Hi Michael,

> So in both these cases for NX and SolidWorks,
> did you have to manually delete the "default fillet"
> star-structured corner and then construct the
> n-sided patch as a separate step? Is there no
> way for the initial filleting to make n-sided corner
> patches from the start?

Yep, I had to delete the star-structure the default filleter doesn't handle n-sided patch that's a seperate feature, NX call it 'n-sided surface' and SW call it a 'boundry surface'

> I think it's actually too light and maybe needs
> to have parameters fiddled with to give it more
> control points to work with in order to hug the
> boundary's normals better.

I'm not a Solidworks expert, I did use a mates seat of the software so I'm not sure how much further it can go, if you like I could have another look.

-

Hi Danny,

> Yep, I had to delete the star-structure the default filleter doesn't handle n-sided patch
> that's a seperate feature, NX call it 'n-sided surface' and SW call it a 'boundry surface'

Hmmm, well that's an interesting difference from ACIS which does an n-sided blend as the default corner making method in the filleter...

Neither kernel really does quite exactly what I'd really like which is an n-sided blend in the filleter but with a good quality NURBS surface result.


> if you like I could have another look.

No, that's not necessary, thanks.

- Michael

ViaCad is switching to a "surface cover" they call it. I think the difference between the more expensive ones, is that ViaCad is not giving control over the degree of the generated cover. The control points get generated the UV, but are very sparse where they need to "Warp". It seems ViaCads follow an "IsoCurve"type of pattern where the others create a nice, full grid and drape as necessary. The ones that trim at the boundry need to extend and twist way out in space to generate the surface, but the sparse points become wrinkles on the nurbs surface.

Maybe Blowlamp can point this to Tim to look at. Control over the UV degree may get a better NURBS out of ViaCad.

Hello,

Interesting result from VSR MultiBlend, to the right. Middle is the NX example, SW the left.

Hi OSTexo,

VSR?? is that Rhino?

-

Hi OSTexo, can you please post the 3DM file of the VSR one? It would be interesting to see if it has any star-like wavy pattern on it or not.

- Michael

It's fascinating how the different software tries to tackle this problem.

In ViaCAD Pro there's a tool to Untrim a surface. So after converting a solid of this test part into surfaces (which remember destroys the original integrity of the nice ACIS produced patch), I Untrimmed this patch and like BurrMan did, switched on control points.

The result is in the picture which to my eyes looks a mess and surely not what I would expect ACIS to produce.
A 3dm file version is also attached.

I hope this discussion helps Michael to engineer a good solution from it all ;)

 

Martin (2).

Hello,

Yes, it's Rhino with the VSR Shape Modeling 2.0 plugin. The plugin is twice the cost of Rhino, but it's paid for itself a few times over with its version 1.1 capability alone. The Multiblend feature is new in 2.0, I thought I'd give it a try. I'm sure that surface can be tweaked to be even better with VSR by someone who actually understands surfacing concepts. Here is the 3dm.

Hi all,

i find this thread very interesting and thought how about using a poly model created in MoI, then beveled in modo and finally bought into T-Splines demo for Rhino. Here are the results:





Regards
Stefan

Do other apps and kernels provide for a multi-radius blend like in this pic?


Martin (2).

Here is the result using cover surface tool on sharfx to make a cover on the hole (I think viacad also have this one), first attempt don't go fine, until i select "export selected only" and "end line as pc" options on export dialog. This way the solid goes flawlessly into Moi, also, sat was exported as v4.0
here is the resulting file









Luis G

Michael, this pic here is a standard G2 fillet done in MoI... it isn't all that bad for a "Star" patch.

Was this done with the 'star' patch routine or was it some special analytical case just for fillets?




I would take the 'lumpy' resultant 'star' mechanism over an empty hole any day!

And perhaps including it could be a stop-gap until a better system is implemented, kinda like how you improved the Network later.
Since the 'star' patch is in the system already, maybe just a non-UI text command would suffice.

But none the less, I want not to sound pushy, I really do appreciate you're open collaboration with us all.

