PDA

View Full Version : Ballscrew Axial Backlash



snapatap
9th Aug 2020, 07:46 PM
The drive end should have thrust bearings or angular contacts. That bearing in the drive end is just a deep groove ball bearing and is just holding the screw up.

electrosteam
9th Aug 2020, 08:34 PM
Thanks for the comments.

My non-drive end is not supported by a bearing.
It is a 32 mm screw so pretty stiff and I have not observed any whipping.

The drive end (I think !) is a deep groove ball bearing, which could be the problem
A strip down will check that the assembly is correct and tight.

There is a multitude of solutions for bearings with controlled/limited axial movement.
I am just seeking comments from others as to what they selected, and why they made that selection.

At 300 mm/min my 10 mm pitch x-axis screw is only turning at 30 RPM.
I think I can ignore all bearing speed ratings and temperature effects and concentrate on the bearing room-temperature static ratings.
But, price/performance comparisons are difficult.

Bearing selection appears to be a black art !

Keep well,
John.

jhovel
10th Aug 2020, 01:51 AM
Given that you only have one bearing on your ballscrew, I would replace it with either a 4-point contact bearing, a double angular contact bearing or - if you can find the space - a pair of angular contact bearing. In any case, it can't be a deep groove ball bearing in the long run: your axial float will get bigger and bigger as the balls and tracks wear.

snapatap
10th Aug 2020, 07:48 PM
wouldn't get to worried about load ratings, more critical is what will fit in the available space. If you let me know what bearing it has and some pics of the mount i can make some recommendations.

electrosteam
11th Aug 2020, 10:02 AM
snapatap,

Thanks for the offer to help.
You are correct on the question of space available.

Correction to the OP:
the suspected bearing is 6204, not 6207.

My investigations lead me to a 7204 bearing pair as a selection.
The pair is 14 mm longer than a single 6204.

I will do a sketch of what I think is the assembly later today and consider appropriate alterations.
This could include boring out the table end cap to accept a bearing sleeve 14 mm longer than existing.
Then boring out the pulley to accept the extended bearing length.

A couple of questions:
a) How does one get the correct preload on 2 x 7204 bearing stack ?

I assume a manufactured selected pair of 7204 bearings with, say, 25 micron end float is out of my price range.
I have a good surface grinder and the patience to go through a few make-install-test cycles to size any spacers required.

b) Would a duplex ball arrangement be 'good enough' in this application ?

Keep well,
John.

OxxAndBert
11th Aug 2020, 10:31 AM
Precision angular contacts would be absolute overkill for the application as you don’t need the precise radial location/support that they provide.

Adding thrust bearings (even a single and preloading the existing bearing to remove clearance) might be all that is needed - but as already said it’s all about what’s practical with your particular installation.

Steve

electrosteam
11th Aug 2020, 06:15 PM
Learning a lot about bearings.
Now know the correct term is 'double row', not 'duplex' as mentioned previously.
And that a pair of angular contact bearings can be arranged either 'front-front' or 'back-back' for different performance characteristics.

Got out the handbook for the original (manual) build of the mill.
The leadscrew drawings for the X-axis and Y-axis show 2 x 6204 bearings.
The drawings do not seem to include any special shims for preloading.

I wonder if the factory purchased matched pairs of deep groove ball bearings, and the pairing has been lost over the years.
Perhaps all I need to do is sort out the pairing !

Put the DI on the end of the ballscrew shaft with steppers de-energized.
Can rotate the pulley CW and CCW and see 0.1 to 0.15 mm end float on the DI.
Did the same thing to the Y-axis and got about 2/3 the movement.
This means I have an easy way to check any new trial assembly.
There is a large spacer between the nut and the pulley, so re-arranging to fit shims in the bearing stack may not be too difficult.

RS Components list a NSK 7204 bearing for $30, so not too expensive.

