MoI discussion forum
MoI discussion forum

Full Version: Trouble with Fillets

Show messages:  1-20  21-35

From: John (ENTANGLE)
9 Nov 2008   [#1]
Hi, just started with Moi but I'm having a rough time getting it to do what I'd like. I am mostly used to poly modeling so nurbs are new to me, I think I have a basic understanding.

Anyway, heres the problem, I have this guitar body:



What I want to do is fillet both of these edges (at different thickness) so that they're smooth, but for some reason, I'm only able to get a miniscule fillet in (around 0.01) on one of the lines. If I remove some of the neck fitting, I can get a nicer fillet on the top, but the line underneath won't fillet.

Confusing huh =\

Attachments:
guitar04.3dm

Image Attachments:
guitar01.jpg 


From: Brian (BWTR)
9 Nov 2008   [#2] In reply to [#1]
John
Just a bit of fun trying to get a result using a different thinking. Not ideal but it may open up some ideas for you.Two sets of sweeps using curved shapes.
Brian
From: John (ENTANGLE)
9 Nov 2008   [#3] In reply to [#2]
That looks fairly decent Brian, but it's not accurate. It still appears to have this "hard flat" top to it, and what I was trying to do was smooth that out with a fillet on both of the edges. Thanks for having a go, nurbs seem tricky to get into =X
From: Michael Gibson
9 Nov 2008   [#4] In reply to [#1]
Hi John, there are a few things in your shapes there that are not conducive to good filleting.

Probably the biggest one are these very tightly curved corners in these areas:





When you have a tight bend like that, it will limit your fillet radius to a small value that is smaller than the radius of the bend.


Here is another way to kind of visualize what happens with a fillet trying to go around a tight bend like that.

If you have a bend here, imagine that the line in the lower-right is the distance of the fillet radius that you are asking for:




Now if you watch as that fillet radius moves across the tight bend, it causes a messed up area of self intersection:




So to make something fillet friendly, you don't want to have tight bends in small areas, instead those need to be an actual sharp edge there.


If I draw in a line like this, and use boolean difference to slice off those little curved corners:



Then you can fillet the top edge quite a bit more, up until it starts to eat up all of the edge surface below it, here is one at radius 0.2 (using MoI v2 beta):




The lower edge has a different problem - that one is probably just not going to fly trying to do it all in one piece due to the area at the end here where the 2 surfaces on either side of the edge are totally flat and tangent to one another:



That would cause the fillet to be totally erased in that region - MoI's fillet engine does not handle that kind of fillet situation very well where the fillet disappears to nothing along some stretch of the edge.

You would need to get that to have some angle in there to make it filletable, or hack off that piece so that you could try to get a fillet along the other regions where the fillet does not have a "disappearing" region to it.

- Michael

Image Attachments:
guitar_fillet1.jpg  guitar_fillet2.jpg  guitar_fillet3.jpg  guitar_fillet4.jpg  guitar_fillet5.jpg 


From: Brian (BWTR)
9 Nov 2008   [#5] In reply to [#4]
John, Michael has the answers.

For interest, here is a rough (very) render of my approach.

Brian
From: Michael Gibson
9 Nov 2008   [#6] In reply to [#1]
Hi John, another kind of technique that can work pretty well for this stuff is kind of along the lines of what Brian wrote previously.

But instead of trying to build a fully curved piece all in one surface what you focus on is in curving one area only and leaving the other side to be filleted.

Here's an example:



Model is attached as guitar_loft_fillet.3dm

What I did in this case was take 4 curves, and then Loft a surface through them, using the Loft Style: Loose option, which tends to reduce the bulgy-ness of the lofted result.

That has kind of built in the curvature of the lower part of the body directly into the surface.

But then I let it come to a sharp edge at the top, when then can be filleted.

This can be a generally good strategy to use - to leave a fillet for finishing off kind of the end of a shape but trying to use surfacing tools to handle the shaping of the interior parts rather than trying to use filleting for the sort of internal pieces of the shape...

- Michael

Attachments:
guitar_loft_fillet.3dm

Image Attachments:
guitar_loft_fillet.jpg 


From: BurrMan
9 Nov 2008   [#7] In reply to [#6]
I never hit on the method Michaels going to produce ....Novice!

I started off fooling with blend (A recent forum favorite) and after seeing Michaels realize that although mine "looks" ok it is not the original shape as I thought.

Let the studies continue.
From: John (ENTANGLE)
9 Nov 2008   [#8]
Woah, thanks for the replys guys, this was fast. I thank you all for helping and offering your advice with this.

Michael, I now understand what you mean about the fillets. I just assumed it would work similar to loops in polygon models (how wrong was I?) I think your first method is what I am looking for. Do you think it's possible for me to "fake" the second lines fillet by using a loft? or would it be possible for me to fillet in certains areas where the surface doesn't go flat?

The second method also looks interesting, and I think I will put it to use on other objects needed for the full guitar. Thanks again.

