Re: What does Tool Radius Compensation do ?

Jon Elson <jmelson@...> wrote:

> If it does not have the G41 or G42 code anywhere in the program, then it
> does NOT assume the CNC is compensating for the tool.

Andrew Werby wrote:

>
> [It doesn't, so is there a way DeskProto could be compensating in software,
> independent of the milling program? Or is tool offset compensation
> irrelevant to my problem?]

OK, how do you make the three toolpaths? Do you rotate the part in the
CAD package, and then generate a new toolpath based on the rotated part?
That must be how you are doing it.

[Actually, I'm rotating the part in DeskProto, (not Rhino) then, as you
say, generating the new toolpath as if each were an independent part. The
idea ( for those who came in late) is to construct a .fully round part by
overlapping three contoured reliefs.]

> I'm guessing you have the
> rotary table on its side, with the axis of rotation parallel to either X
>or Y.
>
> [Right- it's parallel to the x-axis.]

OK, that would be called an A axis.

[If I were actively moving the rotary axis during the cut, it would be more
useful to call it that, but since I'm just indexing on this axis I've been
leaving the A axis at zero for each iteration, so as not to confuse the
MaxNC control software.]

> [Well, I'm still unclear as to how tool offsets function, but as it stands
> I wasn't using any, as far as I know. The question was- should I be? The
> next question: Is it possible?]

If your software doesn't support tool offsets, then it must be compensating
for the tool radius itself. Do you specify the tool size anywhere in the
program?

[Yes, I did, in DeskProto.]

I am pretty sure tool radius compensation is a 2-axis thing, although it could
be made to do 3 dimensional offsets, it gets a lot more complicated. When you
get into 3-D contouring, I think the only program that can adequately program
the toolpath without gouging is the one that has the entire surface description
available, so that has to be your CAD program, not the CNC control.

[I checked with DeskProto (they're actually pretty responsive considering
they're in the Netherlands) and they said, yes- DeskProto does address the
tool-radius offset compensation issue. They said that when running the part
I should leave compensation turned off in the (MaxNC) control software, so
as not to over-compensate. DeskProto is good at contouring, and it deals
with the entire surface description.]

> [Actually, I wasn't changing the names of the axes, just rotating the part
> geometry along its own axis (parallel to the x axis) and instancing it as
> another part. DeskProto put in this facility with its recent upgrade to
> version 2.0, and it was my hope that the various instances would line up,
> since it seemed that the axis of rotation and the axis of construction
> would be congruent if I zeroed y and z to the tailstock center. It almost
> works...]

OK, now I see how you did this. But, it requires you to have the tool
aligned as the PROGRAMMER of DeskProto wants it. So, you need to
see whether there is any documentation on this in the DeskProto docs.

> But, maybe that brings us to the problem, being that if ONE axis
> is calibrated to the CENTER of the ball, ALL axes must be calibrated
> to the same point, the center. If you are applying a 1 mm offset for the
> cutter radius, when you turn the coordinate system, then the other axes
> must also be offset from the same point.

Yes, I think this might be the most logical way, so that the ref point
is the center of the ball. Are you getting too much material left, or is
it taking too much material away? Since you located at the tip of the ball,
I would guess it is leaving too much material, because the tool is one mm
higher than the program expects.

[As far as I can tell, it is neither taking too much or too little; it just
seems to be moving the part over a bit in the Y axis, or maybe in the Z.
I'm not very sure about this though- I'll take a micrometer to it and
check.]

> [Except for the x axis, all axes are calibrated to the center line. It
> didn't seem to matter where the x axis started, except that x=0 should be
> toward the left, where the chuck is, in order for there to be room to cut
> the part. There doesn't seem to be any discontinuity between the instances
> in the x-dimension; the problem seems to be confined to the y-axis, or
> perhaps the z as well.]

Hmmm, yes, I think I see. The X is never affected by the rotation, so it
should be consistant. The Y is affected, because the rotation around the A
axis changes the Y coord of the completed surface.

Anyway, if the problem looks like a 1 mm gap of material between the faces,
then bringing the tool to touch the work, and setting the coords to Z=1mm
instead of Z=0mm as you are doing now, should fix it.

Jon

[I did some experiments like this, zeroing to the leading edge of the tool
in Y instead of the center, and in Z to the ball diameter point instead of
the tip (of my ball-end tool). While I was able to make the problem worse,
neither adjustment made it any better. When I asked DeskProto, they said to
zero to the tip and the center, as I had been doing originally. It's
possible I wasn't doing this accurately enough- I'm going to make a gage
block for zeroing and see if this helps.]

>
> In this particular instance, I think the only thing that needs to be
> addressed is whether you cutter comp off a flat xy plane or off the rotary A
> axis. Cutter comp must take place with your rotary table centerline axis
> being your zero point rather than a flat surface. This may be the reason the
> three panels don't line up.

Right, very good point. The Y=0 must be set with the spindle centered
exactly over the rotational axis of the index head, and the rotational axis
of the
index head must be accurately parallel to the X axis travel. Also, the Z axis
must be zeroed at some known distance above the rotary axis (could be zero,
exactly ON the rotary axis, too). Whatever is consistant with the way the part
coordinates were layed out on the drawing.

Jon

[Yes- "exactly" and "accurately" are words I've never been too comfortable
with (I'm an artist, right?) , but I suppose I could get closer.. But I
have been trying to zero to the rotary (A) axis line- if this is going to
work, that would have to be the axis of symmetry. Thanks for your help with
this, Jon- I'll let you know how it goes.]

Andrew Werby




Andrew Werby - United Artworks
Sculpture, Jewelry, and Other Art Stuff
http://unitedartworks.com