NURBS and Span
All  1-2  3-6

Previous
Next
 From:  Michael Gibson
5897.3 In reply to 5897.1 
Hi Claas, re: "Single span surfaces" - as far as I know it's just a general attempt to reduce unwanted wiggles in surfaces. It's just a sort of byproduct of single span surfaces that they have the lowest count of control points and so with fewer control points comes fewer chances for them to be arranged chaotically.

That strategy only applies very much to a highly advanced and also finicky form of NURBS modeling that's focused on working on 3D swoopy shapes in a "patch by patch" skinning manner, it doesn't really apply to "Solids modeling" techniques where things are more focused on generating shapes from 2D profile curves rather than working with swoopy organic geometry all the time.

The thing is, that for organic geometry where you're working on a 3D flowing surface that does not have very many 2D elements in it, polygon sub-d modeling can be an overall better fit for that really than NURBS modeling. The "use single span surfaces" advice is roughly the equivalent of "use the fewest cage control points" advice for sub-d modeling. With fewer points it's harder for surfaces to get undulations in them.

But these people who are focused on such things with NURBS are doing pretty specialized work and doing things like manipulating surface control points directly and things like that. MoI is not really focused on such specialized and advanced and super high learning curve type of functions, it's a big focus for MoI to make NURBS modeling more approachable to beginning users and not only to some "high priesthood" of advanced highly trained surface aestheticians.

I've really tried with MoI to very much avoid the need for the user to understand all the underlying mathematical details of NURBS such as curve and surface degrees.


> 1. How is this issue be handled in MOI? Do you have some internal
> mechanism working that adjusts dynamically?

No, it's more that I just don't really focus on that one particular style of "surface by surface patching" modeling and instead try to focus more on techniques for generating models from 2D curves instead which are much easier for people to learn.

But if you do a Blend surface in MoI, that does actually make a single span surface in the blend direction currently.


> 2. Why in 2013 is span still such an issue? I am not familiar with the math - personally
> I just imagined that you can seamlessly blend between two spans as it is a math calculation
> for that curve.

Some of it is just almost superstition I guess, and probably over generalization of advice. It is true that multi-span surfaces can be more difficult to edit directly to morph their end shapes since with lots of control points any manipulation of those control points affects a smaller and more localized region of the surface, so when you go to alter the points with single span surfaces you control more of the surface with any edits that you do. With too much localized effect during editing it will tend to just bend a small area of the surface when you try to match it to other ones rather than making the full surface be affected. A localized warping ends up being more like a small undulation or wiggle in a surface which is not the kind of shape that's usually desired in these cases.

But the benefits are pretty specific to that kind of control point level type manipulations...

Just because single span surfaces are helpful for one particular style of modeling does not mean that they're inherently superior in every circumstance. That's where the over generalization can tend to come into play.


With regards to MoI, it has just never been a particular goal for MoI to be a sort of replacement for a $60,000 highly specialized modeling package like Alias Automotive, MoI is more about the complete opposite of that, trying to make some of the more basic and easy to use areas of NURBS modeling more accessible to people without a massive learning curve.

I actually do expect to improve some areas of organic surface style modeling in the future though too, there are some elements of that in place like Blend and Network already, they just have not been a really major focus so far because that style of modeling is just overall more difficult and a lot of people doing organic style modeling are already using a sub-d modeler to do it anyway.

- Michael
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  F.ip2 (CEKUHNEN)
5897.4 In reply to 5897.3 
Hi Michael,

thank you for your input. I think a person that knows the math code behind it can judge this better
than designers that basically rationalize their approach by the visual feedback they get from the tools only.

Car bodies can be very complicated based on the shapes and I was somewhat surprised that the designers
at GM very much use the square tool for a lot.

I guess thats also why the guys doing the CAD work are not called modelers but Alias sculptors.

Personally I had some bad experiences with Alias where certain blends and surface flows were only
possible by rebuilding the design with single span curves and while I know 3D must be incredible complex
I was asking myself if the math code there is simply not yet so refined or if this is just the way how Alias works.


I was just asking because I noticed that in MOI the user if confronted with creative decisions more than
with technical planing of the layouts - which makes MOI really enjoyable to explore design ideas.

