Tolerance and Smoothing

Tolerance and Smoothing

Anonymous
Not applicable
6,162 Views
14 Replies
Message 1 of 15

Tolerance and Smoothing

Anonymous
Not applicable

Why is it such a pain to dial these settings in? Its not just a set once and forget thing, its a continuous battle to get decent code out of this program because of these settings. If something is drawn with an arc, a circle or a fillet, then just use a freakin' G2 or G3 command! I have wasted more time dicking around with tolerance and smoothing to get decent code than I have actually designing the parts I machine.  You go a hair to far on either tolerance or smoothing and your code is 16 million lines with tons of tiny X,Y movements instead of just commanding a G2 or G3 arc motion which results in data starving, faceting and in general jerky, inconsistent motion control. I have very tight machine and it follows the code exactly creating tons of faceting if I dont spend on average 10 to 15 minutes per operation adjusting tolerance, smoothing  and post processing to check the code length. There has got to be a better way guys.  CamBam produces better code in less time, only downfall is it is 2D and cant do adaptive strategies. 

0 Likes
6,163 Views
14 Replies
Replies (14)
Message 2 of 15

LibertyMachine
Mentor
Mentor

Would you be able to share a file that could give us a good example of your settings and struggles?

 

For adaptive, I keep it simple. .0005 tolerance, .001 smoothing. 9 times out of 10 I get good results.


Seth Madore
Owner, Liberty Machine, Inc.
Good. Fast. Cheap. Pick two.
0 Likes
Message 3 of 15

Anonymous
Not applicable

Here it is. Tried .005 and .001, 1245 lines of code for something I could probably hand write down to about 300 lines of code. Whole point of this is so I dont have to hand write the code. 

0 Likes
Message 4 of 15

LibertyMachine
Mentor
Mentor

Looking at your file, I'm actually not seeing anything amiss. Your code isn't poorly toleranced or smoothed. You can see a representation if you turn on Points in the CAM Simulation:
2018-06-28_14h00_30.png

In the above photo, I don't see an abnormal amount of points. Posting out the code, it's the size I'd expect to see from Adaptive.

 

Adaptive is not meant for anything other than a high speed roughing operation. Most machines that would succesfully utilize such a routine would also have certain G codes that would load several 100 lines of code into a buffer and produce even smoother motion. Haas and Fanuc have these options

 

If your machine behaves poorly with an Adaptive cut, might I suggest using 2D Pocket? Doing so produces a program that is 285 lines long:

2018-06-28_14h03_56.png


Seth Madore
Owner, Liberty Machine, Inc.
Good. Fast. Cheap. Pick two.
Message 5 of 15

Anonymous
Not applicable

I run Mach4 with a ESS. Current post processor is the Generic Mach3 Mill/ mach3Mill that I modified for manual tool changes. I notice if I keep increasing the tolerance in the post process screen, it uses more arcs rather than faceted steps. 

Capture.PNG

0 Likes
Message 6 of 15

HughesTooling
Consultant
Consultant

Adaptive contains a lot of spiral moves, you can modify your post to convert these to arcs if they're within a tolerance. I have a heidenhain post I've modified, also does helical moves and the code for the adaptive op comes out as 650 lines.

 

@LibertyMachine Are you still using a post with the spiral mod?

 

Mark

Mark Hughes
Owner, Hughes Tooling
Did you find this post helpful? Feel free to Like this post.
Did your question get successfully answered? Then click on the ACCEPT SOLUTION button.

EESignature


Message 7 of 15

HughesTooling
Consultant
Consultant

@Anonymous As for your question on tolerances, I'll  leave adaptive out as it's going to produce a lot of code and high tolerances increase calculation times. 

 

For 3d finishing ops I find a smoothing of 2 to 5 times the tolerance produces the best arc fitting, might not always be the smallest code but the arcs fit better. Working in mm I use a tolerance of 0.01 and smoothing of 0.02 for bigger parts and 0.005 and 0.01 occasionally 0.001 and 0.01 but calculation times are long.

 

For 2d ops, an arc or rad should come out as G02/03, tolerance should not make a lot of difference if Fusion reports the edge as an arc.

 

JFYI on the Adaptive in your file you used 0.001 and 0.005 total tolerance of 0.006 and in Fusion you get 21kb of code but the toolpaths wonder and you get some sharp moves.

Clipboard02.png

If you use tolerances of 0.0005 and 0.001 total of 0.0015 you get smoother toolpaths but more code, 26.5kb. Now this bigger code will probably run smoother so don't always assume loose small toolpaths are better.

page.png

Mark

Mark Hughes
Owner, Hughes Tooling
Did you find this post helpful? Feel free to Like this post.
Did your question get successfully answered? Then click on the ACCEPT SOLUTION button.

EESignature


0 Likes
Message 8 of 15

LibertyMachine
Mentor
Mentor

@HughesTooling I've updated my post several times, since my first post processor did not have the syntax needed for probing, among many other things. I believe I've not carried over that function. Could you PM me the edit so I can bring it in?


