Curvature cont. corners, no ISDX

[gristle]

Hi. A follow on from my last modelling post (N sided double loop surface).
In the past I used curve through points to create curvature continuous construction curves, but you have very little control over the two vertices at each end. One would assume you could drag these points out (like curve end bulge in Rhino) to smooth the curve, but this is not possible (with a CC constraint). Instead you have to add a point in the middle of the curve and then tweak to get the desired curve.

So... for the last couple of years I have used boundary surfaces to create curve cont. construction. The attached file is a simple slab model; the same approach can be used with more complex forms. It all hinges around using stretch values on 2 sided boundary blends with a CC constraint to get the desired corner. Normally a value of 1.25 to 1.35 seems to result in a curve that is not too flat in the middle. Corners on this model are controlled with OUTER_CORNER and INNER_CORNER boundary blends.

How do other non ISDX users approach this?

2010-08-23_071557_cc_corners_no_isdx-wf3.prt.zip
Edited by: gristle

 

[solidworm]

great approach! i love to see clever surface modeling.
i took a look at the top side only. i think if the spline control points in "top_sweep" feature is constrained properly so that the endpoint's curvature remains constant along the path, resulting sweep will have the required roundness on the corner of the part and makes a good surface. Jeff Howard (in this forum) gave us an exact formula to control the curvature at the endpoint of a spline by adding dimensions to its control polygon, but it can't remember it now, so i added this 60 degree dimension to the sweep profile to almost keep the endpoint curvature constant:

curvature at mirror plane of the part:

60 degree constraint in sweep section:

result without 60 deg constraint:

result with 60 deg constraint:

and the ProE part (WF 5.0)

Edited by: solidworm

 

[mgnt8]

Here's the link for the formula for the radius at the end of the Bezier curve:


[url]http://www.mcadcentral.com/proe/forum/forum_posts.asp?TID=35 749&PN=0&TPN=5[/url]

 

[gristle]

Hi guys. I'll have a look at that, it would be good to not have to trim and add controlled boundary blend to those corners. Cheers!

 

[solidworm]

reference dimensions are not available in sweep in WF5.0 it's grayed out in menu. just wanted to use the relation r = 3 * a^2 / b / 2
it's too late here, i'm going to bed. i hope you'll find a workaround by the morning.

Edited by: solidworm

 

[Jacoolo]

still, great tip

I

 

[gristle]

Good example. As mentioned somewhere in Jeffs posts, smoothest curve is one with less points on it. That is why using BB for construction is handy.

I have not looked further at the relations for controlling the curvature around the corner in my example. Looks pretty useful though, just once I figure it all out.

 

[solidworm]

you made a good use of the pattern feature too. i wanted to dimension pattern a 2 point curve with "Srf Nrm Edge" tangency condition on both ends at the corner of the part but it failed.whats the problem with it? (last feature)
here's the part.

edit: it works with first patterning the points and then ref patterning the curve.

Edited by: solidworm

 

[gristle]

Without looking at the part (bed time here) my guess is you need to group the points and the curve, or as you said, make the curve a reference pattern.

I had a quick look on the PTC knowledge database, could not find any mention of reference dimensions not being available in VSS in WF5. Who knows...

 

[zpaolo]

Good example. As mentioned somewhere in Jeffs posts, smoothest curve is one with less points on it. That is why using BB for construction is handy.

I used BBs a lot for controlled corners, WF 5 also has curvature continuous rounds that really help in this sense. In sketcher one trick to shape CC rounds is to add a middle point in the spline and use it to shape the corner, the results are smooth but slightly different from using drag handles: curvature using BB has a "flatter" part at the very corner, while curvature using the middle point is more "peaked"

Paolo

PS: why you need to replace the corner patch? it already looked smooth from reflection and shaded curvature analysis

 

[gristle]

Yes, the BB corner is flat in the middle, that is why I adjust the tangent/curvature stretch value. You need to tweak a bit using a curvature plot. Only downside is that you have to go into the BB feature to change the stretch value.

