Multi-Channel Configuration (new)

One section per channel, [CHANNEL_0] is optional (channel 0 takes [TRAJ], [RS274NGC], [EMCIO] and [DISPLAY] as usual and reads only the keys it needs from [CHANNEL_0]).

When an axis letter exists in more than one channel, the machine axes are named with a counter in channel order: with channel 0 = X Z C, channel 1 = X Z W C and channel 2 = X Y Z B the machine has X1 X2 X3, Z1 Z2 Z3, C1 C2, W, Y, B. These names are shown by the GUIs, used in the G-code axis words (X2=120, see [cha:multichannel-gcode]) and in all statements that name an axis (GETD(C1), $AA_IM[Z3], POS[W]=…​, TRANSMIT_1_ROT = C1). A letter that exists once keeps its plain word (W300, Y-18).

Channel 0 keeps all stock pin names (motion., axis.L., halui., iocontrol.0.). Channel N >= 1 adds:

The wait marks of channel 0 are motion.wait-mark-active, motion.wait-mark-id, motion.wait-mark-set, motion.wait-mark-abort-partners.

Every joint has the input joint.J.feed-enable, every axis letter axis.L.feed-enable, both TRUE by default (an unconnected pin changes nothing). The PLC pulls a pin FALSE to lock the joint or axis against operator errors, for example Y and Z while a tool change is running:

Homing is not affected by the feed enable. The DMU 60 sim interlocks the head and the gripper arm with two PLC cams (next section): net greifer-ruhe plccam.greifer-ruhe.out ⇒ joint.1.feed-enable joint.2.feed-enable locks Y and Z while the arm is away from its rest, and net z-wechselpos plccam.z-wechselpos.out ⇒ joint.6.feed-enable motion.digital-in-02 locks the arm unless Z is at the tool change position (approached with the tool length offset cancelled, G49 before M6). The changer program does not rely on the lock alone: it reads the Z cam on the digital input and waits for it (WHILE $A_IN[2] == 0) before every swing of the arm.

A cam is a range on an axis: its output is TRUE while the position lies inside, for enables such as "gripper open/close only in the working area" or the feed enable above. The cams are declared in the INI file, one section per cam numbered from 0 without a gap:

The HAL file loads them, after the servo thread exists, with

which loads the real-time component plccam with the names plccam.<name>, adds them to the thread, sets the range and links the position. The output plccam.<name>.out is TRUE while MIN ⇐ position ⇐ MAX; once inside, it stays TRUE until the position has left the range by more than HYST (no chatter on the edge); INVERT = 1 turns it round; .inside is the output without the inversion. The pins .min, .max, .hyst and .invert can also be set from the PLC at run time. The component is documented in plccam(9).

Channel 0’s machine on enables the machine as in LinuxCNC. The task of a channel >= 1 switches only its own joints on and off: the joints of a channel that is off are unpowered (joint.J.amp-enable-out low), follow their feedback and refuse jogs and programs; its task reports ON only while its joints are powered. When the machine goes off (channel 0 off, estop, fault) every channel needs its own switch-on again. Homing a joint of a switched-off channel is refused and home all needs every channel on.

The simplest way to run the machine is one key for everything: mchmi exports mchmi.machine-on, mchmi.machine-off, mchmi.estop-activate, mchmi.estop-reset (one pulse per key press) and mchmi.feed-override (the feed pot in percent), and a netted pin carries its key alone. The G200 configuration (index_g200_postgui.hal) nets them to the halui.N.* pins of every channel:

Take a channel out of a net, or gate it with an input, when its area has to be switched off alone (tool change, robot loading); split the feed net when the hardware has separate pots.

Every channel has its own tool table or tool management. With DB_PROGRAM = mctooldb ([EMCIO] or [CHANNEL_n]) the program mctooldb(1) serves the tools of the channel’s carrier from the file TOOL_DB (one tool per line, key=value pairs): tool number t with sister tools st, station, geometry, and tool life monitoring by piece count or time with set point, actual value, pre-warning and disabled. An exhausted tool is replaced by its sister at the next tool change, the program statements SETPIECE(n) and TOOLSET(id, key, value) and the named parameters #<_tool[id]key> give the program access (see [cha:multichannel-gcode]). mchmi edits the files in its WZ-Liste area.

As on the SINUMERIK a channel can run programs outside the part program: asynchronous subprograms on request and an event program at defined events. Per channel in the INI ([CHANNEL_n], a channel 0 may use [RS274NGC]):

The file names are looked up next to the INI file, then in [DISPLAY]PROGRAM_PREFIX. An ASUP is started by a rising edge on the HAL pin iocontrol.N.asup-<k>, by linuxcnc.command(channel=N).asup(k) or by the panel; iocontrol.N.asup-active and stat.asup show the running program (1 to 8, 100 plus the event number for the event program, 0 none). The request is served while the channel is idle: the task remembers mode and open program, switches to AUTO, runs the file from its first line and, when it ends (M2, M30) or is reset, opens the previous program again and restores the mode. A request while a program runs interrupts it as on the SINUMERIK: the motion stops, the block the program was in is remembered, the ASUP runs; when the ASUP executed REPOSA (also REPOSL, REPOSQ, REPOSH) the program goes on at that block afterwards, read through to there like a run from that line (modal state and offsets re-established, the first move goes from where the ASUP ended straight to the block’s end; spindle and coolant stay as the ASUP left them) - without REPOSA the program stays aborted. A request with the machine off, with the axes not homed, while an MDI command runs or while another ASUP runs is refused with an error ("ASUP k: the channel is not idle"); a HAL edge that arrives while another ASUP or the event program runs is served after it (a tool change request during the retract).

The event program runs the same way at the events of PROG_EVENT_MASK: 1 before a program started with Cycle Start (the program follows when the event program is through), 2 after a program reached its end, 3 after the operator’s reset (not after the reset of an event program itself) and 4 once after the first machine-on with the axes homed. The interpreter reads the number as $P_PROG_EVENT (0 in ordinary programs and ASUPs), so one file serves every event:

A tool changer channel uses this like a cyclic PLC program: its event program retracts from an interrupted change at power-on and after a reset and then serves the tool prepare / change requests of the machining channel in an endless loop (LOOP …​ ENDLOOP polling $A_IN[n]), so the channel always runs; a reset stops it and the reset event starts it again, a request still pending is served right away (the DMU 60 example, wechsler.ngc).

linuxcnc.stat(channel=N), linuxcnc.command(channel=N) and linuxcnc.error_channel(channel=N) address a channel (the keyword is accepted by stock-compatible clients as well; without it channel 0). command.asup(k) starts the asynchronous subprogram ASUP_k of the channel, stat.asup names the running one (Asynchronous subprograms (ASUP) and event programs). linuxcnc.nmlfile must point at the configuration’s [EMC]NML_FILE. New status fields: pose_joint (joint in each pose slot, -1 = none), master_spindle, enabled, synact, ac_marker, ac_param, ac_timer, sa_r, ac_delt, sa_h_number/sa_h_value/sa_h_count, aa_mm, ac_mea, trafo_type, trafo_num, trafo_d, tcs. New commands: synact(op, id) (0 CANCEL, 1 LOCK, 2 UNLOCK, 3 RESET), set_tcs(on), set_axis_limits(axis, lo, hi) (jog limits of the channel’s slot, used by the collision guard).