More slop than a school dinner

I'm a bit pissed off to be honest. Although the Shiz has clearly been looked after very carefully over the years, I know it had about 23 years of use before it was laid up for a further 10 years. 

I have now replaced the X axis bearings and the spindle bearings to address damage due to pitting and / or slight rust. The Y axis bearings looked fine, so were simply cleaned up and regreased before careful reassembly. That seemed to take care of any backlash due to the bearings and the possibility of poorly adjusted belts, loose pulleys etc. 

Furthermore, I investigated and (to some extent) improved the backlash on the Z axis which seemed to be mainly due to movement of the ballnut yoke against the quill, presumably one result of the 30+ years of work. I think I've largely fixed that now.

Yesterday I thought I'd better do some vaguely objective backlash tests to put some numbers to the X and y axis backlash. I have to say that although I was well chuffed at getting the thing up and running and taking a part from CAD to CAM to CNC to metal for the first time (with mostly reasonable looking results), I was less than overwhelmed by the patchy surface finish of some of the curved surfaces formed by the X and Y movements. I'm no trained eye when it comes to CNC machines but even I can tell that it's the result of backlash (lost motion) between the servos and the table. So let's get calibrated here and see what we are dealing with.

Ideally I'd load the table up before doing these backlash tests, perhaps by simply tightening up the table locks a bit. However, I thought I'd start by simply measuring the backlash with the table unloaded and rely initially on just its friction / inertia. After all, that's what most willy wavers seem to do when they are boasting about their machines on Youtube and the various metalworking forums....

Simplest way seemed to me to be:

• Position the table and quill at a suitable position and then use MDI to select G54.
• Zero all axes (still in G54) and then set up a DTI aligned to the desired axis, to pick up movement in that direction. I started with X not surprisingly and used the coarsest DTI - the "Kurt" branded Chinesium thing with 0.001" resolution. The Baty 0.01mm and Mitutoyo 0.001mm jobbies might be a bit optimistic at this stage.
• In MDI:
• G01 X0.5 (move 0.5mm to the right in X direction)
• G01 X0 (move back to origin in X direction)
• Zero the DTI carefully
• G01 X-0.5 (move 0.5mm to the left in X direction)
• G01 X0 (move back to origin in X direction)
• Measure the difference from when the DTI was zeroed.
• The difference in zero readings is the backlash, measured from the table moving one way and back to zero, then back the other way and returning to zero. Approaching the same alleged zero from each direction will unfortunately result in a different table rest position due to said backlash.

Finally, issuing a G01 X0.13 following by G01 X0 repeatedly would result in the needle just moving each time. A lesser value would usually not move the needle. More would consistently move the needle quite noticeably. That is a quick way of demonstrating if the measured value of backlash is about right.

Oddly (and disappointingly) enough, I found a very similar value for the Y axis. Note that these measurements were made near the middle of travel, where the machine probably spent most of its operating lifetime. If I were being really methodical and anal, I'd have repeated this test at different positions along the travel. 

The bummer is that my measured value of 0.13mm is 0.13/25.4 inches or about 5 thou in old money. Bugger and f*ck - that is just completely shite, hence the despondency. Unless I am happy to put up with this, which I'm not, I will have to do something about it. What to do?

Firstly, I removed the Y axis belt drive cover to check if the ballscrew is actually moving axially (against the bearings) when the table is moved. Even with the table locked, the most I could see was less than half a thou, so clearly not a significant factor. 

Then (with the table still locked) I measured the approximate circumferential movement of the 58mm pulley. I found roughly 7mm at a radius of ~30mm, which is equivalent to about 13 degrees. Working back from the measured backlash, 0.13 movement on a 5mm pitch ballscrew is equivalent to about 10 degrees (without any load). That's pretty close agreement, given the rough and ready test method. So it's clear the backlash is almost entirely within the ballscrew and ballnut.

Finally, check that the measured backlash is not just caused by the saddle gibs being loose. If there was sufficient slop, you could imagine the table slewing about and appearing to display backlash.

