X3 mill with rails, ATC, new motor

Lately I've been working on converting an X3 mill, for myself. The eventual plan is to put linear guides on all axes, add ATC to the spindle (using similar method to the Tormach system) with a moving tool rack and replace the spindle motor with something better. To start with I'm working on the Z-axis and ATC - X and Y can wait.

Starting with the motor - this is the one I chose:

Attachment 12613

It's rated by the retailer for 6.5kW, but that's not a realistic rating - it's more like 4-5kW. That power rating is excessive, but the correspondingly high torque rating does mean I can get away without needing to change the belt. I've started rewinding it to control it with my servo motor driver using vector control - so things requiring position control of the spindle will be possible, in addition to getting a large speed range (using field weakening). That'll also allow me to test the idea of linking feed-rate to torque feedback, since the driver will know the torque at all times.

The motor clearly doesn't fit through the cutout in the side of the mill head, so I milled a hole in the top, then cut the annoying protrusion inside off with a slitting saw:

Attachment 12614Attachment 12615

I'll machine an aluminium plate the size of the head to both support the motor and heatsink it. I've calculated that a 5M, 25mm wide HTD belt will be adequate (just about) to couple the motor to the spindle so I've ordered that and some pulleys.

I decided to put linear guides on the column to remove the various (potential) problems with the dovetail slides. To do that I've milled a small amount off the dovetail on one side of the column, then drilled and tapped lots of holes to fit some ground strip upon which the rails the rails will be mounted.

As I found a long time ago when converting the other mill, cast iron cuts well on my CNC router, so I used that to do the necessary milling and drilling:

Attachment 12617Attachment 12618

Drilled array of holes in the back of the column so that I can mount a piece of 20mm thick aluminium to the back of it to improve the torsional stiffness. The ballscrew will be on the front between the rails, so no need to have the column open. I can also conveniently mount the motor drivers inside the column and use the aluminium plate as the heatsink.

Attachment 12623

Now to drill the rail strips - align the vice then drill, pretty simple:

Attachment 12619Attachment 12624

For some reason people seem to like taking pictures of sawing things...so why not:

Attachment 12621

Ready to mount rails - just need to check the join on surface plate, but it seems OK.

Attachment 12620

Simple aluminium plate machined to connect bearings to head:

Attachment 12616Attachment 12622

I've also made a start with the ATC system, but the pictures I took have gone walkies so I'll just describe it. It's basically just a copy of the Tormach system, using an MT3 3/4" collet in the spindle, linked to a drawbar with a stack of disc springs. The springs are pushed with an air cylinder to release the tool. It seem to work fine, but not tested thoroughly yet. The mildly interesting bit will be making the tool rack. I've bought another air cylinder to move the rack into position automatically and the rack can be spun into position with a stepper motor.

Edit, found one picture - here's one of the tool holders clamped in the spindle. It's pulled against a ground plate (which I'll hardened) to get the height repeatable and increase the stiffness:

Attachment 12625

That's about it for now, as I've only spent about 2 weekends on it. I've ordered the ballscrew, which apparently will take about 15-20 days to arrive (trying a different supplier, more on that later), so in the meantime I'll work on the servo driver design, finish rewinding the motor and machine the remaining trivial parts for the head and Z.

I cut out the back plate for column, will leave the tapping for a rainy day:

Attachment 12631Attachment 12627Attachment 12630

The spacing of the bolts is loosely based on the theory here - namely make the stress cones under the bolts intersect to get maximum stiffness. This diagram illustrates that:

Attachment 12639

Made a ballnut mount for Z:

Attachment 12633Attachment 12632Attachment 12636

Remembered the cutout for the ballnut mount in the Z-plate. Clamped directly to mill bed to ensure parallelism.
Attachment 12637Attachment 12638

Here's a picture of the drawbar I made earlier for the ATC:
Attachment 12628
That may not be the final number of disc springs - but there will be quite a lot to ensure they're not operated outside the recommended region to get long life.

