Use poly or nurbs to model a car?

[max854]

i have see many poeple model a cool car and most of them used poly and smooth method . why few people use nurbs beside i also for character most also use poly smooth to create cool model . i feel few and few people use nurbs why? i feel we need not to learn maya nurbs

[joojaa]

Most people dont use nurbse because they do not now how. Its not exactly hard ist just that they dont bother spending more than 10 min on the subject.

[bmcaff]

The answer is simply that Polys are faster and don't require as much technical knowledge as Nurbs. But lets face it, Nurbs are much more accurate than polygons (even in Maya, which is by no means a good nurbs modeler). Cars are a good example of this point . I've seen hundreds of car polygon car models and some are indeed brilliant but most fall into the 'obvious Poly model' category. In reality all cars (and all manufactured products for that matter) are modeled by designers and engineers using Nurbs (or solid models).
The goal of this is to produce perfect, accurate manufacturable surfaces and components.

So I guess it all depends on your models application. For games or non-close up work then quick and dirty polys will do the trick. But for accurate products wih complex surfaces and small details then maybe Nurbs would be better (characters I don't do so don't know).

As for the need to learn Maya Nurbs, I don't think you can dismiss it because you feel that so few people use it. All tools have there application and you should at least know the basics of nurbs modeling because it might come in handy sometime. There is nothing stopping you mixing nurbs with poly and Sub-Ds.

[falloutboy]

Hi just a little question, because i always thought Maya has a good Nurbs - modeler:
Whats is wrong with it?
I took a short look on a Studio Tools tutorial on the alias site and thought to myself: Hey, this is better.
Concretely i am talking about the ROUND toll wich seems to beat the FILLET tool.

[bmcaff]

This is very true... but then Studio tools is geared towards industrial designers who use these features all the time. The tools for secondary surfacing like filleting and rounding are very versatile allowing multiple radii along edges as well as knee radius. Not to mention the fact that you can work to very tight tolerences and produce stitchable surfaces for export to proE, unigraphics etc.

What's wrong with maya's nurbs? well maybe I was a bit harsh in saying it's not good but it does not allow you the degree of control, versatility and accuracy that studio tools does. Having said that Maya is great at what it does and it really is not meant for product or serious automotive modeling. I just felt that alot of great features from studio tools and Power Animator were left out when Maya first came out. I have to a say that for nurbs and the type of modeling I do the work flow in studio tools suits me better (maybe 'cos I've used it for so long).

[falloutboy]

Arrgh!
I have been trying to reply so often, but my answer has never been taken by the board.
i donot know if it is working now, when my FW is off.
Anyway, studioTools are far too expensive for me so i will quit thinking about it.
If Maya does the same job, but less accurate, this may be the best solution for visualisation.
Hmm, just one thing.
Why do i get these little gaps when working with TRIM and FILLET? Any technique i found on the web, that should fix that problem, wasnt perfectly solving this.
Sorry should make a new thread about that.
Greetz
FB

[joojaa]

Ok and again the answer for this (this pretty much would need to be in the faq, it gets asked a LOT)! There is no gap, or there is its very very small (within the set tolerance offcourse). BUT what you are seeing on screen is a approximation of the real nurbs surface not what the surface is, so if you increase tesselation the gap gets smaller, and if the gap is maller than one pixel it will not show up in renders.

To realy view how close the surfaces actualy are try rendering with some micropoly dicing renerer like mental ray (jus set tesselation overrride to fine vith length of 0.5 nas check the view dependent) or Prman.

[falloutboy]

Hi that is right.
But i increased any responsible value to the limit and some tiny gaps are always left.
The best thing i found about that was:
http://www.animagraph.com.br/DeathBrain/Mo...orial\_page2.htm

What can be added is an increasement of curve Precision shaded, but that is only display stuff.
Cu

[bmcaff]

Yes, even with fine approximation there might be gaps in your model where the round edges meet the trim surfaces. This seems to be an underlying flaw with the round operation in maya. I thought that it was just the trim and surface construction tolerences being too loose. But even at the tightest tolerence these gaps still appear.