Burrman, I actually like the result. The outer edges are nice and smooth, it's just the inner sections by the neck that bother me. Would you mind showing how you did it? I can't seem to get anything out of blend.
From: BurrMan
9 Nov 2008   [#9] In reply to [#8]
John,
Yeah, your actually right in that with the blend method and a default value of "1" for a G1 the part where the body meets the top of the neck gets all pinched up, so the result you see here or in my previous post is not what you would acheive, I got a smooth transition with a ".5" G1 but its a drammatically different level of smoothness for the edges.

Possibly redrawing the top curve to be much smaller can produce a different result with blend but I have to say I took Michaels model through to a solid and it actually looks like a guitar! I'll poke around a bit more and If I get anything substantial with blend I'll post back. Have fun!

Burr


From: BurrMan
9 Nov 2008   [#10] In reply to [#9]
I'll stop here as I'll only get nuts now.

If you're being creative then the blend can produce some cool stuff...(the latest alembic?)



If you need to be precise, I must defer to the master with the loft loose example.
From: John (ENTANGLE)
9 Nov 2008   [#11] In reply to [#10]
The last example provided is almost perfect, just a little highly raised. You pretty much nailed it spot on. Thanks again.
From: Michael Gibson
10 Nov 2008   [#12] In reply to [#8]
Hi John,

> Michael, I now understand what you mean about the
> fillets. I just assumed it would work similar to loops in
> polygon models (how wrong was I?)

:) Yup, it is pretty different. Fillets actually do quite a large number of calculations, including making offset surfaces (surfaces that are a constant distance away from the existing ones), and calculating the curves of intersection between 2 surfaces. Then comes some other stuff like creating fillet surfaces that follow those intersection curves, and then doing some additional extensions and intersections between the fillet surfaces and finally filling in corner pieces where different fillet pieces might meet.

Since there are a lot more complex calculations involved, there are also a lot more things that can go wrong than just pushing points around in edge loops... But the upshot is that the final result is a very accurate piece that has a cross-section of an exact circle.

You tend to get the best results if you have pretty "clean" geometry.


> I think your first method is what I am looking for. Do you think
> it's possible for me to "fake" the second lines fillet by using a
> loft?

It would probably be difficult to get good results by doing that. You could put in a "fake" piece with a loft definitely, but it will probably look pretty fake as well because is difficult to produce a smooth result with that method. The loft that you generate will not be ensured to be smooth with the adjacent surfaces. That's another big difference between working with NURBS and working with sub-d polygons - with your NURBS constructions just building 2 different surfaces adjacent to one another with any particular tool does not guarantee that they are smooth to one another. But there are a couple of particular tools like Fillet and Blend that you can use that are set up to build a result that is smooth to adjacent pieces.

The other way that you generally ensure smoothness is to build your model in larger sheets instead of in small pieces or strips when possible. You can pretty easily guarantee smoothness of a single surface construction, like one loft will be smooth inside its own generated result as long as your input curves are also smooth.

So it can help if you build what should be a smooth section as a larger surface sheet instead of trying to do it in diced up bits.

When I saw your result that had 2 surfaces that were both flat, it made me think of building one single larger surface sheet (with that loft method) there instead of that.


But just a note - it is also possible to go too far in the other direction and try to build too much out of a single surface as well. Each "larger sheet" should be a kind of logical component of the model, it can also cause problems trying to build a model made up of many different shapes out of only one single surface as well..


> or would it be possible for me to fillet in certains areas where
> the surface doesn't go flat?

That's probably possible but you would likely need to dice up the model into different pieces to make that happen.

- Michael
From: Michael Gibson
10 Nov 2008   [#13] In reply to [#10]
Hi Burr,

> If you need to be precise, I must defer to the master
> with the loft loose example.

Well, loose loft is not necessarily any more precise, just a different way...

There is often more than one possible route to get a good result! :)

- Michael
From: DannyT (DANTAS)
10 Nov 2008   [#14] In reply to [#9]
Burr, so that's what you sound like :) cool tute dude. Camtasia??

---------
From: BurrMan
10 Nov 2008   [#15] In reply to [#14]
>>Burr, so that's what you sound like :) cool tute dude. Camtasia??

I found that a person never sounds like what I had imagined! I wonder what I used to sound like to you??? He he!
(Since I used other space I could add the audio without worry of size. For me it helps with the understanding of whats happening on screen and the thought process of the creator!)

Yes that was Camtasia. I have another one called captivate which is a pretty powerful tool but not as down and dirty as camtasia. It needs more massaging after the fact. So for a quick record and post its camtasia. Actually If I didnt already have the camtasia loaded I would just use Jing.
From: BurrMan
10 Nov 2008   [#16] In reply to [#15]
Oh Billy!!!!


From: WillBellJr
11 Nov 2008   [#17]
Being a guitar player myself, this thread is well respected!

-Will
From: Matthew (KINGMATTHEW)
11 Nov 2008   [#18]
Hi all,

Maybe this one could also find your interest ...

Image Attachments:
guitar.jpg 


From: Michael Gibson
11 Nov 2008   [#19] In reply to [#18]
Hi Matthew, that one looks great! A lot of details in there, even the little twisted parts at the very ends of the strings wrapped around the pegs, cool!

- Michael
From: Apophis
11 Nov 2008   [#20]
Wow, nice guitar. Very detailed job. Hope we will see some nice render.

Show messages:  1-20  21-35