Ver 9.012 Extrude Problem

[AZRoger]

Single rail extrude is not producing proper components.
The profile I'm using is shaped like an M and is 10 mm wide.
It and the three components are shown in the attachment.
The rails for the three test components are 50 mm, 150 mm and 300 mm long.
The resulting components are 10.112 mm, 10.414 mm and 10.81 mm high.
The components are also 50.154 mm, 150.414 mm and 300.843 long.
I'm using the "Use Vector Cross Sections" option, no weaving and no Scale to exact height.
I discovered the problem in a project with 100 or so components.
However, it is repeatable in this "clean start" project as well.
Any hints? Is this a bug in Aspire? If so, someone else should have seen it too.

[AZRoger]

I repeated the test with Aspire 4.514.
It doesn't do it correctly either, but the errors are smaller.
Heights are 10.112 mm, 10.29 mm and 10.396 mm.
Lengths are 150.154 mm, 150.302 and 300.368 mm.
Roger

[ohiolyons]

If I understand what you are trying to do I got similar results. Aspire 9.012.

my rail is 2.025" long and the extruded cross section measures 2.022"

My M is 0.25" tall and the extruded cross section is 0.2494"

I think the answer the experts are going to tell you that Aspire is artistic based and NOT a solid modeling program.

The differences are incredibly small and quite frankly wood moves more than that due to normal expansion and contraction from humidity.

If 0.112 mm is critical to your work Aspire may not be what you should use for your modeling.

John

[Adrian]

Not to mention the machining resolution of the average CNC router...

[mtylerfl]

...

The differences are incredibly small and quite frankly wood moves more than that due to normal expansion and contraction from humidity...

John

Not to mention the machining resolution of the average CNC router...

So True!

[AZRoger]

OK.

What I'm trying to do is make trim pieces that meet at corners at the corners at a 45 degree angle.
If the short ones are a different size than the long ones, they don't fit very well. 3/4 of a millimeter will show.
Wood does change with humidity - I watch the grain so all the pieces will react the same way.
And my CNC Shark is pretty springy - with any tool pressure at all, it will deviate from nominal.
So I need to compensate when the tool paths are created.

My first clue as to the problem came when a toolpath was going to cut through the material.
The 2D "M" in the previous discussion was the same height as the material. It grew in the model.

I can "smash" the components after the fact and get them to within 0.01 mm of nominal.
That works. It just isn't a step I thought I needed to perform. Now I know. Thanks to all.

Roger

[mtylerfl]

Yeah, it does get a little funky sometimes when working in a pixel-based environment. Plays with my head a little when I'm looking for an exact dimension from a model.

On occassion, I have made a model, then resized it along my "most important" dimension. Mostly I've done it just to make me feel more comfortable! Other times, I'll intentionally make a model larger in one or two axis, to be trimmed more precisely with a profile cut (as in Molding or box sides that will be mitered on a chop saw or table saw).

I think I know where you're coming from.

[ohiolyons]

One thing to try is modeling resolution.

Since Vectric software is pixel based, higher modeling resolutions should yield more accurate results.

It is all about pixels per square inch or MM in your case.

There are tutorials and forum posts out there (the names escapes me right now) that addresses model resolution and pixels per unit of measure.

[AZRoger]

I think my final solution will be to model each section of the trim individually at the highest quality. (See attached screen shot.)
I'll do the Center, a right angled part and a left angled part each with a very small material size. Lots of pixels per square mm.
I can then use Aspire's "array copy toolpath" function to extend the trim piece as long at it needs to be.
So it won't ever be modeled full sized, just carved full size into the much larger actual work piece.

I also poked on the 4 million pixels feature with different sized work pieces. The hope was to use "even" numbers to get better results.
If it's 4,000,000 pixels, then a piece 20x500 should give me a nice round 20 pixels by 20 pixels per square mm. This didn't help.
So maybe it's Mega Pixels - that would be 4,194,304 total pixels. If each side of the work piece was an even power of 2 we'd get nice numbers.
That work piece was 32x512. This gives a nice round (in binary)16 pixels by 16 pixels per square mm. This didn't help either.
There must be some sort of accumulating rounding errors in the calculations. Don't know. But now I know and can adjust to the real world.