Nasty Artefacts with Architectural Materials

[Caveman]

@ schyzomaniac.

Thanks for the suggestion. As I said on the post above yours, the cutoff is not helping at all in removing the noises completely. All it does is lighten them up, which for me is not sufficient

[Caveman]

Hey Maximus whats up with all those frogs. I thought I gave a teapot, didn't I?

Wow! wow! wow! did I really mess up that bad

What problems do you see in the BRDF+Backface culling + trace depth cutoff + transparency, gamma, photo exposure?

According to the mia guide__ "One is however not forced to base the reflectivity on the IOR but can instead use the BRDF mode to set it manually".
And they have used BRDF along with IOR on Transparent Glasses. So, when you said "for a physical glass it should be set to "By IOR". I think it is not correct.

Backface culling is an effect which one may or may not want. So if the Noises are attached to the Backface Culling and we have to turn on Backface Culling on every transparent objects even if we do not want the "backface culling" look, just to remove the noises that it is not good...

Decreasing the Trace Depth cutoff value does nothing but lighten up the noises. So that's not a solution either. I have done some renders with the settings lowered, check them out, above^^

What about Transparency???

I lowered the Gamma because I was getting washed out colors since I was working in a proper linear space so didn't wanted that.

[Maximus]

I'm sorry i've must explained myself bad, the problem of those artifacts is not related to backface culling, even if this is a dirty way to cheat and remove it sometimes, but it will change the look of your render, so nevermind the backface culling.

The tracedepth cutoff helps because it cutoff some rays that will cause that bad feedback but again, those arent the real problem here.

The real problem is an incorrect BRDF value, the fact that you have a slider to play with doesnt mean you will achieve correct results by just putting random numbers in it.
So again, to have a physically accurate glass, the brdf should be set to "Default (IOF)"
The mia_materials or architectural is built for physically accuracy, if you want to play with parameters for artistic look use another shader, if you want to use this shader to achieve materials you should use it like the creator meant (Zap Andersson).

In my last scene i've posted (TeaNewOk.rar) my render doesnt have any artifact, the glass look like a glass, and the color looks just fine, i dont understand what other problem you got?
That render had BRDF set to "Default (IOF)" Tracedepth cutoff at 0 and thats it.

Are you doing tests for your own knowledge or you have a look you want to achieve? because first we should understand what you want to do and then we can discuss on how to achieve it.

Your BRDF 0.01 gives an incorrect angle of incidence wich i've marked here:

http://img852.imageshack.us/i/brdf1.jpg/

See that tiny white line i've added? that is completely wrong, you cut off all the reflections in the middle part going with an ellptical shape, while you should have the greenish floor reflected.
This kind of reflections is unrealistic and not natural, its just wrong. The reflections of the plane into your teapot should follow the teapot shape, like it happens in real life, this is how it should look:

http://img35.imageshack.us/i/brdf2.jpg/

Another thing, remove the ambient occlusion wich can cause artifacts due to low sampling, and create the shader step by step, you had AO turned on, if you want it, add later with higher sampling for definitive render.

I've double checked my scene and it just renders fine, did you take a look at the scene at all? Not the first one i've posted, but the second one called TeaNewOk.rar
this one download/file.php?id=993

I dont understand whats going on in your last render except for the fact that the BRDF is incorrect.

/cheers

Max

[Caveman]

Hey, man thanks for your support/discussion. The question is, if they have given us manual control over the BRDF, and it works flawlessly "Physically Correct or Not" in another package, why isn't it working in Softimage?

And of course, I checked your second file. And it renders fine. But can't use the BRDF without getting those nastiness. It is clearly mentioned in the mia material pdf from Mental Images that we are not forced to base the reflectivity on the IOR but can instead use the BRDF mode to set it manually. I have attached the image showing this example. All of these are rendered with different BRDF values, set with manual control.

Like Zap said "Transparency takes energy from Diffuse, i.e. at 100% transparency, there will be no diffuse at all. But even with 100% transparency I get a mixture of the diffuse color and the transparency color, both in Softimage and Maya. I have to turn down the diffuse weight to 0 to get full transparency. Do you know why? These are the things that are bugging me. And thanks for listening.

Btw I can't see none of your images from Imageshack. I can only see the frog that replaced your image.

[Maximus]

I've changed the link for the images o.O
weird they didnt show up, i see them! tell me if you still cant see i'll try something else.

Anyway back to topic, you should read other parts of the Architectural PDF, i'm gonna copy/paste from there:

Page 8 from the PDF:

"2.3.3 BRDF - how Reflectivity Depends on Angle
In the real world, the reflectivity of a surface is often view angle dependent. A fancy term for
this is BRDF (Bidirectional Reflectance Distribution Function), i.e. a way to define how much
a material reflects when seen from various angles.

