Home Technical Talk

TOPOLOGY - Solving Triangles on Curved Surfaces in Sub-D?

polycounter lvl 10
Offline / Send Message
*Sebastian polycounter lvl 10


I want to model this concept in Sub-D. However, I've encountered a problem that I can't solve topologically.

Can someone help me understand how I can convert these triangles into quads without causing pinching? Some of these triangles are on a curved surface, which makes resolving them without pinching very difficult.

Thanks for your help!

Replies

  • iam717
    Offline / Send Message
    iam717 interpolator
    Should be a solution in here: (everyone should favorite it or book mark it, download it, whatever you can do.)
    Answer is there though, you review the imgs or read the posts.

    If you want me to try in text format, you need to "highlight the cuts, bevel them to make a loop indent it in slightly, and extrude again making an inner support loop", no triangles, everything quads, yes i know in the end its all triangles...
  • Neox
    Offline / Send Message
    Neox godlike master sticky
    What is your target here? Baked down in a normalmap? 
    If so, maybe consider saving your time on this (fun tho) challenge and just do it in texture.
  • HAWK12HT
    Offline / Send Message
    HAWK12HT polycounter lvl 13
    Pretty sure that end cap is on its own. Save yourself some headache and make it in two parts. If making it for game then alpha texture will do the trick. 
  • poopipe
    Offline / Send Message
    poopipe grand marshal polycounter
    that's a really good point actually - if the real world equivalent is made of multiple parts, you generally don't want to try and model it as a single part. 
  • *Sebastian
    Offline / Send Message
    *Sebastian polycounter lvl 10
    Thanks for the responses! I'm currently working on expanding my knowledge in Sub-D modeling. I've already completed some studies (https://www.artstation.com/artwork/K3yzlW / https://www.artstation.com/artwork/QKRXW8 / https://www.artstation.com/artwork/rl0x1a / https://www.artstation.com/artwork/Ovx6lg) and am gradually increasing the difficulty level. However, I recently realized that I'm still lacking some basics when it comes to mesh density and edge loops. So, I decided to go back to the fundamentals.

    Neox said:
    What is your target here? Baked down in a normalmap? 
    If so, maybe consider saving your time on this (fun tho) challenge and just do it in texture.

    Nope. Sub-D only.

    HAWK12HT said:
    Pretty sure that end cap is on its own. Save yourself some headache and make it in two parts. If making it for game then alpha texture will do the trick. 
    Unfortunately, it's exactly this headache I want to deal with. I want to be able to handle this complexity as well.

    poopipe said:
    that's a really good point actually - if the real world equivalent is made of multiple parts, you generally don't want to try and model it as a single part. 

    You're right. I've now decided to change the design to match the model to my current knowledge level.


  • FrankPolygon
    Offline / Send Message
    FrankPolygon grand marshal polycounter
    @*Sebastian Here's some topology routing solutions for slots that run across curved surfaces and need to resolve to quad geometry.

    Support loops and shapes that cross perpendicular to the edges of the underlying curve typically won't disrupt the segment spacing in a way that could cause residual triangles or undesired deformation of the surface.



    The number and spacing [density] of the edges that make up the curve tends to have minimal influence on the sharpness of the support loops around perpendicular shape intersections but having too many or too few segments in the curve can complicate the loop routing or limit the accuracy of other intersecting shapes and the visual quality of the underlying curve.



    As long as these types of perpendicular support loops don't intersect with anything else like other features, surface edges, etc. it's possible to deform the intersecting shape and the loops will still flow around it cleanly.



    Shapes that run parallel to the underlying edges in the curve are constrained by the curve's geometry and it's generally considered best practice to adjust the segment count of the curve so it's existing edges can be used as both the inner and outer support loops for the intersecting shapes.



    These types of shape intersections can usually be deformed along the radial axis of the curve but are still constrained by the adjacent edges that make up the curve. While it is possible to move or deform the shape intersection, within the confinement of the existing edges in the curve, doing so will tend to either deform the geometry of the curve or interrupt the segment spacing in a way that introduces undesired smoothing artifacts like pinching or stretching.



    When working with more complex shape intersections it's often necessary to combine both perpendicular and parallel loop routing around the base of the shape intersection.

    Loops that run perpendicular to the edges of the curve can usually be added easily but it's often necessary to plan ahead during the block out to ensure there's enough segment density in the curve to support both the placement of intersecting shapes and to carry any parallel support loops across the existing edges of the underlying curvature.

    Connecting the parallel support loops in the intersecting shape to the existing edges of the curve provides a place for the loops to run out without interrupting the segment spacing. Which helps prevent unintentional deformation and smoothing artifacts.



    Shapes that use this type of loop routing can generally be deformed the same way as shapes with perpendicular loops and inside corners can often be sharpened by moving the inner most edges of the support loops towards the center of the edge. Potential smoothing artifacts from this type of variable loop width are generally constrained by the inner and outer support loops of the shape and hidden by the edge highlight in the area being smoothed by the subdivision.



    It can be helpful to break down each edge of the shapes in the references and decide whether it needs to be supported by a parallel or perpendicular loop routing strategy. The basic slot shape shown previously can be deformed into something that's close to what's shown in the references. However the bottom corner still needs to be sharpened.

    This can be done by either pushing the existing perpendicular loops closer together (Which will likely result in a pinching artifact unless further simplified.) or the outer edge and its adjacent support loops can be joined together. (Which will likely require merging some of the loops to redirect the flow and simplify the mesh.)



    This is one example of what it could look like when merging the loops in the bottom corner. In some areas, like the inside of the slot, it will require additional edge loops to create a triangular or diamond shaped quad. It may also be possible to merge the top vertex of the diamond quad down into the base of the corner but this could introduce some unintended deformation. Which approach makes sense depends entirely on how tight the corner is and the relative flatness at the bottom of the slot.



    When creating complex shape intersections with repeating features it can also be helpful to isolate the basic shapes and do some simplified block outs to figure out the optimal segment spacing and edge density along the curve to align everything with the loops coming off the intersecting shapes.



    With this approach the more complex features are used to define the geometry density of the less complex shapes and while it doesn't completely eliminate the need to manually redirect the loop flow in some areas it does tend to reduce the occurrence of loop intersections that can't be easily resolved to all quads.



    Depending on what the end use for these models is, it's also worth mentioning that as long as the mesh smooths correctly it's often acceptable to leave some triangles (Or even well supported n-gons.) if they smooth correctly and the amount of work required to resolve to the mesh to all quads is greater than the benefit of having an all quad mesh.

    Recap:
    -Analyze the references to identify if the edges of shape intersections on curved surfaces require parallel or perpendicular support loop routing.
    -Adjust the segment density of curves so the existing edges of the curve act as or connect with the support loops of intersecting shapes.
    -Block out key features first to determine the optimal mesh density for adjacent curves.

    Some additional write-ups that cover loop routing on similar shapes and sharpening the inside edges of slots on curved surfaces, without interrupting loop flow along outer edges of shape intersections:
    https://polycount.com/discussion/comment/2733986/#Comment_2733986
    https://polycount.com/discussion/comment/2761060/#Comment_2761060
    https://polycount.com/discussion/comment/2780449/#Comment_2780449
  • Neox
    Offline / Send Message
    Neox godlike master sticky
    Frank is back! <3
Sign In or Register to comment.