PDA

View Full Version : Improving Poly Flow



Hopper
04-10-2008, 11:31 PM
Ok, so I've got this little project I'm working on. Simple stuff. I've got a clip I need to fabricate and I want a nice visualization to go by. I could do this on paper, but that wouldn't be any fun now would it.

When I put my model into SubD, it comes out clean with no wrinkles, but I'm just not happy with the poly flow. I've tried a handful of things, but this was the easiest way I could assemble it, thus this is the final product for the base clip. Any thoughts or tips would be quite helpful. Thank you.

Pic 1 - No wires showing no anomolies in the surface.
Pic 2 - Polygonal wires showing the haphazard way I assembled this thing
Pic 3 - SubD mode showing unhappiness...

kilvano
04-11-2008, 12:05 AM
Is that a laptop stand? Ive seen loads of companies churning new ones out these days.

Surrealist.
04-11-2008, 12:17 AM
Actually your polyflow is perfect. You just have about 2 times too many.

OK look at the clip as the slit capsule-shaped hole being the center. Let's call this the front. By my count you could use BandGlue 3 times on each side,(one pair of loops each time) both on the top and the bottom. That would reduce those polys by half.

Then looking at it sideways - assuming the slit is the front.

There are two bends. One on the top and one on the bottom. You could use BandGlue here too. Just start at one end and work your way around. Do that once (one pair at a time), then maybe one more time until you still maintain the shape but reduce the polys.

Then do the same on the flat parts (top and bottom) but you could do more rows at a time until you have none in the middle and only an edge at the ends to hold the shape - then of course the ones you left at the bends.

lightripple
04-11-2008, 07:29 AM
You have about 5 times more polys then necessary. Edge weights my friend...

Surrealist.
04-11-2008, 07:51 AM
From my experience and testing, edgweights actually require higher res object resolution at render time to work properly.

This applies to the whole object which is a waste of resources. With subpatch modeling you can control what parts of the object get the detail. If you control your polyflow correctly - that is make sparing use of control edges in the right places - you can actually do a very good job at making lean objects that render quickly.

lightripple
04-11-2008, 08:14 AM
aps and use the edge weight map...

lightripple
04-11-2008, 08:15 AM
That could be modeled with 5x less polys at rendertime.

Surrealist.
04-11-2008, 09:24 AM
The problem that I have had with edge weights is that they produce smoothing errors. Now I have not experimented much with APS on edge weights. If you have some tips and some examples I'd love to see them because I would love to play with them more and this would give me something to work from. However in my experience with edge weights and APS I believe you would still get render errors on the parts of the object that are not rendered at high res. So that is my interest on it. How do you over come that?

lightripple
04-11-2008, 09:32 AM
I havent ran into the ploblems with edge weights that you are talking about...
I will post a scene with edge weights and aps...

lightripple
04-11-2008, 09:37 AM
Are you seeing smoothing errors in the render or just ogl?
I have issues with errors in ogl, but not the render, and usually just modelers ogl...?

lightripple
04-11-2008, 10:27 AM
here is a quik test... pressf9

Surrealist.
04-11-2008, 10:36 AM
I see them in renders. Thanks for the scene I check it out.

Surrealist.
04-11-2008, 11:24 AM
Great. Checked out the scene. Thanks for that. That is a brilliant idea using the weightmap to drive the APS for the edges. I hadn't put those two ideas together before.

The weightmap idea is pure genius. But the box is a bad example and actually points out a problem I have with edge weights - at least I think for now. But to really be able to make a comment on it, I am going to have to do quite a bit more experimenting with it. But on my side of the globe it is past my bed time.

I have a few things to address with this, and this will give me much to work with tomorrow and I will comment on my findings.

Thanks again for sharing this cool technique. Going to keep me busy. :)

Hopper
04-11-2008, 03:11 PM
Is that a laptop stand? Ive seen loads of companies churning new ones out these days.
It's a much smaller piece - only a few inches long, but funny you should mention it. I noticed one of my stands at work looks very similar to this!


OK look at the clip as the slit capsule-shaped hole being the center. Let's call this the front. By my count
Thank you sir. A pleasure to learn as usual.


You have about 5 times more polys then necessary. Edge weights my friend...
agreed. I bandsawed the crap out of it just so I could have enough edges to connect the curves and get 4 pt poly's. This is my inexperience showing again.

