Filter wheel indexing solution

[htngwilliam]

Hi

I am trying to design a filter wheel indexing system. I have a wheel that is sub-divided into 3 positions, 120 deg apart. This wheel will be driven by a stepper motor with a timing belt system with a ratio of 2: 1. I require the wheel to have positioning accuracy of +/- 2 steps.

I face the following issues: -

1) When stepper motor stop, filter wheel sort of vibrate and does not settle until 1 sec later. I wonder how I can improve this

2) Is timing belt system good enough for this kind of precision motion?

3) Is there any way to calculate the system stiffness. I know K [Angular stiffness] = M [Moment]/ Teta [Angular displacement].

 

[drawoh]

htngwilliam,

Are these optical filters? If the optical filters are flat glass, you have no need for accurate centreing. All you need to do is capture the whole aperture.

If you need more stiffness, you can use a heavier duty timing belt. You can use the English series, rather than the metric series. The metric ones are more flexible, and more suitable for high speeds.

Is there any reason you cannot mount the filters directly to the stepper motor? If size is an issue, do you really need full 360[°][ ]rotation? An asymmetric filter mount might give you your three positions, and be more compact than a round one. Just because it is asymmetric does not mean you cannot statically balance it.

Are you or your programmer assuming that the stepper motor starts and stops instantly? An acceleration of infinite violates one of Newton's laws. Try ramping it up to speed and down from speed.

Here is an alternate way to look at your actuator. Consider how you would analyze the problem of me throwing rocks at your filter. In your structural analysis, you would assume that the impact of the rock would be a force or torque of some value operating over some time period. The time period would be a couple of milliseconds at most. Is this what your stepper motor is doing?

How fast does this thing need to index from filter to filter? Perhaps you can slow it down.

JHG

 

[htngwilliam]

Hi drawoh

Thanks for your reply.

1) I need positioning accuracy because the filters are mounted 90 deg with respect to the light source. In order words, any shift will cause an angular error which will affect the light path

2) I wonder whether you have any good timing belt vendors to recommend? I presume if I choose one that can take a higher load, it is equivalent to more stiffness

3) Someone suggest using larger pulleys as it will help to improve the vibration and give good repeatability but I dont understand why this will help, wonder if u can advise

4) I cant mount the motor directly as I need to use a smaller motor to drive a higher load so I am doing a 2: 1 ratio to amplify the torque

5) Currently, I am working with the firmware engineer to use S-curve control

6) Yes, the settling till will be about 0.3 secs at most

7) I wonder how I can balance the mass if I do asymmetric filter mount w/o adding material at the other size to balance it. If I did a asymmetric mount, it might cause unbalance stress to the bearing.

8) I hope to achieve a index time of 0.5 sec inclusive of settling time

9) I wonder how do I actually work out the stiffness of the system by doing theoretical calcualtion? Do I need to analysia all the torsional stresses and shaft deflection?

 

[MikeHalloran]

Turn the outside of the filter wheel into a flat pulley, and drive that with the stepper. You only have to embed two teeth in it to retain positive drive.





Mike Halloran
Pembroke Pines, FL, USA

 

[unclesyd]

Could you possibly use a Geneva drive.

http://www.tangendrives.com/index.html

 

[htngwilliam]

Do I still need a belt system if I use a Geneva drive?

Mike

I dont understand what u mean.

 

[MikeHalloran]

You're probably not running the motor fast enough, and need more reduction.



Mike Halloran
Pembroke Pines, FL, USA

 

[unclesyd]

We use a gear head motor to drive some Geneva mechanism in some of our machines. The ones we use have a Max speed that I can't recall.

I would give Tangen a call.

 

[drawoh]

htngwilliam,

I was visualizing filters rotating in the plane of the filters. 2X[ ]90[°] is still an alternative to 3X[ ]120[°]. If the axis is vertical, there is no need for balancing.

Ask your optics people to work out the effect of an angular error. If your filter is not exactly perpendicular, your light beam will shift sideways by some distance. That distance will be tiny, even if the error is ±2[ ]steps on a stepper with 180[ ]steps per revolution.

