Hi Steve, yeah probably what you are running into there is if you have a high number of lines like that you probably have many of them that are within tolerance of being considered tangent to one another.

Segments that are considered tangent to each other will get kind of glommed together into longer rail pieces. It would probably be better for that rail segment combining stuff to not happen when the segments are lines.

If you had somewhat fewer lines so that they were not within such a close angle of being tangent to one another you might get a more expected result.

If you could post the 3DM model files of your examples I could test them over here and see if tuning up the rail segment combining solves the problem you are showing or not.

- Michael

Hi Michael,

I just used curves from the model Bill posted, have attached.
(Bills model still in there)

- Steve

Hi Steve, also probably 2-rail sweep is just not going to work for this purpose with segmented polyline rails - 2-rail sweep doesn't know how to produce mitered corners, when there is segmentation it's just going to build the equivalent of small sweeps individual sweeps along each pair of rails, I believe that's one of the problems that you were showing.

If you stick with some other tools like 1-rail sweep, lofting (between same # of segments), extrude, revolve, those would probably work better. Actually Network can work as well with the same number of segments on each side, that may give you what you a similar result that you would be thinking of from 2-rail sweep.

The way that 2-rail sweep works, it kind of drags a frame that slides along each of the rails as it moves along them - this surfacing mechanism just does not particularly have a good way to handle sharp creases in the rails other than in special cases like with rails that have the same shape.

- Michael

Hi Steve,

> I just used curves from the model Bill posted, have attached.

You can get a proper result from these curves by using Loft between these 2 pieces, instead of using 2-rail sweep:



That produces this result:




2-rail sweep just won't work in general for this particular purpose - here's an example with a smaller number of segments to show more clearly what will happen:





- Michael

Hi Michael,

The one rail sweeping does not work for this example, there are various problems.

Network gives the same result as 2 rail sweep in this example.


Not to worry, I am OK the way I do this already. I just wanted to point out possible problems to Bill.


- Steve

Hi Michael,

>>You can get a proper result from these curves by using Loft between these 2 pieces, instead of using 2-rail sweep:<<

Yes I know, but those (single)end lines where put in place to simplify the example of the sweep. Try using the actual curves from the edges of the model.

What about the broken edges from the one rail sweep using both end lines?(as shown in the pics)

- Steve

Hi Steve,

> The one rail sweeping does not work for this example, there
> are various problems.

For 1 rail sweep you'll probably want to do it with just 1 profile and a planar rail which will enable mitered corners. That means that you would make something like flared ends by slicing tips off of a more regular initial shape, rather that incorporating the angled end parts directly into the surface construction.


> Network gives the same result as 2 rail sweep in this example.

This is the one that I can probably tune up in v3, it's related to that combining together of things that are within a close angular tolerance of being G1. I think I should be able to detect the case that the inputs are made up only of line segments and make sure that the output has the same segmentation in it.

This is the one that if you had a bit fewer segments in the curves where each line was not so close to being tangent to its neighbor, you would not get that combining effect.

- Michael

> What about the broken edges from the one rail sweep using
> both end lines?(as shown in the pics)

Well, it looks weird but it's similar to the 2-rail sweep case - if you use a rougher curve you can see that you get a kind of individual sweep piece along each segment.

When you select both end lines, that basically makes the sweep do a type of gradual morph between the first shape and the second shape, as things travel along the sweep.

So that shape morphing combined with the added rotation as it steps from one segment to the next, makes it unlikely that the separately created pieces are going to line up.

Things do line up if your rail segments are tangent where they touch, so if they are pretty close to being tangent, they get pretty close to lining up but not quite.

If you use that same segmented sweep sample file I posted earlier but with just 1 rail used you can see how the sweeps are created on each segment individually:





So you're getting the same kind of thing as that, just with smaller gaps due to the smaller angular difference in that other case.

- Michael

Hi Steve, and then you get that kind of odd "broken edge" look in that case because the generated pieces were close enough within tolerance to get joined together.

The sweeper attempts to join together the different generated surface pieces that are created from segmented rails.

If you measure the distance of the broken looking spots, you should find that they are less than 0.005 units apart, so that allows them to be joined despite having that much of a gap.

That might be another good thing to handle specially if the rail was only made up of lines - the joining step could be disabled for that case since it is really meant to join together the results from G1 segments which do then connect up nicely.

- Michael

Hi Michael,

I will stick to my original method of trimming the surfaces once the model is finished(or creating in poly program) when I need quad mesh output.
Trying to create the model directly from facets(in MOI) with the limitations and extra work would make it quite difficult, and potentially lead to having tris when having to cut off ends.

- Steve

And what about this function ;) Lineweb
http://moi3d.com/forum/messages.php?webtag=MOI&msg=3666.1

Hi Pilou,

For this type of construction, no. Lineweb is loft with only edges created, rather than edges/faces. So the same limitations.

