[ISSUE] - Probably bug in the QMC sampler [solved somehow]

[psor]

[Stromberg]

The material did get a checkerboard in the reflections?
Just wondering if i did spot the right bug

[psor]

No the inside is also brown'ish not green'ish! *LOOOL*
Ya next time I'll point it out better.



take care
psor

[Stromberg]

Cant quite see why it should be green tough
But i'l have to wonder why is there a checkerboard reflection in the first render and not in the second?

[psor]

Cant quite see why it should be green tough

*LOL* ... but it is, very slightly - I've to admit.

But i'l have to wonder why is there a checkerboard reflection in the first render and not in the second?

That's the problem you see.



take care
psor

[Stromberg]

Thats what i tought, but then you said the bug was the color on the ceramic bowl... i'm kinda confused here
Is the checkerboard the bug or the wrong color, or both?

[OnoSendai]

Try rendering for longer and see if the artefacts disappear.

[psor]

Erm, how about I send you the scene and
you do it by yourself. I'm on a slow P4 right
now and already did let it render for about
17 hours and to ~15000 s/px.

@Stromberg
It's about the artefacts on the outside of
the tee cup not the color.




take care
psor

[benn]

@psor - email us the scene and we'll try it out.

From our analysis there was a possibility that with the sampling code in 2.2, that noise would go down - then eventually start going up again.

If that is the case then we'll have to look into it.

[psor]

I'm back and have good news.

Last night I did a quick test with the scene. I did a region rendering of the region that
was showing the artifacts. And what I did recognized after they did show up after 7h
that they where way smaller then before. I guess because of the region rendering and
thus the smaller region the sampler had to take care of. Then I did let it render even
longer. It's now at ~10000 s/px after 22h and the artifacts are gone.

And because I do recall that Maxwell Render had a similar issue a long time ago,
I ask you to change the sampler a tiny bit again. It's - at least for me - very bad
to have to wait for a long, long time, just to get rid of the QMC artifacts. I hope
you can understand this. I'm hopefully not the only person that did fall into this
rabbit hole. Well, with that being said, I hope to see a small update in the next
version. So yes Nik, to let it render longer is a workaround but not a good
solution anyway.

Thanks for listening and for your awesome work guys!

edit: I've uploaded an animated .png (3.74MB) so you can see the progress
on the noise and artifacts of the sampler. It's a render history that represents 90
pictures taken every 15 minutes then cropped and resized to make it more obvious.

And yes I know it's not a bug. Anyway it's uncomfortable.

edit: I've added the scene so it's easier for you to test the sampler.

edit: It's not the sampler, it's the integrator. Bidirectional PT strikes again, as usual.





take care
psor

[Jambert]

Up, bug still not fix , Ono, Benn, any news about it?
I use env map / Bidir, 2000 spp

thanks in advance