pmc render -- portions not cleaning w/sample increase?

[johnb4467]

Hi everyone,

I'm working on an interior scene, and rendering with PCM. The overall render is cleaning up fairly nicely when given sufficient samples, but the two candelabras on the sides do not seem to clean up, regardless of how high I set the sampling.
I'll attach a still -- if you look on the screen right / screen left, the candle-holders / candelabras are still grainy, even after I pumped the samples way up to 16,000. The rest of the scene looked fairly decently only at 4,000.

This is for an animation sequence, so I definitely can't have each frame taking 28:03 to render! I'm guessing it's a priority setting of some sort...the only one I can think of (and will try) is the "direct light importance". It could be something else I don't even know about though.

Worth nothing, if I export the candelabra out of the scene and put it in one by itself, it will clean nicely once isolated. I'm guessing it's just not giving enough samples to those areas for whatever reason?

Any suggestions or help would be greatly appreciated -- specifically on this issue, but also on any render-optimization in general. The only setting I've really changed is max depth, by dropping it to 8 instead of 16.

Thanks you!!

John

[TBFX]

Hi John,

Try increasing the Sampling Rate attribute on the candle flame emitters, they are very small lights compared to the other lights in the scene.

Also do a test render of a few frames using PMC before committing to this kernel as although it can converge more quickly I have found there is often undesirable low frequency noise or flicker using PMC for animation.

T.

[johnb4467]

Hi John,

Try increasing the Sampling Rate attribute on the candle flame emitters, they are very small lights compared to the other lights in the scene.

Also do a test render of a few frames using PMC before committing to this kernel as although it can converge more quickly I have found there is often undesirable low frequency noise or flicker using PMC for animation.

T.

Thanks a bunch, T. -- I just got back from teaching & will likely give that a try in just a few minutes.
If PMC doesn't prove to be a solution here...would path-tracing actually be? I don't think direct-lighting is an overly promising option, as it just doesn't seem to give the same quality -- unless there are kernel settings I might be able to highly-tweak?

Thank you for the input!!

John

[TBFX]

If PMC doesn't prove to be a solution here...would path-tracing actually be? I don't think direct-lighting is an overly promising option, as it just doesn't seem to give the same quality -- unless there are kernel settings I might be able to highly-tweak?

Path tracing is certainly an option and the kernel I use the most for animation, however some nice results can be achieved with the direct lighting kernal if it's set to Diffuse mode, though some things like caustics won't work.

Also, just as an aside, if you are doing interiors with direct lighting in ambient occlusion mode you need to increase the AO distance to the size if your room to stop light from your env/sky spilling in where it shouldn't.

T.

[johnb4467]

If PMC doesn't prove to be a solution here...would path-tracing actually be? I don't think direct-lighting is an overly promising option, as it just doesn't seem to give the same quality -- unless there are kernel settings I might be able to highly-tweak?

Path tracing is certainly an option and the kernel I use the most for animation, however some nice results can be achieved with the direct lighting kernal if it's set to Diffuse mode, though some things like caustics won't work.

Also, just as an aside, if you are doing interiors with direct lighting in ambient occlusion mode you need to increase the AO distance to the size if your room to stop light from your env/sky spilling in where it shouldn't.

T.

Hi again.
So, after plenty of experimentation -- you were right about both:
Upping the sampling rate on the candle "wicks" did indeed fix the issue!
And after viewing my render from last night (to see how things were looking without the candles), it is also true that I am getting (I believe) that noise you are speaking of. I'm doing a 10-frame test render with the samples upped on PMC again, but that's really upping my render times.
I'd be happy to use pathtracing, but when I do so, is there anything I need to adjust? Even if I double the sample rate, it still seems like there's a LOT of the image that it doesn't resolve / fill in -- almost like dots everywhere for a lot of it.

I tried Direct-Lighting for a long-while before reporting back here, both on AO and Diffuse, and neither one were really giving me anything in the league of what PMC (or if I can get it working, I would believe) Pathtracing can work out.

Thank you for all of you help thus far!

[ciboulot]

Hello,

after seeing your image, I think you might give a try again to direct lighting.
If you set correctly the settings you might get a result very near to PMC. I think.

Set the glossy samples, specular samples to 25 to begin and diffuse samples to 8. Maybe it will do the trick.

Also, as you saw, the sampling rate setting of each light is very important. I just had a short film with those noisy lights problem. We spent a quite a tmie finding the appropriate settings for each lights.

Ciboulot

[johnb4467]

Here's some updates (with screenshots)...though I'm still not really getting workable results:

I let some really long renders (was sleeping) of the scene set to pathtracing run. Things like the floor and candles remain pretty spotty:


The PMC render, running much less of a time period (but still very long), looks much more refined...however as noted by feedback...doesn't tend to animate smoothly:



And last-but-not-least, I ran a direct-lighting / diffusion version (without changing any other settings, etc), and it tends to look pretty spotty, much like the pathtracing render...and also doesn't fill out any of the areas like the doors in the back of the room:


Is there something I am fundamentally missing here? Would uploading my maya script be of use; maybe it's something simple like scene scale, perhaps (though I have adjusted that setting in the kernel settings somewhat, to no significant result)?
Just in case -- here's a link to the maya file (zipped):
https://dl.dropboxusercontent.com/u/695 ... _setup.zip

[TBFX]

Hi John,

I'll download your scene and take a look at some point over the weekend but I suspect it is just one of those scenes that is very hard/slow to render. It is a totally enclosed, self lit scene with mostly bounce light illuminating what is visible to the camera, pretty much a worst case scenario.

T.

[TBFX]

OK, I had a quick look at your scene and you have tried to set up your main pendant light really physically correct. In other words you have set it up a an emitter that both reflects off the upper housing and then refracts through a specular glass material, however the glass is also only single sided (has no thickness) so it's like having your light stuck in the middle of a solid dome of glass so not actually correct. You could try giving the glass thickness but you would still essentially be lighting your whole scene with a caustic. This will be extremely slow and although you could increase the caustic blur attribute in the render settings and see if that helps I think the best idea may be to turn off this light and make the "glass" surface itself a diffuse emitter, not as accurate I know but will probably speed your render by orders of magnitude.

T.

[johnb4467]

This is excellent advice; thank you so much, T -- especially for taking a look at it so quickly...it really is appreciated when someone takes the time to go out of their way to be helpful like you have.
Well, I suppose I get an A- for trying to be as "realistic as possible", but it also seems that with even Octane it's not necessarily render-smart. I'll take a go at making the glass itself an emitter and see how things progress. Thank you again.

John