linear motor accuracy 1

[subsearobot]

Hello
We are evaluating a system where the mechanical system has been contracted out to an outside vendor. Essentially, we need accurate linear positioning (better than .0005") over a range of about an inch, with a resistance of ~20lbs. no big deal. we also need a peak force of around 200lbs, though not while requiring precision position.

So the contractor selected a big-a$$ linear stepper motor to meet the high load reqs. (its a nema 32 frame, with a 2 start lead screw).

We are looking at the output with a laser based position sensor, and seeing that the positioning accuracy is sinusoidal- where the period corresponds with 1 complete revolution of the motor. from what i can tell, the amplitude is about .003" (full scale travel is 1" here, and typ peak to peak looks about .3%)

does this seem typical for a larger frame stepper? we are loathe to second guess our contractor, so want to understand the issue fully before suggesting alternate actuators. precision positioning is a new world for me...

thanks

 

[mikekilroy]

"SERVO Linear motors" are as accurate as their feedback device. It can be down to +/- quadrature count on the feedback device. So if feedback is a linear scale with accuracy of 1 count = 0.1 MICRON, then your accuracy with decent drive system to support it, is +/- 1 micron.

"STEPPER Linear motors", are as accurate as their feedback device - down to their step resolution. It can be down to +/- quadrature count on the feedback device. So if feedback is a linear scale with accuracy of 1 count = 0.1 MICRON - AND your stepper has at least 1 micron steps, then your accuracy with decent drive system to support it, is +/- 1 micron. Stepper motors typically are limited to a rather crude microstep limit so likely that will be the limit rather than the feedback device.

To say your contractor provided a "linear stepper motor to meet the high load reqs. (its a nema 32 frame, with a 2 start lead screw)" is just wrong. If it is a linear motor, there is NO lead screw involved. So unless you get better data on what you really have, it will be hard to answer your question.

There is repeatability and there is accuracy - not the same animal. You seem to be mixing the meaning. Without knowing which you mean for sure, it will be impossible to answer your question.

Also, since the feedback device is SOOOOOO IMPORTANT to repeatability and accuracy, without stating ITS specs, your question cannot be answered.

http://www.KilroyWasHere<dot>com

 

[mikekilroy]

damm typos... all "1 micron" words in my last post should be ".1 micron"

http://www.KilroyWasHere<dot>com

 

[subsearobot]

Mikekilroy, My apologies. Linear actuator, based on a stepper motor with a lead screw. we are micro stepping 8/ step. From what I can tell, each full step yields approximately .001" linear travel. we used two different drives with similar results.

I am not in question of our testing equipment, that has been verified independently. there will be no feedback on the final system. the laser device has a resolution on .1um (.004 mils). not sure about accuracy, though i would surmise good enough to see what we need.

My question here is:
Is a periodic error of up to .003" in expected position typical for an actuator of that size? I realize this is an extremely general question, but I am fairly ignorant of performance characteristics, and looking for some intelligent input as a starting point. to me, unless I am missing something, it seems like the motor does not work for this application.

the previously attached graph shows the expected position in purple, and the error in green and red. error calculated as

(delta from expected/ full scale travel x 100),

i believe FST was 1.000". X axis shows 800 steps. to be honest, our EE prepared the graph, I am unsure of the method he normalized the data, but to me it is irrelevant- I see a large periodic error. which i am guessing is coming from a less- than precise motor construction. but am hoping for some input from you all that are experienced in this arena.

the graph shows 1 set of data. based on this, I am asking about accuracy here, not repeatability. Repeatabulity comes next!

cheers!

 

[mikekilroy]

Essentially, we need accurate linear positioning (better than .0005")

each full step yields approximately .001"

How do you plan to overcome the impossible? (please don't say FM )

I cannot make enough sense out of the scale or values on X & Y of plot to comment. Title says 6400usteps so if each ustep is .001" seems like FST (I assume this means Full Scale Test?) is 6.4" across X....

Without pitch of the screw (2 start means nothing to telling us pitch - pitch is in units of "/rev; lead will do also - units rev/") and confirmation your stepper is std 1.8 degree /8 per step, no further comment on the cyclic error shown either.

Unless this were a very precision BALL SCREW driven actuator, it will have backlash - typical Thomson/IDC ones are .003-.010" (catalog page 11) so to try to position to an accuracy of .0005" is not possible. I know of no model by anyone who specs accuracy of .0005 in an linear actuator like that. Linear slides can do it but that does not appear to be what you have.

Then lead screws themselves have inaccuracy of typically upto .015" per foot (again reference thomson catalog on leadscrews, page 35). This is typically a wondering around of where you are vs where you should be of +/-.0075" max. Typical often is +/-.005" It CAN accumulate but that is not an issue for your 1" travel. So accuracy w/o feedback to correct for it over 1" should be .00125. Again, how do you expect to get .0005" from all these accumulating errors? You seem sunk before you get out of the starting gate. Sounds like a salesman picked this rather than an engineer. I would not have considered an electric actuator for my first choice here, for .0005" accuracy, let alone repeatability. BTW, what repeatability are you after?

