Experiments in creating necklaces etc for use in Poser's cloth room



  • @morkonan said:

    I used Looper/roper and some others in P11Pro a few days ago, testing them out. One of them had an error, but it wasn't anything much and a secondary script did the same thing, so if one fails, it's no real loss. I think it may have been one of them that had a more complex menu system in it, but the features are in one of the other scripts, too. I know I've used them in 2012 Poser Pro with no fails/hangs.

    Thanks will give them a run and see what develops



  • @amethystpendant said in Experiments in creating necklaces etc for use in Poser's cloth room:

    ...Also, it is supposed to be simulating real world cloth...

    The fact that necklaces aren't usually made of cloth is probably one of the main reasons people don't think of using the Cloth Room for them. But I came at this from the idea that a bead necklace is basically a piece of string. And a piece of string is basically similar to a very narrow strip of cloth. With a bead necklace the beads are usually spaced apart: they can be modelled and assigned to the "Rigid Decorated" group, and will be dragged along with the dynamic string-cloth. I think this was how the dynamic necklaces I mentioned in the writeup (see link in OP) worked. But with a chain necklace there's usually no string, which is why the imaginary string-cloth part of mine is made invisible in the render. Also the links of a chain overlap, so I put a bit of thought into the best way to position each link with respect to the invisible string-cloth quad representing its dynamics. But actual values for the dynamics parameters is something I didn't put much thought into - I'm more interested in whether the basic concept can be made to work reliably (the exploding links when I use the "Rigid Decorated"group is now the main problem).



  • Redoing the 5 pendant necklace so that the dynamic mesh is just made of quads (no triangles), and assigning the chain mesh to the Rigid (not Soft) group seems to significantly reduce the out-of-place chainlinks: I only see two in this render (note that I couldn't see a way represent each chainlink on the dangles with a single rectangle, so there's a difference around the dangleattachment points) - and I note that those are located where there are quads linked by a single vertex. I think my supposition that changing from all-quads-connected-by-edges to triangle-pair-diamonds-connected-by-single-vertices has something to do with the chain-link mesh coming apart if it's put in the "Rigid Decorated" group is worth investigating further.
    0_1486778842902_AllQuads.jpg



  • @3dcheapskate - thinking about it in terms of what the sim is actually doing with the chain-links in the Rigid group (and this post is all just supposition):

    Each individual chain-link (a mesh unconnected to any other) is positioned/oriented a specific way in relation to the initial position of the dynamic mesh. I assume that meshes in the Rigid group shouldn't deform/stretch - just move and rotate along with the 'underlying cloth'. So after each frame of simulation each chain-link mesh in my model should be somehow positioned/oriented in accordance with the underlying cloth. But which part of the underlying cloth, and how? My guess is that Poser might take the average of the X/Y/Z coordinates of every vertex in the specific chain link (in the necklace's initial state), determine the nearest point to this on the underlying cloth mesh, and work out how this relates to the normal and warp/weft directions of the fabric (i.e.a sort of underlying-fabric-space coordinate system for each separate part of the rigid mesh). After a sim frame this coordinate system has moved and rotated, and I guess the specific chain link mesh would be transformed in accordance with this.

    If that is roughly correct, then chainlinks would go astray if the normal/warp/weft-directions at the associated point on the underlying cloth weren't what I expected them to be.

    In this latest all-quads-for-dynamics-5-dangles setup it appears to be two links on the main loop which have dangles hanging from them that go astray. The underlying dynamic mesh around these areas is not the simple one-rectangle-per-chain-link used for most of the links, and in this area there is also a single vertex used to connect the dangling diamond to the two squares in the main loop.

    0_1486780524748_astray.jpg

    Further investigation required.

    P.S.I'm aware (from somewhere, can't recall where - Poser Manual? Secrets Of Figure Creation With Poser 5?) that the Cloth Room works best with a mesh of triangles (and aRtBee's manual investigates in somedepth how various ways of converting a quad mesh into triangles affect the sim). But if I use a mesh of quads does the cloth sim actually convert them into triangles first?



  • @amethystpendant said in Experiments in creating necklaces etc for use in Poser's cloth room:

    but there really isn't enough geometry to do a successful sim. Also, it is supposed to be simulating real world cloth.

