Unexplained Fillet Blend Error

[dzcharlie]

hallo all,

below are images of my issue.

in the first image, you will see a curve outside the object that was used to create a surface recess. (procedure was, project, duplicate, scale, fillet blend)

as you can see right away, there are some funny wobbles along the top side of the fillet. completely unexplained.

then, i moved the original curve to see the effects. the following images show what happens to the surface fillets as i move the curve.
the surface is completely flat. why should there be such drastic errors when barely moving it, especially when one of the surface disappears altogether. as far as i know, there is not any difference between the location of extreme error and a location with a more minor error. either way, there shouldn't be an error at all.

also, the curve was constructed by creating 2 semispheres, attaching curve, then closing curve. the top side of the curve with all the errors is the side that closed on the curve.

very interested to hear what the diagnosis is.

thanks

[gmask]

well my guess would be that you are getting unexpected results because fillet blend is using two curves that have different CV topology due to their being the results of a projection of a curve on surface

[dzcharlie]

hmmm

im not sure i follow. one curve of the fillet is a project, the other as a duplicate of the original and scaled down.
so then are you saying that whenever you project a curve onto a surface, the cv topology changes? i can't think of a reason why that would be the case.

i should also mention that i'm following this model practically identically to digital tutors' "product visualization in maya", and of course, he gets no errors in the tutorial

[gmask]

QUOTE(dzcharlie @ 03/06/09, 06:08 PM) [snapback]302478[/snapback]
so then are you saying that whenever you project a curve onto a surface, the cv topology changes? i can't think of a reason why that would be the case.

Did you look at the CV's of the original curve and the one that was projected on the surface to compare them?

Often maya will add additional CV's to projected curves.. I imagine that the more acute of an angle the more likely it is to happen.. it also seems to happen when the surface it is projected on is irregular. Sometimes moving the isoparms of the surfaces receiving the projection can clear these issues.

[dzcharlie]

k, we're getting pretty deep into this diagnosis, there's gotta be an explanation for this.

i went back and verified the original curve and the projected curve immediately after the point of verification. (there's no way to check the cv of the curve after its been trimmed, right?)

so yes, it looks like it adds some more cv's, but i'm still not sure that is the problem.

below are some more clues, which i think holds the answer.

---

1.

i went back and watched the tutorial with a close eye, and i noticed a discrepancy that might shed some light. the image on the left is of the surface right after it was converted from polygon to nurbs. you can see that the number of faces is 12. THEN in the next chapter, when the curve to be projected is being created, i noticed that the surface had more faces, like in image on the right.

obviously, he isn't working with the same model / file. and can this be the reason why he is not getting the same errors?

---

2.

i went back and redid the whole process, and i noticed something going on with the trim action.

below is an image of what looks to be some kind of trim error. why is the trim not trimming properly along the edge?
i'm not so sure this has to do with the problem, but still this looks no good.

---

3.

could the problem have to do with the closing of the curve?

the original curve was created from 2 separate semicircles, attached curve, then closed curve. i tried 2 different cases. 1 where i attach the 2 semicircles from the lower point, and closing the curve on the top, and the other vice versa.

in both cases, the weird fillet errors occur on the side of the closed curve. but looking at the CV's of the curve, shows how sign of trouble when comparing the top and bottom.

---

still not sure what to make of this. i don't know how good you have to be to use maya nurbs, but it seems like the possibility for strange behaviors and errors are quite high.

[gmask]

QUOTE(dzcharlie @ 03/07/09, 12:43 AM) [snapback]302503[/snapback]
i went back and verified the original curve and the projected curve immediately after the point of verification. (there's no way to check the cv of the curve after its been trimmed, right?)

untrim the surface, then look at the surface curve

QUOTE(dzcharlie)
so yes, it looks like it adds some more cv's, but i'm still not sure that is the problem.

the problem is that when you fillet between two curves with different CV's is that maya has to guess how many to use and it uses tessellation values to make them seamless. It's really a terrible way to work and it is rarely used for production. Even if you do use it ultimately you want to convert everything to polygons and then you'd have a headache cleaning up the mess.

QUOTE(dzcharlie)
i went back and watched the tutorial with a close eye, and i noticed a discrepancy that might shed some light. the image on the left is of the surface right after it was converted from polygon to nurbs. you can see that the number of faces is 12. THEN in the next chapter, when the curve to be projected is being created, i noticed that the surface had more faces, like in image on the right.

obviously, he isn't working with the same model / file. and can this be the reason why he is not getting the same errors?

Well it is a poor tutorial that skips a step.. if that's what happened. Another issue with fillets is that if the end points of the two curves are not in the same place it will not work the way you want it to and it will go askew and produce errors like what you are seeing. You need to untrim that surface look at the surface curve, make sure the endpoints are inline with the other curve. You can edit the surface curve and , delete and move CV's on it. In fact you don't even need to trim that surface to make the fillet.. you can trim it later.

QUOTE(dzcharlie)
below is an image of what looks to be some kind of trim error. why is the trim not trimming properly along the edge?
i'm not so sure this has to do with the problem, but still this looks no good.

