Understanding the positional accuracy of a servo

[astronomerroyal]

Hi,

Q. What determines the positional accuracy of a servo, and what are typical values?

I'm not very experienced with running servos, although I think I understand their basic operational theory.

The new digital servos mention 'increased precision.' Presumably the positional accuracy is detemined by 1) the bit-resolution of the circuitry that receives the PWM signal and 2) the precision of the feedback potentiometer.

Is that right, and do you know of any relevant specs. I looked at Hitec servo specs online and couldn't see anything about positional accuracy. Currently I'm using stepper motors to precisely turn a camera platform, but it's difficult to add a useful feedback element, like a shaft encoder, to something that already rotates so slowly.

What I'd really like is a simple replacement for the stepper - something with feedback so i know exactly where I'm pointing. If i replaced the (1.8degree) stepper with a (360 degree limit) servo, do you think I could get the same 1.8 degree (i.e 200 positions per revolution) resolution out of the servo?

 

[Nigel Goodwin]

I can only suggest you get one and try - but can you easily get a 360 degree servo?.

Assuming the existing system gives enough accuracy, how about adding feedback to the existing stepper system? - link a pot to the movement, although this will only give you 270 degrees. Perhaps you could find a 360 degree pot?.

 

[dknguyen]

They don't have accuracy specs for RC servos.

Also, there are not really 360 degree servos because of the limits on the rotation of the potentiometer as well as the gears have stops (and sometimes not even full round gears).

I also doubt there is much difference between the pots used in analog servos vs those used in digital servos. The measurement circuitry is probably more accurate, but in my mind what it realy reads is:

"Digital servos have increased precision in the measurement circuitry of the pot and/or PWM resulting in increased precision...all other things being equal like potentiometers*

*Potentiometers will vary over time resulting in varying precision"

Can't you use an absolute magnetic encoder? Like the MAE3 at US Digital? It doesn't matter how slowly it rotates then...then it again why does it matter at all how slowly the thing is rotating when as long as the encoder has a resolution greater than what you need?

 

[astronomerroyal]

Servocity.com, for example, sell a variety of servos,some of which have 360' or multiturn capabilities. Some with 'milspec' pots. They even have continuous rotation 'servos' without the feedback pot.

Thanks for the magnetic encoder pointer, I was looking at optical ones on ebay and didn't realise you could get encoders with such fine resolution.

What I'm after ultimately is something really fast but really accurate. Like this,

**broken link removed**

I've actually made a somewhat less sexy 2-axis rig (the controller uses a PIC uC, naturally) for the same purpose (motion timelapses), and although my stepper is fairly precise it's also very slow. I thought a servo type motor (which is what camblock uses) with its feedback would be faster and (potentially more) precise.

Thanks for the replies.

 

[Triode]

but it's difficult to add a useful feedback element, like a shaft encoder, to something that already rotates so slowly.

not about servos, but I've seen a cheap and accurate way of measuring rotation used in printers, it involves an emitter/sensor in a single piece and a clear wheel of plastic sheet with radial lines printed on it. Its accurate enough that the printer usually doesnt mess up the line spacing. And you could even get the needed wheel and sensor from a printer, people throw those out all the time.

 

[shingadaddy]

Analog: Sine / Cosine potentiometers. 360 degree capability. Resolution near infinite.

Digital: Optical shaft encoders (As Triode mentioned). Resolution based on COUNTS PER REVOLUTION. Motors I use have 2000 and 4000 counts per revolution typical. And have an INDEX pulse that occurs once every revolution.

You can also get a QUADRATURE shaft encoder that puts out 2 or 4 bit grey code and do the counting externally.

Also this is precision of MEASUREMENT. The actual positional accuracy of a servo is determined by the physics of the mechanism and the capabilities of the servo drive, effected by mass friction, gain, damping blah blah blah...
All the standard PID stuff.

Steppers of course have a set number of degrees per step that governs their accuracy however you can micro-step them. I seem to remember 1.8 degrees per step being the COARSE step size that was common on some larger motors.

 

[shingadaddy]

Watch for smart motors from Animatics on Ebay. Some are already there now. You can look up their specs on the Animatics website.

**broken link removed**