“An Accurate CNC 4th Axis”…?

Advert

New Replies

Report Bug

Help/FAQs

“An Accurate CNC 4th Axis”…?

This topic has 26 replies, 9 voices, and was last updated 17 May 2011 at 11:04 by Gordon W.

Viewing 25 posts - 1 through 25 (of 27 total)

1 2 →

Author

Posts
27 April 2011 at 13:18 #67612
Tony Jeffree
Participant
@tonyjeffree56510
I read with interest, and a little concern, the article in MEW #175/176 on the construction of a so-called “Accurate CNC 4th axis”.

The main cause for concern was the choice of the athor to use a simple toothed belt reduction of 3.6:1, between the stepper and the spindle, and his calculation that with 8 microsteps, this would result in 5760 steps per rev of the spindle. All very well on paper but…

Firstly, stepper motors are not precision instruments in themselves. The positioning of whole steps on a stepper motor is typically quoted at 5-10%, but if you start going down the microstepping route, all bets are off as the microstep accuracy is highly dependent on the ability of the microstepping drive to accurately position the microsteps (which it achieves by applying varying currents to the two motor coils), and this ability is generally pretty poor. So if you are looking for the best accuracy that you can get out of a stepper based system, then you are better off going for full stepping and uping the drive ratio accordingly, because at least then the accuracy should be as good as the motor is inherently capable of.

Secondly, the accuracy quoted for a stepper motor is under no-load conditions. Apply any rotational force to the motor shaft and it will move to a point where the restoring torque generated by the coils on the magnets of the rotor is exactly equal and opposite to the force applied. So, assuming the force isn’t great enough to cause the rotor to turn to the next step position, then an applied force could cause a positional error of up to 1/2 a step of the motor (i.e., a step position error of 50%). Obviously, there are two ways that this can be ameliorated; you can use a socking great big stepper that generates a gozillion oz-inches or newton-meters of torque, or you can increase the drive ratio to suit the real-life working conditions. In a CNC mill, those conditions can be fairly “interesting”; slap even a modest stepper motor on the end of a 1mm pitch screw and the resultant force moving the axis can be quite large – for example, with my Taig CNC mill (see MEW 120, 121) I figured out that the 120 oz-in motors I used were capable of generating around 100 kilos of force at the tool tip, with a mechanical advantage via the leadscrew of around 125:1. Now, stick a 4th axis with a mechanical advantage of only 3.6:1 on the table of my Taig, and once you start taking a decent sized cut, something is going to give; that something is the posinal accuracy of the 4th axis. So what is the solution? Obviously, use a drive ratio for the 4th axis that is considerably greater than 3.6:1; perhaps slightly less obviously, use a worm drive rather than simple gearing. Why? Because if you use a worm drive with a reasonable ratio (say, 30:1 or greater) it is impossible to drive it backwards against even a small force holding the worm in position. This is one of the major reasons why worm drives, often as high as 90:1, as mentioned by the author, are used for 4th axis drives; they provide, in a single compact package, a large reduction ratio and the important characteristic that they can’t be back-driven by cutting forces.

So, while Mr Gordon’s device is nicely designed and well explained, I fear that building one in the hope of it being useful for real CNC work would be a waste of time.

A secondary cause for concern was the heavy weather the author made of his problems with the quality of toothed pulleys, and the astonishing solution that he mentioned in part 2 of re-machining one of the pulleys after layering on car body filler. Frankly if I had gone down that kind of route I would have kept very quiet about it when I wrote up the project! Treality is that toothed pulleys and belts of a suitable quality are readily available – RS Components for example – however, as mentioned above, this was in any case an example of the old Irish joke – “I wouldn’t start from here!”.

Turning a worm/wheel driven dividing head or a rotary table into a CNC 4th axis is a far more rewarding, and far easier, solution; it will give far greater precision than the one described in the articles, and will not suffer from its very obvious, and in my view fatal, flaws.

