New model & a series of questions

[cstanton]

This post is more of a query than a showing of new work, but I'm going to be presumptuous and post this in the WIP rather than the New Users forum. This is a model of an inline speed skate. This is my second model, the first with curved surfaces, so I have a number of questions. I'm sure all of these questions have been asked and answered before, but I'm still looking for those answers. If anyone can point me in the right direction, I'd certainly appreciate it.

 

1. What is the best way to remove the creases in the model's surface?

 

2. Ideally, I'd like to radius the edges of the frame (the metal structure below the boot), but the short splines connecting the two profile surfaces don't have bias handles and changes to the manipulator properties value don't seem to have any effect, the changed values snap back to the original values. I created the frames by drawing the profile of the frame and then extruding the second side. Would it be better to copy the profile and then add the connecting splines? Or is there some other way to get control of these short connecting splines? Actually, none of the "interior" splines seem to have bias handles, they have bias "arms," but no handles that I can grab. Is that normal?

 

3. Some of these internal splines warp when I make changes to the external spines to which they are connected. Since there's no bias handles, I have to disconnect the splines and reconnect them. Often these spines straighten when I select them and re-warp when they are deselected. I've tried to re-do some of these several times--that can't be right(?). I add tails to these splines and to the connecting CPs before connecting them so they will remain independent splines, but it doesn't prevent the problem.

 

4. There seem to be interior four point patches created by the lateral connecting splines that are not visible and serve no purpose in the finished model. Can these be removed or avoided?

 

5. Would there be problems that I don't foresee at present if I add circles just touching the surface of the frame so I can apply a different material representing the flush ends of the axles? (Rather than connecting the circles to the surfaces with more splines.)

 

6. Five point patches are a mystery. I must be using them improperly because they seem inconsistent. Sometimes they fill in properly on the first try. Sometimes a patch will fill in after repeated failed attempts. And sometimes they don't fill in at all.

 

Thanks,

Curtis Stanton

[cstanton]

These are additional images of the frame with the unfilled five point patches, misbehaving splines and deformed surfaces.

 

Thanks again,

Curtis

[KenH]

1) Creases are caused in many ways. This image (thanks to Colin Freeman) might explain a few things.

 

2) You can get bias handles on some of the splines though? I'm not sure why they aren't showing on the smaller ones. Maybe they're hidden....try it in wireframe mode.

 

3) Can't visualize this.

 

4) This comes by how you have set up the splines. Ideally the splines should flow round the object. Again, I'd need to see the object to see exactly what it is.

 

5) I don't think so. If it's just a seperate object no.

 

6) Make sure they're not flipped. If you made them already and they're still not filled...just use the patch select tool to pick it and press F.

Sometimes it's best to select the five points and press the period key twice to get the FFP tool active. Also, hooks can prevent it becoming active sometimes.

[wwoelbel]

Hi there;

 

Lots of questions huh? I'm not a wizard but I can offer some little bits that I have picked up here and there and some alternate views and suggestions.

 

Internal patches are evil and MUST be avoided at all costs. They create render artifacts. Jump over to the ARM (www.lowrestv.com/arm ???) and check out TinCan's tutorial on internal patches. I had terrible problems initially when building buildings. You would think that if you have a square and extrude it you get a cube (you do by the way). This is fine until you go to the next step and add a window to your wall. When you extrude your wall with a hole in it you end up with internal patches all over the place. I solved this problem by always etruding my walls twice and cutting out the internal structures.

 

On your axel question - a simple 2D circle might just render funny. Consider a beveled cylinder (Yves beveled primatives)?

 

With respect to #3 - I also am having a bit of a time understanding the situation. Could you post an image? I don't understand the math (or perhaps the implementation) but I also do not get bias handles on "end points". I solve the problem by adding a tail. The weird jumping/straightening that your talking about happens to me when I have created a spline (usually a loop) and am welding it onto CPs in my mesh. They straighten out for me when I click on each spline segment around the loop (remeber that each CP has 2 segments - three CPs in a line make for 4 segments).

 

#1 - Those aren't really creases are they? They are more like soft humps. You don't appear to have bad geometry there that would cause a crease. Not sure.

 

I am pretty darned sure that it is not legal to have two 5 point patches adjacent to each other.

