Hi Michael,

Thanks,

I adapted a technique that I'd usually use with a network surface in Rhino.
The Scaling Rail is very flexible because with history on, by editing control
points on the scaling rail, you can change the side profile of the ring after the circle is extruded, rather than having to undo and edit the curve without the cylinder there as a visual reference. And by scaling the control points of the top oval, the top face of the ring will change dynamically as well. I was able to make the top round by scaling the cp's 1D

I'll try to write a tutorial to go with it soon, but if anyone has any questions,
please feel free to ask. The trick to getting the base of the ring to be
smooth is the addition of little circle at the bottom.
Without it, the base gets angular where the curves come to a singularity.

Also, I've noticed that if you try to boolean the surfaces that are out
in space, it can get all messed up, so bring the curves to the origin if you'd like to work with them.

Jesse

Hi Jesse, I was experimenting a bit with network surface in Rhino to see the differences.

One thing I noticed was that the network surface will often have isoparms that flex and bend around, especially as they approach the singularity. On the other hand in this case the sweep generates completely planar isoparms throughout, in some ways this is a bit more predictable in how the surface is constructed.

The other thing is that you don't have to make the scaling rail actually touch all the profiles, profiles will stretch up to match the scaling rail even if you pull it away from the original profiles. So in a sense the rail has more "weight" assigned to it. Network is not like this, everything is more equally weighted there and pulling the "rail" out doesn't cause the surface to hug right against it.

The other thing that is interesting is the implicit symmetry from the center line that you get with the scaling rail - with network surface you have to go through some extra effort to get symmetry edits.

It seems like there may be some reasons to prefer the sweep + scaling rail for some symmetrical shapes rather than network.

At any rate, it certainly looks like the scaling rail boosts up the potential of sweep by quite a bit.

- Michael

>>At any rate, it certainly looks like the scaling rail boosts up the potential of sweep by quite a bit.<<

I completely agree, I was amazed that it did so much with a relatively sparse amount of curve geometry.

With a network surface, if you only used one side rail it would get all whacked out...the reason I mirrored
the scaling rail was to be able to make the curve's end at the base tangent to the other side, so that
the implicit symmetry of the sweep created a smooth juncture at the point of singularity

I know I'll be experimenting to find other ways to take advantage of this tool's versatility... thanks for being a NURB's genius! :-)

-Jesse

Here is an idea I was playing with - plain 2 rail sweep on the left, then addition of scaling rail on the right.



One interesting property here is that each isoparm of the resulting surface is an exact ellipse. It would normally be quite difficult to arrange a sequence of analytic curves like an ellipse to get a smooth result like this.

Just 4 defining curves...

- Michael

Another experiment - this time a wavy pattern is introduced by editing the scaling curve. Still defined by just 4 curves:



- Michael

Hi Michael,

I was working with a similar idea yesterday...:-)
You can get some interesting variations depending on the shape of the scale rail
and where it's placed.

I noticed something that I thought you should know about...
it could be a bug or just the way sweeps work with the scaling rail,
but it could be an issue if you wanted to put more than one profile on a ring.
which is often the case when designing for actual production.
It works great if the profile curve isn't touching the drive rail.
If it touches the quad point, it sweeps only half the circle with the scaling rail.

I don't have the last beta to make a comparison, but it seems like
with the addition of the scaling rail option, sweeps take longer,
even if you're doing a regular sweep without the scaling rail.
Am I only imagining this?

Jesse

Hi Jesse, that looks like another interesting result from a very low number of defining curves.

> If it touches the quad point, it sweeps only half the circle with the scaling rail.

Can you please e-mail the bad one to me? I'll see if I can fix this up.

What is happening is that when you have the profile off to the side, MoI uses the "auto place" mode, where it places the profile for you sort of centered around the rail. MoI does this automatic placement thing any time you have a a planar profile curve that is located outside the bounding box of the rail.

As soon as you move it to the quad point, it is no longer completely outside the bounding box of the rail, so MoI does not mess with the location of the profile in this case, it uses it from where it currently is located.

Then I think the issue is the scaling rail not working right when the profile has a straight vertical line coincident with the line between the spine rail and the scaling rail.

The auto-place works because the line connecting the spine rail and scaling rail ends up intersecting the profile through the middle of the profile in that case instead of grazing along the flat side. If that makes sense...


Re: sweeps slower

They probably are a bit slower now because they now use the cancelation mechanism. There is some overhead to using the thing that allows for canceling stuff in mid-calculation. But the benefit is that you don't have the danger of possibly getting stuck unable to do anything else if you are in a long calculation.

- Michael

Hi Michael,

I'll email you the file...here's something interesting, by tipping the scale rail at
an angle, it will taper the ring, while conforming to the shape of the other curves.
This design isn't the best application of the technique,
but I can definitely think of some that are.. :-)

-Jesse

Scaling rail ring.

