Seems to have asymmetrical movement all along the axis. Varies from 150-250, 175-275, 200-350, 200-450.
If I take out the velocity Feed fwd, it's even more variable, switching positive to negative symmetry and back again.
This is a Warner & Swasey QCM-6 (M-5100). Small machine, very well built like a much larger W/S. Granet bed and headstock. 1984 vintage. I was told by one of the original W/S guys that it was intended for tool room service. Has about 8 inches X travel and about 11 inches Z travel. 2.5" round Z wayrods, 2" X wayrods. Bijur pressure lube. Fibrotakt (German) air operated 8 position turret. Hydraulic drawbar, 2-J collets, Whitnon spindle, A2-5 spindle face (I think). Came with a GE2000 control (for sale). Has about 400 hours on the clock. Slides (wayrods) have seals and ballscrews have bellows, but it all appears to be oiling well.
Will check movement at different positions.
Seems to have asymmetrical movement all along the axis. Varies from 150-250, 175-275, 200-350, 200-450.
If I take out the velocity Feed fwd, it's even more variable, switching positive to negative symmetry and back again.
Hi Charles,
Thanks for the info. Interesting about the "wayrods". I wonder how they handle binding if the rods aren't perfectly straight, parallel, and correct separation. Variation that occurs based on direction, position, and time is most likely a mechanical issue. The servo feedback and control is not likely to have any significant variation. It might be that different tuning causes the motion to be smoother/jerky/slower/faster or such which then cause the mechanics to be more likely to stick or not.
I suppose there could be some issue with the motor or tach. The tach doesn't seem to be doing much. You might disconnect it and see if there is much change and/or things become more consistent.
I suspect the tach not doing its job, this type of amplifier have slow response, and mechanical variations are all combining to work against us.
Regards
TK http://dynomotion.com
Hi Tom,
I will disconnect the tach and try again. Something is causing inconsistent moves. If I do a move several times without changing any parameters, the plot is very likely to look different each time. I will post a series of moves without changes.
Spent several hours removing, cleaning, checking, and replacing every lube metering device. Manually oiled every port with Vactra 2 as specified. Made sure each meter provided oil from Bijur pump. Disconnected Z ballnut. Z slide resists initial movement (stiction), then moved very nicely back and forth. I wouldn't want it any different as it's moved back and forth. After setting a few minutes, overcoming stiction is tough.
The inconsistency in moves is about the same on X axis.
[QUOTE=Charles2;2398630]Hi Tom,
I will disconnect the tach and try again. Something is causing inconsistent moves. If I do a move several times without changing any parameters, the plot is very likely to look different each time. I will post a series of moves without changes.
Spent several hours removing, cleaning, checking, and replacing every lube metering device. Manually oiled every port with Vactra 2 as specified. Made sure each meter provided oil from Bijur pump. Disconnected Z ballnut. Z slide resists initial movement (stiction), then moved very nicely back and forth. I wouldn't want it any different as it's moved back and forth. After setting a few minutes, overcoming stiction is tough.
The inconsistency in moves is about the same on X axis.
Hi Tom,
When I disconnect the tach, the axis tries to run away as soon as the drive is enabled. So I take it the tach is working to some degree. I have tried reading the tach voltage output. My meter is slow to respond, and I have only about 11" of Z travel, so it's difficult to get a good reading. I think I'm seeing 1 to 2 volts, thought it would be more. I will try again with a helper.
I'm posting plots of 5 moves, all with the same parameters. Sometimes repeated moves are even more different.
Hi Charles,
I'm not sure what you mean. No stiction would be better.Disconnected Z ballnut. Z slide resists initial movement (stiction), then moved very nicely back and forth. I wouldn't want it any different as it's moved back and forth. After setting a few minutes, overcoming stiction is tough.
That doesn't really make sense. Removing it might cause it to go unstable and oscillate, but not run away. You might try putting a jumper across the tach connection to guarantee 0V into the amplifier rather than a floating input.When I disconnect the tach, the axis tries to run away as soon as the drive is enabled. So I take it the tach is working to some degree.
Regards
TK http://dynomotion.com
Hi Tom,
While the ballnut was disconnected, the Z slide moved smoothly, evenly, back and forth by hand. Not loose or tight. I would say just right. If it sat for 10 or 20 minutes though, it would take a considerable amount of effort to start it moving. After that it was smooth and even. Direction changes required very little effort. From the plots, I expected more effort to be required to change directions. Maybe stiction is not the correct term.
This machine has no noticable wear. It was at a community college. The 400 hours on the clock may have been from mostly idling. It's really not even broken in, so to speak.
I disconnected the tach wires again and jumpered the terminals. Still takes off when drive enables. Since I only have 11" travel, I may be confusing the beginning of large oscillation with run away.
I feel like I'm missing something, probably something simple.
What did you think of the plots?
Thanks,
Charles
Hi Charles,
Again that doesn't make sense.I disconnected the tach wires again and jumpered the terminals. Still takes off when drive enables.
I'm not sure what you mean by "Enable". The Drives or the KFLOP Axis? You might go back to DAC commands. With DAC=0 does it still run away? Can you find a DAC value where it runs away the other direction? A DAC value where it remains stationary?
You might create a Move Plot of size=0 to capture exactly what is happening.
Actually considering the resolution is 80000 counts/inch they aren't terrible. Have you verified the resolution is 80000 counts/inch? If you move 320000 counts does it move 4 inches?What did you think of the plots?
So the 10000 count test move is only 1/8 inch. With errors of ~500 counts (you didn't include the data) that would be 500/80000 = 0.00625 inches.
What kind of accuracy and speeds are you expecting?