Keep well,
John.

snapatap
11th Aug 2020, 06:59 PM
snapatap,

Thanks for the offer to help.
You are correct on the question of space available.

Correction to the OP:
the suspected bearing is 6204, not 6207.

My investigations lead me to a 7204 bearing pair as a selection.
The pair is 14 mm longer than a single 6204.

I will do a sketch of what I think is the assembly later today and consider appropriate alterations.
This could include boring out the table end cap to accept a bearing sleeve 14 mm longer than existing.
Then boring out the pulley to accept the extended bearing length.

A couple of questions:
a) How does one get the correct preload on 2 x 7204 bearing stack ?

I assume a manufactured selected pair of 7204 bearings with, say, 25 micron end float is out of my price range.
I have a good surface grinder and the patience to go through a few make-install-test cycles to size any spacers required.

b) Would a duplex ball arrangement be 'good enough' in this application ?

Keep well,
John.


A. for this application spacer between the outer races, threaded shaft with nut, just like car wheel bearings, no need for high precision bearings

B. Assume you mean 2 deep groove ball bearings? maybe but i would rather angular contacts.

What sort of mill do you have? I wonder if someone has put 2 deep groove ball bearings in the housing instead of a pair of angular contacts.

electrosteam
11th Aug 2020, 08:21 PM
The mill is a Spanish Kondia Fv-1.
A good solid close clone of the Bridgeport, seems to work Ok with CNC.

The Kondia manual drawings look like 2 x deep groove, and the parts lists them as 6204, deep groove.

Tomorrow I will make some test shims to see if that approach will work.
Prefer cast iron, but at ( say ) 2 mm thick may be too fragile.
My surface grinder has a DRO with 10 micron resolution, so differential shim sizing can be achieved.

When the disassembly occurs, I will check the fitted bearings if they have any orientation markings.
I have sensitive dial indicators and a surface plate, might be able to detect any special inner/outer ring grinding.

Keep well,
John.

jhovel
11th Aug 2020, 09:34 PM
Have a look for a 3204. It's a single bearing with two rows of angular contact balls and races and only 20.6mm wide (6.6 more than the original - and no preloading needed. All done internally at the factory.

electrosteam
12th Aug 2020, 10:42 AM
The 3204 bearing is the double row type I was referring to.
Pricing is not too bad, but trying to ascertain the axial float rating is a bit difficult.

A SKF table quotes C2, C3 etc as referring to axial internal tolerance.
I assume this is the float.
But one commercial source quotes C2 as 'less than normal radial clearance'.

When I re-balled the X-axis nut, my assessment was about 25 micron axial float, so another 25 micron from the bearing would total 50 microns.

A friend is an expert machinist and he suggests I have a chat with his preferred local bearing supplier.

2 x 6204 is 28 mm wide, a single 3204 is 20.6 mm, and the difference can be accommodated with an additional mounting sleeve.

Keep well,
John.

electrosteam
13th Aug 2020, 04:42 PM
Got the bearings out, 2 x labelled PFI 6204-2RS C3.
One net supplier lists the PFI bearing 'C3' as greater internal clearance.

There was a punched (as in burred) shim between the outer rings.
The burrs curve over the ring chamfers without adding to the thickness (but make it hard to measure).
Best estimate of the shim is 100 microns, measured directly with a micrometer and as added to a bearing outer.

Bearing 1:
- outer 13.97 mm,
- inner 13.975 mm.

Bearing 2:
- outer 13.98 mm,
- inner 13.96.

Holding bearing in the hands, one can detect looseness, not sure if pure axial or just twist between the rings.

To sort out the axial/twist question I would have to get a precision block to support the inner on a surface plate.
Then press the outer down and test the height of both rings off the table.

But:
- one bearing has the inner ring thinner than the outer by 20 microns,
- the 100 micron shim for the outer is the wrong way,
- I wonder if a shim for the inner has gotten lost ?
- deep groove balls just seem the wrong selection.

So, still considering a replacement.

