What Are the Best Low-Friction Bearings for Precision Positioning Mechanisms?

[evi7538]

Hello colleagues,

I'm designing a simple precision rotational mechanism for angular positioning of an optical element. I need a total rotation travel angle of about 50 deg at a low speed (a few seconds for the full travel) with about 4000 discrete positions over the travel angle. Any systematic positioning errors can be calibrated so I don't necessarily need an ultra-high positioning absolute accuracy. The most important parameter is high positioning repeatability, or low positioning noise. I need the repeatability below a few arc sec but the system has to be of small size and low cost so I can not use precision positioning stages. I tried a stepper with precision planetary gear and the positioning noise was terrible. I'm now thinking of using a stepper motor with small step angle (0.8 deg) in a microstepping mode (around 100 microsteps per step) and replacing the bearings with some high grade low friction ball bearings. Finding a stepper is not a problem, there is plenty of them available (e.g. NEMA 14 0.8deg steppers) but I need an advice on the bearings. I guess I need a bearing with the lowest friction and the highest friction evenness and repeatability. Basically, the positioning noise of the mechanism is proportional to (motor torque) / (friction noise). There are so many bearing options: ABEC classes from 1 to 9, types of steel, ceramic or steel balls, types of oil, cage types and cage materials (metal, nylon, Teflon). I guess the higher the ABEC class the lower the friction and positioning noise but what are the other right options to get the lowest friction and friction noise?

Are there any other types of bearings with even lower friction than ball bearings? Any magnetic or other types of bearings? I know about the air bearings (those are used in Michelson interferometers) but they need a pump and make the system expensive and bulky so it's not an option for me.

Thanks!

 

[anorlunda]

You can levitate the entire apparatus magnetically, as in maglev trains. https://en.wikipedia.org/wiki/Maglev.

However depending on the total weight, it may be big and power hungry. How much mass are you lifting?

Orientation matters also. Maglev lifts only vertically. What is the thrust orientation of that bearing?

 

[evi7538]

Thanks for the suggestion. The mass is very small but the axis orientation is unpredictable. Also, I'm looking for something available on the market that I can just buy from some supplier.

 

[256bits]

It's no wonder that hard disk drives stay away from stepper motors due to their clunkiness and backlash from the gearing.

In case you don 't know about other methods,
You can also investigate voice coil actuators for the precision they provide.
An investigation of servo mechanisms, and possibly closed loop operation.

(Just out of the blue, an ammeter with its coil mechanism seems repeatable for moving that little arm along the scale ).

You don't say what torque is required though for the manipulation of the optical element.

 

[jrmichler]

A definitive source of information on super precision rolling element bearings, from a top manufacturer: https://www.skf.com/binary/30-12987...precision-bearings-catalogue---13383_2-EN.pdf. It has 422 pages of super precision bearing goodness.

Generally, the largest source of friction is the seals. If you can surround your bearings with filtered air (think in terms of filtration to less than one micron), then you don't need seals. Just lubricate with the appropriate oil per the recommendations in the catalog listed above.

 

[evi7538]

A definitive source of information on super precision rolling element bearings, from a top manufacturer: https://www.skf.com/binary/30-12987...precision-bearings-catalogue---13383_2-EN.pdf. It has 422 pages of super precision bearing goodness.

Generally, the largest source of friction is the seals. If you can surround your bearings with filtered air (think in terms of filtration to less than one micron), then you don't need seals. Just lubricate with the appropriate oil per the recommendations in the catalog listed above.

Tanks for the good link! I think there are also non-contact seals available so they don't create contact friction. But there must be a cage in any bearing that keeps the balls in their place and that is the major source of friction I think. So Teflon cage might be the best option.

It's no wonder that hard disk drives stay away from stepper motors due to their clunkiness and backlash from the gearing.
In case you don 't know about other methods,
You can also investigate voice coil actuators for the precision they provide.
An investigation of servo mechanisms, and possibly closed loop operation.
(Just out of the blue, an ammeter with its coil mechanism seems repeatable for moving that little arm along the scale ).
You don't say what torque is required though for the manipulation of the optical element.

Right, voice coils, servo control and air bearings are used in Michelson interferometers and work great but that system is very expensive.
Steppers magnetically are actually more precise than rotational voice coils due to higher differential torque (the coils are actually very coarse steppers with one step per revolution). But in steppers it's the bearing that limits the precision. ​

 

[jrmichler]

I think there are also non-contact seals available so they don't create contact friction.

There are, and they are called shields. The problem is that shields do not keep fine dust out of the bearing. I have a belt sander that was manufactured with shielded bearings. It did not last very long before the bearings seized up. I replaced them with sealed bearings, and it has been working well since.

there must be a cage in any bearing that keeps the balls in their place and that is the major source of friction I think

Not quite. With the right amount of the right oil, cage friction is very small. SKF has (or had) a bearing calculator somewhere on their site that will calculate the friction for any of their bearing. My recollection is that most of the friction is from seals, followed by rolling friction proportional to the load on the bearing. If low friction is important, you will need to find that calculator and spend some time with it. Do not expect good advice from their tech support service (BTDT and learned the hard way).

