Printable View
when I was testing the touchplate setup on my router, I was getting repeatability to within +-0.003mm. Quick back-of-envelope sums showed that this was equivalent to one microstep. Might this be why this magic number is there? Mind you, my backlash is a bit more than that..Quote:
Originally Posted by Robin Hewitt
As it happens Neal I've just run some more tests. Prepare to fall asleep! Contrary what the fool in the vid says, the Y does seven cycles!
Despite all the grunting (not mine) the steppers don't appear to be losing steps at 2500mm/min.
https://youtu.be/pyuFoJNj5cI
Actually, I don't know what the backlash is yet. I would think it was small but I need to set-up a proper test.
Just measured the back-lash at 0.03mm on each axis. I'm a little surprised by this since I can't push/pull either axis by more than 0.01mm. The motors are coupled to the screws with flexible star-type couplings (see pic) - which are original and in good condition, but I have to say I've always been suspicious of. Now they are robbing me of precious microns!
Seriously though I can live with 0.03mm.
Attachment 22152
I used to tell myself that I could live with a slight backlash but deep in my heart of hearts I knew it wasn't true.Quote:
Originally Posted by Agathon
I move 5um per step. Could I watch it step 6 times without moving the table then sleep at night?
Difficult question...
I'm not sure I can either! After I posted I took another look and the back-lash is at the feedscrew. I think I slackened the pre-load on the X too much and need to tighten up the Y. Watch this space....Quote:
Originally Posted by Robin Hewitt
Decided to strip the ball-screws and repack with new balls:
https://youtu.be/MVdal0BjrpU
Very nice vid. Would you mind giving a link to where you purchased the balls from. TnxQuote:
Decided to strip the ball-screws and repack with new balls:
I have my nuts sprung together one quarter ton. I hold the X screw in tension, Y and Z I crush angular contact bearings together. Everything is one quarter ton and feels glorious.
I sprung my Roland mini-mill to 200 lbf, I don't have a magic formula, I just use whatever seems right.
Do you have a way to inject oil in to the nuts or do you depend on that grease? I started fitting narrow bore nylon tube to carry oil but one of the push fit connectors didn't fit and I wanted to play cutting stuff so I left it out. I think I have to strip back and finish the job. Gravity feed is traditional but would that need wider bore piping? Such fun.
For a moment I was worried that I might have put you off with talk of backlash. Glad to see you back.
Hi Clive, I just ordered them through my local branch of Brammer - 1/8" Chrome steel balls, Grade 100 £10.08 for 500. These nuts have about 0.004" clearance, so I did look into getting balls at 0.127" - but they were very expensive ranging from £58 - £82. As I say in the vid, I realised this wasn't the right way to go with this type of nut.Quote:
Originally Posted by Clive S
The machine's central lubrication system is very comprehensive and delivers oil to the ball-nuts and ball screw in a couple of places.Quote:
Originally Posted by Robin Hewitt
I'll put the x-axis back together later today and see if there's any improvement on the backlash.
Fwiw...
I think You have one of the best manual tools of this type (c frame small mills) ever produced.
Your explorations/experiences/results are similar to mine, on a decent-rigidity heavy chicom 12x" lathe.
Except that mine is about 10x worse in terms of fit, finish, quality in general, of course.
I would opine, while never having seen a Fehlmann in person (is a tour available if I happen to be in Your country ..?),
- repeatability to about 1 micron is very achievable
- resolution to much better than 1 micron is available to You, and probably relatively easy to do
My experience for 0.03 € (I need to make a profit, You know..):
I made treadmill-dc motor-servos with geckos (320) about 2005.
So-so, at 10.000 counts, accurate but noisy, sparks, ozone, some jitter, heating.
I used the treadmill servos (DC motor 180V/6000 rpm, at 68 V dc, so too many amps and not enough volts, but I never need speed anyway)..
and at 1:3 via HTD, 5/15 mm, 1:3 or 16:48 teeth, could reliably index, or incremental-move, 1 micron at a time.
The basic accuracy and stiction and smoothness of the chicom 12x is probably 2x-4x worse than Your mill.
(But it is more rigid. Lathes always are, especially heavier ones, like mine. 350 kg/24" == 2000 KG on a 1.2 m long bed.Maybe not more rigid than a Fehlmann, one of the best ever made.Imho.)
