[JayCop]

When cutting a 5"x5" square my machine consistanly cuts two sides on the Y axis perfectly at 5" and 5 1/16" on the X axis sides. What could be causing the x-axis to repeatedly go over by 1/16"? When each axis is tested individually with a 1" dial indicator both are within a + or - thousandth or two. The repeatability seems to be very good.

I am using CNCRouterParts Rack and Pinion on both x and Y axis. The X axis is dual driven by 420oz steppers. My power supply is 36V 8.8amp. I am also using a gecko 510 board.

Any help will be greatly appreciated.

[PaulRowntree]

What happens with larger and smaller shape dimensions along X?
Does the gcode look right?

[ger21]

What if you cut it in different places on the machine?

[JayCop]

I will test some more sizes tomorrow, and on different places on the machine. I did try some smaller sizes but did not measure them and they are at my shop.

These were cut toward the middle of the Y axis towards the - direction of the X axis but not close to the softlimits.

I will attach the gcode tomorrow when I pull it off my machine computer. The dxf file was made with sketchup and the gcode with vectric cut2D.

[flojor10]

You should try backlash compensation, using this my new machine gets +/_ .0005in everytime.

[JayCop]

What backlash settettings did you use?

[JayCop]

I tested a 10" square and got a different result with the x-axis being short about 1/16". I have attached the g-code to both of the files I cut, as well as a picture of the 10" square with measurements.

[PaulRowntree]

I see that the gcode is using arcs to round the corners of the 10" 'square'. Could this be interfering with your physical measurement of the 'square's ' dimensions?

1/16" seems too large for backlash, but having a constant error on one axis that does not scale with piece size sure looks like backlash.

[JayCop]

The interesting thing is that on the 10" square it is short 1/16" and on the 5" square it was over sized 1/16". I cut the 10" considerable slower than the 5" one and turned my acceleration on the x-axis down to 10.

How can I stop cut2d from rounding the corner? The entry point corner is the only one that looked rounded when I cut it.

[PaulRowntree]

The interesting thing is that on the 10" square it is short 1/16" and on the 5" square it was over sized 1/16". I cut the 10" considerable slower than the 5" one and turned my acceleration on the x-axis down to 10.

How can I stop cut2d from rounding the corner? The entry point corner is the only one that looked rounded when I cut it.

Sorry, I missed that the 1/16" was short on one, long on another. And I also missed that you are cutting a profile outside a square, so the arcs pivot at the corner of the 10x10, leaving a squarer corner than the toolpath suggests without considering the tool as well. Actually, turning down the acceleration can cause corner problems if in CV mode, because the toolpath has to start turning sooner in order to keep the velocity constant, thus rounding it off.

[JayCop]

Should I turn off CV, or turn up accel?

[PaulRowntree]

I always have CV on, and look to the corners to see if it is rounding. To avoid problems, have higher acc (I have seen suggestions to look for the max acc that you machine can support without loosing steps, then reduce it by 10-20% for safety). To see if CV is a problem, just turn it off and look for differences in the quality of the cut (the machine will jerk around more)
CV problems are always at the corners. Large shapes should have the flats at the right positions even if there are CV problems.

Are the X and Y edges of the cuts at truly 90 degrees, or is it skewed at all?

[Senna]

If your math for the X axis steps per inch is correct... then I would guess that your X axis acceleration is set too low. Here's information I received from one of the 'guru's' with regard to acceleration when I was setting up my machine...

The acceleration setting effects how the steppers stop or decelerate. An acceleration setting that is to low will cause coasting down of the axis and the machine to overshoot where it should have stopped. It can also cause the machine to overrun home or limit switches. A general rule is to use between 10% and 15% of the velocity as a starting set point for the acceleration setting. A good visual indication would be to snap "rapidly release" an axis jog key during a jog... that axis should stop motion instantly.

It sounds to me you have the Y axis accel set in the correct range but your X axis accel is set too low and you are consistently coasting down and overshooting that 1/16".

Senna

[PaulRowntree]

The acceleration setting effects how the steppers stop or decelerate. An acceleration setting that is to low will cause coasting down of the axis and the machine to overshoot where it should have stopped. It can also cause the machine to overrun home or limit switches.
Senna

Low acc will cause overshoot when manually jogging (and is scary to see when your gantry is rushing towards an immovable object...), because the controller (Mach3?) can't predict when you are going to release the jog button. I would have thought that limit switches would stop the movement as fast as gantry momentum allows. But when running gcode, the 'look-ahead' predictor knows exactly where it is going on each axis, and does its best to get to the desired position by starting to slow down when 'the end is near'. If you try to maintain constant velocity as well when going around a corner, then the controller has to start slowing down on one axis and speeding up on another before the actual corner.

The upshot of this is that the controller may cut corners short but should never overshoot.
If there is any mechanical slop available, then overshoot may happen because of gantry inertia.

[JayCop]

My plans to try higher accl and possibly turning of CV were thwarted by my controller computers decision to stop working abruptly.

Luckily I have another computer that I just had to load XP onto, and get the settings from the old hard drive on. So I will be able to test again tomorrow.

[ger21]

It sounds to me you have the Y axis accel set in the correct range but your X axis accel is set too low and you are consistently coasting down and overshooting that 1/16".

Lowering the accel will not cause overshooting. Say you want to move 5 inches. There is a fixed number of steps required to move those 5 inches. The control will send the exact number of steps required to travel those 5 inches. No more, no less.

Either the machine is losing or gaining steps, or it's a mechanical issue.

[PaulRowntree]

My plans to try higher accl and possibly turning of CV were thwarted by my controller computers decision to stop working abruptly.

Luckily I have another computer that I just had to load XP onto, and get the settings from the old hard drive on. So I will be able to test again tomorrow.

Bad break, and it always happens when you need it to work (as if there was any other possibility).

But seriously, is it possible that your computer/interface has been marginal and has been giving inconsistent pulse trains?

[JayCop]

At this point I hope it was the computer.

[JayCop]

Tested and it is better with the higher acceleration. Unfortunately it is still a little off.

Does any one have some screen shots of the settings they have in mach3 that you may provide so I can see if I have changed something I should not have?

[JayCop]

I had an epiphany today. I went back to the drawing board and looked at the G540 documentation. I had not used used the G540 xml because well I was lazy and didn't want to have to change all of my home and limit settings. Figuring the settings I had from a previous BOB were fine.

The proof was in the pudding. Now when I cut a 10"x10" square guess what? It is 10"x10"!

Now my next little problem I need to iron out is making the square have true 90 degree corners. I am out of square by about 1/64" on my 5" long machinist square. The corners also have a slight chamfer where the next axis starts to move before the other stops which I guess could be from the acceleration not being all that high ie 20.

Does anyone have a simple way of adjusting the squareness of the axis based on the amount it is out of alignment on the cut outs?

Is there anyway to get true square outside cuts without the slight chamfer while also not killing the speed of circles and arcs?