Edgeweights bother me. It's not the process, it's the fact that I always wind up getting screwed by them. They're still too buggy for me. I have to admit, I'm sure the manner in which I'm using them is probably all wrong, but I've been burned too many times.

Hopper
04-11-2008, 04:37 PM
Ok .. So how's this? That REALLY did the trick! I went from over 2000 polys to 482. Bandglue caused a few non-planar polys but just 3 or four.

If stretching points for alignment will distort your model (in order to get rid of non-planars), can you suggest a better way of doing it?

Thanks guys. I appreciate it.

Oh .. side note: I had that many polys to start (on the flat parts) because I started with a box and subdivided it to use Bend to get the shape I wanted. I know, Spline Tool would have been better. I went back and did it that way and came out with the same shape (minue the million segments), but more points on the curves.

lightripple
04-11-2008, 04:43 PM
looks pretty good...
nonplanars or fine when using subds everthing gets tripled at render...

Hopper
04-11-2008, 05:53 PM
True .. I just hate seeing that liittle "3" flipping me the bird from the statistics panel.

JeffrySG
04-11-2008, 09:23 PM
That topology looks much better to me. You could also try it like this to save a few more polys. Add green edges, delete red edges.

57566

I think you could probably get away with less loops horizontally on the bends. You have six on each bend, you might be able to get away with three or four there.

It would actually be cool to have a challenge here on the forums where we all have to create the same exact object but use the most efficient topology. Then we can see how everyone solves the problem a little differently.

Surrealist.
04-11-2008, 10:58 PM
Ok .. So how's this? That REALLY did the trick! I went from over 2000 polys to 482. Bandglue caused a few non-planar polys but just 3 or four.

If stretching points for alignment will distort your model (in order to get rid of non-planars), can you suggest a better way of doing it?

Thanks guys. I appreciate it.

Oh .. side note: I had that many polys to start (on the flat parts) because I started with a box and subdivided it to use Bend to get the shape I wanted. I know, Spline Tool would have been better. I went back and did it that way and came out with the same shape (minue the million segments), but more points on the curves.


I would not worry about non planars. But if you want a fairly efficient work flow for this you can try this:


Basically create a spline and use Rail Extrude with uniform knots (count your points and enter accordingly) on a rectangle. This will give you a good base to start with.

From there I just used Band Saw pro, add edges and of course Boolean to make the initial hole to come up with the final geometry.

Surrealist.
04-11-2008, 11:02 PM
Subpatched.

Hopper
04-12-2008, 12:02 AM
Here's a test render... I really like how it's turning out so far (simple as it is).

I thought I'd share my painful render times by adding the render window. This is why all my models are butt simple. I can't take the waiting... :D

Hopper
04-12-2008, 12:09 AM
That topology looks much better to me. You could also try it like this to save a few more polys. Add green edges, delete red edges.
Interesting .. I'll give it a spin and see how it turns out. Thanks!



I think you could probably get away with less loops horizontally on the bends. You have six on each bend, you might be able to get away with three or four there.
I did originally, then the bends started getting rough, even after I spaced them out evenly for better flow. I was starting to see the geometry through the render.[/quote]


It would actually be cool to have a challenge here on the forums where we all have to create the same exact object but use the most efficient topology. Then we can see how everyone solves the problem a little differently.
It would definately be a challenge and a great learning experience for us beginners. :)

Hopper
04-12-2008, 12:14 AM
Basically create a spline and use Rail Extrude with uniform knots (count your points and enter accordingly) on a rectangle. This will give you a good base to start with.

From there I just used Band Saw pro, add edges and of course Boolean to make the initial hole to come up with the final geometry.
And I'll bet that took you all of about 5 minutes to create. You'd laugh if you knew how long it took me to make that simple model. You should have seen the circus of point manipulation I did to make the slot. I actually made it in another layer. I created two discs, then welded them together, then multishifted the hell out of it to get the shape and bevels. I then pasted it into the main layer, aligned the top and bottom surface using the Info panel, then welded everything up. And of course, after I was done, I immediately saw the "easy" way I should have done it. :foreheads

Now if I could figure out the Nodes to make a "good" brushed aluminum, I'd re-texture it.

Surrealist.
04-12-2008, 01:43 AM
LOL! thats how I you learn too. If you figure out the brushed metal, post it. I wanna see.

Surrealist.
04-12-2008, 01:44 AM
It would actually be cool to have a challenge here on the forums where we all have to create the same exact object but use the most efficient topology. Then we can see how everyone solves the problem a little differently.

That would be cool.