High precision?

[jakovn]

I am not really making CNC machine but the 3D scanner and I need really high precision rotation! What should I be looking at? Is the answer in Stepper motor with gearbox? The resolution should be at least 0.05 degrees per step, so I thought 1.8 degree stepper motor with 1:40 gearhead but from the manufacturers I got the information that their gearheads can not guarantee that precision. Then I thought of microstepping but controller board manufacturers told me that even 1/16 microstepps are very uneven even with very good controllers and expensive motors.

The stepper will only rotate small laser and will rotate only 1/4 of the full circle, so should I be looking at servo motors?

Any recommendation please!!!

[Xerxes]

I assume that the load is frictionless. You can make stepper very accurate thru calibration. If you know the unlinearity of microsteps, then you can compensate it with software. You could attach an encoder to motor to get precise angle reading.

Servo motor with at least 16000 counts/rev encoder could be also a solution.

[jakovn]

Thanks for the advices! I think I will go with the encoder on stepper for angle correction.
What would be the easyest way for reading data from the encoder? Is there a way for direct interfacing to PC? How would you do it?

[Xerxes]

You could use parallel port as cheap encoder input but it may limit rotation speed. Other option is some PCI card having faster encoder input.

Let us know more about this 3D scanner project. Sounds interesting!

[jakovn]

It is pretty much the same concept as used by
http://www.muellerr.ch/engineering/l...er/default.htm and many others. But since all gear motor manufactures including one used in the project above told me their gearboxes do not provide that accuracy, have backlash and other errors appearing while using gears, I decided to ask here for advice. Encoder system makes sense but even after spending 10 hours reading about them I still don't understand enough to have a plan in implementing it in geared system.

If I understood it well If channel A=1 and B=0 and I compare it to the previous value I get the direction of the rotation if the value changed, and if values change faster than reading of the parallel port I might skip several changes and get an error, right?

One other reason for need for such high resolution is that I want to be able to scan large objects where the stepping errors would be noticable

I have found 8000 CPR encoder at http://www.mouser.com/Search/Product...GbBrEWiQ%3d%3d for 90$ but it is very expensive since I plan to have 4 lasers instead one in the system above

Would you give me advice to use standard dc motors because I guess there will be jittering using stepper motors so if the resolution of the stepper is 200 steps and encoder say 1000 then every 5 steps I would get an error (one motor step would report movement of 5 encoder stepps but maybe also 4 or 6)

[Xerxes]

Encoder produces quadrature signal as function of rotation. Every edge should be counted and skipping pulses leads to error.

8000CRP produces 32000 counts/rev after decoding. You could use that encoder+dc motor as precise servo motor. That's probably the easiest way to go but more expensive than steppers.

Have you considered toothbelt reduction? Zero backlash (i.e. TN profile toothbelt) could provide accurate angles.

[123CNC]

If you are looking at moving the laser rather than the target, perhaps a a Harmonic Drive will solve your resolution and accuracy problem. My memory isn't great, but I do believe they make some crazy reduction ratio's (100:1) with 'zero' backlash.

They aren't cheap, but often show up on eBay, with at least one guy that this seems to be his specialty.

http://www.harmonicdrive.net/

[Mariss Freimanis]

Steppers are excellent open-loop positioners. A quality step motor specifies a +/-3% non-accumulative accuracy tolerance. If a 400 step motor (100-pole motor) is used, the guaranteed accuracy is +/-0.027 degrees (0.03 times 0.9 degrees).

This error is cyclic with only a single minima and maxima over the span of a single revolution. You can therefore reasonably expect a 4-fold or better accuracy over a quarter revolution span. +/-0.005 degree unloaded accuracy is not unreasonable.

The strongest effect on microstep accuracy is motor linearity; its electrical to mechanical angle transfer function. A good motor exhibits better than 1% non-linearity over the span of a full step. Another strong effect contributing to accuracy degradation is hysteresis; a function of holding torque to detent torque ratio. Typically this ratio is 20:1 or better. Its effects have to be taken into account for resolutions greater than 30 microsteps per step on unloaded motors. We haven't done any measurements for thermal effects.

Mariss

[jakovn]

Thank You very much for the answers!

Harmonic Drive seems to be very expensive, toothbelt reduction seems good idea for accuracy but after drawing a bit I realized the system would be very large

I guess I am still by the geared stepper solution + encoder (I figured it out

The last 1,5 weeks I am contacting different companies who produce stepper motors and gears. None of them gave me the prices! Some were extremely expensive (200+ EUR) and India was extremely cheep (20 EUR for motor and gearbox (both from SAIA motors but with different names)).
Some replied ~Their systems do not provide that kind of accuracy~ and did not send me the prices.

Any ideas where in Europe (Austria preferably (or DE)) I could find geared steppers or at least gearboxes, and you have prices or positive experience with them? ... and the source for encoders as Xerxes recommended - with at least 2500CPR (That is 10000 PPR after, right?)

[Xerxes]

You don't need any gearing if you have encoder. Just use high resolution microstepping. You can utilize encoder two ways: to correct position (closed loop motor) or just to get real angle of motor (open loop motor).

Avago and USdigital encoders are probably the most affordable ones.

It might be best to try Mariss' approach first. Then add encoder if bare microstepping was too inaccurate.