Weird Fresnel IOR Results [Solved in post 6]

[dubcat]

I wanted to see how my custom falloff map looked as a graph. I have a scene with 91 planes all rotated 1 degree each. The last plane is using a white material since we can't see a 90 degree plane. I'm using white as Environment and no ColorMap. Credit to adanmq for teaching me this.

- I did a test render with 3dsmax's default falloff map and got some weird results. It's like the ramp is getting brighter and brighter, and when we reach somewhere in 80 degrees, the ramp gets darker. I launched V-Ray 3 and did the same test, the render looked correct.
- In my second test I used the default IOR input with 1.52. The render was correct.
- My third test was with Sigers ComplexFresnel map, Copper preset. The same problem as the first test, the ramp is getting darker at the end.
- Fourth test. default IOR input 4.0. The render was correct.

Does this mean that every time I use a custom falloff map in Corona my fresnel is wrong ? I have never noticed this in my materials before, maybe it's not that important. Just wanted to let you guys know.

EDIT: IOR 999 is the cause, not the custom falloff map. More info in post 5
EDIT2: Fix in post 6

(Hope it's OK that I compared with V-Ray, I never got used to mentalray/iray)

Default 3dsMax falloff map



This is the gradient for Default 3dsMax falloff map. First row is Vray3, second row is Corona. You can see there is something going on in the 80-89 degrees region.



Fresnel IOR 1.52. Working as it should.



This is the gradient for 1.52. First row is Vray3, second row is Corona.



Sigers ComplexFresnel map, Copper preset. Weird spike is back.



This is the gradient for Sigers ComplexFresnel map, Copper preset. First row is Vray3, second row is Corona.



Fresnel IOR 4.0. Working as it should.



This is the gradient for 4.0. First row is Vray3, second row is Corona.

[romullus]

Just to be sure, when you rendered with Corona, did you set MSI to 0?

[Ludvik Koutny]

If you use falloff node in Fresnel mode with GGX in Corona, it won't work correctly. It's known limitation.

[maru]

This is "the spike"?
https://forum.corona-renderer.com/index.php/topic,4542.45.html

[Ondra]

If you use falloff node in Fresnel mode with GGX in Corona, it won't work correctly. It's known limitation.

This - Corona has different built-in fresnel model to make the BRDF physically correct (reciprocal)

[dubcat]

Just to be sure, when you rendered with Corona, did you set MSI to 0?

Default render settings. Setting MSI to 0 didn't change anything.

This is "the spike"?
https://forum.corona-renderer.com/index.php/topic,4542.45.html

I tried "Legacy (A7) mode", same results.

If you use falloff node in Fresnel mode with GGX in Corona, it won't work correctly. It's known limitation.

Ok, thanks :)

EDIT:
Think I found the problem.
It's 999 IOR that is causing the problem. This is why 1.52 and 4.0 did not have the problem.

This render has no custom maps, only 999 IOR. The black box is showing where 999 IOR is tainting the ramp.



Here's the mysterious spike.
From this day forth it shall be known as...."The 999 Spike"



I made a IOR Gradient list for people who are interested. The red numbers correspond to the IOR value.

[dubcat]

I'm double posting, because this is worth its own post.

I found a fix!!!

Corona Default Max Falloff



Gradient



Corona Siger Complex Copper



Gradient



This is what you do:
Add a CoronaColor map into "Fresnel IOR" and copy paste 100000 into R G B, this will disable IOR.
999 still has values. Even 10000 has values.



I have added a Material Library.

I did a test with Sigers "Paint Metallic Arctic Silver" material, it was 10 RGB darker before the fix (In the problem area).

[romullus]

Now take into account the fact that fallof map isn't linear in 3ds max, also plug fresnel into glossiness like someone suggested... there's lot of movement have to be done just to have correct(?) base.

If you use falloff node in Fresnel mode with GGX in Corona, it won't work correctly. It's known limitation.

This - Corona has different built-in fresnel model to make the BRDF physically correct (reciprocal)

So what does that means? Is Corona fresnel has bug/limitation or it is physically correct? I'm kinda lost now.

[Ludvik Koutny]

From what Ondra told me, basically GGX required some different fresnel implementation, that's different from Max's native one, so if you for example set FresnelIOR of CoronaMTL with non-mirror reflection to 999 and plug in falloff node with fresnel 1.6, and compare that to CoronnaMTL with same settings but IOR of 1.6 straight in the material, you will notice the falloff node one to cause weird white fringing around the edges. IDK if the one in the CoronaMTL is anyhow better, but from the way i understood it, it's necessary for GGX to work correctly :)

[Ondra]

microfacet models like GGX have fresnel factor built directly into it, and it is very difficult to separate it out. We tried, and what you see is the result ;)

[siadebisi]

I still don´t get how to use curves 100% physically correct if not by disabling IOR with 999 Value. Do i have to use 1.6 while using the curves or is there a way to get the exact IOR Value from refractiveindex by multiplying a special value. Gold f.ex. has a value of 0.3 which isn´t usable at all. I read that you use a different system, but what exactly to we have to do to use it the right way? A physically correct sample material for plastic would be very nice :)