Microradian motor shaft angle control

There may be an aerospace quality resolver that would do that, but the ones that I know of are about an order of magnitude too noisy and about an order of magnitude bigger than your itty bitty motor.

Jewel bearings may not be the way to go. A properly working jewel bearing leaves the shaft not quite centered in the jewel. This ambiguity in shaft position and 'normal' angle will complicate any attempt to measure the 'radial' angle.

Thanks. I dimly recall that the shaft rolls around the interior of the sapphire doughnut, so I suppose you're quite right.

I might also try two encoder sensors mounted at (say) 90 degrees from each other, which would allow me to take out tilt and decentration of the encoder wheel. Avago makes some nice 2500-7500 pulse/rev incremental encoders that might be able to be hacked like that.

The other nice thing about multiple pickups, ISTM, is that it would be possible to sort out motor jitter from encoder jitter--the motor jitter could be made common mode and the encoder jitter is all 90 degrees out of phase.

Then there's Mr. Newton, who constrains some of this stuff pretty tightly. (I have lots of ideas, just not much motion control experience to help me weed out the stupid ones.)

Cheers

Phil Hobbs

One arc second is 1/1296000th of a circle; I don't think that a 2000 count encoder is going to have the accuracy you need.

For that matter, one arc second is 1/1296000th of a circle; if you put a one meter lever arm on your shaft one arc second of rotation will move the end of the arm by a hair more than one wavelength of red light.

Not that I'm trying to naysay or anything...

Thanks.

I'm trying to do Radon inversion of phase shift plots from beams rattling round inside a rotating prism, with a 40-50 mm path difference, and am operating at 350 nm. So given that I want my data equally spaced, I get right up into that territory.

A 6000 line encoder with Moiré readout, with the tilt and decentration fixed, needs about 100x-200x interpolation to do the job. My hope is that the actual irregularities of the motion are on a coarser scale than that (e.g. the first 10 harmonics of the number of motor poles), so with a bit of help from Newton's laws, I can do pretty well by interpolating. I don't mind measuring and remembering the error of the individual encoder lines, and with independent readouts at different phases, I'm hoping I'll have enough info for the job.

How good are bearings?

Cheers

Phil Hobbs

...

Bearings are better when moving continuously without reversing. I built a laser interferometer to provide position feedback for the carriage of a ruling engine driven by a leadscrew*. Position had previously been measured by a rotary encoder on the leadscrew. The first thing we learned from the interferometer was that each time the carriage was restarted, it actually backed up a few microns before going ahead. The behavior was traced to the dynamics of oil films.

Jerry ____________________________

• Two inches diameter, and the support for one side of the carriage.

Interesting, thanks. What did you wind up with by way of bearings? Preloaded ruby balls? Steel balls? Precision sleeves? The more detailed the wisdom, the more useful in this case.

Cheers

Phil Hobbs

I thought you were quoting 1000 lines, and given the size of of your motor figured you couldn't go to a larger encoder. 6000 lines may do it

-- I'd check noise, though. You can get some encoders with impressively high resolution if you don't mind something that's over a foot in diameter. And note that I say "resolution" -- getting that sort of accuracy is going to be a huge science project no matter what you do.

Whenever I ask a mechanical engineer that question he says "how much money do you have?"

If you're designing the machine for production -- even low volume -- you can find a custom bearing supplier and explain what you want and why, and they will help you specify a bearing that they can sell to you for huge piles of money.

If you really want to know more ask me off list and I will ping my mechanical engineering friends for names. Expect a few more zeros on the price over equivalent-sized skateboard bearings.

I supplied oil under pressure to the existing bearings to keep the film from collapsing when the machine was still. Normally, the turning shaft wedges oil into the gap at the bottom. Large machines like power-plant generators and turbines settle into metal-to-metal contact when still for a while and suffer damage with every restart. Typically, power plants use "turning gear", electric motors that keep machinery like peaking generators moving slowly when not in use. They must know about pumping just the oil because that is done in the thrust bearings of vertical turbines. (It takes a lot of power to put a generating station on line.)

Correcting the anomalous motion wasn't high priority once the interferometer was in the feedback loop. Only rarely were lines needed very close together. After all, the carriage motion was erratic only in detail. The accuracy of the leadscrew guaranteed the accuracy of the carriage's gross motion.

Jerry

Take a look at Sony BH-20 Encoders. A little larger than 9mm (16mm dia). With a little interpolation they will get you in the range. At 115200 lines, a 16x interpolation gets your resolution below an arc-second and is easily accomplished with a simple lookup table. Also I' not sure if your low speed will work, but consider a self generating aerodynamic bearing like laser scanners use.

Thanks, those are really cute.

Cheers

Phil Hobbs

Passive air bearings are very good and can be very stiff. Some large pumped air bearings are called hovercraft. Pumped oil bearings work well also.

Jerry

Thanks. I'm trying to keep this gizmo light on the hardware and heavy on the software. Sony doesn't seem to want to talk to mere mortals about their encoders--you have to register before they'll even send you a data sheet. I'll probably have to wear a pigtail and bang my forehead three times on the floor before they'll speak to me.

Cheers

Phil Hobbs

they are tough getting information from, and you may find that they don't really understand what they have. Stay away from their interpolation electronics as they are pricey.