If I was to go the route of creating the edges for facets in MOI, then I would create the model, extract the needed edges, then convert them to polylines(Using the rebuild to polyline script), then have those polylines imported into a poly modeller, then create Gordon/Coons surfaces, or even use (maybe) sweep functions in the poly modeller.

- Steve

Well for me editing poly structure of an imported moi model inside C4D was far faster then the slice up
method shown in the above post. Unsure if importing as an .lwo into C4D works better then riptide plugin as .obj.

Model I tested was a rail swept model similar to the one above.

Hi Flashfire, if you have time could you post some screenshots of your process?

Marc

I just looked for edges which would cause the test model to improperly hypernurb and dissolved them in C4D. Then
tested hypernurbs and it looked great. Unfortunately I find you almost have to do some cleanup like this if you plan on using
a mesh smooth or hypernurb method in 3D poly software.

Before doing this in C4D I used optimize to weld points of the original model.

Hi Flashfire,

> Unsure if importing as an .lwo into C4D works better then riptide plugin as .obj.

If you're going to be editing it, it shouldn't make any difference one way or the other.

But if you're going to be rendering the model directly you should use .obj format for that since that will preserve vertex normal information which helps the shading look the same as the original NURBS model, since the normals used for shading actually come directly from the original NURBS surface in that case.

I don't think that C4D knows how to read the vertex normals out of an .lwo file (until the last few years it was not really something that was found in LWO files), so it will instead cook up the vertex normals for smooth shading by averaging the normals of surrounding faces which is not really as nice as using the normals from the original NURBS model.

If you edit the model at a vertex or face level, it will invalidate any existing vertex normals and cook up new ones though, so that's why if you are editing it, it won't make any difference.

- Michael

There's also TopoGun (http://www.topogun.com/) which is a retopologizing program which is dedicated to this task of generating a quad only sub-d friendly topology.

Also 3D-Coat (http://www.3d-coat.com/) has a variety of retopo tools in it.

- Michael

-------------
Quote FlashFire
>>Well for me editing poly structure of an imported moi model inside C4D was far faster then the slice up method shown in the above post.<<
-------------

The examples in this thread are extremely simple. The surfaces have very little difference in length and the Mesher in MOI will actually give you a result that can be easily edited in a poly modeller.

-------------
Quote FlashFire
>>Model I tested was a rail swept model similar to the one above.<<
-------------
If that surface you show in your last post was your "Test", well, that is even more of a simple example than the ones show previously, as your example appears to be a single surface.

For a simple example, showing the kind of problem with quad mesh creation, lets look at this simple block with 2 curved surfaces with fillets.



Now that is easy to create in MOI and not much more difficult to create in a poly modeller. But if you output a mesh directly from MOI from the model as it is, then the mesh would not be so easy to edit into quads.

My point was nothing will be perfectly setup out of the box, output just has to be edited if you plan on using a mesh smooth or hypernurb function. Normally I use riptide, but just thought I'd try .lwo. It's good for like Mike said, when you need to
edit anyway. Or a quick and dirty way.

But I wanna have a go at editing that button model using C4D to make a clean model for hypernurbs... ;)

Hi Steve,

> Now that is easy to create in MOI and not much more difficult
> to create in a poly modeller. But if you output a mesh directly
> from MOI from the model as it is, then the mesh would not be
> so easy to edit into quads.

Yup, but that's also a good example of a model where there would not be any particular purpose (that I can tell anyway, let me know if there is one) for why you would need that model to be all quads.

It's got all rounded parts in it already, it's not like a rough mesh that you're trying to smooth out by applying sub-d smoothing on to it.

It does seem to be fairly common for people to want to get things as all quads even if there is not any particular benefit for their particular case, I guess because they've heard that it's good to have all quads if you are going to be doing sub-d modeling.

But when you create a model like you show there, it's all finished up, it's not like you need to apply sub-d on to it additionally. In cases like that there is no need to have all quads, you just export that model from MoI and take it into your rendering program and render it.

In general the output from MoI is oriented towards that kind of use where you're going to take the model into your rendering program and render it. Taking it in as a sub-d control cage and then applying smoothing to it is kind of a weird thing to do to CAD models since that sub-d smoothing actually kind of melts and changes the shape.

If you already have the thing in the shape you want, then you're all done with the model... Running sub-d smoothing on something that is finished does not just automatically "improve" it, it can mutate it and even potentially introduce ripples and bumps in it if the topology is not organized in a sub-d friendly layout.

By the way these comments are not necessarily directed at you in particular, these are just some things that come up fairly often with a variety of people, particularly people who come from a sub-d background where they're so accustomed to only working with quads.

- Michael