value for the env. rays, this is because the environment light here has a big impact in the render quality, if you lower too much the env. rays you loose details from the HDRI (in our case, the sun contribution).

Let's see it with a better AA


As you can see the result is better (and a bit faster) then with the sky/sun system.
This is due to the optimized environment sampling.

Just out of curiosity, here's the equivalent render done with the FG "force"



For the interpolated IP we will use the two approachs used before:

- Finding detail level




- Defining sampling quality



- Finding (low) detail level


- Adding details with AO




- Example 2: Car rendering

Let's see something more, the HDRI workflow is used very often in car renderings.
Here's a free model from http://www.3dmodelfree.com/models//25017-0.htm.
You can download the scene from here:
http://www.lamrug.org/resources/scenes/ipcar09hdri.mb
http://www.lamrug.org/resources/scenes/ipcar09hdri.zip
You need also the probe_16-20_anglemap.hdr frome here:
http://gl.ict.usc.edu/skyprobes/
With the IP you don't need an additional "direct" light anymore.
As already seen we can try our steps:

- Finding details level





- Defining sampling quality


As you can see the render as a very good level of detail, and we can reach it with few steps.

With better AA



Now, we can try to add some materials, to see how much it affects rendertimes.

As you can see the time to generate the IP it's pretty much the same.

We can take the chance to use the "secondary interpolation", this method mix the non-interpolated IP and interpolated, it can use the first for the diffuse materials and the second for the reflective/refractive materials. It can be very helpfull to improve the quality vs speed ratio.
I'll use the non-interpolated method for this:

First render with the IP only (render stopped after IP generation)
importon emitted = 150.000
interpolate = never
IP rays = 64
env. rays = 300
passes = 3
depth = 3
save the IP with a name
Second render with the final image
importon emitted = 150.000
interpolate = secondary
IP rays = 500
env. rays = 600
interpoints = 128
passes = 4
depth = 4


As Bart already said, for now the IP is best suited for diffusive lighting, the reflective/refractive behaviour is not computed, you can see it in the last render (the interior part of the car and the illuminator group lack the refractive contribution (i.e. caustics) coming from the HDRI, they're very dark and there is no way to light it up without tricks.



*in Maya 2008 you can't use the IBL node, you may use the mib_lookup_spherical instead
in 3ds Max 2009 you can use the skylight

coming next: Complex Interior Test

[seq]

hi dagon,

can you please answer a few questions:

1. is the amount of importons dependent on the image resolution that we render?

2. when we render out an ip map do we need to use the same resolution as for the image rendering?

3. do we need to render an ip map for every viewpoint? e.g. we have one room and one total view and four mid shots. the light situation is the same, only the exposure settings are changing.

thanks for any answers. your input here is just great!

[Remydrh]

I'm doing something wrong. While I get some indirect lighting, the objects not facing the light are always black regardless of max depth and indirect passes. And it crashes a lot, but that's neither here nor there.

[bart]

Remydrh,

Using ipsimple.mb from above in Maya2008 SP1/Ext2? I notice that I can see the environment in your window. Did you turn the shutter speed way up in the tone mapper?

seq,

1. We have two ways to control importon emission count, using "importon density" or "importon emitted". The latter supercedes the former if they are both used, and it is resolution independent. However, density is indeed a resolution dependent setting, as it specifies the number of importons emitted per pixel. It can be greater than or less than 1.0, and the default is 1.0 (or 1 importon emitted per pixel). Currently, using "importon density" is the recommended way to go since the whole effect is meant to take advantage of view-dependence.

2. So if you use "importon emission" you don't need to render at the same resolution to store the IP map. But you should render at the same resolution if you use "importon density". Either that, or increase your density, when you render at lower resolution.

3. Yes. Importons determine view-based importance at each importon intersection point, where the IPs end up being located. The idea is that you should get a good speed/quality tradeoff because of this calculation which is view-dependent. That said, if two shots are similar in position, you might be able to get away with sharing an IP map.

[Remydrh]

I actually just whipped together a scene myself (maybe the problem) because I was using something more complex before and was getting the same issue where I had completely black areas. So I decided to go back to the basics. But I still end up with black areas. It's slow going because it crashes after the second render. Autodesk's report inbox will be full by the time I figure this out tonight.

It's physical sun and sky at default. I just didn't kill the background. With FG it's fine, dark some but not dark like it is with the IPs. Side question, does the portal light care about IPs or just FG? Not using it but the though had crossed my mind.

And so far IP rendering is so slow compared to FG I am still not sold on its performance. I was hoping I could generate a map and run with it (kind of like photons) but since it is importance sampled based on view I'd have to rebuild when my camera moves or is that misleading?

[bart]

version of Maya? There's a factor of pi difference for the environment. Maya 2009 is brighter and more accurate. What kind of tone mapping?

