Re: [CAD_CAM_EDM_DRO] Safety -- servo vs. stepper

Carl Mikkelsen wrote:

>Building my own servo amps might be the answer. I don't see how I
>can get feedback from a Gecko drive to help me know that a problem
>is occurring. I'd need encoder feedback to the software, a model of
>what the drive signal should be, and possible also tachometer
>feedback to give me a redundant way to detect motion.
>
>

I have an interface that allows my Universal Stepper Controller to power
and read the encoders directly, and it also repeats the encoder signals to
the Gecko drive. So, it takes care of sending the encoder signals two
places.

>I have some nifty quadrature decoder, pwm generator, A-to-D
>converter, CPU chips available for the servo smarts, but I'm not
>really comfortable building the high current, high voltage H-
>bridge. What did you do for your home-built servo amps?
>
>

Those are a bit of a mess, I was trying to pack all the functionality of
an old-style analog velocity servo amp like a Servo Dynamics onto
a 4.5 x 6" circuit board. Amazingly, it works, and quite well, but
is VERY complicated.

I have just designed a new, much simpler servo amp that takes in a
PWM waveform, and a controller board that generates the PWM
waveforms.

>With a good E-stop circuit, is it customary to have a fault on one-
>axis stop motion on non-faulted axes? [sounds like from Leslie M.
>Watts' emails that ALL faults trigger a e-stop on all axes]
>
>

Yes, absolutely, and stop the spindle, as well. What purpose would be
served
by having the other axes continue blindly along their path when one axis
has stopped?

>It seems valuable to have several features in a servo drive:
>
>1) over current detection, where over current is a dynamically
>changed quantity based on the current acceleration and static load
>models.
>
>

That is overly complex. I know of no amplifier that does this.

>2) redundant velocity feedback, matched against the current expected
>velocity
>
>

I know of no amplifier that does this, but some machine tool controls
DO perform these types of checks. It is easier if you already have
DC tachometers needed for the velocity servo amps.

>3) a true closed-loop control, rather than a step-and-direction
>interface, so that faults which result in unexpected performance
>problems can be sensed by the control software
>
>

Yes, I agree!

>4) limit switches -- if only I could find a way to implement them.
>For a hexapod, these would need to sense all conditions what would
>occur as the result of a single or possibly dual axis failure. With
>the kinematics of my hexapod, this would most likely result in a
>joint angle error before it would cause an actuator length error.
>
>

I don't know how valuable these are. When everything is working normally,
software limits are better. When a G-code command is given (either through
MDI or in a program) that exceeds the software limits, the move is not
even started. In EMC, I think it will halt as soon as the look-ahead sees
the block that exceeds the limit. When things are NOT working right,
ie. a runaway, the limit switches would have to trip so far from the
actual limits
that it wastes the machine's work envelope.

>5) sensing in the servo drives to assure that the host is actively
>generating control inputs. This could be through a watchdog time or
>other means.
>
>

Yes, all my systems have watchdogs - but they are never used. EMC and
Linux are just really reliable. Still, I like to know they are there.

>The thoughtful responses have gotten me thinking more intensely
>about how to design for safety.
>
>Are there multi-axis commercial servo amplifiers that approach Gecko
>pricing?
>
>

HA HA HA! Not even close! The closest thing I know of are Copley Controls,
which run some $400 an axis in small quantity.

>Does anyone have a proven H-bridge circuit that takes direction and
>PWM input, plus an analog signal representing a dynamically
>adjustable current trip point, and delivers motor outputs and a
>fault signal?
>
>

Well, pretty close. My PWM-input servo amps have a user-settable
current limit, as well as a fixed fault trip limit. It can be run with
either
sign-magnitude PWM (as I do with my UPC board) or with 50% duty-cycle
PWM signals (% duty cycle = -velocity to + velocity). The 2nd mode
dissipates power in the motor and drive by delivering a triangle wave
curen to the motor all the time. That's why I prefer the 1st method
(sign - magnitude). These amps are good up to 160 V and 20 A, and
I'm selling them for $100 each axis. See my web page at
http://jelinux.pico-systems.com/pwmservo.html for more info on this.

Jon