I compared a simple model constructed in maya initially and then rounded in Maya & studio tools and rendered in Mental Ray with fine approximation so I could get a fairly objective result. The Maya round was generated at the tightest tolerence while I got 2 tolerences in studio tools (one at ProE settings and one at general CAD settings). The tight tolerence of ST is of course ment for producing solid stitchable nurbs for export to ProE which is a bit unfair to Maya so the loose tolerence is more equal to the tolerence offered by Maya. Here are the rendered results plus close-ups. (all including the close-ups were rendered at 1600x900 and resized to 640x480 for web)

Maya
Maya closeup
Alias ProE
Alias ProE closeup
Alias General
Alias General closeup

Even in the like-for-like comparison Alias seems to be much better. (Actually for production surfacing (not rendering) most of the rounding and blendings is done using UG or ProE, I remember the engineers I used to work with were always complaining about the accuracy of Alias rounds so we started doing all of that detailing in UniGraphics).
This, while it is usually not a problem for the likes of animation and such, is a big problem for high res visualization.

The question is, is there some way of getting the same results from Maya as Studio tools, and if so, how? Also, has the Nurbs engine been changed from that used in PowerAnimator (same as ST)?

[falloutboy]

Hi,
first of all i am not familiar with ProE or UniGraphics.
Seems to be some ST comparable stuff.
But, altough i am german, i got 99% of your post;)
So maybe it makes sense to be Maya that inaccurate, because acceptable results can only be reached, when increasing the values a lot, which means the machine to slow down, while you are often having crowded scenes in Maya( compared to ST, hmm...thats what layers are for).
Just an idea- maybe a bad one, because once the fillet would be stitch(able)ed/ attach(able)ed( what you cannot do with the results in Maya), you wont get into these troubles anymore, right?
But thanks a lot for the pics wich are calming me, so i am not too stupid, to manage these things.
To me these results are not acceptable, because you can even see this gaps from a normal distance.
But i know now: Thats the way it is.
FB

BTW. POWERANIMATOR has really been better with this?

[joojaa]

Well actualy the precicion of your world comes to play in here! In both proe and alias the worldspace is much higger than mayas. So If your object was 100 units wide instead of 10 youd get better results. Atleast for me it always does! Yeah uits acrobatics but works.

By the way ive never ever seen a program that does as good rounds as proe... Studiotools comes close but proe is better.

[bmcaff]

Absolutely true, ProE does a great job on round blends, especially where multiple blends converge. This is a great big pain in ST.
As for worldspace, I guess it forces Maya to tighten tolerences. Good point.

[falloutboy]

Hey, cool.
Thats exactly the hint i got from GNOMON.
Scale up> Fillet> scale down.
Cant we increase the world space?
Must test it, when back at my Desktop.
THANKS A LOT for all your replies, because i really want to get more familiar with NURBS.

[falloutboy]

But wait (sorry, i cannot test it right now), wont there be a problem to keep everything in the right postion when scaling it down together?

[bmcaff]

group it!

[falloutboy]

Hi.
No, back at my desktop i am just getting the results, that i expected, because things are scaling down relatively inside the group.
Gnnn, maybe its my influenza holding me back to solve this possibly simple problem.
please help

[bmcaff]

Aaah, ok. Myabe Delete the history for all the surfaces and then group and scale.

[joojaa]

Hooold on guys im actualy having some slight problems on this thread... So let me summarize and explain better. (while the tesselation picures on this thread are nice the clonclusions drawn from them are total crap, sortof barking at the wrong tree! Sorry been to busy to think! Or check every claim for that matter)

I Will do an explanation as soon as i can get theese pictures done and uploaded! Then im going to link this to the faq.

Sorry the fault is entirely mine... I should know better.