Regards,

Tony
Advert

27 April 2011 at 13:18 #38042
Tony Jeffree
Participant
@tonyjeffree56510
Only accurate under no-load conditions, I fear.
27 April 2011 at 14:33 #67618
John Stevenson 1
Participant
@johnstevenson1
Tony,

You echo my exact same sentiments with one major difference.

I went down this same design path some years ago but built a much larger unit. Powered by a large type 42 motor with 4:1 reduction on 1″ wide L series timing belt to a slave D1-3 spindle to take the chucks off the lathe, in short a real hefty bit of kit.

However imagine my surprise when doing two keyways at 90 degrees to one another to find I had two very slow helical slots !! It turns out what was happening was that the stepper when in the parked position was accepting power from the driver at half current which is normal and in addition the the more micros steps you use the more torque is drained off. The result was the cutting forces were over coming the detent torque of the stepper motor at rest. If I rested my finger on the large pulley I could feel it click over as it was cutting.

Outcome of this was the whole unit was scrapped and a large H/V rotary table was converted to do the same job, a job it’s still doing to this day.

As regards the pulleys perhaps a change from T5 to HTD type would be better as T5 is more for driving as opposed to positioning.

John S.
28 April 2011 at 16:01 #67658
Spurry
Participant
@spurry
I have recently completed a DivisionMaster conversion of one of Arc’s 6″ rotary tables as per the gospel (and pictures of John S).

There have been references to light and heavy cuts. If rounding off, say a piece of MS 3/16″ thick with an 8mm end mill what sort of depth of cut should the DM powered RT be able to cope with? It seems that this setup would not pull the skin off a rice pudding, but not sure if I am expecting too much.

TIA

Pete
28 April 2011 at 17:14 #67659
Tony Jeffree
Participant
@tonyjeffree56510
Posted by Spurry on 28/04/2011 16:01:38:

I have recently completed a DivisionMaster conversion of one of Arc’s 6″ rotary tables as per the gospel (and pictures of John S).

There have been references to light and heavy cuts. If rounding off, say a piece of MS 3/16″ thick with an 8mm end mill what sort of depth of cut should the DM powered RT be able to cope with? It seems that this setup would not pull the skin off a rice pudding, but not sure if I am expecting too much.

TIA

Pete

Pete –

There’s no substitute for trial and error.

Regards,

Tony
29 April 2011 at 09:38 #67675
Les Jones 1
Participant
@lesjones1
Hi Pete,

I am not clear whether your statement ” would not pull the skin off a rice pudding” referrers to the belt driven version or the rotary table version. In your question you do not sate the diameter of the disk you are asking about. The larger diameter the disk is the larger the torque required so cutting say a 15″ diameter disk you would have to take much smaller cuts than for a 3″ disk.

I had only skimmed the first part of the article on the 4th axis CNC and assumed that there must be some kind of brake mechanism to be shown in later parts. On reading it properly I could see no mention of this so I totally agree with Tony’s and John’s comments.

Les.
29 April 2011 at 11:47 #67683
John Haine
Participant
@johnhaine32865
My own experience of measuring repeatability and accuracy of my CNC converted super-7 cross-slide bears this out. i am using 8 “microsteps / step” and certainly don’t obtain the predicted accuracy or resolution, and the repeatability depends on the applied load. Seeking advice on the Mach forum it was pointed out that stepper motor accuracy is poorly specified for microstepping, since it is achieved by applying linearly interpolated current values to the windings between full steps and the torque/current curve is probably non-linear so there is, even in the best case, a non-linear relationship between demanded and actual position. This will be made worse by the effect of cutting torque and variable friction since I suspect that the full holding torque of the motor probably isn’t available. I’ve seen adverts for specially made steppers with a linear microstep characteristic in the US “digital machinist” magazine to improve performance, but no idea how much these are or whether available over here.

