formula not following in G02

[stagemachines]

Hello All, I am new to cnc  and gcode.  I am impressed with Mach 3 and the Hobby cnc gear.   I am building a custom machine.   I am finding that the formulas don't follow the X and Y Axis.  The machine has only two axis.  A rotary table (A axis)  and an linear axis (B axis)on that rotary table(A axis).  The tool is on that linear axis( B) and is constrained to a line that passes through the center of the  of that rotator (axis A).   A axis is the rotate and B axis is the linear arm. X and Y axis are active but are not connected to any motors.  This will be a plasma torch. It looks a little like a Banjo.  All axis (X Y A B) are homed to the center of the rotory table. It is in the R and D phase.
Idealy it will use just X and Y in the g code. 
The formulas are as follows.
Rotate      A= arctan(y/x)*57.2957795    ( it took me a bit to understand the default is in Radians. )
Radius on the arm    B=((x^2)+(y^2))^.5  ( simple Pythagorean theorem)

For any given point the formulas work great.   I have scaled the both axis and they run well, don't lose any steps, and return to home with very little backlash.

When I run gcode. ( my test has be a 5 inch square).  I am using 2dnc to  generate the gcode.  No tooling off set.  My machine will draw the square.  The corners are square and at right angles.   The sides of the squares are cutting as slight arcs. ( Coming off the line .25 or so) All 4 sides of the square have this  and the crown is away from the origin.  Opposing sides are parallel at the  5 ".  I have the same results any place I have where ever I cut this on the table.   The  A and B axis seem to follow in G00 And G01. with the error mentioned . But will not follow in any G02.  The A and B axis just drop out and stop. When the tool path progresses  back into G01 after G02, Mach 3 will pick A and B back up ( in the correct coordinates ).  If you "Hold the feed " and restart It will move the A and the B. This will only last for that block only.  Is this a bug?  Is it processing problem in my computer?  Is this a ABS or Inc (G90/91) issue.  I have tried it  both ways. Or even the G90.1/91.1 issue ( I think I have tried all combos)

Here are some questions. 

Does any one know why the G02 drops the formulas?
Do formulas process in ABS (Machine Co) ?
Will Mach 3 be able to process the feed rate?
Will I be able to use offsets with this or will  Xand Y all ways have to homed at the origin ( center  of the the rotate)?
Any thoughts would be helpful-

[BR549]

I don't think FORMULAS were ever fully developed. You may want to consider doing all the conversion inside the CAM post.

Just a thought, (;-) TP

[!returnmethere]

It should be updated and improved. But by and large, not bad!!!

[stagemachines]

Yes, I am very impressed with the program so far.  Unfotunaltly the whole machine is predicated around  the easy programing.  I was only willing to move ahead with it when I saw that you could Jog X- Y  in step and MPG and have the Polar machine move in the A-B.  I am willing to give precodeing in the compiler a shot.  I am using 2dnc  and lazycam.  I don't see how to do that in either.  I have looked a G16. G12.1 and g112. Any thought on what compiler I might try?== Adrian

ps- are the G codes listed in the list you get when you press the G-code button the only ones supported by Mach 3?

[stagemachines]

Oh, and thank you both for your help so far.- Adrian

[BR549]

Yes on the Gcodes on the list.  Lcam while free is an unsupported orphan. IF you are building a 0ne arm bandit plasma table please note it is an old idea(;-).

One problem you face is MACH doe not do kenimatics of a joint so it is not easy to predict the end point in 2d space. I know formulas have been tried in the past.  Don't remember if any of the projects are still around.

For a CAM for plasma I would recommend Sheet Cam .

Hope that helps, (;-) TP