hi letterworks (and everyone else),

I’m tied up all the time, reviewing magz, keeping my website uptodate, learning Poser, doing stills, doing animation, doing tutorials and doing Rendo as well. No sweat, it’s just; what to do first.

I render the animation out overnight, there is still lots of room for improvement but it gives you some idea. Project 011 Scene 02 (these moves, this gown, 18 sec from 286 in total) Take (=version) 12. At first I had to get the lighting right (IDL didn’t work out, I invented softboxes etc), then make the moves, then conforming clothes didn’t work out, then dynamics showed all issues nanette is running into as well.

I found out that conquering Cloth Room is: take another perspective. While Pose Room – conforming clothes included – is anything Virtual, Cloth Room is everything Reality. 

Pose Room is everything about morph-dial and bone-driven mesh deformation without any link to reality. You can have a rope standing up and have Vicky walk right onto it to the top. You can conform figures A to B, which means that they share the same morph and bone drivers, and that the B-ones are master. So people conform clothes to Vicky, pose Vicky, and take a shot. But they also could had clothes conform to Vicky conform to a skeleton, and pose the skeleton. Or, conform the skeleton to Vicky conform to clothes, and pose the clothes.  It’s in the naming too. If a table has its legs named as a horse you can make it gallup just from your pose library. If the horse, or the table, has the legs named as Vicky you can make them wear jeanz, and walk in them. If you really take care, you can have the table wear jeanz and gallup at the same time, point and click.

I’m referring to this as people demand cloth room to recognize conforming clothes for further processing. Poser only recognizes bone-driven meshes and calls them figures. Non-bone-driven meshes are props. We regognize figures as clothes, beings, vehicles, buildings, gnarly trees, whatever. Poser doen’t. Skeleton conformed to Vicky conformed to Jeanz. Is Vicky clothes? How can Poser know?

Cloth Room is Reality. It has gravity and air. 

Take a piece of cloth, pump up the density (increase mass), zero the air-resistance to eliminate the air-effects, drop it from some height and look up your first class highschool mechanics formulas, and the Earth gravitational constant. They apply (with a quirk. Tell me which).  Check whether your animation speed is 30 fps, and you can predict when the thing hits the floor. Density is kg per m2. Office paper is 0,080, is on the package (80 grams it’s called, per m2 that is. A0 flipover/poster size is 1 m2). I’m a 6 feet guy , I guess my pants take about 1 m2 of cloth, when I put them on a kitchen scale it reads say 250 grams, so the Poser setting would be 0,250. Just measure up your own clothes, and you know.  

Air resistance is similar.  It’s the force generated by an air flow of 1 m/s through (or at) a 1 m2 cloth surface. You can measure it. Take a cylinder (Poser primitive), make it long enough, put it horizontal als a flag pole, and attach a piece of cloth. Just collide to the pole, and one row of vertices in the constraints group. Now you’ve got a flag. Drape it, after some frames it will hang down properly. Now you can apply wind force, from the menu, make it from aside. The wind will blow it aside, gravity will pull it down and the angle that results from the simulation tells you the relationship. Vary resistence and density and the angle will exactly vary as expected. 

Now redo the first experiment, take the cloth up but now you give it meaningful density and air resistance values. Like the default ones, which are for table cloth. Drop it, and note that it takes a bit longer to hit the floor. If you drop it from larger heights, and note the intermediate results, you can observe that it reaches a constant velocity going down. The cloth moves through the air, the resistance generates a force upward. Gravity pulls it down, and increses the velocity. Therefore, the air resistance force will increase, until it equals gravity. Then the forces are in equilibrium, and the thing will not accelerate any more. Just another way of deriving the air resistance value.

Under water, some cloth absorbs water and becomes heavy. Increase density a lot. All cloth will resist water far more than air. Increase ‘air’ resistance a manyfold. Apply the ‘wind’force generators at will. Now you’re into fluid mechanics, and able to dress up your mermaids.

Are cloth parameters weird and at magic scales? No, they’re right from the physics book, in standard physics units. Something with force per m2 in most cases. 

How many frames does one need to settle a piece? Like life, it depends on density and air resistance. Large density and low resistance will speed things up, as thing will come down faster. And vice versa.  

More on this tomorrow, the animation is still rendering and I’ll have to walk the dogs.

By the way, this was not about theory, This was about me discovering Cloth Room. It’s not Virtual, it’s Reality. For me this implies that when I get weird results, it’s either in the simulation algorithm, or in the interpretation of reality. I do need a proper understanding of reallife cloth, to bring Cloth Room to life.

In Pose Room, I am boss. In Cloth Room, reality is.  For some, Pose Room and Cloth Room are both about clothes, and one is an advanced case of another. For Poser, Poser Room is about figures, and just for me: some figures are clothes. For Poser, Cloth Room is about cloth simulation, and just for me: some cloth are clothes. I know about clothes, Poser doen’t at all.

For me, Dynamic Clothes relate to Conforming Clothes as a formula 1 car relates to a bike. Both have two pedals, gear switching in the steering wheel, sporty clothes are handy and for both a helmet is required. Both have their use, and Ferrari (read: SM) is not going to build a “ F1-car for bikers”, and a handy “F1 dashboard wizard” will never happen, I guess. I would appreciate it, but we have to learn to live without, I’m afraid. Sorry.

Till then, we’ll have to experiment, to find out, to share. This was my attempt, for today. 

HAH......!!!!! That was the little tidbit I needed! Thanks, aRTBee! Now I can start putting together a table of settings without a horrendous amount of simulate and guesstimate.

put them in a spreadsheet because you will have to correct them a bit after I've revealed the Gravity Quirk tomorrow (cliffhanger).

this is a GREAT thread .... GREAT ! I tell ya

aRtBee, thank you so much for sharing your experience and your insight. You are really helping me understand this stuff.  Yes, DPH and I fell into the quad trap, I probably led him astray, because I was overconfident. Also the little well-meaning tutorials about how you can adapt conforming clothing to the cloth room and quickly set up something as a hybrid. That is what we tried to do. It seems that is the road to frustration.