The main reason for microstepping seems to be to allow better driving of the motor at higher speeds without exciting resonances. For accurate positioning looks like it’s better to have a large mechanical reduction (worm/wheel) and rely on whole steps.
29 April 2011 at 12:37 #67685
John Stevenson 1
Participant
@johnstevenson1
The other problem is that every time you increase the microsteps you reduce torque so at some point under load you don’t have enough torque to make the move.

Stiction also plays a big part in that nothing happens until suddenly it moves so may steps in one go.

Ideally full or half steps are the best options if you can get the mechanical advantage.

John S.
29 April 2011 at 14:46 #67686
Spurry
Participant
@spurry
Les

My query relates to the use of Arc’s 6″ geared rotary table. One of the jobs was to round the corners of some MS 1/2″ thick plate to a radius of 8mm. From memory the plates were 2″ x 2 1/2″. Cutting full depth, I found the maximum infeed per pass to be about 0.1 mm or less.

Next up was some 3/16″ x 3/4″ MS bar with a radius of 8.50mm on each corner. I was just curious as to what experienced users of the Divisionmaster/Arc rotary table had found to be a suitable depth of in-feed.

So in both cases cutting was taking place very near to the centre of the RT, as the radii indicate.

Pete
29 April 2011 at 15:32 #67688
Les Jones 1
Participant
@lesjones1
Hi Pete,

I do not have a stepper / divisionmaster on my rotary table (6″ Vertex) but I have considered putting a stepper motor on it and designing my own controller as it sounds like an interesting project. (No criticism of Tony’s Divisionmaster.) I have used my rotary table for cleaning up disks of 2 mm steel and 20 mm chipboard both about 16″ diameter. I did not find that any great effort was required to turn the handle but I was taking very light cuts.

Les.
29 April 2011 at 15:46 #67689
Tony Jeffree
Participant
@tonyjeffree56510
One thing to bear in mind is that the Divisionmaster controller can only drive up to 2A/phase; if the motor that the OP is using is rated at significantly more than 2A then the controller won’t be driving the motor as hard as it can go. Also, be sure to set the current level on the controller appropriately…I’m pretty sure that the default setting is less than the full 2A/phase capability.

Regards,

Tony
29 April 2011 at 23:53 #67716
Spurry
Participant
@spurry
Thanks for the replies gents. Much appreciated.

Pete
2 May 2011 at 11:06 #67843
joegib
Participant
@joegib
Posted by Les Jones 1 on 29/04/2011 15:32:35:I do not have a stepper / divisionmaster on my rotary table (6″ Vertex) but I have considered putting a stepper motor on it and designing my own controller as it sounds like an interesting project.

I appreciate that you have a background in electronics so may prefer to do your own thing. But it may interest you and others to know that a DIY dividing head/rotary table controller has been developed by Steve Ward (aka ‘Kwackers) and is described/discussed on the CNC Zone here:

Rotary Table Indexer

The main block of documentation is at Post No.45 and the most recent PIC firmware /Manual versions can be downloaded at Post No.420. This project has been around since 2006 and is now pretty mature, relatively cheap to build and has good documentation (not to mention the exemplary support Steve has given via that thread). A video of one constructed by M. Parker-Lisberg can be seen here:

M Parker-Lisberg

— and another build by C Raynerd is available here:

C Raynerd

Joe
2 May 2011 at 11:33 #67844
MICHAEL WILLIAMS
Participant
@michaelwilliams41215
To achieve a high accuracy plus high resolution in an electronic dividing system you have to have an absolute standard of accuracy in the system somwhere . In a mechanical dividing head the dividing plate is the absolute standard and this plate still has to exist in some form in the electronic version . Using a normal stepper motor effectively as the standard is unsound practice and dividing results will almost always be of uncertain and variable accuracy .

