What's wrong with my G2 patch strategy?

[eobet]

I've been using Alias for many years (as a student), but I still take way to long time with seemingly basic stuff:

As you can see, I'm trying to fill in the blanks between nice, single span, 5 degree surfaces, which up until this point I have had no problems with cotinuity or uneven CV distribution.

But it's always the last small holes which get me. The top one has crazy spans, and on the bottom when I tried explicit control, it naturally failed and when I try manual alignment I get this (the blend curves were attached at 45 degrees because I didn't think that narrowing a patch to be infinitly thin would work, as the align tool suggests in the image).

Can anyone just draw a few colored outlines suggesting how I should have gone about this, or perhaps have some profound basic tip that I might have missed?

[TheHangman]

Hey eobet

In my point of view, your patch surf. it way to big, making it way too complex, leaving you with not enough control CVS and as you may know the more CVS you add the heavier become the geometry and its difficulty to control remining in this case you only need 3 hulls per side giving you 5x5 patch surface. 

I have drawn something for you hope it helps. Again, if you can not get it right share only these surfaces, so I can have a go for you and give a better analyse of the situation.

Regards

[eobet]

Thank you very much for that image!

I'm always hesistant to add more patches, since it naturally makes the model more complex (and difficult to modify), but I guess in this instance it's really needed.

I will try to work it out and let you know.

[TheHangman · Accepted solution]

Hi eobet

I have thought about and even simpler way.to build the patches for you. 

check this out.

hope you like it.

[ravenzep]

Just split the surface. As easy as that. Your patch layout is fine. And make sure everything is one span.

Good luck!

[TheHangman]

Hi ravenzep

Sounds like you liked one of my approaches. I tried to give the better explanation as possible as requested.

Cheers. 

[ravenzep]

🙂

[eobet]

Well, here's the first result I got (without explicit control) when I introduced some new (imo rather nice) curves to split the patches with:

I'm going to leave it at tangent for now, because it's not a production model anyway (yet).

Also, I don't really know what I could learn by posting the .wire and being provided with a fixed file, unless you record a live video of you actually fixing it and uploading it to Youtube or something, and I can't really ask for that. 🙂

Some day, I'll take another Alias course (when I can afford it).

[TheHangman]

Fair enought 

[ravenzep]

Why did you introduce curves to split surfaces? You should just use detach, which will split the surfaces at any ISOCURVE you want....This would be the standard procedure, and if you press ALT it will split surface right at middle.

Also make sure you are using CATIA V5 construction settings....Alias works quite efficiently with catia v5 settings.

It s quite natural, that you will not achieve G2 in all 4 sides of a surface when aligning. You should massage the CVs and try to achieve it manually. Alias likes to work with this workflow, and it achieves good results. For example, ICEM surf, doesnt like this approach so much, but neverthless, icem's align works differently than alias.

Another thing to note, is that if you dont see G2 green combo, it s not the end of the world. Make sure you have G1, and then work your way out to G2, by using for example zebra mapping to see how the surface is flowing....If G2 is like 0.125 out of tolerance it may not be so bad.... Look at the highlights, maybe that s more important than just a G2 number. Many many times we achieve G2 and the surfaces are bad anyway. So, make a wise use of what G2 means and at which areas it is really important (the most visible ones) .

Finally, I will give you a tip which I tell everyone, and is what I think that makes the click:

Don't model from curves, but rather, model from surfaces. Think in terms of surfaces, and not curves. This will make all work downstream a lot easier.

That's why you see many times over extendend (overbuilt) modeling techniques. They are not modelling from curves or surface edges, but rather from the surfaces itself. I ll give you a little example:

Instead of thinking about a curve, dont see it as a curve, but rather as a intersection between 2 overbuilt surfaces. Get it? This is what is behind the overbuilt surfaces you see in many models. You tweak those 2 surfaces to fine tune the curve you need (intersection curves of the 2 surfaces), Which is a totally different approach than building those 2 surfaces from a common curve on a common edge.

Dont' model from trimmed edges. NEVER. A trimmed edge is the worst kind of geometry that you can have. PLus it will make your modelling too rigid, not allowing great solutions. Try to duplicate a curve from a trimmed edge. Look at those nasty CVs and spans....So you are building geometry from this crap? You will only get bigger problems downstream. Garbage in garbage out. How would you solve this? Easy.... Instead of trimming, project curves on surface as normal curves, and not as COS. (As a curiosity, it is how Icem works. There are no COS in icem. They are all natural curves and natural surfaces.) Watch the deviations, and make sure G0 is kinda near 0, doesnt need to be exactly zero. When you later aling by projection this surfaces to the other underlying surfaces, you know already if it s gonna be easy or not to align G0, because you know already how close it was from the curve u have projected earlier. then u can trim, from the surface edge. Many times, in such case, what is hard to achieve is G0 , and not G2. You control that from the curves you use and their projection on surface. You can tweak the curves very little as you project and watch the tolerance from the underlying surfaces. Something you cannot do if you are modelling from trimmed edges. This offers a totally new level of flexibility and power.

I can tell you from my personal experience, that it takes many years and a lot of effort to be good at nurbs modeling (class-A). So, don´t give up. If you practice a lot, you ll be able to solve your problems. You just have to work a lot, and examine pics u see in the internet in which you can see patch layouts, and try to understand why are they like they are. Why did they do that? That s how I have learnt, and i never took a lesson from anyone. I use alias since the old days it was called power animator, Alias Wavefront. Good old days. 🙂

Well, this is all i could think of to try to help. Read carefully what I described above.... The dark secret is all there.....LOL

I think you are on a good path, judging from your patch layout. Keep working 🙂 Nothing comes easy.

Good luck and have fun.