Re: [CAD_CAM_EDM_DRO] Re: age old question

cnc_4_me wrote:

>I have a question about belt reduction for servos...Assuming the
>servo has enough torque in each example below. What setup will have
>the least positional error. I am looking at this from the prospect
>of using cheap DC motors as servos. Or asked in another way, what
>setup with these cheap motors will give better positional accuracy
>with the same control...
>
>Assume we want 1000 rpm on ballscrew.
>
>1) 3000 rpm servo 100 oz-in continuous, with 3-1 reduction...
>
>2) 2000 rpm servo 150 oz-in continuous, with 2-1 reduction...
>
>And finally, were will encoders give us the best control on these
>setups, on the motor, or on the ballscrew...
>
>
>My personal thoughts are the 3-1 reduction with the encoder on the
>ballscrew will give better positional control...
>
>

Putting the encoder on the screw, with a proper helically-slit coupling,
will give
the best accuracy, as any errors in the belts and sprockets don't affect
the encoder. This probably doesn't matter with rolled-thread ballscrews,
but with good ground screws, mediocre belts and sprockets definitely can
affect
accuracy. I buy sprockets with bores smaller than I want and rebore ALL of
them, as practically every one I have ever bought is at least .010" off
center, and
some were wildly off!

>With the encoder on the ballscrew the control will see a higher line
>rate per rev of the motor.
>
>

No, it is the other way around. With reduction between motor and
screw-encoder,
for every motor rotation, the encoder only goes through a PARTIAL turn.
So, you
may want to get a higher line-count encoder for this purpose. For
instance, a 1000
cycle/rev encoder gives 4000 quadrature counts/rev. With a 5 TPI
ballscrew, you
get 4000 * 5 = 20,000 counts/inch. That is a resolution per count of
.00005"
(50 micro-inch). That is certainly enough for VERY smooth motion, even
at very low
speeds. Unless you are making telescope mirrors by CNC, you are not
likely to
need more.

Jon