#353 tool radius comp error

[Dewey Garrett]

Tool radius compensation error originally reported on dev mail list.

msg: "Tool radius not less than arc radius with comp"

observed in 2.5.4 and master 9c339d7

Reproduced with simplified example:
$ cat arcfail.ngc
; simple test using g42.1 try with/without block_delete
; interp_convert.cc CHKS fabs cos A_ang

g17 (xy plane)
g21 (mm)
g40 (cancel cutter radius compensation)
g49 (cancel tool lengthoffset)
g90 (absolute distance mode)
g94 (units/min feedrate)
g54 (Coordinate system 1 default)

o1 if [ #5400 NE 0]
(debug, use no tool for this test: g42.1)
m2
o1 endif

f 100
g0 x 5.0005 y 5.0005
g0 x 5.0005 y 0.0000
/ g42.1 D 10
g2 x 0.0000 y 5.0005 i 0.0000 j 5.0005
g40
m2

---

msg: "Tool radius not less than arc radius with comp"
debug output using debug print statements below:
emc/rs274ngc/interp_arc.cc:119 arc_radius=5.000500 tool_radius=5.000000
ok
emc/rs274ngc/interp_convert.cc:695 A_ang=-1.570696 cos=0.000100 fabs(cos)=0.000100 TOLERANCE_EQUAL=0.000100
(fails)

---

example debug print statements:
diff --git a/src/emc/rs274ngc/interp_arc.cc b/src/emc/rs274ngc/interp_arc.cc
index 5011430..4d3bbdc 100644
--- a/src/emc/rs274ngc/interp_arc.cc
+++ b/src/emc/rs274ngc/interp_arc.cc
@@ -116,9 +116,11 @@ int Interp::arc_data_comp_ijk(int move, //!<either G_2 (cw arc) or G_3 (ccw arc
a, end_x, b, end_y, arc_radius, radius2,
abs_err, rel_err*100);

+fprintf(stdout,"%s:%d arc_radius=%f tool_radius=%f\n",FILE,LINE ,arc_radius,tool_radius);
CHKS(((arc_radius <= tool_radius) && (((side == LEFT) && (move == G_3)) ||
((side == RIGHT) && (move == G_2)))),
NCE_TOOL_RADIUS_NOT_LESS_THAN_ARC_RADIUS_WITH_COMP);
+fprintf(stdout,"ok\n");

/ This catches an arc too small for the tool, also /
if (move == G_2)
diff --git a/src/emc/rs274ngc/interp_convert.cc b/src/emc/rs274ngc/interp_convert.cc
index 4319334..d40d589 100644
--- a/src/emc/rs274ngc/interp_convert.cc
+++ b/src/emc/rs274ngc/interp_convert.cc
@@ -692,7 +692,9 @@ int Interp::convert_arc_comp1(int move, //!< either G_2 (cw arc) or G_3 (ccw ar
double AB_ang = atan2(center_y - end_y, center_x - end_x);
double A_ang = atan2(cy - end_y, cx - end_x) - AB_ang;

[Dewey Garrett]

sourceforge changed _FILE_,_LINE_ in fprintf above

[Sebastian Kuzminsky]

The arcfail.ngc program needs "g91.1 (relative arc centers)" in the preamble.

[Sebastian Kuzminsky]

Chris and I discussed this on IRC some: http://psha.org.ru/irc/%23emc-devel/2014-01-28.html#00:39:02

We both came to the conclusion that arcfail.ngc is supposed to be rejected by the current interpreter because the compensated arc is smaller than the smallest allowed arc: the arc radius is below the minimum arc radius of 0.001 mm. (See http://www.linuxcnc.org/docs/2.5/html/gcode/gcode.html#_center_format_arcs )

Chris pointed out that the error message could be much better, and that there may be an improvement possible that would allow the program to be accepted.

[Chris Radek]

quoted from the IRC log: a separate check for this case that gives an error like "Arc entry move is degenerate because arc and tool radii are nearly the same", or a special case that lets it decay into a line, would be the two possible fixes I can picture

I'm leaving this open but postponing. Possibly we can fix it in a later branch, one of these two ways.