[guru_florida]

See picture. Some of my 90 deg corners are being cut off. Circled red. Top cut where the tool does a 180 is correct. This isnt even the worst shortcut. Later it gouged about 3/4" off the mark. Cutting with a 1/4" 2Fl EndMill.

This happened before but isnt common. I cut the same part yesterday with no problems. I've cut many other parts without issues - including aluminum. The machine has no backlash that I've ever seen - parts are typically <0.5mil within tolerances.

I did notice the GUI was slow before starting. Also, during Touch Off the automatic reload of the file took 2-3 mins where it typically takes 10-20secs. Wierd!?!

First 5 step passes (50mil/pass) were fine and within tolerance, machine just got lazy suddenly.

I was in G64 Continuous Path mode. I am not sure what my P is but it is the default value. This would have to be a pretty bad P to be this bad! I will switch to G62 or at least specify a tight tolerance on P.

I didnt get any realtime errors and never have since tuning the realtime system years ago.

Any suggestions?

C

[geekinesis]

Hi
Yes I think you have to specify the p tolerance otherwise g64 on its own reverts to best speed rather than tightest accuracy in corners.
The p value will still allow for best speed but will restrict the distance the tool can move away from the path.

I use a laser so I set it for speed rather than accuracy because I need the laser to have a constant speed and I dont mind it cutting a corner slightly to achieve this

[jbeug]

I am having a similar problem with cutting circles. Sometimes they are perfectly round and sometimes they are not. It does not seem to matter if I use G64 or not.

[guru_florida]

I think your issue may be backlash, have you tested it? Are your elipses aligned with N-S, or E-W? Or are they oblong in a diagonal? Backlash tends to create circles with a bulge in one of the diagonal axis. While misconfigured EMC "axis scale" setup tends to create perfectly symmetrical elipses aligned with an axis.

Can you describe your linear bearings and screws a little more?

I am having a similar problem with cutting circles. Sometimes they are perfectly round and sometimes they are not. It does not seem to matter if I use G64 or not.

[guru_florida]

Thanks.

I am going to try again today and add the P term (I know better). I kicked myself yesterday. I was so mad that I just shut the machine down without looking at the the 3D preview output, it would have shown me if the motion planner actually cut corners.

Hi
Yes I think you have to specify the p tolerance otherwise g64 on its own reverts to best speed rather than tightest accuracy in corners.
The p value will still allow for best speed but will restrict the distance the tool can move away from the path.

I use a laser so I set it for speed rather than accuracy because I need the laser to have a constant speed and I dont mind it cutting a corner slightly to achieve this

[jbeug]

I have checked the backlash and it is minimal, only a couple thousandths, and I set backlash comp in the ini file. It does seem to have a bulge. I know the scale is correct because it moves the exact commanded distance.

I have heavy duty recirculating linear rails with ground ball screws.

I am using a pico systems universal step control board and gecko 203V drives. I wonder if the pid tuning is incorrect or the accelerations are wrong??

[guru_florida]

If you are using G203Vs then there are no PIDs since you are using steppers. If the acceleration is too high then you might be missing steps, but I am guessing you are running at a reasonable acceleration.

The backlash compensation, if not done right, would also create odd circles though I am not sure exactly what they would look like. How are you getting backlash of a few mils (thousandths) with ground ballscrews? (What's the C rating, if rated?) You should be solid. Backlash comes from the ballscrew not the slides - cross-axial slop and resulting vibration comes from the slides. So without a load your backlash should be less than 1 mil.

Checking scale by individually commanding an axis is good. So that eliminates that as the reason. Try taking the backlash compensation out and draw circles, better or worse?

I have checked the backlash and it is minimal, only a couple thousandths, and I set backlash comp in the ini file. It does seem to have a bulge. I know the scale is correct because it moves the exact commanded distance.

I have heavy duty recirculating linear rails with ground ball screws.

I am using a pico systems universal step control board and gecko 203V drives. I wonder if the pid tuning is incorrect or the accelerations are wrong??

[Andre' B]

If this is the result of g64 P settings you should see the rounded corners on the screen plot, red line.

[guru_florida]

I know. I was so angry at losing the part I shut the machine down and gave up and later thought about that.

I did do the part again last night and it worked out great, used a P of 3mils. EMC UI was back to it's normal responsiveness as well. I think something in the computer was messed up too that choked the motion planner thus it started cutting corners hard. I've done a lot of parts without a (G64) P and it worked fine. I have to replace my controller motherboard at some point.

Thanks for the input guys.

jbeug: Any luck/progress on your issues?

C

If this is the result of g64 P settings you should see the rounded corners on the screen plot, red line.

[jmelson]

I have checked the backlash and it is minimal, only a couple thousandths, and I set backlash comp in the ini file. It does seem to have a bulge. I know the scale is correct because it moves the exact commanded distance.

I have heavy duty recirculating linear rails with ground ball screws.

I am using a pico systems universal step control board and gecko 203V drives. I wonder if the pid tuning is incorrect or the accelerations are wrong??

Ok, backlash is the first thing to check, but there are a couple other things that can cause this.
First, measure diagonally (to the machine axes) across the circle. If one way is different than the other, it is a strong indication your X and Y axes are not perpendicular. the best way to fix this is to correct the machine alignment. There are other ways by using non-trivial kinematics, but that is probably more complicated.

And, of course, how not-round are they? A few thousandths of an inch? Or, a lot more?

If the PID settings are not good (Yes, with the Pico Systems USC there IS a servo loop, even if it is only between the computer and the USC's step generators) it will generally cause grinding noises from the motors. But, you can check the following error in Halscope while jogging around to see if there is any significant lag. Mostly, just plot ppmc.0.encoder.00.delta (velocity in raw step units/servo period) to see what the commanded velocity is (also have the scope trigger on this). Then also display pid.0.error and blow the vertical scale up to see the small details. Something like 200 u (micro) per division is about right. If the error is never more than the equivalent of a few steps, then everything is fine. If it is many steps, then tuning the PID will be beneficial.

You can contact me for advice on how to do that, if needed.

Jon

[guru_florida]

Well, I've been using the G64+P command and my cutting has been excellent. Looks like that was the primary issue. No probs on the computer bogging down again (which I believe was the cause emc2's motion planner couldnt do a decent precision even without the P.)

But...lightning struck and blew out the computer controller. (literally) So I am putting another together. lol