 

[evi7538]

There are, and they are called shields. The problem is that shields do not keep fine dust out of the bearing. I have a belt sander that was manufactured with shielded bearings. It did not last very long before the bearings seized up. I replaced them with sealed bearings, and it has been working well since.

Thanks for the advice. I found that ASF offer rubber molded non-contact seal (type 3RU) and they claim that it does not make a contact and does not increase the torque while still protecting from the dust. That is probably the best sealing option I think:
https://dpk3n3gg92jwt.cloudfront.net/domains/ast_units/pdf/ENB-04-0556.pdf

Not quite. With the right amount of the right oil, cage friction is very small. SKF has (or had) a bearing calculator somewhere on their site that will calculate the friction for any of their bearing. My recollection is that most of the friction is from seals, followed by rolling friction proportional to the load on the bearing. If low friction is important, you will need to find that calculator and spend some time with it. Do not expect good advice from their tech support service (BTDT and learned the hard way).

Thanks, I found that calculator but for the low load that I have it gives zero for the frictional moment :) I think it depends on the application. I think in my case of very light load (~10-20g - weight of the optical element and the motor shaft) and non-contact seal the cage might be the only major residual source of friction.
https://dpk3n3gg92jwt.cloudfront.net/domains/ast_units/pdf/ENB-04-0573.pdf

 

[evi7538]

I think I found what I need: ceramic bearings from Boca Bearing. They claim that "Because ceramic is a glass like surface it has an extremely low coefficient of friction and is ideal for applications seeking to reduce friction."

 

[Tom.G]

Wind-up pocket watches and wrist watches use shafts with pointed ends riding in a conical depression in sapphire bearings, they are also called jewel bearings. They have very low frictional losses because the lever arm at the point of contact is extremely small.

For heavier duty, I think if seen them with a sapphire ball embedded in the shaft ends; which you may need to handle the axial load with random orientation of your design.

Try: https://www.google.com/search?&q=sapphire+bearings
over 2 000 000 hits

Cheers,
Tom

 

[evi7538]

Wind-up pocket watches and wrist watches use shafts with pointed ends riding in a conical depression in sapphire bearings, they are also called jewel bearings. They have very low frictional losses because the lever arm at the point of contact is extremely small.

For heavier duty, I think if seen them with a sapphire ball embedded in the shaft ends; which you may need to handle the axial load with random orientation of your design.

Try: https://www.google.com/search?&q=sapphire+bearings
over 2 000 000 hits

Cheers,
Tom

That's a good find, thanks Tom!

 

[Mech_Engineer]

I'm designing a simple precision rotational mechanism for angular positioning of an optical element. I need a total rotation travel angle of about 50 deg at a low speed (a few seconds for the full travel) with about 4000 discrete positions over the travel angle.

I work in this design space, you're right to be concerned about repeatability and hysteresis with sub-arcsecond angular resolution. Have you considered a flexure-based positioning solution? 50 degrees of travel sounds pretty large for optical positioning unless its a high-speed scanning application, but I've used flexure designs similar to flexure pivots with some success in sub-arcsecond applications.

Something like this: they're fricitonless so there is no positioning hysteresis, you just need to take into account the torsional spring rate in your actuator's design.

 

[Mech_Engineer]

In case you're interested, here is some reading on the subject as well regarding use of flexures for alignment tasks when high precision/accuracy and limited travel are needed:

"Slit Diaphragm Flexures for Optomechanics", Vukobratovich, Richard, Proc. SPIE 2542 (1995)

Many precision motion applications in optical engineering require limited ranges of travel. Often the linear range of travel is less than 1 mm, and the range of rotation is less than 5 degrees. Conventional rolling element bearings are not a good choice for providing motions of this type. The relationship between friction and torque in a rolling element bearing is non-linear for small ranges of travel (1,2). Hysteresis and rapid wear are other problems associated with rolling element bearings used for small ranges of travel (3,4).

 

[evi7538]

In case you're interested, here is some reading on the subject as well regarding use of flexures for alignment tasks when high precision/accuracy and limited travel are needed:

"Slit Diaphragm Flexures for Optomechanics", Vukobratovich, Richard, Proc. SPIE 2542 (1995)

Mech_Engineer, thank you so much for the links. I've seen custom-made flexure designs in some Michelson interferometers but I did not know that flexure bearings are available as standard parts. This looks like a perfect solution to my problem!