Timing belts are totally standard. Any power transmission vendor will have them.

Search W.M.Berg for Geneva mechanisms. These are an alternative to timing belts. Also, consider a gear drive.

Consider all the hardware needed to mount a timing belt. By the time you get everything attached, and make everything rigid enough that your vibration problem goes away, the 2:1[ ]reduction probably is not saving you very much size and weight. I would do a design layout with just an appropriately sized stepper motor, in addition to whatever it is you are doing now.

You do know that steppers are sized by rotor intertia, rather than torque, right?

JHG

 

[htngwilliam]

Hi

1) I dont understand why no balancing is required. Mechanically, on a free body diagram, the forces and cg will be offseted.

2) I wonder what is the positioning error for timing belt system? +/- 2 steps?

3) Is it possible to drive the filter wheel 2 direction based on timing belt system? I think there will be backlash issue.

 

[unclesyd]

Still another senior moment.

Here is the ccompany (Tangen) that I referenced in my previous post and left out the url.

http://www.tangendrives.com/

 

[MikeHalloran]

Htngwilliam, you appear to be in over your head. It's time to rent an engineer with experience applying stepping motors to dynamic systems.





Mike Halloran
Pembroke Pines, FL, USA

 

[drawoh]

htngwilliam,

If your rotation axis is vertical and your system is unbalanced, gravity does not affect its position. If the axis were anything other than vertical, gravity does affect the position.

An advantage of unclesyd's Geneva mechanism is that your system would remain in place even with the power off. Steppers lock in place, but they consume power doing it. All of that power is given off as heat.

We actually do not know how big your steps are. The 180[ ]steps per revolution I noted above, is fairly standard with stepper motors, but there are lots of exceptions, and lots of clever stepper drives.

The correct timing belt, with good design, can have close to zero backlash and error, but these things are relative. If it really matters, you can order a harmonic drive with a chart showing the measured error of that specific mechanism. For your application, this almost certainly is gross overkill.

How much error can you tolerate?

JHG

 

[htngwilliam]

Hi

1) Yes, now I understand. I would say in static, the wheel could be unbalance but during dynamics state, it doesnt affect.

2) Our firmware engineer program with 20k steps where the step size is 0.018deg.

3) I will talk to the optics engineer whether this shift is critical

Thanks for all the help. I totally have no experience in this area and I appreciate all the comments and advise given.

 

[unclesyd]

One way to alleviate some of the problems you have mentioned is to use multiple stations of each filter like 6 instead of three. This helps in multiple ways, it ameliorates problems with inertia, control, radial speed, shortened transit time, etc.
This would apply to both a Geneva or timing belt system.

 

[drawoh]

htngwilliam,

2) Our firmware engineer program with 20k steps where the step size is 0.018deg.

Always ask about the tolerances for optics. I had a chat with a machinst at a local university, early in my career. He thought it was hilarious how engineers would specify housings for windows and filters at Ø1"+.002/+.001. Windows and filters are cut from round glass bars. The cuts are not necessarily perpendicular to the axis. It is not as if windows and filters have to be accurately located, anyway.

You can very crudely estimate the sideways shift through incorrectly rotated filters. Assume the light travels straight through the filter, i.e. perpendicular to the face. If your filter is 3mm thick...

SHIFT = 3mm[×]sin 0.036[°] = 0.0019mm.

I believe, although I am not 100%[ ]certain, that the error will be less than this.

JHG

 

[MikeHalloran]

Interference filters also shift wavelength when off-axis.
That may be a more severe problem.



Mike Halloran
Pembroke Pines, FL, USA

 

[unclesyd]

If you have the time you should look at these direct drive roatary tables.

http://www.intellidrives.com/Default.htm

 

[htngwilliam]

Hi everyone

Thanks! I think I will build the prototype based on the feedback. After which I will check out what is the settling time. I wonder whether it is possible to calculate the expected belt elongation over time if I am given the tensile modulus. I am aware that the tensile modulus is a function of belt length and the tensile force.