So I built very heavy supports, mounts, belts, using HTD8-30, taperlocks, 17 mm shafts on 220V ac brushless servos, of 750 W, 10.000 counts, at 1:2.
Results are weak in accuracy.
Pulleys are poor, relatively, and belts may or may not be poor, relatively.
But..
Using these relatively-expensive servos, 700€ landed, I do get 1 micron indexing or relative movement.
But it is not smooth, consistent, steady, and no-way no-how a dial-to-size solution.
But..
Theoretical resolution is 0.2 microns, and the fact is, the screw itself *will* move by 0.2 - 0.n micron increments, every single time. Increments, not accuracy.
Screw will always move in angular terms. Some waviness from belts/pulleys., ie it is not perfectly linear/smooth/accurate, transmitted to saddle x axis, with "some" bounces of uncertain size, perhaps 1-2-10-20 microns, depending on gibs, tension, position, oiling.
With light (typical manual) gibs, oiled, imho, about 0.5 microns resolution is perfectly reliable for me. I cannot measure this, yet.
I will, and may put in 0.1 micron glass scales (thats the plan, anyway).
I opine You can easily get 0.5 microns, or better, resolution from one of the best mechanical mills ever made, via servos.
Once pre-loaded on an axis,
Led readouts, on the servos, show errors, which are always zero, at position, because the servo has == 10 Nm / 10.000 counts / rev, or 20 Nm at screw, about 10x more than any possible resistance.
At 20 Nm the push force == 2000-3000 kgf.
The saddle bounces because of these mechanical errors, none of which relate to stepper or servo as such,
- the x mount is not rigid-enough and not mounted in-plane with the screw thrust
- sticktion
- gibs
- poor screw (I knew this)
- yoke connection to saddle is poor re:rigidity. Much worse than I thought, just saw this 2 days ago.
With a heavy load/stiff gibs, everything bends noticeably on the x axis.
Everything bends always, I/one just did not see it well enough in the past.
I got great results with soft gibs, semi-heavy x axis lock, like manual turning, since 2005 or so.
The current ballscrew on x is 14x more rigid than the original 16-17 mm acme screw (fixed-fixed mount, in tension, rigid for 2x, half free length for 4x, thicker for == 2x).
So I expected 6x more rigid belt drives to do better.
They did worse, or same, with much better servos, and more rigid/tighter gibs.
Probably, softer gibs==manual, would deliver better results in resolution. I now think.
All this is related.
I got 20x, perhaps better, higher mrr or Material Removal Rates, than I have ever had, last week.
Industrial level results.
2.5" 63 mm ISO30 face mill, 4 inserts. Cutting thick tool steel, full width.
3.9 mm deep, !!! 63 mm woc ie full face, 45 mm/min speed, 550 rpm.
That is a relatively 63 mm thick end mill, 3.9 mm deep !! for milling terms, full width.
The machine was worked hard, but quite happy, when I achieved balance.
Balance was hard / impossible to maintain.
New face mill, axxx something, new inserts, I just hold the facemill front in the 12" 4-jaw chuck, clocked to 0.01 mm or better tir.
- gibs are now too tight, after adjusting
X-axis Screw is now 0.750 " roton, rolled.
The new x axis screw is TBI taiwan, 32 mm, 5 mm, with a 60 mm thick yoke.
I need major work before it is in, ..
new saddle plate (ends are now milled, hooray !), yoke all 6 faces, yoke bored, supports of no particular accuracy.
My opinion(s);
1. Go to servos.
2. Go direct drive. There must be a reason all manufacturers use direct drive.
3. Use a bigger coupler.
4. Make anything used in the motion-control train very very heavy in steel.
5. Use highest-resolution servos you can, while being relatively economical, and having sufficient speed in khz/mhz for your controller.
Mine is a csmio-ip-s, 4 MHz.
A 3000 rpm/10.000 count servo, is 50 revs/sec, = 500 kHz.
I could not care less about top speed, and may use something near top acceleration, in the future, or not.
Today I use about 1/5 top speed and top acceleration, and both greatly exceed the best stepper systems I used in the past.
"Good" steppers as in fast nema 23 steppers, 68 v, gecko 203v, centipede hw pulser/controller (excellent hw and timing).
At that, the stepper made in tests 10.000 rpm, over 4-6 secs acceleration, with no torque, no-load.
Geckos fault 380 kHz-400 kHz+ or so.
