Hello Guest it is March 28, 2024, 07:47:12 PM

Author Topic: toolpath drifting during cut  (Read 6737 times)

0 Members and 1 Guest are viewing this topic.

toolpath drifting during cut
« on: March 22, 2011, 10:50:50 PM »
hi, i'm running a 2160 on a large woodworking machine with 10hp atc spindle.

i'm having a problem when cutting 3D relief files, in that the axis doing the large number of 'raster' passes slowly drifts, so the relief ends up slightly trapezoidal. The machine isn't out of square, it's just that each pass shifts a small amount from the previous one.

The overall length of each pass is correct, it's only the start and end positions that vary. I have tried generating the toolpath so that it rasters in the other direction, so the other axis is doing all the passes, and the problem occurs in the same way. The drift is always towards the 0 end of the axis.

The error seems to be consistent and linear and always in the same direction, so I'm guessing it can't be random noise. The enoders are A/A- B/B- Z/Z-, I've tried running them single ended but it made no difference. I've also tried powering them from the Galil 5V line rather than the servo amps' own supply. The fact that the same happens in both the main axes seems to suggest it's not an encoder/cable /drive problem.

If I look at the live data record in smarterm or galiltools, with the motor energised the encoder position is perfectly static- if I push against the machine as hard as I can it resists the movement, and I can see the (small) position error appear then disappear as the error is corrected, and the encoder returns to exactly the original position.

I've attached a picture- the inner rectangle is a relief cut with 0.5mm stepover, rastering in the Y axis (up/down in the picture). I also did a larger file as an 'air cut' with the spindle off (to eliminate noise from that source), and you can see the larger rectangle where the extractor brush cleared a path in the dust, showing the angle again.

[by the way, the 'choppy' edge you can see is mostly because it's a ballnose doing a relief, so it's at varying heights when it reached the end of each pass, which makes it look uneven.]

Any ideas?... I'm running out!
« Last Edit: March 22, 2011, 11:01:42 PM by steve_p »

Offline kcrouch

*
  •  193 193
  • In way too deep!!!
    • View Profile
Re: toolpath drifting during cut
« Reply #1 on: March 23, 2011, 08:11:25 AM »
Steve,
Does your machine always come back to the same position? If so it would appear that your table isn't flat. Or there is lost motion (backlash) in your table drives.
Kenny
Having way too much fun! Something must surely be wrong.
Re: toolpath drifting during cut
« Reply #2 on: March 23, 2011, 09:05:05 AM »
Steve,
Does your machine always come back to the same position? If so it would appear that your table isn't flat. Or there is lost motion (backlash) in your table drives.
Kenny

Hi Kenny,

thanks for the reply. The position drift is in the X and Y planes (it might also be in the Z but I haven't yet checked for that), so the flatness of the table shouldn't really affect anything. I did check the table height when I was levelling the machine, and measured from the spindle nose the entire 10x5 foot bed was within 0.1mm.

I can't understand why the length of each pass is exactly right (presumably meaning that the encoders are reporting position ok) but the entire pass has moved! There is no mechanical slipping in the drive train as far as I can see (pulleys are keyed)

When doing, say 500 passes at 0.5mm stepover, the axis that is doing the passes gradually shifts a total of maybe -10mm. If the file is twice as long, I get twice the error (ie same angle.) Surely backlash wouldn't add a cumulative error? One axis is R+P, the other a 32mm ballscrew, but the same is happening whichever way I orient the toolpath, to whichever axis doing the back-and-forth passes. And it always moves towards the origin of the axis, never the other way.

I'm also having problems with the file just stopping randomly, which I've mentioned on the other thread about this. Often this is accompanied by the Galil reporting a timeout error 'while waiting for the controller'- Mach just seems to stop sending commands, but doesn't always crash. This is often several hours into a job. The spindle is still spinning in the cut where it stopped- it just looks like a feedhold, but is terminal. I can also sometimes see Mach 'pausing' for maybe half a second while running code, at random points, at which points the machine suddenly stops for that period of time, causing a judder, but then catches up when the code runs again. I've been able to improve this by increasing the queue buffer level in the plugin. The reason I'm mentioning these other things is because I'm wondering whether I could be having problems with the data record transfer between Mach and the Galil?

Thanks for any thoughts!

ps I built a new PC at the weekend in case the one I was using was a cause of problems- it's a Pentium dual core Gigabyte board without on-board graphics, and has only Mach, Smarterm and Galiltools installed on a new XP Pro installation on a new HD.




Offline kcrouch

*
  •  193 193
  • In way too deep!!!
    • View Profile
Re: toolpath drifting during cut
« Reply #3 on: March 23, 2011, 09:26:07 AM »
Steve,
Set up a dial indicator on the edge of the table and watch the encoder counts as you push to the side. You need to see that the encoder zero matches the actual table zero. Now push the other way and check again. If the dial indicator is different in both directions, you have found the trouble. You may need to have all your scan passes start from the same edge to eliminate this play.

As to the disconnecting, I have seen it when one or more of the following programs are installed, Skype, Quicktime, Acrobat reader,  any virus scanner software. They often attempt to "phone home" to check for updates and can disrupt the stream of data to the ethernet port.
BTW, what port address are you using? Is the cable a direct connection or do you have a switch and other devices on the network? Please advise.

