Announcement

Collapse
No announcement yet.

Grid artefacts in smoke sim

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Grid artefacts in smoke sim

    hi there,

    I´m back with some more problems:

    I jsut simulated some smoe and I´m getting visible artefacts in the rendering.
    I´m using the trick with the peak transparency diagram to give my smoke more detail, I also disabled "light cache" in the colors/transparency rollout and tried decreasing the step size in the rendering tab as suggested in the manual, but none of that helped. I know I already had success with lower grid resolutions, but I don´t remember what I did differently...

    Here is a render with about 500 k cells:

    Click image for larger version

Name:	PHX_Smoke_Grid_Artefacts.jpg
Views:	1
Size:	69.2 KB
ID:	880941

    And here one with about 9 mill cells:

    Click image for larger version

Name:	PHX_Smoke_Grid_Artefacts2.jpg
Views:	1
Size:	76.1 KB
ID:	880942

    I also tried the tip from the manual with drag particles and rendeirng with PHXfoam as fog, but with the default settings (fog=1cm) I just got one blurry blob and when I started lowering the fog size to something like 0.01 cm, rendertimes got really high, I actually stopped the rendering so I coudln´t even see if it would in theory look good enough.

    So, where do I go from here?
    Increase grid resolution?
    Increase steps per frame?
    Any other solutions?

  • #2
    You might try to enable smoothing on the smoke channel from the Input panel and say how that looks.
    Svetlin Nikolov, Ex Phoenix team lead

    Comment


    • #3
      seems like you are using opacity curve, in the first phoenix versions we were preferring this method, until multiple complains about artifacts appeared.
      we changed the default opacity settings to be simple smoke, the opacity curve is as an invitation for the artifacts.
      about the particle method - the fog method can very easily overload the memory, that's why we introduced the point mode, use it.
      however in both cases you will need about 50M particles and more. you can see the "ink in water" scene that demonstrates this technique, and one more important point - prepare at least 500G free HDD space for the particles.
      ______________________________________________
      VRScans developer

      Comment


      • #4
        Oh, yes, this looks workable!

        Click image for larger version

Name:	PHX_Smoke_Grid_Artefacts_smoothing.jpg
Views:	1
Size:	54.7 KB
ID:	854581

        I´m losing some detail, but I figure I can play around with the smoothing settings and see if I can find a compromise there.
        Other than that, the only way to get rid of the artefacts would be to increase the grid resolution, right?

        I actually really like the look of the opacity curve, and the artefacts are also visble If a leave the opacity curve at its default value.
        I can also try to set the smoke to simple smoke instead of smoke.

        But I think I can work with the smoothed smoke channel, maybe add some drag particles and render them with point shader to get a "grainier" look.
        Last edited by ben_hamburg; 07-11-2014, 04:02 AM.

        Comment


        • #5
          Yup, and the artifacts sill won't go away completely after increasing the resolution. Btw, if you are using the official build or a nightly older than a month, you might want to try a new nightly - the smoothing has been optimized and will take significantly less time.
          Svetlin Nikolov, Ex Phoenix team lead

          Comment


          • #6
            Yeah, using simple smoke gets rid of the artefacts.... But I also lose a lot of detail:

            Click image for larger version

Name:	PHX_Smoke_Grid_Artefacts_simple_Smoke.jpg
Views:	1
Size:	53.5 KB
ID:	854582

            I also simulated a couple of frames with 50 mill cells and opacity curve:

            Click image for larger version

Name:	PHX_Smoke_Grid_Artefacts_50mill_opacitycurve.jpg
Views:	1
Size:	46.1 KB
ID:	854583

            So I guess I can answer my own question: higher grid resolution does get rid of the artefacts.
            So I´ll probably try to find a compromise between grid resolution, opacity curve and smoothing to get the best result with the least sim time and artefacts...

            Comment


            • #7
              Hm, not sure yet...did another test with 8 mill cells and now think I prefer this with simple smoke instead of opacity curve+smoothing...I think below this resolution I just don´t get enought detail without artefacts.

              One other thing: My scene scale is in real world units (1 cm), so maybe this is also problematic? I remember it beeing a problem with liquids.
              Wasn´t there a way to change the scene scale in the Phoenix sim without changing the scene scale in the whole scene?

              I´m on one of the latest nighlies (less then a month ago I think).

              Comment


              • #8
                You can multiply the scene scale form the Dynamics panel, but for smoke and fire it will just change the.. feeling of scale - the smoke will take more time to move through the grid with a bigger scale, but there are no problems like boiling liquids etc. with fire/smoke.

                5-10M cells for a smoke sim without artifacts is about the lower limit You may also try the Spherical sampler instead of the Linear (from the Rendering panel) if you are fighting minor artifacts.
                Svetlin Nikolov, Ex Phoenix team lead

                Comment


                • #9
                  hehe, yeah, always pushing the limit of whats absolutely necesseray. Gotta finish three more shots today, and Phoenix is only a part of them...I´d love the luxury to work in a team on shots like these...

                  Thanks for all the tips, I think I´m good now and I found a solution I can live with for the short amount of time I have.
                  I just remember having done another sim in the past, where I easily got away with a much lower grid resolution, but maybe I just dreamt that...

                  Comment

                  Working...
                  X