We tried an alternative skirt from Freestuff, made for Sydney. Just a cone with the top removed and made of small triangles. One run of the cloth sim for me with default settings, with the same settee and pose on Dave's Esther, and a perfect drape first time. No poke-through, no strange behaviour, and the preview consistent with the simulation results. Just one row of polys constrained around the waist and it didn't fall off. I think there is a lesson there. We are re-thinking the wardrobe for our story now. Fortunately David has awesome modelling skills, so we can fix meshes and triangulate them if needed.

What I don't understand is why during the simulation I got an almost perfect drape in preview with the original dress, and at the end, even if I step back to earlier frames, I get a totally different disaster. Surely that is a bug in the cloth room? It just seems wrong.

Quote - this is a GREAT thread .... GREAT ! I tell ya

Great because some great people with positive experiences are willing to share, for which I'm very, very grateful!

I'm learning this stuff myself, so don't have much to offer.

I like to figure out how things work by starting really simple.

I loaded a Poser box and a one-sided square. I rotated the square so it lays flat, and suspended it over the box. I enlarged it so that it's edges were outside the box. I then clothified the square and let it drop on the box.

It falls straight to the ground unless you start enabling some of those other collision detection options.

Further experiments led to some unexpected results.

A couple things I found out but don't really understand yet:

1. The square folded in half, diagonally, to varying degrees for no reason. In the worst case, it folded up a lot.

2. When it landed on the box, quite often the preview showed serious poke through. But on rendering, there wasn't any.

Quote - 2) When it landed on the box, quite often the preview showed serious poke through. But on rendering, there wasn't any.

 

That's really odd, I never would have thought to render when I seen poke through.

hi bagginsbill,

just a quicky at sunrise after breakfast.

1. did you really use the hires square or just a simple one? The latter is far too low on vert/poly density to do any reasonable sim testing. Hires cloth against lores cube is exactly what cloth room was made for, originally, so all defaults only should do fine.

2. preview has a problem handling the sim result detailts well. See nanettes post, and other threads. Don't know why, yet. Pokethrough in preview does not appear in render, and vice versa. Having no pokethroughs in preview just gives a nice starting point for drape, that's it. The good news: you can use quite low-quality renders, and area renders are real life savers out here. That's my current approach anyway.

general: you're in reality, aka the Poser understanding of it. If you drop cloth from a height, airstreams will take effect, especially at the edges, then creep inward. Air layers will make it hover just above the table, and make it slip at low table-to-cloth sheer resistance. Think about clothing the table, in the garden, at a quite summer day, and you warp the cloth from the second floor down. What happens? Think rough wood versus smooth plastic for the table, makes a difference. So: zero the airdamping to get rid of (most of) this effect. Poser will not let you reach real zero BTW, as one cannot have no air resistance in real life. Increasing cloth density makes it jeans like instead of thin linnen (the default values). You know from life what to expect.

happy draping :-)

Quote - It would be nice if the the cloth room could recognize conforming clothing, and simply wnhance it further, rather than insist on be a whole separate technology.  that it could simply read the different body parts of the clothing, and drap them around the body parts of the figure in a natural way.

dph

Actually, to some degree it can and does. Unfortunately it will only recognize a part of the clothing like a hip or Abdomen. On the other hand if all you want to do is drape a skirt hemline across a pair of crossed legs, it comes in handy.

Eclark, what DPH and I have been trying to do is exactly that: trying to drape a skirt over some legs in a sitting down position. Not even crossed legs, just bent legs. We thought it would be handy too. It didn't turn out that way.

The densities of the mesh in the skirt and the legs were about equal. Checking all the collision options made no difference. We are persevering though. It is essential for our project that we get this to work.

Quote - I'm learning this stuff myself, so don't have much to offer.

I like to figure out how things work by starting really simple.

I loaded a Poser box and a one-sided square. I rotated the square so it lays flat, and suspended it over the box. I enlarged it so that it's edges were outside the box. I then clothified the square and let it drop on the box.

It falls straight to the ground unless you start enabling some of those other collision detection options.

Further experiments led to some unexpected results.

A couple things I found out but don't really understand yet:

 1) The square folded in half, diagonally, to varying degrees for no reason. In the worst case, it folded up a lot.

The cloth simulation is based on vertices, not on polygons, so it really moves only the vertices of the cloth around, without caring too much if it actually forms a square. However, when moving the vertices around the square might get nonplanar, and has thus to be divided into two triangles. Note that a 3d renderer usually has a similar problem (nonplanar polygons), but it normally does not matter, since it is only interested in what is visible and what is not visible, so it can still maintain a continuous 4-sided polygon for shading etc. However for the cloth simulator, especially when polygon-collisions are enabled, The simple information on what is behind and what is in front is not enough, since the exact 3d-coordinates are needed. Hence the cloth-simulator really has to divide the n-sided polygons into triangles.

Quote -

1. When it landed on the box, quite often the preview showed serious poke through. But on rendering, there wasn't any.

I cannot say for sure, but the preview probably uses the triangulated mesh of the cloth-simulator, while the renderer (having more knowledge about where those triangles came from), can use the original (4-sided) polygons; i guess it does exactly that and renders the square and not the 2 triangles that were used to generate the deformation of the square.

Quote - At first I had to get the lighting right (IDL didn’t work out, I invented softboxes etc), then make the moves, then conforming clothes didn’t work out, then dynamics showed all issues nanette is running into as well

Yeah. It'll do that to you even when, you know, you've got the theory all learned up and stuff... Are you ready to join our ever-growing club yet, aRtBee?

