Servo-PWM-DSP

I have been thinking about stirring up this issue for a while and then I
saw David Howland's nice post about an I-O board so I guess it is time
to do it.

A "traditional" servo system has a "controller" that outputs typically
+/-10 VDC to the analog servo Amplifier. The servo amp internally
monitors motor current (torque) and gets a velocity and acceleration
feedback from a tachometer attached to the motor. Internally in the
servo amp, the feedback loops are tweaked for stability and accuracy.
The actual position of the motor or table is fed back to the
"controller" from an encoder (shaft or linear).

When designing a hobby system, the "controller" is the easy part. A PC
with an analog output board and an input card to read the encoders will
basically meet our requirements. There are lots of surplus brush DC
motors around that can be attached to a ball screw or other loads.
Similarly, attaching a shaft encoder or a linear DRO is straightforward
and mostly a question of finding the right surplus one and how to
mechanically mount it.

I see very little discussion in the mail list about using tachs. They
are little harder to find but again but that is not a problem and the
mounting is a mechanical issue again. But why so little discussion?
Further, I have seen no comments at all about servo amplifier design and
tuning. Has everyone found a perfect sources and there are no
problems???? Or maybe everyone is using steppers??

After this long introduction, I finally have arrived at the meat of the
subject:

A proper servo amp is difficult to design, difficult to tune and
requires a tach for best performance.

With today's very fast processors, it is possible, and it is being done,
to measure the time between each encoder pulse, calculate the actual
speed and use that value in a DSP filter for the servo amp. That
removes the need for the tach. With DSP, the servo tuning can be done
in software instead of with screwdrivers.

Now for the final step: Why not go digital all the way? If we used a
switching amplifier, class-D or PWM, there would be no need for the
analog difficult portion of the servo amp, just a smoothing filter on
the out put to keep the RFI out of the motors. (I forgot about the
current sense, but that is not a big deal to add a little A-D to measure
it.)

A DSP-PWM servo amp basically would have a small digital control section
with a pair of heavy duty output switching devices. This might be
reasonable as a hobby project and it would be a lot easier than an
analog servo amp.

How does EMC fit into this. It is my impression that it will put out
the +/- voltage in servo mode and read the encoders through an I-O
card. Is it actually using the DSP calculations internally and if so,
what specs are required for the servo amps since if that case, the
internal amplifier feed back loop is not needed?

Comments?

Bertho Boman
Vinland Corporation