POST parameter: bug or incorrect behaviour?

[cecofuli]

Hi,
the image talks for me. It seems that the changes in the Post Processing rollouct affect in some way the Corona core.
But, in my opinion, this behaviour is very odd and incorrect.
Try to think...


*(1) Ok, I like my rendering, but it's too dark.
*(2) I raise the Exposure compensation during the test rendering, in realtime. Now it's better and I like.
*(3) I want to change the resolution. I stop the render, change in.. I don't know... FULLHD
*(4) click render, but now exposition and quality is COMPLETELY different than the previous render... (O_o)
*(5) I'm shocked!

It's impossible work in this way!
Or am I wrong something?

[Ondra]

the max sample intensity parameter is relative to the current color mapping, so starting render with different exposures will result in different max sample intensity, and in turn to different amount of bias in the picture. This is so that for example MSI 10 results in the same amount of bias in dim interior shot and in bright exterior one.  But once you change exposure after the render has started, you cannot change the already rendered bias.

[cecofuli]

If I understand correctly: max sample intensity parameter is exposure dependent.
And, if yes, Exposure compensation is not very useful, because, every time I restart the render after some MSI modification in the previous test in Realtime, I'll obtain a different version, EVERY TIME! As in my attachment

In VRay I can change the Exposure in the VFB and, every time, I obtain the same image. Because the Post modifications in the VFB don't affect the sampling.
Only the Color mapping (Rehinard etc...) influence the images.

In conclusion, isn't there any solution of my problem? 

So, if I want to obtain every time the same image/quality solution, what is the correct workflow?
Do I must to start (and change :-( ) every time with the default setting:

* Exposure compensation:0.0
* Highlight compensation: 1
* Contrast: 1.0

And change their as I want? It seams to me so painful method...

[cecofuli]

No idea how to help me? =)

[Ondra]

I am still not sure what you are observing is a bug or not, but what you should do is to iterate rendering with different exposure controls to get the result you want, not reset exposure or anything like that

[cecofuli]

Keymaster, I don't know if it's a bug or now. But I want to understand how use Exposure compensation parameter.
I use it during a rendering (Render A) and I change it in realtime from, I don't know, 0 to 3, obtaining a nice render (Render B). I stop the render, i change anything (resolution, or some HD cache parameter etc... ) and I start again. The new image (Rendering C) is complete different from (Rendering B). So, how can I use correctly the Exposure compensation parameter if I don't know what happen=  In my opinion, we need a Exposure compensation parameter that doesn't affect the MSI or any other Corona core parameters, but only the final image, as Highlight and Contrast do!
Or I'm lost something? =)

[Ondra]

you can iterate, adjust exposure of render C, render over, eventually adjust exposure of render D, etc.  I could do exposure-independent MSI, but that would complicate GUI too much.

[cecofuli]

Keymaster, this is what I wanted to tell you... =) Just a exposure-independent MSI! This it would the best solution.
I know, I can Iterate (bad and not professional solution in my opinion) but as you can see in my test, with Exposure compensation: 2.5 I have not only a different "exposition", but also the quality of GI is very bad!!!

EDIT: look in third image, the ceiling and in the base of the right wall. There are a lot o splotches...  :-(

[cecofuli]

I recorder a new video. I hope this can help :)

[maru]

I think we must wait for a camera object with exposure controls. In current version it's best to change exposure in post.

[cecofuli]

maru: I think the same!

But thank. Now I understand better why Exposure compensation makes different render with the same value ;-)

[cecofuli]

Ok... loot at my next example...  How can I brighten up the image???

Keymaster, please: do exposure-independent MSI!

[maru]

it's probably because of your highlight compression @ 30

[cecofuli]

Unfortunatly no. See the image

[maru]

You are right, something is wrong here. From what I see it doesn't happen when using CoronaLights.

[maru]

And looks like changing max sample intensity changes a lot.

[cecofuli]

You are right!!!  I'm doing more test, wait...

[cecofuli]

Mmm... Why does Max sample intensity change so much the exposition?
What do I do to obtain the correct quality lighting solution in these scene?