Do you want to send me your simple scene in mail? If its obvious, then we should mention it here for others who might run into the same issue.

[Remydrh]

I'll send it. I might have a setting wrong. I am also trying one of the HDRs listed above and I am getting a similar "dark side of the moon" issue where the rest is blowing way WAY out but backsides from the window are pitch black.

It's Maya 2009.

[seq]

Remydrh,

Using ipsimple.mb from above in Maya2008 SP1/Ext2? I notice that I can see the environment in your window. Did you turn the shutter speed way up in the tone mapper?

seq,

1. We have two ways to control importon emission count, using "importon density" or "importon emitted". The latter supercedes the former if they are both used, and it is resolution independent. However, density is indeed a resolution dependent setting, as it specifies the number of importons emitted per pixel. It can be greater than or less than 1.0, and the default is 1.0 (or 1 importon emitted per pixel). Currently, using "importon density" is the recommended way to go since the whole effect is meant to take advantage of view-dependence.

2. So if you use "importon emission" you don't need to render at the same resolution to store the IP map. But you should render at the same resolution if you use "importon density". Either that, or increase your density, when you render at lower resolution.

3. Yes. Importons determine view-based importance at each importon intersection point, where the IPs end up being located. The idea is that you should get a good speed/quality tradeoff because of this calculation which is view-dependent. That said, if two shots are similar in position, you might be able to get away with sharing an IP map.

thanks Bart. Great help.

we still have problems when rendering with interpolated IP a scene with a coloured glas surface. as you can see in the following image, we get strange artefacts on the green glas surface in the back of the room. do you know why this is happening?

we use 100.000 importons, max depth 4/4, 600 IP rays, 200 interpoints and 1000 env rays.

the second image shows the same scene with non-interpolated IP.

thanks you for your help.

[dagon]

And so far IP rendering is so slow compared to FG I am still not sold on its performance. I was hoping I could generate a map and run with it (kind of like photons) but since it is importance sampled based on view I'd have to rebuild when my camera moves or is that misleading?

IP is slower but more accurate (as you can see in my test with environment sampling, this is simply... impossible with FG, even with brute force FG)
actually is a little bit complicated to find the good ratio quality/speed... hopefully it will be solved in the next releases

Side question, does the portal light care about IPs or just FG? Not using it but the though had crossed my mind.

portal light was developed to help the "environment sampling" with FG
with IP you don't need it, the importance sampling works pretty well with environments

I'll send it. I might have a setting wrong. I am also trying one of the HDRs listed above and I am getting a similar "dark side of the moon" issue where the rest is blowing way WAY out but backsides from the window are pitch black.

It's Maya 2009.

depth value?

[Remydrh]

Hey Dagon,

Which image were you using as your HDR? There is a sequence on the site. I picked a middle image.

Depth is 4 and 4 indirect passes. It doesn't light the scene anywhere near as evenly as yours. The back sides (not facing the window) are all completely black. There's a checkbox somewhere I'm missing isn't there?

Do you have that file using the HDR available with the settings for 2009? I have the HDRs already but I am wondering what the differences might be. It's like pathtracing with poor depth.

[dagon]

Hey Dagon,

Which image were you using as your HDR? There is a sequence on the site. I picked a middle image.

Depth is 4 and 4 indirect passes. It doesn't light the scene anywhere near as evenly as yours. The back sides (not facing the window) are all completely black. There's a checkbox somewhere I'm missing isn't there?

Do you have that file using the HDR available with the settings for 2009? I have the HDRs already but I am wondering what the differences might be. It's like pathtracing with poor depth.

probe_16-20_anglemap.hdr
you need to use the "angular" mapping of IBL (and unchecking "infinite" you can also rotate ora move a little bit down the sphere)
yes i have the files, i'll upload it asap

oh, are you using the importon density or the importon emitted?

[Remydrh]

I'm using density, but it's fairly low. So about 90,000 importons. I also used angular but left infinite checked on. I'll check for the files when you upload them. Thanks!

[dagon]

I'm using density, but it's fairly low. So about 90,000 importons. I also used angular but left infinite checked on. I'll check for the files when you upload them. Thanks!

may you try with the importon emitted?
i think there is a bug in the density option (you can try also with many different density values)

[Remydrh]

I will try emitted, but I tried .1, .2, .3, .4 and 1.0 densities. Depends on the machine I'm on. Some workstations are better than others. I can get a nice smooth result, just not well lit.

[dagon]

I will try emitted, but I tried .1, .2, .3, .4 and 1.0 densities. Depends on the machine I'm on. Some workstations are better than others. I can get a nice smooth result, just not well lit.

ok, so you're getting the same dark problem with different density values?

[Remydrh]

Yep, always darkness.