question

Willie Frazier avatar image
Willie Frazier asked

Chattering, 'noisy' specular?

Hey all,

I am having a real tough time with something that I feel shouldn't be so difficult to address, so apologies in advance if any of the frustration and/or helplessness I feel at the moment comes across in my question, but with a deadline creeping and nothing I am doing having a tangible impact, I am starting to feel it.

Ok, so here it is. I am doing a product visualization project. The product in question has buttons which are a sort of smoky chrome. The overall color is dark, but the reflections are pretty sharp (I have an actual demo unit so that I can dial in the look of the materials). Up until the end of last week, I was rendering only close up stills, and everything was looking good. I then rendered a rather lo-res 360 degree spin around animation, and again... things looked good. Then, on Friday, I rendered out a mid-res animation (1285x796). After 14.5 hours, when the render was done, I took the exr sequence in AE to evaluate, and it was then that I saw for the first time, the specular highlights on the button bevels was chattering like crazy as the product slowly rotated toward the camera. Ugh.

So I looked into it, took a dive into my AOVs and saw that they all looked great until I got to direct specular and that is where I saw the sort of jagged white line along the button bevel. Ok, cool... I now know where the problem is, and after digging through the documentation it seemed like there would be 2 ways to address it. First, increase the specular roughness (I had it at .1). I increased the value gradually until I got to .4 at which point is seemed like it might be better, but the material now had a more blurry reflection than its real life counterpart - and to be honest it seemed like there might still be a little of that jaggedness along the bevel edge, but it was being hidden more, and I am not sure that the chattering effect would have gone away in an animation.

The second thing I saw to adjust were light samples. Now, I have 8 lights in my scene, one spot, and 7 quad lights housed in soft boxes. 3 of these are fixed, ambient lights, and the others are positioned to the product some dramatic bits of light and dark areas. All of the lights were at 3 samples, so I gradually adjusted up, waiting longer for each render as I did, and no matter how high I set those values (at one point I was all the way up to a ridiculous 12 samples), the render showed no perceivable change.

So, now I don't know what to do... this is my first project with Arnold, so maybe (hopefully actually) I am missing something obvious, and I really hope there is someone here who can shed some light on what I might be doing wrong.

My current settings settings (initial animation settings in parenthesis):

- Camera AA - 7 (6)

- Diffuse - 5 (4)

- Specular - 7 (4)

- Transmission, SSS, Vol indirect all at default 2.


Here are a couple of screenshots (can't show the whole product as it has not been released yet)

Notice the area around the edges of the rectangular buttons, especially the long one furthest down in the frame: This is from the initial render, but as stated no amount of sample tweaking made any noticeable difference to the look (I took shots after ever change to observe and compare differences, and was dismayed to see none)

ls3-ss4.jpg

Now here is a shot from my boosting spec roughness to .4:

roughness-up.jpg

It does seem to diminish the severity of the issue, but I honestly think its just due to there being less contrast between the Botton's top surface and the bevel. It could be in my head, but I feel like I still see some of that jaggedness which would lead to chatter when animated.

And just to show there isn't anything particularly odd about the shape, here is a close up of the Gouraud shaded button geo:


screen-shot-2020-11-15-at-95148-pm.png

Thanks so much in advance for any suggestions.

noise
10 |600

Up to 5 attachments (including images) can be used with a maximum of 2.0 MiB each and 9.8 MiB total.

Lee Griggs avatar image
Lee Griggs answered

(Too much text for a Monday morning) ;)

- Camera AA - 7 (6)

- Diffuse - 5 (4)

- Specular - 7 (4)


What if you raise the Camera (AA) even further?

Do you need so many Diffuse/Specular samples? What are the numbers in brackets (I hope not Ray Depth)?

Do you really need all of those lights too?

I take it there is no bump on the buttons?

Another tip is to render larger and then shrink it in post to avoid issues like these.

I don't want to complicate things further seeing as this is your first Arnold project but it might be worth experimenting with Adaptive Sampling. The variance filter could help detect how many Max. Camera AA samples you need.


1 comment
10 |600

Up to 5 attachments (including images) can be used with a maximum of 2.0 MiB each and 9.8 MiB total.

Oh, and also the adaptive samples idea is interesting.


and correct, the buttons are smooth.

1 Like 1 ·
Paul Carter avatar image
Paul Carter answered

If it's just on the bevel I think it's too fine to resolve at that distance. Only the camera samples will help this on the rendering side. Another option is to alter the bevel width. Another would be to remove the bevel and use round corners.

4 comments
10 |600

Up to 5 attachments (including images) can be used with a maximum of 2.0 MiB each and 9.8 MiB total.

Yeah, it was a CAD model (first time working with one, I do usually model myself). The bevel is fine, but the overall topology makes it too big a time suck to remodel.

Given the ‘fineness at distance’ argument, I wonder if it will be resolvable. That would be catastrophic frankly... do you think rendering higher res and scaling down might help (as Lee suggested)? The scale of perspective would be the same, but if the area physically occupies more pixels, it seems it could help.

