Hello everyone.
I'm not currently an MoI user because it doesn't have some of the features I need, but I have used it and really appreciate the care that's put into its development. I am however, a ViaCAD Pro user and that's the reason I'm replying to this thread. I just wanted to point out that the shape posted by Michael, can be blended in ViaCAD if you use the right option. It worked OK for me with a blend radius of up to 2.4 before giving an error message. To do it successfully you need to use the Variable Blend tool and select the Fixed Width option within that tool. I'm including the SAT file I made, for your scrutiny.

All the best.
Martin.

Hi Martin, that's definitely an interesting alternative. But the example you posted is actually kind of lumpy and not so great in that area, the surface has kind of gotten bunched together.

Here I've turned on the metallic lighting in MoI and examined your model. Notice these darker bands, which are actually little lumps in the fillet surface:






Compare that to this result created in MoI with radius = 3:






So not only does MoI allow you to go to a higher radius that ViaCAD is not able to reach, but also the surface generated for this particular case seems to be higher quality as well.

That's just this particular case though, there are quite a lot more cases where ViaCAD handles filleting better than MoI does, but not completely 100% of the time as this example shows.


- Michael

Hi Michael.
That's interesting. ViaCAD won't let me do a radius larger than 2.4 - it throws an error if I try 2.5. As for the surface quality, I must say it does seem OK within ViaCAD Pro. I've included a couple of screen shots of the shape with the Surface Analysis -> Environment Map applied to it. Do you think what you're seeing is something to do with exporting it as a SAT file?

Martin.

Hi Martin,

> Do you think what you're seeing is something to do with
> exporting it as a SAT file?

No, most likely it's just that the analysis display in ViaCAD is just not detailed enough to pick up on the small sized little bumps.

I've seen similar problems in other CAD programs as well - remember that the shaded display is actually generated by creating a polygon mesh approximation of the true surface. If too few polygons are created they can actually miss little ripples and you just don't see the true surface quality.

One of the nice things about MoI's viewport display is that it puts a much denser set of display polygons up than you typically see, (including putting polygons along every surface knot line) so it is able to more accurately display problematic surfaces that other programs just kind of gloss over.


In ViaCAD if you set the object resolution to "super fine" from the default of "very fine", you can start to see the bumps a little bit more, here I've re-created your fillet in ViaCAD rather than importing it from your SAT file and you can start to see a bit of the lumps in these areas:



see the traces of those 2 vertical lines in there? Even at ViaCAD's "super fine" resolution there are actually still not enough polygons being generated to be able to see it properly though. But you can see it a bit easier if you move the view around with the highlights running through that area.


It's very unlikely that ViaCAD would produce bad SAT output, because SAT is the native format of the ACIS geometry kernel that ViaCAD uses.


- Michael

Hi Martin, also those little bumps are related to why ViaCAD cannot make a radius greater than 2.4 - what's happening is that the fillet is starting to kind of cave in on itself at the top. At radius 2.4 it's kind of awkwardly bunched together and at radius 2.5 or higher it becomes not just bunched but actually has a portion of it kind of inside out with a self-intersecting surface.

It's not unusual for various NURBS surface fitting mechanisms to begin to get unstable with little bumps and ripples when they start to closely approach that kind of self-intersecting state.

That same kind of thing can happen in MoI as well, but in some cases like this one MoI's filleting mechanism is able to detect that the fillet is going to be self-intersecting and it stops the fillet outside of the area where it would become collapsed and puts in a different kind of blend patch in there. It isn't able to do that in every fillet case but with extrusions it seems to work pretty well.

- Michael

This is a excellent and very informative thread, thanks to all, especially Michael for the detailed explanations on handling situations like this!

-Will

Agreed. Thanks for furthering this thread. Although every nurbs modeler has it's strengths, weaknesses and differences in areas such as filleting, I think MOI has the happy medium of not too much but overall just enough. Besides, this program isn't meant to become the next Rhino or other high-end nurbs program.

