Large Tolerance Changes - Small vs Large cuts. HELP!

Hello Friends,

I’m having an issue with my machine…

I cannot get an accurate cut between two different size squares. I adjust for one size, and then cut the other and the tolerances are off.

  1. I start by cutting a 10mm x 10mm square
  2. I measure and find that the caliper reads 9.5mm x 9.5mm
  3. I adjust my steps/mm
  4. I cut a 10mm x 10mm square
  5. I measure and find that its 10mm x 10mm… Great!!
  6. I cut a 20mm x 20mm square
  7. I measure and find that its 20.4mm x 20.4mm… wtf??

I’m not sure what is happening here. I’ve tried adjusting belt tension, wheels, feeds and speeds, etc… but no luck. if i calibrate for the larger square, and then cut the smaller square, the smaller square is too small.

I cannot seem to get consistency between different sized cuts.

Any advise is welcome.


You’ll also need to adjust for each endmill (diameter, geometry, material, coating, stickout, etc), material you’re cutting, deflection, etc.

Don’t drive yourself nuts calibrating.
Nature of the machines.
Calibrate for your need, and adjust your CAD design to accommodate any dimensional issues.

What are you making?

1 Like

10x10mm is very small, any backlash, runout or tool diameter variations (they rarely are nominal value) or mechanical deflection factor will be large.

Make a test pattern, carve a 100x100mm pocket and another 100x100 outside cut.
Do a couple of them, at various feed/carve rate.

As Neil says, dial your machine in to meet your needs. We all have to :wink:

Thanks for your feedback. I’ll calibrate for the 100 x 100 and make adjustments to my solidworks model accordingly moving forward.

To answer your question, I’m working on a robotics project that requires a vertical cosine wave. I’m using HDPE, and a 2mm flat tip 2 flute bit.

I have a Meterstick, and i do 700mm long test patterns. .2mm over that range is close enough for me.

Can you share more details? Now I’m curious.

I just got some new belts in… polyurethane w 4 strand steel core. I’m rigging it up tmrw for Labor Day :slight_smile: . I’ll let you know if there’s an improvement. I’m hoping it alleviates some of the stretch from the neoprene stock belts

I can’t really, sry man. Give it a year when that full patent goes thru, then I can show you. I’m just knocking out the full prototype right now on a DIY build CNC.

1 Like

Now I’m REALLY curious. Please share when you can.

Will do… here’s my custom build. Dubbed Goliath


So I put the new belts on

IMO way better that the stock neoprene GT2.

I was able to use my calculated steps/mm (320 steps/mm X&Y)

I got 99.7mm x 99.7mm for a 100mm cube on both Axes. And 9.7mm x 9.7mm on a 10x10mm cube for both axes.

With the neoprene I was at 335/330 steps/mm to compensate for the stretch.

1 Like



Are those the steel wire belts?
What controller?

It’s a leadshine 3 axis 1/64 microstep. I have a 4th single axis hooked up for 3D print/4th axis projects tho. Running the whole thing off a arduino mega for simplicity… low bit/clock speed on the 2560 processor, but it does the job. Works with the mega grbl branch. The uno doesn’t have enough GPIO for what I need.

Yea the belts are steel core. It’s linked above.

I use the 9mm steel core on my machines. I don’t need a link, but I don’t think your link is correct.

Try that one… I have to use the 6mm the way I have it setup. I have the belts running under the wheels in the v-slot. I’d like to use a larger width like yours, but it would require significant mods for the idler wheels to mount on the gantry

If/when I do a tear down/ rebuild, I’d also add closed loop steppers and probably design my own controller. But I need to focus on getting the prototype for the patent first…its easy to get distracted with other projects

Right now with the new belts, I’m getting a repeated 0.3mm under size. With some tuning I can probably cut that in half. And then just account for the undersize in my cad model.

1 Like

Interesting, using the v wheels as idlers as well?
I would be concerned that the belt is lacking sufficient engagement with the pulley.

Seems to be doing fine. After adding 0.3mm to my cuts in both X and Y, I’m getting 0.05mm error.

I’m trying to get a clean sinusoidal wave, but I’m having issues. I can get clean cuts on the HDPE in EASEL with my calibration cuts, but when I switch to meshcam and upload my gcode, the finish pass produces fuzzies b/c the chip load changes. If I hook up my VFD to the controller, will meshcam maintain the same chipload by reducing the rpms on the finish path?

I’m tempted to manually code the sine wave. Anyone have experience with programming a parametric move directly?

Chipload is calculated without regard to the DOC. Is your finish pass just removing less material axially?
You should be able to control a vfd with a PWM signal from the controller. Are you setting the rpm manually now?
What are you using to do the 3d design?

Thanks for the reply.

The finish pass is still taking some meat off, but yea, it’s removing far less material than the rough pass, hence the fuzzy/gumming. When I run the rough pass, the problem is that it creates steps in the wave, cutting each wave dip separately. I would like to have it run a continuous motion, and slowly increment the amplitude of the wave. I.e I have a sine wave that’s got a 6mm amplitude , and a 10mm period. If I’m going by my current settings which cut the HDPE cleanly, the router would follow the same period, and increment the amplitude 0.5mm each pass.

I’m using solidworks 2018 for the CAD design, and setting my bldc spindle driver to a rpm setpoint manually

My issue is that even if I get the VFD hooked up to my controller, I want a very high res, which means a uber low step in the wave. I don’t think I can run my spindle slow enough to account for that chip load.

Do you know of a cam program that will run a spline cut for the z axis? Or is there a way to set this up in meshcam?

My backup is to program it manually…

Without really knowing what the design looks like my recommendation would be to use Fusion 360 for the CAM. You’ll have much more control than Meshcam. With a ball end mill, you should get really smooth contours.