when using smooth constraint why are curvature handles hidden?

[cekuhnen]

is there a reason why smooth constraint hides the curvature handles? it is for me important to adjust tangent and curvature values to sculpt smooth constraint / g2 transitions like in the right case but without it I need too add afit point to bring the curve down in the mid image!

[schneik-adsk]

If you don't edit or activate the curvature handle before adding the smooth constraint you will run into this block. If you activate or edit the handle before applying the constrain you can edit the handle.

[cekuhnen]

I am lost here:

"activate the curvature handle before >> adding ???? << the curvature handle"

[schneik-adsk]

Sorry... before adding the smooth constraint. I edited my reply. Next time I'll not post after my second Maß.

[cekuhnen]

With a spline alone I can edit tangent + curvature handle
When adding a tangent constraint I can edit both handles.
When adding a smooth constraint all I can edit is the tangent handle?

Why is that so? Is there a technical reason for it? Will we later get access to it? G2 without a G2 scale influence seems problematic to me.

True when not doing any handle activation the G2 transition in Fusion is equal Alias with one major difference

I can only adjust the tangent value through the handle and not the curvature handle/scale/profile and that handle change often makes drastic and illogical changes to the curve because well the G2 scale factor (alias term) I cannot address. Mini changes in the tangent handle often have drastic results already and I am sure I could smooth things better if I could massage the curvature handle as well.

[dunderhead]

Won't smoothing in most cases render G3ness of academic interest only? At some point in the manufacturing or rendering pipeline triangulation happens -- and smoothing and refinement will overide the importance of G3?

I'm asking because the curve in your example can be very accurately reproduced in Bsplines using only 3 (three!) control vertices (and symmetry and points for the linear pieces).

These three points were a piece of cake to place.  Working with B-splines is so much easier and fun for curve fitting.  As you can see, I got a continous, but non-differentiable curvature diagram that's pretty darn close to the one you made in artwork in Alias (but not G3)! See below.

What was not a piece of cake is the multiple hoops you'll have to go through to do sketching with B-splines as currently incarnated in the sculpt enviroment!! (I found yet another issue with the basic 'face' command: it won't allow me to draw an n-gone for n>4 involving the axis of symmetry, no big deal, though.)

For me the real issue in sketching/sculpting is how ineffective it is to put even 2D drawings together in sculpt. I really would like to draw vertices and edges directly.  Would that address some of your concerns as well?

[cekuhnen]

G2 and G3 are just tools to position in Alias CVs so you can watch your curvature comb.

G1 is a terrible surface finish which you see mainly with a lot of mass produces products.

G2 is good G3 however is better because with it you can better make sure that you acceleration of your curves is perfect.

Thats not so easy with only G2

However you can fake things by aligning your CVs so that the result will look like G3.

That for example you can do in Inventor to get a faux G3: 

http://cadsetterout.com/inventor-tutorials/autodesk-inventor-surfacing-curvature-continuity-sketch-c...

The problem however is that it is fake, the moment curves change the fake G3 will break instantly.

G3 is mainly used in transportation but also luxury product design because with G3 you can perfectly make sure your surfaces

have even smooth bend profiles and your high-lights will be even in size and running/flowing over an object.

In TS nor in any poly modeler can you create an NGON over a mirror axis. That simply does not work. You can only mirror edges along the axis.

However TS is pretty slow with NGONs also when the poly model is pretty easy:

Capping the right opening is at the speed of a button, creasing it yellow edges in the left image entering a number.

Fusion works quite a lot on this. So I switched to a different workflow which has certain advantages.

In Blender I can weight crease edges Maya and such do the same but those are dump poly modelers and not offering the precission TS has.

Sadly the precision requires time to compute. Here you see a Blender model where edges have different crease values:

But and here comes the glory of Fusion, because it can turn TS into BREPs we can do the same by just creating only the sides, then patching the caps with a surface tool

stitching everything together and then we can perfectly precission round the edges wit the fillet tool. No poly modeler can do that, besides Mesh Fusion add-on in Modo which

produces terrible dense poly meshes...

So the downside of TS is that it can be slow, but because we have the timeline and perfect TS to NURBS tools it is really not a downside actually.

