Resonances in a scanning mechanism

[logbook]

I am trying to understand the resonances in a scanning mechanism being developed where I work. The mechanism is driven by a stepper motor via an offset bearing and pushrod to give a reciprocating motion of a few degrees. The stepper motors are only of the order of 2 watts, so we are not talking about a large mechanism here.

The existing mechanism gives clear resonances as heard, and seen on a sound meter. With the sound meter a few inches from the mechanism the resonant level is up at 90dB compared to the non-resonant level at around 74dB. Hence the resonances are very pronounced.

Talking in terms of pulses per second to the stepper motor (pps), I can get clear resonances at around 250pps,500pps,750pps,1500pps. There are others at lower speeds as well.

Should I deduce that the resonance is actually at the 1500pps rate and the others drive the 1500pps resonance due to the harmonics in their rectangular pulse shapes?

This is a real problem for us. We have just spend loads of money with a contract mechanical design house to get a smaller mechanism made. This is half the size and weight but resonates in a similar way to the previous mechanism. Evidently the same mistake has been made again. What should I be looking for? (other than competent mechanical designers!)

My brand new mechanical design book on dynamics and statics has very little to say about resonances.

For me I think that this full metal design has loose couplings and no damping. The stepper motor pulses kick the mechanism hard and the whole mechanism is available to ring like a bell at certain frequencies. I therefore feel that some damping is needed but I can’t put my finger on where it should go. For example, if I put rubber mounts on the pushrod to the mirror the worry is that the position of the mirror will be uncertain (wobble). Is there such a thing as an adjustable damper so the response can be tuned up rather than calculated?

 

[GregLocock]

I don't think you can deduce which frequency your problem is just from the data you've given. You could record the problem on a PC and get the spectrum using Goldwave, for example.

Finding resonances in mechanical systems is quite hard, particularly if the system is very small compared with the wavelength of the problem. If you are lucky it will be a structural resonance of one part, in which case you can redesign that part with a different resonant frequency.



Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[electricpete]

"Talking in terms of pulses per second to the stepper motor (pps), I can get clear resonances at around 250pps,500pps,750pps,1500pps. There are others at lower speeds as well.

Should I deduce that the resonance is actually at the 1500pps rate and the others drive the 1500pps resonance due to the harmonics in their rectangular pulse shapes?"

If you have done a sweep and have confidence those are the only resonance within that range, my personal opinion is that you have come to the right conclusion.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[GregLocock]

1st 2nd 3rd and 6th harmonics only? funny shaped pulse.



Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[MikeHalloran]

It might be helpful to think of a stepping motor as a small rotary hammer.

When the current pattern in the coils changes, if there is enough current, the rotor changes position almost instantly. For a size 23 stepper, typically about a millisecond. For any given stepper and load combination, there is really only one step rate where it just cleanly goes to the next position and _doesn't_ ring.

A book and a thesis about "Phase Plane Analysis of Stepping Motors" or some such came out of the University of New Hampshire some years back. I think the authors are Taft and Gauthier. The material is opaque at first, but concentrated study provides a new understanding of stuff that is covered only empirically elsewhere.

You may be able to make some improvements with mechanical damping and/or phase current shaping, but I think if you want the mechanism to be able to scan at anything other than one rate defined by the mechanism itself, e.g., rate sweeping or arbitrary rates, you are using the wrong mechanism.

Start looking at voice coils.






Mike Halloran
Pembroke Pines, FL, USA

 

[logbook]

Well I did some surfing around this forum and found a recommendation about a book by Hartog (mechanical vibrations) so I have ordered that. Thinking abut the problem some more the sound is so loud that it ought to get picked up by a simple microphone (which I have) and a scope (which I have) and then I should be able to FFT it on the scope (I think it can do that) and that should tell me definitively if it is the 1500pps (1.5kHz?) that is the key suspect.

I agree with Greg that the harmonics are fishy and I will check the other frequencies more carefully if I get the chance. I am sure there is at least one more set of resonances below 250pps but I didn’t measure those with such interest.

Hopefully Hartog will arrive in a few days and will be enough to answer a few basic questions.

Is it reasonable to suppose that a gap in a bearing/pushrod will resonate at a lower frequency if the gap is bigger?

 

[GregLocock]

Yes... but not always. When you've got your mic and scope set up, and you've identified the frequency of interest, try tapping various parts of the mechanism with a pencil, screwdriver or hammer, that may allow you to home in on the part that causes the resonance. If it doens't then things are more complicated.


Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[TPL]

I'm not sure that your assumption of an increase in noise results from an excitation of mechanical resonance is really valid. It’s a good starting point, but before spending more money, I think you need to verify the origin of the noise that is causing concern.

What is your objective? Is it to remove the noise or is it to remove the ‘resonance’?. If this is a variable speed/excitation system, then you need to ensure that all mechanical natural frequencies are above or below the minimum or maximum frequencies of excitation – not easy (otherwise any changes you make are likely to just chase the ‘resonance’ to another frequency in the operating range).

If it is safe to do so, maybe you could run the mechanism at a ‘resonance’ for an extended period of time – this should result in a fatigue failure pointing towards the ‘problem’ component. If you don’t get a fatigue failure after a certain amount of time, then you could ask the question ‘do you have a problem that needs to be solved’?

