CoronaPhysical Mtl & Physical Mtl Questions

[Jpjapers]

Recently ive been confused by a few things in the the Corona that over the years ive learned to live with.
couple of questions; For the attached example, Physical material is on the left and corona on the right.

1. Why does inverting a map in the bitmap node result in a different output than if i invert it within a corona colour correct node?

2. Why do the same PBR maps connected to both a corona physical material and an autodesk physical material render differently to each other?  Ive attached an example compariso_galv.jpg. On the left is a physical material, the right is a corona material. Both seem slightly different colours and also seem to be reflecting differently. It seems corona assumes metals have an IOR of 2-3 because i had to set the physical material to 2.5 to get a close match. Is it just down to a different implementation of metalness?

3. It appears to me that physical materials render much faster and clean up much quicker with the gpu denoiser. At least during this testing thats what i saw. For instance in the attached comparison, the corona material on the right is far more blurry than the physical on the left which has cleared up in the same number of passes.

Not a huge difference but im looking at the viability of using physical materials instead of corona for better compatibility with our pipeline at work and just noticed a few subtle differences.

[romullus]

#1. Corona colour correct is not linear, if you want to get the same result as with invert in Max's bitmap or output nodes, then you can use curve adjustments in CCC.

#2. If you want metal, then switch Corona physical to metal node. 2-3 IOR is not metal.

#3. what denoising you are using and what image filtering. If that's Corona HQ denoiser, then filtering also should be set to HQ, otherwise it will severely underperform. I didn't notice dependency on material type, but maybe that's because i almost exclusively use Corona physical.

[Jpjapers]

#1. Corona colour correct is not linear, if you want to get the same result as with invert in Max's bitmap or output nodes, then you can use curve adjustments in CCC.

Good answer thank you, i did not know that.

#2. If you want metal, then switch Corona physical to metal node. 2-3 IOR is not metal.

I think you slightly misunderstood my intent. Im saying that to match the coronaPhysical metal in with a physical material i have to set the physical material IOR to 2.5-3. Coronas metal mode locks the IOR of course, which is physically incorrect anyway because metal ior has a fairly broad range range and absolutely cannot be defined by a single arbitrary number. I would hope this is something that gets changed in future so we can get physically correct metals.

#3. what denoising you are using and what image filtering. If that's Corona HQ denoiser, then filtering also should be set to HQ, otherwise it will severely underperform. I didn't notice dependency on material type, but maybe that's because i almost exclusively use Corona physical.

Its the Nvidia GPU denoiser in IR. It just seems to handle the physical materials faster for whatever reason.

[/quote]

[romullus]

I think you slightly misunderstood my intent. Im saying that to match the coronaPhysical metal in with a physical material i have to set the physical material IOR to 2.5-3.

Maybe it would better if you would show side by side comparison with the same settings then, i.e. metal mode is on, metalness is mapped.

Coronas metal mode locks the IOR of course, which is physically incorrect anyway because metal ior has a fairly broad range range and absolutely cannot be defined by a single arbitrary number. I would hope this is something that gets changed in future so we can get physically correct metals.

You can switch to complex IOR mode which is more physically correct for metals.

Its the Nvidia GPU denoiser in IR. It just seems to handle the physical materials faster for whatever reason.

Maybe try to switch to Intel AI denoiser, i think it shows slightly better results than Optix. Anyway the difference in denoising quality in your example is really huge, perhaps it's worth to post it as a bug?