Servo systems (was: re:cleaning, lovejoy, black box tach....)

Jon, Mariss, all,

I've just tried to retitle this thread, hope it "takes". It was becoming
not so much a thread, as a tangled wad of string. I would like this
discussion to continue, though. I wish to extend my limited knowledge of
servos.

I understand the basics of the subject. I've written software PID
control loops, though they have all been low bandwidth (i.e. temp
controllers). The questions I want to pose are not of the "what parts
should I buy?" kind, but rather the "how does this really work?" kind.

This velocity feedback versus position feedback issue intrigues me.
Here's what I think I understand: The output stage of the servo amp
controls the armature current and, therefore, motor torque. The G-code
specifies ending position and feed rate. The fact that we are trying to
do coordinated multi-axis motion implies the need for tight control of
velocity and acceleration.

So lets see, we have, at least, position, velocity, acceleration,
torque, current, and time to deal with. There are, of course, well known
mathematical relationships between these and we have a computer at our
disposal to do the math.

It seems obvious to me that independently measuring position, velocity,
acceleration, and time, is, at least theoretically, redundant. Since we
are using digital electronics we "know what time it is" for free. So,
then, it would seem that all we really need is one of: a tach, a
position encoder, or an accelerometer. (No, I'm not seriously suggesting
we stick an accelerometer on the table).

The setup that Jon and others are using clearly works, but, to me at
least, it seems redundant and confusing. If I've got this right, there
are two servo loops controlling each motor. One consists of the amp,
motor, and tach. The other is EMC, amp, motor, and position encoder.
This "extra" complexity makes it harder to think about, and I wonder if
it doesn't make the system harder to tune.

At very low motor speeds I suspect that neither a tach nor an encoder
give us much useful information. My guess is that cutting nice smooth
interpolated circles depends more on getting the servo loop damping
tuned just right than anything else.

I look forward to you explaining just how, exactly, I'm missing the
whole point. (Oh, I just love learning new stuff.)

Jeff

On Wed, 29 Nov 2000 16:24:32 -0000, Mariss wrote:

>--- In CAD_CAM_EDM_DRO@egroups.com, Jon Elson <jmelson@a...> wrote:
>>
>> I had trouble following this thread, I hope this is the original
>message
>> that spurred the question at top.
>>
>> My feeling, not having tried the Gecko servo system, is that there
>> are two differences in the system. One, is that it is not a closed
>loop
>> system, in that the CNC control has no direct feedback from the
>encoders.
>> The Gecko Servo unit DOES have feedback to IT, but it can't tell the
>> CNC control where it is, all it can do is try to keep the machine
>where
>> it has been told to put it. If there is a fault (crash or
>overload) or an
>> emergency stop, the machine position may no longer correspond to
>> where the CNC control thinks the machine is. A servo system with
>> position feedback to the CNC control will not lose position under
>these
>> conditions.
>>
>> The second difference is under extremely slow movement, which ocurrs
>> any time you interpolate a circle, or when you are following an
>angle that
>> is almost parallel to one axis. The stepper-> servo system, or ANY
>> system without a DC tach, cannot make smooth movements that leave
>> it between encoder pulses for a significant time, because that is
>the only
>> form of feedback, both for position and velocity. A DC tach allows
>> smooth motion even when it is several seconds between encoder
>counts!
>> If you have a very fine (high resolution) encoder, this may be no
>> concern. If the encoders are of relatively low resolution, it
>could cause
>> wavy lines in machined surfaces.
>>
>> Jon
>
>Hi,
>
>Jon, sorry, I jumped into this thread at about the 40th "Re:" level
>with my 2 cents worth. Point by point:
>
>1) The encoder is free to be used by the PC as well if you wish for
>direct feedback.
>
>2) The drive has a "fault" output readable by the PC in the event of
>a crash.
>
>3) Smooth movement between encoder edges. In either case the motor is
>moving between encoder edges, smooth or not.
>
>4) Seconds between encoder edges. Let's do the math. Assume 1 second
>between encoder edges, a modest encoder resolution of 500 lines and
>an agressive tachometer Kv of 100V/1,000 RPM.
>
>It would take 2,000 seconds per revolution, for a motor speed of
>0.0000083 RPM. The tach feedback voltage would be 0.00000083V, or
>less than 1 microvolt! My point here is taking brush noise and other
>factors into account, a really good tach has a signal to noise ratio
>of no better than 25 db.
>
>Mariss
>