I don't know if it is 'doable' but you might want to consider mounting the entire mechanism on a shaker table (especially if you can make a jig to allow shaking in x, y and z axes): fire up the table at the frequencies of interest and carry out a detailed visual examination (with no power applied to your motor) using a strobe that is slightly detuned from the excitation frequency. If you really do have a resonance, you might be able to ‘see’ the resonant component moving relative to other parts of the jig.

 

[MachineryWatch]

All good posts. In addition to the microphone data I would use a carefully placed very small accelerometer to measure vibration (very small to avoid mass loading the device). You should be able to use your FFT capable scope to look at the data to very quickly determine both which components are vibrating and at what frequencies it is vibrating. Design modifications are best undertaken when you eliminate the guess work! The small accelerometer (and whatever power supply is needed) can be rented from various places. If you are in the US you could call The Modal Shop in Cincinnati, describe the size of the device you are testing and they could recommend the proper sensor that they could rent you from their inventory. Their number is 800.860.4867.

http://www.machinerywatch.com

 

[logbook]

Thanks Greg. Nice idea.

TPL, the idea is to reduce the noise primarily but an undamped resonace will additionally give a scanning error which would be difficult to isolate.

Turns out the scope didn't have FFT capability. The other one does but it is not real time (takes 10 seconds to calculate). I have hired an analyser which should arrive the day after tommorow.

I'll have to look and see if I can get a small accelerometer as well. As I recall I went for a microphone rather than an accelerometer because a three-axis accelerometer is more complicated(?)

 

[MachineryWatch]

For what you are doing I wouldn't concern myself with getting a triaxial accelerometer. The benefit of triaxial is not worth the added complexity and expense from what I can tell of your application.

http://www.machinerywatch.com

 

[Warpspeed]

Gosh, this reminds me of when I was a keen but poverty stricken engineering student many summers ago, and built myself a complete XY plotter using stepper motors. The mechanism was rigid, and the software even worked, but oh the unbelievable noise !!!

The secret seems to be in how you drive the stepper. If you pulse it brutally it will behave rather like an impact air wrench, and behave accordingly. Trying to drive something with an impact wrench is never going to be a silent or a gentle process, no matter how the resonances are carefully rearranged, hehehe.

You need to research "micro-stepping". Instead of just switching the motor windings hard between supply and ground, they need to be driven with sine waves originating from a pair of digital to analog converters. The motor windings can be driven with sine and cosine waves (displaced ninety degrees). Motor torque will be really smooth, and the stepper will function all but silently.

It can also be driven to a position and held, in between "steps", hence the term micro-stepping.

I eventually bought a commercial XY plotter, but the learning experience and the challenge made the whole protracted drama well worthwhile.

 

[logbook]

Warpspeed, I am using a commercially available stepper driver card and just giving it logic pulses to advance to the next step position. I therefore have no control of the waveshape. I am already micro-stepping the stepper (to make the movement smoother). This gives 400 steps per rev rather than 200.

 

[Warpspeed]

It seems then that you have no choice.

Microstepping digitally is a cheap option that doubles the number of available motor steps by simultaneously energuising pairs of motor windings, but it is not really true microstepping.

With a three phase mains connected induction motor, you have three sine waves energising the triple windings phased 120 degrees apart. Because of the characteristics of sine waves this produces a true smoothly rotating magnetic field. No jumps or steps, smooth torque, and no noises.

Here is an application note how to digitally drive stepper motors with sinewaves:

http://www.st.com/stonline/products/literature/an/8700.pdf

 

[GregLocock]

As an alternative have you considered putting a flexible coupling in there?

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[Warpspeed]

That might work, unless positioning accuracy under varying load torque is an issue.

 

[logbook]

I didn’t have much luck with a tie-clip electret microphone clipped to the scan mechanism but a studio microphone plugged directly into an audio FFT analyser showed some interesting, if complex, results. The pulse per second frequency does appear most of the time, but I have seen harmonics and sub-harmonics as well.

A resonance at 167Hz gets stimulated at 167pps but when pumped by 334pps the 167Hz still gets excited. I have even seen the 167Hz come up when driven at around 1400pps. Since I am half-stepping the motor I suppose it is reasonable that the divided-by-two sub-harmonic is present.

I did try tapping the mechanism as suggested to see where it rings, but there seem to be so many excitable elements that the situation is far from simple. At the lower pulse speeds the response has so many harmonics the response looks like equal amplitude harmonics in a comb.

Since hitting the stepper hard makes it ring hard I turned the voltage drive down so the current ramps up more slowly (due to the stator inductance). Reducing the drive voltage from 40V to 15V knocked the resonant sound level down by 5dB, a good improvement.

 

[MikeHalloran]

Half stepping is always going to excite the structure because of the torque ripple.

You can make some improvement by shaping the coil current rise and fall, especially damping the induction spike on turnoff. Microstepping takes that a step farther by sinusoidally modulating the coil currents to achieve smoother motion. Modules to do it are getting cheaper all the time.









Mike Halloran
Pembroke Pines, FL, USA

 

[GregLocock]

If you use a program like Goldwave you will be able to produce a waterfall plot as you change the stepper speed. That's the way we usually identify resonances and excitations.

http://www.prosig.com/signal-processing/SoundSteering.html

Cheers

Greg Locock

Please see FAQ731-376 for tips on how to make the best use of Eng-Tips.