Here's a hard one....

.... and even harder to describe without images.

3 datum planes, X_DIR, Y_DIR, Z_DIR. I make an extruded surface
sketched on the X_DIR plane both sides. The profile is like a tilde
~ , or a flattened S shape lying on its side. IOW it goes from convex
to concave. I will call this surface "S".

In the Z_DIR plane I do two sketched datum curves, similar but not
quite symmetrical. They were but now they're not, thank you designer.
They are projected onto the surface "S". From the sketching plane of
these 2 curves surface "S" looks like a rectangle.

So we have 2 similar-looking curves projected. On each one I make a
variable section sweep surface, normal to projection, picking the
Z_DIR plane (the one I sketched the curves on) as the direction
reference. The sketch for each surface looks like an arrow head with
the aspect ratio the wrong way around. So from the center of the
sketch there's one short line going up at 3 degrees to the vertical,
and another going down at 3 degrees to the vertical. IOW, one segment
like " / " when the center is at the bottom, and another segment " \
", when the center of the sketch is at the top. LIke a chevron on its
side. The apex of the chevron runs along projected curve. The lengths
of the two segments are the same.

If I offset this surface, shouldn't I expect the edge of the chevron
on the offset surface to lie on the same surface "S", as the original
variable section sweep? It's out by .003" in some places.

 

snip

Graham,

Man, that is really hard to follow. I read it carefully twice, and I'm
still not sure I totally get it.

I think we're dealing with the intersection of non planer surfaces,
and an irregularity when projected out. This may or may not make good
geometric sense. Could you email it to me?

 

... and even harder to describe without images.

3 datum planes, X_DIR, Y_DIR, Z_DIR. I make an extruded surface
sketched on the X_DIR plane both sides. The profile is like a tilde
~ , or a flattened S shape lying on its side. IOW it goes from convex
to concave. I will call this surface "S".

In the Z_DIR plane I do two sketched datum curves, similar but not
quite symmetrical. They were but now they're not, thank you designer.
They are projected onto the surface "S". From the sketching plane of
these 2 curves surface "S" looks like a rectangle.

So we have 2 similar-looking curves projected. On each one I make a
variable section sweep surface, normal to projection, picking the
Z_DIR plane (the one I sketched the curves on) as the direction
reference. The sketch for each surface looks like an arrow head with
the aspect ratio the wrong way around. So from the center of the
sketch there's one short line going up at 3 degrees to the vertical,
and another going down at 3 degrees to the vertical. IOW, one segment
like " / " when the center is at the bottom, and another segment " \
", when the center of the sketch is at the top. LIke a chevron on its
side. The apex of the chevron runs along projected curve. The lengths
of the two segments are the same.

If I offset this surface, shouldn't I expect the edge of the chevron
on the offset surface to lie on the same surface "S", as the original
variable section sweep? It's out by .003" in some places.


Not if you're projecting "normal". The only place it'll be right is where the surface projected upon is tangent to the direction plane. After that, all distances are some sine function. As the frequency gets smaller, the distance between offset surfaces, in the area of sharpest slope, get smaller and smaller. IOW, the distance is variable, depending on the slope at any particular point: greatest where slope is zero, least where slope is greatest. I think you can avoid this by making the swept section normal to the trajectory, not to the projection plane.

David Janes

 

But the projection plane represents the mold face: that's why I use
that feature. The Var Sec Swp surface is a datum surface used later on
to drive the tangency of a boundary blend, 2 of whose boundaries are
the projected curves.

Anyway I worked out pretty much what you did, except I thought of it
in terms of 2D sections and how they would offset. So instead of an
offset surface I copied and pasted/transformed them. Of course I had
to re-route about 20 features....