Re: Low cost Servo Controller

OK, you guys have managed to drag me out of the woodwork on this one.
Due to numerous Real Life issues, I've been very swamped, but I really
want to chime in here, because this is the group project that I've been
waiting for.

I'm just going to hit the key points that I can help with or that I am
curious about.

Most of you have seen my work on the PC-based DRO, and those of you who
haven't can see it at:

http://www.mcs.net/~kulaga/dro.html

This is the same card that Dan Mauch is selling.

I think that it could serve as a good starting point for a servo
controller card. I'd knock it back down to 2 axis, add 2 DAC outputs,
a few digital I/O channels for limits, home, etc, and generally add alot
more in the way of buffering and protection. I know that we sort of
discussed this here on the list some time ago, but I haven't seen very
great enthusiasm, and being pressed for time myself... you know the
story. Anyway, there was some interest from Fred regarding setting up
EMC to work with my card. Each card would support 2 axis, and users
requiring more could just add cards in a modular fashion. It would be
a little demanding on ISA slots for more complex machines, but any PC
controlling something that complex would be dedicated and probably not
running sound cards and such.

I was also thinking that we could dovetail this project with all the
talk about handwheels as controllers. Add one of my 4-axis cards for
reading the handwheels, strip EMC down to just the PID loop stuff, add a
little code to read the handwheels and feed commands to the servos, and
you'd have a nice manual interface for your CNC machine. I don't know
nearly enough about Linux, programming, or the Real Time patch to do any
of this myself, but I'm sure someone could figure it out. It wouldn't
be as clean a solution as running it all inside EMC, but it would be
plenty useful.

The debate about linear versus rotary encoders is one that I thought I
could add something of interest on. Now, any backlash in the system is
going to be annoying, and sometimes more than annoying. However, if you
use linear encoders for feedback, that backlash will kill servos as a
usable solution. For a PID loop to work right, the feedback to the loop
*must* be exactly what the motor is doing. I have, however, seen this
sort of thing done, and it resulted in a stable system. The trick was
to use both rotary and linear encoders. The rotary, mounted directly
to the motor shaft, would provide feedback for only the Derivative
portion of the PID loop, with Integral and Proportional coming from the
linear. It is not a perfect solution, but it does work. Still,
backlash should be kept very very small, or the system is impossible to
tune. Hands down, the best solution is zero-backlash ballscrews. It
may be more expensive to retrofit, but the headaches and added cost of
reading multiple encoders per axis is *far* more expensive and
difficult, IMHO.

When it comes to the amps, I'm not much of a help, but can anyone say
whether it is possible to convert an analog signal to a PWM signal? It
would be impossible for EMC to handle a software generated PWM signal,
and the only way I know how to do it is by using a dedicated motion
controller chip which is set up to output PWM. It's gotta be possible
to convert one signal to the other, right? Especially given that we
can decide here and now what level the signals will be (industry
standards of +/-10VDC for analog, and 0-5VDC for PWM).

OK, my brain is feeling pretty tapped right now, so I'll cut this off.
I'll be monitoring the list daily, and I'll try to post if I've got any
other thoughts. I know there was something else I wanted to say, but I
can't seem to remember what it was.

-Tom Kulaga