1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.
Dismiss Notice
Like RaceDepartment on Facebook.

Realistic smoke

Discussion in 'Racer Physics and Technical' started by fsr, Jan 13, 2011.

  1. fsr

    fsr

    As I was testing the drift Supra, the weird smoke got on my nerves, so I played a little with the settings. Seeing as real tire smoke doesn't really have velocity, I started by setting that to 0. Anyone who's ever seen a real drift or burnout knows that the smoke basically just hangs in the air, slowly ascending and dissolving. Unless there's wind, it basically doesn't move much at all. So I basically just plotted in my guesstimated values, and imho they were pretty much spot on right away :)

    Try replacing the smoke section of particles.ini with this and let me know what you think...

    Code:
    smoke
    {
      ; foreach car
      ; Car tire smoke (light; smoke from burning engines uses a different particle system)
      ; Max #particle at once for a system
      max=1000
      ; Lifetime and variance (actualLifetime = life-life_var..life+life_var)
      life=2
      life_var=4
      startvelocity=0 0 0
      startvelocity_var=0 0 0
      ; Start size
      size=1
      size_var=0.5
      ; Growth of particle (m/s)
      growth=1
      growth_var=0.5
      max_alpha=0.1
      alphafade=linear
      ; texture=data/images/smoke.tga
      texture=data/images/smoke_fake_lighting.tga
     
      influences
      {
        airfriction=1
        wind=0.05
        gravity=-0.02
      }
    }
    
    I do however, wish for the smoke size to increase in relation to slip angle. That way we could get more smoke when lighting up the tires. It would also be nice to get some smoke from lateral slip as well, not just longitudinal...
     
  2. Alexander Knoll

    Alexander Knoll
    NEVER GIVE UP!!

    Looks fine...THX...
     
  3. Cool, but in which beta of Racer do you use this settings?
     
  4. I guess all smoke starts small and then disperses, just with more SA you get more smoke particles being born, which increases the apparent size/intensity of the smoke that is there.

    Perhaps the SA a particle is generated at would dictate the life of the particle too.

    So spreading/enlarging is linear and fixed.

    Particle count is fairly high to start with, but maybe 50-100% higher at high slip ratios.

    Life is tied to SA/SR it was generated at (rough approximation for the heat it was generated at for now)


    I think that would look nice and makes sense in my head :D

    Dave
     
  5. I still think the smoke in Racer 050 was the best.
     
  6. fsr

    fsr

    Nah, my settings for the newest versions are pretty sweet ;)

    [​IMG]
     
  7. KS95

    KS95
    RACER Moderator

    I always thought that the way the smoke is made in GTA IV is the best I've seen, it's a lot more gradual than in Racer.. hmm.
     
  8. Except it doesn't look realistic. The smoke doesn't disperse like IRL.
     
  9. fsr

    fsr

    DavidI: What's unrealistic about it? Unless it's windy, the smoke doesn't actually disperse that fast. Basically, it's fully adjustable in the particles.ini file, and you can specify whatever you like.

    KS95: I don't recall the smoke in GTA IV, but I can say that when drifting, it's not gradual at all. It basically pours from the tires and hangs in the air for a long time...

    Here's an example. Now, compare that with my screenie and tell me again that Racer's smoke is not realistic ;)


    [​IMG]
     
  10. I see a lot of difference; the Racer smoke is uniform throughout, the smoke in the picture is very random.

    Now that may be due to a real track surface having variances in surfaces that Racer doesn't have.

    While I disagree on the smoke, that cars looking awesome. Good job
     
  11. fsr

    fsr

    Well, the smoke in the picture is probably as random as it gets. I've seen it a lot less random...

    Having said that, I'm not saying it's perfect in Racer, but it's not bad, and settings play a big role. You could try these settings with the default fx interval of 20:

    Code:
    smoke
    {
      ; foreach car
      ; Car tire smoke (light; smoke from burning engines uses a different particle system)
      ; Max #particle at once for a system
      max=1000
      ; Lifetime and variance (actualLifetime = life-life_var..life+life_var)
      life=1
      life_var=1
      startvelocity=0 0 0
      startvelocity_var=0 0 0
      ; Start size
      size=0.1
      size_var=2
      ; Growth of particle (m/s)
      growth=0.5
      growth_var=4
      max_alpha=0.2
      alphafade=linear
      ; texture=data/images/smoke.tga
      texture=data/images/smoke_fake_lighting.tga
      influences
      {
        airfriction=1
        wind=0.0
        gravity=0.1
      }
    }
    As you can see, they produce a bit more randomness:
    [​IMG]

    Then compare to this:
    [​IMG]
     
  12. It's not the same. The smoke has puffs and billows in reality, and the Racer smoke seems to spread out downwards through the surface of the track, unlike the real stuff which seems to react to the road. Real smoke is affected by the stream of air coming off the back of the car, it's stirred up in the wheel arches and poured around the sides. It doesn't just puff up, it whirls around.



