1.47 migration to 1.5rc rQMC noise threshold issues

[res3d]

If I had my noise threshold set to .001 before in 1.47 what sould it be in 1.5?
my renderin time at .001 are enourmous and the quality when I go to 0.01 is not as good. Any thing in bwteeen just takes toooo long.

Help!!!

[panthon]

I am also interested in that, since i dont seem to find an acceptable group of settings while keeping rendertimes under control. Either i get lots of noise or gigantic rendertimes...Im using adaptive qmc more and more, but i found adaptive subdivision more reliable.

[studioDIM]

I found 0.01 works in most situations, here, when used in unclamped mode.
I normally raise max settings, and the sampler does step through them (as it was generally not the case before, hence the lowering of threshold values).
To aid in the less contrasted situations, i normally lower threshold by 0.001 at a time, generally never going below 0.005, with 0.007 being the "high quality" setting.

My 2cents, though, not law

Lele

[res3d]

render time in 1.5 and Antialiasing

That does help.
I was using Mitchel-Netravali for my anti aliasing filter in 1.47 with good reults. Now when I use it the rneder times quardruples so I had to use area set to 1.5. Using area I don't get as good results.

Any suggestions?

[panthon]

I found 0.01 works in most situations, here, when used in unclamped mode.
I normally raise max settings, and the sampler does step through them (as it was generally not the case before, hence the lowering of threshold values).
To aid in the less contrasted situations, i normally lower threshold by 0.001 at a time, generally never going below 0.005, with 0.007 being the "high quality" setting.

My 2cents, though, not law

Lele

That's the conclusion im reaching too. So u check "use QMC sampler threshold" and tweak the rqmc sampler settings?

[studioDIM]

That's the conclusion im reaching too. So u check "use QMC sampler threshold" and tweak the rqmc sampler settings?

That depends.
If i am having issues ONLY with the AA, i change only the AA sampler.
If i am havig issues with the AA not distributing enough, let's say, QMC GI samples in dark areas, but the AA is done properly, then i either raise the GI subdivs above the max of the QMC AA, or lower the rQMC threshold.
The first method will ensure that VRay will spend more time only in that particular phase of the shading, and not for everything else(shadows, and so on).
I forgot to mention one thing, Which res3d found out painfully by himself:
I NEVER use any AA filtering with QMC aa.
As it is a purely oversampling method, i do trust it to do a more accurate, and a LOT faster, job at antialiasing and blending my pixel values, than the filtering could.
I DO use filtering with adaptive subdivision, where i WANT the image to be undersampled and interpolated by the filter where needed.

Hope i didn't mess it up more for you :P

Lele

[panthon]

I NEVER use any AA filtering with QMC aa.
As it is a purely oversampling method, i do trust it to do a more accurate, and a LOT faster, job at antialiasing and blending my pixel values, than the filtering could.

Lele

Thanks for the info!
I will try this option of not using AA filtering at all to see how it goes!

[dlparisi]

I forgot to mention one thing, Which res3d found out painfully by himself:
I NEVER use any AA filtering with QMC aa.
As it is a purely oversampling method, i do trust it to do a more accurate, and a LOT faster, job at antialiasing and blending my pixel values, than the filtering could.
I DO use filtering with adaptive subdivision, where i WANT the image to be undersampled and interpolated by the filter where needed.

Lele,
Could you elaborate on this a little? When I set it to Adaptive QMC image sampling and the filter to none I still get lots of jaggies and moire problems. What settings are you changing to avoid this or am I missing something .?

[emptyvoxel]

Here's Vlado's answer to the original question:

http://www.chaosgroup.com/forum/phpB...ic.php?t=16393

0.001 for qmc in ver. 1.47 was a really good setting, now in 1.5, do i have to use 0.001 further, or is 0.01 now the "magig" number?

0.001 for the old QMC sampler would translate roughly to 0.005 for the new one.

[studioDIM]

Well, the normal way of doing AA led oftentimes to turn down bitmap filtering, to help keep textures sharp in situations where AA filters tended to blur them.
I get no moire (not any more than usual, that is) if i leave filtering of bitmaps at 1-pyramidal and use QMC AA with no AA Filter.
Bear in mind that if you have other QMC sampled effect in the scene, and went towards a UniversalSettings setup (all subdivs at 1, for instance), you WILL need something like 10 or 12 max subdivs.
Sometimes even more.
If raising them doesn't help much, then the problematic areas are below the threshold, and you will need to decrease that.
Jaggies on HDR reflection edges will always be there for the reasons that have been explained by Vlado and Joconnel in his beautiful tutorials (where he explains also how to do a little post to help clean them up).
The only way i know of to avoid them is by color clamping.

Lele