Another mechanical issue that can cause inconsistency is backlash in the ball screw and/or bearings. Have you checked backlash? By making tiny slow moves and measure actual distance moved with a probe.
Regards
TK http://dynomotion.com
Hi Tom,
I'm using SWE and +5v from Kanalog JP8 to switch a 3-32 VDC solid state relay, which completes the E-Stop circuit, which enables the motor drive (amplifier), preventing the motor drive from starting up or drifting before Kflop is fully powered up.
With the tach wires disconnected, and jumper on terminals, the axis takes off as soon as the motor drive is enabled, before I have time to enter a DAC move.
320000 counts moves the axis 4 inches.
I would hope to have .0005-.0010 accuracy.
200-300 IPM should be fine on this small machine.
At this point, how do I make specified distance moves to check backlash?
DAC1=0 move plots and data to follow.
Thanks,
Charles
The plots above are not DAC moves of course. They are Step Response moves of zero size. Data following.
Data21.txt
Hi Charles,
That's good, but as the Link in my post #67 explains it is better to have both the SWE keep the amps disabled and a separate disable to keep the amplifiers disabled. That allows after power up to keep the amps disabled whenever the KFLOP axis is not enabled.I'm using SWE and +5v from Kanalog JP8 to switch a 3-32 VDC solid state relay, which completes the E-Stop circuit, which enables the motor drive (amplifier), preventing the motor drive from starting up or drifting before Kflop is fully powered up.
KFLOP initializes the DACs to zero so by the time the motor drives are enabled the DACs should be commanded to zero. So I don't understand this. There may be some small offsets so a DAC command of zero might not be exactly zero which could cause some drift or gradual acceleration.With the tach wires disconnected, and jumper on terminals, the axis takes off as soon as the motor drive is enabled, before I have time to enter a DAC move.
Good that confirms the 80,000 counts/inch resolution.320000 counts moves the axis 4 inches.
Things will need to improve by ~10X to achieve thisI would hope to have .0005-.0010 accuracy.
200-300 IPM should be fine on this small machine.
You could use Console Screen MoveRel commands: MoveRel1=80 should move 0.001 inches. MoveRel1=8 should move 0.0001 inches.At this point, how do I make specified distance moves to check backlash?
The Plots do not indicate any run away. Instead a tiny oscillation of +/- 4 counts ( +/- 0.00005 inches) at 60 Hz. Was the Axis running away?The plots above are not DAC moves of course. They are Step Response moves of zero size. Data following.
Regards
TK http://dynomotion.com
Hi Tom,
Thanks again for your help.
So if the motor drive is disabled whenever the Kflop axis is not enabled, does that mean DAC moves are not possible?
I may be able to use the Zero pot on the axis drive to balance the DAC=0 position.
As to the plots, the axis was not running away or drifting because the tach wires were connected at that time.
With all power off, indicator set to check Z axis relative to spindle face, I found about .0002-.0003 backlash or deflection using a 3 foot long 2 x 4, moving positive to negative and negative to positive. X axis shows no indicator movement while trying to shove the cross-slide back and forth by hand. It was not as easy to use the 2 x 4 on X.
Thanks,
Charles
Hi Tom,
If I understand correctly, when using the "and" logic circuit with SWE and opto output to control drive (amplifier) enabling, the opto output is turned on with a "C" program only if the Kflop axis is enabled. Is that correct?
Thanks,
Charles
Hi Charles,
No. Normally you would have a forever loop monitoring KFLOP's Axis Enable and if becomes disabled for any reason to then disable the amplifier to avoid any drift. But if you want drift and DAC commands then don't run that program.So if the motor drive is disabled whenever the Kflop axis is not enabled, does that mean DAC moves are not possible?
Yes.I may be able to use the Zero pot on the axis drive to balance the DAC=0 position.
The idea was to diagnose the "runaway" with the tach disconnectedAs to the plots, the axis was not running away or drifting because the tach wires were connected at that time.
That seems quite good. Not sure what the 2x4 was used for, I'm surprised "shoving" shows no deflection whatsoever.With all power off, indicator set to check Z axis relative to spindle face, I found about .0002-.0003 backlash or deflection using a 3 foot long 2 x 4, moving positive to negative and negative to positive. X axis shows no indicator movement while trying to shove the cross-slide back and forth by hand. It was not as easy to use the 2 x 4 on X.
Yes, normallyIf I understand correctly, when using the "and" logic circuit with SWE and opto output to control drive (amplifier) enabling, the opto output is turned on with a "C" program only if the Kflop axis is enabled. Is that correct?
Regards
TK http://dynomotion.com
Hi Tom,
I checked both X and Z for backlash using the MoveRel commands. Found .0002 in Z and .0001 in X. Used .001 moves positive and negative and also .010 moves.
Thanks,
Charles
Hi Tom,
I was able to adjust the drive Zero pot to minimize drift with the tach disconnected and DAC1=0. Still seems to build up and try to move after a few seconds. I guess if the Kflop axis enable is activated in time, it will hold position. I ran a couple of plots. See below. I keep forgetting to limit the time to 1 second so the data won't be too large to send. Will rerun and send that.
Data for run with above parameters except .8 second time to keep data file size small enough to post.
Data23.txt
Hi Charles,
I would now tune the axis to see how good it can get without tachs. With such a low P gain of 0.12 an error of several thousand is needed to create the output of several hundred to achieve the acceleration you are testing of 5 in/s^2 (or 0.013G which is fairly low). Optimize P and D. I would test a larger move as your system can be more unstable when moving faster.
If you zip the data file it should be ~4X smaller.
Regards
TK http://dynomotion.com