Are there known issues regarding frame-to-frame flickering when using multiple meshlights?

[Anonymous]

We have an outdoor scene using 50 meshlights (50 unique objects, 50 mesh_lights).

In a 15frm test, 2-3 frames (sometimes more) will show surfaces receiving light from the meshlight objects/lights will be darker. On subsequent renders, which actual frames this occurs on will randomly vary. After trying all sorts of things, we decided to condense all 50 unique objects into 1 object and use just 1 mesh_light. Result: No flickering frames.

Has anybody else noticed this behaviour?

We're using 5.1.0.1. Could this issue have perhaps been fixed in a later version?

Thank you,

[lee_griggs]

Does increasing the number of mesh light samples help?

[Anonymous]

Hi Lee,

We tried that too and it did not help.

Thank you,

[lee_griggs]

How were the objects modeled? Are they water tight? Normals all ok? etc

[Anonymous]

They were model in Maya, normals are good. Don't know what you mean by "water tight".

The light-emitting surface of the meshlight itself is a simple light bulb polymesh.

The light-receiving surface (the one that's flickering) is the inside of a light 'hood' so to speak,

like a cord lamp hanging from a ceiling,etc...which is cone-like and has modeled thickness between

the outside and inside of the lamp. All in all the geometry is pretty simple. This is not the first time

we've seen flickering like this and despite analyzing here and reading post on this site, have never

been explained why this occurs.

Thank you very much for looking into this and I hope you find a solution as it's really a drag to deal

with since it's random.

[Stephen.Blair]

"water tight" ... are the meshes totally closed, with no gaps or holes where light could "leak" out ?

[Anonymous]

Ahh, ic, a closed surface...

I just loaded this into Maya as I was not the modeler who built it.

Indeed the surfaces are *not* closed...the light-receiving hood has a thickness but it's not closed at the top.

The light-emitting bulb is made of three surfaces who are indeed are *not* closed either.

I'm guessing that could cause issues?

Thank you,

[Anonymous]

Thoughts?

[Anonymous]

Just to add...

So we did a test where we combined all 50 individual objects into one objects and used only one meshlight and that had an enormous beneficial effect...but it did not entirely eliminate some very minute flickering from frame to frame. We might get by with it but client might also object.These were the same open-surface objects as before. I'm planning on creating a new test object that only uses closed-surface primitives spheres and cones and see if that has any effect.

Thank you,

[Anonymous]

So a couple of updates on this...

Instead of 50 objs and 50 meshlights, we combined all of the objects into one object and use one meshlight. Results are vastly better. However, there is still some small amount of flickering, mostly noticeable in distant objects. I'm told when projected on the big screen, it's objectionable. However, combining the objects and using one meshlight is *definitely* the way to go. Btw, these objects are the original 'open' surface objects. 1) of 2)

[Anonymous]

So we created a custom test object, using Maya spheres and pipe primitives (pipes are reshaped into light cones/shades,etc). This created a more uniform image to evaluate. These were all combined into a single object mesh and again using just one meshlight. However, we still see this very small but objectionable flickering. This is creating a very urgent situation for us. We'd very much appreciate your help solving this. Will try to attach the assfile from this test. Thank you.

[Anonymous]

meshlight-flicker-example-rh.tar.gzimage-example.jpg

Attached is an assfile example of this latest test using spheres & pipes,etc.

At the present time, we normally use v5.1.0.1 but these last few tests were with 5.4.0.0 and this assfile was generated using 5.4.0.0.

Also included a simple jpg to get a feel for the content.

Thank you very much,

[lee_griggs]

Why are you using mesh lights and not spot lights for this?

[Anonymous]

There are spotlights in the scene to create the actual light cones that one might expect. What you have is a simplified part of the entire scene and what are test objects. The real scene is far more complex and has many light types. Using meshlights in this scene context allows the lighters to get around fixed limit of number of allowable lights in a single assfile render.

Thank you,

[Anonymous]

Hi Lee, do you guys have any updates on why meshlights are flickering in 5.4.0.0?

Thank you,

[madsd]

Jim,

I was not able to make the ass file flicker.
I instanced it 50 times in the scene and made a solid block of white lamps.

No blinking, I do notice the lack of .exr files, I wonder if you are losing a texture connection now and then and that its completly random, I cant say for sure, since I dont sit on your machine.

[Anonymous]

Hi Mads, just to help clarify...

Earlier 'flickering' did accurately describe what we were seeing.

After combining all the objects into one object and using one meshlight, it not an on/off thing,

not a blinking thing but a very slight dimming on random frames.

If it were a missing texture, one would think we would see that in the queue logfile which we don't.

This still appears to be an Arnold bug.

Thank you,

[Anonymous]

Hi again Mads,

Referencing the actual question above...are you saying there are no known issues regarding meshlight flicker in general?

Thank you,

[madsd]

I don't know.

http://masd.dk/temp/gif/qwe.gif

Here, I see nice expected uniform noise on the ground as frames are ended premature, there are 1000 mesh lamps.
I have a hard time seeing any dimming down, maybe I need to render some series and do a difference pass on them and scale it up so even tiny changes gets magnified, is this the approach needed to observe the pulsating?

[thiago.ize]

How random is this exactly? I understand that when you render your 15 frame test 2-3 frames are flickering, but if you re-render will the flicker be on the same frames? If the answer is no, then something is definitely going wrong because Arnold should render identical images each time, warts and all -- can you provide two images of the same frame one with and one without the flicker? If on the other hand the answer is yes, then can you export two .ass files, one with the flicker and one without (ideally an adjacent frame) so we can see what is going on? The single frame you provided us isn't enough to detect a flicker.