Three-axis motion with GRBL

For an upcoming project I want to get three axis motion working with linear stages which use a stepper motor and a trapezoidal thread to move in a very accurate way.

Stepper Stages

photo: ebay listing

I used three of these stages from celinahotoin on ebay. They have a 200-step-per-revolution NEMA11¹ motor and a Tr6x2(P1) trapezoidal thread. The thread has a 1mm pitch but a “double start” so it moves 2mm per revolution resulting in 100 steps per mm (not 200 like the listing said).

As shipped the bearing end plates were a little misaligned on two out of the three stages, causing them to jam near the far end of travel. To fix this, just slightly loosen the two screws on the end plate, send the platform out to ~1mm from the end of its travel and then tighten the end plate screws again. No more problem!

The actual length of travel is about 103mm. I’ll post some accurate measurements later when I write about 3D printing a chassis for these.

GRBL board

I combined them with this GRBL driver board from lunyee on ebay.

In the photos the board is identified as LY-GRBL3-10086-V11 but the board as shipped is identified as LY-3Axis-4.0-V2.0. It is otherwise the same so far as I can see.

photos: ebay listing

USB configuration

Plugged into a USB-C port, it shows up as a USB device with descriptors:

  bDeviceClass            2 Communications
  bDeviceSubClass         0 [unknown]
  idVendor           0x0483 STMicroelectronics
  idProduct          0x5740 Virtual COM Port
  bcdDevice            2.00
  iManufacturer           1 tomeko net
  iProduct                2 LUNYEE_4axis_Control

… and supporting a CDC data interface². In Linux it gets recognized as a serial adapter and appears as /dev/ttyACM0 (etc).

GRBL build

The $I command reveals it is running a custom build of GRBL 1.1f:

Monport
[VER:1.1f.20230316:������������������������]
[OPT:VMZHL,35,254]

some � omitted for clarity

CPU

This particular board uses a Gigadevice ARM Cortex CPU, probably a GD32F303VCT6 but I’m damned if I can read the inscription on this one even under a microscope.

I probably should have found an ESP32 based board if I’d wanted to do some software development, but this one was cheap and ticks all the boxes. If I get around to implementing polynominal trajectories I’ll probably make up my own board based on an Raspberry Pi Pico / RP2040 or similar.

Stepper Drivers

The board comes pre-populated with three A4988 drivers. They have little heatsinks on top already and have current limit trimpots which need to be set up to suit your stepper motors, which means turning the trimpots gently to achieve a V_REF which corresponds to a healthy maximum current for your specific driver modules and stepper motors.

As shipped the board is set up to microstep where each microstep is 1/16th of a full step. Since my linear stages have 100 steps per mm this works out to:

Microstepping is not without its drawbacks, eg: reduced torque and repeatability, so we might end up going with 1/4 or 1/8 steps instead. This is controlled by the stepper driver jumpers hidden under the drivers themselves.

GRBL Settings

GRBL keeps a bunch of settings in NVRAM and these are the ones I changed:

By keeping the limit switch activation point about 1mm from the end of travel and reducing homing pull-off to 1mm ($27=1) the stage limits can be set to a very neat 0 - 100 mm.

For some reason the Z axis is inverted in hardware. I checked my stepper wiring many times and it just is. So I invert X and Y ($3=3) to match, so that 0 is with the stage at the stepper end and 100 is with the stage at the far end. Homing direction is also inverted ($23=7), so the homing cycle travels to the negative X, Y and Z direction.

In my case the Z axis is just like the others so I’ve set its max rate to be the same as that of X and Y ($112=2000).

Axis Connectors

Each axis has it’s own JST XH2.5 with the following pinout. Except Y2 which is a reversed pinout of Y1. The steppers had bare wires and I had some spare pre-wired headers with Grey/Blue/Purple/White wires, so I just made up some frankencables.

Using motors with pre-installed connectors would save a lot of messing about with solder and heatshrink.

Limit / Home switches

When the board first wakes up, it doesn’t know where the stepper stages have been left. So it has to perform a homing cycle, and to do that it needs³ some switches to tell it when each axis is ‘home’. Optionally you can also provide limit switches for the other end of travel, but in this case I’m happy to trust the steppers to count correctly and use soft limits ($20=1, $130=100, $131=100, $132=100).

The home and limit switches wire to two-pin “dupont” (aka “pin header”) connectors on the board. For now I’ve just used microswitches in a 3d printed plastic carrier which clips to the stepper, eventually the microswitches will get housed in a less ugly way.

Board Power

The stepper drivers are powered from the DC connector, which is 5.5mm OD / 2.5mm ID⁴ It is center positive and expects 7 - 36V.
I found a 19V 4A supply to suit. Don’t forget to turn on the switch!

The CPU is, or at least can be, powered from the USB-C port while the motor power is off. There are a number of other power connectors on the board whose purpose is not entirely clear but are presumably for spindles, lasers, etc. I’m hoping to repurpose one of these for variable lighting power.

There are three red power LEDs on the board which might be helpful to know about:

• A: Motor Power: labelled PWR
• B: USB-C Power: no label.
• C: Either Motor or USB-C Power: labelled 3V3

Does it work?

Yes! The three axis can be controlled independently or together using G-Code commands. Note that GRBL starts a homing cycle with $H instead of the more common G28, and depending on build options it might not let you do much until you run a homing cycle.

Next Steps

• printing a case for the board and some parts to hold the stages together and more elegant end stop switch holders.
• implementing smoother kinematics by sending way too much Gcode.
• work out how the 12V “laser pwm” port works.