Function Request - Draw line from equation inside Sketcher

[abdullah]

I think 10 points might be enough, but talking with other loudspeaker builder I can check if we might need more or less points.

Yes, I used Bspline to connect all points, besides the curve do not exactly pass in the points and think for a real purpose is accurate enough.

I liked your model to check Ajinkya's latest PR (point on b-spline). Actually it seems to be that with this new ability, you will be able to do much of what you want with about 3 or 4 well selected sections:

Screenshot_20230121_195122.png (56.76 KiB) Viewed 688 times

Here there is a single B-Spline of 4 poles. The first three vertical lines and the last one have the endpoint, point on object (point on B-Spline). As you can see this (much simpler) B-Spline passes through all the other endpoints of the other vertical lines, even if there is no constraint at all on them. This could simplify the sketch quite a lot:

Screenshot_20230121_200151.png (22.7 KiB) Viewed 688 times

Of course, this selection of which segments to constraint, as well as the number of poles, is for me "random", as I do not know much about the real problem you are trying to solve. But it may suggest 4 uniformly distributed vertical lines may be enough for approximating the exponential...

I think we will have this new ability in master pretty soon.

[onekk]

...

It seems that posing the correct questions, some results have been achieved, or in the near future will be achieved.

There are only "correct questions" to be posed and maybe "correct answers" will arrive. This thing is happening quite often last times.

As usual with "correct people" in the discussion the level will be "decently high".

Thanks to @chrisb and @abdullah I think the discussion is near to be solved.

Regards to all.

Carlo D.

[lordsansui]

care to describe it in the wiki of Sketcher RestoreInternalAlignmentGeometry?

I can try with some guidance, I don't have access to it, when I open that page it didn't recognize my user used in the Forum. Take also into account that English isn't my main language.

[lordsansui]

Of course, this selection of which segments to constraint, as well as the number of poles, is for me "random", as I do not know much about the real problem you are trying to solve. But it may suggest 4 uniformly distributed vertical lines may be enough for approximating the exponential...

I think we will have this new ability in master pretty soon.

Nice, thank you for your feedback about future changes and new stuffs.

Also providing more feedback here for who is following the discussion, while trying to implement in a real loudspeaker model with Bspline it didn't work well, I can't use orthogonal constrain between a line and and Bspline, so It was not possible to proper emulate the offset profile nether to guarantee correct thickness for the wood, at the end It was needed to implement as a discrete exponential profile with 10 segments connected with straight lines. It's a good approximation and work well.

005.png (61.67 KiB) Viewed 623 times

006.png (12.55 KiB) Viewed 623 times

If I change the area size of the Mouth (bigger) and the Throat (smaller) the model adjust itself automatically, see below

007.png (17.84 KiB) Viewed 623 times

[chrisb]

How about using tangentially connected arcs?

[abdullah]

Nice, thank you for your feedback about future changes and new stuffs.

Just merged Point-on-BSpline support. So that is in master now.

Bspline it didn't work well, I can't use orthogonal constrain between a line and and Bspline

Yes, this does not exist yet, but I think Ajinkya (@jnxd) wanted to pursue this path and he is fast...

so It was not possible to proper emulate the offset profile nether to guarantee correct thickness for the wood

There is some work to create static offset curves in the pipeline. Hopefully over the year.

[lordsansui]

How about using tangentially connected arcs?

I would still need a good way to determine the proper thickness to get a good point, not being able to constrain versus bspline didn't gave me alternative.

Using discrete segments is good enough for a while, if needed we can increase of decrease the amount of segments, let see we the updated tool we get in the future so we improve the models.

[lordsansui]

Nice, thank you for your feedback about future changes and new stuffs.

Just merged Point-on-BSpline support. So that is in master now.

Bspline it didn't work well, I can't use orthogonal constrain between a line and and Bspline

Yes, this does not exist yet, but I think Ajinkya (@jnxd) wanted to pursue this path and he is fast...

so It was not possible to proper emulate the offset profile nether to guarantee correct thickness for the wood

There is some work to create static offset curves in the pipeline. Hopefully over the year.

A lot of improvements going on, very exciting.

I'd like again to thanks to everyone that contributes in this discussion witch leads me to find a good solution with the tools currently available.

[jnxd]

Bspline it didn't work well, I can't use orthogonal constrain between a line and and Bspline

Yes, this does not exist yet, but I think Ajinkya (@jnxd) wanted to pursue this path and he is fast...

so It was not possible to proper emulate the offset profile nether to guarantee correct thickness for the wood

There is some work to create static offset curves in the pipeline. Hopefully over the year.

"Orthogonal" can be temporarily achieved by creating a "tangent-at-knot" constraint with a line and then making a perpendicular constraint with that line.

I'm slightly skeptical of offsets, because in general they are more complex. For example offsets to ellipses aren't ellipses and offsets to B-splines cannot be described by NURBS in general. So we might be taxing the gcs infrastructure a whole lot by saying "make an offset to this curve and make that offset pass through this point".

As for b-splines to represent exponential functions, note that bsplines are polynomials, so they can't do that exactly. However, we only need a few degrees to approximate them well, since taylor series of exponentials converges quite fast. Depending on the use case (how much smoothness you want for your speaker) other alternatives like the line segments you're using could either be sufficient or lacking. I suspect you're planning to 3D print it anyway?

[lordsansui]

The major ones are made by plywood as you can see below.

P_20170305_101222.jpg (115.62 KiB) Viewed 469 times

The examples I posted in this thread are: Altec816 and Altec817 as you can also see below, they have exponential horn. There are others project but I'm not planing to cover all.

altec816.jpg (47.47 KiB) Viewed 469 times

Altec 817 - 01-me.jpg (72.87 KiB) Viewed 469 times

Discrete horn / line segment probably give enough information for the user to bend the wood. People use different Technics to bend wood like indicated below.

Kerflor-Timberflex.jpg (191.73 KiB) Viewed 469 times

One thing that people do a lot is to machine some panels using CNC, I personally didn't had the needs to do it, but people who likes to use FreeCAD could learn more about his integration with CNC machines. There are so many potential here.

post-15193-13819738616936.jpg (603.58 KiB) Viewed 469 times