Diagnosing source of shudder / stutter / jerk / shake

[bartniswonger]

Hi all - I am having trouble with the Y axis of my router shuddering when doing short moves. It seems like something is binding, the question is how to determine what.

The machine is a an old Maxym router - an odd cantilevered design but heavily built. The Y axis runs on a pair of THK linear rails, 16" on center. There are two trucks on each rail, about 8" on center. The Y axis is driven by a chinese 2010 ballscrew with a plain ballnut (not backlash limiting). The steppers are driven by a Gecko g540.

The shudder started after I tore down the Y axis, disassembled and cleaned the bearing trucks, and replaced a 1/2" acme screw with the 2010 ballscrew. Now I am wondering if cleaning the bearing trucks allowed them to have too much side to side play, allowing the set of them to rack, and therefore bind. Or is it an issue with the ballscrew, and particularly the lack of backlash limiting.

The shudder only happens during a series of short moves. I can move rapidly (I have only gone 300 ipm) and it appears very smooth during acceleration and at speed. It is also smooth during arcs (G2, G3). I use spiral tool paths a lot, for aesthetic reasons, and my CAM software tends to make them into a lot of G1's with deltas of .0005" or so. These used to be fine prior to my "fixing" things, and they are still fine on the X axis (which has similar linear guides, and a 1/2" acme screw). The shudder happens anywhere on the Y axis.

I was worried there was some interplay between the axis motion, so I stripped out all the Y and Z movement in a test file. The machine shudders. I ran the same program, but swapped to the X axis (and then to Z), no shudder. So it seems like it is isolated to the Y axis.

I have played with acceleration and max speed. Bringing the acceleration way down (less than 2 ips in Mach3) does not eliminate the problem. Using my phone as a seismograph, I found that in Mach3 - 5ips acceleration produced max vibrations of >300 gal. 2ips produced vibrations of >175gal. I tried LinuxCNC - it is much smoother, it turns out. Acceleration of 10ips in LinuxCNC produced max vibrations of <125gal. The steppers do not stall until over 15ips. This is with a max velocity of 180ipm in both Mach and LinuxCNC. I am using LinuxCNC now, but you can still hear the machine banging around so I would like to find a way to tighten it up.

I have also tried to measure backlash in the screw. I did this by moving my Y axis from 0 to 1 and indicating that spot with a dial indicator. I then moved from 1 to 2, and back to 1. My indicator says I am within .001" of where I started, which is within the error of my indicator. So I think the backlash is within at most .001", and I am not sure how to measure it more precisely. Is that enough to cause shudder?

I should also note that I have lubricated all these parts with grease - Mobilux EP2. I started using way lube on the ball nut, but I could hear the balls rattling around and finally switched to grease. It is quieter now, but does not seem to have made much difference.

Another note - the end seals on the bearing trucks were trashed - a gooey waxy substance rather than rubbery. I have new seals, but they have not yet been installed. Could it be these trucks are worn out?

I am not sure how to break the problem down further. How can I rule out either the ballscrew or the linear guides? If it is the guides, could I load them laterally so they were less able to rack as a unit? That is to say, push them outward from the center between the two rails, then bolt them to their mounting plate. It seems like it would cause more wear on the inside of the rails, but it seems like it could work, maybe. How much to load them? Maybe I do not care about guide wear in a low-production machine. I know I would rather not replace all the trucks to find that it was the screw all along.

Thanks for your time!

Bart

I have attached some images

1) The machine itself, sorry about the clutter. It worked very nicely to machine the brass seen on the table, no shudder, but just nice straight moves.

2) The axis in question

3) A closer view of the linear guides and the ball nut mount.

Similar Threads:

• Need Help!- Y-axis Shake
• Y Axis Shudder
• Shake up at OneCNC?

[louieatienza]

Hi all - I am having trouble with the Y axis of my router shuddering when doing short moves. It seems like something is binding, the question is how to determine what.

