Author Topic: Ideas wanted on proving/testing a CNC machine  (Read 2973 times)

Offline sparky961

  • Hero Member
  • *****
  • Posts: 844
  • Country: ca
Ideas wanted on proving/testing a CNC machine
« on: May 20, 2016, 05:15:07 PM »
So, I know I haven't been posting progress on my CNC machine build, but honestly most of the stuff I've been doing is wiring and tuning.  That doesn't lend itself to a really exciting thread with lots of pictures.  That said, it's been going pretty well and I haven't had to ask many more questions on the LinuxCNC forum to keep me going lately.

I have the X, Y and A axes under servo control and from my simple testing and setup I'm repeatable to within a few thousandths on each of the linear axes and many fractions of a degree better than I need to be with the rotary table.  What I'm wondering from you guys is if you've come across or seen any good, methodical test procedures for initially validating a machine.  Perhaps certain shapes and patterns to cut that test specific characteristics, or loops to check repeatability and such.

Just a heads up though before you suggest some awesome tests that will check within a few millionths reliably.... I do still have considerable mechanical issues that will be sorted out along the way.  I know there's a ton of backlash in X and Y (probably on the order of 0.020 to 0.050 on a bad day), and a little bit in A.  The Z axis, which I have yet to set up, suffers from being a worm drive rack with a clock spring counterbalance (like a drill press but with worm drive on the pinion).  This isn't likely to change but I might tighten up the spring WAY in the future or maybe something else creative.  All of this means I won't be taking heavy cuts that rely on the rigidity of the machine and anything that would test the machine's ability to cut in both directions at this stage is pointless.  It will have to be treated as though it's an old worn out machine making cuts only when the backlash is taken up.  I will be measuring the backlash and adding it as a parameter but I don't expect to have good results changing directions during a cut.  I should get acceptable results changing directions and coming back into a cut though.  I have future plans for anti-backlash nuts, or maybe ball screws if I'm so inclined.

Offline Swarfing

  • Sr. Member
  • ****
  • Posts: 417
  • Country: gb
Re: Ideas wanted on proving/testing a CNC machine
« Reply #1 on: May 20, 2016, 05:57:19 PM »
Sparky my last machine I fine tuned by doing series of squares and circles at a given distance (circle in a circle for example). Over a given distance you can work out the percentage it is out, then apply that percentage differenCE to your tuned numbers for the given axis. To measure the backlash ask an axis to go X measurement in one direction then back again. Set up a peck drill MOP with a pen in the collet. See how far out it is and add the detail to the compensation setup (don't know what control software you are using?).

Hope this helps?
Once in hole stop digging.

Offline sparky961

  • Hero Member
  • *****
  • Posts: 844
  • Country: ca
Re: Ideas wanted on proving/testing a CNC machine
« Reply #2 on: May 20, 2016, 06:16:05 PM »
It was so obvious to me that I didn't mention it, but I will here for those of you who aren't familiar with my setup. 

This is a combo mill/drill/lathe machine, now running LinuxCNC (or working toward it anyway).  I'm starting out with the 4-axis mill configuration, then when I'm happy with it I'll duplicate and modify for a lathe setup with proper axis names, the ability to work with diameter values, and any applicable G-code differences for lathe.

Offline efrench

  • Full Member
  • ***
  • Posts: 213
Re: Ideas wanted on proving/testing a CNC machine
« Reply #3 on: May 20, 2016, 10:19:20 PM »
One way to see if you're really making circles is to drill a hole through the workpiece into the spoilboard. Put a tight fitting pin in the hole. Mill a set of circular grooves (from close to the pin to as far away from the pin as possible). Rotate the workpiece 90 degrees.  Run a sharpie (or use Dykem) around the edges of all of the grooves.  Rerun the gcode.   I'm sure with a little math you should be able to determine any errors to a few microns :)