Exposure compensation doesn't work when I restart the render...
But if I change the MSI (from 10 to 50 for example), the scene is much bright with the same Exposure compensation
Also, if I use th esame  MSI (100 for example) with different Exposure compensation (from 1 to 4 for example) the brightness is very similar... (with Exp = 4 a little bit brighter)

I think there are something strange in the Corona Color mapping...

[Chakib]

Weird bug indeed...

[cecofuli]

With Corona light rectangular:

With Exp: 4.0 -> If I change MSI from 2 to 100 I've a little bit more bright solution

With MSI = 10
If I change Exp from 1 to 4 starting from 1 and change in realtime, the lighting brightness is very similar if I start directly from 4... the mystery increases!

(*) about the Exposition problem. I obtain a different result with the Corona Sun, but not with Corona Rectangular Light
(*) CoronaMtl emission has some problem with the Corona Color mapping and/or MSI parameter

[patoaltaco]

Cecofuli,

You are going to write "corona the complete guide" ?  :)

[cecofuli]

Ehh... Maybe =) 
One of the most parts I don't like VRay is the color mapping. Especially the famous problem with sub-pixel mapping OFF
I'm looking for a good render in CPU (cause RAM amount problem on the GPU and plug-in support).
Corona is a good renderer but it's have, as we can see, some problem too in the color mapping area :(

[cecofuli]

Here another VIDEO where you can see better all different light sources: CoronaSun, Portals, Rectangular CoronaLight and Emitters.
Only Rectangular CoronaLight  seems not to be affect by Exposure compensation (or very slightly)

SCENE DOWNLOAD

[Ondra]

ok, this is weird, MSI should not influence the result so much. I'll need to take a look at it when I am sober.

[cecofuli]

Keymaster yes, I think the same. But it could be a good start to have the Exposition parameter MSI indipendent ;-) maybe in some nightly build?
Becous eI would like to use for some interior render with artificial lights and geometry emitters.
But without this option and the MSI corrected it is difficult or impossible to do.

[cecofuli]

Maybe, this idea?

[cecofuli]

Ok, after some test,t his is my conclusion:

if you want that Corona takes into account the Exposure compensation (with Emitters Geometry an Sun), you have to raise up the MSI up to 200/400.
Corona Lights are not so influenced by MSI.  This is the only way how to do with A4.

Now my question: are there a big differences in terms of quality/errors between  MSI: 10 (default) and  up to 200/400?

[cecofuli]

[Ondra]

the exposure is preset before rendering in each case?

[cecofuli]

Yes Keymaster :) I change it before the launch of the render!
I clean up the scene and I post soon!

[Ondra]

ok, post it, this looks like easily fixable bug

[cecofuli]

See the attachments

[Ondra]

ok, I did some experimenting and thinking, and I concluded that the error is not in the way MSI interacts with exposure, but that it is set too low. MSI should be a parameter "how bright are the fireflies permitted to be". But setting it to 10 classifies most of your scene as fireflies (because it is relatively hard scene, lots of indirect lighting). The problem does not happen with rectangular lights, because they make the scene easy (lots of direct lighting). Point the rectangular lights into the ceiling and you will get the same result.

Anyways, because the scene renders fine with high MSI, I think the best solution is to set the default MSI higher

[cecofuli]

Ahh, ok, now I understand.
But, from your point of view:

(*) what is the best setting for the MSI Default values? Because in my scene it was between 200/400.
(*) Are there disadvantages / advantages by using so high values?

Thanks

[Ondra]

It might differ from scene to scene. If 200 works, use 200. I will test later to determine new default

Advantage of high MSI is less bias, disadvantage is more noise, it is simple ;)

[cecofuli]

I did some test and with this scene. Corona is veeery slow  compared to VRay. See the render.

(*) VRay: 1h with noise threshold 0.003 = PERFECT. I can in 30 min with noise threshold 0.005 and the noise is little more, but not too much.  (BF+LC)
(*) Corona: 1h min. so much noise! (PT+HD)

Did I do some error?

