The reason I asked abut curve continuity in another thread is because I'm interested in the possibility of MOI having abilities like Max's patch modeler.

In Max you can take a 4 sided patch, grab any one of the sides and drag/clone it and it makes a new patch connected to the last one. All the lines (even ones that cross each other) are tied together by 4 way handles. Move one point and it affects each patch that it is a part of. Meaning you can't break your smooth surface unless you specifically want to.

Not a lot of people like patch modeling in the sub D world, probably because sub D box/ NURMS modeling is more robust. But for solid modeling I think it would make a lot of sense. You can very quickly make organic shapes without much planning ahead. The only down side is that patches have to be 4 sided. But that's not that bad.

Any thoughts on adding this type of functionality to MOI?



EDIT: If I import my patch model into MOI (convert to Max nurbs and export as SAT) I can better illustrate what I want.

If I grab a point and drag it up in Max all the surfaces remain smoothly attached. In MOI (after separation and turning on show points) I can grab the points, but they are not connected between the surfaces and it makes a crease.

I want some way for points on two surfaces that touch each other to be connected and remain smooth. Maybe this is a pipe dream as I don't understand the math underneath the software very well. But this would make modeling much more free-form and less like a puzzle.

thanks

Hi Shaun, I guess I kind of look at that older style Patch modeling as a somewhat of a dead-end compared to sub-d modeling.

One problem is that I think you only get tangent continuity with that kind of a system, not curvature continuity. That basically means that you have kind of one level of smoothness but if you make the surface reflective you will probably see that the reflections don't look smooth because of the lack of curvature continuity.

Also the process of editing points directly on corner junctures like that tends to introduce bulges and a kind of bunching near the edges of the full surface - I think it's quite a bit harder to get really smooth flowing larger surfaces under that kind of editing as compared to sub-d cage editing.

There was a company that was working on producing a CAD modeler using a kind of similar technique though, with a few added twists (like any number of sided patches I think), you might want to check it out: http://www.freedesign3d.com/ I'm not sure if it is still under development or not though, it seems to have run out of steam.

But making a new toolset for an alternate style of 3D point cage oriented modeling will be a lot of work, so probably nothing will happen in that area for a while yet. Then when I do get a chance to focus on it, I think it would be hard to justify sinking in a large effort on that kind of area instead of sub-d modeling...


Back quite a long time ago there were also some free stand-alone tools for doing that kind of patch modeling as well, I think it was called hamapatch? http://www.eatonhand.com/images/spatch.htm I'm not sure if those are still around anymore. You can kind of see in the link here that the models have a kind of bulgy/ridgy nature to them, which is fine for some kinds of modeling like cartoon characters but not such a great quality for product design.

- Michael

Yup. I got my start with hamma patch and spatch. Moved on to Animation Master before I finally got a real 3D job.

Maybe I'm the only one left who likes patch modeling. It's not being developed further in Max either :(


But still there must be some way to "glue" surfaces together so moving one common point changes both smoothly.

Hi Shaun,

> But still there must be some way to "glue" surfaces together
> so moving one common point changes both smoothly.

It depends on what you mean by "smoothly" - yes for maintaining tangency but more difficult for maintaining a higher amount of continuity than that.

The main issue is that a mechanism that tried to glue little separate surface pieces together would just result in the same kind of somewhat lumpy editing that I'm sure you've already seen with the patch modeling...

- Michael

One thing you could look at is to just rebuild the surface in MoI, using the max surface.. You could extract edges from the max model and re-surface them, or you could project some lines on it to then use those to resurface.. Then you can pull the points on the new surface.

I guess the overall point isn't rebuilding the patch object in this example. I'm looking for more flexible solid modeling of organic shapes in general.

Say i have a complex dome where several surfaces meet at the top in MOI and my boss decided the dome should be higher. I would have to re draw all those curve separately and make new lofts (or networks or what ever) all over again. I would much rather be able to grab the common point and drag it so it stretched all the surfaces of the dome at once. Less planning ahead, more sculpting like.


None of this should be taken as criticism of MOI. As a person new to solid modeling I'm just bouncing thoughts off this very knowledgeable community.