-jdk-

Hi Jesse - wow, very nice result!

So let's see - is that a one-rail sweep + scaling rail?

How would you have done this without the scaling rail? I guess maybe NetworkSrf with 4 curves in each direction, or it looks like 2 rail sweep with 2 wavy rails.

I suppose if you have a one-rail sweep where the central spine rail is on the mirroring plane of the object, then a one-rail sweep + scaling rail is actually equivalent to a 2-rail sweep between mirrored rails.

- Michael

> Does a "generic transformer tool" like "bend, fold" can be imagined in Moi?

Hi Pilou, unfortunately these are quite difficult to do on solids.

It is difficult to bend the solid in an accurate way so that the different surfaces that make up the solid are still connected together properly at the end, with no gaps between what are supposed to be shared edges.

So this won't be possible to do in MoI for quite a while. Rhino V4 has some cool new tools to do this stuff, and it is easy to share data between MoI and Rhino so for a while you will need to use Rhino along side of MoI to do these kinds of operations.

- Michael

<won't be possible ...
So imagination must replace it (for the purists) :)
In general there are always many ways for make something :)

> In general there are always many ways for make something :)

For now what you can do is draw it in the "bent" shape from the beginning instead of thinking about bending it later on.

- Michael

It is a 1 rail sweep with a scaling rail.

>>How would you have done this without the scaling rail? I guess maybe NetworkSrf with 4 curves in each direction, or it looks like 2 rail sweep with 2 wavy rails.
<<

The last method is how I'd do it in Rhino.
It would be a 2 rail sweep with the same cross section profile as I used in MoI.
a circle for the finger hole and a wavy curve offset outside
of it -which would be more complicated to draw than the scaling rail curve.

The result of this kind of sweep in Rhino is somewhat unpredictable.
The waviness is an effect of the sweep, rather than a predetermined shape that
you can easily control,

-Jesse

Hey Jesse this looks great, I'm learning a lot of tips from this thread, thanks.

Hi,
I am a jeweller and am at last getting the hang of the workflow to make the sort of ring like the one of Jesse's above, building up one section on the circumference with booleans and then arraying it around the ring . I also got carried away with sets of boolean difference holes. When I come to add and subtract these decorations , if I do them all at once it never seems to finish calculating. If I do them one by one it still takes a long time for each.
Is ths what I should expect? It seems to be a fundamental ingredient of ring design, to gradually build up more and more complicated units and then build them together by using mirror or array, so do I expect it to take a long time?
I have not got the newest system but it has over a Gig of memory and a reasonable graphics card, so am I just over impatient!!

regards Tim.

Hi Tim, yes unfortunately the booleans do quite a lot of calculation so you can expect that they will tend to be slow.

> but it has over a Gig of memory and a reasonable graphics card

The boolean calculation speed is entirely dependent on your CPU speed, unless you run out of memory but I think you'd be fine at more than 1GB. What CPU speed do you have for your computer?

One thing you can do to speed things up if the speed is really getting in the way of doing your work, is to work on surfaces instead of solids, and then use Trim to cut only the surfaces that you know are slicing into each other.

Booleans basically do a lot of automatic Trim and Join operations for you, but they will tend to do a lot of checking for any intersections between 2 solids. You can kind of get more of a "fine-tuned" control over the process if you do the trimming and joining between surfaces yourself. But it is a lot less convenient, it is probably better to just wait a few minutes until the booleans finish.

- Michael

Hi Everyone,

I've been playing at making a fitted wedding band, but not getting the results that I want.
No doubt it's more to do with my lack of knowledge of how all of this works?

I've been trying to do it as a sweep, as that seemed to me to be the logical choice??
But I kept getting splits in the sweep where the band turns.
Here's 3 different versions that are progressions of each other.
Could someone please advise me of the best way to go about this kind of construction.

regards Colin

Hi Tim,

Just ask on the forum if you have questions about jewelry modeling in MoI, I may not have the answer, but someone will, so it will help others who want to learn, too...and then if enough jewelry designers ask questions, Micheal will know what tools to make for us! ;-)

I'm not sure if this is correct, but it seems that Boolean-ing parts in small groups seems to go faster (towards the end result) than doing them all at once.

Since many jewelry models are symmetrical, you can Boolean subtract things like azures for stones on one side of a ring and then trim away the other half of the model, slicing it right down the middle with a cutting plane, then mirror over the finished side and join them together. I've also build just a quadrant of a model and then mirrored or circular arrayed to complete it.

Another thing to keep in mind,- if you're CNC milling your models, Boolean unioning may not be required at all. I use ArtCAM which handles STL's that touch each other as one entity, so when you import the file to make a tool path,,it doesn't matter if the model was Booleaned unioned or just an assembly of parts before it was meshed.

Hope this helps.

-Jesse