Feedback on Improving Precision in XYZ Gantry Dispensing Machine

Hello everyone,

I'm new to machine design and have encountered some precision issues with my XYZ gantry setup for a dispensing machine. I am hoping to get your insights or suggestions for improvements.

Requirements:
• Precision: The machine should place a dispensing head with 0.3mm accuracy.
• Workspace: 400 x 400 x 250mm (X-Y-Z).
• Mobility: Designed to be lightweight and have a small footprint so it can easily be moved by one person.

Current Issue:
• There seems to be an X-Y error of approximately 1mm.

Design:
• Drive System:
◦ X axis: single motor, belt driven.
◦ Y axis: single motor, belt driven.
◦ Z axis: single motor, timing belt rotates a lead screw.
• Bearings:
◦ Y axis: Two 16mm parallel shafts with LMKL16UU bearings.
◦ X axis: Two parallel MGN12H linear guides, each with a single carriage.
◦ Z axis: A single MGN12H linear guide with one carriage.
• Frame: Constructed from t-slot extrusions, with machined ends to ensure controlled length and perpendicularity.
• Assembly: I used a large flat plate as an alignment jig during assembly.
• I have attached screenshots of the CAD designs and photos of the assembly to provide a clearer picture of the setup.

I think it must be a problem with the squareness of the axes to the frame. The electronics/motors are working fine and each actuator is moving the correct distance (within 0.2mm).

Potential issues:
• Frame Straightness/Flatness: The extrusions seem to have a flatness of 0.3mm or lower over their length when placed on a flat plate, but I'm
questioning if this is enough.
• Structural Rigidity: The machine is rigid however if I push one side of the Y axis forward and the other backwards I can cause the y axis assembly to
yaw. Is the separation of the Y axis shafts too large for these bearings?
• Alignment: Y axis bearings are not perfectly perpendicular; one side has about 0.3mm of clearance between the Y-axis and the endstop when homed. I don’t think this alone could explain the 1mm error however.

I'm considering replacing the aluminium extrusions with custom machined plates to enhance precision, though I was trying to avoid this to reduce weight and cost.

I have seen CNC routers on here that are using aluminium extrusions which would suggest they are suitable for precision applications. Also it seems most consumer 3D printers use these aluminium extrusions so I thought they would provide enough precision for my application.

Can you see any examples of poor design or poor component/material choice which could be responsible for the precision issue?

Any thoughts are greatly appreciated. Thank you!

Structurally, it looks OK to me. The Y axis is always going to be at risk of twisting due to its relatively thin nature vs. it's height/width. Bolting it up good & hard should mitigate that, depending on what sort of load you're purring on the Z axis tool.

You say the 1mm error is in the X-Y; but is it in the X, or the Y, or both? And over what distance do you see the error: Is it cumulative, or does it suddenly appear after a certain distance moved? e.g. if it were in the X axis over the full length of travel; if you move it all the way to the left, then all the way back to the right - does the error disappear, or is it still there? If the former, I wonder if your steppers are actually moving the distance you're expecting them to move. You could potentially dial that out in software, if it's the case. If, however, the machine doesn't repeatably go back to the same place after a long X or Y move, then it's likely to be the mechanism that's either flexing, bending, of something.

Does it always come up short? Or can it be short OR long. If the former, maybe the stepper is losing steps? That could be due to a sticky spot in the mechanism, or insufficient drive power, or running them faster than the CPU can guarantee all the steps will be processed. If it could be short or long, then again I would suspect frame flex/warp.


FWIW, I'm no expert - just thinking of possible causes.

PS: If you think the base frame is a parallelogram shape, rather than a true square/rectangle; adding precisely drilled right-angle-triangular plates underneath would eliminate this for sure.

OK so it is a sort of pick and place machine with no real loads at the end effector (dispenser end)? And the positioning errors there when you just jog it about with no load on the effector?

I would say the machine you have build should hold a much tighter position from a mechanical point of view. Yes with a single drive at the centre of the gantry and what looks like 2 unsupported rails either side with a fair span between them it will rack if you push the uprights in opposite directions, as you have discovered. So back to needing to know a bit more about the issue and how it occurs.

Motors calibration
Have you calibrated the stepper motors? You need to tell the software how many steps should be sent per mm of travel - this can be calculated based on the microsteps on the driver (are you microstepping?), the steps per degree on the stepper motor itself (what motors are they?), the pitch of the ballscrew (where used), and any ratios on the drive pulleys. Or just iterate the value via trial and error for now if that is a bit much to work out for now.

Software
What software are you running for the controller? Is it ethernet, USB or parallel port ?

Drivers
What is the driver specification and what voltage are you using?

Stalling
What speeds are you running at and might you be missing steps? Run it very slow to make sure this is not the issue. It will get to the commanded position but just take longer (for testing).

Backlash
If you manually push the effector part back and forth, side to side etc. do you feel backlash anywhere across any parts of the machine? This must be pretty close to zero, or at least way less than the accuracy you are aiming for.


Quick tests . . .
If you suspect or want to rule out machine flex then temporarily:
1. clamp the base down to securely using G-clamps or similar to a very sturdy table (create shear panel on the base and stop it moving)
2. clamp a stout piece of wood (or spare extrusion if you have it) diagonally across the gantry side legs to reduce flex sideways [red bracing]
3. stopping the racking is not easy without modifications so maybe put a DTI up to each gantry leg and ensure they are both moving the same amount

Attachment 32429