:-)

Hey Blowlamp,

""""""""""""The result is in the picture which to my eyes looks a mess and surely not what I would expect ACIS to produce. """""""""""

That happened to me if I tried to "move" a control point after the points were on. The model wouldnt undo from it. It could be display driver related.

Hi Luis.

Good to see you here from the ViaCAD/Shark forum!

Sorry though, but I can't get any of that to work for me - any chance of a bit more detail please?


Martin (2).

Hi Burr.

That picture is the result of un-trimming the n-patch after first converting the solid to surfaces. I also included the file of the patch for anyone interested to look at.


Martin (2).

Hi OSTexo, thanks for posting that VSR blend - that's one of the best star type blends that I've seen. There is no hint of lumpiness between adjacent left-right neighbors of the filling surface, so it doesn't have the kind of residual star pattern artifact that seems to be usually present in this kind of blend.

The main quality problem that I see with it is that it has a somewhat flat-ish shape through most of it and it kind of has a bit of a sudden bend in it right when it comes to the outer boundary. This is interesting since with other star blend mechanisms it's this area that is actually well shaped.

Also the surfaces generated by VSR are very light - they are higher degree (degree 5) single span surfaces instead of cubics. This is a fairly different kind of structure than the regular star blend mechanism which is usually based off of making degree 3 Coons patches for each surface. I think that VSR is kind of specialized in its focus on working with single-span surfaces, undoubtedly some general techniques coming from ICEM Surf coming into play there.

It's not exactly perfect but I think it does show that the star blend mechanism is not necessarily 100% fatally flawed.

Using only single span surfaces would probably have some downsides in other situations though, like if you were trying to fill in a hole where the edges were sort of wiggly and not so uniformly shaped a single span surface would probably not match up very well then...

- Michael

Hi Stefan, the t-spline result is very interesting! The interior region of each area seems to be nice but it kind of looks a bit odd at the borders between every shape, I think there may be some fairly tight little bends in the shapes where pieces meet up.

Notice here how dark the "fillet" surfaces are right where they meet up with the planar area:



I think there is some kind of little small sized lump or tiny dip in those areas, it kind of makes the juncture between all the cylinder shaped fillets and the adjacent planes to be kind of more accented then they normally are, that's kind of what catches my eye as being slightly weird looking with this example.

I also extracted some of the edges of those surfaces and examined them in Rhino using a curvature graph and yeah you can see that there is a sudden spike in curvature right at that plane/fillet juncture area, so there does seem to be some irregularity in shape in these plane/cylinder juncture areas when using this technique, you kind of basically lose the accuracy of planes and cylinders with things all glommed together into big surfaces like this. Here's the curvature graph of one of those surface edges:



A normal cylinder/plane juncture would not have that sudden spike in curvature where they connect together, there would normally be no curvature for the planar part, and then only a constant circle curvature for the cylinder part, this has a kind of little dip or bump where the 2 shapes (planar and cylindrical) come together.

So it kind of opens a bag of worms of problems in the plane/cylinder areas where the regular NURBS objects are well behaved...

- Michael

Hi Burr,

> That happened to me if I tried to "move" a control point after the points were on. The model
> wouldnt undo from it. It could be display driver related.

That happened to me too, I think it's just a bug in undo in ViaCAD, probably not driver related.

- Michael

Hi Martin,

> Do other apps and kernels provide for a multi-radius blend like in this pic?

You mean with 2 different radius values coming together at a single juncture? Yes, a lot of CAD programs do allow for that, and I believe it's even supported in the geometry library that MoI uses, I have a general plan to make that possible in the future by some method in Fillet, probably something like some button where you could say "new fillet set" and then pick a new sequence of edges to fillet with a different radius value without leaving the fillet command.

- Michael

Hi Mike,

> Was this done with the 'star' patch routine or was it some special analytical case just for fillets?

It's a general star-patch hole filling routine that is calculating that in MoI.

And yeah it's not exactly horrible but if you generate a heavier tessellation of the model you'll see a kind of star pattern of ridges radiating out from the center point. Often times the ridges can be fairly pronounced especially when the boundary is not quite so uniformly shaped as in these fillet cases.

- Michael