In use, best-optimum std nema 23 steppers make about 600-700 rpm, in == 0.2 secs (lathe, needs acceleration, as much as possible).
Small nema 23 servos, and the bigger nema 34 servos on the lathe, deliver;
== 0.02-0.04 secs to 3000 rpm, at load.
About 10-50x better in acceleration, real-world.
About 5x more accurate.
Imho, accuracy is very important.
Acceleration is very important, or important.
Top speed is totally irrelevant.
what happens is that servos are excellent at many small tiny moves, like 3-d contouring, or modern high-speed milling toolpaths.
Or very,very accurate tiny moves, that steppers cannot do directly.
A stepper at 1/10 microstep (2000 steps/r) has about 1/10 rated torque iirc (gecko, Mariss).
So a 3 Nm stepper has 0.3 Nm at 0 rpm (best case) at 1&10 microsteps. It is, in essence, a spring.
A servo at 10.000 steps has 3x rated torque, from 1 step at 0 rpm to 1 step at max rpm, say 3000 rpm (for the 3 secs max peak torque).
The servo of 1.3 Nm (400 W, similar cost 290€ / axis), 68V DC/AC, has 1.27 Nm cont, == 4Nm peak.
So the small servo has approx 4 Nm vs 0.3 Nm stepper torque of a similar size, about 13 times more, and over 50-100x more torque over 1000 rpm vs a stepper.
All this with a cheap small economical Nema 23 400 W servo, of course.
If comparing to more industrial-type stuff, ... well...
My 750W - 220V ac driven servos .. are exactly the same in use.
Just 10.000 counts vs 5000 counts,
220V vs 68 V,
3.x /10 Nm vs 1.27 Nm.
My 0.2 micron (now) step size lathe has 10.000 x 2 / 5 mm = 0.25 micron resolution, theorical/electronic.
= 7.5 m / minute, 0.125 m second.
Typical free length is less than 100 mm = 10 cm, on z. 24" minus 12" chuck 180 mm, minus ts, == 220 mm, == 400 mm.
About 250 mm free length between 12" chuck and ts at shortest extension.
And plenty length for me, never needed to remove the gap from the bed, or the ts maybe once, minor stuff.
I don´t usually do e. gunstuff, or long spindles, and when I do, easy mounts exist.
My lathe is a technology demonstrator, mostly.
When cutting anything, clearances are usually 2-3-5 mm, for me, now, and less when everything is dialed in.
The absolutely only need for high acceleration /speed in lathes, for me, has always been leaving threading, at the end.
And it is extremely important to be very, very, very consistent, of very high torque, at very high speed in both start-time and pullout-time, in ms, when pulling out.
E.typical.
Threading at 500-800 rpm towards hs, steel, about 1-1.5" D workpiece.
Typical, maybe 8 passes.
Every pass makes the end bit divot deeper, and the pullout point is always deeper, and more rigid, each pass.
Any tiny error or delay, makes the threading tool dig-in, if there is any bend (there is always bend), slop, backlash in the whole x axis drivetrain.
Any tiny dig-in, exponentially increases the error until failure of something, unless the tool is pulled out fast enough, strong enough, to avoid failure.
In this scenario the great benefit of servos is,
-vastly faster acceleration for pullout
-vastly higher force of pullout
-vastly more accurate/repeatable pullout point and action
The relevant part of the pullout is perhaps 0.1 - 0.5mm in length, aka most of the thread depth, and takes maybe 0.1 secs with a stepper, and maybe 0.05-0.01 secs with a servo.
But the servo exerts 10-50x the force, at 10-50x the acceleration, and 5x or more accuracy.
It is of note that a very tiny 0.01 - 0.02 mm mechanical slop with steppers, or machine error, can snap the tooltip, as the dig-in is exponential and mechanical and increases exponentially.
This does not usually happen with servos, because no matter what most of the tooltip is already out of the workpiece, with the same mechanical error condition, simply leading to a slightly rounded pullout edge.
E.
I used 50 mm / 2" thick tool steel for the mount plate for the Z axis ballscrew.
About 140x200x50 mm.
Not because it is "stronger" but because it is very much stiffer, and will repeat better.
I think 10x more rigid than "typical", don´t really know, and it only cost == 20-30€, maybe 15 kg in mass for the mount plate alone.
My z axis screw is 32 mm.
Yours will be similar, I think.
32 mm screw == 1600 kgf push force, rated, static (weakest rating).
Example/anecdote.
