1. Join Date
    Aug 2009
    Posts
    70
    #1

    Physical Correctness

    Hi,

    i just gave Nox a try, and i have to tell that i am not very impressed. Speed is rather mediocre, (for example indigo renderer seems to be 2x - 3x faster) but most importantly, i cannot understand one thing. I saw a tutorial about Nox - 3ds Max workflow, and there was a hint "Global illumination usually renders too dim, so we have to correct it". I cannot understand how this is possible. If its supposed to be unbiased renderer (with some simplifications, as you stated) then there should not be such a thing as "too dim GI". There should only one GI intensity, the correct one. If you leave GI intensity in the state, where it needs to be tweaked by user, then you introduce very significant bias, way more serious then for example bias caused by interpolated GI technique. Stable and correct GI intensity is essential.

    From my comparsions with other unbiased renderers (Indigo and iRay), it does not seem to behave correctly.

    One more problem i see is a weird behaviour of daylight system. It looks too oversaturated, like if it was rendered non-linearly. I know its a dependent on camera response curve, but it still seems too much.

    And lastly, for some reason, shadows from sun are completely hard and crisp by default. I do not have Nox explored deeply yet, but soft sun shadows should be there by default. Sun disc is not a point light.

    Anyway, i wish you good luck. There is never enough of renderers to keep competition on their feet and running forward. I am believe that in final version, Nox will be production ready, stable and fast renderer. Evermotion was always a mark of quality. ;-)
  2. Join Date
    Aug 2009
    Posts
    70
    #2

    And i have found a first bug

    I have set location in Nox daylight system to central Europe. Have set GMT to 1, which should correspond with central European time. And i have set the date to 12th August, which should be late summer. According to Nox daylight system, sun would be on horizon already around 18:00, when in reality, in such a month, sun goes down about 21:00-21:30. So there seems to be something very wrong somewhere.
  3. Join Date
    Aug 2009
    Posts
    70
    #3

    Some response would be nice...
  4. Join Date
    Jul 2007
    Posts
    532
    #4

    Sorry for long time waiting... but first day of release and there is still a lot to do. So...

    GI - it's intensity is evaluated for the whole image and has nothing to do with pixel colors or differences of intensity between pixels. There is no doubt that this image intensity is not correct. However for beta it is temporary left as is. The reason for this is that metropolis light transport (mlt) is quite a difficult to implement technique - there is no simple guide for creating one. During development of mlt I had a lot of doubts and misundersands. Often resolve came out after a lot of tries of intuitevely changed code. It is possible that this way of GI evaluation will become a dead end in NOX and we will go back to old good path tracing. What can I say, it's still a beta.

    As for sky color... I know that. It's part of code that some day will be rewritten.

    Sun size - it's not true, that shadow is too hard, unless you had some bug. Its shape (circular) and direction (cone of directions) of light is controlled and set as on Earth by default (0.533 degrees).

    As for sun position on sky... august is a summer time. So another hour of gmt. With that settings in NOX I get a sunset near 8-9PM.
  5. Join Date
    Aug 2009
    Posts
    70
    #5

    Quote:
    Originally posted by: marchewa
    Sorry for long time waiting... but first day of release and there is still a lot to do. So...

    GI - it's intensity is evaluated for the whole image and has nothing to do with pixel colors or differences of intensity between pixels. There is no doubt that this image intensity is not correct. However for beta it is temporary left as is. The reason for this is that metropolis light transport (mlt) is quite a difficult to implement technique - there is no simple guide for creating one. During development of mlt I had a lot of doubts and misundersands. Often resolve came out after a lot of tries of intuitevely changed code. It is possible that this way of GI evaluation will become a dead end in NOX and we will go back to old good path tracing. What can I say, it's still a beta.

    As for sky color... I know that. It's part of code that some day will be rewritten.

    Sun size - it's not true, that shadow is too hard, unless you had some bug. Its shape (circular) and direction (cone of directions) of light is controlled and set as on Earth by default (0.533 degrees).

    As for sun position on sky... august is a summer time. So another hour of gmt. With that settings in NOX I get a sunset near 8-9PM.
    Hi, thanks for all the info. I can provide you with some references from other renderers to help you set the daylight system straight. About MLT: Indigo for example uses it, and they do not seem to have any problems with it, intensity is same as with pathtracing, but it converges actually slower than PT in most of the standard Archviz scenes. That together with the fact, that MLT needs a lot more setup makes is controvert to use. If you are going to switch to another method, i suggest you consider Bidirectional PT, as it should be a lot more efficient than standard PT.

    Oh, about daytime, that was my mistake, but some daylight saving time checkbox would be handy. ;-)

    And about sundisk size... i might have done some mistake, i will test it on some scene where the effect would be more obvious

    Anyway... Good luck!
  6. Join Date
    Jul 2007
    Posts
    532
    #6

    Looks like you know what's what. That's great

    I'd be grateful for your help. I'd be happy to see those references you said. I'll send you my e-mail on PM.

    Thanks

    Also for smooth shadow from sun... I believe sometimes sun visibility may be a bit distorted by atmosphere and shadows may become a bit softier than in technical evaluations that NOX is based on.
  7. Join Date
    Jan 1970
    Posts
    119
    #7

    mlt is useful in case, that you have a lot of glass,sss or dielectric materials. MLT is searching for complex areas. sure, its slower than classy PT but useful for fast previews and indirect lighning,glass,sss,dispersion.. classic PT is more efficient for long rendertimes. MLT usually stuck after few hours on blotchy noise and noise is dissapearing very slowly.

    but this can be tweaked by touching mutation probability.

    check this thread, very useful http://www.indigorenderer.com/forum/...php?f=3&t=2213
  8. Join Date
    Nov 2003
    Posts
    37
    #8

    I like the GI compensate, but yes you need the Physical correctness for a starting point.
    Last edited by Invaderzim; 2011-07-14 at 05:51 AM.
You need to be logged in to leave a comment. You don't have account? Register now.
Best of Evermotion
The best works from the Forum Gallery picked by Evermotion staff