1. ## Backlash...bring the flames.

Perhaps I should introduce myself. After lurking here for nearly a year, purchasing an IH manual mill and later completing the CNC conversion from an IH kit, I finally have some comments and questions for yall. Cruiser, Bob, and Runner have nearly become household names to me, yet you guys have never heard a peep from me, so here I am. Bob, I love your web-blog, keep up the great work. Cruiser, nice work on the belt drive and the z-axis mod. I am located in Bend Oregon, If you ever come over to go skiing, drop me a line, and we can have a beer.

I am an ameture machinist, having had an introduction to machining at Oregon State University (GO BEAVERS!!) where I studied to be a Mechanical Engineer, I now work for Cessna Aircraft here in Bend.
I have many questions to ask/answer, and a lifetime of tooling crashes ahead of me. (shockingly none so far.....crap....well that goofed it up..)

OK, enought rambling. Backlash..... I have approximatly 0.003-0.005 system backlash on my x and y axis, and approximatly 0.010 on my Z. the "backlash" is of course not due to the ballscrews, but rather the torquing of the ball-screw required before the axis slide begins to move and the screw actually begins to turn in the ballnut. I refuse to turn the backlash compensation in Mach on, as the stop-go movements are rather annoying, and certainly don't do any favors to your program run time.
I have set my gibbs so there is approximatly 0.0005 in lateral movement across the joint when the table is forced in-plane. (ie x-axis is pushed in y direction, and the movement at the end of the dovetail is measured with an indicator.)
So my quesiton is this, a certain bit of "torquing" in the ballscrew is inevitable prior to movement, but is 0.003-0.005 backlash excessive? If i interpolate a 0.5 inch hole it reliably comes in at 0.493 +_ 0.001 and round as can be. Repeatablility is outstanding, but the inability to hit the desired dimension is irritating. (note: I have double checked the actual cut width of the cutter used, as well as trying different size and shape holes.)

Does anyone have any suggestions for reducing this backlash without excessively loosening the gibbs.
Are the gibbs adjustments really that much of a dance to achieve optimum adjustment?
The Z axis is un-copensated for its weight, and the gibbs are set so there is about 0.001 "rock" when a lifting load is applied to the nose of the head. (any looser and the spindle would come out of perpindicularity alignment with the table.)

If any of you guys familiar with the mill have any suggestions for me, I am all ears.
Thanks!!!

JeepJake

2. do you have the IH screws on your mill? if so i would think that you didn't preload the nuts enough if you have that much back lash. i forget what the setting is but i think its 1/2 turn of the adjustment nut gives you about 120 lbs of preload. mine are set to exactly the recommended spec and when i checked backlash initially i had none. i am about to recheck it.

3. Runner,

Yes, I am using the ballscrews and ballnuts supplied in the IH kit. I preloaded the nuts to the maximum recommended preload. the backlash is ceartainly not coming from the ballscrews.

If I power down the machine and turn the screws by hand, I can observe the pully moving several degrees before any rotation is detected at the cross slide. as one can imagine this is reduced when the table is closest to the drive pully, as there is not a long distance of ballscrew that must be torqued to generate movement.

comparable to haveing a long extension on a socket wrench, you must provide more angular displacement at the handle to achieve the same torque at the socket than you do with a short extension, due to the length of the extension. not more torque mind you, but more angular displacement.

Jeepjake

4. Jake,

theres only a few places you can generate backlash in your system. screw torque is not one of them. for what your implying the screw would need to stretch, this is not possible with the amount of force on this screw.

my gibbs are tight and my axis moves freely. i don't think you set the preload correctly on the ball nut. to do this properly you need to snug the adjustment nut against the spring washers (make sure these are all seated against each other tightly, mine had a tendency to seperate and bind and not seat) then once all the slop is taken out, then you turn your adjustment nut .5 turns to preload against the spring washers.

5. Runner, thanks for the suggestions on the ballnut, I did observe the same issues when I was torquing them. I am, however, certain that there is no backlash there.

Perhaps I am not describing my situation best here, you are 100% correct, there is no way on this green earth that the screw is stretching any measureable amount under free movement, and in the traditional sense, the system IS backlash free.