 

Could you post a side profile (or a rotoscope that you may be using) of the frame? You have a lot of "harsh" geometry in there. Id like to take a whirl at remodeling it but I know that there are certain bits that need to be accurate. Perhaps I can see what I can do and then offer some alternate suggestions on how to get a kinder/gentler model.

 

Bill

[ypoissant]

1) By adjusting biases and/or by applying the porcelain material.

 

2) Splines with only 2 CPs cannot have their bias adjusted. Those can only be straight lines. Because you extruded, the connecting splines are 2CPs splines. What you can do is extrude 2 times more and get rid of the internal patches that will be created because of that. Then you can tweak the biases to get a beveled effect.

 

3) This is a real-time glitch, The spline doesn't actually warp. Just press spacebar to refresh the display and they should get back to normal.

 

4) Indeed, those are to be removed and avoided. TinCan tutorial about internal patches will help you.

 

6) For valid 5pt patches, all the 5 points should be in a convex organization. 3 CPs aligned on a straight line as you have at several places will not work. Apart from that, sometimes, you may have to flip the normals of the 5pt patches to have them fill. If you find a 5PT patch will not allow you to click the 5PT-patch button, then hide everything else, reselect the 5 CPs using the grouping tool and then your 5pt patch button will be available.

[cstanton]

Thanks for the quick and useful replies. I've printed this thread and now that I've recovered from yesterday's all-nighter I'll try to implement the suggestions. (An inefficient way to work in the long run, but it sure is easy to get caught up in this stuff.)

 

I just found Tincan's internal patches tutorial; that will be helpful. (Internal Patches - What to Avoid )

 

Could you post a side profile (or a rotoscope that you may be using) of the frame? You have a lot of "harsh" geometry in there.

 

Ah, the rotoscopes. That was a whole 'nuther can of worms. I tried to be extremely careful when I shot the photos, but I never was able to get the different perspectives to scale so they matched exactly. I ended up winging it quite a bit when I drew the model. I tried to scale the rotos to actual size; you can see the scale of the frame in the second group of images. (The frame is 14.26" long.)

 

Anyway, thanks for the offer. I used tga files because I wanted the alpha channel. The files are over 1MB so I posted a couple at:

Left side (4.56MB)

and

Front (1.16MB)

If you would like another arrangement or another format let me know. All the replies seem to suggest that I'm going to have to redo the patches on the frames, so even the suggestion of another strategy would be helpful.

 

Thanks again for all the replies,

Curtis Stanton

[cstanton]

I copied the frame profile into my 2D drawing program and tried to diagram a new mesh for the frame on a layer overlay. The new mesh eliminates the 5pt meshes that weren't working. I have the feeling I still may be on the wrong track with too many splines ending at some of the junctions.

 

I added questions marks at some of the junctions that I thought might not be good according to the Colin Freeman diagram because multiple splines are ending at the junction. Would it be better if I made two of the splines continuous at each of these junctions and peaked the CP?

 

Thanks,

Curtis Stanton

[cstanton]

Deleted image, sorry, I don't know how to remove the entire post.

Curtis

[cstanton]

Deleted image

[cstanton]

This is best profile and mesh that I've been able to come up with. There are more splines and CPs than I'd like, but there don't seem to any surface artifacts that I can see.

 

I started over from scratch. I gave up on figuring out the principles of mesh creation. After trying to approach the mesh systematically several times I just started adding and moving CPs and adjusting biases in shaded/wire-frame mode until the artifacts were gone. The whole thing seems very fragile. Adjusting a CP just a small amount can add or remove an artifact in the narrow curved sections. I ended up changing the shape of the left cut-out, so it doesn't seem like much of a technique, but it may be time to move on.

 

Curtis

[KenH]

That's looking better. I think you were right with alot of the question marks on the image above this one. But if it works and there's no creasing then why change it. It's not an animating part.

[Rodney]

Curtis,

You're definitely going to have some problems with those areas where more than 2 splines intersect.

 

Also, since the both side are mostly identical (I'm assuming) I'd suggest making half and then Copy/Flip/Attach.

 

Rodney

[wwoelbel]

Hey Curtis;

 

I am sooo sorry. I forgot to check back on this thread. I'll pull the rotoscopes this evening and give it a shot (even though you may already have it fixed).

 

Bill

[cstanton]

OK, I'm going to have to re-think the mesh. The outside surfaces are a little simpler because I didn't extend to the outside the plates that join the two legs of the frame to the boot, but there are still junctions with more than two splines. In the mean time. . . .