John, you make me smile. Really.

I'm into massive computing, physics simulation included, since 1976. I'm into home graphics fiddling (graphc card drivers, POV raytracing) since 1986. I'm into more enhanced 3D (3D Studio, then MAX, then Cinema4D) since 1996. I won prizes, the shorts are on my (olde) website (the Mirror-in-Mirror one with the Smarts took over 1000 hours continuous ! rendering time on a 4 PC cluster for the final result only).

I also spend my time on Bryce, MojoWorld (panorama on my new site) and I started doing Vue some time ago. Recent result is in my Rendo gallery as well. I started a new site in April, and have over 120 manazine reviews and 1200 artist links added in already. Its growing 3 magz issues a month, and the editors are asking for more.

I also spend a few years on chairing and organizing a Dutch graphics community, with tens of active menbers, based on real (non virtual) gatherings and projects. As one of the outcomes I gave you my friend and rendo artist RocSerum for a full size front-page interview on this site (see the 2010/11 Winter Favorites post about Cornelis on my own website to save you looking for the links). The organizing etc (money raising!) took so much time that I had to stop it only to be able to do anything creative myself again.

In the meantime I looked at Daz Studio, and worked with P3dO developer Yarp to get the non-Poser DS filetypes into his tool, plus some additional features I needed for getting the grips on my 50Gb content library.

So, I rebooted Poser in July this year and restarted all serious learning from the bottom up. That's the way I work. At the bottom is my PC itself. As a result, I gave you a thorough PC-memory-management tutorial by Oct-1. It's in the Rendo Tutorial section as well, 150 hits from there, 900 hits in total. All other findings will end up in tutorials at the same level of quality. 

Working from the bottom up, I started Camera, then all Lighting and I'm into Posing and Animation right now. Reallife poses and moves for reallife characters at photoreal quality in 7000x5000 size stills and DVD-quality shorts. I'm about able to replicate Rianeli's photos in Poser, lighting (and shaowing!) included. Show me you cn do it too. It takes him half a day to get the lighting done in a real studio with a real model, you know. And all of it will end up in tutorials for you.

Cloth Dynamics is just a side kick at the moment, but I give it priority for you. I only spend two weeks on it, between finding out that Conforming was not the way in my short (mid September) and refocusing on Posing (end September) as that was my main issue in the first place. The animation I will show you (when the thing has finished rendering) and everything I present on this issue, is the result of only that, starting with empty hands two weeks earlier. Now you know.

This is spare time only, in case you wondered. I've got a fulltime day job as a project management consultant. And I'm into critical comments on gallery results, for those who like that. I'm into various forum debates, like a recent one on gamma correction and file types. I learned a lot, especially from bagginsbill, but from others as well. Perhaps I can give him something back on this topic, or anything else. I'm no guru, but I'm thorough and determined. And I will get there. With you, I hope.

So, dear John, you make me smile. No hard feelings, no annoyance. Poser has six rooms only and I will eat them one by one completely. Hence, this is your own question in return: are you ready to have me as a member of your ever-growing club? Then better fasten your seatbelts, as the engines aren't even warm yet. 

I'll be back with you this evening, over say 8 hours or so.

See you around, happy draping. 

Quote - hi bagginsbill,

just a quicky at sunrise after breakfast.

1. did you really use the hires square or just a simple one? The latter is far too low on vert/poly density to do any reasonable sim testing. Hires cloth against lores cube is exactly what cloth room was made for, originally, so all defaults only should do fine.

I specifically and intentionally used the one-sided square. After so much effort on reverse-engineering the material room nodes, I am convinced that if I don't understand how a simple scenario works in isolation, however contrived and useless it may be for real work, I have no hope of understanding a larger system of components interacting with each other. As an example, I never would have discovered how transparency and specularity interact with each other in the material room if I had not done precise measurements of specularity without any transparency, and transparency without any specularity. There are some surprising cases in the material room where .5 * 1 = .7. Similarly, my discovery of how to get at the W coordinate (of the U/V/W texture coordinate system) for dynamic hair was based on absurd and useless test cases that had no value in actually shading hair, but revealed a ton of information about how the hair node works. I now control the Hair node like nobody has ever done before.

The importance of these contrived and over-simplified scenarios, such as a single clothified quad falling on a box, is to verify hypotheses, allowing me to make accurate predictions of cause and effect. Then when applied to real use, I do not have to wonder at which of 20 factors is the cause of unexpected results.

In this case, I was trying to understand this: We have an expectation that enabling the additional collision options will do something. We assume we know what "object vertex against cloth polygon" means, but without testing that assumption, we can't explain to Nanette why turning all of them on still results in the skirt passing through the knees.

My single square test did something very similar in some cases, and not in others. Since I am able to reproduce the problem with even a single polygon, it stands to reason that the solution will likely be revealed by manipulating that single polygon and its cloth settings. I have the benefit, as well, of a simulation that finishes in .2 seconds, allowing me to do hundreds of trials in an hour instead of just a few trials.

Quote - 2) preview has a problem handling the sim result detailts well. See nanettes post, and other threads. Don't know why, yet.

Right - that's why I mentioned it. Until I said it, nobody else had. My point was not that I have an answer to why, nor do I need one. The important thing to know from the single poly case is that you cannot trust what you see in preview. Don't even bother posting preview screen shots to show a poke through problem. The preview is somewhat irrelevant. Always show a rendered version of the problem.

And, related to making the rendered geometry visible to each other in test cases, it's a good idea to use some standardized materials that are designed specifically to make contours easy to see - perhaps even some false-color ones would be in order. I'm developing these to share so we can all better understand what we're looking at. Frankly, the semi-transparent and textured skirt Nanette shows is very difficult for me to see what sort of mis-folding is happening there.