To demonstrate the condition I am speaking of, take this example to the extreme. lock down your gibbs TIGHT. Turn off your machine and turn the motor pully by hand. There is no way that you are moving the slide, nor are you appricaibly stretching the ball-screw axially, but you can easily get about 10 or so degrees of rotation at the drive-pully, (and thus the encoder) while there is no turning of the ballscrew inside the ball-nut some 20 inches away. This is the "backlash" I am talking about. It is not so much a backlash as it is a lack of rigidity in the drive system (hence my thread title...bring on the flames, I knew i was opening a can!! )

I agree there is no "backlash" in the system in it's proper definition, my error in terminology, sorry. but there is certainly a percieved motion by the controller (since the encoder is mounted to the drive pully on the screw) when in reality, there is a few degrees of twist being produced in the screw before there is any rotation of the ballscrew in the ballnut (at times a good 20 inches away, and the farther away, the greater the error). Again this can be more easily observed if the gibbs are tightened down.

The only true way to eliiminate this error is to mount the encoder on the free end of the ball-screw, ensuring that the rotation at the encoder is an accurate representation of the rotation inside the ball-nut...well that seems a bit excessive. Or use larger diameter ball-screws, thus increasing their torsional rigidity.

Things I will try:
1. I will interpolate a hole at +y-x extreme table position for a worst case, and a -y+x extreme which will shorten the distance between the encoder and the ball-screw to represent a best-case. i should see a difference in accuracy realitive to commanded motion between the two. I will report my findings.

as an aside, I am using 30wt motor oil as Way-lube, I can't imagine that friction due to poor lubrication is an issue, do you?
thanks!
Jake

6. Jake,

i wonder if you and i are experiencing the same problem. i just ran through the tests that mach gives in their manual. and for the life of me i can't find any back lash in the axis. 10 degrees of motion on the pulley is alot of motion with no movement. that motion has to go somewhere. if it is backlash then it should show up someplace. this type of motion is no good period. i am going to take my covers off an try turning the pulley by hand and see what i get. right now i've got an indicator mounted against the spindle so wheel see if i get any motion with even the slightest movement.

steps poer inch on this machine are 20,000.

oh i just checked and my pulley is tight so any twitch in the pulley relates to a twich in the indicator. back to my argument that something is not tight. are your bearings seated right? did you put the shim between the bearings? is your ball nut tight to the mount? can you take it apart and reset the preload on the ball nut? somewhere in there is where your getting lack of motion. i would be that your going to get the same error whether its close the drive or not. also mounting the encoder on the far end of the screw will make no difference. getting a reading from a linear scale would be more accurate.

7. I've had similar problems on my knee mill, tracked down to flex in the bearing supports. I replaced the original cast aluminum mounts with solid steel mounts, and the problem went away.

Have you used and indicator to check for axial movement of the screw? The problem has to be one of:

1) Backlash in the nuts
2) Movement of the nut in its mount
3) Axial float in the screw, due to inadequate pre-load in the bearings, or flex in the mounts.
4) Radial play in the pulley(s)
5) Insufficient belt tension

It is *not* the screws twisting - they simply don't do that under anything close to normal machining loads.

Regards,
Ray L.

8. It could be that my Gibbs are just set a bit too tight.

at risk of messing with your machine's setup, really crank down on your gibbs, a lot, make them overly tight, just to make more clear the issue I am experiencing.
make a witness mark with a paint pen or something non-corrosive between the ball screw and the ballnut on your slide.
when this is done, you can twist the drive pully several degrees, which turns your encoder, indicating to mach that the machine has moved. hold the pully in this position and check the witness mark at the ball-screw, it will not have moved at all (rotationally) thus the percieved motion at the encoder has been "absorbed" by the rotational deflection along the length of the ball-screw.

maybe I need to loosen my gibbs a bit, and use a thicker way oil to maintain a better plane bearing, reducing friction at the slide, thus reducing the amount of pre-torque is needed in the ball-screw prior to moving the slide.

since I have about .004 inch error, 20,000 lines per inch, this amounts to 60 encoder lines observed at the drive pully before rotation at the ballscrew on the slide is observed, translating into linear movement at the slide. i cant remember the TPI of the ball screw, so I don't know what this comes out to as far as rotation.
The reason that the encoder on the free end would provide more accurate position data is that the feedback response to the actual position is not "damped" by the rotational displacment required to enact motion, the free end of the screw is not under load, so would provide an instantanious indication of the actual motion seen at the slide ballscrew. essentially taking the feedback sample out of the load path. (the same reason the encoder is on the screw now, rather than on the motor, only putting the encoder on the free end of the ball-screw would be taking it one step further.)

Have you interpoalted a hole, then measured it to determine how accurate the hole came out ?
Thanks for all the discussion Runner, we will certainly get to the bottom of this.