This push force is == equivalent to lifting a SUV with the screw.
And You want to try to bend as few microns as possible, while lifting the suv.
A typical 32 mm screw has 54 kgf/um or 540 N/micron rigidity.
I am by no means a "servo zealot".
Steppers are very easy and can provide excellent repeatability, accuracy, force, positioning, of relatively low rpm, very reliably and cheaply.
But they have low dynamic range.
This means either accuracy, or speed/power, but not both.
So the ideal apps for steppers are
e.g.
telescope mounts of low rpm and very high geared resolution/repeatability,
cutter grinders, similar,
saw accessories,
microscope accessories,
cnc mills of typical hobby shops needs, some jobshop uses,
cnc jewellery stuff,
etc.
E.
I have made all my x-axis mounts/stuff behind the lathe .. so it looks stock from the front, and can be used manually.
New screw connects to yoke of 60x120x120 mm.
It will be about 100x more stiff than now (because I use a temp. crap lashup to bolt to the current saddle plate, (was better in the past, pre new servos) of temporary/test use).
With direct drive, I will get 0.4 microns/step vs 0.2, but no wind/spring/bounce from belts, and vastly less from winding error / screw, and much less bend from yoke.
And no pulley error, belt error.
Belt drives with small belts are great for steppers - but the wrong choice for servos. Imho, Imhe.
I have a comment.
I have seen credible comments from guys with experience, that the grease used in milling spindles, had great effect.
Re: heating at higher rpms, power used, etc.
Like 30-50% of the power went to heat, vs 5%, with better grease.
And noise decreased 50% or so, by tfar method.
(That Feels About Right).
Kluber Isoflex 15 is the gold std, for high end machine tool industrial spindles.
It is very expensive.
You need very little, about 1.5 cm3 for a big spindle bearing.
I bought a 50 gm tube, for 1 micron spindles I am making, 4 of, experimental/commercial test samples, with real abec 9/iso 2, bearings in 25 and 40 mm D.
I will make 4 test spindles, 2 of 25 mm and 2 of 40 mm, with both bearings, and see how they work.
Yes, the spindle at 40 mm is now hardened, ground, polished, has 2 microns tir (in spec, just) but I will use a soft lap and diamond paste to reduce the error a bit.
Anyway, better grease has reportedly made a major difference on spindles.
If You want some, I am happy to mail some Kluber "unicorn snot value" grease to You, free.
I doubt the grease has any major effect on accuracy .. but ..
I think it quite probable, perhaps, the grease can/will point out the next error in the chain.
Contact me any way You want, if You want to try some.
One email is greystoneprecision at the google mail dot com.
I do know for a fact, that modern cnc lathes (tools) use P4 bearings or better, since the packaging on mine says so.
And my factory training.
And these bearings do very much benefit from rgw better grease, mostly at higher speeds 2-5k and up.
Personally, I doubt it matters at low speeds, but think I might well be wrong.
CNC machining is often about corner cases, and stuff works different to what one might expect.
The tiny balls in bs supports and nuts actually run really fast at their surface speed.
Anyway, You want a bit, You can have some.
No conditions at all.
Quote:
Originally Posted by Agathon
Hi Hanermo2 and thanks for your very comprehensive and convincing post on your experience of servos. I'm coming to the same conclusion. Steppers are fine for hobby machines, but for anything a bit more industrial a servo is the way to go. Fehlmann certainly dropped steppers within a couple of years and went to servos.
The Fehlmann is a slightly odd machine in that it's a drilling machine with milling capability - so despite it's industrial specification in terms of mass, the design is a compromise, with milling, and then CNC, as an afterthought. With that said they sold well in their home market. The latest generation of Fehlmann machining centres are built round a conventional square slide-way for the Z axis, but they still produce a manual and semi-cnc/manual version of the Picomax: http://www.fehlmann.com/en/products/...picomaxr-21-m/
I was interested in what you mentioned about direct drive. When Fehlmann switched to servos (see pic) they mounted the motors remotely - not sure if it's a belt or gear drive, but suspect belt. The latest semi-cnc/manual of the drill mill that I have also has remote servos - I'm not sure about their machining centres.