Here all this poly modeled and then in Fusion turned into BREPs and trimmed! P like Perfecto!

I actually agree on that it is really a downside that in TS you cannot extrude just edges. But I was told the nature of TS does not allow it.

I wish there would be a poly mode and a TS mode in Fusion. Thats why I continue concept sculpting outside of TS because it is faster.

But that is ok - the data can easily be transfered. Recently for school I did a web cam arm fixture in Fusion brought it to Blender created

more organic joints, brought everything back into Fusion - my parts all were correctly positioned and dang used the timeline to create the

data for the Makerbot. 

I researched this process quite well. Blender/Modo offer tools Fusion can never have and Fusion as much Blender Modo will never be able to have.

And while all in one application would be fantastic this process because data can be send two ways (missing part is OBJ export from Fusion TS) works spot on.

I never had any scale or position issue.

BTW the curvature comb in your example would be a shock to every Alias designer. You have pinched points and inflected ends. Specifically the last are a no no.

What might look good to our eye is via math and combs often not the same!

TS is also as far as I know only G1 and a STEP export is heavily subdivided with isoprams. But from what I have understood sofar in Fusion internally the data is pretty slim.

What you later send to the printer mill cnc etc will be poly data anyway. So question is how far can Fusion become the designer / engineer work together on this.

Even Apple designs in Alias and makes the manufactring data in Unigraphix. 

[dunderhead]

Thanks so much for your kind reply and all the illustrations, especially the Blender ones!!! I see why my curvature crossing zero is a fumble, it'll create a dimple than can be seen in reflections!

I tried my G3 skills using the new B-spline tool in the unreleased new version of Inkscape. Needless to say, that didn't go so well either when I examined the result in F360 with curvature combs!!!

However, I still fail to see why G3 per se is important. Sabin (of Doo-Sabin subdivision) with others write "For graphical purposes, we suggest, curvature behaviour is almost irrelevant" in "Curvature behaviours at extraordinary points of subdivision surfaces" which has some really funky examples of oscillating or diverging curvatures. I admit to not understanding this very well. In practice I find the tangent handles quite bothersome, both as visual clutter and as tools to work with, I'd vote for CV splines as you suggested, I think.

The problem I had constructing a polygon is a minor bug: my initial n-gon didn't cross the line of symmetry but I used part of it as an edge.  But for n=4 it worked.

[cekuhnen]

With G3 you simply have more points on a curve / surface to smooth out the design better.
You can eye ball it with G2 but it will be much harder to make it perfect!. Also when the surface we made a fake G3 transition to changes you have to manually adjust it so it looks like G3 again. The focus here is just on the combs.

G2 is pretty good for most everything thats true. G3 is just a tick smoother and better.


One important aspect is also in Alias Apple for example makes the concept surface. But it will be rebuild in UX. So essentially you want to work with a workflow that later the engineer can use to rebuild your shapes in their application. ICEM in transportation design loves single patches which is why using curves with edit points in Alias is a no go.


So I guess it really comes down to what you really only need.

[dunderhead]

Thx for the references to icem etc ! Interesting, I get it (finally) --- it's as a user parameter that G3 is important in your field. It's going to be a challenge to integrate all of this into a not too intimidating UI however.

I've struggled a bit just to learn how to use the simpler stuff already there, but there's nothing proper documentation wouldn't fix in this regard (the current spline sketch manual entry http://fusion360.autodesk.com/resources/akn/view/NINVFUS/ENU/?query=spline is basically content free)!

[cekuhnen]

G2 or G3 in Alias is pretty easy.

Again G1 to G3 only defines how many CVs you have on a curve to make sure you get a certain curvature comb. Thats all. The only sculpting issue is that G1 needs less space and G3 needs more space to make nice and smooth acceleration combs. Thats honestly it.

I think they make more out of it than it actually is. So from a UI point this should not be a problem.

I am not sure why also Inventor does not offer G3 for splines and such I also cannot find any information about this regarding Solidworks. I know ICEM Catia has at but then again they are often used for Trans design.