The machine is a an old Maxym router - an odd cantilevered design but heavily built. The Y axis runs on a pair of THK linear rails, 16" on center. There are two trucks on each rail, about 8" on center. The Y axis is driven by a chinese 2010 ballscrew with a plain ballnut (not backlash limiting). The steppers are driven by a Gecko g540.

The shudder started after I tore down the Y axis, disassembled and cleaned the bearing trucks, and replaced a 1/2" acme screw with the 2010 ballscrew. Now I am wondering if cleaning the bearing trucks allowed them to have too much side to side play, allowing the set of them to rack, and therefore bind. Or is it an issue with the ballscrew, and particularly the lack of backlash limiting.

The shudder only happens during a series of short moves. I can move rapidly (I have only gone 300 ipm) and it appears very smooth during acceleration and at speed. It is also smooth during arcs (G2, G3). I use spiral tool paths a lot, for aesthetic reasons, and my CAM software tends to make them into a lot of G1's with deltas of .0005" or so. These used to be fine prior to my "fixing" things, and they are still fine on the X axis (which has similar linear guides, and a 1/2" acme screw). The shudder happens anywhere on the Y axis.

I was worried there was some interplay between the axis motion, so I stripped out all the Y and Z movement in a test file. The machine shudders. I ran the same program, but swapped to the X axis (and then to Z), no shudder. So it seems like it is isolated to the Y axis.

I have played with acceleration and max speed. Bringing the acceleration way down (less than 2 ips in Mach3) does not eliminate the problem. Using my phone as a seismograph, I found that in Mach3 - 5ips acceleration produced max vibrations of >300 gal. 2ips produced vibrations of >175gal. I tried LinuxCNC - it is much smoother, it turns out. Acceleration of 10ips in LinuxCNC produced max vibrations of <125gal. The steppers do not stall until over 15ips. This is with a max velocity of 180ipm in both Mach and LinuxCNC. I am using LinuxCNC now, but you can still hear the machine banging around so I would like to find a way to tighten it up.

I have also tried to measure backlash in the screw. I did this by moving my Y axis from 0 to 1 and indicating that spot with a dial indicator. I then moved from 1 to 2, and back to 1. My indicator says I am within .001" of where I started, which is within the error of my indicator. So I think the backlash is within at most .001", and I am not sure how to measure it more precisely. Is that enough to cause shudder?

I should also note that I have lubricated all these parts with grease - Mobilux EP2. I started using way lube on the ball nut, but I could hear the balls rattling around and finally switched to grease. It is quieter now, but does not seem to have made much difference.

Another note - the end seals on the bearing trucks were trashed - a gooey waxy substance rather than rubbery. I have new seals, but they have not yet been installed. Could it be these trucks are worn out?

I am not sure how to break the problem down further. How can I rule out either the ballscrew or the linear guides? If it is the guides, could I load them laterally so they were less able to rack as a unit? That is to say, push them outward from the center between the two rails, then bolt them to their mounting plate. It seems like it would cause more wear on the inside of the rails, but it seems like it could work, maybe. How much to load them? Maybe I do not care about guide wear in a low-production machine. I know I would rather not replace all the trucks to find that it was the screw all along.

Thanks for your time!

Bart

I have attached some images

1) The machine itself, sorry about the clutter. It worked very nicely to machine the brass seen on the table, no shudder, but just nice straight moves.

2) The axis in question

3) A closer view of the linear guides and the ball nut mount.

Disconnect the ballnut from the Y and try to run a program and see if you have the same problem. Maybe the bearing block on the fixed end is binding. Are you using a coupling to drive the screw abd does it allow for misalignment? Does tgat axis slide freely with no binding? Also just make sure your connections from drive to stepper are good.

[alan_3301]

in your mdi type:
g64 p0.001

this will try to maintain programmed feed rate by sacrificing path accuracy up to 0.001" you can raise this number depending on the accuracy you hope to achieve.
It greatly reduces vibration due to acceleration and deceleration of very short segments.

read this page for more info. LinuxCNC Documentation Wiki: TrajectoryControl


hope that is your problem.

