Eliminating Backlash, Part 2: Refinements
If you're looking over this page, I'll assume you have a machine already converted that has the "good" parts (like ballscrews), you've gotten rid of most of your backlash (let's say you're down to 0.005" or less), but you still have too much backlash. If instead you either, wonder why you should eliminate backlash (or how much can your stand), you want to build a CNC machine from scratch, or you are converting a manual machine that has measurable backlash, try Part 1.
I Installed <Insert "Good Part" Here>, And Still Have Backlash. What Should I Do?
First, make sure you do in fact have the Good Parts on your machine and that they are installed in the most advantageous way. If you have any doubt, you might just quickly peruse Part 1.
Okay, so you have the good parts and still have backlash. How to go about diagnosing? Simply put, there are just a few areas that can be sources of the backlash on a given axis:
Let's look at each one, describe how backlash can develop there, and discuss what you might do to eliminate that cause if you diagnose it as a problem.
Let's start here because it is the easiest to work with and likely the source of a fair amount of your problem. I'm assuming you're not running commercial ball bearing linear slides, but have conventional ways of one kind or another with adjustable gibs. Start by making sure the slideways are properly lubricated. I hate to leave out that essential point, but while we're on it, be sure your ballscrew, ballnut, and angular contact bearings have proper lube as well. Remember, any unnecessary friction can translate into forces that want to bend or deflect something and can lead to backlash.
Unfortunately, adjusting your gibs is a matter of art coupled with much trial and error. I will try to provide some insights. The recommendation is to run the gibs as tightly as possible but no tighter. Easier said than done. If they are too tight, that yields the unnecesary friction that then leads to flexure or stick/slip which leads to backlash. If too loose, there can be slop in the system which translates to backlash and in the worst case, binding and more backlash. In the end, you need to creep up on it by gradually tightening the gibs and taking backlash measurements at each tightening. The backlash should gradually reduce until you've gone too far, at which point it'll jump back up. Back off in small increments and try again to find the sweet spot, being better informed about when to stop the next time around. I have often wondered if measuring the torque applied to the adjusting screws wouldn't be a way to make this a little more systematic and scientific. It can take you literally hours to get this right the first time, but it's worth it if you're chasing out the last increment of backlash.
Of critical importance is balancing the forces if you don't have tapered gibs. Tapered gibs have a single adjustment screw that varies tension along the whole length of the gib. Non-tapered gibs use multiple adjusting screws along the length of the gib, and the challenge is to get them all about even. Again, I wonder whether a digital torque wrench wouldn't be a God send to balance that all out properly.
Trial and Error: Tighten, Measure Backlash, Rinse and Repeat
In terms of more analytical approaches, I have two suggestions. First, you can simply adjust your gibs while systematically measuring backlash and quit when you have as little backlash as possible and it is still possible to turn the leadscrew. Note that minimum backlash may not turn out to be at maximum gib tightness, so that's why we're measuring backlash each time we tighten the gibs.
Set Gibs Based on Slop Perpendicular to the Axis
A second analytical approach is one I was told is the Bridgeport factory procedure (and is also recommended by Fadal) for setting up the gibs. Use a 0.0001 reading indicator and measure the slop in the slide. Example: For the X axis, place the mag base on the end of the saddle and put the stylus on the table. At that end of the table push and release. Then pull and release. The differance is the amount of clearance in the slide. Repeat at the other end of the saddle. Adjust gib in a like new machine with little wear to give a reading of 0.0005 (note, Fadal recommends 0.0003", while Pyramid, a rebuilder, recommends 0.0004"). A machine with more wear will have to be checked with the table closer to the end of travel. The same procedure is used to set the saddle to knee gib. There must be some clearance for the oil film and that film also helps dampen vibration. On a machine with hardened and ground box ways and turcite on the moving member the procedure is to set the clearance to almost nil. 0.0001 is a good number. This is again a practice of checking backlash while setting the gibs, the difference is the Bridgeport factory knew what backlash to expect on a newly manufactured machine.
Prototrak wants you to adjust their lathes so the cross slide has not more than 0.001" of slop perpendicular to the axis.
Use Your Load Meters
Now suppose you have a CNC machine. Are there load meters on each axis? If so, you have a shortcut to consistent gib adjustment because the load meter will tell you how tight you have them. I plan to install load meters on my upcoming IH mill CNC conversion for this and other reasons. The load meter is just an ammeter on the axis DC supply before it gets to the driver board, so it wouldn't be hard to add these. One account I read suggests setting the gibs so that your axis load is about 30%.
Use a Torque Wrench
Southwestern Industries (ProtoTrak) sets the gib tightness on their CNC lathes using a torque wrench. They recommend 15 in/lbs of torque be all it takes to turn the ballscrews.
Linear Slide Adjustments
If you are running ball bearing linear slides, and you have two of them on an axis, are they truly parallel, or are they binding up because they're not? The latter will result in flexure if you overpower it with a strong motor and leadscrew combination. You've got to get them parallel to an acceptible standard,
Leadscrew & Nut
The gibs are adjusted to best effect, and you're still on the hunt for backlash. What's next?
Is everything bolted up tight with no play or flexure back to the machine? Just check it out carefully, perhaps even disassembling and reassembling to make sure everything is torqued well. Make sure the leadscrew runs parallel to the direction of axis travel, or you're going to get binding at some point that may force flexure and therefore backlash into the system. If you have access to do so, try to place your indicator against the mounting points of the ballscrew and ballnut in order to check for small amounts of flexure where there should be none. If you find some, you either need a beefier bracket, beefier mounting method, or less friction in the system (gibs too tight? everything lubed properly? ways in good shape?).