One question I have is, why isn't there a form of cage modeling for solids? If you do a Show Points on a sphere in MOI something very similar to a control cage appears where you get the elastic type of modeling I'm looking for. I want to be able to grab a section of this cage and extrude it, or weld two points together.

I know it's really a trimmed surface underneath, not a mesh. But there must be some way get closer to sub D techniques.

Michael,
(MoI is the greatest thing since sliced bread, I'll enjoy my remaining trial days and put it on my Christmas list)

There must be some way to at least introduce a simple command, even an addition to the Align tool set.
One that would apply tangential alignment to the edges and intersections of selected associated mesh networks.

I'm guessing that the first control points on either side of a network edge point would be averaged and moved to line up with each other through that edge point.
In operation, it would take no more time and effort than the process of Joining the mesh edges when you're done with them.

Even though, you'd still have issues with curvature continuity, you'd be providing a way to remedy those awful creases and lumps right at the seams.
These show up and gives a solid that unwanted "panel-ized" appearance. It's been the only disconcerting thing to me I've run into learning how to use MoI.

Maybe some type of averaging of point positions could smooth out those control points along the mesh, outward from the target edge.

What you're wanting can be done and indeed Solidworks (sometimes) does just this. Unfortunately what you can end up with is patch edges that are tangential (G1) continuous to adjoining surfaces but the patch itself can become like a mountain range trying to average these out across the whole patch.

Hi Shaun,

> I'm looking for more flexible solid modeling of organic shapes in general.

Organic shape modeling is not a strong area of solid modeling - solid modeling excels more in semi-mechanical man made type shapes where the shapes are driven more by 2D profiles and have pieces cut out from them.

If you want to make organic shapes, sub-d modeling is overall much more suited for that.

At the same time sub-d modeling is poor in doing mechanical shapes with cut out parts on them.

Different kinds of 3D modeling technology have different strengths and weaknesses, and your best bet in general is to try and leverage each technology in its strong area, rather than trying to do everything in one single tool. Try to use the best tool for each particular kind of a job rather than using a tool that is not well suited for your particular job at hand.


> I would have to re draw all those curve separately and make
> new lofts (or networks or what ever) all over again.

Usually that doesn't actually take much time though - one of the good parts about NURBS modeling is that you can make a lot of stuff happen from a relatively small number of curves and it is quick to draw curves. So recreating some area of your model is usually not so labor intensive in NURBS modeling as it would be in polygon modeling where are dealing with a lot more little elements since you are dealing with individual vertices on the 3D shape directly.


> I would much rather be able to grab the common point and drag
> it so it stretched all the surfaces of the dome at once. Less planning
> ahead, more sculpting like.

Sub-d modeling works exactly like that, so if that's the kind of technique that is important to you, you probably want to be using a sub-d modeler that is based on that technique instead of MoI.


> None of this should be taken as criticism of MOI.

None taken - in fact I want to be very clear that MoI is not really intended to be a replacement for sub-d modeling techniques. If you need to use those techniques there are already a lot of sub-d modeling programs out there that you can use already to do that.

MoI is focused on different techniques than that, but that's actually why it is so useful as a companion program - because the techniques that you can use with MoI are very strong in modeling mechanical shapes, exactly where sub-d modeling is very weak.


> One question I have is, why isn't there a form of cage modeling
> for solids?

It's because of how trimmed surfaces work - after you do some booleans, the edges that you end up with do not necessarily correspond to the control points of the underlying surfaces anymore - the edges are more like markers that mark regions of the surface as holes. Because they're just markers and don't drive the shape of the surface, you can't pull an edge to deform things, the thing that you can pull are the control points of the underlying surface.

You can see those control points on a solid by using Edit > Separate - you can always turn on control points for an individual surface. But because you can't edit a surface by pulling on a trim curve where 2 trimmed surfaces are touching, if you pull the underlying surface control points you will usually open up a gap between what used to be a joined edge on the solid.

For more information, please see this FAQ:
http://moi3d.com/wiki/FAQ#Q:_Why_does_show_points_work_for_some_objects_but_not_others.3F

That has a bunch of illustrations that should help explain why a solid made up of trimmed surfaces is not easily smooshed around by point editing.