Excellent Jonathan - this is something I will also be attempting soon, so will be following with interest.
.
BTW if you are going to use linear rails/bearings on all X and Y and using the "fill in" the dovetail method you used on the Z, are you going to lose ~60mm in the Z due to the height of those units?
.
Cheers
.
Chris

Quote:

---

Originally Posted by Washout

BTW if you are going to use linear rails/bearings on all X and Y and using the "fill in" the dovetail method you used on the Z, are you going to lose ~60mm in the Z due to the height of those units?

---

Something like that, yes. I could get the Z-height back again by adding a spacer block under the column, but clearly how that affects the strength needs to be considered. Also, I'm tepmpted to replace the existing table with something bigger...

Really cool,

We defo needed some cnc "mill conversion" logs on here, loved the bolt spacing theory so thanks for sharing that, cant wait to see the rest, keep up the good work !

EDIT: Just a quick question, how come you went with ali for the column back plate rather than steel?

EDIT 2: I dont know if it will apply to you, but i recently watched a build log where the guy filled in any voids in the base and so on with a mix of stone and epoxy, will you be doing that as well or?

.Me

Quote:

---

Originally Posted by Lee Roberts

EDIT: Just a quick question, how come you went with ali for the column back plate rather than steel?

---

1) I can mount the motor drivers in the column - the aluminum plate can be used as a heatsink, so saves buying that.
2) Aluminium is easier for me to machine.
3) I had a 1.50x0.74x0.02m sheet of aluminium 'lying around' and I'm not sure where to buy steel plate for a reasonable price.

Quote:

---

Originally Posted by Lee Roberts

EDIT 2: I dont know if it will apply to you, but i recently watched a build log where the guy filled in any voids in the base and so on with a mix of stone and epoxy, will you be doing that as well or?

---

I'm considering it - it seems worthwhile, but I might be better off just making a new base.

Quote:

---

Originally Posted by Jonathan

1) I can mount the motor drivers in the column - the aluminum plate can be used as a heatsink, so saves buying that.
2) Aluminium is easier for me to machine.
3) I had a 1.50x0.74x0.02m sheet of aluminium 'lying around' and I'm not sure where to buy steel plate for a reasonable price. I'm considering it - it seems worthwhile, but I might be better off just making a new base.

---

Ahh fair enough then.

I looked at a build on the zone the other day re making a base, on that build, they used a lump of granite for the base, what are your thoughts on doing that?

.Me

What sort of motor are you using for the spindle?
And how on earth does it achieve such a high power density?

Quote:

---

Originally Posted by embraced

What sort of motor are you using for the spindle?
And how on earth does it achieve such a high power density?

---

It's the same type of motor as I used here - permanent magnet synchronous motor (PMSM), with an outer rotor. The high power density partly stems from the magnets, which replace the rotor magnetising current in, for example, an induction motor, resulting in lower losses. There is more to it than that though - e.g. differences in cooling.

This is how far I've got with rewinding the motor - two phases done, but I do have a low resistance from one phase to the stator, so I'll probably have to re-do that one and be more careful.

Attachment 12661Attachment 12660

Cut out the motor mount for Z from 20mm thick aluminium yesterday:

Attachment 12669Attachment 12670Attachment 12659Attachment 12658

This morning I cut the motor mounting plate for the spindle motor:

Attachment 12657Attachment 12666

This is where it fits:

Attachment 12665

The large pulley (60T) isn't all the way on the spindle shaft, but this shows the general idea. The spindle pulley is keyed and the motor pulley fixed using four grubscrews, which should be sufficient...more on that later.

Attachment 12664Attachment 12668

The spindle pulley is much bigger than the original gear, so I had to mill a bit off the top cover to clear the pulley and belt:

Attachment 12663Attachment 12662

Z-ballnut mount (again):

Attachment 12667

I just need to make the Z-ballscrew bearing mounts, finish rewinding the motor and make the driver, decide if I need an encoder on the spindle and then I can continue with the ATC.

a cracking thread again jon.
im looking forward to seeing how this performs,thanks for posting ;)

Quote:

---

Originally Posted by deisel

im looking forward to seeing how this performs,thanks for posting ;)

---

You're not the only one!