9. put an indicator on the end of the screw - lock the table... Turn the screw - does the screw move in and out? If it does then it is the thrust bearings. if it doesn't - it is the ballnut/mounting play.

sam

10. Originally Posted by HimyKabibble
I've had similar problems on my knee mill, tracked down to flex in the bearing supports. I replaced the original cast aluminum mounts with solid steel mounts, and the problem went away.

Have you used and indicator to check for axial movement of the screw? The problem has to be one of:

1) Backlash in the nuts
2) Movement of the nut in its mount
3) Axial float in the screw, due to inadequate pre-load in the bearings, or flex in the mounts.
4) Radial play in the pulley(s)
5) Insufficient belt tension

It is *not* the screws twisting - they simply don't do that under anything close to normal machining loads.

Regards,
Ray L.
belt tension should not come into play as that is "out of the loop" for the IH cnc kit, the encoder is directly hooked to the ball-screw, so any slop in the belt is elimnated from the feedback loop. (but mine are tight)
Looks like I've got some homework to do, i have not measured any of the other above suggestions, thanks Ray, I will do so just to make sure. I just set up the machine, and was very cautious to properly and evely torque EVERYTHING, so I am 99% certain based on my assembly process there there is any issues with "true" backlash. I am confident that the system is indeed backlash free in terms of axial motion, the symptom I am describing is more of a system rigidity issue.

Thanks, I will check your suggestions tonight.
jake

11. This sounds interesting, My first thought was possibly the drive end bearing load but runner got to that one. If it isn't there then I'd say loosen the Gib's and DO NOT use the side grubs at all ! They are not practical on cnc system, and they put deformation in the surface. These Ball screws will torque flex quite a bit if gib is too tight, and I'd have to say that your reference implies too tight of a gib. If the ball screw wasn't in the picture you should be able to slide the axis by hand, I'd start by backing them out till they are obviously loose then just push them in with thumb or finger and shaking axis some with finger pressure then set the gib there TO START WITH. Then recheck the backlash. Take a mirror and flash light and look under table at the mount and look for any movement there like HImy implied.
Actually First I'd Take the Gib's out completely and look them over real good and make sure that something didn't go seriously wrong like Galling and file off the bumps from using the grubs, In fact Take them out and store them in tool box permanently ! The simple act of this one thing may cure your problem altogether ! The torque needed to twist these ball screws without moving the axis is not as much as we might think and If your judging these backlash numbers by the DRO then Its Torsional motion being represented by the DRO.
as far as the stop and go with the backlash on, go to general config and set the pintle accel to lowest dec. acceptable, I have .00001 There are a few other settings that could have an effect but this one baffled me for a while, without understanding it is easy to make it erroneous !
One more note about the gib grubs, if the gib were set lightly, and then grubs were brought in and nudged into side of gib, then you would have two quarter size points of friction right off, then try to adjust gib and it won't move because of the indent from grub or ball, which would give a tight slide but still able to rock slightly. If the grubs had been tightened to test the system then those indents will be serious and any indent will produce an outie ! If the grub had been set very tight then it could distort the gib in that area with a discernible warp or bow. Take them out and loose them in tool box !

12. Cruiser,

excellent points, i have been using the grubs as a method of tightening the Gibbs (in fact the primary method) just as you have described, they are loose, and set using the grubs. After i take the gibbs out and ensure there are not craters from the grubs, I will use the gibb position screws to do all the adjustments to elimnate those friction points. this makes perfect sense to me as to why my gibbs may be creating too much friction.

I will task myself tonight (or as soon as practical) to measureing the before and after results.
The torque needed to twist these ball screws without moving the axis is not as much as we might think
I couldn't agree more, in fact this was one of the first things I noticed about the system and was honestly surprised that the error was not greater.

This discussion has led me along the path of thinking it may be benificial to move the encoder for increased accuracy. (at least for the z axis where the weight of the head is the biggest contributor to the torsional twist related "backlash")-- try it, turn the machine off then try to move the head to the top of the column by manually turning the drive pulley, you will rapidly become annoyed that the drive pully springs back half way each time you move it , your servo has to do the same thing!--
if the encoder was moved to the free end of the ball-screw, it would have the same effect as moving the encoder from the drive-motor to the ball-screw. This move eliminates the error associated with the belt. A move to the free-unloaded end of the ball-screw would eliminate the error of the ball-screw, further elimnating feedback associated error by removing the sensing equipment from the load path of the driving equipment.
I will keep everyone informed on the merit of this task...(this may take a while) some may not care about .005.. but I am too anal to let it go LOL.

Thanks again (to all!!) for the excellent suggestions, I will keep updating on my findings.

Page 1 of 3 123 Last