0 Likes 0 ·

A larger render would help. Adaptive sampling is now useable since the latest release, definitely worth using. For resolving difficult geometry, you need at least 10 camera samples. This is where adaptive sampling is good. You can also turn down the diffuser and specular samples with more camera samples.

Try setting camera samples to 4. Diffuse, specular, transmission and light samples to 1. Turn on adaptive sampling at set it to 20 or 30 and set the threshold to .25

This assumes you're on the latest release.

1 Like 1 ·

Sweet, I will try exactly that.

I just got Arnold about 3 weeks ago, and am running R23 SP1. I’ve not seen anything pertaining to an update, so presumably I’m latest/greatest.

Im at 1:23 am in my neck o the woods, but I will give this a try tomorrow.

0 Likes 0 ·
Show more comments
Willie Frazier avatar image
Willie Frazier answered

Hi Lee,
Just the right amount of text for a Sunday night though ;)

Seriously though, sorry for the long post, I was just trying to be thorough.

The numbers in parenthesis are the numbers I originally rendered with, the ones before those are the settings I have been increasing as I try to eliminate the the problem.

Rendering larger and scaling down is an interesting option. What do you find works best with Arnold in for that? I’m thinking 4x but maybe 2x is enough?

The client wanted moody and dramatic, so the lighting is kinda what is, but the lights are rather focused, there’s no more than 2 or 3 of them on any given area.

I did see some more understated noise in the darker areas of my indirect specular pass, which is why I increased that number. It’s still there a little bit but I’m hoping I can denoise out the traces which remain.

And yeah, I can and probably will increase camera AA more (I was told in a video lesson I watched that it wasn’t great to lean on that parameter so much as it tends to be expensive... but whatever it takes at this point.

2 comments
10 |600

Up to 5 attachments (including images) can be used with a maximum of 2.0 MiB each and 9.8 MiB total.

I would really just experiment with both Camera (AA) and resolutions and see which pays off with regards to render times. Obviously, the higher the better but then your renders will take longer.

With still images, I will often just render at a higher resolution so that I don't have to worry about sampling settings too much, and usually it pays off.

Keep us informed of your progress and results. This sort of thing is usually very scene-specific and would be useful for other people here too (including me) :)

Paul's advice about adaptive also seems sound.

1 Like 1 ·

For sure I will. This is the same model I was working on in the flake shader thread. Once I got one in person, it turns out that it wasn’t really a flake, but an super subtly almost ‘orange peel’ like bump under a snooty clear coat.

1 Like 1 ·
Willie Frazier avatar image
Willie Frazier answered

So, an update - I managed to achieve a result I liked by doubling the render resolution, and was actually able to lower the camera samples to 5. It really was just a matter of there not being sufficient pixel density in the areas in question. Of course there was some irony involved when the client (whom I sent the original render to just to evaluate it for any other notes) didn't see the issue I did when viewing the asset in context. Still waiting for a consensus from his team, but since the per-frame render time jumped from 3m/40s average to 15m/30s average. So, if I can avoid that giant render, all the better I suppose.


I couldn't experiment with adaptive sampling as my C4DtoA is 3.1.1.1, and is listed as Up To Date, so I am guessing the 3.2 version is only available in developer / tester circles.

3 comments
10 |600

Up to 5 attachments (including images) can be used with a maximum of 2.0 MiB each and 9.8 MiB total.

I use 3ds Max, but doesn't the option Lock Sampling Pattern helps reducing that flickering? Also, you may want to see if scale the render in something like Fusion or Nuke would work. You can use a [Log to Linear], [Reformat] and [Linear to Log]. I usually render 2k and the scale it up sometimes to a 4k. That perhaps could be a solution for your situation.

1 Like 1 ·

The 3.2 version is available to everyone. Go to https://www.arnoldrenderer.com/arnold/download/

0 Likes 0 ·

"C4DtoA 3.1.1.1 is listed as Up To Date"

Something's definitely not right with querying the latest version in R23. I tested in R21 and works there, but not in R23. I thought I fixed that already, but seems there's still something's wrong. I'll take a look.

0 Likes 0 ·

Write an Answer

Hint: Notify or tag a user in this post by typing @username.

Up to 5 attachments (including images) can be used with a maximum of 2.0 MiB each and 9.8 MiB total.

Welcome to the Arnold Answers community.

This is the place for Arnold renderer users everywhere to ask and answer rendering questions, and share knowledge about using Arnold, Arnold plugins, workflows and developing tools with Arnold.

If you are a new user to Arnold Answers, please first check out our FAQ and User Guide for more information.

When posting questions, please be sure to select the appropriate Space for your Arnold plugin and include the plugin version you are using.

Please include images, scene and log files whenever possible as this helps the community answer your questions.

Instructions for generating full verbosity log files are available for MtoA, MaxtoA, C4DtoA, HtoA, KtoA, and Kick.

If you are looking for Arnold Documentation and Support please visit the Arnold Support site.

To try Arnold please visit the Arnold Trial page.

Bottom No panel present for this section.