I finished rewinding the motor the day before yesterday, so here it is ready to varnish:

Attachment 12698


Checked the back-emf constant was correct by spinning it in the lathe and measuring the peak voltage with oscilloscope. Then powered a small light bulb with it for general amusement...


Attachment 12699Attachment 12697

I tested the motor with my driver last night, just open loop control - will post a video if anyone's interested.


I started a job at university yesterday, so progress will be slower for now. I'll instead concentrate on finishing off the motor driver design and making them.

Quote:

---

Originally Posted by Lee Roberts

loved the bolt spacing theory so thanks for sharing that, cant wait to see the rest, keep up the good work !

---

Since you seem to like bolts, here's my justification for putting four setscrews in the motor pulley...

A setscrew works by applying a force which results in a contact force between the opposite side of the shaft and bore. Based on the tightening torque of the setscrew, we can calculate this force and if the co-efficient of friction between the shaft and bore is known, the holding torque can be determined - approximately of course.

For one screw it's very simple - the axial force on the screw is equal to the contact force (C) between the shaft and bore, so:

.
.
.
.

.
.
.

The motor shaft diameter is 12mm, we can fit M5 grubscrews, they're black and according to this list the tightening torque is about 4.7Nm, so substitute that lot in:

.

With multiple screws, we have to resolve the forces to find the effective contact force. Define the angle as the angle between the two screws, then resolve the force on one screw in the radial direction and double it to get the radial force for both screws:

.

I put the screws spaced at 90° because it makes milling the flats on the shaft easier, but this formula shows that the smaller the angle between the screws, the greater the holding torque, so 60° would have been better. For 90° that reduces to ., so now we have .. That torque is a little greater than the motor rating, but shock loads could impose much higher torques so I added another two screws further along the shaft to double the load capacity.

Some pictures I forgot to post of testing the motor:

Attachment 12780Attachment 12781Attachment 12782

A week ago I made some progress with the power drawbar in preparation for the ATC system, here's a quick video to demonstrate:

https://www.youtube.com/watch?v=GQS89FdYg-c

More to follow in a couple of weeks, other projects getting in the way...

I got the ATC working this weekend - at least in principle:

https://www.youtube.com/watch?v=ku9RGQl2llc

http://www.mycncuk.com/attachment.ph...id=13399&stc=1http://www.mycncuk.com/attachment.ph...id=13400&stc=1

I've made quite a few parts for the carousel etc, but they're all quite simple so it didn't take long. There's a few issues I've mentioned in the video - the main things are to add sensors, change the coupling between the carousel and gearbox for something stiffer and check the clamping torque upon the toolholder. It's currently just controlled using a few lines of gcode - the air solenoids are currently controlled from LinuxCNC via relays just using the M64/M65 commands and delays. When I've got time I'll do it properly, so sensors are checked at each stage (using M66) to avoid mishaps.

Looking good, it will be interesting to see it working under real load conditions but it looks like a first on this forum.

Well done. ..Clive

Now that's porn,16 tools..it looks smart..i vaguely remember you mentioning mesa/pico hardware for this mill,did you bother.
will the turret be bidirectional and take the shortest route to the next tool or..not that it matters it seems quick enough.
nice work.

Quote:

---

Originally Posted by Clive S

Looking good, it will be interesting to see it working under real load conditions

---

I almost forgot - I did try cutting with the DC motor and it seemed fine, but that's not a good comparison considering how much higher the torque rating of the new motor is.

Quote:

---

Originally Posted by diesel

i vaguely remember you mentioning mesa/pico hardware for this mill,did you bother.

---

For now I've decided it's too expensive for what it gains - more and faster I/O. If I need more inputs/outputs I can add a 2nd parallel port and until I use servo motors the frequency limit of the parallel port is adequate. One temptation for using a motion controller is it could fit in the back of the column, so I'd be able to get all the electronics inside the column.

