"Thunk" during reversal of Z-axis

[torinwalker]

I'm well on my way to getting this machine running. Starting from a bare-boned chunk of iron with stock motors and encoders, I've brought this Frankenstein to life. Now I'm in the tuning phase, trying to eliminate following errors and get everything just so. It's an Ex-Cell-O 605 CNC.

I'm using a 70V Keling 7220 1400W supply, AMC 30A8 PWM brushed-servo AMPs (which, in the case of the Z-axis has been tuned to a critically damped square-wave per the servo-loop tuning instructions.) The motors are Electrocraft E727 38 006 (about 1.125 HP each) with tach feedback and 500ppr encoders. All axis are 10:1.

I've tuned the Z-axis so that errors are less than 2.5m on the scope (viewed from 500u/div), and multiple sequential moves in the same direction are as smooth as glass.

On reversal however, I get a huge oscillation. I believe I've tried compensating with backlash up to .004 (or was it .0004?) in hopes that EMC would neglect this part of the error tracking, but I still get a "thunk" when the axis reverses. No amount of D appears to make this stop without introducing instability (as far as I have experimented thus far.)

It is the "THUNK" that concerns me. I also tried moving the ballscrew shaft looking for axial play, but no movement was apparent.

I figure this is simply a tuning issue, and that I've screwed up the tuning parameters so that while I think I've tuned them properly, I have in fact only half-tuned, or I've tuned them to some mysterious harmonic that manifests in proper movement only in the same direction.

If tuning parameters are suspect, let me know and I'll post the z-axis params. From memory (I'm on the train right now), they're something along the lines of:

p: 200
i: 15
d: 1.72
FF0: 0
FF1: -.0142
FF2: 0.003


Torin...

[Al_The_Man]

What mode are you operating the drives in?
Al.

[torinwalker]

I'm using the drive in Velocity Feedback (Tachometer) mode.

I'm on my way home right now - as soon as I get to the machine, I will verify my settings and re-post them if they're incorrect.

I don't get any kind of "thunk" on my Y axis, but then the Y-axis is large and heavy, as you well know. I believe I also get it on my X axis, but without the pronounced noise (but I still see the glitch on the graph though.)

I've also lowered my acceleration to as low as 4 in/s^2.

What velocity(ies) would one normally tune at assuming the motor cannot be removed from the axis?

I'm trying to tune over the entire range. I'll start with, for example, the highest velocity and tune so that it scales all the way down to 1IPM. I still get the occasional following error, but I figure with enough tuning I'll be able to iron out the last few kinks.

Which brings up a separate question which probably belongs in a separate thread, but since you'll probably answer it anyway, Al, I may as well ask it as a sidebar question here: is it possible to remove the x table without completely dismantling the servo/ballscrew mechanism? I need to make sure the meters on the inside of the X-table are receiving fluid in case it has an effect on tuning.)


Torin...

[Al_The_Man]

I have always run my AMC drives in the Torque mode, and eliminated the Tach altogether.

Originally Posted by torinwalker Which brings up a separate question which probably belongs in a separate thread, but since you'll probably answer it anyway, Al, I may as well ask it as a sidebar question here: is it possible to remove the x table without completely dismantling the servo/ballscrew mechanism? I need to make sure the meters on the inside of the X-table are receiving fluid in case it has an effect on tuning.) Torin...

It's been quite a while, but I believe you have to strip it right down.
I think the 'Dog Bone' in the middle for the X & Y nuts are one unit?
I did a number on both the ones I had and replaced all the metering devices at the same time, as well as completely strip and clean the whole lube/pump system.
My rapids are set fairly conservatively.
I just remembered, I had an issue with the Z on mine about a year ago, it was the arm that went from the nut to the quill had flex due to a lose allen head bolt.
Al.

[Al_The_Man]

I remember now, IIRC the table comes off quite easily if the ballscrew nuts are off and the gibs removed.
Al.

