[Feature Request] Radial array/polar pattern in Sketcher

[Crow]

Let's assume it was 6 weeks before christmas and I want to model this:

Ha! joke is on you! Its six weeks and three days, and you did model it!

Just kidding, Thank you for the thoughtful reply. And thank you for making the example. I've taken a look.

Helix_Advent.PNG (60.15 KiB) Viewed 1680 times

I was able to helix sweep a surface of your model along the correct axis. And yes, this is a way to accomplish this kind of geometry. In this picture I'm using a subtractive primitive to remove your 2mm extrusion. I tried and failed to use the "pocket" function.

Helix_Advent_2.PNG (60.43 KiB) Viewed 1680 times

Taking a look at the resultant feature tree has me worried though. I'm getting "topological Naming Problem -anxiety" thinking about how this is just a simple demo piece.

So if it's OK with you guys, I'm going to continue to be of the opinion that a parametric polar array feature would be a good addition to the Sketcher workbench. Again, not making any demands, and with no expectations.

[drmacro]

...I'm going to continue to be of the opinion that a parametric polar array feature would be a good addition to the Sketcher workbench. ...

A sketch array tool will have the same TNP possibilities.

[adrianinsaval]

A sketch array tool will have the same TNP possibilities.

Not really, in there he used the face from the pad for the helix, that's not very robust (and I wouldn't want a helix that takes many seconds to recompute to break and reference the wrong face)
If the array was inside sketcher he would have used the sketch for the helix directly, so no TNP. And you could say that he should use the helix on the individual sketches and make a pattern from the helix... but that gives shitty results:

Three simple sketches, five operations in the tree.

I don't think that was the intended shape but yeah, it can be made simple like that. It's computationally intensive to make the helix and patterns though (I assume making a single helix feature would be less intensive), and it has real trouble with tangent stuff. I'm not very satisfied with this result:
example.pngexample.FCStd
I had to put the teeth profile something like a milimiter (0.1 wasn't enough) inside the octagon so it doesn't give broken results with the pattern

(the parallel discussion is driving me nuts )

[drmacro]

...
Not really, in there he used the face from the pad for the helix, that's not very robust (and I wouldn't want a helix that takes many seconds to recompute to break and reference the wrong face)
If the array was inside sketcher he would have used the sketch for the helix directly, so no TNP. And you could say that he should use the helix on the individual sketches and make a pattern from the helix... but that gives shitty results:

Any parametric change to the array in the sketch that causes an edge to be added or deleted will cause TNP.

[adrianinsaval]

Any parametric change to the array in the sketch that causes an edge to be added or deleted will cause TNP.

In features that reference geometry from the helix or specific edges of the sketch yes, but the helix feature itself only references the sketch as a whole, whatever happens inside the sketch won't break the helix or at least not from TNP

Edit: a file speaks 1000 words so I made two similar bodies using both methods to get a cross shape used in a helix then disabled recompute and added circles to the sketches, open the file and recompute and see which one breaks from TNP
Spoiler alert: the one with the simpler sketch and PD pattern + pad breaks, the one with the hand made pattern inside the sketch works no problem

[freedman]

Maybe this is more powerful;
COPY TO VERTEX
Select a target sketch object, for instance a circle, then draw a selection box around any sketch elements or group of vertex, then click the copy icon. Any vertex inside the selection box gets the target sketch object applied to it, I suppose there would also be constraints applied.

This way we could do selective work. Also using polygon to the nth (pic) would give us a constrained circle by applying the method to the hole polygon and you then have an array tool.

The selection box vertex could contain any shape, we could also place an unorganized group of vertexes and do the operation.

The first selection sketch object could be limited to a single vertex and line combination, like a circle or a line.

This might be easier for the Devs to program.

[Crow]

Maybe this is more powerful;
COPY TO VERTEX
Select a target sketch object, for instance a circle, then draw a selection box around any sketch elements or group of vertex, then click the copy icon. Any vertex inside the selection box gets the target sketch object applied to it, I suppose there would also be constraints applied.
(...)

I'm not really qualified to comment on how viable this is, but I like the thought process.

I'm concerned that the proposed method does not maintain a parametric array feature though (as is the case shown in the example from NX). If you use copy to duplicate the desired elements around the array, then they will not (I assume) be updated when you modify the "Master" lines/curves.

The same goes for the n-sided polygon. Would it be possible to edit the number of n ?

Also, there is a need to define an anchor point (in the master geometry) to constrain to the polygon vertexes. And in the case of a circular pattern the rotation of the slave-clones has to follow the circle (normal to sides of polygon?).

(the parallel discussion is driving me nuts )

Lets keep any further discussion in this thread, since it's the oldest.

[freedman]

I'm concerned that the proposed method does not maintain a parametric array .The same goes for the n-sided polygon. Would it be possible to edit the number of n ?

I'm not sure the goal is parametrics, thats what happens when you apply the polar array in solids.

The copied objects would have constraints so they would all update together. I like this because it gives us more options, we already have an array tool.

[Crow]

I'm not sure the goal is parametrics, thats what happens when you apply the polar array in solids.

I know, and it's a very good idea in terms of making the sketch creation more convenient, but I believe some of the people on this thread had reservations about the Sketch Array not being parametric (I suspect for the sake of making the final model more robust?).

The copied objects would have constraints so they would all update together. I like this because it gives us more options, we already have an array tool.

Oh ok, you're completely right. I see now that the copy function creates some (=) constraints that keep the "Slave Geometry" defined. I was confused because this for some reason does not happen when I use the Sketcher Symmetry (Mirror).

I put together a crude array using copies in the suggested way. I acknowledge this was a sub-optimal way of defining the geometry and there is a lot of construction lines to create an "anchor" in the middle of the triangles and a radial line to the center of the array. The latter is to rotate each triangle so the round corner faces out. I guess I could have referenced the circle or polygon edges but I assume the copy function won't be smart enough to copy an angular constraint to another reference for each copy.

Copy_based_array.PNG (97.02 KiB) Viewed 1392 times

Copy_based_array2.PNG (103.24 KiB) Viewed 1392 times

Copy_based_array3.PNG (123.33 KiB) Viewed 1392 times

Copy_based_array4.PNG (82.66 KiB) Viewed 1392 times

_
So I tried another setup where I have the anchor on the middle of one of the triangle lines, and then made this line tangent with the circle that comes with the polygon. That works and it's certainly much better than the previous version, but the tangent constraints have to be added individually as suspected. Still though, the slave triangles update nicely from the original once everything is set up.

Copy_based_array5.PNG (35.91 KiB) Viewed 1392 times

[adrianinsaval]

I believe some of the people on this thread had reservations about the Sketch Array not being parametric (I suspect for the sake of making the final model more robust?).

It's more about avoiding unnecessary complexity for the solver (solving time does not grow linearly)