If you list the model no of what you have, some folks on here may be able to look up the specs for you and offer data that pertains to your hardware rather than just generic typical specs. Maybe backlash is not of concern in your application, but most of these units do have .010" or more backlash - new from the factory.

http://www.KilroyWasHere<dot>com

 

[subsearobot]

thanks
I think the expectation was that the accuracy would be attainable using microstepping. or perhaps, management loosened the accuracy requirements to .001".

that said management outsourced this design to a contractor, as we are swamped with other projects. I believe that the contractor has other designs based around this motor, and are re-using said designs. but they may not have understood our requirements for precision. there is a serious lack of scope control with this project. I am coming into it late in the game, so I don't know the full story.

Backlash is not technically an issue (though I am not a fan). position is zero-ed at one end of stroke, every cycle, after a backlash compensation movement has taken place. so it's not the absolute position we are interested in, it's the delta.

the 6400 corresponds to microsteps (8*800), while measurements were recorded only on full step increments. I am still studying the graph too- it was created by another engineer, not specifically for sharing. but i thought it showed the periodic error well enough for an illustration.

there is no model# tag on the stepper... the leadscrew is 5/8" (or perhaps 16mm nominal?)OD. linear travel is stated by the contractor as 6.5mm / revolution.

my ignorance aside, I am very interested in this issue- I studied mechatronics in school, though my career thus far has not, until now, given me an opportunity to delve into precision motion control.

 

[mikekilroy]

so things are beginning to jive more. Good for you jumping in and learning Mechatronics!

6.5mm/rev * rev/360deg * 800step/chart * 1.8deg/step = 26mm or 1 inch in electrical terms close enough.

So your cyclic error is 4 cycles /chart or 1 CYCLE PER LEADSCREW REV

So your job is to find WHY you have a once per screw rev cyclic error. I have no clue what the amplitude (Y axis) is in real life terms so cannot comment if it is good or bad. EVERYTHING that moves will have all kinds of cyclic errors that look awful if blown up high enough. If we assume this cyclic error is large enough to be objectionable due to your asking, then it needs to be reduced. But having a handle on what its size is in absolute rather than relative terms will help. If this error is .000005" p-p then, sure it is fine and you won't get rid of it.

Next ask 'what causes 1x per rev cyclic repeating error?' I hear 2x per is usually due to a bearing problem. 1x per rev does not jump at me as typical cause.You know now it is not due to the linear guides in the actuator. Maybe the stepper was dropped at the factory and it has a bent shaft?

What does the straight purple line in the graph represent? maybe it is another clue?

I try to stay away from stepper motors and only use servos so do not have enough experience to relate some stepper related malignity; no 'common' one that I would have heard of jumps at me. You tried 2 drives; I would find it hard to believe the drive any clue about 1x/rev cyclic error so I would say it is not drive related. Can you turn the screw by hand and FEEL if there is a hard spot? Altho again, higher torque from a hard spot would even cause cyclic position error in a stepper since it moves in position steps. so Forget my thought of a bent shaft....

If your chart/scope shows this cyclic error happens at same locations along the 0-1" travel, you could learn something by removing the motor, turn it 45 degrees, put it back on, and see if the next chart moves this 45 degrees? Then you would know if it is in the motor or the screw maybe?

http://www.KilroyWasHere<dot>com

 

[itsmoked]

Mike a LPS for your continued teaching of motion subtleties. Thanks.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[subsearobot]

It looks like the amplitude is about .003". My partner cleaned up the graph, and labeled it a bit. the % error on the right is full scale error (error magnitude over full scale travel, which is 1")

to another way of looking at this, is that every revolution, the deviation from command varies by 1.2%.

we have a couple of motors here; I'm looking deeper...

 

[mikekilroy]

Have you asked the MOTOR mfgr what their max error is per rev?

Have you asked the actuator mfgr what their max error is per rev of screw?

Both are perfectly valid questions they should answer for you; may save you some time if the answer comes back on one of them as 1.2%....

If you want, post the step motor mfgr and part no, and the actuator mfgr and part no - maybe someone here has this info already...

My gut feel is any stepper driven LEAD screw actuator will easily have .003 in-accuracy or worse just as designed.... I think maybe you are trying to make that hardware do better than it was designed for.

http://www.KilroyWasHere<dot>com

 

[mikekilroy]

Thank you Mr. Smoke.

http://www.KilroyWasHere<dot>com

 

[subsearobot]

unfortunately, the motor tags are not present. on top of that, there is political resistance to us asking the contractor for the motor information.

we have a hunch that it is a Koco Motion/ DINGS stepper. if it is, they publish repeatability as +/- .006", and accuracy as .0006"/in. I am not seeing that accuracy over an inch; will be calling in the morning to clarify.

thanks for all the info, Mike. it's very helpful to get some background info and advise to look into in sorting this out. Our feeling is that this motor is not up to par, but want to be sure before we go to the contractor with that opinion.