Attachment 22222
Having reassembled the ball-screw, replacing it proved interesting since the nuts, or rather the keys in the nuts, didn't want to go into the key-way. The key-way or the keys were obviously minutely out of line with each other. A rub over a stone got them to slide in, but the results from the screw were disappointing - still not sure why. It could be an alignment issue - which seems strange given the precision nature of the table and the screw, but it's the sort of thing I've regularly encountered with precision machine tools - a lot of fettling takes place when these things are built. Anyway, I decide to but the nuts on the in the original orientation which seems better, but I can't get the backlash better than 0.01 without sacrificing the smoothness of the screw. I will tinker some more today.
I've had 20 years experience rebuilding machine tools and generally I've found that even if a bearing's tracks look good through a loupe they may well be slightly rough when under load. This is sometimes due to poor fitting (hammering the inner) and tiny imperfections are left on the bearing tracks which will only show up under a microscope. I suspect that the ball-screw and nut assembly have similar issues. I'm not about to replace them so will live with the 10 microns of backlash.
Thanks for your offer of the Kluber Isoflex. This is a product I'm familiar which since it's specified by some of the manufacturers I deal with. I actually use an SKF LGMT2 which has a similar specification and was more readily available at the time I needed it. In fact the Fehlmann's central lubrication system covers everything from the spindle bearings through to the ball-screws and everything in-between. The Vaseline used was just to get the balls to stick to the nut during assembly - it will soon wash out.
If you're ever in my "neck of the woods" you'd be more than welcome to visit.
I forgot to say that it would be nice to see a photo or even a video of you CNC machine(s).
David
I forgot to ask about the above which intrigued me. What control system are you using? Are you planning on linking the scales back to the controller in a closed loop? Can this be done?Quote:
Originally Posted by hanermo2
Spent the evening trying different permutations of the ball-nut assembly and the best I can achieve on both axes is 10 microns. If this were a conventional manual machine I'd be overjoyed with 4 tenths of a thou! In fact the very expensive Swiss machines that I sell have a backlash from new of around 0.01-0.02 mm.
Anyway, the machine is running very nicely and the back-lash compensation on Mach3 is working well.
As I frequently say to my customers "it's a metal cutting machine not a grinding machine" so I don't expect nor want to machine to tolerances better than +/- 0.005. I certainly don't think it will be a problem in respect of climb milling.
More soon...
Yes, the scales can feed back to the controller, at least thats what the makers say.
CSMIO-IP-S on lathe, from cslabs.
Somewhat expensive, very good, very good value for money.
Even if I only got dros from the scales, it would be good enough for my use via sw macros.
I am aiming for extreme resolution, repeatability, and fairly well willing to spend money.
Quote:
Originally Posted by Agathon
Swiss backlash 0.01 mm new ??
I would have expected your picomax to have zero to 0-1-2 microns backlash.
I suspect the bs fixed end bearings are not properly adjusted/preloaded, or ..
.. your ballnut is loose/failed somehow and has no preload.
Easy enough to test.
Dti screw end, while loading back/forth with a prybar of some type, separately on screw end and on nut end/assy.
One should have a major deviation vs the other.
Quote:
Originally Posted by Agathon
I've checked everything and it's all correctly adjusted - first thing I thought of was movement in the pillow-block bearings but there's no movement here. If I take the ball-nut preload one increment higher on the vernier the screws become notchy. It's probably just down to wear and tear. We'll see what the performance is like in practice.
You are being very vague, is this sprung or a simple crush?
Crush.Quote:
Originally Posted by Robin Hewitt
It was a bit of a bodge springing mine, I added a spacer and 2 Belleville washers. Well 6 spacers and 12 washers to do all the bearings and nuts. But the reward was out of all proportion to the effort.
Attachment 22229
So, if I'm interpreting that correctly, you have a double ball-nut set-up like my machine?Quote:
Originally Posted by Robin Hewitt
I could do something like this although it would mean making a new retaining flange. I have to say that apart from grinding spindles, I'm not a fan of spring-loaded solutions. I'm also not convinced that backlash of 0.01 is much of an issue. My conventional mill has about 0.25 backlash and it's only a problem with heavy climb-milling or slot milling and then it can be overcome by reducing the feed rate and the depth of cut. Mind you the table and slide alone must way 200kg - mass helps, as do well adjusted gibs.
I meant to have asked what the rest of you mill consists of Robin - got any pics?
I don't have a recent pic, here's an old one...
Attachment 22230
Very neat, especially the Z axis. Wouldn't mind seeing some details of this as I'm sure I'm going to have to go down this route eventually.Quote:
Originally Posted by Robin Hewitt
What sort of backlash have you achieved on the x and y?