Also some previous discussion and explanation here:

http://moi3d.com/forum/index.php?webtag=MOI&msg=4052.6
http://moi3d.com/forum/index.php?webtag=MOI&msg=4050.3

 

> If you do a Show Points on a sphere in MOI something very
> similar to a control cage appears where you get the elastic type
> of modeling I'm looking for. I want to be able to grab a section
> of this cage and extrude it, or weld two points together.

Those are techniques that you'd use in a sub-d modeling program - if that's the style of modeling that you want to do, then you should use a sub-d program to do it - they work exactly like you're describing!

MoI is focused on a much different way of working that than - I kind of describe the sub-d way as more like sculpting where you're manipulating the 3D point cage, and MoI's method is more like drawing where you focus on drawing curves to use either for building shapes or for cutting away areas of your model.

Organic shapes tend to work well with the sculpting method, and mechanical man-made shapes tend to work well with MoI's "construction" type approach.


> I know it's really a trimmed surface underneath, not a mesh.
> But there must be some way get closer to sub D techniques.

Not really - the whole mechanism of how trimmed surfaces work is very different from sub-d modeling, since a trimmed edge is a marker that lives on the shape instead of being something that defines the actual surface shape.

In a polygon modeler there are no trims - every edge that you see is also a natural edge of the face and so pulling the edge around is the same thing as pulling the face around.

However, this same mechanism of trimmed edges is exactly why Boolean operations work so much better with NURBS modeling than with polygons - because with NURBS booleans when you intersect objects the underlying surfaces don't change at all and only new trim curves are calculated on the surface. This greatly helps to avoid complexity in the booleaned result. With polygon modeling booleans mean dicing things up into a bazillion little fragmented facet surfaces which is one of the reasons why it doesn't work very well.


So really the main thing is to use the right tool for the right job!

If you want something that works exactly like a sub-d modeler, then you can get that already by just using a sub-d modeler... It's more of a focus for MoI to actually work differently than a sub-d modeler so that it can have strengths in different areas and work better for the right kind of jobs.

- Michael

Hi Mike,

> There must be some way to at least introduce a simple command,
> even an addition to the Align tool set.
> One that would apply tangential alignment to the edges and
> intersections of selected associated mesh networks.

In v3 I do want to work on a tool for aligning a surface to be smooth to a neighboring one.

But when you're building a bunch of surfaces in a NURBS modeler like MoI, it's rare for the control point structures of each surface to be the same, so the alignment process can't just move control points of each surface in line with each other, it is more of a process of treating the target surface as an analytic smooth object and adapting the surface to be modified to the other surface's shape, nothing really to do with its control points.

If you have 2 surfaces that touch each other and have the same control point structure where they touch, that's probably an indication that you should have built those surfaces as one big single surface right from the beginning rather than as small individual surface pieces.

One common problem with people that I tend to see with people coming from a polygon modeling background is that kind of construction method of building an object out of a lot of little separate patches sitting side by side - usually instead with NURBS modeling you want to build a much larger single surface for something that is supposed to be smooth and then form the final edges by cutting away portions of that large extended piece, rather than an approach of building all the surfaces directly along their final edges.


Here are some links to some discussions on this kind of fundamental NURBS technique of creating extended pieces and then cutting them to produce the final model, browsing these links may help:
http://moi3d.com/forum/index.php?webtag=MOI&msg=3855.7
http://moi3d.com/forum/index.php?webtag=MOI&msg=3855.9


> These show up and gives a solid that unwanted "panel-ized"
> appearance. It's been the only disconcerting thing to me I've run
> into learning how to use MoI.

Do you have an example of that kind of unwanted panelization?

It may be the result of using that kind of patch-by-patch building approach instead of the "Build one large extended surface that then gets trimmed" approach - it often takes polygon modelers some time to get accustomed to thinking about using trimming and cutting as a first class modeling method since it is something you avoid with polygon modeling but it's the main method to do things with NURBS modeling.

Once you get the hang of building more extended simple pieces and then cutting them to get your final boundaries rather than trying to surface the final boundaries directly, you can get better results and things actually go more quickly as well.