That's typical.. it's called display tessellation. If you look in the attributes of the trimmed surface you will see settings that increase the accuracy of that display tessellation. This tutorial should have explained this.. if it doesn't then tsk tsk on the author.

QUOTE(dzcharlie)
could the problem have to do with the closing of the curve?

well if there are two CV's on top of one another as a result of the closure then that's not helpful.. but I don't think that's the problem.

QUOTE(dzcharlie)
i don't know how good you have to be to use maya nurbs, but it seems like the possibility for strange behaviors and errors are quite high.

You don't unless you plan to be an industrial engineer in which case you won't be modeling in Maya. Nurbs are handy for certain things like organic surfaces that need to be tweaked alot but ultimately you won't be finishing whole models in nurbs for production. Personally I'll use nurbs to rough out surfaces because the flexibility of them for that is great but then I convert it to polys soon after.

Nurbs are more difficult to work with and the possibility of user error is much higher Trims are a pitfall.. there are ways around using them but it takes way more steps to accomplish the same function but the end result is predictable... trims are for the lazy.

[dzcharlie]

hey gmask

really appreciate your feedback and patience. i'm definitely learning a ton from this hurdle.

QUOTE
Another issue with fillets is that if the end points of the two curves are not in the same place it will not work the way you want it to and it will go askew and produce errors like what you are seeing. You need to untrim that surface look at the surface curve, make sure the endpoints are inline with the other curve. You can edit the surface curve and , delete and move CV's on it. In fact you don't even need to trim that surface to make the fillet.. you can trim it later.

i'm not sure i follow. what do you mean that the endpoints from each curve have to be in the same place? for ex, endpoint on curve1 is 5 cv's to the left, so endpoint on curve2 has to also have the endpoint 5 cv's to the left?

also, why on earth would projecting a curve onto a surface produce more cv's? i deleted them to get it to look closer to the original curve, and that made the fillet even worse.

after some more examination, i noticed that not only does projecting add cv's to the curve, but it actually created some cv's off the curve and on the surface. moving the original curve will produce all sorts of unexpected cv's on the resulting projection. in the image below, a cv at the bottom left was created outside the curve. why can't it project as expected? arghhh

if this is true, that maya has to guess when fillet between 2 curves with different cv's, it completely renders the tool unreliable, which has me at a lost. i am modeling / rendering for the purposes of industrial design, and i was turned onto maya because someone suggested that maya had strong capabilities in both nurbs and polygons and switching between the 2 is a great flexibility.

the tutorial i am following advocates this, and claims to use nurbs for its strength, such as precise rounds, and trims. i'm also learning alias studiotools, and i was under the impression that the trim tool was very powerful there. being that maya was made by alias, i am quite surprised at how finicky nurbs is.

thanks

[gmask]

QUOTE(dzcharlie)
i'm not sure i follow. what do you mean that the endpoints from each curve have to be in the same place? for ex, endpoint on curve1 is 5 cv's to the left, so endpoint on curve2 has to also have the endpoint 5 cv's to the left?

well that would help matters but what I'm talking about is if the endpoints were on the left side and the others where on the right side (this is an extreme example) then the fillet would probably be twisted.. I think this can also happen if the curve directions are opposite of one another.

QUOTE(dzcharlie)
also, why on earth would projecting a curve onto a surface produce more cv's?

I dunno but that's what happens.. a better approach might be to project the curve and use it as a template and then draw a curve on surface so that is exactly as you expect it to be.

QUOTE(dzcharlie)
if this is true, that maya has to guess when fillet between 2 curves with different cv's, it completely renders the tool unreliable, which has me at a lost. i am modelling / rendering for the purposes of industrial design, and i was turned onto maya because someone suggested that maya had strong capibilities in both nurbs and polygons and switching between the 2 is a great flexibility.

the tutorial i am following advocates this, and claims to use nurbs for its strength, such as precise rounds, and trims. i'm also learning alias studiotools, and i was under the impression that the trim tool was very powerful there. being that maya was made by alias, i am quite surprised at how finicky nurbs is.

Maya is not a CAD or CAE program.. nobody uses it for that. Now Studiotools is used for design and engineering but from what I understand the accuracy of the tools in Maya for those purposes is sub standard so switching back and forth is a pipe dream. If anything you import models from Studiotools into Maya and leave it at that.

Besides that I'll still assert that trims are lame.. they produce discontinuous surfaces which from my point of view tend to create more problems that they solve. If you want nice nurb patch models that are quad based then trims are out.. they break those rules.

And just a heads up.. when you get to booleans don't be surprised when you find that are terrible in Maya.. other programs do it better. You can get them to work in maya but they are finicky.

[dzcharlie]

thanks gmask,

i get what you're saying about trim. what alternative methods are there for cutting a hole out of a surface?

[gmask]

I believe the technique is called "socking".. I believe the term is descriptive of what it might look like when you try to blend two cylinders together when they meet at a perpendicular angle. Just imagine that instead of another cylinder it's just a hole.

anyway in order to create a surface with a hole in it.. you can create the trim on it and use it as a template. Then you would draw curves on the surface that originate from the 4 corners of the surface and terminate at the 4 corners of the hole. From there you add additional curves and then loft between them to get 4 panels.