Quote - general: you're in reality, aka the Poser understanding of it. If you drop cloth from a height, airstreams will take effect, especially at the edges, then creep inward. Air layers will make it hover just above the table, and make it slip at low table-to-cloth sheer resistance. Think about clothing the table, in the garden, at a quite summer day, and you warp the cloth from the second floor down. What happens? Think rough wood versus smooth plastic for the table, makes a difference. So: zero the airdamping to get rid of (most of) this effect.

Right - I did set the air damping to zero. In fact, I set everything to make the square very rigid and super heavy at first. It fell pretty flat. Later, still with no air damping, I managed to get it to fold up, not because of air damping, but because the detection of intersections between the clothified square and the box were not behaving in a way that would be classified as "reality". Specifically I saw a slight bend that should not be there, and then I did things to accentuate that anomalous bending until it was truly absurd. I would post what I did, but I forgot how I did it after so many trials. In this early stage, I'm pretty much just creating questions, not answers - questions that we didn't even know to ask.

Example: The top of the box is a regular grid of square polygons. At the time when the cloth square strikes it, it is very close to flat and all of the box polys are well inside the square. Why did some "catch" and others not?

Quote - Poser will not let you reach real zero BTW, as one cannot have no air resistance in real life.

I'm not disputing that (yet) but I have to ask how you know that assertion to be true? And I would like us to be really careful (instead of careless) about what is meant by 0! We all know that .000001 is not zero and if that is the lower bound it is accurate to say the air damping cannot be zero. But is it relevant? Is it measurable? (In my work-place, every statement we make and act on must be clear, accurate, relevant, and measurable.) If the minimum air damping is such that it alters the position or velocities by one part in a trillion, then that's certainly not something to argue is the cause of a major folding in half of the quad. It's not relevant if the non-zero true value is unmeasurable.

I would like to see an experiment done in which the falling cloth time is demonstrated to be other than the solution to 1/2 a t^2 = H, where a is acceleration of gravity, t is time, and H is initial height. If you can show me that the expected fall time is off by a measurable amount even when air damping is set to 0, then I'll believe that it cannot be set close enough to 0 to be able to exclude that factor from consideration when performing an experiment.

Quote - Increasing cloth density makes it jeans like instead of thin linnen (the default values). You know from life what to expect.

That's an interesting statement, but here's my counter. I "expect" that when I set up a single polygon 10.25 inches on a side that weights 50 Kg, has the maximum possible stiffness (fold resistance, stretch resistance, etc.) and I set air damping to 0, that it does not fold up at all. That is not what happened.

(Sorry if this Batman reference is unknown to non-Americans. Trust me, that was funny.)

Some things to note:

Compare the intersection curve of the render to the preview. Totally different.

Look at the box side that pokes through the square. The square is not reflected - we see reflections of the floor instead.

Quote - Riddle me this, Batman.

(Sorry if this Batman reference is unknown to non-Americans. Trust me, that was funny.)

 

teehee

:O)

Quote - Rendered.

Some things to note:

Compare the intersection curve of the render to the preview. Totally different.

Can't we chalk that up to smoothing?

Laurie

Quote - In this case, I was trying to understand this: We have an expectation that enabling the additional collision options will do something. We assume we know what "object vertex against cloth polygon" means, but without testing that assumption, we can't explain to Nanette why turning all of them on still results in the skirt passing through the knees.

Any time I do a cloth simulation, I always make sure all 3 of those boxes are UNchecked, and I have never had any problems with clothes poking thru like Nanette. During the sim, it maky look like there's poke-thru, but that's just because the figure is usually lagging 1 frame behind the cloth during sim calculations, then, when finshed and I scan all frames with animation slider, it looks perfect in all frames. Just make sure there's no initial poke-thru in your modeling and all should go well. I am still learning myself about the cloth room, but I seem to be more successful at it then others and I believe it's because some people overcomplicate things. I only know what works for me, and I would think it will work for others as well.

Oh, you're just lucky BR ;o)

:oP

Laurie

There is no doubt of the reason those options are options - they are time-consuming techniques that are usually not going to affect the outcome.

However, it's really wrong to assume they're never needed.

Here's what happens in my test case when none of those collision options are checked.

The square has four vertices. None of these ever comes in contact with the box polygons. Therefore it simply falls to the ground.

This is the most common form of poke through - caused when a relatively large cloth polygon passes through one or more small figure polygons that are completely contained within the cloth polygon boundary.

The solution should be to enable "object vertex against cloth polygon" (let's abbreviate this as OVACP.) And sometimes it works, and others not.

I don't have all my ducks in a row, but OVACP does not behave the way I expect. Is it a bug or are my expectations incorrect? That is the first question that needs to be answered, and it leads to many other questions.

If OVACP cannot be trusted, what's the point of tweaking the parameters - you're going to be dealing with poke through perhaps in any case and pushing the parameters to try to stop the poke through will likely make the cloth behave strangely in other ways. Finding out when OVACP fails, and how to deal with it is key in my opinion.

For those who have the option to model, understanding what makes OVACP work or not work is going to be very valuable information. How valuable is personal, and depends on the value one places on a day's work.

For those using a 3rd party mesh, being able to look at it and in a few seconds decide if you are going to able to use it clothified or not is also valuable. Right now, nobody can give any definitive information on this question.

(Related to this is our general knowledge that long, thin polygons will often create artifacts in Poser. We know to model for Poser and that means don't do that.)

If rendering with and without smoothing answers that question, then no. My previous render had smoothing on. This one has smoothing off.

