For better Arnold - Chapter 1/50: bad texture sampling

[Sunny_Suppy]

First of all, let me say that I'm a fan of Arnold because it used to bring me joy in the past. That's why I spend time to do this things right now even though I don't use it in my production because its lackings and weakness. Days by days, months by months, years by years, I constantly express my opinion to the people who do RnD and creating Arnold on tweeter. But I don't understand why they never listen to me. Even though it's too obvious that my opinions are only because I wish the good things to Arnold and I believe your longtime big customers want the same.

So now I decided to do a series about 50 important flaws of Arnold in here and I will try to update gradually. I also don't understand why I have to do this when I already have better renderers but I really hope that the developers will take this matter seriously and responsibly with a product that has been considered the best in the market in the past. I hope my effort will not be ignored. Thanks!

* And now this is the first part of 50 posts - bad texture sampling

- Arnold sample very bad texture at the surface that tends to tilt or be perpendicular to the camera

You need to look at the .ERX file for better detail look. And this contains project + alembic also :

https://drive.google.com/file/d/1ksJNBEt5vk2TRgbv2LdKQ61y35FFMDS3/view?usp=sharing

[CiroCardoso3v]

Find this extremely interesting. I look forward for the next posts.

I think would be interesting to see the shader for the leaf. Any chance you can export a FBX so I could test this in 3ds Max?

[Sunny_Suppy]

Thank you. I just update the chap2/50 .

About the model, I send you link of alembic:

https://drive.google.com/file/d/1X_O1R3iEbwHodjGxP59JVgrhdP2qP95F/view?usp=sharing

[thiago.ize]

I'm not sure what the issue is with anisotropy that you're reporting. Is it that the same settings in c4dtoa produce different results in corona/3delight vs arnold? Or that arnold is unable to generate anisotropy? By adjusting the specular anisotropy setting in the Leaf shader, I can get results that match corona/3delight. Here's my render where I made the anisotropy=0.

[thiago.ize]

For the veins issue, I switched your render from using blackman harris filter to gaussian and that sharpened it up. Here's gaussian plus specular anisotropy=0.

[Sunny_Suppy]

You really cannot see the different ? This is what I want to say:

[Sunny_Suppy]

There was a mistake when I went through the old file from version 6 and forget to check the parameter. Thanks for check it for me. I comfirm that anisotropy works good but texture sampling still fail.

[Sunny_Suppy]

Whatever sampling method you use Black Harris or Gaussian, and whatever level of shapern/blur, the results get same issue.

For those cannot see the difference due to low quality compressed image :

[thiago.ize]

If you want sharper textures, https://docs.arnoldrenderer.com/display/A5AFCUG/Sample+Filtering points out you can try raising the texture_max_sharpen. I don't know if c4dtoa exposes this directly, but you can still do it by typing the following into the "user options".

A small increase usually does the trick (I found raising it from the default 1.5 to 3 worked here). It's set to a low value so that textures don't consume too much memory.

[Sunny_Suppy]

Thanks so much to find out. That works. Setting options.texture_max_sharpen to 4 make me satisfied. Anyway this is very important for detail look, but also a secret. Because it's hard to know that it exists in diagnostic section when I cannot access to https://trac.solidangle.com/arnoldpedia/wiki/texture_max_sharpen to read it. Ok this problem is solved.

[thiago.ize]

We've updated the documentation so that it now links to the proper page.

[marcus.jungerstam]

This is interesting.

Is this with mip-mapping in all 3 engines? Or is mip-mapping turned off?

looks like you are getting a low res image. You could try to force Arnold to pick a more high res tile.

[CiroCardoso3v]

Corona doesn't support Mip-mapping.