You can also try setting the indicator's magnetic base on the table, and the indicator tip on the ball groove of the screw. Try to move the table by hand. You should not be able to move it far at all (<0.0005" on a commercial CNC machine). If you have too much play here the ballnut to ballscrew connection isn't tight enough (need preload? ballscrews or nut too worn?) or the nut mounting may be loose.
Do you have a beefy enough leadscrew? This is largely a function of diameter. Check out what companies like Hiwin are using in their Bridgeport conversions. Did you buy a little wimpy ballscrew intended for a much lighter application? If so, it is probably flexing when you try to generate too much force with it.
I'm hoping you find nothing of value here, because if you do, and you change anything, you have to go back and re-adjust those darned gibs. Doh!
Leadscrew and/or Nut Mounting (aka OMG those bearings are expensive!)
Okay, we spent a lot of time on this in Part 1. But maybe you bought cheaper bearings than will suffice for your application. You might want to save this for last, because if all else fails, you will face the expensive proposition of upgrading the (probably already expensive bearings) to some better (even more, possibly much more) expensive bearings. Try to come up with a way to measure just the play in the ballscrew versus the bearings. Disconnect the nut from it's mounting, position a dial indicator to read the end of the bearing end of the ballscrew, and try to push and pull on the far end. Can you read any play there? Given some of the preload numbers we've talked about (125 lbs to 500 lbs), you might need a fair amount of push/pull force to tell. Do the best you can. If there is too much play here, you may have a problem with the ballscrew mounting, very likely the bearings.
You can also run a test with the ballscrew turning. Most ballscrews with have an indentation in the end. You can rely on the end to be true, but you can place a ball bearing in the indentation and then indicate off the ball bearing. The ball bearing can be held in place with a little dab of grease or superglue. This is one of the best ways to measure the motion of the ballscrew in its bearings.
Make sure the bearings are installed properly. Are they facing each other properly in the DF configuration or is one flipped? Check the torque on the lock nut for proper preload. What about the mounting block for the bearings? Are they tight in there, or can they move around? Check the spacers, shims, and fits of it all. Remember we talked about how in the most precise designs a lot of this requires ground precision?
How about the ballscrew and ballnut? Is there play there? Try the opposite of the prior measurement. Leave the ballnut mounted, but remove the bearings supporting either end. What kind of play is there? Is the ballscrew used? Could it be worn? If so, consider reloading some oversized balls, but start in small increments. BTW, there are outfits that will do this for you at a nominal charge and it is a pain, so consider using one. If you have double ballnuts, how about adjusting the preload?
Motor Drive Mechanism
Spring couplers have a bad reputation where backlash is concerned, especially if a lot of torque is required. Oldhams are better if you are direct driving, but a timing belt drive is the best. Make sure your belts are not overly worn and are tensioned properly. Are both pulleys mounted well on their shafts? Commercial CNC machines will use very tight keys and a taper lock onto the shaft.
Are we talking about an exotic drive using gears? Fraught with peril from a backlash standpoint. Is it at least a harmonic drive?
Rack and pinion? Very low precision. Try to preload the pinion on the rack and hope for the best.
If you have a servo related system, don't overlook the possibility that the tuning needs tweaking or that the encoder has some degree of mechanical backlash relative to the motor for some reason even though the motor has no backlash with respect to the screw itself. Perhaps the mounting for the encoder is loose in some way. If the encoder mounting can move or flex, it will affect the feedback it is giving about what the motor is doing, and this can be a phantom source of backlash even when all else is working well!
Machine Flexure and Rigidity
If you are feeding too much force in for the machine, something is going to flex. It could be your ballscrew, the mounting for the angular contact bearings, the ball not mountings, or some essential part of the machine that is supporting things. If you get hunting backlash below a thousandth, all this has to be considered. Try some lower speeds for your axis travel and see what effect that has--less force is fed into the system at lower speeds. Try to devise some ways to measure flexure with your dial indicator. There's a weak link somewhere there and you need to find it and beef it up.
I read an account one time of a machine that had a cracked ballscrew mount. The suspicion was the riggers cracked it moving the machine. It took a lot of force to make the crack flex, so this fellow took quite a while to track down his backlash, but he eventually did find it.
Best of Luck on this Part!
I know, it's a finicky and painstaking business, and a bit of a black art. Try to follow the path of the forces: the motor turns the ballscrew, which is held in position by its bearings. The ballscrews moves the ballnut, which is held by it's mounting bracket. Said mounting bracket transfers force to the axis to be moved, and that axis is travelling in ways that apply forces to keep it aligned along a proper and true direction of travel. If all of those things are happily working in concert with the minimum required forces, life is good. If they're fighting amongst one another, that fight is force applied to no good end, and it could well result in backlash. Your job is to track down those little fights and put a stop to them. Approach it as a systematic program of tests and experimentation. Keep a notebook. Measure everything. Try it as many ways as you can think of. Try to think logically about what it all means and how it all fits together. At some point, you will decide things are good enough. Be happy at that stage!
Do you want to be a better CNC'er?
Get Better Tool Life, Surface Finish, and Material Removal Rates.