Need help: crawling/sparkling reflections

[Dman3d]

Having problems with reflections "crawling" or "sparkling" I have tried everything I can think of but need an assist. Probably something simple Im overlooking, but my brain is goo at this time.

Rendering from IRMap and LC

10 Frame excerpt, look in the hilighted area
http://www.dman3d.net/Temp/ref_sample.mov

Set up:
LWF - Gamma method

Material:
http://www.dman3d.net/Temp/Counter_Mat.jpg


Render settings:
http://www.dman3d.net/Temp/VRay_set-01.jpg

and

http://www.dman3d.net/Temp/VRay_set-02.jpg

[studioDIM]

if it has to do with QMC sampled material effects, try simply lowering the QMC threshold in the QMC sampler rollout.

set it to something silly like 0.0005 and testerender.
If the ants go away, you know the path to follow.

Lele

[Dman3d]

Using .85 amount and noise threshold to .003, will try lower

[studioDIM]

then it's most likely due to that.
0.003 doesn't work well at all for animation in my own experience.
Need it definitely lower, so the sampler will move towards higher sampling rates.

Lele

[Dman3d]

Will do! Thanks, note also that it is isolated, not throughout the rendering.

[Dman3d]

Well this did not seem to work, as far as I can tell there was very little change. Any other ideas? I will check smoothing too. Maybe lower amount and thresh?

[studioDIM]

for it to work you will ALSO need a high enough AA max subdivision level and /or a high enough subdivision level for glossies.
I left it implicit.
Try Vlado's universal settings (QMC 1-100, no AA filter, threshold around 0.00025-0.0001).
If the noise is still there, the reason is not glossies.
And it may lay in bad geometry or GI sampling gone wrong.

If you could post an image, though, it'd be helpful

Lele

[Dman3d]

errr... the first post includes a 10 frame segment showing the problem, is the link not working? I will try pumping the samples for the mat too, but other surfaces with roughly the same values seem to be behaving themselves, will try applying same mat to a different object see what happens.

[studioDIM]

nope, not working here:404.
sorry if i speak by assumptions, i'm a bit all places today

Lele

[Dman3d]

Whoop, bit of a typ-o there, caps/not caps, got it fixed. The QMC solution seems to address it, but of course the render times become much higher.
Any suggestion on how to get around using the "universal settings?" I tried making the reflection slightly more blurry but that did not quite clean it up and of course is not correct. Think of Adaptive QMC but something more like 1 and 8, think that might be enough? With .85 and .0003 or so? I will also try creeping up the material samples too, kinda low but, other blurrier materials are fine with fewer samples, but better to sample the crap out of the one mat rather than the whole scene.

But Also using gamma so I will also switch to Linear only per Vlados suggestion (or point) here, wonder if I can get away with increasing the mat samples without resorting the universal settings?

Cant test it right now, not at work at the moment.

http://www.chaosgroup.com/forum/phpB...ic.php?t=14747

studioDIM wrote:
I haven't had time to test it yet, but could you please confirm, Vlado, that glossy based speculars will resolve properly even by leaving the settings to their defaults on the materials?

Yep, glossies will be fine and correct. However, this also means that because of the unclamped colors and lack of color mapping very thin hilights may appear somewhat jagged. There is no way around this unfortunately, as it is the correct result.

Best regards,
Vlado

Thanks for the assist, cant wait to try the ideas out tomorrow

[DaForce]

Try (just for testing) not saving the glossies with the LC.

And since you will be testing this stuff out ... make sure to use render region so you dont have to render whole frame.. save you some time

[jacksc02]

uncheck use LC for glossy rays.

[studioDIM]

doh, i never thought of that one (lc for glossies).
But wouldn't that affect all the other surfaces (which look "good enough") as well?

Lele

[DaForce]

it can only get better

but yeah it would

[Dman3d]

Yeah Im only sending 10 frames I suppose 2 would be fine so I can compare the two, still shot is not the issue, only when its moving.
Gonna start trying this now. First I will go back to the beginning and uncheck glossy rays which makes it faster. Moral of the story dont use a tiny detailed profiles