Thanks for taking the time to explain all of this, Michael.
Even though I'm not a licensed user of MoI I like to look in on this forum as it's friendly and informative. In fact, it's not unknown for me to direct some of the more interesting topics and innovations here, over to the ViaCAD forum. So I hope none of you mind an interloper in your midst.
The only thing that's still a little puzzeling, is that none of these artefacts seem to show in the Surface Analysis tools that are part of ViaCAD Pro. So does that mean it's just a display 'feature' rather than something fundamentaly wrong with the geometry? I include a couple of Zebra screen shots for reference.

Martin.

Hi Martin,


It's most likely an issue with ViaCAD's surface analysis tools - as I mentioned before ViaCAD is just not creating a dense enough mesh to pick up on these ripples which are quite small in size.

But also on top of that zebra stripes by their nature are rather better at displaying how things connect at surface edges where there is actually a break in continuity. It's not so good to detect little ripples that are located within a single surface. Those ripples are actually not continuity breaks, the surfaces have full continuity throughout their interior areas.

It's easy to jump to the conclusion that full continuity also means full quality, but that's unfortunately something that many people are easily mistaken about - you can have full continuity but problems with bumps and lumps.

If you look closely in your zoomed in zebra analysis shot, you can see some of the jagged artifacts that are coming from polygon outlines:




Those polygons are larger than the little bumps, you're not going to be able to see the bumps unless you get polygons that are small enough so that they actually follow the surface more closely.

Even then Zebra is not really the best tool for looking for bumps, glossy specular highlights tend to be better for bump analysis but only if you have a dense enough mesh.


Also as I mentioned, those little bumps are actually related to the reason why the fillet fails in ViaCAD if you go even a little bit larger in radius - the next little bump up in size changes from a kind of awkwardlly bunched area into a full fledged self intersection which is kind of more traumatic.

If you backed off just a slight amount in radius probably those bumps would go away as it would avoid some of the bunching that is causing it.


But I'm pretty sure you've got those bumps in ViaCAD as well (with the radius set to just before it fails), you're just not able to see them because the display mesh is not fine enough even at its "super fine" level.

I've seen the same kind of problem (can't see tiny surface details due to coarse meshing) in other CAD programs as well, it's not particularly unusual. But one of the nice things about MoI's regular display is that it is detailed enough to be able to see those kinds of things.

- Michael

Hi Martin, here's another example examining the fillet surface you posted with zebra stripes in Rhino.

If you have a kind of loose density mesh like this for example:



Then it's not too apparent:



Now notice what happens if I up the density by many many times so that the analysis display uses little tiny polygons that hug the surface more closely. With a mesh like this:



The accuracy of the analysis display gets increased and you can see some of those details that were previously getting essentially "blurred out" by a rough mesh:




I don't know if you have any controls available to you in ViaCAD to turn the display mesh up to a really high density as shown above, or to display the outline edges of the display mesh that is currently being used. If you do have controls like that available then crank them up until you see that the generated mesh has very small polygons as in the second example above, and that should help you to see these small surface details. In that example I told Rhino to break down any polygon that was larger than 0.025 units in size.


Also if you're trying to notice bumps using Zebra (which as I mentioned is not really my favorite tool for that job) try to set the zebra stripes to be thinner if possible.


- Michael

Hello,

Hi armin, and Michael... thanks for your input, I just have no luck with my hardware setup then, I wish I could have some more time to send bug reports etc, maybe I'll give Viacad a third try later this year ... I'll first have a look to the Alibre solution now.

Hi Ralf-S,

This t-flex product seems quite complex for my needs, but thanks for sharing ... I'm really scared by all this historic based modeler, of course I understand all the advantages, but it's not for me ...
I'm really not a tech guy.

During my Viacad tries, I also discover that in some scenario, MoI is giving a better result (as Micahel explains) ... so I end up with a huge waste of time, testing for every fillet scenario, which tool is giving the better result ... maybe also a matter of experience :P ... but very not productive.

Interesting discussion you got here.

