Re: [CAD_CAM_EDM_DRO] access to Ametek motors- suitable for Servo application?

Kim Lux wrote:

>It should be noted that in a typical PID loop, such as what might be
>used in a Gecko driver, the voltage/current applied to the armature is
>proportional to the error, ie the difference between the count stored in
>the Gecko and the position sent by the encoder. Thus, NO motor is going
>to stay exactly on position when manually rotated when standing still.
>
>How much the motor will rotate depends upon the error count and the
>higher the encoder resolution, the more steadfast the motor holding will
>be in terms of allowable deviation.
>
>We've got several types of servo motors setup up with Gecko 320s, some
>of which are Amateks as being discussed. WE've also run various encoder
>resolutions, from 100 counts per rev to 1000 counts per rev. The motors
>with the low resolution encoders feel really loose at standstill. Put a
>high resolution encoder on the same motor and it feels tight. This
>happens with all motors including Amateks.
>
>

Thanks, that helps define the problem better. I corresponded with Mariss
about this, and he said the I term of the PID would drive it up to full
current
in a fraction of a second. Well, my experience did not show that happening.
Your comment about the resolution of the encoder seems to confirm that.
I wonder what the reason for this is, as I'm sure Mariss had this
working fine
at his lab. But, some combination of parameters seems to defeat the
integral
term of the PID. Does the G320/340 even sense the motor current, other than
for fault trip? I wonder if motor inductance or encoder jitter could
somehow
keep resetting the integral term? The motors in my local client's machine
have 250 cycle/rev encoders (I think), so they are in the low part of
that range.
But, 250 Cycle/rev = 1000 counts/rev, and that gives 2.8 counts/degree.
If the shaft can move 15 degrees, that is 42 counts! Maybe the
anti-diter circuit
defeats the integral term when at standstill?

>Turning up the gain until just below resonance helps too.
>
>

Yes, clearly, and we have it as high as we can. Also, the raw gain of the
PWM output stage is proportional the DC supply voltage. He was running
his unit on 36 V, which is a bit marginal.

>It makes a difference if you are pushing on the table or manually
>turning the motors. I can't measure any deflection of the table with
>the G320/Amatek setup that we use. And yet the motor will turn slightly
>if you put enough torque on it.
>
>

I'm talking about grabbing the motor pulleys. As my servo system was
designed
to deliver about 1000 Lbs linear force, pushing on the table will be
insignificant.
I also can move the motors by hand, but the resistance is quite strong,
even at
very small displacement. On my client's machine, there is almost no
resistance
to turning the motors about 10 to 15 degrees from the center of the
null. Then,
it begins to resist strongly. If you leave it alone, the motors will
twitch every
5 - 15 seconds, moving perhaps 5 degrees at a time, randomly, within
that window
of 'looseness'. This seemed totally wrong for what is supposed to be a
precision
positioning system.

Thanks for the comments! I'd still like to get back to my client's machine
and make it more precise.

Jon