- Michael

Thank you very much Michael for your thoughtful explanation.

I used CorelDRAW for graphic design with its anti-aliased Beziér curves for the last 20 year so when I found MoI I took to it like a duck to water.
It feels comfortable to work with.
Consequently, I've used Sketchup for a few years, and though it is a polygon based modeler, their approach is based upon building shapes out from each other.
It's easy to draw a shape in 2D on the planar surface of one object and refine it into a more complex one. It's strong suit lies in its "Follow Me" tool, though that tool is limited. I've used a myriad of plug-ins to do features like skinning and more organic shapes.
Ah... but Sketchup can get "polygon heavy" very quickly and it was frustrating dealing with beaucoups of misplaced edges and permanent "sticky" elements.
Sketchup does have strong abilities in Boolean and sweep functions, so I know what you mean by creating larger master objects and cutting away from them to refine the model.

Somewhere in my self-teaching and mindset I desired to craft shapes of an exact nature to be satisfied, so I got used to creating meshes and piecing them together.
I was crafting this pedestal sink basin last night and first used a sweep of four profiles between two closed rings, one a super-ellipsoid made with the Conic tool.
The results were okay, but had ripples in it that radiated from the center. The faucet shelf was okay, but I desired a more flat shape.
I created the lines to make that shape, but would have to resort in using Meshes instead of the Sweep. And in succession, Joined them together.
It left me with the hard edges I've run into before. Wish I could get rid of them, as they show up in renders.

I've also considered fabricating the bottom half of the basin from a solid sweep, then making one of the inside and Trimming it out to form in inner bowl with the Boolean function. Then applying a Fillet to the top rim and bottom hole.
However, I'm often now apprehensive to go that route as both the Fillet often fails to work on the edges of more curved shapes and sometimes the Trimming function does not leave the inside wall that I wanted.
I can do a reach-around with other Boolean functions with some cutting and pasting though.

It's not all a deal breaker. MoI does the job, and it is as easy to operate as the natural media application known as ArtRage.
Thanks for providing the "NoSave" version. I'll be able to feed my MoI addiction until the funds are there.

(In pic below: The left ones are made from Sweeping profiles. I wanted a more specific shape for the back so I made the right ones from Network meshes - Notice the visibility of the sharp intermediate edges)


Mike

Hi Mike, yeah SketchUp is great for boxy objects, but trying to deal with curved objects is awkward with it since it is hard to manage piles of little edges and facets that are trying to approximate curved areas.

re: panels - can you please also post the 3DM file that goes along with your screenshot so I can inspect the actual geometry?

But basically if you built it by making something like each quadrant as a separate Network surface, then that's normal that they won't naturally be smooth where they touch - that's because when you build any kind of shape, whether it's a curve or a surface if it just comes to an end it won't have any knowledge of what kind of shape is supposed to be coming after it's end. Usually the solution to this is to build it as a larger surface all in one shot rather than in totaly separate pieces.

Here are a couple of examples that show the difference with making a shape that stops at a single point as compared to a longer continuing one:
http://moi3d.com/forum/index.php?webtag=MOI&msg=1398.18
http://moi3d.com/forum/index.php?webtag=MOI&msg=1398.19

Like I mentioned previously, I do want to add some tools in the future for smoothing out a surface that was built in a piece by piece method, but even at that point you will likely be better off using the "build a larger smooth surface in one go" method since it just naturally makes a smooth object and building in individual separate patches naturally makes shapes that come to an abrupt end.

The other ways that you can get smoothness is by leaving some space between objects and using the Blend command which puts in a smooth connector surface in between the space between 2 edges, or by building things with some initially sharp corners and then using fillets to smooth them out.

But for your sink to get a smooth shape most likely your best bet would be to get the whole revolution done in one surface creation command rather than only doing one quadrant at a time.

- Michael

Thanks Michael,

That makes very good sense and it will be an interesting and fun challenge to implement the "whole piece" strategy.
I also discovered that you can make meshes out of larger sets of curves! As long as they form some type of logical grid.
I knew about this ability already, I just have to think more ahead to use it.

And thanks for the examples.

Mike

Hi Mike, thanks for attaching the example file.