The backlash is clearly between the ballscrew and the connection with the saddle.

I suppose the most "obvious" approach would be to get the table and saddle off, get the ballscrews out on the bench and take a closer look. But to do that would be quite a task. I've already had the X and Y axis drives and pulleys off and that wasn't a walk in the park. I'd have to do that all over again - and much more besides. Besides, the next step would involve some significant and unwelcome effort.

A quick estimate of the table mass focuses the mind. I estimated the equivalent dimensions of the table at around 4 x 40 x 120cm (those approximations allowing for the negative contribution of the tee slots etc), which is a volume of about 38 litres. As cast iron has a density of about 7, that equates to about 270kg. Call it a quarter of a tonne, give or take.

There is also the question of how to get the ballscrew out. Unlike the "normal" AN-S model, my AN-SB model has the leadscrew stationary (fixed to the saddle) and the ballnut moves along it, fastened somehow to the bottom of the table, taking the table with it. Clearly I'd have to remove the ballscrew with the table, after liberating both ends from their bearings. Then somehow get the table onto a bench without buggering the ballscrew by applying the 270kg weight of the table to it. I'd much rather not take that route - surely there must be a less painful solution - or at least one involving less effort.

And then there would be the saddle itself. Removing it simply to get the ballscrew out doesn't actually look necessary - on these Shizuokas, the leadscrew actually sits outside the machine, so it may be accessible enough to service without having to remove it and the saddle.

So, taking the path of least resistance, I thought it best to tackle the Y axis first.

From what I have seen of the Z axis ballscrew and the parts list exploded views of the (very similar) AN-S machines, the ballnut is almost certainly a double nut affair, with a ground shim between the nuts to remove backlash. On the Y axis, it appears that the ballnuts are held captive within the cast lug on the saddle, by means of 2 collars, each held in by 4 hex socket bolts.

Access to the front 2 bolts isn't so bad but the rear 2 are on the almost-impossible-side-of-difficult. Getting an Allen key in there and undoing the tight fixings looked a bit of a challenge. I tried and gave up on the front collar but managed to shift all 4 of the rear collar bolts. That reveals the end of the ballnut assembly. 

I was hoping that I would be able to lock the table and drive the ballnuts out of their bore by turning the ballscrew (by manually moving the toothed belt), then reshim the gap by adding shimstock (about 5 thou, then), before plopping them back in. Beyond that, whether or not I would be able to get the X axis ballnuts into a suitable position to repeat the trick would remain to be seen, even if I succeeded with the Y axis.

*** LATER***

I managed to push the ballnut out of the saddle lug - almost no effort required. Sure enough, the nuts are located between the 2 collars and prevented from rotating by a large keyway. There are a pair of semicircular shim pieces between the ballnuts. Here's one of them:

The key was a bugger to remove. I have to admit that I ended up using a centre punch to drift the end up and out of the slot. Didn't seem to be any other method on the cards, as it was well tight.

With the key removed and one of the shims back in place, tightening the nuts against each other ensures zero backlash. If the shims are the correct thickness, the key will then simply drop down into both keyways. If not, the keyways won't line up. 

Here the shim is trapped between the nuts. Although the key is not engaged with the right hand nut, you can see that it would drop right down directly into the keyway. In this case, they are spot on. Certainly a lot better than 10-15 degrees...

My expectation was that the shims would be a loose fit, requiring some additional gapping to take up the slop but in fact, they were a good tight fit. That was what I also found when I dismantled the Z axis ballnut assembly, so it looks as if the ballscrew wear on these machines is negligible. Pretty impressive after 33 years.

That's encouraging. And it suggests that the backlash movement is between the ballnut assembly and the 2 collars, so that the ballnuts are actually moving axially within their bore. If so, I could deal with that by simply packing out one of the collars to ensure the ballnut assembly is under compression ie fully constrained. That would be a fine result. Best way to find out - rebuild the assembly with some form of packing and see if that has got rid of the movement. I imagine a 0.8mm dia stainless steel TIG welding rod would form a suitable ring-shaped packing piece. Let's find out....