 

A couple people mentioned that I might have animation problems if I just laid circles on the surface of the frame for the bolt heads and ends, so, in the interest of modeling experience, I made an axle bolt (sans threads.) My question is: Can each bolt be a boolean cutter and cut its own hole in the frame? (They're Allen bolts and the frame fills the wrench hole in the bolt if there's no hole in the frame.)

 

I extended Tincan's internal patch removal process to create bevels on the frame by extruding the frame profile twice instead of once and shrinking the outer surfaces. It worked well enough, but not flawlessly as the bevels on the concave curves of the cut-outs were reversed, so they and the bevels on the end curves had to be tweaked by hand.

 

Curtis

[wwoelbel]

About the axel bolts - why bother with the boolean cutter? If you wanted to be uber detailed, you could model the hole and then put the tiny bevels on it or you could just put the axel bolts in place. No bevel, no hole, no cutter, just two overlaping meshes.

[cstanton]

Hello Rodney,

 

Thanks for the input. I'm trying to break out of lurker mode. This is part of the project that I mentioned at the Siggraph party

 

Curtis

[cstanton]

About the axel bolts - why bother with the boolean cutter? If you wanted to be uber detailed, you could model the hole and then put the tiny bevels on it or you could just put the axel bolts in place. No bevel, no hole, no cutter, just two overlaping meshes.

No, I'm definitely not trying to be uber-realistic. Part of the reason I'm attempting this animation is to try to simplify a complex subject. I got a little carried away with the bolt, but when it was finished I thought the hex hole was pretty cool--newbeitis I guess. If the boolean cutter thing is a problem I'll just take out the wrench hole.

 

Thanks,

Curtis

[wwoelbel]

I really know absolutely nothing about boolean dodads except that they are tricky to use. What *I* would do is just take your modeled bolt and place it where it goes and forget about the hole itself. Give it a try and see if it meets your vision. I think that you'll be pleasantly surprised.

 

[EDIT]

You said "wrench hole". Arrgh. I misread your post. I thought that you were talking about the hole in the frame for the axel to go through!! Please disregard completely. I'm obviously going blind

[/EDIT]

 

I took a quick shot at the frame for you. The difference between what you did and what I did is that I added some additional CPs as needed so that I would have nice square-ish 4 point patches and a few 5 point patches that did not share any geometry. The problem you have is very similar to modeling a pelvis (one wiast hole with two leg holes). I added some geometry so that I could get 4 five pointers seperated by a nice set of four pointers. The model that I am attaching is in no way complete. I have roughly set the bias on the left-hand side and mostly ignored the right. I selected the entire model, peaked it, and then began tweaking those biases and the cp positions. Take a peak at what I did and see if it makes sense. Of course, if something looks stupid, let me know so that I can learn as well.

 

Bill

 

p.s. It will probably crab about the rotoscope. Just hit cancel to bypass it.

Skate1.prj

[Rodney]

Its great to have you posting here in the forum Curtis!

I had a great time at SIGGRAPH.

 

Keep posting!

Rodney

[mtpeak2]

Here's a solution I came up with for you.

[cstanton]

Here's a solution I came up with for you.

Interesting, thanks guys. Very clean. I'm going to use this information to try a slightly different approach. Mesh creation is turning into a strategy game. I hope it becomes less time consuming with experience.

 

I appreciate your time and effort,

Curtis

[cstanton]

After looking at the other frame meshes, my big idea, because the frame was all flat surfaces, was to start with an over-sized mesh of regular squares. I laid out the profile over the mesh, connecting it to the splines of the mesh. I then removed the parts I didn't need. Because I had the previous examples in mind, I guess it's not too surprising that I ended up with a very similar result.

 

Thanks for the help,

Curtis

[cstanton]

Mark,

 

How did you get your adjacent 5pt patches to fill in? I had the same basic grouping. All the patches filled in on the first side of the model, but when I extruded the second side I couldn't get the second patch of either of the adjacent pairs to fill in. I tried all of Yves suggestions from an earlier post but nothing worked. It doesn't matter which patch of either pair I choose; the first patch of the pair always fills in, the second patch doesn't. I finally resorted to adding a second horizontal spline to separate the patch pairs. Adding to the puzzle, as I said, all the 5pt patches filled in without any problems on the first side of the model from which the second side is extruded and which looks identical.

