Noise in the reflection table: bug?

[Ludvik Koutny]

If your light sources are large (like windows) then caustics will not converge that slow. Try to set Max sample intensity to about 20, enable shadows in your glass material, and render close up of the chair shadow. I think the shadow will not be opaque, but will converge fast ;) The extreme clamping scenario usually happens only in cases with very small and bright light sources in a very dark environments ;)

Max. sample intensity (MSI from now on) does not make shadows opaque, or disable caustics. It just throws away samples that are significantly different than the average of samples taken from the given point. So if you have very large light, like window, where it will be very easy to calculate caustics even using regular path tracing, then not many samples will have significantly different value than the rest, so not much data will by thrown away and you should get transparent shadows with soft caustics. ;)

[cecofuli]

Again, noise in DOF, like in my old child bedroom. Also, fireflies on transparent chairs.
About DOF, I think I cannot do nothing. 8 hours and 450 passes are too much. I think Keymaster must to do something =(
For fireflies, if I disable the light, everything is ok. But, in the light option, I don't have "disable in reflection/refraction"
Corona parameters as default ( PTS= 32 , MSI 20 )
Some suggestion?  I cannot send to my client rendering with this noise...

[Ondra]

PT samples are 16 by default, right? And you should lower it if you are using DOF, to about 4

[cecofuli]

No, 32, as in the interactive Guide. But I try with 4 in a crop. What about the white fireflies?

[Ondra]

you can try for example disabling the visibility of the light, or lowering max depth

[cecofuli]

Better with PT= 4. Max depth = Ray Depth? Now is 25, by default. Usually in VRay is 5. Or am I wrong?

[cecofuli]

The only way to "solve" (It's better say reduce)  the fireflies is changing "Internal res" from 2 to 4.

[cecofuli]

Or use this free tools. But, obviously, it adds "blur" to the  rendering.

Defly - firefly and hot pixel noise removal tool

[maru]

Or render in huge resolution with "do antialiasing" disabled and then downsample in photoshop.

[Polymax]

Also, fireflies on transparent chairs.

This is Alpha-4 bug! In the last builds is fixed!

[Ondra]

It does not necessarily needs to be. If there is a complicated glass structure, then since on every surface you have both reflection and refraction, the number of possible paths grows exponentially with ray depth, and you get something with fractal behaviour. If you the have thousands of possible ray directions for single pixel, the probability that some of them hit the sun randomly grows, and so you get frequent bright spots. This is physically correct behaviour, and Corona should converge to the physically correct result in the long run, although it is certainly undesirable for this kind of renders.

[Polymax]

Here are the fireflies on transparent objects (post from mantis).
Or is it not?

[Ondra]

which ID is this, this looks like good old NaNs

[Polymax]

0000231

[cecofuli]

With the new version, noise and fireflies. But after 9 hours, the noise is too much.
Any suggestion?