Smoothing applies a smooth, curved displacement interpolating across inter-polygon boundaries, within a single welded mesh containing at least two adjacent polygons. This is a single polygon, so smoothing cannot affect it.

Hmmm...wonder what in the world is going on then ;o) Does Poser do some sort of internal triangulation? If it does, than Poser would see that as two triangles. Or at least treat it as if it were two triangles anyway ;o).

Laurie

I can't recall if someone authoritative like stewer ever gave info on internal triangulation or not.

I have to get to work, but I have an experiment in mind that can perhaps give some info on that.

Suppose internal triangulation does happen. Given a quad, there are two diagonals available to minimally triangulate. Another possibility would be that the quad is made into four triangles meeting at a new hypothetical centroid. (I doubt this but we can test for it.)

Now if you were going to build the triangulation code, you'd have a choice of choosing the diagonal randomly, or choosing it based on vertex order. The latter is more likely. A third, very unlikely choice, would be to use world coordinates to make the choice.

All of these have differing implications that can be tested.

Suppose instead of using the Poser primitives, I build some procedurally, with great precision.

Suppose I build, from identical vertices in world space (A, B, C, D), four different quads:

ABCD

BCDA

CDAB

DABC

If I were to simulate and render these, and I get different outcomes, even though the four points are the same, then we know vertex order is a factor, and further, we'll probably get some evidence of triangulation choice method for one diagonal or another.

Further, we can rotate these so that they align with the other variations. This will reveal whether or not it is model space or world space that influences things. Doing this requires great precision (rotating about the center of the square very carefully) so I would not attempt this with dials. I would do it procedurally, via a script.

We'll see what that reveals.

it's rough and ready so don't be to hard on me - tis a quick test to see what line of modeling would be the way to go.

setup - 4 prop boxes, 1 ground plane.

1 single Quad plane

1 quadded plane

1 tri'd plane (the Quadded Tri'd)

1 X Tri'd plane (as above)

all set to : 10 frames drape. 30 frame anim, both collide - Object vertex against cloth polygon & object polygon against cloth polygon - set to on. all other settings left at default.

each cloth set to collide with it's box and the ground plane.

once again, this is a rough and ready test, not definiative results.

Actually it kind of does. Note that Firefly is based on a REYES-algorithm (which i know from the internet and therefore must be true), which essentially means it subdivides your geometry always, you like it or not. If you have an older non-REYES (ie. Poser 5 or so) around, you can try render with that and it should rather look like the box on the right or like your preview. (Image made in blender, not REYES, but allows to subdivide manually[left box]).

However in any case the firefly renderer does not show what the  simulation calculates, so the preview is this case more close to the simulation than the render.

You are quite right to point that out, MG. REYES micropolygon generation (is that called the slice or the dice - can't remember) is certainly at work in the render and not the preview. I made the communcation assumption that Laurie was referring to the "smooth polygons" render option, because she used the word smoothing. Testing for MP dicing is easy - change the min shading rate to a high number.

That might be the reason why turning the shading rate off and rendering, I still get some smoothing...lol. Still thinking tho, that Poser might be doing some trangulation of it's own, because how else would it handle non-planar quads? Every figure is bound to have some of those, don't they? not a modeler of figures

Laurie

Yes, you are right. Setting the shading rate up, gives the preview or the sharp L. I am not sure what the shading rate actually is, clearly not pixels; The image is made with a min shading rate of 50000, 5000 still shows interpolation.

I believe twisted quads are actually treated like a membrane - they really take on saddle shapes. In the renderer, not in preview.

On the previous page, MG pointed out (and I think it's true) that the cloth sim is just moving vertices with simulated springs and tensioners between them and in a strict sense triangulation doesn't have a lot to do with this.

However, in order to perform the vertex against polygon collision tests, particularly the OVACP, it has to consider the twisted shapes of the cloth polygons during the simulation. Does it treat them as triangles or as membranes? If triangles - which kind? Does it choose one diagonal or another, or does it test against BOTH triangulations? (That's what I'd do.) Or does it have some in-between thing it does with pretending that it's dealing with a membrane, but not with the high density that the renderer does?

Lots of questions.

Of course if you strictly model in triangles, all these questions are moot! You control everything and there should be no confusing poke throughs with OVACP.

I've certainly been getting better results with an X tri'd model for dynamic cloth than I have been with quads ;) At this juncture, I'm most interested at what Poser is doing internally with cloth more than anything...lol.

Laurie

This is a REYES parameter that determines the average area of the micro polygons. I set it to 20, the largest permitted value. This makes them roughly between 4 and 5 pixels across - big enough to see.

Looking then at the intersection of the poke through, I expect to see some straight segments. I don't.

PS: Shoot me now. I'm really surprised at how little I can predict in Poser. Obviously it's incredibly hard to figure out what is truth.

Poser's only predicable behavior is its unpredictability...lmao.

Laurie

Quote - I tested increasing the min shading rate.

This is a REYES parameter that determines the average area of the micro polygons. I set it to 20, the largest permitted value. This makes them roughly between 4 and 5 pixels across - big enough to see.

...

I case you do not know: you can set the shading rate to larger values with Dimension3D's 'Render Firefly'-script that comes with poser.

Quote - > Quote - I tested increasing the min shading rate.

This is a REYES parameter that determines the average area of the micro polygons. I set it to 20, the largest permitted value. This makes them roughly between 4 and 5 pixels across - big enough to see.

...

I case you do not know: you can set the shading rate to larger values with Dimension3D's 'Render Firefly'-script that comes with poser.

Aha! I did not know. Eeeexcelleeeent.

Hi folks, back again, interesting new posts.

I’ll answer to them in this post, and a next.

