Cutting Problem, I Need Some Help.

[shilala]

I took time to do a test run of the raised panel effect I made, and for some reason the curved top makes her come unhinged a bit.
Trig-wise my guess is that it should be okay. I had to think a bit to get the curves to become an equal spread offset. The shapes seem clean, the nodes match and they're nice and clean (or so I think).
I was hoping you guys could take a look and give me some ideas?
I'm thinking it may just be my toolpath settings?
Here's a link (https://www.box.com/s/jx3qc8o027p2elxbg2nj) to the Aspire file and toolpaths I used to cut this. I'm hoping anyone can download them from Box, the files were too big to zip and attach.
I also zipped the toolpaths and attached them here, I'm not sure what good that'll do, but I know it's standard procedure. :)

Here's a few pics to explain what's up...

Here's the panel.

http://i925.photobucket.com/albums/ad96/scottshilala/C399FD98-E1BD-4DCB-A1BC-D8DCEE897138-10963-00002259EF128202.jpg

Here's the problems, a raggedy edge where it "pecked" the edge. Also the toolpath started on the outside, then skipped over to the inside, and left that divot you see in the middle of the slanted rise.
I should say that I'm running a BT48 PRStandard, 3ips, 1/8" ballnose.

http://i925.photobucket.com/albums/ad96/scottshilala/15B409BF-955C-4616-AE76-EC0903ED1CBC-10963-00002259FA38D42F.jpg

And a closeup...

http://i925.photobucket.com/albums/ad96/scottshilala/615B11EE-3AA0-489D-808E-D86DA79FC054-10963-0000225A02D6A924.jpg

Thanks guys, any help or ideas will be MUCH appreciated.

[oddcoach]

It looks like you problem is caused by the tool rastering and not following the shape you want. from the pictures it looks like the problem only exists on the first cove. you need to add 2 profile tool paths with a ball nose tool with the same radius you want for the cove. Just select the innermost and outermost vectors and make the depth of cut equal to the radius of the tool.

[srwtlc]

Hi Scott,

If you look at your 3D view, you'll see that there are a bunch of pixels extending above the surface that you want cut and the toolpath is following them up and down.

First, I'd jack up the resolution of your model in the "Set job dimensions and origin" dialog to at least "Very High 7 times slower". You can go higher by holding down Shift when clicking the setup button, but things can slow down depending on your computer.

Then, if you create a zero plane for the size of the model or with a rectangle larger than your sweep by creating a flat profile with 0 base and then make your sweep and you should see that the pixels are not as bad. At that point, you could bake the two together. Then recalculate your toolpath again. Apply the appropriate vector boundary offset for the radius of your tool (-.0625) and that offset toolpath should be better. The offset strategy has a bad habit though of making many up and down moves when reaching the top of a profile like you have there (bowls and spheres do the same, drives me nuts) so I usually just go the raster strategy route. I had to go as far as -0.075 to get rid of the vertical moves. You could increase/change the size/shape of your sweep cross section to allow for that if it doesn't machine the way you would like it to (because of losing a small amount of the profile due to a larger offset).

I did modify the cross section a little by slanting the vertical walls a bit. May have helped some.

New file on SkyDrive...http://sdrv.ms/118lcqN

[bleeth]

You can get rid of a lot of that with smoothing on your model also.

[shilala]

Thank You for all of that, Scott.
It took me awhile to understand everything you said, but I chased it out when I looked at the file you made up.
I have no idea why the zero plane model makes a better boundary for the sweep model. I'm guessing the pixels are much smaller and match the pixels inside the model, so the sweep model can't take liberties? Understood or not, I'll make use of the trick.
I can see why slanting the sides of the profile helps. It should stop the "peck" and make a rise, or force the path to go around the model. The bit has a "defined something" to work on rather than the nothing it had on a straight wall.
Baking the two makes really good sense. That smoothed everything really nice!

I wonder what would happen if I didn't connect the rail nodes during the sweep? It seems like I'm forcing the toolpath to try to work with spaces that aren't equidistant, so they get all goofy, even if it's not "seen". When the rails line themselves up, it seems to stretch to points so that they're equidistant. Then maybe the toolpath would just go around in circles?
It'll probably just make a big mess, but it's worth trying just to see what happens. :)

I'll cut your model a little later and see how it goes. It looks WAY cleaner than mine, for sure. I'll do a little more messing around, too.
Thanks so much for your time, brother. It's very appreciated!!!

