While calibrating and measuring the backlash on our Phlat3/CNCUSB combo, I encountered an issue with the settings not updating/storing backlash values.

Since the values were so small, I was unable to visually see that the backlash compensation was not working. It was only after several iterations of setting, measuring and re-checking the mechanical set-up that I realized that the inputted values were having no effect. And yes, I had the newest software/firmware installed.

I decided to start fresh using the Phlat settings that Mark supplied for the new rollers.
After loading the settings file, I set the backlash compensation for 0.5” so that I could visually check that it indeed was working, (it was). I then re-did the calibration, measured for backlash, set and re-measured the machine and was generally happy with the measured results.

Dynamic set-up:

The foam & plywood parts that I am cutting are stacked onto each other into a group, and then the group is mounted to an aluminum part that has the same profile shape and bolt-hole pattern. The aluminum part was water-jet cut and it’s pretty close to perfect in regard to my original CAD drawing.

My priority is to get our current prototype back in the air so I adjusted the Phlatprinters settings just enough to get the foam and wood parts to match up to the aluminum part, (good enough for now).
All of aforementioned parts match, and they match the aluminum part. But the tool does not come back to the “zero” start point.

I am not going to tout myself as some CNC set-up expert, but this isn’t my first barbeque. My initial take on what I have encountered is that it is NOT entirely a mechanical issue; I think it is software related.

The more I think about it the more it hurts my head.
This is essentially an open loop system, its counting steps to figure out where it is. If the system is indeed having mechanical slippage/slop then the START & FINISH points of a complex profile shape should not line up, period. - Not unless it was a closed-loop system that could determine where it was and try to compensate.

P.S. - Another thing I discovered was that when I zero out the “Y” axis the “Z” axis value will change –quite annoying.

 

I have had similar experiences with my machine. I am running the CNCUSB combo on my machine as well using Marks settings file, although I am working mostly with EPP foam and 1/16 plywood. I attributed my home return errors to slight movement of the materials I'm cutting on the drive rollers. All of my cut parts match up perfectly from sheet to sheet so I'm really happy with the accuracy and repeatability of the machine. I do notice the home returns are randomly off. Sometimes they hit spot on, sometimes the home return will be off by 1/4 - 3/8 inch.

I hadn't noticed the backlash issue you're describing. I had added .015 backlash to my X axis & .012 to my Y axis a while back. It didn't seem to help the square edges I was seeing on the 1/4" circles I have in some of my parts layouts. Last week after doing a hard drive change in my computer I use to run my machine I did a software reload, imported Mark's settings and ran a trial cut. Near perfect 1/4" circles without even doing a calibration check. This leads me to think there could be something not yet quite right with the backlash software.

 

I tried with very low backlash value and it works fine. What do you mean that setting is not stored? If your tool does not return to start point then you're probably loosing steps. Most common reason is, that "Jog Decelerate" setting in not checked and machine is loosing steps when you jog.
If you're using version 20110204 then backlash should work.

When you zero "Y" axis, "Z" axis value should not change. What exactly are you doing?

 

Thanks for the quick diagnostic Kroko! I'll take another look at my settings file when I get back home this evening. My jog decelerate setting may very well be the issue. Thanks.

 

Hi Kroko,
I have the version you mention.
There were times when I would input a backlash value, export the settings file and then proceed to re-measure with the new backlash value. Even though the settings file would show the value, the machine would not perform the backlash.
I got the backlash to work by starting over with the basic settings file that Mark provided with the new rollers.
Parts are cut pretty well, when the tool is returned to 0,0,0 it might be off some of the time. It is odd to me that the parts are being cut OK, but it does not return to 0,0,0.
Regarding the “Z” axis change..
Let’s say I have zeroed out all three axes, and then I decide to re-position my “X” & “Y”.
– If I jog the “Z” plus one inch, (as to not scratch the material) then move the “X” & “Y” to their new positions, when I individually zero the “X” axis just the “X” axis value will be re-set to 0.
When I individually zero the “Y” axis the “Y” axis value will be re-set to zero AND the “Z” axis will change from +1” to some random plus or minus number. ** I should have noted that ONLY the value of the "Z" axis changes, the "Z" axis does not move to a random coordinate value.

 

When machine has freeplay it is important that you always approach position from same direction. Some issues that you're having can be because of this.
Perhaps someone can explain this better than me.

 

Can you try to cut version with backlash again but with very very low speed.

 

Thank you. I'll do some tests now.
This really looks very strange.

 

I changed test g-code a little. I removed pockets. Can you try to cut this with normal speed and backlash active.

Attached files BacklashCalibrationTool-NoPocket.cnc (19.7 KB)Â

 

I will find what is wrong. I hope soon.
Thank you for all tests you made. It is great help.

 

Here is experimental version that might solve the issue:
http://www.planet-cnc.com/files/Files20-20110612_BETA.zip

I can not test it myself on real machine but I've done a lot of simulations.

Please try to cut original test pattern and see if problem is solved. Fingers crossed.

Thank you
Andrej

 

So, what happened next?
Did the beta code make a difference?
Will Lisa marry Bob?
Did the network run out of money and cancel the show?