Niggling problem with my DIY CNC

[stugrey]

Hi everyone, I hope this is the right place for this thread and that you can help me. I've built a MDF machine based on the buildyourcnc.com designs and using the hobbyCNC controller and motors. I use an old PC with Ubuntu and EMC2 installed.

Everything has been progressing nicely and I've slowly been eliminating any wobbles etc. in the machine. Yesterday I tried to cut out a slot within a slot, i.e. 3mm depth by 27mm width then another 7mm depth but only 22mm wide. The deeper slot was nested inside the shallow one.

The results really baffle me! In the X axis it is perfectly straight but in the Y axis it seems to drift by a certain amount as it gets deeper resulting in a stepped profile at one end and the inner slot breaking out the outer slot at the other end.

I initially thought that maybe it was missing steps or something so reduced the steps per second from the the PC by half. The only effect this had was to change the direction of the stepped profile (in the first try the stepped profile was at the -ve Y end and in the second it was at the +ve Y end.

I've uploaded a picture that might help show what I mean:
http://www.flickr.com/photos/stugrey/3077155008/

Many thanks

Stu

[CarveOne]

I am using the Ubuntu and EMC2 as well. If you will post your g-code file I will cut it on my machine and see if there is something in the code doing it.

In AXIS simulation mode you can click on each line of g-code and AXIS will highlight each line segment on screen. If each step you see is a represented as a unique segment on screen it is because the steps are built into the g-code.

Also check the Z values to see if it changes value at all in the areas that have the steps. If these Z values are not constant the cutter will move vertically.

If no step segments are seen on screen then there is another mechanical or electrical reason for the steps.


CarveOne

[stugrey]

Thanks for replying. I haven't tried simulating it yet but here is the G-code, let me know what you think. (had to remane it from .ngc to .txt)

Cheers

Stu

[grimy]

If I can jump in here and offer a test posibly. I was looking at your nc code and simulated it, it simulates ok but I did notice that rapids are being used, perhaps if you change the rapids from G00 to G01 as a test to see if the steppers are dropping steps during the rapids?

[springlakecnc]

Dude, if your machine has a head that slides up and down VS. a quill, your head is out of tram.

[stugrey]

Sorry springlakecnc, I don't really understand what you are saying, I'm new to all this. In my machine my router is jammed into a hole in a platform jutting from the z axis.

grimy, I'll definitely try changing the rapids to G01 to see what happens. Problem is now the power switch for my controller has started sparking and smoking so I cant test it out!

[CarveOne]

I looked at the file in Axis simulation and see equally spaced layers for each Z step down. I don't think your problem is in the gcode file. It's 23*F in the workshop right now so I'll have to look at it again after work today when it's a little warmer out there.

The problem steps all appear on just one end of the pocket in your photo. Strange. Seems to be related to moving the cutter in only one direction in the material.

CarveOne

[Miguel-]

It might be a resonance problem with Y axis. I had a belt driven axis with steppers and no microstepping that had this problem. At high speeds everything was ok but when driving certain speeds the resonance occurred and the motor missed steps (even when there was no load on system).

Nowadays I have a 1/8 microstepping and the problem is gone (actually it went away wit 1/2 microstepping but it goes smoother wit /8 ).

Oh yea -the g-code was fine

[springlakecnc]

1) Draw a horizontal line on paper
2) Draw 3 more horizontal lines below it perfectly parallel to it about 1/4" apart
2) Now draw a vertical line, through the middle of the horizontal lines, only off a few degrees, like maybe 10 degrees from being perfectly perpendicular, like an 80 degree angle vs a perfect 90 degree angle to the horizontal lines.

The vertical line is your z axis. The tangency point of each intersecting line point is NOT directly above the other, making steps

Your up and down axis is not square with the x&y axis! Get it?

One end will show steps, the other end will not, as in your picture!

Hope this helps you out,

Buck,
Spring Lake cnc llc

[stugrey]

Everybody, thanks so much for your suggestions, they have been a lot of help. after many re-runs/repeats I think i've solved it.

In the config files for EMC2 where i've been setting the accelerations etc there is also a setting called STEPGEN_MAXACCEL, in the file it states:

# NOTE: the step generator module applies its own limits to
# acceleration and velocity. We have discovered that it needs
# to have a little "headroom" over the accel by the trajectory
# planner, otherwise it can fall slightly behind during accel
# and later overshoot as it catches up. In the long term we
# hope to come up with a clean fix for this problem. In the
# meantime, please set STEPGEN_MAXACCEL below to a few percent
# higher than the regular acceleration limit MAX_ACCELERATION

So up till now i've had it at ~10% over max accel. I dropped the accels etc to try and help and also dropped the difference to about ~3% and ran the code. The result was much worse than the first time! I then changed the difference to ~30% and ran it and it is nearly perfect. I guess "a few percent" in the note doesn't mean 3%, it means %30%!

[CarveOne]

That's cool. Good to know also. It's already too cold here to do anything in the workshop this evening for you anyway. Glad you have resolved it.

CarveOne