As everyone always wants photos, attached is my obligatory non-informative contribution.

388057

Keep well,
John.

OxxAndBert
13th Aug 2020, 04:50 PM
Given you’ve got a surface grinder you could always just grind a smidgeon off the side of one of the outer races to create some preload.

Steve

electrosteam
14th Aug 2020, 12:18 AM
Steve,
Grinding is tempting.

I could put the bearing on the table with the 100 micron shim under the outer ring.
The inner would be sucked down to the chuck, leaving the outer somewhat higher.
A grind right over would leave the outer width less than the inner width (up to 100 micron difference).
Then assemble the pair without the shim -- tempting !!

Keep well,
John.

electrosteam
14th Aug 2020, 02:11 PM
Developments this morning.

I have a nice benchtop inspection DI with a Mercer 0.0001" indicator by Arthur Murray in Melbourne.

Put the bearings on the anvil, can can pick the differences in the thickness of the rings noted earlier:
(A) inner high by about 2.5 microns (approx 1 div),
(B) inner low by 18 microns (5 - 10 div).

Tried adding the shim between the bearing and anvil, and pressing the inner down flat:
(A) inner low by 75 microns (approx 30 div),
(B) inner low by 75 microns ( 30 div).

Found a gauge block that spans the inner and not the outer.
The small set of gauges I have is Mitutoyo, and the blocks are only general purpose ground on the non-gauge sides, so had to micrometer select one that was parallel.
When tested in the DI, can easily pick up the inner being higher by something like 30 to 50 div, but not reliable because the outer ring would rock around.

I will repeat all the above with a fellow machinist this afternoon as a check.

It is easy to see why I was getting such a large backlash on the mill.
Keep well,
john.

electrosteam
23rd Aug 2020, 06:26 PM
The 2 x 7204 angular contact bearings can be purchased as a matched pair that provide virtually zero axial float.
But, only available with open construction necessitating seals with an oiling facility to be added

I was told the 3204 C2 is virtually unavailable in Australia (at unspecified high cost).
The normal range is specified as 7 to 25 microns, and available at a reasonable cost.

Fitted a FAG 3204 with appropriate spacers to allow for the reduced width.

The shaft float when jogging X +/- was about 40 microns, much improved.

Operating the X-axis with rapid moves of +/- 100 mm from the zero position shows the actual on-table backlash as 60 microns.
This is an acceptable value and I don't think I will pursue further improvements.

I have another FAG bearing and will soon look at improving the Y-axis.
Keep well,
John.

electrosteam
26th Aug 2020, 06:32 PM
A) I thought I should check the backlash using a Mitutoyo 1 micron/division Dial Indicator, 200 per rotation, 1 mm total.

This Di is a difficult beast to hold steady, but I made a special rigid mount hanging off the mill head
The photo shows the arrangement.
There is a box angle plate on the table, the DI positioned to allow pass-by of the box, and a parallel that is slid back for the pass-by, and carefully angled in to the tip for the measurement.
A setup as stiff and repeatable as I could quickly arrange.

Test started with the DI upscale, G0 move (600 mm/mim) to X+100 mm, G0 to X0, note the DI reading, G0 X-100, G0 X0, note the reading.
I take the backlash to be the difference in the two readings.

Consistently got about 70 microns for the backlash.

B) Then, something interesting.
Instead of G0 commands, I tried G1 F100 to +/- 50 mm.
Now consistently getting 90 microns.

I have no idea why this would be.
Any suggestions ?
I did go back and check the value at G0 and 600 mm/min, still 70 microns.

C) I then repeated the exercise with the original Dial Test Indicator, typical common unit labelled Insys.
10 micron /div, 800 per rotation, 1 mm total.
This is my goto DTI on the mill.

Got a virtual repeat of the values from the DI.

D) I then tried jogging using the CNC manual commands with 0.1, 0.05 and 0.01 mm steps.
Now getting about 100 microns change on the CNC before the DTI detects the movement.