I tried this in Spaceclaim with radius 3 and turned the display quality to max (guessing this adjusts the display mesh) and turned on zebra stripes.
Spaceclaim is Acis based, and yes There seems to be a problem with the fillet here too.
It's probably no surprise but Spaceclaim also struggles with radius of 3.
I got the impression this is indeed a real fillet problem, not just a mesh coarsness probelm wen turning the model around.



Michael, I think it's great how you take these things into consideration so the meshing is gives "as accurate as possible" display of the surface.
Your commitment to software quality gives us a better chance to be confident on our model quality!
It's also a pleasure to use Moi because it gives such a direct access to the model.

If anything i would like to see in Moi, - maybe a dynamic sectioning tool for investigating shapes. I think of a kind of "clipping plane + intersection line at the same time" type of thing.
Perhaps also a 2d, flat on screen, curvature graph to se the surface continuity where sections go through two or more joined surfaces.
... just some thoughts

-Fredrik

Anybody tried IronCad or SolidEdge with this case?

Thanks,
d3

Yes, if anybody has the chance, would be interesting to see if the Parasolid kernel does it differntly.
Solidworks too uses Parasolid.

-Fredrik

It fillets up to 2.5 but it's also struggling, you can see this colour plot.

Yes, I'm finding this interesting too. It's an eye-opener that the 'Big Boys'. such as ACIS and Parasolid are both struggling a bit with this one and MoI seems to be taking it in its stride. I assume that as they can't radius this part much beyond 2.4/2.5 they must be hitting some kind of natural limit within the geometry.
If that assumption is correct, then does it mean that MoI's Blend tool is making some subtle changes to other parts of this solid that ACIS and Parasolid wouldn't consider to be acceptable?
I'm not trying to whip up an argument, I'm just wondering why there's a difference.

Martin.

Hi Martin,

> It's an eye-opener that the 'Big Boys'. such as ACIS
> and Parasolid are both struggling a bit with this one
> and MoI seems to be taking it in its stride.

To be fair, it's much more common for filleting in ACIS/Parasolid to handle cases that MoI can't, particularly in areas where fillet sections are coming together in a complex corner juncture.

But yes this case does show that the big MCAD kernels are not better in every single case 100% of the time...

Filleting is a really complex area of calculation with a lot of different steps that go into it. The Solids++ kernel has a self-intersection resolution mechanism in it which is what allows it to handle this particular case better.


> I assume that as they can't radius this part much beyond
> 2.4/2.5 they must be hitting some kind of natural limit
> within the geometry.

Yes, that's true - it's due to the tight bend. You can't physically place something that has a large width to it around a bend that is tighter than that width without it causing bunching or self intersection, here's an example:



In this case when the fillet gets over about 2.5 units in size it then runs into this situation where the fillet kind of starts to bunch up and kind of fold over itself in this manner.

That kind of bunching can cause problems in various cases in MoI as well, but in some cases like extrusions it is able to be resolved pretty well.


> If that assumption is correct, then does it mean that MoI's
> Blend tool is making some subtle changes to other parts of
> this solid that ACIS and Parasolid wouldn't consider to be
> acceptable?

MoI's fillet is not modifying anything about the original solid other than the trimming away of some area where the fillet is going.

But it does modify the generated rounded fillet surface - in the area where the regular fillet would not fit here:



It stops the fillet pieces with the exact circular cross-sections outside of that self-intersection zone and instead puts in a different kind of surface blend in that corner which adapts itself to the shape more rather than only having only circular cross sections in that spot.

Both ACIS and Parasolid will put in other similar kinds of non-circular cross section surface blends in other circumstances like for some kinds of corner patches at the ends of edges, they just don't have something that recognizes that the fillet is going to cross over itself at some inside region of the edge like this case.

- Michael

Thanks again for your detailed reply, Michael. It is very much appreciated as I don't as yet even own a licence for MoI!

Martin.

Hi Michael,


Yes, this is somewhat true but then the big MCAD kernels do offer more options to achieve the desired result, take NX as an example, you can choose from Edge Blend, Face Blend, Soft Blend and Styled Blend (Blend=Fillet), each command tackles fillets differently, then you have finer control also, but you do have to pay 100 times more than what MoI costs.

Cheers