    I've been pondering on these two observations,which seem difficult to refute.But Ithink I've come up with something ! ;o)

    Consider the steel-balls-and-springs analogy that's sometimes (often?) used to explain the maths used for cloth simulations. If you want to simulate a normal piece of cloth (be it a curtain, tablecloth, towel, toga, ballgown, etc) you'd definitely want a lot of geometry in order for this steel-balls-and-springs maths to make the mesh seem to behave like real world cloth.

    But if we forget about the real world cloth and consider just the steel balls (vertices) and springs (edges)implementation, then I think that we have enough geometry to simulate (very roughly) what's effectively a piece of string.

    Of course, I may be wrong, and I may have misunderstood something or made some false assumptions. But it's still looking promising to me.


  • Poser Ambassadors

    @3dcheapskate Thanks, interesting stuff .



  • @3dcheapskate said in Experiments in creating necklaces etc for use in Poser's cloth room:

    P.S.I'm aware (from somewhere, can't recall where - Poser Manual? Secrets Of Figure Creation With Poser 5?) that the Cloth Room works best with a mesh of triangles (and aRtBee's manual investigates in somedepth how various ways of converting a quad mesh into triangles affect the sim). But if I use a mesh of quads does the cloth sim actually convert them into triangles first?

    Simulations indeed work best with triangles. In quads the four vertices in general are not in one plane so there can be ambiguity about the shape of the surface between the nodes.The program therefore has to assume a shape step by step and this can lead to jumping between two modes each step and prevent convergence. If the mesh is triangle, there is no such uncertainty.
    To suppress instability in a simulation usually reducing the time steps helps and the increase of damping. Reducing time step makes the 'corrections' that are needed less severe and lets them act over a shorter range travelled so the amount of energy introduced in the system is (quadratically) reduced. Damping takes energy out of the system. Nature's laws stipulate that equilibrium is where the amount of energy in the system is minimal.

    I really do like the results with this gold chain necklace. I did try something similar I think in Poser 7. Results were such that I abandoned the project. Good to see this works our better now.



  • @F_Verbaas

    It makes sense that calculations are easier, and the results likely to be a bit more accurate, if the mesh is in tris. A tri is the lowest number of vertices a face can have and any three tris sharing edges are always going to be a face. All tris contain vertices that are also coplanar, by definition, and it is not possible for a tri to have its face's surface warped. Tris are easier to calculate lots of things for, vs quads, for this reason and others.

    For quads, higher mesh density helps to combat the problem, immensely. That's why dynamic cloth should be more dense than its conforming counterpart. (Helped by the new sub-d routines, this isn't such a terrible problem anymore.)



  • @F_Verbaas said in Experiments in creating necklaces etc for use in Poser's cloth room:

    I really do like the results with this gold chain necklace. I did try something similar I think in Poser 7. Results were such that I abandoned the project. Good to see this works our better now.

    Spurred by your comment I tried them in Poser 6. They seem to work fine (at least the three simple loop necklaces that I uploaded to ShareCG, already linked in OP). The sims ran smoothly, taking about 1-2s per frame for the smallest necklace and around 5s per frame on the largest one - that's on a several years old Dell Inspiron laptop. Had to adjust the gold material, but here's a render at frame 90:

    0_1486971536765_Render 2.jpg



  • ...and even the latest 5 pendant necklace (with chain mesh in 'Rigid' group) almost works in Poser 6, and only around 2s per frame. The orientation of some links seems to be wrong, and the central triangle pendant seems to have become detached from red triangle underlying it.

    0_1486973885987_P6.jpg



  • @morkonan/@F_Verbaas - regarding quads/triangles, I'm still puzzling over why a dynamic mesh made of quads seems to perform better than one made of triangles when it comes to attaching a 'Rigid Decorated' group to these necklaces. I think it's time to go back to the simple loop necklace, which used a loop of rectangles, and try it with various triangular meshes.



  • Test4DynamicNecklace (a Poser 6 PP2) uploaded to ShareCG if anybody wants to play.