But first of all, here is my animation as promised http://www.artbeeweb.nl/artbee0110212.wmv =  3Mb wmv http://www.artbeeweb.nl/artbee0110212.avi = 22Mb avi (cinepak)

both medium quality, for some reason I couldn’t get DivX working and high quality took too much time. I needed Poser back for answering Bagginsbills remarks.

Second, completely off topic, for all of you having fun with simplifying things, this is the best short I’ve seen in some time. I ran into it yesterday. Now I like cubes even more. Pixels, by Patrick Jean - http://vimeo.com/moogaloop.swf?clip_id=10829255 

Bagginsbill really started to cut the beast to pieces, in order to get a thorough understanding, and publish every single step. I couldn’t agree more on this principle, I look forward to everything coming out of this, and I’m happy to support. 

= = =

While he starts with the simplest cloth object possible, I prefer the hires plane for a good reason. That is, we are into a simulation of reality. And reality is a very fine grained or even continuous thing. Even when you divide a m2 of cloth into 10.000 quads of 1cm2, then still 1cm is 100 times as wide as the 0,1mm thickness of a thread the cloth might be made of.

Next to that, simulations in general have a problem: they are inaccurate and unstable at the boundaries, the edges (years of personal experience and observation. I got my MSc on this.) 

Good sim routines are succesfull in limiting the inaccuracy, the instability, and prevent it from rippling inwards. Note that the Poser sim routine might not be the best available.

One way forward is to get rid of the edges, by performing calculations on a sphere (meteorologists can do that), or by extending the object at hand (financial planners can do that, they move the sims end out of sight). A second way forward – since edge width is determined by the number of nodes/vertices and not as a portion of the overall object – is to raise vertex density at the edges. The last way forward is to distrust any conclusion derived from anything else but the inner 2/3 of the simulated thing.  

My issue with the ultra simplified 4-vertex cloth approach is that it’s edge only. This might turn out great, I’m really curious how the story develops. But in the meantime I do ask myself whether this approach is really going to help us at larger scale issues.

While writing this, I realize that there is a scale difference on scene effects between nodes in Material Room and vertices in Cloth Room. Material nodes do have an overall effect, they can change the appearance of a whole object. Like a raw potatoe effects the menu. Therefore you need to know how to cook the potatoe in order to get your third star from Michelin. But cloth vertices are more like the cells in the potatoe.

Of course, microbiology will help us to understand how to cook our potatoes better. But this does not mean – for me at least – that doing microbiologic research yourself will make your meals tastier, and bring you that extra Michelin star. Good cooks will have some basic microbiologic understanding, and some microbiologists might be proper cooks. But that’s about it.

In other words, although I welcome the detailed approach I’m not that sure it might give us the results as we got from a similar approach of Material Room. Like Pose Room, Material Room is about Virtual, while Cloth Room is about Reality.  = = =

I do like Bagginsbills materials suggestion, but as the thread has developed since then it might be questionable whether the renderer itself can help us in our understanding of cloth sim details. 

Having that said we should realize that in the end only the renderer result is what we work for. Whatever it takes, the sim result should look good in there and nothing else.

Then bagginsbill observes anomalities. His rendered results are clear to me, as far as the cloth behaviour goes, sorry for that. The quad (was it really?) had been divided into two triagles, and a mild fold has become manifest as a simulation edge effect, as mentioned above.  Again, any clear understanding will be an understanding of the sim edge artifacts only. I’m afraid it won’t help us out on the mainstream issues.

= = =

I’ll address the gravity and zero air issues later, tests are running now to meet bagginsbills high  standards. Hope I’ll pass.

To BionicRooster I can only say: just look at one of my previous posts in this thread. When I suffer poly-explosion or sim-time explosion, the checkboxes are the fix. If it ain’t broke, don’t fix it.

Indeed, when you have no initial poke-through all goes well. As long as you stand still or move so gently that any pokethroughs can be fixed by the sim routine. Buth when you enter the danger zone, by more exaggerate moves for instances, the routine does well with some side arms.

happy draping

OK so now we pretty well understand why twisted poly pokethrough looks the way it does. Moving right along.

Moderators: Please sticky this thread.

Been playing with dropping a cloth on a chair.

However, no matter what I've tried, the cloth always passes through itself somehwere in this test. You can see an area where this happened here near the closest seat corner.

hi all,

I spend last evening and this morning on running tests on gravity and air-resistance, and more. I'll share my findings in this and consecutive posts, interesting stuff ahead. I reran all my original tests with better controlled settings, and I do have to admit that not all my opinions could hold. That's good.

Why are gravity / density and air-resistance important (at least, for me). Because a lot in Cloth Room is about the interaction between the sim routine and 3D meshes, that's virtual world. And most parameters are about cloth interacting with itself or the collision target. That's halfway virtual and real. But density and air resistance do exist since the Cloth Room itself is filled with gravity and atmosphere, which cannot be turned off. These factors not only are the main drivers behind the Cloth Room sim anyway (without gravity no drape!!) but they are (the only?) hard link from Poser into our real world.

So while some of us seek more understanding at the vertex level, I seek my first understanding in this area. That's why.

So I investigated on:

• does Cloth Room gravity really match Earth gravity? Yes it does.
  So you can continue to derive cloth densities from your own clothes, armed with a ruler and the kitchen scale. 
• does Poser handle gravity consistently? NO IT DOES NOT.
  Our pet monster has a Big Gravity Quirk hidden inside. It harms doing animations in a storytelling Poser scene. I don't know why it's there. I guess it's not a bug, but a feature :(
• does ZERO air resistance mean ZERO indeed? Yes it does.
  I was at fault there. I could not reproduce my earlier findings.
• does the sim routine has serious edge artifacts? Yes is has.
  It might hardly effect stills, it might have a mild effect on normal results in animations, but it will definitely trouble bagginsbill in his fundamental queste. 