    For what it's worth, I think your smoke does look better than the normal settings, but it's still just a representation of smoke rather than something really convincing.
     
  13. fsr

    fsr

    Sure, there are of course many factors, but simulating the airflow in the wheelarches to make nice smoke is a bit further up the road imo :p

    As of now, this is as good as it gets, and considering the limitations in the way smoke is implemented, I think it looks pretty good. Of course I would love a better simulation, where the smoke actually starts as millions of tiny particles pouring from the tire surface and creating beautiful smoke effects. That would require a hell of a lot of hardware power though :D
     
  14. KS95

    KS95
    RACER Moderator

    I know it's not gradual like that, at that point the tires are pretty damn hot.

     
  15. I agree, it's absolutely the fault of the way the smoke is implemented as a system, not the way that the system is set up. The more I look at real photos, the more diversity in smoke there seems to be. Sometimes it can look really heavy and dirty, other times it's almost creamy.

    When I illustrate it like this, we have something real, something realistic, and something wrong.
    They are reality, v050, and v0.8.29

    [​IMG]
     
  16. fsr

    fsr

    Well. the main difference between 050 and 0829 is the texture file itself. I still like the new smoke better :)
     
  17. Stereo

    Stereo
    Premium Member

    The texture is a little uniform, but shader would need a lot of work to make it look good with anything else, since they're all being drawn in the same orientation. Maybe replacement of the single smoke frame with an array of them, selecting one at random for each particle. And simulating actual smoke is a huge pain in the behind... I've seen papers on it and they're happy if they can get 4-5 frames per second on high-end CUDA processors. Usually with a static frame of reference and motionless source, convecting the smoke upwards (eg. a cigarette). To put that into Racer, it'd need to simulate the air movement all around the car, on a fine enough grid to look good (probably 0.1m or so), which is a pretty big deal when you consider how large the track is. Though I've seen some work on sub-voxel turbulence before the render pass, so it might be possible to raise that to 0.5m without totally destroying the sense of smoke. Unfortunately I saved all those papers on my laptop which is non-running at the moment.

    On the other hand, it would make car aerodynamics easier if we were simulating air movement :p no need to guess at drag coefficients and wing locations, you could measure directly how much air it's pushing out of its way.

    Maybe in another 5 years...
     
  18. It is gradual at first then once you get going it pours off. I am not sure how they do it exactly. They also have some form of tire wear..
    GTA IV does not generate as much smoke as all the pictures you have been posting. Perhaps that is what they think is real for street cars with street tires or perhaps they wanted more but had hardware limitations. *shrug* You would have to ask them.

    Either way, I really like how the GTA IV smoke is handled near the wheels. Looks good imo.

    I have the PS3 version of GTA IV and have grabbed some quick vids. I have to apologize for the crappy recordings, I just cannot find on short notice good recording software. FRAPS does not recognize my capture card's software and the software's built-in function complains about this being protected content. I am stuck with other lame stuff that has bad performance on my PC that can normally very easily handle this sort of thing.:frown:

    Anyway here are the vids!
    First just a little circle w/ only the driver's tire exploding in one of my fav cars:
    http://www.divshare.com/download/13854473-793
    Then along a street with me getting distracted by something tempting:
    http://www.divshare.com/download/13856060-04e
     
  19. Maybe it's more to do with what ur smokin :wink:
     
  20. Hiya FiftyOrange, good of you to illustrate this point. 050f is pretty real looking, and yet the 08.29 looks good too-but too uniformed. DavidI had told me to reduce my smoke settings for my ancient pc, and so my smoke (in 6.6.5.5) looks a little like the 08.29 and dissipates quickly like a breezy day :) but to an extent I can see why this is something that should be at least given some attention to, after all for the final-final version...it should compare to the "big name games". BTW, in some sections of NFS:Undercover there is a mist almost like a fog. The tire smoke is real- on my Wii...Whee! :)