The above is for direct indexing . Where a worm and wheel of modest number of teeth and standard accuracy is used the situation is a little improved but still not very good . To achieve high accuracy this way requires an exceptionally accurate worm wheel with a high number of teeth . With a high tooth number a stepper can be used in coarse resolution mode and its intrinsic accuracy matters less since the wheel is now effectively the reference standard .

To achieve a certainty of high and repeatable accuracy the easiest way is to use a rotary encoder of high resolution (4096 is acommon number and easy to use) and high accuracy . Photo etched glass disc types are quite low cost and super accurate . You can either buy a complete encoder or just the disc .

The indexing head can still be motorised in many different ways with or without feedback from the encoder and a very simple interface can give you CNC control .

There is a principle to recognise in all discussion of this topic in that you cannot ‘contrive’ greater accuracy in a system which is not accurate in the first place . More generally the errors in the output accuracy of any system are the transformed and added result of all the input errors in the system . Output accuracy can never be better than input accuracy and it is in general not as good .

Edited By MICHAEL WILLIAMS on 02/05/2011 11:46:30

Edited By MICHAEL WILLIAMS on 02/05/2011 11:58:56
2 May 2011 at 11:51 #67845
Tony Jeffree
Participant
@tonyjeffree56510
Posted by MICHAEL WILLIAMS on 02/05/2011 11:33:12:

You are looking at this dividing problem backwards . To achieve a high accuracy plus high resolution electronic dividing system you have to have an absolute standard of accuracy in the system somwhere . In a mechanical dividing head the dividing plate is the absolute standard and this plate still has to exist in some form in the electronic system . Using a normal stepper motor effectively as the standard is unsound practice and dividing results will almost always be of uncertain and variable accuracy .

The above is for direct indexing . Where a worm and wheel is used as well the situation is a little improved but still not good .

Actually, you are wrong.

The situation is greatly improved if you use a worm and wheel; the positional error of a stepper motor isn’t accumulative, so the accuracy of positioning improves by a factor equal to the worm/wheel drive ratio.

Direct indexing using full steps of a stepper motor gives you a positional accuracy of ~ +- 5% depending on the quality of the motor (and also gives you a very limited number of divisions, as there are only 200 discrete positions that the motor can index to). Stick that motor on the end of a 90:1 worm drive and the +- 5% reduces to +- 0.055%, modulo the accuracy of the worm and wheel.

This fact leads to a very usable means of improving the accuracy of dividing plates on a dividing head or rotary table; you cut one set of dividing plates as accurately as you can, then use that set with the reduction drive on your dividing head to generate a second, more accurate, set.

Accurate, manually operated division systems that work on that principle have been around for far longer than rotary encoders. The only fundamental difference between a manual dividing system using a high reduction ratio and the stepper-driven equivalent is that you push a button rather than cranking a handle.

Regards,

Tony

Regards,

Tony

Edited By Tony Jeffree on 02/05/2011 11:53:16
2 May 2011 at 12:02 #67846
MICHAEL WILLIAMS
Participant
@michaelwilliams41215
I’d already edited my posting to explain this better . We obviously croosed over in the server somewhere .
2 May 2011 at 12:09 #67847
MICHAEL WILLIAMS
Participant
@michaelwilliams41215
Hi Tony ,

I think we agree on most of the above even if it was a bit of a scramble getting there .

On the subject of using worm and wheel dividing to make more accurate division plates from less accurate ones this only works if the worm wheel itself is of high accuracy . I’m sure you are aware that errors in the wheel just get copied onto the work piece .
2 May 2011 at 12:36 #67848
Tony Jeffree
Participant
@tonyjeffree56510
Posted by MICHAEL WILLIAMS on 02/05/2011 12:09:58:

On the subject of using worm and wheel dividing to make more accurate division plates from less accurate ones this only works if the worm wheel itself is of high accuracy . I’m sure you are aware that errors in the wheel just get copied onto the work piece .

Hi Michael –

That is of course true (and I did point that out in what I posted, although maybe not as explicitly as you state here).

