Re: DAC resolution

[Jon Elson <elson-at-pico-systems.com>]

Dave Engvall wrote:

> List:
>  
> Anyone care to comment on the DAC resolution necessary to adequately 
> control a servo system on a high end hobbist milling machines? To put 
> it another way: at what resolution can one see the difference in 
> performance? I'm assuming 1 KHz updates using emc.

There are ways to calculate this effect.  It is a non-linearity in a 
supposedly true analog
system, ie. quantization error.  For a 12-bit DAC, that would have 4096 
discrete
velocities, but only 2048 in each direction.  So, it would have a 
resolution of
1/2048 = .049 %  (My PPMC card with its 16-bit DAC is admittedly overkill,
if there was a DAC with all the features of that chip with, say, 
14-bits, it would
have been perfect.)

Anyway, you can calculate the raw positioning error from the above info.
Say you have a max speed of 60 IPM or 1 IPS, and the servo loop checks
position error every .001 S.  So, 60 IPM is selected by a nearly full-scale
output.  If the error is one full DAC bit, that is 1 IPS +/- .049% or
1.049 IPS.  At 1 IPS, that means that in .001 S, it will have traveled
0.001049", a positioning error of 0.000049" (49 uIn), which is MIGHTY
small.  it gets more complicated to figure out what effect this has on
servo loop stability, and how much you have to lower system bandwidth
to keep it stable.  But, unless you are running a machine at 1000 IPM
or pointing telescopes or diamond-turning optics, even 12-bits seems
good enough.

Jon