> On Saturday, April 24, 1999 8:01 PM, Gerhard Fiedler wrote:
> I'm not sure whether PID control is actually the most adequate way to avoid
> ringing and overshoot (the mechanical, spring-like part of it you
> described). I would guess that limiting the change of the output, in some
> smart way, might do the same thing, without feedback.

  Well I agree in theory! To change the velocity of a fixed mass by a fixed amou
nt will take the same amount of energy every time
you do it. So if you know the galvo's velocity and mass you can calculate how mu
ch energy to apply to get it to move and stop where
you want.
  If you drive the galvo with a fixed current source then energy is just a funct
ion of time. If you know the acceleration profile of
the galvo then:

Velocity = time * Acceleration
Position = time * Velocity
OR
Position = time squared * Acceleration
If you know position then you do not need feedback

  The problem is knowing exactly the performance of your galvo. How fast does it
 accelerate.  There are many factors. Here are some:

  The acceleration of a galvo is non linear. It depends on the relationship of t
he rotating coil's magnetic field, to the fixed
magnet's field. This relation ship varies as the galvo turns and the fields chan
ge position.

  Galvos and mirrors being mechanical will vary slightly from unit to unit. On h
ome made and inexpensive galvos there can be a
relatively large difference between galvos

  Galvos and mirrors, especially home built ones, are not perfectly balanced. Gr
avity will aid acceleration in one direction and
oppose it in another, of course once you cross center the oppose and assist dire
ctions swap

  Because the Galvo are at right angles to each other in order to produce the co
rrect x-y movements one galvo will be effected more
than the other.

  What this all means is that to do it right you must store a lot of information
 about the galvo's performance to improve on the
basic PID algorithm. In experiments I performed several years ago I was able to
increase the performance of a test galvo by about
15% over the basic PID filter. I used a table lookup approach and had about 4k e
ntries in the table. The tables output went through
an algorithm that looked at the next two positions as well as the current positi
on. The problem was the system had to be tuned for
each galvo. Change the galvo with another unit of the same make and model and pe
rformance went to 90% of the PID value. That problem
could have been fixed but I stopped the research when I realized that the spring
 was the source of most of the problems. Removing
the spring gave a 200% increase in performance. However without the spring, feed
back became necessary to keep the galvo from
slamming into the stops whenever the smallest amount of current was applied.
>
> actually it is the differential part that is responsible for smooth stops,
> and the integral part often is responsible for oscillations :)  it's
> function is to compensate the position error inherent in a proportional (or
> a pd) regulator.

oops! You are correct. My apologies, I should have spent more time proofing befo
re I hit the send button. btw laser galvo's usually
do not use the Integral term because they hardly ever stop moving.

> >The first differential component allows the servo amp to see into the future.
> that's not quite true... control would be easier if we actually had a device t
hat could do this :)

I apologize to any list readers who thought I was trying to say that it was poss
ible to see into the future . . No body can do that
with the possible exception of Bill Gates. . I Only meant that the effect was li
ke being able to see into the future.