I did detailed measurements under controlled conditions, and I did renders. Here they come, in separate posts. Sorry for flooding.

Does Cloth Room Gravity match Earth Gravity?

In short, I took a piece of cloth, turned it into lead, disarmed all air effects, dropped it from a height and measured positions at all frames. This gave me velocities, and acceleration.

Earth gravity acceleration is sort of a constant. Converting 1 sec to 30 frames gave the Poser equivalent. Measured acceleration and (converted) Earth acceleration were the same, within (5%) measurement accuracies. 0.01092 vs 0.01089, less than half a percent difference anyway.

This implies that we safely can get our rulers, scales and books out, to obtain values for Poser Cloth parameters. 

By the way, I just found out that the best website ever on Poser cloth, in the good old days that was: www.poserfashion.net, has disappeared. I therefore found out that the main pages of that site, are in my Library. So I will do you a cloth parameters table.

In long.

Piece of cloth is the hires square, raised to 19,6 mtr to give it exactly 2 sec = 60 frames for impact on the ground at 9,8 m/s^2 Earth gravity. Sim is set accordingly, all collision options switched off. Cloth parameters were set to max, thats 1000 for fold / sheer and stress resistance, and 1 for stretch damping and cloth density. cloth / static / dynamic frictions were set to 0, as was air-damping.

Reality test: a pack of 500 sheets of office paper measures 5cm, so 1 sheet has a 0,1mm thickness. Lead has a specific density of 11340 kg/m^3, therefore a sheet of lead as thick as office paper will have a density of 1,1340 kg/m^2. A similar value can be obtained with a sheet five times as thick (half a mm) and a specific density of 2000 kg/m^3. This is what stone (as used in corn-mills) is about (all: wikipedia). So, now we know what density 1 stands for.

The sim was run and as expected, it took about 60 frames to hit the floor.
Then I used a box to measure positions per frame, I could find no other way to find out. This introduced some measurement errors, for rounding (3 decimals = mm only), for the Front Cam is not good in spotting falling horizontal cloth position (no mesh thickness) and because the sim artefacts coased irregularities at the cloth edges.
Then I calculated velocity, according to V(f) = [ H(f-1) - H(f+1) ] /2, H= height, V=speed, f= frame number. Note that when you take adjecent frames, you get the speed at halfway the frame which is a minor shift in time. This way is more accurate.
The I obtained acceleration the same way: A(f) = [ V(f-1) - V(f+1) ] /2. It varied a bit due to the measurment errors mentioned, but did not fall down so air resistance was taken out indeed, within accuracy.

Note that due to this method frames 1,2, 59 and 60 (impact) could not produre meaningful results for acceleration. The average result over all other frames read 0.01092 m/f^2. The measurement error was obtained as about 5%.
Thats in frames time. The sim was done at 30fps, so I checked whether changing the value in the Animation Window made a difference. It did not. For the sake of it, I checked whether the scale of the cloth made any difference. It did not. 

Earth gravity reads 9.800 m/s^s on the Earth surface. There are differences, it's 9.832 at the poles and 9.780 at the equator as the Earth is not a perfect ball, and it varies a bit with underground and surroundings too. But 9.800 is a decent average.

That value, converted to 30fps, equals to 0.01089. 

And although the measured value resembles more to the gravity found at the poles or in the Rocky Mountains than with Earth average, in my point we've got a match. 

Poser Gravity Quirk Around

Poser does have an alternative for gravity calculations. Its in menu > Scripts > Utilities > Gravity, at least in mine. 

So I took a cube, raised it 19,6 mtr as in the cloth experiment, and fired the script. It created a nice set of Y-Move keyframes, the animation window showed a nice parabolic curve and all positions could be read out as Y-positions on a per frame basis.

I could see the cube felt faster than the cloth (I put them next to each other) and the cube hit the ground at about frame 56 instead of 60.

So I did velocities and accelerations again and I measured a gravity acceleration of 0.01310 with an accuracy within 2% (as I could read out the positions, measurement errors were less than with the cloth).

So, this way of dropping things if off the Earth gravity constant, and hence off the Cloth Rooms value too. This means that when you've got cloth and props and drop them both in a no-air environment, the cloth will lag behind for this reason only. So that's the quirk. It annoys me. No one uses the script for stills but this effect is considered harmful for meaningful animations in scenes with cloth and props. Like in story telling.

Where does this value come from? I don't know for sure, but I do know this:
Earth gravity converted to 30fps reads 9.8 / 900 = 0.01089
Earth gravity converted to 25fps reads 9.8 / 625 = 0.01568
The average of those reads 0.01329, which is within the 2% accuracy of the measured 0.01310.  So my wild guess is that someone out there could not decide how to convert earth values into Poser, should I take NTSC/60Hz/US-UK-based 30fps or should I take PAL/50Hz/French-Europe based 25fps? So he took a middle position. It's not a bug, it's a feature.

But unfortunately, someone else out there decided upon 30fps for the Cloth Room.

I don't know, I'm not pointing fingers. perhaps some can just change this one parameter in the script.

happy dropping :-)

Can Air Resistance really be zeroed out?

yes, just type 0 in the field. I incorrectly took the edge distortions (see nexts posts) for air stream / turbulence effects without further investigations, and perhaps there were some mishaps in my original measurements. So, to bagginsbill, sorry for this and thanks for your remarks.

Resizing the cloth did not change the previous results. This does not say much, as an increased surface also will give an increased mass, and effects of air are made by the ratio of damping over density. twice the mass and twice the damping will give a same result.

So I brought down density from 1 to 0.01 and then to 0.0001. It made no difference, same drop-down positions, speeds and so on. So, when you set air damping to 0, there is no additional residue distorting the results.