    0_1486976433543_Test4DynamicNecklace.png



  • Here's a simple test I should have done much earlier. I started with the 128 link simple loop necklace, and did four additional variations of the red mesh:
    1 - The original mesh of rectangles connected by edges.
    2 - Mesh 1 but with each rectangle split into two triangles with diagonals all pointing the same way.
    3 - Mesh 1 but with each rectangle split into two triangles with diagonals forming a zigzag
    4 - Each rectangle from mesh 1 converted to a diamond and connected to adjacent diamonds by a single vertex.
    5 - Mesh 4 but with each diamond split into 2 triangles.

    0_1486982780072_BackToTheBasicMesh.jpg

    It didn't seem to make much difference (in this very simple test) whether the mesh was made of quads or triangles.

    But the single vertex connections between the diamond shapes in meshes 4 and 5 are probably making it difficult for the simulation to settle down. What F_Verbaas said a few posts back probably applies here too.

    For meshes 4 and 5:

    • Using a starting position that's much closer to the figure, and thus avoiding the excessive momentum that gets built up in the initial drop (as pointed out earlier),might help.
    • Finding a better set of parameters to use with them might also help.


  • The chain links also seem to distort when I use meshes 4 and 5 (diamonds joined by single vertices).

    0_1486986053351_distort.jpg

    When I use meshes 1 to 3 they don't.

    0_1486986075757_ok.jpg

    I think this is also probably related to the F_Verbaas'comments about the sim not converging.



  • @3dcheapskate said in Experiments in creating necklaces etc for use in Poser's cloth room:

    @morkonan/@F_Verbaas - regarding quads/triangles, I'm still puzzling over why a dynamic mesh made of quads seems to perform better than one made of triangles when it comes to attaching a 'Rigid Decorated' group to these necklaces. I think it's time to go back to the simple loop necklace, which used a loop of rectangles, and try it with various triangular meshes.

    Edge loop calculations? Cleaner material group borders? Dynamic items come in all flavors, some with better topology than others. IMO, anything "dense" that is either all tris or quads deforms fairly well, but quads are handled better in rendering in Poser. (Firefly)

    A sub-d will quad anything, but I've come across some terribad meshes that just couldn't be saved by the mighty sub-d...

    @3dcheapskate said in Experiments in creating necklaces etc for use in Poser's cloth room:

    The chain links also seem to distort when I use meshes 4 and 5 (diamonds joined by single vertices).
    I think this is also probably related to the F_Verbaas'comments about the sim not converging.

    Diamond-shaped faces are bad, in general. They cause puckering effects during deformations/smoothing and can cause the same during soft-shading effects. In normal modeling, they're to be avoided unless it's absolutely necessary. One can, somewhat, and only a little, counter these effects if the surrounding area is set up right, with a good general flow that helps to prevent those artifacts. But, it's darn difficult to do. If you're familiar with V4's mesh, as an example, there are several diamonds in key places, some apparently intentionally put there to contribute their effects to dimpling or linear deformation tendencies.

    But, IMO, diamonds suck... unless they're the only thing that can save your topology, then they... still suck, but are necessary. :)


  • Poser Ambassadors

    Clothified quads folded diagonally and rendered in FireFly.

    Ewwwwww.

    0_1487005817446_Clothroom Quads Folded Diagonally.jpg


  • Poser Ambassadors

    Compare with clothified triangles.

    0_1487005889819_Clothroom Tris Folded Diagonally.jpg


  • Poser Ambassadors

    Clothified hex triangles. No ugliness and it has the correct stiffness and shape that I wanted.

    0_1487005987379_Clothified Hex Triangles.jpg


  • Poser Ambassadors

    Note - my last three postings have nothing to do with jewelry. Just reacting to suggestion that we should use quads. I would suggest to AVOID quads unless you have a compelling reason to use them. I'm speaking here regarding 2D sheets of cloth, not necessarily applicable to a simulated 1D chain.



  • Agreed that quads in dynamic cloth are generally a bad idea. I've had that sawtooth effect in a few renders where I've used dynamic clothing made from quads, e.g. here http://3dcheapskate.deviantart.com/art/Spirit-Of-The-Wind-496304864 just under Lucy Liu's right breast

    0_1487006920601_jaggies.jpg

    Still not sure regarding this necklace thing though.