Hi Scott,

If you look at your 3D view, you'll see that there are a bunch of pixels extending above the surface that you want cut and the toolpath is following them up and down.

First, I'd jack up the resolution of your model in the "Set job dimensions and origin" dialog to at least "Very High 7 times slower". You can go higher by holding down Shift when clicking the setup button, but things can slow down depending on your computer.

Then, if you create a zero plane for the size of the model or with a rectangle larger than your sweep by creating a flat profile with 0 base and then make your sweep and you should see that the pixels are not as bad. At that point, you could bake the two together. Then recalculate your toolpath again. Apply the appropriate vector boundary offset for the radius of your tool (-.0625) and that offset toolpath should be better. The offset strategy has a bad habit though of making many up and down moves when reaching the top of a profile like you have there (bowls and spheres do the same, drives me nuts) so I usually just go the raster strategy route. I had to go as far as -0.075 to get rid of the vertical moves. You could increase/change the size/shape of your sweep cross section to allow for that if it doesn't machine the way you would like it to (because of losing a small amount of the profile due to a larger offset).

I did modify the cross section a little by slanting the vertical walls a bit. May have helped some.

New file on SkyDrive...http://sdrv.ms/118lcqN

[shilala]

You gave me the idea to break up the profile into a number of pieces riding inside concentric vectors. At least the inside-most and outside-most parts of my profile.
If I break up the sweep into smaller pieces, and do a number of sweeps, it may not compound the problems.
Then I could bake everything, and the different coves should be pretty well defined to take them to toolpathing.
Hmmm. :)


It looks like you problem is caused by the tool rastering and not following the shape you want. from the pictures it looks like the problem only exists on the first cove. you need to add 2 profile tool paths with a ball nose tool with the same radius you want for the cove. Just select the innermost and outermost vectors and make the depth of cut equal to the radius of the tool.

[shilala]

Scott, I tweaked your file a little bit and got things cleaned up pretty nice.
The baked component stacked for some reason and was twice as deep as it was supposed to be, so I deleted that bake and rebaked.
I did away with the offset completely, because it left a groove on the surface, outside the vector. It doesn't show in the 3d view, but it's very prominent and wouldn't fly at all.
There's still the little groove in the center of the slanted rise. I'm hoping it'll iron out when I make this bigger. The sweep needs to be 1 1/4" wide for the desk panels.
Whew. Thanks again, man. This has been a very simple thing that became a big pain in the ass. I still can't believe it's so convoluted.

[srwtlc]

Scott,

Without the zero plane, the tool can/will tend to follow the thin possibly jagged edge pixels and cause tool marking along the edge. If your model resolution is low and the model larger, this can be worse.

It's not so bad in your situation, but if you don't connect the rail nodes, you can end up with a twisted effect with frame like sweeps, but you have to have the same amount of nodes in/out for it to work best. What you did by breaking the sweep up into sections and then baking together is the way to handle that situation. There was/is a tutorial video on doing that exact thing somewhere. Changes to that are coming.

The stacking issue happens when you have two of the same components checked/turned on and they are added together. Just uncheck/hide the one you don't want shown or delete it if you no longer need it.

Your sweep rails were that far apart (1.25"), but the cross section was a bit smaller. Don't know if that is an issue or not, but you could try 'scale cross section to width' in the sweep form or just draw it to scale.

[shilala]

Scott, I still never broke the profile into chunks. I used a lot of the things you mentioned, tweaked, and the whole thing is absolutely perfect less a problem in the slanted cove.
The toolpath still splits and cuts from the outside, then the inside, but now it goes down and cleans all four corners of the sweep first. It's odd. When it meets in the middle, there's one bit pass that goes just a bit deep on one side and the top, then it leaves a bit width high on the bottom and the other side.
It doesn't show in either preview, 3d modeling or toolpath preview and I have both cranked up to maximum resolution.

I ran one panel. I'm not happy that it isn't perfect, but it's driving me nuts, so I kind of decided to back off it.
I guess I'll stop and break the slanted cove out of my profile, sweep it by itself, then bake it into the finished product I have already.