Seth Madore
Owner, Liberty Machine, Inc.
Good. Fast. Cheap. Pick two.
0 Likes
Message 9 of 15

Anonymous
Not applicable
My main concern comes when there are 20 to 30 little tiny XY moves trying to make a corner. Ive whitnesed this on a few parts Ive machined when I didnt doublt check the code and I left it at a "tight" tolerance. A simply 90° corner with say a 1/4" radius will not be round but rather have 5 to 10 flat facets. Ive also have these super large code files bog down my little controller as it commands back to back to back .0001" moves when my steppers cant even make those small of moves. Im limited down to about .0005 moves, any less and my steppers physically do not move as my system is all direct drive. Eventually Ill get servos but nothing I have machined yet has to be within .0001".
The other part of this that concerns me is that adaptive is ment to keep a cutter tooth engaged at all times correct? And remove large quantities of material. If its making a faceted corner, to me, thats not maximum material removal and its not keeping the cutter engaged as well as it could.
I machined 2 flanges from 316SS, one with their typical tolerance settings and one where I screwed around for an hour or better finding what I felt was optimal settings that created mostly arcs instead of straight line moves. The end product of the second flange was overall better. The tool ran cooler, virtually zero chatter compared to the first code which chattered and squealed occasionally with all the tiny straight line moves. The surface finish came out better on the second one and my cutters edges looked better as I changed cutters from the first one to the second one for comparison sake. Didnt help that about 2-1/2 hours into the first one with the unmodified tolerance, my machine actually straight up crashed into the part, ruining the cutter and the part. And its not like I stopped the machine to change anything, it just all the sudden plunged in the wrong place.

Hopefully these example help you guys understand why Im not very amused with the tolerance and smoothing settings when on my old school CamBam it can generate a 2D path with arcs without having to setup a tolerance or smoothing feature.
0 Likes
Message 10 of 15

Steinwerks
Mentor
Mentor

@Anonymous wrote:

The other part of this that concerns me is that adaptive is ment to keep a cutter tooth engaged at all times correct? And remove large quantities of material. If its making a faceted corner, to me, thats not maximum material removal and its not keeping the cutter engaged as well as it could.

Adaptive style toolpaths is to keep the tool engaged with a constant stepover within the tolerance specified. If you are running full-bore load on the tool, even a small variation in cut width can end that tool instantaneously. So considering that arc fitting (Smoothing in HSM) inherently increases the tolerance value it can increase the tool engagement into dangerous (for the cutter) territory.

 

That said, if the toolpath doesn't fit within the tolerance for the arc, it simply won't fit, and the toolpath uses the standard tolerance definition to develop as many linear segments as necessary until it reaches another portion of code that will fit in the arc tolerance definitions (this is as I understand it, Smoothing may increase the linear tolerance but I'm not positive on that).

 

Inner corners are tricky as well since stepover cannot be calculated as if it were linear; the tool is taking a wider radial cut turning an inner corner than it would traversing a straight contour or outer radius. This is especially true when using a very small minimum cutting radius going into a corner that is technically smaller than the tool (equal tool radius corner with stock to leave, for example). The linear moves translate into sharper transitions than would fit in the arc described in the Smoothing setting.

 

You can also check out this thread which has some great info as well: https://forums.autodesk.com/t5/fusion-360-computer-aided/understanding-smoothing/td-p/6636189

Neal Stein

New to Fusion 360 CAM? Click here for an introduction to 2D Milling, here for 2D Turning.

Find me on:
Instagram and YouTube
Message 11 of 15

kb9ydn
Advisor
Advisor

To some extent it's inevitable that the adaptive strategies will produce bursts of short line segments, just because that's how they work.  You can make it produce more sparse code my using a loose tolerance with a large smoothing value, but the tool paths produced won't necessarily be "smooth".  This may be ok if you have a machine control that is severely challenged in processing speed or by how many lines of code it reads ahead of time to plan its moves.  I have an old Bridgeport EZTrak that's absolutely horrendous in this respect, to the point where it's nearly unusable for adaptive paths.  It's so bad in fact that if the feed rate is above 20IPM it will come to a complete stop after EVERY move.  It's the craziest thing, but then again it was made in 1996 and used a PC based controller with 486 processor running at like 33MHZ.  So you can't expect much.  Anyway, I digress.

 

After reading through this I'm wondering if maybe the problem is with how Mach4 is configured.  Have you checked the setting for look ahead lines?  I know very little about Mach4 but I do know it has a look ahead setting and maybe some others for trajectory smoothing.

 

 

C|

0 Likes
Message 12 of 15

rhdfmail
Advocate
Advocate

Bumping this one...

@HughesTooling  What mods in the post is needed to make it output arcs?

Had one part today where none of all the settings for smoothing/tolerance  i tried resulted in any kind of arcs...



0 Likes
Message 13 of 15

Laurens-3DTechDraw
Mentor
Mentor

If you can attach the part here we can take a look if it's your post or the CAM being the issue?

Laurens Wijnschenk
3DTechDraw

AutoDesk CAM user & Post editor.
Found out the hard way is the best way to win.


0 Likes
Message 14 of 15

rhdfmail
Advocate
Advocate

Because of NDA i had to create a new model with only the feature that gave me headache. 
It was a prototype so the surface-finishes wasn't that important this time, but next time I do something similar it might be an issue.

To verify I  created different strategies to check wich ones, if any actually makes arcs. 
2D Contour: arcs
2D Conour: arcs, except in the link-moves

Ramp: Line-segments
Blend: Line segments

 

0 Likes
Message 15 of 15

Laurens-3DTechDraw
Mentor
Mentor

Ramp by nature will be tricky to create arcs. But using steep & shallow's arc fitting would create a lot better results already.

Blend is half-finished as a toolpath so that sort of explains that one.

Morph usually is somewhere in the middle. It's not great at smoothing/point reduction, yet it's not as tricky as ramp.

 

Depending on your control you might actually like evenly spaced points more than the arc fitting in Steep and Shallow.

Laurens Wijnschenk
3DTechDraw

AutoDesk CAM user & Post editor.
Found out the hard way is the best way to win.