Quote:

---

Originally Posted by diesel

will the turret be bidirectional and take the shortest route to the next tool or..not that it matters it seems quick enough

---

Yes - that shouldn't be a problem. I've currently tuned the carousel motor to run with quite low top speed and very low acceleration as the low stiffness of the coupling means it oscillates too much if I run it faster, so it will certainly be quick enough when that's fixed.

Believe it or not, I found some time to work on this project again!
Decided I should do some sort of drawings for the X & Y axes, so here's the rough idea. I say rough as most of these parts are just blocks to show the general arrangement of things - I've not added detail:

Attachment 17006Attachment 17007

The X axis uses a bed from a different machine. I'm adding HSR35-1100mm rails to this and SHS25 rails to Y. These rails sizes are based on what I've got good deals on. For the ballscrews I'll be buying TBI DFU ballscrews, similar to Z, sized to get a reasonable stiffness for each axis.

As a 'first pass' look at it, I've tabulated the stiffnesses of each component, then combined these (as springs in series) to get a value for the stiffness of each axis:

Attachment 17008

This implies that the stiffness is dominated by that of the ballscrews, as these are by far the smallest numbers in the table, so it is worthwhile to invest more in 'large' ballscrews as improving these should have the biggest effect on the overall stiffness of the axis. For the table above I've calculated the stiffness's based on DFU2005 on X and Y, with angular contact bearings on both ends of the X-ballscrew, to compensate for the lower stiffness of such a long ballscrew.

This analysis is still too basic though, as it assumes the forces are all applied at the respective axis center of stiffness, since moments aren't considered, so I need to re-calculate including the 'pitch' and 'roll' stiffness of each axis. If this shows that the ballscrew stiffness is still significantly lower than the linear bearing stiffness, then I'll consider using even bigger ballscrews.

So far I've started machining the bed and steel spacer strips to mount the rails, that's about it...

So today, I got from this:
Attachment 17034
Via lots of this:
Attachment 17035
To this (cute little HGH15 bearing is for scale):
Attachment 17036

Next job is to drill and tap more holes than I care to count in the bed, which I can barely lift as it weighs more than me, so that's going to be fun... then I'll make the aluminium plates which fix on the steel strips that poke out of either end. Finally, I plan to get both sides of the bed surface ground so that the rails have a nice accurate surface to mount upon.

I also started drawing how I plan to mount the X-axis ballscrew drive end:

:applause:Fantastic build Johnathan, thanks for posting, always get very envious looking @ your projects.
Regards
Mike

Thanks Mike.

Today, I milled two slots!
4mm wide by 5mm deep slots in the steel strips shown in the previous post. The slots will have simple covers placed into them.
Attachment 17059

Next I've devised a bit of a mad scheme to drill the array of M6 holes in the bed:
Attachment 17062Attachment 17060

Then checked that it was sufficiently parallel - couple of whacks with the nearest hammer and it's now within 0.05mm over 400mm.
Attachment 17061

Now to continue the drawing...

After another few hours drilling, I just managed not to capsize my milling machine, so now I've got all the holes in the bed:

Attachment 17087Attachment 17088

Just to visualize, I placed the rails on with the bearings at the spacing I plan to use. Satisfyingly huge:
Attachment 17089

Lots of tapping now...

Oops :(

Attachment 17091

Ok, broken tap situation resolved...

I intend to buy the fasteners I need for this tomorrow, so time for some rough calculations to check what grade I should use. From the THK documentation, I gather that the tightening torques for the M8 and M6 fasteners in the rails should be 30.4Nm and 13.7Nm respectively. They don't recommend a grade of fastener (as far as I can see), so we are left to work out the tensile strength required ourselves.

First step, use the torque value to calculate the tensile force on the fastener. This is the same formula as working out the axial force you get for a given torque on a ballscrew, so I don't even need to look it up (although can be more accurate by including more in this):

.

This gives the force, for a given torque (tau), coefficent of friction (mu, ~0.2 for steel) and pitch. Machine screws are graded based on the tensile stress, so we need to divide this value of force by the cross sectional area of the screw to arrive at the stress, in N/mm^2. The stress area is roughly the minor diameter of the fastener, so roughly the diameter of drill you would use. The area is then clearly just the area of a circle.