Kenny
Having way too much fun! Something must surely be wrong.
Re: toolpath drifting during cut
« Reply #4 on: March 23, 2011, 12:32:02 PM »
I've just tried an area clearance to a fixed depth, with the tool doing exactly the same passes as in the relief file, but the only difference being that the Z axis is being held stationary rather than following the 3D surface. It's perfectly square! So it seems unlikely to be anything mechanical...

I'm wondering whether the Z motor could be a source of noise. I did have a problem before with the Galil dropping comms all the time, it turned out to be a broken screen at one end of the Z encoder cable, once I fixed it the dropouts went from once every 2 minutes to once every few hours, I guess the single-ended screen was acting as an antenna, picking up all the motor's noise.. I guess I'll try replacing the cable since I have a spare. Could noise trick the Galil into thinking it was receiving encoder counts, despite using differential encoders? Could noisy pulses always send it in the same direction?

As to software, I've got nothing on it but a fresh install of Windows XP SP2 (installed from an old 2006 CD, I haven't even updated it since I didn't want to connect the machine to the internet..), the two Galil programs, and Mach3 installed. No antivirus, and all updates and power management and screen savers turned off. I'm using a 3 foot crossover cable directly, with the PCs ethernet card set to 10Mb only (the 2160s maximum.). I've currently got the data record polling set to manual in the controller registration, since it seems more stable than when it was on auto. I also found an improvement when I unchecked the 'event driven DMCCommand' box. (what is this?)

Thanks for your input

Offline kcrouch

*
  •  193 193
  • In way too deep!!!
    • View Profile
Re: toolpath drifting during cut
« Reply #5 on: March 23, 2011, 12:42:39 PM »
Steve,
Is the power to your cutting spindle located anywhere near the Z axis encoder cable? If so you may try relocating it or removing it from the wireway temporarily. You may have noise on the grounds. Yes, I have seen machines with differential encoders lose counts in only one direction. Not sure why that occurs. If you have one, try using an oscilloscope on the shields of the encoder wires to check for noise.
Lastly, is your z-axis perpendicular to the table surface?
Sorry, but I need to keep asking, maybe I'll get lucky and we'll find the trouble.
Kenny
Having way too much fun! Something must surely be wrong.
Re: toolpath drifting during cut
« Reply #6 on: March 25, 2011, 05:25:14 PM »
The drift in X and Y happens both with and without the spindle motor running- It only disappears when the Z height is held stationary (which is a bit odd, since the Z servo is still working to maintain position).. but as soon as the Z axis starts running up and down to cut a relief, the X and Y start to drift. So I'm suspecting noise from the high power Z motor supply is getting back into the Galil through either the encoder cable, or a GND wire.

I've bypassed the encoder cable with a spare one running straight out the control cabinet door and over the top of the machine to the Z motor, so it's nowhere near any other cables, but it made no difference.

I did scope the encoder grounds, relative to the chassis ground, and it did seem to have up to 2V spikes, so it's likely to be something to do with noise. I've also attached a photo of the scope screen showing the encoder A and B outputs, I'm not sure how much noise to expect, so it'd be good to hear from other people what they think.

The grounding setup as it currently is (as set up by the machine manufacturer, and as recommended by the motor/drive supplier) is that the AC servo motor phase wires come from the drives with both an internal earth and a screen, which are both tied to the main earth busbar in the cabinet. At the other end these are both terminated at an earth connection on the motor body. The encoder cable screen is also attached to this point, which runs back to the drive and terminates at the chassis. The encoder GND is also connected to the cabinet chassis through an earth wire on each drive. The GND of the encoder is internally connected to the GND of the analogue input within the drive (both of which are tied together in the Galil anyway- I thought it would be a good idea to isolate the GND of the encoder drom the drive chassis and just run it straight in to the Galil inputs, but as I said, the Galil ties the encoder GNDs to the analogue output GNDs, whiich end up at the drive anyway.) I've attached a picture of the manufacturer's wiring diagram, I'd be interested to hear if there's anything that needs changing when using the Galil.

Thanks
Re: toolpath drifting during cut
« Reply #7 on: March 25, 2011, 07:52:25 PM »
did you scope the x or y motor to see if this noise was on them too?
i didnt get my daily call today... lol
if its not broken yet, fit more gadgets!
Re: toolpath drifting during cut
« Reply #8 on: March 25, 2011, 08:24:30 PM »
did you scope the x or y motor to see if this noise was on them too?
i didnt get my daily call today... lol

It's just gone midnight, so I officially didn't ask you any questions for a whole day!

yeah, the encoder traces all seemed pretty similar on the scope. Although my checking didn't feel very systematic, since when you're checking for noise that might be on the grounds that are connected to earth, where do you put the gnd clip of the scope..? I didn't post a pic, but with reference to the chassis, the encoder OV had regular spikes of up to 2V, which I guess could either register as pulses or effectively reduce the level of a real pulse so it wasn't recognised.

What I don't get is that the drift is so linear- I did a cut today of a relief with a 0.5mm stepover, the drift on the Y passes is a really nice straight line, so if it's noise causing it, it's really regular, which seems odd if it's coming off the Z motor which is doing a very irregular relief with some flat areas.  ???