why G17 G18 G19?

[T-bolt]

i would like to see a technical article dealing with why standard g-code restricts arcs to the 3 major planes. i realize that with some newer controllers you are able to define custom planes. i just wonder why arcs couldn't have been defined: xyz start, xyz end, xyz center. i work with visualmill and wonder how much smaller my programs would be if the cam system was allowed to fit arcs freely.


thanks,

jeremy hill

(i posted this question at mmsonline a week ago, and no one yet has replied)

[HuFlungDung]

Hi T-bolt,

This would just be a guess, but I suppose the limitation was natural due to the inability of many controllers to be able to perform an interpolation of this type.

In my mind, the stickler on this issue is the feedrate of the three axis during an interpolated movement such as you have postulated. Normal 2 axis circular interpolation has a very predictable relationship between the position and the speed, defined by the trigonometry of an arc. Normal 3 axis linear interpolation also has a simple formula which can calculate the 3 feedrates to move between two points in a straight line.

Now, when I think about crossing planes while doing a circular interpolation in 3 axis, my brain shuts down For example, cutting a vertically oriented arc, while moving from the X+Y+ quadrant to the X-Y- quadrant (a 45 degree plane rotation), the controller needs to generate a feedrate in X and Y that contains both a circular and a linear component to the feedrate. This would be a highly intensive calculation that would have to be "customized" for every conceivable rotation of the plane you want to move in.

It could just be a matter of the raw processing power needed to calculate the continuously varying feedrates to each 3 axis increment. Lots of controllers are still being manufactured with cpus from the early 90's.

[hardmill]

Im going to say because it has been a standard for who knows
how long. I do applaude your train of thought.
With current advancements I would'nt be surprised to see
something along this line in the future.
Good job, way to use your melon.

PEACE

[HuFlungDung]

Maybe someone could invent a 6 axis machine that has an XY table mounted atop another XY table, and dedicate one table to linear movements, and the other one to circular movements. Combine the two together, and T-bolt would have his dream machine. Mind you, someone would have to write software for it.

[Bill]

Hey guys there was a mill that HFD is talking about, when I was working at Rockwell in the jigbore department we had a rotary mill, the table standard xyz, but the head/spindle was a rotary table with xyz movement.

Bill Johansen

[T-bolt]

here's where you guys can maybe straighten me out a little. it seems that it wouldn't have been hard for the controller to establish a cutting plane based on the arc coordinates, since arcs are always planar entities. then the move would be standard circular interpolation. it seems that this would be a very logical development, since the code required to make most contouring passes could be magnitudes smaller, and the block processing speed of the controller wouldn't be nearly as much of a factor(read: hsm).

thanks for the input,

jeremy hill

[HuFlungDung]

The trouble is Jeremy, that every motion actually has to be factored across the 3 servo amplifiers. Certain simultaneous movements are possible only because the mathematical relationship is well defined.

For example, a 3 axis linear move, has three different feedrates that can be computed for each axis so that they arrive at the endpoint at the same moment. The 3 axis drives are unaware that the other ones exist, only the controller establishes the relationship between them, by factoring the output vector into its components.

For a 2 axis circular interpolation, one axis has a sinusoidal relationship to the feedrate, while the other axis has a cosinusoidal relationship. Again, neither axis "knows" what the other one is doing.

I am not certain how simple it would be to create a feedrate that is not linear and not circular. Essentially, the curves you are describing would be ellipses, parabolas or hyperbolas within the Cartesian coordinate system, so these types of functions would need to be incorporated into the controller software to produce the desired feedrate factor and smoothness. At present, the simplest way to do this is with interpolated segments that are approximately correct in what they represent.

Look at it from another aspect. Suppose you have software that is capable of doing what you want. Then, the problem becomes "how do you prevent infinite programs?" Indeed, if you break free from the Cartesian coordinate system, then there is no limit to the cutting planes, and whether they should be radial, or planar patterns or what. I think that once you have made a choice about what kind of path you intend to cut, that there is not very much functionality in cutting only arcs "across planes" because this would very seldom pan out to be the actual situation.

[balsaman]

I just came back from robot programming training. The robots are typically 6 axis. They are able to move the tool in arcs in any plane using all 6 axis simultaiously based on 3 programmed points. Start, middle, end. Quite cool to watch. I think it's done the way it is in CNC because of the limitations of controllers/software back from the 60's when the whole cnc thing was new. The standards were set then, and we are still stuck with them.

Eric

[balsaman]

And don't forget, we have just had software that is capable of doing this kind of 3 axis 3D toolpaths in the last 10 years, and the gcode produced needs to run on 30 year old mills.

Eric

[coldplugs]

Originally posted by balsaman And don't forget, we have just had software that is capable of doing this kind of 3 axis 3D toolpaths in the last 10 years...

This is true for PC based software but mainframe APT/360 and later APT variants had powerful 5-axis capability starting in the late 1960's. The output CL files had enough info to do arcs on any plane, if the postprocessor and machine tool controller were up to it. [I'm very old, so I remember this stuff]

If you were refering to the software in the CNC controller, sorry to butt in.

[balsaman]

I was refering to CAM software.

Eric

[CoolHand]

OK, correct me if I am wrong here, but isn't cutting an arc from xyz1 to xyz2 possible right now?

My machine will cut helixes in any of the three planes, thus giving 3D arcs in pretty much any direction.

Granted, its only a 3 axis machine, so the tool is always comming from the top, but you can get a path that curves in pretty much any direction.

Am I missing something here, or is this uncommon?