I'll just be copy/pasting this finished job onto all the separate panels of the desks, so I guess it's no big. Yesterday I convinced myself that I'd have to repeat the whole process on every panel I designed. That wouldn't even be a big deal but my back is killing me and this shop stool is shortening my patience dramatically. :)

Here's where we are so far. It's a HUGE improvement.

http://i925.photobucket.com/albums/ad96/scottshilala/94372DD0-F160-49E4-BF86-317FE709CA84-141-0000001175D69CFC_zps0e3ec29c.jpg

[srwtlc]

You could try a raster cut instead of offset to see if it clears it up. Maybe the panel is moving after relieving some stress and if that part of the cut is done towards the end of the file, things may have cupped/warped a little. I don't use the offset strategy much due to some of the very issues you're having. Some places it's good, but others not so much.

[myxpykalix]

Scott,
sometimes it's better to just walk away from a problem for a few hours or a day THEN come back and look at it with a clear mind. Frustration tends to keep you from seeing everything clearly. The answer is right in front of you....
(Now I sound like Master Po..brady;) ) but he's usually right

[shilala]

You could try a raster cut instead of offset to see if it clears it up. Maybe the panel is moving after relieving some stress and if that part of the cut is done towards the end of the file, things may have cupped/warped a little. I don't use the offset strategy much due to some of the very issues you're having. Some places it's good, but others not so much.
This is panel number 7, actually. I ran 6 tests (jogged 2" and overlapped a couple so I could just see the top and bottom) on 4 sides of "not really scrap" cherry. It shows up in all of them.
The raster might work to some extent, but the problem is definitely in the sweep. It's just a little "twist" in the center of that angle, the same kinda of thing that happens if you don't connect your nodes, but very small in comparison. I think it's just the program trying to rectify the fact that things just aren't completely equidistant, piling up on one side and sinking in on the other.

If I could mask that area and smooth it, that'd do it, I think. (I had guessed that's what the "clear area of selected component inside inside selected vectors" function does, but it's not like masking.) I haven't had cause to use smoothing, so I don't know if it adds or just subtracts. If it will fill that area, I can easily guess where to smooth.

Breaking the profile apart didn't really work. It just went goofy. I didn't follow the same steps, so I still need to do that.
At this point I'm working on a good panel and hate to wreck it, but if that's what it takes, so be it. Right? :)

Thanks for sticking with me, brother!!!

[shilala]

Scott,
sometimes it's better to just walk away from a problem for a few hours or a day THEN come back and look at it with a clear mind. Frustration tends to keep you from seeing everything clearly. The answer is right in front of you....
(Now I sound like Master Po..brady;) ) but he's usually right
I've slept on her twice and walked away a whole bunch of times, Jack. I agree with Brady 100%. I gave up and was going in the house for lunch a little bit ago and had a whole bunch of lightbulbs, turned around and here I am. I guess I'll have lunch some other day. :D

[shilala]

Welp, it's time for intermission. I broke my last 1/8" bit.
On a positive note, I found 1*, 1/8", 1/4" shank coated carbides just like the onsrud 77-102 for 38 bucks. They're right here. (http://www.cuttingtoolsource.com/tapered-ball-nose-end-mills.html)
I ordered 4 of them cause shipping is 15 bucks flat. If they ever show up I can get back to torturing myself.

[rcnewcomb]

Here's a writeup I did awhile back about adding a zero plane: ->Link (http://www.vectric.com/forum/viewtopic.php?f=28&t=6474)

[srwtlc]

Thanks Randall, I was lookin' for that for a better explanation than I could come up with. ;)

[shilala]

Here's a writeup I did awhile back about adding a zero plane: ->Link (http://www.vectric.com/forum/viewtopic.php?f=28&t=6474)
Awesome write-up, Randall. Scott had me ironed out really well, but you put an answer to why it worked.
I had guessed there was a sharp vertical wall outside the model with odd pixels radiating into it and nothing to hold the tool back.
I wonder what would happen if the software was coded so that dragon land had zero plane pixels floating on the whole thing, then just got displaced when you dropped the boat in?
I've not done enough stuff to see where that approach would create a negative issue, but I can certainly see a ton of places where I can use this trick.
Thank You!!!