I'm chewing through a tough problem, and wanted to know if anyone could shoot down or confirm a few ideas I've been bouncing around, and help out with their experience.

I have two cylinders with the top face of one connected to the bottom face of the other. The cylinders can spin about the Z axis (think two cans, stacked one on top of the other, spinning the top can). Is there any way to measure the relative rotation of one to the other with +-.005 degree precision? The cylinders will not spin more than +-5 degrees relative to one another. The system has to be mounted to the side of each can, however there may not be anything rigid between them (they should be able to rotate freely), so 'soft' strain gauges are ok, metal braces with potentiometers are not.

Ideas I've had so far:
Gyroscopes paired with accelerometer to make an IMU
Electronic compasses (difference between magnetic norths)
Inclinometer (?)
Fiber Optic Gyroscopes

 

That +-.005 degree requirement is a killer: IMHO, you'd be pushed to measure that on 'stationary' targets...

Unless you can get away with strain gauges, best I can think of is crossed optical gratings for the 'vernier' combined with optical encoder wheel for the 'coarse' reading.

 

.01 degree precision would be acceptable, I could just clamp my apparatus down

 

It's going to be one heck of a fine optical grating but that's about the only logical solution I can see.

 

Does anyone have experience with any of the methods I suggested in my post? I don't think a custom optical grating is in my skill level yet...

 

If you have anything built up already, perhaps you can take a photo of it and post it here. If not, maybe a sketch of the arrangement would be helpful in moving this discussion along.

hgmjr

 

Although you have specified the system 'must be mounted on the cans' presumably the information is required elsewhere?

So can you not use an external optical sensor to read a vernier mounted half on each can?
Since there is only very limited angular displacement a linear vernier would do be sufficient to measure this.

 

What is the diameter of the cylinders? Potential angular resolution should increase with increasing diameter.

 

An identical line scale mounted on each can and optically sensed could be used. One sensor would feed the UP line of a counter and the other would feed the DOWN line of the same counter. Any difference from zero would indicate the degree of displacement.

 

hgmjr is right -- post some pictures, dimensions, and drawings. You'll get better answers. You first say 5 millidegree resolution; then change this to 10 millidegree. Changing resolution specs can change the problem significantly, so you probably should also specify what exactly is being solved and what the constraints are. You should also give the budget involved -- if this is an industrial project with reasonable funding, then this opens up solutions that won't be available to a hobbyist on a limited budget.

Also state whether the measurement has to be absolute or relative. An absolute measurement is where the system can power up and tell you the offset of one part to another without any initialization; this usually involves things like encoders that have fixed scales on them -- you're essentially "reading" that the encoder says 23.45° angle on a scale. A relative measurement is when the system is started up, some angular position is defined to be zero, then subsequent measurements are with respect to that reference point. For example, digital readouts for machine tools often work this way (or a machinist's electronic caliper -- you have to first set the zero point; the electronics just count divisions for you).

IOW, the more information you can supply, the better answers you'll get.

My personal recommendation would be to, if you can, design the system so that an off-the-shelf part like a rotary encoder could be used. Your comments indicate that you may not have such freedom because you're trying to retrofit an existing device.

The "obvious" solution that comes to mind is to put an optical source on one device along with a sensor, then mount a transparent scale with divisions on the other device. Relative movement lets you count divisions. 0.01 degree resolution is conceivable depending on the radii involved -- you're talking 36,000 divisions around a circle. Maybe these folks have some parts you could use -- there are lots of sources on the web along with design information.

As to a gyroscope, I'd inherently shy away from something like that because of the complexity and cost (even a ring laser gyro would probably be too complex, massive, and costly). Besides, they provide acceleration information which you have to integrate twice to get position information (which adds to the complexity).

Magnetic compasses would be a challenging design -- even getting a compass to read accurately to 0.1° would be hard and you'd have to worry about tiny stray fields.

Oh, also consider a centuries-old technique -- use the optical lever. It's low mass, effective, and cheap.

 

Without a diagram or picture of some sort, the members of this forum are reduced to playing the Internet version of twenty questions.

For example:

Can I assume that one of these cans is attached to a motor and is being spun at a given RPM while the second can is free-spinning and suspended a few millmeters above the driven can? Your task is to induce a rotation in the free-spinning can that not only matches the rate of rotation in the driven can but the two cans would be in synchronous rotation such that if both cans had a marking on the side of them these two markings would maintain a fixed angle between them that would not vary more than +/- 0.005 degrees.

How much time are you alotting yourself to accomplish this fascinating challenge you have taken on?

hgmjr

 

Looking through all the sugggestions many are potentially useable, close or will lead to something.

If you've got the room mount a mirror on each, shine a laser beam at them then detect the reflected beam at some distance as it passes.

 

Sorry about 'with-holding' information, it wasn't my intention, I didn't realize the problem was more in depth than it seemed when I first proposed it to myself.

This is on my budget, hobbyist.

The cans can only rotate freely between +2° and -2°. They have stops after that. They do not spin, their movements are from external forces acting on them.

I guess I'll keep working on this until I'm done.

10 millidegree resolution would be a good proof of concept. If I can get it to < 10% error, I'll be satisfied.

The system only needs to be relative; the angle output should be the angle of rotation of the cans relative to each other (top can to bottom can).

Wouldn't MEMS gyroscopes work? Perhaps in an IMU configuration?

Here is an extremely rudimentary sketch I threw together to illustrate my point. I can fashion something in Solidworks if it's required.

I just realized when I drew this, the bracket on the left pointing to the two cans is a little misleading; it illustrates that I cannot have anything rigid between the cans (that would ruin the purpose of the experiment).

http://personal.stevens.edu/~jlupinsk/cans.png

 

This is on my budget, hobbyist.

The cans can only rotate freely between +2° and -2°. They have stops after that. They do not spin, their movements are from external forces acting on them.

I guess I'll keep working on this until I'm done.

10 millidegree resolution would be a good proof of concept. If I can get it to < 10% error, I'll be satisfied.

The system only needs to be relative; the angle output should be the angle of rotation of the cans relative to each other (top can to bottom can).

The picture is helpful. You mention that the two cans do not spin. From your description, I take it that each of the cans are free to rotate but are mechanically constrained to do so only within a 4 degree span.

Can you reveal the application? That would perhaps give our members a better idea of what you are attempting to achieve.


hgmjr

 

That does make it 356* simpler.

 

http://www.ruhle.com/rotary transducers.htm would work but you need deep pockets.