 

[Mech_Engineer]

Mech_Engineer, thank you so much for the links. I've seen custom-made flexure designs in some Michelson interferometers but I did not know that flexure bearings are available as standard parts. This looks like a perfect solution to my problem!

Pay close attention to your travel requirements and the flexure bearing's fatigue rating. A travel of +/- 25 degrees could be a bit farther than recommended for those bearings but they may be a good starting point.

 

[MRFMengineer]

Lots of great bearing options have been listed already.

If you want to invest some time and money on the motion control side adding a rotary encoder, brushless motor and motion controller+motor drive, would do quite a bit to improve repeatability over a stepper.

 

[evi7538]

Lots of great bearing options have been listed already.

If you want to invest some time and money on the motion control side adding a rotary encoder, brushless motor and motion controller+motor drive, would do quite a bit to improve repeatability over a stepper.

Thank you for your suggestions. I don't understand why a brushless motor would be better than a stepper in microstepping mode. Would a stepper with similar size and current have higher angular differential torque (since the same holding torque is applied over a small step angle as compared to a motor with similar magnetic torque where the angle is much larger?). My understanding is that the positioning noise is proportional to the torque noise from bearings divided by the differential magnetic torque, so for motors with comparable bearings, size and current the stepper should have lower positioning noise. Am I missing something?

I can use an encoder on brushless or stepper motor but I don't think they can have positioning accuracy down to a few acrsec. Or can they?

 

[Mech_Engineer]

The main problem with micro-stepping is you don't have closed-loop position feedback, and the smaller the step the more likely a stall condition or missed steps will occur. Speed can also be an issue with a micro-stepping, in addition to potential vibration during operation.

A direct drive rotation stage will struggle to achieve arc-second resolution, it depends on what encoders are available on it.

 

[evi7538]

The main problem with micro-stepping is you don't have closed-loop position feedback, and the smaller the step the more likely a stall condition or missed steps will occur. Speed can also be an issue with a micro-stepping, in addition to potential vibration during operation.

A direct drive rotation stage will struggle to achieve arc-second resolution, it depends on what encoders are available on it.

Understood by at an arcsec precision how would you implement a closed-loop feedback? Encoders are not good enough at that level of precision.

I agree about speed and missing steps but in my case of very low speed this is not an issue. As long as the magnetic torque per microstep is a few times larger than the bearing friction torque there will be no missing steps or stalling at low speed. The point is, as I said above, steppers have much higher magnetic torque per the same angular increment compared to servos or brushless motors. therefore the same bearing friction causes proportionally smaller angular positioning error/noise.

 

[MRFMengineer]

Arc second resolution is readily achievable with an external optical encoder, several companies make them, such as Renishaw: https://www.renishaw.com/en/atom-encoder-series--37564 (~$500?)

Even if you used a stepper, at least you could monitor the actual position not just commanded position. There are several methods for reading the encoder position.

The encoder should be rigidly mounted close to your sample, not on the motor if possible. I am not sure how your system is laid out.

 

[evi7538]

Arc second resolution is readily achievable with an external optical encoder, several companies make them, such as Renishaw: https://www.renishaw.com/en/atom-encoder-series--37564 (~$500?)

Even if you used a stepper, at least you could monitor the actual position not just commanded position. There are several methods for reading the encoder position.

The encoder should be rigidly mounted close to your sample, not on the motor if possible. I am not sure how your system is laid out.

Thank you for the link, that is interesting. Still too big and expensive for my design. If nothing else works I can still use those, but before I go there I want to explore simpler/cheaper solutions. I plan first to try a fine pitch stepper in miscrostepping mode with high grade ceramic mini-bearings, that is the easiest thing to do. My calculations tell me that it should be good enough for a few arcsec precision. But if that is still not good enough, I plan to try jewel or flexure bearings with a stepper.

 

[MRFMengineer]

What is the microstepping resolution for your motor? What motor are you using?

 

[evi7538]

What is the microstepping resolution for your motor? What motor are you using?

It's a standard NEMA 14 with 0.8 deg full step and 4 oz-in holding torque. I'm driving it with 16-bit DAC + power amp so I can make as many microsteps as I need, and I need about 100 per full step.

 

[Tom.G]

It's a standard NEMA 14 with 0.8 deg full step and 4 oz-in holding torque. I'm driving it with 16-bit DAC + power amp so I can make as many microsteps as I need, and I need about 100 per full step.

As suspected, there is a drawback to microstepping; torque decreases inversely with number of microsteps.

Incremental torque/microstep as the number of microsteps/full step increases
Microsteps/
full step % holding
torque/microstep
1 ... 100.00%
2 ... 70.71%
4 ... 38.27%
8 ... 19.51%
16 ... 80% 9.80%
32 ... 4.91%
64 ... 2.45%
128 ...1.23%
256 ... 0.61%
The table shows the significant impact of the incremental torque/microstep as a function of the number of microsteps/full step.

from: https://www.machinedesign.com/archive/microstepping-myths
found with: https://www.google.com/search?&q=stepper+motor+torque+versus+microstep

Cheers,
Tom

edit: fixed typo at 16 steps

 

[Baluncore]

But if that is still not good enough, I plan to try jewel or flexure bearings with a stepper.