[innovatenate]

The curvature handle controls the "radius of curvature" of the spline entity at that specific fit point.  The radius of curvature may be thought of as the radius of the best-fit circle at the fit point. The Curvature Handle dictates the Radius in the below images. This corresponds directly with the height of a curvature comb.

From Alias Help:

From Fusion 360:

When you specify a Smooth Constraint (G2 Continuity), you are also specifying the height of the curvature comb.

The presence of a smooth constraint AND the curvature handle dimension create over-constrained, redundant sketch controls.   

Also keep in mind that there are no reference dimensions, only active, driving dimensions in Fusion 360. For this reason, I believe the curvature handle is disabled when a smooth constraint is added. This will prevent a user from getting the "over-constrained" error message.

I hope that helps.

Thanks,

[jiang_peng]

Hi Claas,

If you add a smooth constraint between a spline and a curve, that means the curvature value of the spline at the connecting point is same with the curvature of the connected curve.

The curvature handle means that the curvature value at that fit point of the spline is the radius of the curvature handle arc. So in this case, if we still allow curvature handle, that will be confict: we will not able to determine which curvature value we shall follow. It's the reason why the curvature handle is hidden when a smooth constraint is built on the end point of the spline

thanks

[dunderhead]

Nathan, your 360 sketch is pretty, what are the yellow stipled lines and the markings 1/7 and 1/70?

May ask two further little questions?

1) If I activate both tangent and curvature handles, then I'm exactly getting G3 (curvature acceleration control) by adjusting the tangent handle length, right? And this control is inverse: the longer I make the handle, the less acceleration? Also, in this case, the algorithmists at Autodesk have choosen to loosen the 0th order constraint on the the node so it slides a bit while adjusting the third derivative? Here the problem is that because already by introducing one tangent handle, you're overconstraining (there are already as far as I can see six nodes involved in computation of a quintic or such spline) and by introducing a curvature handle, the whole thing is even more overconstrained, so your algorithm relaxes the node position, while converting the tangent handle length to directly control G3 while still controlling the first derivative (tangent)? If so, this is quite clever I think.

2) I've haven't figured out how to extend my own (3D) splines, except for accumulating extra points near the end point, on the spline - and then sliding the end point - and then spacing the extra points out. Is that the way to do it?? I'm still in vain right-clicking the end point just because I'd expect a menu item that says "extend spline" and I feel a little dumb about that!

[dunderhead]

As I played around a little more, I encountered a situation where the position of the node does not change when adjust the tangent or curvature handles! But then I managed to have the curvature arc become a circle that defected!!!! It's center hops away from the node, sliding along the tangent handle. I'm not sure whether these situations are hidden gems or ugle bugs.

[cekuhnen]

Here is another video, I hope this explains my point.

Please take a look at it, as it from my point of view seems to show a short comming or problems of the reliance of only spline and constraints as drawing tools.

1. To stabilize sketches I have to constraint everything, this adds extra work which in the concept phase I might not want to invest. (prototyping is of course different but that phase follow afterwards only)

2. If the smooth constraint uses curvature handle value of the primary curve then I have two problems: 1. I have no controll over G1 and G2 of my blend / secondary curve. And 2. to get the blend curve right I have to adjust my primary curve and thus change the primary curve.

So for me while the spline tool in many areas is pretty fantastic specificaly for curve network designs and loft, it seems also to be a limiting tool when you need certain control. So this might again be a case for us to get CV curves, if the spline curve cannot allow you to have different tangent and curvature handle values for primary and secondary curves!

https://drive.google.com/file/d/0Byzv_NlyKp_2dmc1b1dCaWZJZEk/view?usp=sharing

[cekuhnen]

I think there are some bugs left particular with the curvature handle.

[dunderhead]

Nice video. In my understanding, it's already possible to fiddle with the curvature change even with an imposed smooth constraint: the tangency handles appear with their length determining the curve acceleration -- the meaning of the length of the tangency handle being exactly the same as in the case with both active tangent and curvature handles: to determine the acceleration (the 0th, 1th and 2nd derivative being fixed by the smooth constraint on shared nodes). In your case, it seeems, your spline has only two points and F360 won't accept G3 (is that a bug?) changes. Add a middle point to the spline and it'll work (for me): I can work how fast the curvature comb grows out of the end points of the spline. But the G1 part of 'smooth' (G0, G1, G2) breaks easily and then only the curvature will sync, not the tangents... Other funny things happen, too.

