Re: Proper size of capacitors for transient suppression?

> of the following types (ground is by a seperate #6 AWG Cu. bus):
>
> DC voltmeter leads
> Connections to remote Enable relay coil
> +5, +12, +24 volts DC lines
> E-Stop contacts
> MPG encoder leads

Dc voltmeter use a 10mfd cap.
Connections to remote enable relay if it's just driving a relay coil
don't work about it.

All DC lines need 100mfd to 1000mfd caps a the end fartherist from
the source.

E-stop contacts. Hopefully this is back through an opto to give you
some noise immunity. Probably won't be a problem but a .1mfd cap
would not hurt.

MPG Encoder leads....now this one could be a problem. It's the wires
that should be shielded and isolated if possible (our good ole friend
the optoisolator. I would work to try and keep the grounds of the PC
and the input signals out of a loop condition.

Your grounding approach should work (big honkin' ground lead to a
common point ground). It is accepted practice to have signal grounds
and power grounds and to send voltage and signals down twisted pairs
or wires where one is a ground. Twisting wire tends to cancel out
noise since the wires act like a winding on a transformer wire with
bucking coils. The more twists per inch the better the noise
rejection.

>
> Except for the MPG leads (not yet installed), the longest distance

of

> any line is less than 10 feet (3 meters). I used #18 wire for
> everything to keep the resistance to a minimum.

Wire is funny stuff. At DC it looks like a resistor and resistance
is an issue. #18 for your power runs and power grounds are good.
For low voltage logic levels they create termination problems, and
the bigger wire has different characteristics, like higher
capacitance (yes, wires have capacitance, inductance, and
resistance.) By far the biggest problem with getting low level logic
signals down a pair of wires is the capacitance. Since larger wire
has more surface area (like a plate on a capacitor) bigger wires
running alongside other wires tend to have more crosstalk (I say tend
because the spacing of the wires has to do with it as well as the
frequency and rise times on the signals). In short You don't buy
yourself anything by running low current, low voltage signals in
bigger wire. Voltage loss is a factor of the raw DC resistance and
the amount of current involved. At a few milliamps it takes a lot of
wire to have enough resisitance to matter. Take #26 wire. It has
43.6 ohms of resistance per 1000 ft (.044 ohms/ft). At 20ma the DC
voltage drop would be .088 volts at 100ft. Think about your phone.
The actual signal on the wires is small and you can run that miles.
They use #20 or smaller for most runs. The run a "balanced pair"
meaning that the pairs are twisted and not ground referenced. Any
noise (and there can be volts of noise) is on both conductors and
cancels out. If either of the wires get ground referenced you will
get a loud "hum" on the line.

>
> The pendant also has the keyboard emulator embedded. I have heard

of

> people connecting each terminal through a capacitor to ground (56
> terminals? Ee Gads!)

Treat the keyboard signals like its an extended keyboard to the
computer. Shielded wires. If you have false triggering then
consider making sure the DC power source for the keyboard is well
filtered. The problem here is that the PC ground and keyboard ground
offer a ground loop opportunity. Since you already have a low
impedence ground via your star grounding scheme the ground from the
keyboard if connected to the PC ground might setup a ground loop and
be a noise conduit. Hopefully there is enough debounce in the
signals to reject noise. If noise becomes an issue you will probably
have to buffer the signals back to the computer. I don't think the
individual keys will need caps back into the chip as long as the chip
is close to the keys.

Glad to see you are close to switching things on. We all await your
results!

It sometimes counterproductive to try and determine where noise is
going to occur. Take care of the obvious and deal with the
circumstances as they arrise.

>