So I took your curves for the outer surface and joined the elliptical looking pieces together so that there were 6 curves with 2 closed ones (see attached 3DM file). Those curves are all already all nice and smooth with each other (they may have been cut up pieces of one original curve to start with?) so without needing to any other editing other than that joining, you can then build a Network through that whole thing at once, giving this result here:



Looking at that from the underside:



And now compare that to the same outer surface shape in your original model that you did as 4 separate networks rather than all at once:



So that's a pretty good example right there of how to avoid that kind of crease lines appearing between panels - it's generally easier to make things smooth when you build the piece you want to have smooth as all one big surface more like this...

Hope this helps! Sink forms can pretty difficult models to handle where they kind of thicken and change shape in the back part.

- Michael

Thanks again for your time in answering all these basic questions. Now for one more!

What determines the number of points on a surface when you click show points?

For example a default sphere shows 86 points. Since there are an infinite number of points on a surface I have a feeling MOI shows the least number of points possible in order to represent the curvature of the surface but still be workable. Is my assumption correct?

"What determines the number of points on a surface when you click show points?"

It's the number of control points of that you have yourself created when you have drawn your curves ;)
So your number of clicks if you have used a mouse ;)
Then multiply when you make a command betwen these curves (Sweep for example)
And surely yes the minimum for the surface will be the more approached according the curves

For the sphere I believe the number of points is the less reasonable number for a primitive for future rich deformations ;)
Because a sphere's surface can have only 28 control points !

I wonder if there would be a way to make a Rebuild command for surfaces like there is for curves.

Then you could get more points to work with if you wanted them.

> I wonder if there would be a way to make a Rebuild command for surfaces like there is for curves.

It might be easier to edit the number of control points or edit the points in the original edge curves before re-Meshing.

Michael,
Here is the same sink done in ONE MESH action. I'm amazed!
I took the profile shapes and added profile rings to allow the Mesh function to follow along.
So the path follows from one drain ring, around the sink, and back to the bottom drain ring.

I now have a seamless solid shape with (nearly) perfect surface contours.
Thanks for the insight!

Mike

(Framework on left consisting of open shaped vertical profiles and closed rings to follow them. Single Mesh result on right)

Hi Shaun,

> What determines the number of points on a surface when
> you click show points?

It depends on what technique was used to construct the surface - some surface construction techniques will create a surface that directly inherit the control point structure of the curves that were used to generate them, while some other construction methods may go through a kind of fitting process where the surface is refined until it reaches a particular level of accuracy.


> For example a default sphere shows 86 points. Since there are an
> infinite number of points on a surface I have a feeling MOI shows
> the least number of points possible in order to represent the
> curvature of the surface but still be workable. Is my assumption correct?

Yeah, actually a sphere is a special case - the sphere internally uses the minimum number of points needed to make an exact sphere surface but that does not allow for very good squishing, so when you turn on control points for a sphere it actually shows you the control points for a rebuilt version of the sphere with more points around it. In the future I want to make the number adjustable.

You can actually build a custom sphere with a particular number of points in it right now - to do that you would set up an equator circle and a vertical arc (see attached 3DM file) then use Rebuild on them with the # of points mode to set those curves up with the number of control points you want to have in the final sphere, and then create the custom sphere by using Construct > Revolve > Rail revolve - pick the vertical arc as the profile curve and the equator circle as the rail path, and then pick the revolve axis as the line between the endpoints of the vertical circle and you'll have a sphere with that number of points in it as the result.

If you want to build a flat plane with a particular number of points in it, you can use the Extrude command with the "Set path" option for that, see here:
http://moi3d.com/forum/index.php?webtag=MOI&msg=1313.2

- Michael

Also re: spheres - MoI will only show you the "auto-rebuilt" control points for a sphere that is a full surface that does not have any trimmed out areas in it.

If you trim away a little piece of the sphere and then turn on control points, you'll then actually see the sphere's natural control points which look like this:



These points don't deform well though - it's basically 8 chunks kind of glued together and if you disturb these points you can easily make a sharp crease appear between the sections - that's why MoI will show you the control points for a rebuilt sphere instead when it is able to.

- Michael