adding rendering parameter - pixel samples

[pclaes]

Hi,

I am trying to add the pixel sampling parameters to control the pixel samples of one object in the scene. I have tried adding it to the shader and to the object, but it does not seem to influence the outcome at all.

My problem is the following:

I am using a glass material on an object with pbr rendering, I have a reflection sphere around the object (with a constant material) and I have some depth of field going on.

The depth of field requires about 6 x 6 pixel samples to look good, but the object with the glass requires around 15 x 15 samples to get rid of some of the noise otherwise produced.

1) My question and assumption is that I want to set the pixel samples (or increase the quality in another way) for the object that has the glass material on it. So where do I add the rendering parameters? On the shader, or on the object? Or can I not do this?

2) The reason why I am using PBR and not micropolygon rendering for this is that the scene renders in about 2 minutes with PBR, and around 45 minutes with micropolygon rendering. I am guessing I am doing something wrong with micropolygon rendering. When rendering with PBR I do seem to lose the color accumulation that takes place over the density of the material. I would prefer to stick with micropolygon rendering as it has proven a bit more stable, but with the time it takes per frame it is just not possible.

thanks for any tips.

[symek]

Hi,

I am trying to add the pixel sampling parameters to control the pixel samples of one object in the scene. I have tried adding it to the shader and to the object, but it does not seem to influence the outcome at all.

My problem is the following:

I am using a glass material on an object with pbr rendering, I have a reflection sphere around the object (with a constant material) and I have some depth of field going on.

The depth of field requires about 6 x 6 pixel samples to look good, but the object with the glass requires around 15 x 15 samples to get rid of some of the noise otherwise produced.

1) My question and assumption is that I want to set the pixel samples (or increase the quality in another way) for the object that has the glass material on it. So where do I add the rendering parameters? On the shader, or on the object? Or can I not do this?

2) The reason why I am using PBR and not micropolygon rendering for this is that the scene renders in about 2 minutes with PBR, and around 45 minutes with micropolygon rendering. I am guessing I am doing something wrong with micropolygon rendering. When rendering with PBR I do seem to lose the color accumulation that takes place over the density of the material. I would prefer to stick with micropolygon rendering as it has proven a bit more stable, but with the time it takes per frame it is just not possible.

thanks for any tips.

Well, I don't think you can control pixel sampling on per object basis, in other way then simply rendering it in separated layers with different settings in Mantra ROP. As to glass artifacts, I think you can control ray tracing quality (rays varies) independently from a pixel sampling. This way you can save some CPU cycles, increasing rays amount but with stable pixel sampling.

Does it have sense?

sy.

Edited April 16, 2008 by SYmek

[pclaes]

Well, I don't think you can control pixel sampling on per object basis, in other way then simply rendering it in separated layers with different settings in Mantra ROP. As to glass artifacts, I think you can control ray tracing quality (rays varies) independently from a pixel sampling. This way you can save some CPU cycles, increasing rays amount but with stable pixel sampling.

Does it have sense?

sy.

Thank you for clarifying, to a degree it does make sense, I will try increasing the ray tracing quality.

I think I will try to split it in different passes as much as I can.

What I wonder is then: When can I use those rendering parameters from the edit parameter interface? Are those parameters there to be used mainly for creating integration with other renderers? (I assume these are the kind of parameters Jason is using for his Indigo render support)

So far I have used them to add reflection and refraction limits to the glass shader in 9.1 (because they removed those properties since 9.0) and also added the soho_autoheadlight parameter to one my rops. Initially I thought those rendering parameters were similar to rendering flags, like "mantra -n 3" or something similar. But this seems not the case.

I have started rendering with raytracing and that seems to give a compromise between the micropolygon and pbr. pbr is fast, but does not pick up the color in the density. Micropolygon is fast for the environment (depth of field, mblur, displacement) but very slow for the glass object. Raytracing is fast enough and accumulates color in density, but seems slow for depth of field.

thanks again for your help.

[symek]

Thank you for clarifying, to a degree it does make sense, I will try increasing the ray tracing quality.

I think I will try to split it in different passes as much as I can.

What I wonder is then: When can I use those rendering parameters from the edit parameter interface? Are those parameters there to be used mainly for creating integration with other renderers? (I assume these are the kind of parameters Jason is using for his Indigo render support)

So far I have used them to add reflection and refraction limits to the glass shader in 9.1 (because they removed those properties since 9.0) and also added the soho_autoheadlight parameter to one my rops. Initially I thought those rendering parameters were similar to rendering flags, like "mantra -n 3" or something similar. But this seems not the case.

I have started rendering with raytracing and that seems to give a compromise between the micropolygon and pbr. pbr is fast, but does not pick up the color in the density. Micropolygon is fast for the environment (depth of field, mblur, displacement) but very slow for the glass object. Raytracing is fast enough and accumulates color in density, but seems slow for depth of field.

thanks again for your help.

I'm not an expert in this (Mario help!), but I think that well written glass shader should perform in both engines quite similar. There are some minor changes in settings that have to taken into account. For example in case of heavy ray traced scene, micro-polygon cache size is extremely important because lets you avoid constant recreation of geometry after every bucket is done. In raytrace mode this setting is not so important as I would suspect (I mean its ray trace equivalent). Another thing you can try is to increase bucket size for MP mode. There are dozens of such small issues that can help to speedup render times. Perhaps some Master will give you better explanation.

cheers,

sy.

Edited April 16, 2008 by SYmek

[Jason]

What I wonder is then: When can I use those rendering parameters from the edit parameter interface? Are those parameters there to be used mainly for creating integration with other renderers?

Unfortunately there is no way to know which Rendering Parameters are only applicable on a global versus object level. This has often been requested (at least by me) but right now it requires prior hard-to-get knowledge. It seems as if each of these parameters should know

SYmek is right, though: Pixel Samples is a global parameter unfortunately; we often wish for it to be possible to vary this on a per-object basis (think of objects motion-blurring through Volumes, where you want smooth motion-blur, and not oversample the Volume).