So, e.g, the M6 fasteners are 1mm pitch, 13.7Nm and steel, so the force is 13.7*2*pi*0.2/0.001=17kN. Divide that by the stress area, 17000/(pi*2.5^2)=865Mpa. Bolts are graded such that the first number is 1/100th of their tensile strength rating, so e.g. a 10.9 grade fastener is rated for 10*100=1000Mpa. A rule of thumb suggests to only tighten a bolt to about 80% of its tensile strength rating, so in my case I will be OK with 10.9 grade fasteners for the M6 screws (especially as the actual stress for the given torque will be a bit less due to things I missed out for brevity).

Now I just need to remember to bring my torque wrench home...

Also noticed that RS have the caps to help prevent debris getting into the linear rails, so I'll get some of those.

good to see your doing something at last!

if you need a lift with ought let me know. Although I about give myself a hernia the other week lifting the motor ontop of my mill.

matt

Today I almost finished tapping the bed and made some bits for other projects / people. That's about it ... will buy some steel flat bar I need tomorrow to finish off the bed.

Quote:

---

Originally Posted by kingcreaky

if you need a lift with ought let me know. Although I about give myself a hernia the other week lifting the motor ontop of my mill.

---

Thanks for the offer, but what would be most helpful from you is if we could come to some arrangement whereby I get to use your new mill for a few hours? The XY carriage piece may be manageable on mine, but the bed piece is going to be too big.

I made some progress - over the weekend I machined the end pieces for the bed:

Attachment 17179Attachment 17180Attachment 17181Attachment 17182Attachment 17183Attachment 17184Attachment 17185Attachment 17186Attachment 17187Attachment 17188Attachment 17189Attachment 17190Attachment 17191Attachment 17192Attachment 17193

I have now left the bed with "Leytoner De Montford LTD" in Leicester to get both sides surface ground. First time I have needed someone else to machine something for me...

Next step is to acquire a straight edge asap to align the rails and decide on what bearings to use for the ballscrews (not sure on angular contact vs taper roller) so I can machine the housings.

Jonathan,

this thread makes me wish I had space to set up my RF45 clone and finish the conversion... I too did a huge part of the machining on a round coloumn mill. Will post photos when I find them

Just wanted to chime in on the bearings front. I would suggest you go with angular contact bearings and ones specific for ballscrews or ones specifically made with medium preload as standard.

Angular contact bearings have lower thrust rating than taper rollers but you will not need to add any spacers to get the preload to them right and also have lower friction. Additionally they take up less axial space. There is also a lower limit to the ID of taper rollers which from memory is 14.989mm (imperial).

Did you mention what size/grade ballscrews you are using?

Quote:

---

Did you mention what size/grade ballscrews you are using?

---

Briefly in post #18 - I'll be getting TBI brand ballscrews with DFU nuts, C5 grade with preload. I have already put one of these on Z. For the diameters, likely 20mm on X and Y based on the stiffness calculations in the same post, but not certain yet.

Quote:

---

Originally Posted by komatias

Just wanted to chime in on the bearings front. I would suggest you go with angular contact bearings and ones specific for ballscrews or ones specifically made with medium preload as standard.

---

The ones specific for ballscrews are what I've been looking at (i.e. 60 degree contact angle, preloaded), but I'm somewhat put off by the pricing, even on eBay. This document compares the virtues of triplex angular contact bearings vs a pair of tapered roller bearings and seems to conclude that the main limitation is the additional friction. In terms of stiffness, they are pretty much the same. I don't really consider the additional torque to be an issue, as that is a secondary consideration compared to the bearing axial stiffness - in essence I can just use slightly higher rated motors, which wont be a concern if I use my own. I will use at least 15mm bore bearings, so the available sizes are not an issue.

Some time ago, I did buy some angular contact bearings from aliexpress with 12mm bore that were supposedly matched pairs. Although they have the markings of matched pairs, they don't actually seem to be so, hence I am a little wary of trying that again with more expensive bearings, e.g. BS1547TN1 or BS1747. Interestingly the load rating of ballscrew specific bearings is similar to that of similarly sized taper roller bearings - probably the stiffness requirement dictates a certain load rating.

