Re: [CAD_CAM_EDM_DRO] question about servo amplifiers

skykotech wrote:

>I have some brush servo amplifiers.
>
>A.M.C. 25A8K and Copley 422
>
>Both of these have switches for four modes:
>
>current mode
>voltage mode
>IR compensation mode
>tachometer mode
>
>Which is the best mode for use with a motor with encoder on a cnc
>mill running EMC with a vital systems PCI card that outputs +/-10V
>control signals and reads encoder information from the motors?
>
>

Do you have DC tachometers on the motors? If so, tachometer mode is
best. If not, then current mode is most likely the way to go. IR
compensation
mode is for speed control, and can cause dynamic instability in positioning
systems unless the tuning is done very precisely.

>I guess what I am asking is with a setup like mine, how would these
>drives in say current mode with my non-tachometer motors compare
>with the same drive in tachometer mode with a motor that has a
>tach? If EMC and the vital systems card is closing the loop around
>the motor and driver anyway, what does the tach feedback to the
>amplifier really buy you?
>

It allows the servo loop to be totally real time - ie. a continuous-time
system. The tach gives continuous velocity feedback, whereas the
encoder gives sampled feedback, and the sampling rate slows down as
the velocity decreases. If the encoder is the only form of feedback, then
the servo response has to be made "softer" or the loop will break down
into discrete jumps. This becomes inevitable at some low speed, no
matter what you do. With a tach, the computer is only providing the
positioning loop, and can have lower gain, while most of the loop gain
is in the tach loop, which is stable at any speed, down to zero, because
the tach is not sampled. You can also tune a velocity loop to be very
stable and quiet at standstill, while a torque loop with only encoder
feedback has to bounce back and forth between encoder counts.
(There are various schemes such as a deadband to eliminate the
servo hunting.)

> Is it that it allows a faster response
>time? If so, what type of difference are we talking about?
>
>I am trying to decide if it is worth my time to develop a digital
>tachometer that generates a voltage based on encoder readings.
>
>

It is definitely NOT worth your while. Even if you did such a thing, it
would provide no benefit, because an encoder is not a tach! it has no
feedback between quadrature counts, so it has to "guess" at the
instantaneous velocity. The computer would essentially be doing the
same thing anyway. And, doing a digital (or analog) encoder to
tach converter is not as straightforward as it might seem. The kicker
is the variable sampling rate provided by the encoder. The converter
would always be working on the interval between the last two encoder
counts, which is some variable amount of time behind the CURRENT
instantaneous velocity. So, if a sudden acceleration were to develop,
and the counts are coming in at exponentially decreasing intervals,
the estimate of the current velocity would always be low. You can't
fix this with a differential, as it will make it unstable for every isolated
encoder count that comes in, treating it as a rapid acceleration, when it
may be a VERY slow drift. This is the beauty of the DC tach, that
it reports TRUE instantaneous velocity in a continuous-time fashion.

The only way to beat this problem is with a very high resolution encoder,
which is not that hard to come by any more. If the encoder resolution
is on the order of 10 microinches or so, then the count rate is always
sufficiently high that a tach becomes superfluous.

Jon