Swoosh !!!

bsharp:

Wow - that went right over my head. I'll let you explain that theory along to Brian.

Sage

The picture below is an interesting example of how the arc in the profile should have been trimmed to stay within the starting area.
The profile should never cross its self. Although there is times when you may want to cross a profile like in a figure eight it should not be possible to do so using left or right cutter comp.
It should only recreate offset periphery geometries for the first area that can be painted "Flooded" with a brush "Tool" the size of the current active tool.
The specification of the G4* offset word would determine in witch bounds of the area to fill.
A determination of open ended and area type geometries would need to be determined.
Geometry specifying a G4* word and which does not intersect itself would be considered open ended.
Geometry intersecting itself without the G4* word would be a center line tool path no offset needed.
Geometry intersecting itself with the G4* word would need to be "Flooded" like stated above.

If we can get that working in the X Y plane then maybe we can move on two all three. 
 
     

It's cutter comp, not comp / pocketing / magic bad code correction.

It's hard enough to just get it to offset correctly. Comp shouldn't try to decide what I'm trying to do, it should just offset the coded path. That's it.

It's cutter comp, not comp / pocketing / magic bad code correction.

It's hard enough to just get it to offset correctly. Comp shouldn't try to decide what I'm trying to do, it should just offset the coded path. That's it.

Yea but look at the picture in my last post. It did it wrong. I gave my Opinion on how to make it work correctly. What is yours?

Quote from: ger21 on October 25, 2008, 11:14:12 PM

It's cutter comp, not comp / pocketing / magic bad code correction.

It's hard enough to just get it to offset correctly. Comp shouldn't try to decide what I'm trying to do, it should just offset the coded path. That's it.

Quote

Yea but look at the picture in my last post. It did it wrong. I gave my Opinion on how to make it work correctly. What is yours?

I am afraid you did it wrong,  there are rules with any form of machining e.g. you have to drill the hole before you can tap it.

Well tool radius compensation is the same.

The golden rule is that the tool must always travel in the same direction tangent to the part, it must never reverse its direction, this means when you are working with angles less than 90 degree the point must have a tiny rad on the end, if not the path stops and the direction is reversed.

Graham

If I have to code the exact tool path and not the outline of the part then what is the point of advanced cutter compensation?
It sounds like a simple problem "you just need to offset it" but it is not.

Please read.
http://www.linuxcnc.org/handbook/gcode/diacomp.html

From the source:
"The cutter radius compensation capabilities of the interpreter enable the programmer to specify that a cutter should travel to the right or left of an open or closed contour in the XY-plane composed of arcs of circles and straight line segments. The contour may be the outline of material not to be machined away, or it may be a tool path to be followed by an exactly sized tool"

"Open ended and Closed" Sound familiar?

Thinking that it is simple is probably why it still does not work. And it probably will not work correctly until someone realizes this. 
Fire up an old GE550 control it will do the same crazy stuff. It took even the likes of GE quite a few years to figure it out.
 
All the new controls do pretty much what I said. I program machines every day that have no problems cutting a simple circle at any orientation or throw an err if the projected offset geometry intersects itself or trim the geometry to fit the tool. The simple fact that this trivial aspect of the control is not refined shows the immaturity of it. To make the cutter comp work correctly it will  take a little wider eye view to see than with what has been looking at it. Mach is to promising to be held back by such a primitive gaping wound.

I gave my opinion on how to make "Advanced cutter compensation" work. And I am sure you know what they say about Opinions!
"Everybody has one" And I showed you mine. 

Expecting it to work on a closed path is asking a lot. I think you're misinterpreting what they mean. If you read farther along, they talk about using an entry move. This is contrary to using a closed shape. Imo, it's probably almost impossible to comp a closed shape without a lead in move.

Looking through the EMC manual, I don't see anything different then it's always been, and have a feeling EMC won't do what you want either. Mach2's comp was based on EMC's, and mach3's advanced comp improved on it. Perhaps EMC has caught up and passed Mach's implementation, but I'll have to see it to believe it.

I don't think anyone's saying you have to code the comped path. You code the exact part, but you also need to code the correct lead-in move(s), and not try to do something that is invalid for comp.

General Method

The general method includes programming an alignment move and two entry moves. The entry moves given above will be used as an example.

I think I just was not explaining my self enough.
The cutter comp in EMC is atrocious. Basically non existent. I think it was just never fully developed.
And after looking at the EMC source code I am amazed it works as well as it does.
What it mainly lacks is the curve fitting part of the complete cutter comp package that most industrial controls nowadays have.
I believe that is what ART was trying to do with Advanced Compensation. Advanced compensation sometimes works but is buggy and problematic. I sat and thought about this for a few hours and noticed that in any profile that you would use compensation on would be ether a open ended or closed loop profile. Open ended profiles are not a problem. But it is when you get into complex closed profiles it gets buggy. Now with any closed loop profile you will only have one intersection. If for some reason the tool to be offset is to big for the profile this will cause another intersection in the generated offset profile. If you simply stop at that second intersection witch will be at the radius of the tool and continue on from that point within the current profile. It would be no different than milling an internal square with a 1 inch cutter and expecting the corners to be perfect 90'S. Would you do that of course not so doing this wouldn't be any different.
I have attached some pictures to explain a little about what I am talking about
The first three sets are of common offset geometry and show how Compensation in the software generates the tool path "Offset Geometry". The last is and example of incorrectly compensated geometry "tool to large for profile". My suggestion would be to trim the geometry at the second intersection like I had said. But it could ultimately just call an error and refuse to run the code. Anyway it would be a good idea to implement ether to prevent scrapping a part or parts by simply punching in a wrong tool table entry.

The last is and example of incorrectly compensated geometry "tool to large for profile". My suggestion would be to trim the geometry at the second intersection like I had said. But it could ultimately just call an error and refuse to run the code. Anyway it would be a good idea to implement ether to prevent scrapping a part or parts by simply punching in a wrong tool table entry.

While either of those two options would indeed be nice to have, I wouldn't expect to see either. I think most comp users would just rather see the bugs fixed, and I really don't consider this a bug. In your "correct" solution, while you don't get a gouge, you're still not getting what you really want. Isn't it pretty simple to just not use the wrong size tool? Probably an awful lot of work for Brian to check for tool table typos.

That's the thing I think it need's a little more than just a few bugs fixed to get it to work right. You shouldn't have to fight with the thing to make it work. Like I said I don't use compensation on the Mach mill so I never realized how bad it actually works. Figured it would be a simple thing for the Mach burning table like with the Mazak Lasers and MG plasma cutter but nope Mach instantly turns into a pain in the but as soon as you call a G4*. But even with the sad excuse for cutter compensation it is well worth the money. I do hope it gets fixed but I am not going to hold my breath.

Thanks for all the post to my topic
I am done with it.   

Other than a few minor bugs, it actually works pretty good for what I do, and I use it all the time.