Out of curiosity how do applications like SolidWorks approach the span issue? I know that you can put down the
profile curves etc as well, but it seems to me from the short exposure that they do not have such a high focus
on limiting edit points on a curve like Alias Automotive does.
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
Next
 From:  Michael Gibson
5897.5 In reply to 5897.4 
Hi Claas,


> Personally I had some bad experiences with Alias where certain blends and surface
> flows were only possible by rebuilding the design with single span curves and while
> I know 3D must be incredible complex I was asking myself if the math code there is
> simply not yet so refined or if this is just the way how Alias works.

It's hard for me to know for sure what the particular issue was for that case without examining the actual geometry used.

It's possible it could have been some kind of limitation or bug in Alias, but also possibly your rebuilt geometry was less wiggly or had some other change in form that avoided some particular kind of shaping problem. But I can't really know without seeing the specific example.

The overall code base for Alias is pretty ancient at this point, and such an old code base just usually has lot of various clunky things about it.


> Out of curiosity how do applications like SolidWorks approach the span issue? I know
> that you can put down the profile curves etc as well, but it seems to me from the short
> exposure that they do not have such a high focus on limiting edit points on a curve like
> Alias Automotive does.

SolidWorks is more similar to MoI's orientation in this regard, they really focus a lot of effort on generating shapes from 2D sketch planes and working more on solids throughout their workflow rather than trying to focus on building surfaces one little small patch surface at a time filling in some complex wireframe outline. They have added more surfacing tools over time but they're more often used for building a cutting surface or something like that, rather than trying to build a big model out of a quilt of individual little small surfaces that have to connect to one another smoothly.

So as far as I know there is no special focus or worry about "single span"-ness in SolidWorks either, it's more focused on mechanical part design and not on modeling broad swoopy forms made of some patchwork of surfaces.

Just in general the SolidWorks type solids modeling approach is much more "mainstream" than the Alias Automotive "try to build a big complex surface patch by patch" type modeling. Alias Automotive is a much more heavily specialized tool and that whole process requires a lot of training and experience to be able to do stuff with it. Meanwhile the SolidWorks type of approach that generates models from 2D curves tends to deliver a lot more productivity for a lower amount of learning curve.

One other general factor though is that the kinds of models that are being worked on with Alias Automotive are also though just fundamentally more difficult models because they are not so well defined only by 2D profile curves, so they're not such a good fit with the traditional 2D-profile driven solids modeling approach. I guess that's why Alias continues to exist at this point at all. But also once you get into the territory where your model is not well defined by 2D profile curves then the whole sub-d modeling toolkit can become a better way to form those kinds of organic models rather than trying to do them with NURBS at all. Usually that's what I steer people towards using instead of MoI if they're trying to build some kind of form that is not very well defined by 2D profile curves.

- Michael
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged

Previous
 From:  Michael Gibson
5897.6 In reply to 5897.1 
Hi Claas, or I guess looking into it a bit more, the single span thing you're referring to is more about certain Alias tools only working when curves are touching each other at precise locations like they have to all meet up at knot locations?

Some limitation like that has more to do with the specific mechanism that whoever implemented that command decided to use, certain kinds of mathematical surface construction techniques might have requirements like that, but it's also often possible for there to be different approaches for doing the construction that just go about it in a somewhat different way that does not happen to have the same limits, possibly by doing something like a fitting process.

Sometimes the limited way may have some special benefit like the surface generated by it has the exact same control point structure as the original curves used. But the bad part is that if the command only works with a very specific kind of structure it means you have to know quite a lot of details about how NURBS work before you'll be able to really be able to use it very well...


Some of the other focus on single span surfaces might come from automotive people from an ICEM Surf background, which is a surfacing system that focuses only on using Bezier surfaces which are the equivalent of single span NURBS surfaces, and focuses more on manipulating surface control points quite a bit. It's again a highly specialized tool with a high learning curve and cost involved...


- Michael
  Reply Reply More Options
Post Options
Reply as PM Reply as PM
Print Print
Mark as unread Mark as unread
Relationship Relationship
IP Logged
 

Reply to All Reply to All

 

 
 
Show messages: All  1-2  3-6