This will be a many-part series covering my efforts at building a CNC Rotary Table (RT). Please forgive any obvious mistakes. Yes, of course you are going to get my biased opinions. Yes, there will be photos and drawings. No, there won't be detailed plans: they are not really the point of this exercise. Updates will happen every few days.

The background to all this is the rather long thread here at CNCZone about a Backlash-Free Rotary Table, started by Zoidberg
from Sweden in Jan-2009. It went for 80 pages, which is not bad. There were some passionate arguments about technology, especially on just how you can get (make) 'backlash-free' reduction in a home shop or hobby context. Shall we summarise by saying 'opinions differed'? But we all had fun.

Attachment 270536

Let's remove any doubts right up front: this is NOT about hanging a stepper motor on the crankshaft of a manual rotary table. Such tables do not have the bearings for continuous CNC operation, plus they usually rely on manual brakes to stop them from moving under load. As a manual indexer they may be fine, but they are NOT a full CNC Rotary Table.

For the most part, the discussion was about the means for gearing the motor rotation down to to the table shaft. Arguments pitted ordinary gears against worm drives against toothed belts. There were some esoteric variants, but generally they were just
'variants'. Some felt that all the options have backlash, while others (myself included) argued that modern toothed belt profiles
do not. A few felt that gears could be made backlash-free, while others argued for special worm drives. Various rather expensive
commercial variants were brought into the discussion, but it was generally conceded that they are unrealistic for a hobby
construction. The issue was never really settled to everyone's satisfaction.

Attachment 270538

The really big problem here is the rather large reduction ratio needed between any drive motor and the table itself in order that
fine positioning can be had at the rim. Of course, if you don't want fine positioning at the rim then life becomes far simpler, as
shown here. OK, maybe that is a shade too simple? Mind you, for pottery it may be just fine.

Attachment 270540

If you have really simple requirements you can even dispense with any reduction, and just go with a chuck on a stepper. That might be adequate for woodworking for instance. In this context, the motor options we normally have for a hobby construction are either a 1.8 degree stepper motor or a DC servo motor with a 500 line encoder. There are others of course, but these would be the most common.

To give a concrete example, let's consider a 10 micron accuracy at a 50 mm radius. This 10 micron accuracy 'sort of' matches the linear accuracy one might expect from the XYZ axes of a good (hobby) mill. To get that sort of resolution (not accuracy)
requires an angular resolution of approximately 0.0115 degrees (0.69 arc-minutes) at the table. A 1.8 degree stepper would
require a 157:1 reduction to get there (ignoring microstepping for now). That is a lot of reduction. It also places some very tight
limits on allowable backlash, and that is where the trouble lies.

The problems with ordinary gears are fairly well known. First, they have to have some clearance in order to turn, even with the
very best design and manufacture, and this clearance makes for backlash. Second, practical size constraints within a rotary table mean that gear ratios of more than (say) 5:1 are unlikely. To get the required reduction with gears would mean a rather long gear train, and the concensus was that the backlash at the end of such a gear train would be ... bad. To be sure, not everyone agreed.

This does not even consider the problem of getting a perfectly concentric pitch circle on each gear. To be sure one can get very
close, but even 'very close' leaves room for backlash.

A worm gear reduction of about that much is far more possible, but even so there are real problems with backlash. You still need some clearance so the worm can turn against the wheel, although the clearance could be very small. What might be more difficult is (again) the machining error found in the pitch radius of the wheel. Any slight variation in effective pitch radius would mean that the table could be rather tight in some positions and rattling around in others.

It was suggested that you could have a spring-loaded worm driving the wheel. That is true, but there are now two design problems to deal with. First, you have to have the worm really tightly constrained along its axle. That means good miniature bearings.
Second, you have to allow the worm and its support bearings to rotate slightly around some pivot point (or axle) without the
slightest trace of 'rattle' in the rotation bearings. That puts a lot of precision bearings in a very small volume. It's possible, and I did spend some time on such a design, but it is difficult. Well, I thought it was difficult, anyhow.

Of course, since you have a sliding movement between the worm and wheel, there are also problems of surface wear and lubrication. To get lubrication between the worm and wheel you need clearance, even if it is only a few microns. That means some backlash. No lubrication and you get much faster wear. Not easy. Yes, I know most small manual rotary tables use a worm and wheel, but they don't get spun as much as a CNC rotary table would (so there's less wear), and they all come with quite powerful position clamps at the rims, which have to be used when machining anything on them. In effect, they are really indexing tables, not full dynamic rotary tables.

Attachment 270542

Finally, we have toothed belts. The old (or original' designs of 'timing belts' such as L & H and variants thereof had all manner of gaps between the pulley teeth and the lugs on the belt. Yes, that meant you could have backlash, and variable backlash at that.
However, we have come a long way from the early L and H profiles. Modern profiles such as the HTD, AT and GT series are far better, and they have been designed expressly to zero out the backlash. It's all in the belt profile and how it meshes with the pulley teeth you see. Googling will get you lots of vendor explanations, so I won't expand any further.

Attachment 270544

Belt stretch has similarly been improved with steel cores (the string inside the rubber), fibreglass cores, Kevlar cores and carbon fibre cores. The belts are meant to operate under some tension (to seat the belt teeth into the pulleys) and they can deliver a surpisingly large number of kWatts are very high speeds. This means you could string a couple of toothed belts together and
still have negligable backlash on a rotary table. However, you would still have the problem of that 157:1 reduction ratio. A
realistic limit for a single stage of reduction in the confines of a rotary table might be 6:1. Three high-ratio stages would be needed, and you would have to tension every one of them. That's not so easy either.

Despite all these problems, I decided I really needed a rotary table, and it had to be better than a converted manual
indexing table with lots of backlash. OK, make that 'I really WANTED to make one'.

Before closing this chapter, I had better draw a distinction between resolution, accuracy, linearity and repeatability. I will use
angular rotation here as it is most relevant.
* Resolution is the minimum step the hardware is capable of doing. You might for instance have a resolution of 0.001 degrees.
* Accuracy is effectively the difference between what you ask for and what you get. In this context it is a bit complex, so we will
pass on for the moment. Suffice to say that resolution is often (usually) finer than accuracy.
* Linearity is part of accuracy: does a command for a 1 degree rotation give you exactly the same movement when the start angle is 0 degrees compared to 180 degrees for instance? Usually the answer is 'close but not exactly'. An off-centre drive pulley can for instance cause cyclic non-linearity.
* Repeatability is another part of accuracy, but it pulls in the concept of hysteresis. If you tell the system to go to +90 degrees, then have it wander off for a while doing things, then again tell the system to go to 90 degrees, repeatability asks whether it come back to exactly the same position as before. What the target angle is does not matter here; it's whether you get there or not. For instance, if you approach the 90 degree position from two different directions (CW and CCW), hysteresis usually means you do not get to exactly the same position. So repeatability is complex as well.

In the next few Chapters I will discuss my options for bearings, housings, spindles and so on. After that comes a provisional
design, a few major changes (which altered the entire project), further engineering details like motors, and so on. As part of the
series I will go into measured performance specifications.

Cheers