However, I take issue with your statement that you cannot contrive greater accuracy in a system that is not accurate in the first place. If that was universally true, then it would be impossible to construct measurement systems that are more accurate than the ones we had before, and we would not have the instruments we have today.

Regards,

Tony
2 May 2011 at 13:01 #67850
John Stevenson 1
Participant
@johnstevenson1
Are we in danger of chasing ghosts here ?

If you go to the Divisionmaster legacy site still on the web at

Divisionmaster Legacy Site and under examples of use, near the bottom where it shows a big Hoffman dividing head it reads:-

Top right shows gear cutting in progress, cutting the seventh tooth – he was cutting four 27-tooth gears simultaneously. Bottom right shows the display of DivisionMaster at this point – the actual position is 93.330 degrees. The theoretical position should be 93.333 degrees; John said that the positional error on any one of the 27 moves was never more than 0.008 degrees.

can anyone measure an error of this magnitude in the home workshop – or even, pray, an armchair ?

John S.
2 May 2011 at 16:59 #67860
Steve Garnett
Participant
@stevegarnett62550
29 seconds of arc? Easy!

Actually, no it isn’t. I’d have to raid the optics lab at work to do any measurements that accurately. And I think I’d have to do it there, as well – don’t have a stable optical bench at home. Also, a surprising number of things can disturb a measurement that small…

But I agree with Tony completely – in most disciplines, the vast majority of advances in scientific measurement in the past have relied entirely upon contriving at ways of analysing, and subsequently reducing the errors in previous measurement methodologies.

I wonder how easy it would be to contrive an experiment to determine how positionally accurate any given stepper motor actually is, within reason? I wouldn’t have thought that it was beyond the bounds of possibility…

Edited By Steve Garnett on 02/05/2011 17:02:28
2 May 2011 at 17:13 #67861
Tony Jeffree
Participant
@tonyjeffree56510
Posted by Steve Garnett on 02/05/2011 16:59:41:

I wonder how easy it would be to contrive an experiment to determine how positionally accurate any given stepper motor actually is, within reason? I wouldn’t have thought that it was beyond the bounds of possibility…

Edited By Steve Garnett on 02/05/2011 17:02:28

A lot easier just to look at the spec sheet

Regards,

Tony
2 May 2011 at 17:34 #67864
Steve Garnett
Participant
@stevegarnett62550
Oh come on, Tony. You’ve done enough electronics not to trust one of those, haven’t you?
2 May 2011 at 17:46 #67866
Tony Jeffree
Participant
@tonyjeffree56510
Posted by Steve Garnett on 02/05/2011 17:34:34:
Oh come on, Tony. You’ve done enough electronics not to trust one of those, haven’t you?

I’m afraid that is true….

Regards,

Tony
5 May 2011 at 09:59 #68008
MICHAEL WILLIAMS
Participant
@michaelwilliams41215
Hi Tony ,

However, I take issue with your statement that you cannot contrive greater accuracy in a system that is not accurate in the first place. If that was universally true, then it would be impossible to construct measurement systems that are more accurate than the ones we had before, and we would not have the instruments we have today.

My statement applied to things like the dividing head under discussion – whatever the fundamental accuracy of reference that is used ie stepper motor / dividing plate or wormwheel then that does indeed set a limit on what accuracy is achievable with the device as a whole and no amount of ingenuity of design will change that fact .

Ever increasing accuracy in more modern measuring systems still fundamentally comes down to finding more and more accurate and stable reference standards . We have moved on over the years from measuring for instance distance in strides to standard bars to wavelengths of light but the higher levels of accuracy in practical measurement are only possible because these improved reference standards have been found .

Best wishes ,

MW
17 May 2011 at 01:07 #68650
John Stevenson 1
Participant
@johnstevenson1
Any comment on the practice of building the teeth up with body filler then re-machining ?

John S.
Author

Posts