Many materials exhibit this behaviour. Glass, water and other dielectric materials with fresnel
effects (where the angular dependency is guided strictly by the Index of Refraction [this means turn the option "Default (IOF), by doing this your BRDF will depend on your Index Of Refraction]) are the
most obvious examples, but other layered materials such as lacquered wood, plastic, etc.
display similar characteristics.

The mia material allows this effect both to be defined by the Index of Refraction, and also
allows an explicit setting for the two reflectivity values for:
• 0 degree faces (surfaces directly facing the camera)
• 90 degree faces (surfaces 90 degrees to the camera)
See the BRFD section on page 24 for more details.

Another:

Page 19 from the PDF:

2.4.3 Refractions

The transparency parameter defines the level of refractions and refr color defines the color.
While this color can be used to create “colored glass”, there is a slightly more accurate method
to do this described on page 37. <-- we'll get there later
Due to the materials energy conserving nature (see page 7) the value set in the transparency
parameter is the maximum value - the actual value depends on the reflectivity as well as the
BRDF curve.
The refr ior defines the Index of Refraction, which is a measurement of how much a ray
of light “bends” when entering a material. Which direction light bends depends on if it is
entering or exiting the object. The mia material use the direction of the surface normal as
the primary cue for figuring out whether it is entering or exiting. It is therefore important to
model transparent refractive objects with the surface normal pointing in the proper direction.
The IOR can also be used to define the BRDF curve, which is what happens in the class of
transparent materials known as “dielectric” materials, and is illustrated here:

(here there is an image, check it out in pdf so i can avoid to post!)

and again:

One is however not forced to base the reflectivity on the IOR but can instead use the BRDF
mode to set it manually:
here there is your image



Different types of transparency
The left cup again acquires it’s curve from the index of refraction. The center cup has
a manually defined curve, which has been set to a brdf 90 degree refl of 1.0 and a
brdf 0 degree refl of 0.2, which looks a bit more like metallized glass. The rightmost cup
uses the same BRDF curve, but instead is set to “thin walled” transparency (see page 10).
Clearly, this method is the better way to make “non-refractive” objects compared to simply
setting refr ior) to 1.0 as we tried above.

Again, having brdf at 0.1 or 0.2 in my scene, with cutoff at 0, it just render fine to me, but lets continue

Now comes the most important part
Page 24 from the pdf:

2.4.6 BRDF
As explained in the introduction on page 8 the materials reflectivity is ultimately guided by
the incident angle from which it is viewed.

There are two modes to define this BRDF curve:

The first mode is “by IOR”, i.e. when brdf fresnel is on. How the reflectivity depends on
the angle is then solely guided by the materials IOR. This is known as fresnel reflections and
is the behaviour of most dielectric materials such as water, glass, etc.

The second mode is the manual mode, when brdf fresnel is off. In this mode the
brdf 0 degree refl parameter defines the reflectivity for surfaces directly facing the viewer
(or incident ray), and brdf 90 degree refl defines the reflectivity of surfaces perpendicular
to the viewer. The brdf curve parameter defines the falloff of this curve.
This mode is used for most hybrid materials or for metals. Most material exhibit strong
reflections at grazing angles and hence the brdf 90 degree refl parameter can generally be
kept at 1.0 (and using the reflectivity parameter to guide the overall reflectivity instead).
Metals tend to be fairly uniformly reflective and the brdf 0 degree refl value is high (0.8
to 1.0) but many other layered materials, such as linoleum, lacquered wood, etc. has lower
brdf 0 degree refl values (0.1 - 0.3).

And last:

Page 36 from PDF:

2.5.2 Quick Guide to some Common Materials

Here are some quick rules-of-thumb for creating various materials. They each assume basic
default settings as a starting point.

2.5.2.1 General Rules of Thumb for Glossy Wood, Flooring, etc.

This is the kind of “hybrid” materials one run into in many architectural renderings; lacquered
wood, linoleum, etc.

For these materials brdf fresnel should be off (i.e. we define a custom BRDF curve). Start
out with brdf 0 degree refl of 0.2 amd brdf 90 degree refl of 1.0 and apply some suitable
texture map to the diffuse. Set reflectivity around 0.5 to 0.8. <---this is the case where you use "manual mode"

And now Refractive (transparent) materials:

Page 37 from pdf:

2.5.2.4 Glass

Glass is a dielectric, so brdf fresnel should definitely be on (they mean to check the "Default (IOF)" option so you will use the IOR you set before to control the BRDF, so you must not use "manual mode" for Glasses. The IOR of glass is around 1.5.
Set diffuse weight to 0.0, reflectivity to 1.0 and transparency to 1.0. This is enough to
create basic, completely clear refractive glass.

If this glass is for a window pane, set thin walled to on. If this is a solid glass
block, set thin walled to off and consider if caustics are necessary or not, and set
do refractive caustics accordingly.
Is the glass frosted? Set refr gloss to a suitable value. Tune the refr samples for good
quality or use refr interpolate) for performance.

It goes on explaining some tips for colored glass, wich is basically the fallof color and distance, go ahead if you want.

