Precision angle (0.5° or better) with a motor to test angular sensor

[TheCamel]

Hi,



For a project, I am using a Hall effect absolute rotary encoder (AS5048B from AMS: http://www.ams.com/eng/Products/Position-Sensors/Magnetic-Rotary-Position-Sensors/AS5048B). It claims to have 14-bit resolution, but I need to prove the precision is at least 0.5° (required for my project). In order to prove it, I was thinking of using a motor (I was more specifically thinking about servos or stepper motors) to precisely position the magnet, then read the sensor value and compare it with the angle imposed to the motor.

However I don't know how to achieve such a high precision level with a limited budget. Is it worth it to invest in a $20+ servo (like this one http://www.tohobby.com/high-precision-tr213-servo-180-degree-rotation-13kg-accuracy-of-0-5-degrees.html)? Should I use a stepper motor with a larger step than my resolution and use gearing to obtain such a small angular precision? Is there a way to achieve this with a 'normal' DC motor?

I am a bit confused on how to obtain a such precision without having to pay too much, as my budget is limited to $100 for the all project (including these verification means).

Thank you in advance,
TheCamel

[mikeselectricstuff]

A stepper will give 200 or 400 steps, and the accuracy of each step position should be well under 0.5 degrees.

[Yansi]

By salvaging some old hard drives (five and a quarter inch or eight inch), one can get sometimes rather bizzare 5phase high precision steppers out of the disk drives...

These have usualy 500 full steps/revolution, and microstepping is possible.

http://www.orientalmotor.com/technology/articles/2phase-v-5phase.html

[Rerouter]

well straight off the bat as the non linearity is quoted at 0.8 degrees, there is a possibility for any one point to be out by 0.8 degrees and still be in spec,

as for measuring it, i would think a piece of rod, a bearing and a big piece of mm grid paper would be the most repeatable, and reliable, align the sensor with a cross on the grid paper, and note the X and Y vs where it sits, the firmer the rod, the less error introduced, so something like 6mm threaded rod would work a treat,

[mikeselectricstuff]

Don't forget you don't need anywhere near 720 positions  to show 0.5 deg accuracy 

[daqq]

If you're on a budget, go for ebay. You don't need a to go with a precise motor necessarily - you need a precise measurement system to get the position - try finding a rotary encoder with enough counts to satisfy your requirements: such as this:

http://www.ebay.com/itm/New-ABZ-three-phase-Encoder-1000P-R-Incremental-photoelectric-Rotary-Encoder-/321348779657?pt=LH_DefaultDomain_0&hash=item4ad1e13e89

. At 1000 pulses per rotation you are getting a measurement resolution fine enough I think. There are other, more precise types as well. The cheap option might be to couple this to a cheap motor and the magnet.

[Paul Moir]

I'm with Rerouter.  If you need to prove it just once why build a rig that will likely take longer to build than to measure it manually?  Also what proof do you have that a stepper or whatever has been made to 0.5 degree accuracy?

If you're doing this at a University, drop by the mechanical engineering department or physics if they have a machine shop to build experiments.  If they're not terrible, they'll have something to accurately and easily measure the angle.  Otherwise graph paper should be fine.

[f5r5e5d]

mechanical tolerances, magnet uniformity will make building to a high accuracy spec difficult

getting a sufficiently resolving, repeatable assembly you can then trim, calibrate for accuracy

you need to understand the distinctions, techniques used in measurement re precision, resolution, repeatability, accuracy

angular sensing could be done with the very big hammer approach - most have easy access to multi megapixel cameras today, with a optically patterned/structured target I bet code for extracting image rotation running in Matlab/Ocatave or Python could be found
 

with rotation center near the image center even the camera optics don't have to be that great -  just rotationally symmetric - I'd still use something better than the one on your phone though

[Richard Crowley]

I have seen many surplus stepping motors for sale that have a reducion gear-head which likely exceeds your requirements.
For example:  http://www.allelectronics.com/make-a-store/item/smt-107/stepper-motor-w/gear-train/1.html



http://www.robotshop.com/en/12v-17a-416oz-geared-bipolar-stepper-motor.html

[f5r5e5d]

as long as you only turn in one direction, and the gears aren't too eccentric and don't have high spots and bind changing the torque load...


vernier protractors could be cheap, even 1 degree marked miter protractors could be read to 1/4 degree with care

[dannyf]

I need to prove the precision is at least 0.5°

The simplest would be a rotational sensor (optical encoder for example).

If you are really on a budget, you could use a capacitive / inductance sensor. Measuring the output frequency allows you to measure the capacitance / inductance -> relative position.

[max_torque]

Clamp a 1m long bit of stiff plastic tube to the shaft, Measure how far linearly you need to move the end sideways to change the output by 1bit.  Then use basic geometry to work out the angle.   You can easily measure the length of the stick and the linear distance to sub 1mm,  which is better than 0.05deg resolution in your measurement error!

[pmbrunelle]

I wouldn't even bother trying to use an angle reference.

Instead, rotate the magnet at constant velocity. Attach the magnet to a heavy flywheel that spins fast... maybe the end of a bench grinder shaft?

If you can estimate the inertia of the bench grinder, and you estimate the torque ripple from the electric motor, you can have an idea of the possible velocity variations within a revolution, and from the velocity deviations, you can calculate the worst-case position deviations.

[mathsquid]

Clamp a 1m long bit of stiff plastic tube to the shaft, Measure how far linearly you need to move the end sideways to change the output by 1bit.  Then use basic geometry to work out the angle.   You can easily measure the length of the stick and the linear distance to sub 1mm,  which is better than 0.05deg resolution in your measurement error!

I was about to make the same sort of suggestion.

[f5r5e5d]

a bit awkward walking the ruler measurements and hand entry, trig calcs all the way around the circle - with expected nonlinearites from geometric tolerances and magnet field you really want to measure/calibrate the full rotation range