To devteam: Shadows and Realroad tex's

Discussion in 'Technical Archives' started by GTrFreak, Jun 1, 2012.

  1. GTrFreak

    GTrFreak Registered

    Joined:
    Dec 7, 2010
    Messages:
    376
    Likes Received:
    7
    Hi there guys,

    First of all, b85 improved rF2 in almost every aspect and I want to say thanks for that particular reason. I know graphics are a priority now so I thought of informing you with some things I've been thinking about:

    Shadow bias settings

    However I'm not into the modding itself I helped some modders several years ago to promote their work, by making renders of cars for them (mainly for GP4). Because of this I know there's something called shadow bias (in 3dsmax at least) In short, the higher this number, the greater the distance between an object and the 'starting place of the projection of the shadow'. I can't put this in other words but hopefully the image attached will show what I mean.

    My question to the devteam:

    I know this bias is there to prevend any shadow glitches, flickering or prevend something I think called Z-fighting issue (?). But I think the bias is to much noticable. Is this about to change or is this a standard setting which will be kept, because that would be a shame really..

    In this screenshot of an OW'er in pCars the bias is noticable on the wing endplate (below the small wing) but under the suspension it's not even showing up. I find this a way better balance between functionality (not flickering or glitching shadows) and graphic quality.

    barely noticable in pCars
    View attachment 2654

    Very noticable in rF2
    View attachment 2651

    Realroad textures

    Now, something else which has been discussed recently is how the Realroad textures look. In b85 the tex is getting influenced by sunlight as it's specular map has been altered. I think this is a good thing, although it looks way off at the moment. I think the realroad textures' specular map isn't the worst thing, I think the overall texture isn't the perfect one either. When you look at screenshots of a well rubbered-in track it's like someone just smudged black all over the place, and, even worse, it's white when you watch towards the sun.

    Example of b85
    View attachment 2653

    I've thought of this for a moment andthis is what I think of it. In some sort of way we need a texture which contain tire 'tracks' (?). Lines of rubber, not just a random black spot, but a nice collection of skids, lines of rubber in one texture. Something like this:

    View attachment 2648

    You can clearly see (even when you left out the environment and kerbs) the direction the cars have been driven in.

    Now, when we have a basic texture of some sort, you can think of how a specmap would have to behave like. I think, when you watch lots of photos of racinglines, the racingline itself is cleaner then its edges, so the edges would be a bit darker in the diffuse texture and the base of it a bit lighter. IMO the more smoothened base of the racing line needs to reflect light (no too much, making it white) and the darker outside of it not nearly as much. That way you can get this effect, screenshot borrowed from K Szczech (thanks mate ;)):

    View attachment 2649

    The main difference in this last comparison is the angle at which the sunlight is reflected (let's say) a hundred percent VS the angle at which the sunlight is reflected less than 20% (or so). By this, I mean, when you see the racing line directly under the sun it needs to reflect, but when the angle is like the line is not under the sun anymore the falloff (or maybe the decrease of reflection) needs to be very fast/hard/steep (hope you're getting my point here). Little explanation:

    View attachment 2650

    In other titles with newer generation graphics (with more advanced shaders) you see that skid marks in low sun angles are reflecting extremely much light, which isn't correct either. Example of pCars:

    View attachment 2652

    I think Realroad is a fantastic feature graphically aswell as physically but even though it's a unique selling point of rF2, it doesn't need to be overdone. Just make it noticable but subtle.

    Would be nice to get some thoughts of the devteam, especially the guys working on the graphic side of all this.

    Thankyou

    I'm struggling to keep this clear for everyone because of English not being my native language but I hope you guys understood my points, think of all this and will share your opinion about all this. Thanks for reading
     
    Last edited by a moderator: Jun 1, 2012
  2. K Szczech

    K Szczech Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,720
    Likes Received:
    45
    Look at the shadow of right mirror:

    [​IMG]

    It really depends on angle, so you would need two similar screenshots with the same sun position to compare shadow quality in two games.

    Although there are some areas in rF2, where there's light bleeding near corners, where it's not justified in my opinion. Like at edge between ground and building - it gives me the impression that shadows are casted by backfaces, while in pCARS they're casted by front faces and therefore there's less light bleeding.
    I think it's always better to let front faces cast shadows, because it makes shadows more solid.


    Shadowmap have some quality limitations and no game engine will help with that. You will see some games have light bleeding while others have shadow bleeding - depending on which way they went with shadow bias.

    One could take screenshot from one game and say - "it looks good in situation A", then take screenshot of another game and say - "it looks good in situation B". So then he could say: "So take solutions from these two games and you will have perfect shadows" - but it just doesn't work this way.


    And as for my screenshot with Murcielago - it's much more zoomed in than F1 photo on the left, therefore reflection on the road seems wider :)
     
    Last edited by a moderator: Jun 1, 2012
  3. GTrFreak

    GTrFreak Registered

    Joined:
    Dec 7, 2010
    Messages:
    376
    Likes Received:
    7
    I see what you mean concerning shadow bias. I am fully aware of it being one way ór the other but I find it not being optimised enough. I'd like to know if there are people who agree with me and ultimately some answer from the devteam concerning these settings.

    as for your screenshot about readroad specmap, I see what you mean but I think the shot itself explains well what point I like to make. Because in b85 the realroad tex's show up white 50% of the time while you don't have the sun directly above the road as near as much..

    Edit:

    About shadowing: Don't know anything about game graphic engines but do you think it's something to do with scene scaling or some sort? I mean.. It's got to be done better right?
     
  4. K Szczech

    K Szczech Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,720
    Likes Received:
    45
    Shadowmaps have limited precision. It takes a smart, dynamic shadowmap bias to minimize this effect. It will not dissapear completely though. Thin objects will always cause troubne.

    At the moment I believe shadows in rF2 use one static bias for all cascades, but so does Skyrim for example, and that's why it runs into some problems with NVIDIA's PCF filtering sometimes. I'm not trying to explain it in detail here, I just want to point out that various elements of graphics engine and graphics hardware may affect each other. Game programmer needs some experience to find a way through all that.

    These two images show what kind of probmels programmers need to deal with:

    [​IMG]
    [​IMG]

    In first image you can see that there's a gap in shadow near the wall - like the one on our screenshots. But there are more serious problems there with shadows on surfaces that should be lit.
    Changing shadowmap bias will increase that gap near the wall but will also solve our problems on other surfaces.


    There are other, better ways of solving these problems - like calculating bias dynamically based on the angle between surface's normal vector and light direction. Surfaces facing the light source need only a small bias.
    But that kind of quality improvements usually come later. For now, shadows in rF2 have other problems than just precision :)
     
    Last edited by a moderator: Jun 1, 2012

Share This Page