Anyway, i've tried doing a glass with 3ds max and mental ray, and it happens the same artifacts if you use a BRDF of 0.01 (wich have really not any sense, this value you used is totally wrong, it is too small and i dont get why you used it, if you go with 0.1 it works)
Anyway 3ds max had the same problem 0.01 brdf set to manual and a glory of artifacts came out, so its not Softimage related problem, and i dont have maya here but i'm pretty sure it does the same thing, i will check it anyway

BRDF set to manual should be used to create materials like wood, metals, ceramic, stone, and whatever, because those materials can have different BRDF incidence angle and you can use manual mode to get the right look or personal look.
However when it comes to glasses, they need to respect physical rules, Index Of Refractions, and BRDF should be sete to "Default IOF" you shouldnt go wild with it

To answer this:
"Like Zap said "Transparency takes energy from Diffuse, i.e. at 100% transparency, there will be no diffuse at all. But even with 100% transparency I get a mixture of the diffuse color and the transparency color, both in Softimage and Maya. I have to turn down the diffuse weight to 0 to get full transparency. Do you know why? These are the things that are bugging me. And thanks for listening."

I am not Zap, but i guess the reason is that there is always a small part of energy stored and conserved into diffuse even at 100% transparency, that is the reason why you need to turn it completely off by using the "Diffuse Weight" parameter, wich control not the color of the diffuse, but the emission the diffuse has, hence if you put it at 0 you will have 0 color diffusion.

I hope that helps, actually architectural material is a beast, and probably the best material ever made between all 3d render engines once you get used to it you can achieve every look you want in small fraction of time!

/cheers

Max

[Mathaeus]

well, with all due respect to guy that made this shader, render engine, whatever...
I'd believe there is just something wrong in design of this thing. As far as I can recall, 'Fresnel' reflections became popular in 3d, sometime after the 'Titanic' movie (1997 or something), and even with free raytracers of this time, like POVRay, we were able to render 'sharp' reflections/refractions, together with that nice falloff... Before, people were rendering just 'flat' reflections everywhere
Now, what causes these artifacts, this can be.... everything, maybe some not-so-happy optimization, maybe some internal mechanism to avoid NaNs, maybe...
I really won't go with this 'physically-accurate-is-only-approach'. AFAIK, both methods rely on 'some math' around angle between camera ray and surface normal. Too much simple for glorification, for my taste.
Someone needs to fix that, that's all. And that 'someone' isn't any of posters in this thread.

[Caveman]

Thanks for the post Maximus. I can see the link to images now and I have checked them out.

Atlast I have solved the issue completely(if you can call that an issue). Even with unrealistic Manual BRDF settings, like for instance 0.01 and Refraction Cutoff set to the default value of 0.0,1 I am not getting those noises anymore. I checked all the possible causes for those noises, and narrowed it down to the Transparency Color. The Transparency color needs to be white, if any color is used in the Use End color. And Use End Color needs to be turned off if we are using the Falloff with any kind of Transparency Color.

Silky Smooth Render
Case Solved

[Caveman]

Hi! guys need to post this. It seems that my previous post about solving the issue is not completely correct. The Transparency color can be used too, without having any kind of noises. The trick is to tweak the Reflectivity, Manual BRDF (if you are using it) and Falloff Distance until you have eliminated the issue.
In Maya, mia material doesn't have any noises regardless of what values you have put in these fields. I wonder why it behaves like this in Softimage. I hope that Metal Ray 3.9 solves this problem. ;)

[origin]

man 3 page topic about teapot
I just checked, 3ds max is behaving the same
so I believe maya will do also same "bugs" since they share same dll libraries
maybe scene scale or whatever is diffrent in those softwares and this is affecting cutoff value

[Caveman]

man 3 page topic about teapot
I just checked, 3ds max is behaving the same
so I believe maya will do also same "bugs" since they share same dll libraries
maybe scene scale or whatever is diffrent in those softwares and this is affecting cutoff value

Actually, this topic was never about "teapot". It is about a possible bug which I have encountered in Softimage. And I didn't wanted to conclude the thread with the wrong information that I had posted earlier. So, I thought I need to correct it to some extent, so that users with same issues who might come across this thread while searching for a solution, may find this thread a bit helpful. You dig And, I don't see any rule in this forum, stating that there shouldn't be a topic with 3 pages in it, so I guess I am OK.

[Hirazi Blue]

No, but there is a rule, that any thread featuring teapots should at least be four pages long...
(a lesser well-known obligation of the "Sarbanes-Oxley Act")

[nas]

What i also find strange in 2011 onwards are very fine, almost pointelized dots in the default arch mat. Since working in linear workflow this has appeared. I i turn off all the tickboxes( Preferences>Tools>Display>Colour Management>Apply to) it goes away. However i'd like to keep working in linear workflow for all it's other benefits...

I made the material black to illustrate the problem further.