Robust galvanometer coils and the spinners in gyrotheodolites are suspended by a thin vertical bronze tape. If you fix one end of the tape in tension, mount your item part way along a tape, then turn the far end with the motor, you will get low bearing noise with a ratiometric angular reduction.

 

[jrmichler]

Can you mount your optical element directly on the motor shaft? No external bearings means no bearing friction. Since your load is only 10-20 grams, you can overhang them beyond the end of the shaft. You might possibly want to exchange the motor bearings for high precision bearings.

 

[evi7538]

As suspected, there is a drawback to microstepping; torque decreases inversely with number of microsteps.

Incremental torque/microstep as the number of microsteps/full step increases
Microsteps/
full step % holding
torque/microstep
1 ... 100.00%
2 ... 70.71%
4 ... 38.27%
8 ... 19.51%
16 ... 80%
32 ... 4.91%
64 ... 2.45%
128 ... 1.23%
256 ... 0.61%
The table shows the significant impact of the incremental torque/microstep as a function of the number of microsteps/full step.

from: https://www.machinedesign.com/archive/microstepping-myths
found with: https://www.google.com/search?&q=stepper+motor+torque+versus+microstep

Cheers,
Tom

Right, I know. But the differential_torque = dTorque/dAngle is always the same regardless of the number of microsteps. And then the positioning noise is the bearing_torque_noise/differential_torque which is independent of the number of microsteps. The incremental torque decrease in microstepping mode only becomes a problem when you need to move high loads or at high speed (or both).

So to get the lowest positioning noise I need a stepper with the highest differential_torque and bearings with the smallest friction. To get the highest differential_torque I need a stepper with the highest holding torque per full step (which is limited by size and current in my application) and smallest full-step angle (0.8 deg seems to be the smallest angle available on the market).

Can you mount your optical element directly on the motor shaft? No external bearings means no bearing friction. Since your load is only 10-20 grams, you can overhang them beyond the end of the shaft. You might possibly want to exchange the motor bearings for high precision bearings.

Right, that is exactly what I'm doing right now: the bearings in the motor are of a standard size 5mm ID so I'm replacing them with the same size high grade ceramic bearings

 

[Mech_Engineer]

In my experience using a stepper motor for rotational positioning is challenging when micro-stepping at around 30,000 micro-steps/rev. (~43 arc-sec per step), some drives can reach 128,000 micro-steps/rev but that is for smoothness not incremental motion resolution. A micro-step at the arc-second scale would equate to 1,296,000 micro-steps per rev., and I'm not aware of any reliable methods to micro-step at that level. Remember when micro-stepping lost steps are a real problem and you constantly have to re-home the motor (or use an encoder) to maintain positioning accuracy.

I think you need re-evaluate your positioning plans, my feeling is replacing bearings in a motor (no matter how exotic they are) isn't going to get you there.

 

[evi7538]

In my experience using a stepper motor for rotational positioning is challenging when micro-stepping at around 30,000 micro-steps/rev. (~43 arc-sec per step), some drives can reach 128,000 micro-steps/rev but that is for smoothness not incremental motion resolution. A micro-step at the arc-second scale would equate to 1,296,000 micro-steps per rev., and I'm not aware of any reliable methods to micro-step at that level. Remember when micro-stepping lost steps are a real problem and you constantly have to re-home the motor (or use an encoder) to maintain positioning accuracy..

I don't need a few arcsec positioning resolution, I only need approx. 40 arcsec per microstep which means that with 0.8 deg full-step stepper I need only 50 microsteps per full step. That should not be a problem with my 16-bit driver. But what I do need is the repeatability of positioning within a few arcsec. And yes, the motor will be re-homed after each working cycle.

I think you need re-evaluate your positioning plans, my feeling is replacing bearings in a motor (no matter how exotic they are) isn't going to get you there.

Well, we will see, I will find it out next week. It's the easiest way because I have all setup ready. If it still does not work then I will need to redesign with jewel or flexure bearings which will take longer time.

 

[Mech_Engineer]

I don't need a few arcsec positioning resolution, I only need approx. 40 arcsec per microstep which means that with 0.8 deg full-step stepper I need only 50 microsteps per full step.

This sounds a bit more feasible, I was originally thinking you needed arc-second resolution which wouldn't be possible with micro-stepping alone.

Good luck!

 

[evi7538]

I got pretty good results with NEMA 14 stepper and ceramic ABEC5 bearings: the positioning noise is around 3 arcsec RMS, good enough for my project