Re: [CAD_CAM_EDM_DRO] Servo heat-sensors or paranoia? servo protection questions

R Rogers wrote:

>
>Jon Elson <elson@...> wrote:
>
> A fuse before the drive should blow before the motor fries.
>
>
>This is not likely to work, until the drive burns out! Due to the
>switching action, and the fact that a stalled
>motor develops no back EMF, the voltage across the stalled motor is near
>zero, therefore the power consumed
>is small (voltage x current). While the motor may be getting 20 amps
>(at a few volts), the drive will be drawing
>just a few amps at the DC supply voltage, so that power in is close to
>power out. So, the fuse will have little
>current through it, and will NOT blow. (For instance, I test my PWM
>servo amps at 20 A into a nearly dead short,
>an ammeter shunt. With 30 V in and 20 A out, the amp only draws about 5
>A from the power supply.)
>
>Jon
>
>/////Hi Jon, I see, the current between the drive and servo is isolated from the DC supply side of the circuit. So, a fuse there is moot? Then what type of protection is possible in the event of a stalled servo using a Gecko? Would it be effective to fuse between the drive and the servo instead?
>

A fuse will prevent fires and circuit traces being burned off the board
if a transistor
fails in the drive. But, a fuse is a terrible way to protect against
overloads in a servo
application. The higher-end drives have what they call RMS protection.
They integrate
I^2 over time, as I^2 will be proportional to motor heating. So, these
drives might
allow 20 amps to flow for 2 second, 10 amps to flow for 8 seconds, but
only 5 A
to flow continuously.

A properly selected time-delay fuse would protect the motor fine. The
problem is
that the design of the Gecko drive cannot handle the arcing that would
happen
at the moment the fuse blew. It would have a high probability of
popping transistors
in the drive. That probably applies to Rutex and similar drives, as well.

>
>Also, what is actually happening with an encoder when a servo is at idle? There is 5v, ground, A channel and B channel. Is there a constant A or B signal back to the drive or is it simply a fast pulse and the drive waits for another to move, correct, etc.? Seems there could be some type of device that monitors and in the event of lost encoder activity which would unlatch the E-stop circuit.
>
>

At idle, most drives tend to rattle back and forth across some deadband
or region where the DC gain of
the servo loop is unable to cause motor movement, due to friction. This
is usually a very small distance,
on the order of +/- 2 encoder counts or so. A circuit to detect a
failed encoder without causing false
trips could be quite hard to do. Comparing a differential encoder's
outputs to make sure the two
complementary signals are actually at opposite logic states might be a
first step to detect broken
cables, popped fuses and actual failure of the encoder circuitry. But,
that would totally fail to detect
an axis jammed against the travel limits.

One possible scheme would be to put a very small value current sensing
resistor in series with the
motor wires. Because this resistor is rapidly switched between ground
and the DC power supply
voltage, circuits to monitor the current are a bit tricky. But,
monitoring the TEMPERATURE of
the resistor with a PTC thermistor might be quite simple! These PTC
resistors have a very abrupt
rise in resistance when they are heated above about 50 C. You could
string them in series and
glue them to some .1 Ohm (for low-current motors) down to .01 Ohm power
resistors for big motors.

This would basically implement a true RMS current limiting scheme. The
string of PTC thermistors
could be wired in series with the e_stop switches, if their cold
resistance was low enough. This is
so simple, I'm probably going to put it on some of my machines! I will
see how much it would cost
to add it to my PWM servo drives for the next revision.

Jon