[torinwalker]

Al,

Thanks for the X axis info. I still need to remove my X-axis to inspect the remaining meters.

Incidentally, I figured out how to clean those meters - there's a fabric inserted into the non-valve end that comes out quite nicely with a very narrow set of tweezers. The rest can be cleaned with carb cleaner, and they're as good as new! I actually took three of them completely apart to reveal a pin, rubber diaphram, brass plate, and the end cap. All but one went right back into service without a hitch. The one that I couldn't get working was replaced. If I were a shop however and uptime was a concern, I'd be replacing these things. They're very time-consuming to clean.

So I finally set everything up and ran my first g-code program to move the axes back and forth in a continous loop so I could more easily tune on the fly - I think I've tuned my X and Y quite nicely with rapids up to 180ipm. There's just one thing now that bugs me: That darn "thunk" when the Z axis switches direction. Now that I can run the Z-axis in a loop, I can listen and home in on the source of the thunk - it's coming from the gear case that houses the - helical gear? - that joins the Z motor spindle to the Z-axis ball screw. Now here's what I think so far...

There's a little oil window facing front just to the left of this helical gear. On top of the gear case are two little 3/8 hex-keyed threaded plugs that I surmise are for putting oil, or some kind of lube, into the gearcase. I removed the left one and dabbed the end of the allen key into the hole to reveal a blue goo - grease! This raises some concern - shouldn't it be oil if there's an oil window? If grease was meant for the gear case, there'd be a grease nipple in place of the threaded plug. I also don't see an oil port on the bottom of the gear case for oil to drain.

You have an excello, do you not, Al? Is yours a 605? Would you happen to know what is inside this gear case and by chance what it's supposed to be lubed with?

The reason I'm asking is, this "thunk" I hear may be the result of a badly worn gear - or perhaps because the wrong lube is being used. Perhaps... just perhaps, putting oil into the case may reduce or eliminate the thunk. This, of course, is a wild stab in the dark.

I don't think it's a tuning issue. No matter what acceleration I use, PID parameters, or whatever, the thunk is constant. I feel pretty comfortable tuning the axes, so while I could be wrong, I don't think it's a tuning problem at this point.

I'm posting a video shortly.

[Al_The_Man]

If it is the gear box you are talking about there is a gear shift lever that moves from 1:1 to around 6.3:1
The sight glass should show the oil level when I believe it is in 1:1?
I spoke to ExcelIlo at the time I overhauled mine, which had the original gear oil in it, they said they had gone over to using automotive transmission fluid, which I have been using ever since.
I have a NPT filler plug as well as NPT drain plug.
The only think I had to repair was the worn bronze gear shifter, I silver soldered a new shift fork on to the old plate as they did not have a spare at the time.
Al.

[torinwalker]

No, I'm talking about the transfer case on the Z-axis. Your machine isn't a 605, is it?

I've created a video to illustrate:

YouTube - 00053.MTS

Torin...

[Al_The_Man]

Sorry no, its a 602!!
Al.

[torinwalker]

Al,

Did you see the video? (I fixed the link.) Does your z-axis look anything like mine? I've never seen a 604 - is it the same as a 602?


Torin...

[torinwalker]

I tried something interesting - I wanted to see if there was a big gap between the gears that connect the horizontal and vertical gears, so I wrote a little gcode program to move the axis back and forth between 2.000 and 2.005. The very interesting part is there was absolutely no observable backlash - when the horizontal shaft turned, the vertical ballscrew shaft turned at precisely the same time. If there was a large gap between the gears, I could account for the very large thunk when the axis switches directions, but now... I don't know what it is. If the z-axis has no play, it must be a software tuning issue!

I guess I'm back to square one.

[lumberjack_jeff]

The noise on the video sounds almost like I envision the sound of ballscrew bearings taking up for one missing ball on reversal. Not a "thunk" so much as a bunch of "clacks" a few microseconds apart.

Can you take the servo off and rotate the shaft manually?