 

Curtis

 

Edit: I should have mentioned that adding the 2nd horizontal line to separate the 5pt patches allows all the patches to fill in. I was just wondering if there's something I'm missing. Was I just lucky on the first side of the model, or is there a technique for filling in adjacent 5pt patches?

[wwoelbel]

I am curious about the 5-pointers as well. I have been operating under the premise that 5 point patches were not allowed to be adjacent.

 

Bill

[Mr. Jaqe]

They CAN, but it takes me a few tries before it actually works and make sure you don't have hooks connected to any of the cp's, you'll have to un-hook the hook, create the 5-pointer and then put back the hook. Everything else should work, just give it a few shots. Just remember to hold shift if you click-select.

[cstanton]

OK, I finally got the recalcitrant, adjacent 5pt patches to fill. Still seems like Voodoo though. I'd been either shift selecting or drag selecting in all my previous attempts, but the second patch of one pair filled in when I turned off "show bias handles"(?) and then shift selected the CPs. The second 5pt patch of the other pair required more "messing around." It finally filled in when I shift selected the splines next to the CPs rather than the CPs themselves. Don't know why that worked this time, I'd tried that several times before.

 

Unless there's a more systematic way of doing this, I'm not sure it's worth the effort.

 

Curtis

[NancyGormezano]

Still seems like Voodoo though

 

I 100% agree. This is an extremely frustrating problem. I really wish there was a reliable way to conquer these stubborn 5 pt patches. I too go through all sorts of dances & incantations - and then maybe...finally, the green patch donut shows up. It is not always the same trick that works, which is really frustrating. And sometimes its just a matter of time. The same dumb action I had been doing repeatedly to select the cps and not working then decides to eventually miraculously work.

 

I try not hiding, then hiding everything but the 5 cps. Then try selecting cps clockwise, counterclockwise, with the lasso, with the rectangle, starting on different cps. Close the model, open the model, blah blah blah.

 

I do not know how to document this problem to A:M reports, in a way to help isolate the problem. But I can't believe that Hash doesn't know about it.

 

Sometimes I just have to recreate some of the surrounding splines - looks the same but I guess it eventually resets something. And maybe it's not really the same?

 

Sorry for pouring my sad pitiful heart out here ... I live in fear of 5 point patches.

 

I too, want to know what the secret formula is.

[pleavens]

Someone suggested inverting the 5 point selection and then re-inverting to select ("." x2)

 

Doing so has worked well for me.

 

Phil

[NancyGormezano]

Okey dokey - thanks, Phil - I will add that to the trick list - I believe I have tried that - but can never be sure ...

[cstanton]

Someone suggested inverting the 5 point selection and then re-inverting to select ("." x2) 

 

Doing so has  worked well for me.

 

Phil

That's a good tip. It seems to work most of the time, but it doesn't always work on adjacent 5pt patches. The interesting thing is that, if after doing the invert/invert, the 5pt patch icon doesn't activate, you may notice that one of the 5 CPs won't be highlighted. If you select the spline in the adjacent patch connected to this CP and detach it, your target 5pt patch usually will fill in on the next try. Of course, you've just created a new problem in the adjacent patch, but it's interesting that even when invert/invert fails, it seems to highlight the problem CP.

 

 

Bill, I managed to create the hex holes using a boolean cutter. I haven't had any problems--yet--of course, I've only done a very simple animation of the model.

 

These flybys are different formats of the same clip of the skate in its current form. The model still needs work and this simple animation shows that I have a lot to learn about lighting in A:M, but it's obviously a skate and I feel pretty good about that (amazed actually.)

 

Broadband fly-by (Windows Media, 1.14MB)

Dial-up fly-by (Windows Media, 390KB)

Quicktime fly-by (1.19MB)

 

Thanks for all the help,

Curtis

[NancyGormezano]

nice skate - glad to see -

 

(Also have to report the double invert selection on 5 pt patches seems to be working very reliably for me! yayyyyyyy! even on adjacent patches - No more fear... Life is good again)

[wwoelbel]

Curtis;

 

Glad that you got things spiffed out. I am really sorry about the mis-information about the 5 point patches (not allowed to be adjacent). I hope that that was actually true of some aged version of AM and that my wandering mind didn't just invent it!

 

Bill