Corona IES light noise

[Dippndots]

Not really sure if this is should go in bug reporting, but I've noticed a couple weird things occurring when using IES profiles within a corona light.

First, a sealed box with single corona IES light generates lots of direct light noise when an HDRI is added to the corona environment light slot. How is this possible, surely it shouldn't be going through walls? The first attached pic below is 10 passes, no hdri on the left half.

Secondly, in a different scene, multiple Photometric IES lights generate very little direct light noise, but multiple Corona IES lights create lots of it, to the point where 200+ passes doesn't clean up the noise. The first 2 pics are Photometric and Corona at 32GIvAA with 2LSM the 3rd is Corona at 16GIvAA/6LSM (suited to clean up the extra direct noise being thrown out).

Sphere vs disk on the corona IES didnt make enough of a difference in my opinion. The photometric is just a point, soft shadows option is off.

What is going on here? Anyone run into this before?

[Ondra]

even though the environment map does not contribute to the image, there is no way of knowing that unless you explicitly sample it. So the rendering gets more noisy.

second scene: it is simple - photometric IES lights are fake (they are point lights, not area lights), so they can be faster in some scenes

[maru]

First, a sealed box with single corona IES light generates lots of direct light noise when an HDRI is added to the corona environment light slot. How is this possible, surely it shouldn't be going through walls? The first attached pic below is 10 passes, no hdri on the left half.

This definitely shouldn't be happening but these "hot pixels" or light leaks are a known problem and they will be probably fixed in near future.

Secondly, in a different scene, multiple Photometric IES lights generate very little direct light noise, but multiple Corona IES lights create lots of it, to the point where 200+ passes doesn't clean up the noise. The first 2 pics are Photometric and Corona at 32GIvAA with 2LSM the 3rd is Corona at 16GIvAA/6LSM (suited to clean up the extra direct noise being thrown out).

This is (probably) expected, and as you did, increasing LSM should help in such cases.

Sphere vs disk on the corona IES didnt make enough of a difference in my opinion. The photometric is just a point, soft shadows option is off.

I do not understand this one. Can you explain?

[Dippndots]

Ok that makes sense re. the evironment map Ondra. The photometrics in my case were much slower.

Sorry Maru, what I was trying to say is that I had come across some examples on the forums where disk versions of IES light were throwing out significantly more noise compared to the sphere option. From the tests I did the disc had more noise than the sphere but not enough that I felt sphere's were a solution to the problem.

I guess its just finding a balance, mixing corona light and photometrics seem like a bad idea as photometrics in my case moved a lot of noise to the indirect pass, where as indirect noise cleans up very quickly for me with corona lights, so its hard to tweak the settings one way or the other.

I'll keep tweaking/testing and post a final result later.

[Dippndots]

GIvAA of 20 and LSM of 8 seems to be the sweet spot for me, still needs close to 300 passes to clean up the noise though.

[maru]

So if you want to keep the same quality, but lower the amount of passes (=you do not need that much AA), you can simply increase GIvsAA further.

[Dippndots]

Ok, I just noticed that since my indirect noise was cleaning up faster than direct, I lowered the GI. Is this the right way to go about it?

[maru]

Ok, I just noticed that since my indirect noise was cleaning up faster than direct, I lowered the GI. Is this the right way to go about it?

Nope. You will get exactly the same ratio. :)

GI quality = GIvsAA
direct quality = GIvsAA*LSM

So if you change GIvsAA only, you are also changing LSM proportionally.

[Dippndots]

Yeah sorry, should've mentioned I was raising LSM as well.

I've gone with  GIvAA of 32, LSM of 8 and 175 passes and have a similar result as previously, but also shaved about 10% of the render time. So success in the end!

[steyin]

Yeah sorry, should've mentioned I was raising LSM as well.

I've gone with  GIvAA of 32, LSM of 8 and 175 passes and have a similar result as previously, but also shaved about 10% of the render time. So success in the end!

Those are my typical settings when using IES for final renders.