For my simple purposes, I'd like to fiddle with splines without the visual clutter of the tangent handles (I'd like an option to not show the non-activated handles) and to easily extend them. That'll help in situations where you're simply tracing something existing.

[cekuhnen]

The handles are quite nice less for tangent but specifically for the curvature as it visually is a nice guide.

But I find often a UI clutter when you have handles from to splines next to each other or overlapping tangent / curvature handles.

Selection can be slow and tricky then. But I am not sure how you could visualize this differently.

The handles work different then G1 G2 CV movements in Alias.

[innovatenate]

The yellow dashed lines are my "fake" dimensions for  the curvature comb height. I didn't want to ruin my pretty image in MS Paint. 🙂

Item 1 - The short answer for G3 is "sort-of." You are influencing acceleration with the length of the tangent handle, but you can't really specify the value. It would be an "eye-balled" G3 at best.

Let's back up a bit and just look at what the tangent handle is for reference. Each control point on the spline has a tangent handle. When the spline is selected, the handlebar is visible as a line segment whose endpoints may be manipulated. Moving the end points of the handlebar changes two attributes of the tangent handle, the direction and the magnitude.

From Calculus we know the derivative of the curve's function at a point is the slope of the line tangent to the curve at the point. The direction of tangent handle is a tool that provides real time manipulation of the slope. 

The other attribute of the tangent handle, the magnitude or length, controls the an attribute called derivative length.  The resulting impact on the curve might be called the influece, weight, or acceleration. So why can't you specify a value and get precise G3 continuity? Let's dive a bit deeper.

Fusion uses classic NURBS (Non-Uniform Rational B-Splines) as its internal representation of sketch splines. NURBS are comprised of control points, (also called control vertices or CVs), and a knot vector. In a Fusion sketch, the control points and knot vector of a NURBS curve are constructed to satisfy the fit points, handlebars, and curvature bars provided by the user. The control points themselves are not exposed directly because they are more difficult to work with in a constraint-based sketch environment. For example, dimensioning an interior control point of a NURBS curves does little to control the physical dimensions of the resulting product. The number of dimensions/constraints to define all degrees of freedom for an interpolation/fit point curve are greatly reduced when compared to a CV curve in a constraint-based environment.

The mathematics of converting derivatives and fit points to an appropriate NURBS curve are well established, and Fusion uses some of these established techniques as part of its spline computation engine, but Fusion does not expose derivatives, directly. Derivatives are tangent vectors to the curve, and are therefore useful in establishing curve direction, however the magnitude of the derivatives has a less "tangible" effect.

buh-dump pshhh 🙂

Consider that if an appropriate derivative is chosen for one configuration of the curve, it is unlikely that the same derivative will be appropriate as the curve is modified, either directly by the user (e.g. inserting, removing or moving other controls) or by a constraint solve. To solve this problem, Fusion normalizes the derivative magnitudes. The magnitude of a derivative is a local change of curve length per local change in parameter.

The spline engine can use the neighboring fit points and its knowledge of the parametrization of the curve to get an estimate of the change of arc length per change of parameter. That estimate becomes a scale factor that converts a handlebar length to a derivative length. As the spline is manipulated, the handlebar length stays the same, but the scale factor used to compute derivatives from the handlebar changes in a way that doesn't unduly affect the character of the curve.

This makes precise values for acceleration more difficult to specify. This is why I say "sort-of." If you're okay with eyeballing it, you can get really close.

Is anyone sleepy yet? 🙂 

Item 2 - As soon as a spline is 3D some of the functionality changes. I'm not sure that it is possible to use the Extend command on a 3D spline at this time. I think this might be an item to add to the IdeaStation.

@cekuhnen I agree with dunderhead about adding the extra fit point for the reasons outlined above. The fit point spline should adjust the scale factor based upon the existing dimensions and constraints and allow you to maintain the character of the curve.

I hope this helps better outline how the spline tool is working.

Thanks,