Re: Analog Spindle

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

Dave Engvall wrote:

> Hi J, D, R:
>
> Clearly Ray's searching capabilities exceed mine...that is good because it
> demonstrates there is some code to start from.
> My simple brain envisions another axis defined as S that has PID control and
> expects to use the dac channel one past the defined axes. E.g if one defines
> X, Y and Z in the normal manner then going X Y Z S would get spindle
> control. The scale would be number of counts from the encoder for one
> rev...and some sort of timing would be involved as feedback to the PID to
> determine speed and deviation from the target speed rather than position as
> in the other axes.

Well, for a first cut, I think just open loop control, where you have two
choices:

1.  direction and spindle start/stop is handled by a forward and a reverse
solid
            state relay, and
     a 0-10 V signal controls the motor from 0-full speed.

2.  There is only a spindle run relay, and direction is controlled by the
polarity of
        the analog speed signal, so that you have -10 V .. 0 V .. +10 V
represent
        reverse full speed .. stop .. forward full speed.

I think this would get a lot of people what they want for milling machine type
work.  It would also be pretty good for most lathe work, with the extra need of

having a spindle encoder for threading.  but, the spindle does NOT need to be
a true servo.  As long as it can hold speed within a few % of commanded, that
should be fine.  I really think that trying to pound the work environment into
the EMC mold by making a fully servo-controlled machine, where spindle
position is tightly controlled, too, may cause all sorts of problems.

It seems that all the machines I've seen allow the spindle to run open-loop,
and then slave the axis motions to the spindle rotation.  One of the problems
in not doing this is how to deal with the spindle starting over at zero every
revolution.  Except when threading or performing some similar operation that
has to be sync'ed to the spindle, the spindle encoder is ignored.  When
beginning
a threading operation, the other axes attempt to move in such a way that they
pass
through some control point at the exact instant that the spindle passes the
index
position.  This is really not very hard to do, but I don't know how much
hacking
would be needed inside EMC to handle this.  There are indications that the
earlier
Sun version of EMC that was demoed at GM Powertrain had a spindle encoder and
did hard tapping.  I don't know how much of that code has been lost in the
porting
to Linux.

Jon