All posts related to V3
#391967
hopefully that works

just change the material on the plane (already in the mxs... options are additive,coating ,normal ,normal 5050)

@gmenzel this is why I was asking for the torus scene mxs ...... and not the sofa mxs
Last edited by photomg1 on Wed Aug 24, 2016 3:35 pm, edited 1 time in total.
#391968
I don't know if we are mixing topics. On that scene I can see weird things in the coating one, just on fire, works fine on production, wich is normal. And I still can't see energy increasing, or at least not as clear as in the 2 balls example.

I'm not sure about coating, but I can't see any problem on additive blending.
#391970
eric nixon wrote:Gmenzel, Can you post an UP TO DATE example? I'm not seeing any issues here since MW 3.2
Really? I'm using 3.2 too and could easily reproduce the effect.

Just set a 225-lambertian BSDF in one layer and a BSDF with 100% reflectance, roughness=0, nd=10 (purposefully extreme to make the effect more obvious), fresnel=on in an additive layer on top. Put two spheres next to each other and assign the material. Light the scene with a low-intensity sky dome (sun off), so that the background appears as a mid-gray. Done.

If there was no light amplification going on, no part of the rendering should appear brighter than the sky background.
#391971
gmenzel wrote: Really? I'm using 3.2 too and could easily reproduce the effect.

Just set a 225-lambertian BSDF in one layer and a BSDF with 100% reflectance, roughness=0, nd=10 (purposefully extreme to make the effect more obvious), fresnel=on in an additive layer on top. Put two spheres next to each other and assign the material. Light the scene with a low-intensity sky dome (sun off), so that the background appears as a mid-gray. Done.

If there was no light amplification going on, no part of the rendering should appear brighter than the sky background.
Thank you, know I could reproduce it. I'll check it, I think it's a corner case, and this should not work this way.
#391973
I can also reproduce this outside of maxwell in another engine.Also using an additive approach, which leaves a quandary.

How exactly are we supposed to create energy conserving materials in maxwell that also display the correct fresnel properties but at the same time avoid this issue? which to be fair is the problem gmenzel has been alluding to all along and I'm sure why coatings were brought into this discussion.

I can do it elsewhere ,so this is in part to do with how maxwell is setup.
#391975
luis.hijarrubia wrote:Well, maxwell additive blending is conserving energy, or should be.
As I understand it, Maxwell only limits energy exceeding the allowed range (to avoid the computational problems that occurred with earlier versions). Otherwise, additive layers wouldn't work at all.
#391976
photomg1 wrote:I can also reproduce this outside of maxwell in another engine.Also using an additive approach, which leaves a quandary.

How exactly are we supposed to create energy conserving materials in maxwell that also display the correct fresnel properties but at the same time avoid this issue? which to be fair is the problem gmenzel has been alluding to all along and I'm sure why coatings were brought into this discussion.

I can do it elsewhere ,so this is in part to do with how maxwell is setup.
This is what I mean when I say that this is not just some bug that can be fixed. Rather, it is an inherent limitation of the current material system. That's why I revived this thread, asking whether V4 would bring improvements in this regard.
#391977
What happens if you turn of force fresnel, and have a realistic nd for a laquer around 1.5, or do either of those.

I think the additive layer should be 100%, like you set it.

Base layer can be a brighter colour though.

I expect its possible to calibrate materials with a calibrated scene like the simball (assuming that it is still calibrated) or a calibrated hdri which are rare. I think its useful to have the creative freedom to push materials beyond reality a little if desired.

Or maybe it just doesn't work, like you say.. I'm not sure.
#391980
eric nixon wrote:What happens if you turn of force fresnel, and have a realistic nd for a laquer around 1.5, or do either of those.

I think the additive layer should be 100%, like you set it.

Base layer can be a brighter colour though.

I expect its possible to calibrate materials with a calibrated scene like the simball (assuming that it is still calibrated) or a calibrated hdri which are rare. I think its useful to have the creative freedom to push materials beyond reality a little if desired.

Or maybe it just doesn't work, like you say.. I'm not sure.
Any measure that diminishes reflectance will of course decrease the effect. By lowering nd, lowering refl.90 or increasing roughness, you can make the effect all but disappear. All that doesn't change the fact though, that energy increases whenever the output of one BSDF is added to the output of another BSDF without some form of normalization, which seems to be how the additive blending mode works. Tom's statements earlier in this thread seem to confirm this. So it looks like we just have to live with this limitation until NextLimit adds something like a 'stacked' layer blending mode, or a procedural falloff map to do this manually.
#391986
that energy increases whenever the output of one BSDF is added to the output of another BSDF without some form of normalization, which seems to be how the additive blending mode works
But it doesn't work like that, maybe you could say its normalizing to 101% given extreme settings.

Please if you can post a specific test from the latest maxwell so we can rule out caustics of caustics, lighting or geometry issues.

3Dm, I hope we see a 'production ready' version of MW3 soon, .... long before MW4.
#391987
eric nixon wrote:But it doesn't work like that, maybe you could say its normalizing to 101% given extreme settings.
It doesn't? Do you have inside information or what is your reasoning here? Limiting, i.e. cropping, the energy at the maximum of the allowed range is different from normalization. The 'normal' blend mode does that with fixed layer weights, which means that, mathematically, energy can never exceed 100%.
eric nixon wrote:Please if you can post a specific test from the latest maxwell
You can build a test scene that reproduces the effect in literally two minutes, as others here have already done.
eric nixon wrote:so we can rule out caustics of caustics, lighting or geometry issues.
With caustics switched off, simple sky-dome lighting and only two sphere primitives in the scene, I think we can rule out all those potential issues. But by all means, try it yourself.
#391989
t doesn't? Do you have inside information or what is your reasoning here?
My reasoning is that you were talking nonsense; as though there is no normalization when there blatantly is.

Your way of responding in general seems unreasonable. You raise an issue but wont check it with the latest build for the benefit of all who want to know. I cant test this right now. No equipment here..

I have no inside information, just my experience which I am sharing with you. I genuinely thought this was fixed because I haven't seen any issues lately, and therefore the issue cant be that bad can it?
  • 1
  • 3
  • 4
  • 5
  • 6
  • 7
Will there be a Maxwell Render 6 ?

Let's be realistic. What's left of NL is only milk[…]