Re: Single Servo circuit

[Ray <rehenry-at-up.net>]

On Sun, 22 Oct 2000, garrett wrote:
<S> 
> I'm using a small pittman gearhead motor with integral
> encoder.  I don't have any specs on the motor, but
> it's about 4" x 1.5".  The encoder has 500 lines.  The
> counter chip can do the quadrature X4 multiplication
> to give 2000 "steps" per rev.  For the system you
> looked at, was the 2000 the line count? or was it
> after the quadrature  multiplication? 

Good question. I think that they were using a real 2k ppr but I'll have to
ask.  

> > Were your choices of PID and FFn trial-and-error or
> > did you use a
> > tuner?  What are your impressions of the accuracy,
> > stability, and such of
> > the amp/motor/hardware that you use?  
> 
> Very much trial-and-error.  In fact I'm quite sure
> they are not even the "best" values :) 
> I don't think that the Analog portion or the circuit
> is "up-to-snuff".  I also think that the gearhead
> causes problems due to the backlash.  

If the gears aren't between the motor and the encoder, this shouldn't make
any difference.  If they are, then that would be a real source of problems.

>The accuracy of
> the system does not seem to be terrible.  On the emc
> display it holds position within a few thousandths of
> an inch (I haven't checked what this translates to in
> degrees of shaft rotation).  The steady state error
> improves if I add just a teensy bit of integral, add
> too much and the motor runs away.  I'm just playing
> with an unloaded motor right now, and if I add almost
> any derivative, the motor oscillates.  I wonder if the
> backlash of the gears causes this? 

If my memory serves me, Jon spoke of a similar experience with
derrivitive with his servo setup.  I think he found that an FF1 of
about 6-8 worked much better than derrivitive.  You can read more about
his experience at:

www.linuxcnc.org/handbook/part2/tuning/pidtune.html

> I think the next thing I will do is use the amp in
> torque mode.  That is control current rather than
> voltage.  I have read that this setup is easier to
> tune.

Let us know how you do this and what you find.  Another one of those xfig
drawings would be great.

Ray