E) In the real world of cutting metal, feeds tend to be around 100 mm/min, so perhaps I must rate the backlash at 90 microns, not the 60 previously advised.

Keep well,
John.

pippin88
26th Aug 2020, 07:23 PM
I don't understand why are moving to x100 and how you are measuring backlash that way.

The way I do it:
Firmly mounted indicator.
Move towards (into) the indicator until indicator is well in contact.
Now you tell controller to move away from the indicator until the indicator reading actually changes. (Usually easiest to do this in steps).
The difference between the CNC DRO and the actual movement is the backlash.


Stick slip can be a factor in different readings at different speeds.
Depending on steppr / servo resolution you may also have errors / differences also. With steppers, microsteps are not as precise as a full step and if you are trying to read a difference within the microstep realm you will not be accurate.

electrosteam
27th Aug 2020, 02:51 PM
The +/-100 mm at 600 mm/min. and +/- 50mm at 100 mm/min were aimed at duplicating real-world situations that affect jobs.
The backlash is the difference between approaching the Touch-Off zero position from the two directions.

You mentioned stick/slip, and I think you have to add inertia, and I have a belt drive ( stiff and short, but still a belt ).
I have always wondered about the track in the ball nut that the balls take, bottom of the valley or up one of the sides.

Our methods for short-range jogs whilst in contact with the DI are similar, but could be slightly different.
It appears you move slowly in, whereas I was jogging with high acceleration, followed by a high deceleration, but the maximum speed reached is still very low because of such short steps.

I will investigate further the question of the microsteps.
Cannot do much about it as the machine needs microsteps to get acceptable parts.

Keep well,
John.

electrosteam
7th Sep 2020, 10:20 PM
As the X-axis bearing change went Ok, I went ahead to check the Y-axis.
Discovered the same 2 x 6204 plus 100 micron spacer between the outer rings.
Changed to the 3204 and tested the backlash as previously described for the X-axis, and got similar results.

So, to complete the trifector, I tested the Z-axis.
Got somewhat similar results.

Summary of backlash test results:
X-axis:
28 microns - rapid moves at 600 mm/min,
66 microns - moves at 100 mm/min,
120 microns - jog in steps of 10 microns at 6 mm/min.

Y-axis:
38 microns - rapid moves at 600 mm/min,
49 microns - moves at 100 mm/min,
90 microns - jog in steps of 10 microns at 6 mm/min.

Z-axis:
45 microns - rapid moves at 180 mm/min,
53 microns - moves at 30 mm/min,
70 microns - jog in steps of 10 microns at 6 mm/min.

I observed an interesting difference in the mill movement responses comparing the X, Y axes to the Z-axis when jogging in 10 micron steps.
For the X and Y axes, the mill would not move until a number of steps were entered, then linear per step.
For the Z axis, observed a small movement for each of a few entered steps, then linear per step.

Then I tried an actual cut test.
Programed a circle Dia 74 mm and 5 mm deep on CI stock Dia 75 mm.
No backlash compensation in the computer.
The photo shows the flats observed at the quadrant points as alternately the X and Y-axis hit their reversal limits as the tool tracked around the circle.
The flat is about 5 mm wide.
If you do the trigonometry, this width flat infers a radial step in of about 80 microns.

Repeated the test with backlash compensation entered into the computer and a cut of 0.2 mm.
X-axis 66 microns, Y-axis 49 microns, Z-axis 53 microns.
The photo shows the result, a flat that is extremely small, and cannot be felt by touch.
Tried running a DTI around the test piece, but the background fluctuations masked any flat detection.
These fluctuations reflect the limitations of doing circles with stepper motors on an old mill.

This, I think, completes my investigations under this subject.
I am happy that the mill is performing as good as I can expect.

Keep well guys,
John.388559388558

Edit: The Z-axis is still the original factory-fitted trapezoidal (metric equivalent to Acme) screw.