LuxCoreRender Forums

Dade wrote: ↑Sat Nov 24, 2018 10:56 am Could you use a condition based on "samples per pixel" statistic ? It is exactly the same problem of BCD denoiser warmup up period (and the same solution could be used here too).

Done: https://github.com/LuxCoreRender/BlendL ... 991fca133d

Dade

I'm following your steps and trying to reproduce the problem here but it works all the times

It's been happening quite often this end.

I try to use very little clamping (maximum values up to 1) and prefer to turn the light values down, rather than clamp.

I've noticed that the closer I get to 0, the more often the issue appears to occur. Maybe I'm creating negative clamp values, if that 's possible?

Done: https://github.com/LuxCoreRender/BlendL ... 991fca133d

Can I try this. Is it just a case of copy paste the code to somewhere?

Thanks

See this thread: viewtopic.php?f=5&t=666&start=20#p7061
You'll need the latest LuxCore binaries: viewtopic.php?f=9&t=736

B.Y.O.B

Thank you

Hi

I've been trying out beta 3 and beta 2 for the Suggested Clamp values.

If the amount of samples required to set the suggested value is 2, is this enough?

With beta 2 I notice that if I let the render run for about 5 to 7 samples then I tend to get a slightly lower Suggested Value than if I let it run for just 2 (could it be finding more light with the more samples?).

The result of the slightly higher value with Beta 3 or just 2 samples tends to allow for a few fireflies.

It could just be that my scene has become a bit messy and is causing a problem within itself, but I thought I should just mention it.

Thanks

....as a work around for my last post, just adjusting the light group values enables reducing the fireflies; is this approach the right one?
Thanks

Hi Racleborg

I would not use light groups and changing their gain for the purpose of fireflies extinction

I usually reduce image resolution or view port, latter is more comfortable, and adjust clamping manually. Depending on sampler type it takes several seconds to get an idea how final result looks like with current clamping.
Often image changes its brightness, even if if auto brightness is disabled, until result stabilizes. But then if you like the result settings are close to a good choice.
In manner of samples spoken, wait until you have 20 samples, better more.
I use 1920x1080 at 20% what leads to fast renders because its area is 1/5*1/5=1/25 less than HD what means render is about 25 times faster on CPU. openCL adds compiling time, this again is often shorter for small images too.

FarbigeWelt

Thanks for all the very helpful info.

I like the idea of ' 20% 1920x1080 render and set clamp manually'. I'll try that out.

Cheers

By the way, you can enter multplications or divisions in entry fields like 1000/2 or 2000*2 if you don’t want calculate yourself.