Re: Home brew servo board

[Jon Elson <jmelson-at-artsci.wustl.edu>]

Darrell Gehlsen wrote:

> What is the current status of the home brew servo board?

I am working hard on a parallel-port connected scheme for
having DACs, encoders and digital I/O modules that plug into
a ribbon cable that connects to a PC parallel port.

> I just had a thought for an add on for running steppers from EMC.
> If you use the servo board to feed +/- 10 volts to a Voltage Controled
> Oscilator and drive the stepper driver with the output of the VCO that would
> give you a smooth step signal.

This is just slightly short of brilliant!  The only reason it ISN'T brilliant
is that it has been done before.  But, it might just completely solve
the frequency / pulse jitter problems with EMC!  The only tricky part
of this is to make a circuit block that converts a BIPOLAR signal
to frequency (step) and direction signals, and works very cleanly
around zero.  A simple analog scheme might tend to hunt maddeningly
near zero - in fact I would expect it to.  A better approach would
be an all digital scheme.  You have a fast digital counter, and a
preload value tells it how often to produce a step.  There would be
a bit (or count value) that prevents it from stepping, and a direction
bit.  This is pretty close to what freqmod is doing, BUT, a chip can
do this a WHOLE lot faster than a CPU.  So, if you had a 10 MHz
clock, you would have the ability to program step pulses in 100 nS
units, about 100 times finer than with freqmod.  The chip could
also have a position counter built into it, so it keeps track of relative
motor position.

I have been doing some designs for a 4-channel quadrature encoder
counter and parallel-port interface one one chip, using the $18
Xilinx Spartan XCS10 series chips.  I think we could make a
stepper version with at least 2 channels of this type circuit on
one of these chips.  The software would look essentially like
the servo version of EMC.  The one gotcha is that an emergency
stop, or any other fault condition would cause an abrupt stop,
and the motor position would no longer be in sync with the
position counter.

> You could then either use an encoder for feed
> back to EMC or for an open loop system just feed the VCO output back to EMC.
> Now all you have to do is "tune" the PID for your steppers. What I haven't
> figured out yet is if a VCO will accept the -10V for input or if you would
> have to convert the signal some way.

I would recommend that anyone building such a system would
put quadrature encoders with a cycle count at least 1/2 the
number of steps/rev they will be using on their motors.  In
quadrature, that will give 2x the steps/rev of the motor.  Just
a little dead zone in the feedback (like 2 counts) would eliminate
hunting.

I'd be glad to work on this as soon as I get the analog output and
quadrature modules working, if there is sufficient interest.

Jon