From what I've read/calculated the stiffness of generic single row <60° angular contact bearings in matched pairs is low compared to the screw/not or simply unclear/unspecified. There's a formula floating around for calculating the bearing axial stiffness for given dimensions, which I can do from the data on the SKF site, but the resulting values don't give much confidence. I fear that if I use these the stiffness of the bearings would be noticeably less than the ballscrews, so when combined in parallel with the nut & screw stiffness, it would not be a well balanced system, hence I want to at least get 60° contact angle bearings, if not roller bearings.

The stiffness you get from the bearings is also dependent on the preload class you use. Just like in the ballscrews, you can have zero backlash (P0) but you can also have P3 and P5. And of course there is the precision class and clearance to them too. The more accurate, the stiffer the bearings can be claimed to be.

You are right about the price though. On My KRV I went with heavy preload and spent around £240 for 2 sets.

You should also think about the stiffness of rotation perpendicular to the axis of the ballscrew. this will govern which way round you put the bearings. I forget which orientation is stiffer (face to face or back to back)

In the end I bought IBC bearings with 60deg angle medium preload. You can find all you need here: http://www.ibc-waelzlager.eu/dmdocum...t_Bearings.pdf

The ones I would have bought had I not bought the ones I did are: http://www.qualitybearingsonline.com...ct-20x47x15mm/

Another slightly overlook aspect is the locking nut you use should be as good as you can get it. The squareness of the thread to the face is not great in some of the chinese nuts I have come across. In my case I also wanted to lock on the front face as opposed to on the thread. Too a while to find anyone in the UK what had them.

Quote:

---

Originally Posted by komatias

Another slightly overlook aspect is the locking nut you use should be as good as you can get it. The squareness of the thread to the face is not great in some of the chinese nuts I have come across. In my case I also wanted to lock on the front face as opposed to on the thread. Too a while to find anyone in the UK what had them.

---

My plan there was to make them - not been a fan of any cheap ones I've bought and making my own at least means it's under my control. Or maybe buy these.

Today's big news is that I collected the newly ground milling table. They did it really quickly - dropped it off at about 9am on Monday and it as done by midday. Will post in more detail later - for now, here are some shiny pictures:

Attachment 17246Attachment 17247Attachment 17248Attachment 17249

Note the size comparison with the old X3 table :)

Jonathan,

I used these: http://www.spieth-maschinenelemente....ts/series-msr/
Much better than I would have been able to make or get made.

Can't quite see in the photos of the table. It seems that the marks are circular? Did they use a Lumsden grinder? If so then do check the flatness and parallelism they quoted. It is nowhere near as good as surface grinding that I would have thought they would do for a machine table. But it may of course be good enough for what you want....reminds me that my machine could do with a regrind this year.

Thanks for the link to the locknuts.

Regarding the griding, yes it was done on a Lumsden grinder.

For the parallelism I've checked with a micrometer at several points roughly down the centerline of the rails on both sides. The range of the readings was 49um, but most were within 20um or so. I'll post the readings tonight. That is worse than one would expect for a commercial milling machine, but I think it is very reasonable for what I need (being realistic about other tolerances in the system) and certainly better than the mill I have. Also consider they only charged £50.

Flatness is harder to check - I've tried bluing a parallel and seeing how the blue transfers to the bed at places across the width and length. It seems fine, but the parallel is only about 200mm long. I do have a ~600mm long surface plate. No idea of it's history though so I'm not sure if there's much value in trying to check flatness with it?

Quote:

---

Originally Posted by komatias

Jonathan,

I used these: http://www.spieth-maschinenelemente....ts/series-msr/
Much better than I would have been able to make or get made.

Can't quite see in the photos of the table. It seems that the marks are circular? Did they use a Lumsden grinder? If so then do check the flatness and parallelism they quoted. It is nowhere near as good as surface grinding that I would have thought they would do for a machine table. But it may of course be good enough for what you want....reminds me that my machine could do with a regrind this year.

---