topic Re: TS CV layout density and also G2 edges in Fusion Design, Validate & Document Forum

TS CV layout density and also G2 edges

cekuhnen — Mon, 15 Aug 2016 01:53:22 GMT

left TS Fusion - right MOI SUBD

comparison - I was just curious.

What I like about the MOI result is the light CV count but as one can see Alias clearly states that edges are not cleanly G1.

Also the highlights show some noise in the MOI parts.

But I was curious if the CV layout can be changed in Fusion actually to make it lighter.

With bigger models the load can be an issue.

Also any idea of G2 at one point is possible with TS? The main reason to ask is because without G2 the curvature graph is less usable with G1.

Re: TS CV layout density and also G2 edges

jakefowler — Mon, 15 Aug 2016 05:43:21 GMT

Hi @cekuhnen

The extra CVs are the result of the 'capping' done at star points on the T-Splines body. This is how T-Splines maintain G1 continuity at the star points - MOI's conversion doesn't appear to worry so much about the continuity across patches, which allows them to produce lighter results.

Unfortunately we don't offer a way today to control this. I don't know if it would even be technically possible, unless you were willing to sacrifice G1 continuity.

I'll follow up with the T-Splines geometry team to see if they're working on anything new in this area. There have been discussions about whether we can make caps G2, but my understanding is that it's an extremely big project, and therefore probably won't be coming the near future I'm afraid.

Thanks,

Jake

Re: TS CV layout density and also G2 edges

cekuhnen — Mon, 15 Aug 2016 11:43:58 GMT

@jakefowler

A lof of apps offer a SUB-D to NURBS and even claim G2 but well Alias often states differently. MOI does not claim it - I used used it to show actually the G1 quality of T-Splines.

To the naked eye the NURBS surf seems fine but when evaluating problems can be visible.

The G1 curvature comb with T-Splines is like using a comb in Alias with a curve that has many segments - it somewhat is pointless or very hard to use it which is why I brought up G2.

But I can understand that the math behind that might be an issue in terms of computation needs and performance.

Re: TS CV layout density and also G2 edges

jakefowler — Mon, 15 Aug 2016 12:44:12 GMT

@cekuhnen

This post from the T-Splines forum has some good background on the math/philosophy behind the capping geometry:

http://www.tsplines.com/forum/viewtopic.php?f=9&t=30144

Regarding curvature combs - there should only be G2 discontinuities on the edges that enter a star point. Are you referring to this, or other edges along the surface? All other portions of the surface should be G2-continuous (the sharp corners you get in the red line running along the comb are indicating G3 discontinuities; to avoid this, T-Splines would need to use degree-5 NURBS underneath rather than degree-3 - again, something that's been discussed, but again, a very large project!).

Thanks,

Jake

Re: TS CV layout density and also G2 edges

cekuhnen — Mon, 15 Aug 2016 13:01:45 GMT

@jakefowler

I must have overlooked the fact that in simple non star patches the edges are actually G2.

The problem however is that also the graphs on a surface - so not along the edge - are hard to use.

For example here I rebuild a cross section with 3 curves and the comb is nice and smooth along the curves

That is why I asked about the CV layout. If you compare actually the combs in Alias between the T-Splines and MOI result

you will see that MOI is actually showing a lot less noise.

How-far this with T-Splines will actually result into also being visible in the manufactured product I do not know.

I have no CNC or injection molding tool to verify that.

I also know that in Alias I use more a class a technical surfacing approach - not that I want that in Fusion.

But the combs are just hard to read in TS.

My main curiosity is to get better surface vis options when sculpting with TS without having going too much towards Class A nightmares.

But I am not sure if that is possible.

Can the comb maybe be interpolated or smoothed to filter out the noise?

Re: TS CV layout density and also G2 edges

jakefowler — Fri, 19 Aug 2016 10:46:13 GMT

Hi @cekuhnen

I've discussed these issues with the T-Splines core team. Some good news: we have an update in the works that will both reduce spans and improve curvature characteristics at star points. Here's how your example looks before and after the upgrade (the spans nearest the star points have been eliminated; there are still some curvature inflections near the star points, but they are considerably more subtle and more localised):

What's more, this new star point capping should also improve modeling performance significantly. This update should be released around late September.

Regarding G2 caps: according to the team some early research has been done on this, and it is theoretically possible, but it would actually result in surfaces with a lot more control points. The balance is achieving technically smooth results that also look natural. As mentioned, it's a large project, so probably won't come in the immediate future unfortunately.

Another tip: the non-G2 region is localised to the T-Splines faces directly surrounding the star point. So (exact) subdividing the faces surrounding the star point a few times will decrease the size of the non-G2 region and improve surface quality:

The downside is that this obviously increases the number of surface spans, which could impact performance.

Hope this provides some good news at least,
Jake

Re: TS CV layout density and also G2 edges

cekuhnen — Fri, 19 Aug 2016 11:12:58 GMT

@jakefowler Thats great looking forward to the update since this will be a topic for me presentation at AU!

Re: TS CV layout density and also G2 edges

jakefowler — Fri, 19 Aug 2016 15:45:35 GMT

Oh cool! Not sure if I'll be there this year, but if I am will definitely come watch.