The Smell of Molten Projects in the Morning

Nets connecting Kicad components get their names in three different ways, each treated slightly differently by Kicad-to-HAL.

The simplest case happens with a Global Label attached to the net, as with the All-home signal on the right, which can be referenced anywhere in the schematic :

The XML file carries the Global Label text as the net name:

<net code="154" name="All-home">
    <node ref="joint.3.100" pin="20"/>
    <node ref="joint.2.100" pin="20"/>
    <node ref="joint.0.100" pin="20"/>
    <node ref="joint.1.100" pin="20"/>
    <node ref="dbounce.100" pin="2"/>
</net>

The order of the nodes within the net depends on which one Kicad encounters first, so Kicad-to-HAL sorts the destination references in ascending order:

net All-home <= dbounce.0.out => joint.0.home-sw-in joint.1.home-sw-in joint.2.home-sw-in joint.3.home-sw-in		# dbounce.100

The ASCII-art arrows around the HAL source reference set it off from the rest of the clutter, with the comment (off-screen right for you) giving the original Kicad source component reference.

Kicad generates a name for anonymous nets, like the one between the 5I25 GPIO pin and the debouncer, from the first component it encounters:

    <net code="617" name="Net-(dbounce.100-Pad1)">
      <node ref="hm2_5i25.0.gpio.013.100" pin="2"/>
      <node ref="dbounce.100" pin="1"/>
    </net>

Kicad-to-HAL replaces those names with a less noisy one in an ascending sequence:

net N_159 <= hm2_5i25.0.gpio.013.in_not => dbounce.0.in		# hm2_5i25.0.gpio.013.100

The anonymous Kicad net between the parameter setting the debounce delay starts with a similar name:

<net code="127" name="Net-(mux2.100-Pad1)">
    <node ref="parameter.106" pin="1"/>
    <node ref="mux2.100" pin="1"/>
</net>

Kicad-to-HAL extracts the source and destination nodes from the net to produce a HAL setp command for each destination:

setp dbounce.0.delay 3		# parameter.111

Kicad nets can get a name from a Local Label accessible only on that particular schematic sheet:

Those Kicad net names include the sheet name as a prefix:

<net code="130" name="/Jog Speed/Angular-Motion">
    <node ref="mux2.100" pin="3"/>
    <node ref="or2.112" pin="3"/>
</net>
<net code="132" name="/Jog Speed/Vel-Per-Second">
    <node ref="scale.104" pin="1"/>
    <node ref="mux2.100" pin="4"/>
</net>
<net code="138" name="/Jog Speed/Vel-Per-Minute">
    <node ref="scale.101" pin="1"/>
    <node ref="scale.104" pin="4"/>
    <node ref="scale.100" pin="1"/>
</net>

It’s not clear which characters HAL regards as valid, so KIcad-to-HAL smashes the Kicad name flat:

net _Jog_Speed_Angular-Motion <= or2.12.out => mux2.0.sel		# or2.112
net _Jog_Speed_Vel-Per-Minute <= scale.4.out => scale.0.in scale.1.in		# scale.104
net _Jog_Speed_Vel-Per-Second <= mux2.0.out => scale.4.in		# mux2.100

Kicad-to-HAL sorts the HAL net list by name to make a specific net easier to find.

The XML netlist file contains a zillion single-node nets, one for each unconnected pin in the schematic. If the Kicad net name doesn’t start with Net-, KIcad-to-HAL will include a note in the HAL file:

#* Named net with single pin: hm2_5i25.0.100 read_gpio

Prefix the pin name (when you create the symbol) with an asterisk to suppress that message. This will be most useful for HAL function pins intended for an optional thread connection:

Kicad-to-HAL does not generate HAL nets for Kicad nets with -thread in their names, as those generate addf commands.

The Mesa Electronics 5I25 FPGA card can implement a vast number of hardware I/O controllers, but I’ve only set up the few Kicad symbols required for my simple Sherline mill. The LinuxCNC hostmot2 doc gives the details.