I replaced the corner patch because I was messing around to see what I could do with the corner. The zebra stripes are a bit flat in the middle, so by adjusting the cross curves you can change the surface middle. On reflection, probably would not use a pattern though because it ties down the stretch value of all the BBs used in the cross curve construction.

 

[zpaolo]

Yes, the BB corner is flat in the middle, that is why I adjust the tangent/curvature stretch value. You need to tweak a bit using a curvature plot.

What I meant is that when you shape the corner using drag handles and BB the curvature plot is flatter in the middle, i like this.

Paolo

 

[Jacoolo]

such smooth corners as introduced here are widely discussed among different CAD apps

see links below

[url]http://aliasdesign.autodesk.com/learning/articles/details/Ad vanced_Sculpting_134931/[/url]

[url]http://www.solidsmack.com/solidworks-tips/the-iphone-3g-mode l-take-2-plus-surfacing-resources-and-coolness-from-mark-bia sotti/[/url]

[url]http://www.solidsmack.com/solidworks-tips/solidworks-surfaci ng-tutorial-3g-iphone-model/[/url]

 

[solidworm]

here's one of jeff's parts from my archive.

he has applied the relation for the side of VSS that neighbors the double curvature surface and solved it for b:

(k=curvature of a bezier curve (= ProE spline with 4 CVs) at it's end points)
file:///L:/IDM downloads/Zips/2009-01-13_221909_paraboloid-vss_blend.prt.zipk=3/2*a^2*b
he describes:
a = the length of the control polygon segment cv0 -> cv1
b = the distance to cv2 measured perpendicular to cv0 -> cv1
he has measured curvature at both ends of the surface edge and has assumed a curvature distribution in between (the graph) which is an approximation. so along the path 'k' is read from the graph and 'a' is known dimension. so no reference dims are involved.
this relation is used for one side only because the other side has zero curvature and the relation is satisfied automatically by the chosen CV layout.

now which is easier?
1-finding a straight forward relation for curves with higher number of CVs.
2-pushing PTC for G2 in VSS. (can it really be done? what is the definition of surface continuity?)


Edited by: solidworm

 

[mgnt8]

OK one question:how to create Known Dimensions? How did he get kd5 in there?


G2 curvesshare a common center of curvature at the join point. There are tangent contraints in VSS. I supposea curvature constraint would absolutely require aspline attached to the trajectory intersection, but would you need a minimum number of cv's?

 

[solidworm]

whenever a dimension is created between reference entities in a sketch,that is a known dimension.the demension between origin curve and chain 1 in VSS sketch is the known 'a' or kd5

>"G2 curvesshare a common center..."
what about G2 surfaces? i guess the required and sufficient condition is that two section curves along the common edge should be G2. (min and max curvature directions)


Edited by: solidworm

 

[mgnt8]

Sorryto keep beating a dead horse, but I'm just not seeing where the known dimensions come from.


Let's say I'm in sketcher for an extrusion. I dimension a line to a datum for example (Sketch > Dimension > Normal). The resulting dimension has the suffix sd#. Let's say now I dimension a fully contrained line.Sketch > Dimension > Referenceresults in a reference dimension with the rsd# suffix. The other options are Perimeter and Baseline and these both result in the sd# suffix.


What specific steps must I do to get a known dimension with the suffix kd#? Is there a menu pick for "Known"? Do I specify it after creating?


In Jeff's part, when I delete the relation in Var Sect Sweep 3, kd5 is frozen and this message comes up:


"Reference or known dimensions are not allowed for this operation. Please select again."

 

[solidworm]

add two datum points on a plane, insert a sketch on that plane and reference the points then add a dimension between the points.the dimension will be a kd, it is already a defined dimension.
it doesn't give that error, for me.



Edited by: solidworm

 

[design-engine]

greg! you should remember/know 'known' dimensions in sketcher from you surfacing class!
Edited by: design-engine

 

[mgnt8]

Cripes! I knew I shouldn't have skipped the first day. I'll call Steve to schedule a refresher.


I see now. kd's are dimensions of references not geometry. Sorry to hijack the post.
Edited by: mgnt8