There's a build log from many years ago, I will try and link it...
http://www.mycncuk.com/threads/651-W...ht=Warco+Major
I fitted zero backlash double nuts, big disappointment, so I sprung them and it was like magic.
The springs mean that I have no backlash up to 1/4 ton but if I ever pass that loading presumably everything will go horribly wrong.
I do get some sideways slop in the quill if I forget to pack it out with thick, icky grease. Open to suggestions.
I might give the belleville washer thing a go. The Fehlmann's vernier adjustment is clever, but I think a grub screw lock would be better as it would allow finer adjustment. Quality manufacturers have a nice trick of fitting a pressed-in brass plug in the adjustment nut before threading it so that you have a formed locking piece that doesn't damage the thread and needs little pressure to lock.Quote:
Originally Posted by Robin Hewitt
Regarding your quill slop, there isn't really a satisfactory solution. The fit of a cylinder into a bore is something that requires a degree of precision and careful fitting at the point of manufacture - surface finish is critical. Any fool with a cylindrical grinder and enough patience can make a cylinder to size and parallel to a micron or two, but getting the bore honed for a perfect transition fit is another issue altogether. Most of the Chinese machines fall at this point, although I have to say that Myford's VMC (Taiwanese) had a pretty decent fitting quill - though not to Swiss, German or the best US and GB standards (has to be said though that most of the machine tools produced in the UK were not of a particularly outstanding fit in this respect).
You could split the casting and put a clamp in place, but this is a pretty crappy solution as you're going to get lots of point contact at the clamp position (the bore will be an oval) and you'll still have the original clearance further up. Alternatively you could get the quill hard-chromed and ground to a nominal size slightly larger than the clearance, but then you have the difficult honing issue mentioned above. Not worth doing either of these, better to start again with a higher-grade machine and transfer your clever CNC additions to it.
If you've got space there's been some fantastic 1980s CNC machines on eBay recently including a couple of Deckel FP3/4s for about £1500! Fab things with superb ball-screws and servos (probably old brush types though).
I meant to have complimented you on the really super job you've done on the casings for the drives - very professional.Quote:
Originally Posted by Agathon
Most kind. This is the magic of CNC, it frees you from straight lines and you enter a wonder land of flowing curves. If I only had 2 more axes I could do blobs.
I have some parts from an older machine looks like PICOMAX100 that was being disposed of by a machinery dealer .Quote:
Originally Posted by Agathon
It used sevo motors and timing belts to drive the ball screws .
when I arrived the controls had been chucked in a skip , the sevos removed and the intention was to strip the rest to get rid of the 3000 K that was in the way .
I still have the sevos , the SF 32 spindle , some SF30 tooling , the roller slides , massive leadscrews ( maybe 30mm ) with ball nuts and bearings .
So true sevos when others used steppers ! - well as you say , it IS Swiss.
What a shame it was scrapped - I would have considered buying it.Quote:
Originally Posted by C1Geoff2020
I've only just started working on the old Fehlmann again after a couple of years of inactivity. My machine is a pain as it was originally only two axis. The Picomax 100 was a three axis machine that used the quill rack and pinion driven from a worm and wheel belt driven from a servo as the Z axis. It sounds like a terrible idea when one thinks about the average drilling machine rack and pinion, but being Swiss there's virually no backlash. As I mentioned earlier in this thread, the extremely clever engineer I bought the machine from had made a worm and wheel drive (the aluminium housing seen on the right hand side) to drive the pinion. However, I have found that because he used a fully throated worm wheel it has been almost impossible to remove the backlash.
Attachment 27456
I will probably end up using a ball-screw to move the quill, but I have recently reinstated the machine's original worm and wheel drive (which employs a gashed bronze worm wheel) and have backlash of about 0.01mm - not bad. I shall see how it works in practice once I have mounted the stepper motor to the original 90º cross-helix gear drive to the worm shaft. There is significant backlash in this gear, but it will be unimportant in terms of the ability of the quill to remain in position because of the worm and wheel - the backlash is constant so it can be accounted for in the software. As I say, this is almost identical to how it was originally done on the three axis machines, however, it has proved difficult to employ a belt drive to the worm shaft due to differences in the head castings - see drawings.
I should have said Geoff, if you want to sell any of the Fehlmann bit - especially the SF32 - tooling let me know. I've PMed you.
David