The hm2_5i25.0 component (upper right) carries the pins and functions pertaining to the card itself:

If you have two such cards, you’ll also have an hm2_5i25.1 component. Kicad will annotate both with a zero, which Kicad-to-HAL then strips off before attaching the pin names.

The read and write functions must attach to the first and last positions of the servo-thread component, respectively, to make the thing work correctly.

The read_gpio and write_gpio functions are normally not connected; the * prefix suppresses a Kicad-to-HAL message about unconnected named pins. The code will strip the asterisk before connecting the pins, but that part is … lightly tested.

Similarly, the schematic shows the watchdog output connected to one of the on-card LEDs. I’m not sure how to test that and wouldn’t see it inside the PC case; it might work as described.

A hm2_5i25.0.gpio.000 Kicad symbol (middle) collects all of the HAL pins related to a single 5I25 GPIO pin; edit the three-digit pin identifier to match the hardware pin. Although the out and in names look backwards, they indicate the hardware direction: you write to the out pin and read from the in pin.

The Sherline uses four stepgen blocks, so that’s the only other Kicad symbol on offer:

The nets connecting those three components come directly from the LinuxCNC pncconf configuration utility. Fancier machines surely require more attention to the coefficient details.

Connecting a true parameter to the step.invert_output and direction.invert_output pins flips the polarity of those signals. The pins are aliases of the corresponding GPIO invert_output pin, with the key advantage of not requiring you to figure out which GPIO pin corresponds to the FPGA’s output pin for each axis.

The 5I25 card requires a pair of loadrt components to load the drivers and configure the FPGA:

The one-page demo schematic uses the same 5I25 configuration as the Sherline, with the LEDs driven by gamepad buttons to provide something to look at:

I should force-fit an LED into the Sherline driver box just to have something easier to see.

The LinuxCNC HALUI “component” has so many pins that I divided them among about 18 Kicad symbols, each collecting the pins related to one function:

The Kicad references (halui.mode.100) use periods as separators, not dashes, because HALUI runs automagically and has no thread functions.

The symbol name (HALUI_MODE) serves as a unique identifier that should match the reference stem. I used underscores just for pretty, but it doesn’t make any difference.

All of the Kicad symbols include the StripAnno field to remove the Kicad annotation, as described earlier. The Kicad annotations now start from 100, because I’m in the midst of tinkering an automagic way to produce the proper start-from-zero HAL numbering with everything properly sorted; this is very much a work in progress.

So the full name of the resume pin on the HALUI_PROGRAM component sporting Kicad reference halui.program.100 is just:

halui.program.resume

This corresponds to the description of that pin in the HALUI doc, without a numeric identifier: there is only one HALUI.

The net connecting the gamepad button to that pin:

net _Gamepad_Buttons_Program-Resume <= input.0.btn-base4 => halui.program.resume

A few more examples from the nets near the AND2 gate:

net _Gamepad_Buttons_E-Stop-A <= input.0.btn-top2 => and2.0.in0
net _Gamepad_Buttons_E-Stop-B <= input.0.btn-base => and2.0.in1
net _Gamepad_Buttons_E-Stop-Reset <= input.0.btn-base2 => halui.estop.reset
net _Gamepad_Buttons_Machine-On <= input.0.btn-pinkie => halui.machine.on
net _Gamepad_Buttons_Manual-Mode <= input.0.btn-base3 => halui.mode.manual

Some HALUI component references include numbers and letters identifying specific joints / axes / spindles, all of which you must edit as needed and none of which should be subjected to Kicad’s renumbering whims:

The HAL commands producing the nets connecting the AND2 gates to the HALUI pins:

net N_065 <= and2.14.out => halui.axis.x.minus halui.joint.0.minus
net N_066 <= and2.15.out => halui.axis.x.plus halui.joint.0.plus

The nets lack any specific name, so Kicad-to-HAL replaces Kicad’s default gibberish with the next integer from an ascending series. If your HAL configuration requires more than three digits worth of anonymous nets, feel free to tweak the format string.