[Polymax]

can you share the scene? I want to test the it.

[cecofuli]

It's in the attachment some post above. Ready to Corona. Just hit render

[lacilaci]

hd cache max records 100?

[Ondra]

it is *1000

[cecofuli]

100 cause I remember the old Corona scene and the problem in the ceiling. I'll try with the HD defualt setting. =)

[Polymax]

Standard material for? Why gray enviroment? Floor have a thin geometry!
I have compared the Vray and Corona on a complex interior, with complex lighting:http://forum.corona-renderer.com/index.php/topic,433.0.html
Vray(BF+LC) lost heavily in all parameters.
Corona work fine in this case also!

[cecofuli]

Polymax, nice result!
Standard material because I do this test with VRay and I override with VRay Material. I forgot to override in Corona too
Thin geometry cause Vray don't have problem.
Gray environment because... I don't know, some old setting...
but... can you scare the scene corrected? Thank you so much!
10 min is an incredible time!
Did you use a geometry emitter like in my original scene?
Bacause in VRay I use a VRayLight as a mesh. Corona doesn't have this option and I used a geometry emitters

[Polymax]

Of course, I attach the scene.
I would like to not divert Keymaster just so, he does not have much time for it!
Further, please be careful in their tests, check all parameters! For the Corona is important and thin geometry and enviroment and materials.
Thank you for your understanding!

Yes, I use yours Lightmesh in Corona, but I tessellate her (just in case).

[cecofuli]

Polymax, you are 100%. But with my  test =)  we discovered that the MSI is too low by defaul and in some scenes (both of my scenes) was too low.
Now I try to "study" your scene =) Thanks

[cecofuli]

Polymax, sorry, but... what's you PC? In 13 minutes, with my i7 970 esacore @ 4.0Ghz I've a lot of noise!
I open your file and hit render...  Also the rendering is darker.
I redo the render in VRay in 30 min.

[Polymax]

I have I7-980 (3,91 GHz).
What version of Corona are you using?
I use daybild from March 25.
Verify the settings:

[cecofuli]

I have the official A4 Mar 18...
I'm not able to use the daybuilt...
Yes, same setting of you.
In attachment after 26 min...  :-(
You use MSI = 10 with Exposure 4.5 . From my test, better to use MSI= 200 or more (for the rendering brightness)
Very strange...

[Polymax]

Ok, I understand your problem! Please set the parameter "Writable for # passes" in the HD cache setting to 1 and try again!
Alpha 4 build have a bug with CameraClip and HD Cache, but now is fixed!

[cecofuli]

O__O what's this parameters?? Now all is perfect...
I read the on line help but... it doesn't help me =)

[Polymax]

Ok,solved! :)
ps: On the forum discussed this issue ;)!

[cecofuli]

Ok, it's the HD cache and Clipping plane problem. Thanks

[cecofuli]

And.. what about the weird and bad antialiasing problem near the bright part of the ceiling? Remember me the baaad old sub-pixel mapping and clamping problem in VRay. there is a solution?

[Ondra]

And.. what about the weird and bad antialiasing problem near the bright part of the ceiling? Remember me the baaad old sub-pixel mapping and clamping problem in VRay. there is a solution?

set internal resolution > 1 and then go up with higlight burn until the aliasing disappears

[Polymax]

And.. what about the weird and bad antialiasing problem near the bright part of the ceiling? Remember me the baaad old sub-pixel mapping and clamping problem in VRay. there is a solution?

is another question ;)

[cecofuli]

And.. what about the weird and bad antialiasing problem near the bright part of the ceiling? Remember me the baaad old sub-pixel mapping and clamping problem in VRay. there is a solution?

set internal resolution > 1 and then go up with higlight burn until the aliasing disappears

Ok, I'll try. I'll do more test and I'll post here ;) Thanks

[Stan_But]

cecofuli, if you still have a 'dancing' with post processing parameters, then you must set post contrast to '1' and don't change it more. In this case even MSI 20 will work properly, but may give some 'dancing' with renders in views when using only nondirect lighting

i hope i wrote all correctly)