Air damping is proportional to atmospheric density, and to the relative speed of the cloth to the air, perpendicular to the cloth. You can move the air by using wind generators, and you can move the cloth by animation. By reducing air damping you can make the cloth move as on the moon, or in vacuum (nice for poser testing). By raising the damping you can make the cloth move like under water, or like on Jupiter. My totally unsupported current wild guess it that Cloth Room has an atmospheric density of 1 atmosphere, as on the Earth surface at sea level. I'll have to find out.

The catch is this: for proper Moon or Jupiter scenes you've got to alter Cloth Room gravity too. No way to do that, at the moment. A similar thing holds for under water scenes, as water provides an extra upward force proportional to the volume of he object. Food for thought.

Quote - I've been experimenting a bit with various mesh topologies. I don't have a definitive conclusion yet, but I'm leaning toward homogeneous (all quads cut on same diagonal) triangles.

I've come to the same conclusion after some experiments last night. At first a friend and I thought X tri's but nope...diagonal tris seemed to work while causing the least concaving.

Laurie

Edge Artifacts in the Sim Routine

Simulation routines do hate edges. Of course this is crystal clear to you, but for those who want to tell their friends, this is why. You can see some example in the way I calculated velocities and accelerations in the gravity post above.

The velocity in frame 2 was derived from the positions in frame 1 and 3. But since there is nu frame 0, the velocity at frame 1 either cannot be determined or has to be derived in another way. Frame 1 is the edge and so is frame 60, the edge  width is 1 (frame).
The acceleration, the 2nd pass in the run, at frame 3 was derive from the velocities at frame 2 and 4. So, the acceleration at frame 2 became at question given the debate on the velocity in frame 1, and the acceleration at frame 1 became questionable even more. The same holds for frames 60 and 59, and the edge width has become 2 now.

Because I only needed one average value I decided to ignore the values in the edge, but for cloth sims this is not the best idea. You do want the edges of a dress or towel look good, do you? And the sim runs hundreds of times, as per frame, and steps per frame, and internal iterations per step. You can't afford to loose all that info. 

This is the main puzzle for simulations. Not the algorithm itself, but how to deal with the edges. how to keep the edge width limited, how to reduce the effects in the edge area, and how to prevent the edge effects from rippling into the main area. As I was loosing 1 frame at each pass, on both sides of my calculation. After 20 passes I would have lost 40 frames out of 60, and had to take the average of only 20 values. This would have raised my measurement error, made the result less accurate and less fit for decision making, to say the least.

The Poser Cloth Room sim has to face two kinds of edges. One: the boundaries of the cloth mesh. Two: the limits of the parameter values. They're different kinds of limits, but anyway.

So I ran the sims for gravity and air damping investigation. The cloth did not collide to anything but the ground in the last frame only, and as discussed above air effects were sort of ruled out. Initial density was 1, the lead sheet.
During the entire fall of the cloth, the edges showed serious effects (all renders will follow in the next posts). At the touch on the ground, a few of them faded out but most others remained. Essentially, the effect was on the front edge only. About.

But, when I did the air damping sims at low cloth densities (0.01 and 0.0001 instead of 1.0) those edge effects were completely absent, at any frame. So I started in alter the density value from its upper limit (1.0) down. At 0.9 I found a reduced distortion, on 0.8 I found a further reduced distortion but shifted to the side edge. At 0.7 the further reduced effect had shifted to the back edge, and at 0.5 the effect was almost gone at all.

Hence, my impression is that the Cloth Room sim is not only sensitive at the cloth boundaries, but might be sensitive to the limits of the parameters as well. And why did the effect remain after dropping on the ground, I had expected the cloth to flatten out.

At second thought, the fold, sheer and stretch parameters were still at their very max. The cloth stiffness might have prevented the final settlement. So, I reduced those parameters as well, from 1000 to 10 while keeping density at 1. This made the effects disappear indeed after the ground impact, at max cloth density. They did not disappear that much in the earlier frames, while falling down.

All effects gradually grow from the edges in, from about the start till ground impact, and tend to worsen over time.

So, what can be inferred form this?
To me it means that extremes have to be avoided in parameter values, I'll do with the lower half of the allowed ranges. It also means that some artefacts which are seen in my results and others might originate from these edeg anomalities. In a still, one has the option to make them fade out by just adding a few frames after the final pose is reached. I don't know. It also implies that density might be the final dial to iron things out a bit. Might. Don't know for sure. Further investigation is required. Will do, in due time.

To me this also means that experiments with low resolution meshes, and parameters at their limits, might produce anomaly-driven results only. They may be of limited value for our further understanding and solution of the issues that were flagged. Any extrapolation should be made with great caution. Research at the vertex level might reveal interesting results, but might turn out to be th wrong way home as well. There is a health warning in there. 

Time for pictures, and time to stop. From now, I will focus on the issues as in the title of this thread, and just try to shine some light into the Cloth Room in general.

Green - in all further images - cloth at extreme parameter settings (sheer damping 1000, density 1), showing artifacts at the edges. Yellow - a copy with adjusted parameters.

This image: extreme sheer etc settings but density at 0.0001. This is frame 100, after ground impact at frame 60. Note the Green cloth still shows considerable effects while the Yellow one is about flat. Note that the Green sheet shows effects at the front side only.

This made me infer that the simulation suffers from edge effects, and that density, or the extreme setting of it, might play a role. 

Green - in all further images - cloth at extreme parameter settings, showing artifacts at the edges. Yellow - a copy with adjusted parameters.

Similar image, but now density is up to 0.5 (Yellow). there is a minor glitch visible halfway the front edge.

Green - in all further images - cloth at extreme parameter settings, showing artifacts at the edges. Yellow - a copy with adjusted parameters.

similar image, now density = 0.9 (Yellow). the differences with full density are minor.