[bartniswonger]

Thank you for the idea of disconnecting the ball nut. I just did it, and everything moves smoothly - meaning when I run my test program there is no shudder or any kind of hesitation while I was holding onto the ball nut. Additionally, the cantalevered beam with the X axis moves smoothly - it is heavy and I cannot accelerate it very fast by hand, but I also cannot feel any kind of binding. Grabbing it on the end - so I have a 5' lever arm - I cannot detect any play in the linear bearings - it feels very tight, and moves immediately with no rocking.

I am not sure testing the Y axis stepper and ballscrew unloaded is a fair test to determine the bearings are the problem. But I am no expert! It just seems like the load on the ball nut could cause the balls to bind with lots of back and forth, and I may not be able to recreate that load by holding the nut with my hand.

As for connections to the stepper - I am attaching a photo of the connection - it is a standard coupling that allows some lack of alignment. I think it is pretty close to aligned, certainly I see no evidence of lack of alignment. Everything seems tight.

[bartniswonger]

alan - You appear to have nailed it. I must say I am totally flabbergasted. I have been using Mach3 for a year or so, struggling with Constant Velocity settings and roughness/shuddering - LinuxCNC appears to be quite superior, or I really am missing something in Mach3.

I put in G64 P0.0001 and then ran it at 160ipm - the vibration is less than 100gal on my phone-seisomograph and I cannot hear any clunks. That is twice as fast as I was running it before and it appears to be fine.

Thank you!

[alan_3301]

You're welcome.

I struggled with this the first time I ran a 3d rastering toolpath, with millions of tiny segments.
I believe the default settings of linuxcnc years ago was exact stop mode, where the machine would actually start and stop on each line.
My machine tried to shake itself off of the table.

Be aware that you can program multiple g64's anywhere in your program you want, so you could specify a higher number during roughing, and tighten it up on finishing.

I forgot to say, the P# is in machine units, so inches or mm, whatever your config is setup with.

Also, if that was really 0.0001 that you entered and not a typo, I'm suprised it helped. You should try 0.001 or even 0.005 it really speeds things up.

[bartniswonger]

I did 0.0001 - I tried 0.001 as well, there was not much difference between the two, so I went with the smaller one.

What confuses me is that I had Mach3 set with .01" for its constant velocity mode and it has always worked ok, if a bit rough - you could see the shake in the cut in certain places. But the new screw and/or cleaned bearings really did not work well in Mach3. I now need to go back and check my settings to be sure I didn't change something in Mach3 - although I am getting used to linuxcnc.

Thanks again!

[ger21]

What confuses me is that I had Mach3 set with .01" for its constant velocity mode

I would disable this setting in Mach3, as it can probably cause a lot of stuttering in your case.
I'd disable all the CV options except "Stop CV on Angles" and play with that to get the results you need. Start at 89 and go down from there as needed.

[samco]

What version of linuxcnc are you running? I am glad it is performing well for you. The G64Px.xxxQx.xxx really does give you very nice control over path following. The next big release of linuxcnc (2.7) Will have a N lookahead trajectory planner - it is very exciting.

The mach CV Dist Tolerance_____Units.. isn't really path following tolerance it is 'how far from the end of the segment do I start the blend for the next segment. It will follow the path closer but not dynamically like Linux will. If you tell linuxcnc to follow the path within .005" it will go as fast as it can while obeying that tolerance. (ie - the mach tolerance doesn't exactly relate to path following)

Turning the CV Dist Tolerance_____Units off is interesting also - it probably works fine for lower speeds - but higher speeds mach seems to combine segment in an unpredictable way.

here is an image of linuxcnc plotting the path mach3 was taking - the path is back and forth - you can see one way the path 'jumps' over quite a few segments while the other way it follows closer.
http://wiki.linuxcnc.org/uploads/Screenshotmach.png

Here is some info I compiled while helping test linuxcnc new TP
LinuxCNC Documentation Wiki: NewTrajectoryControl

sam