> It is VERY bad to ground the person, it involves risk of electric shock
> from anything not connected through a fault current breaker, including the
> project's power supply. As far as I know, the only place where you can
> ground the operator safely is in a production environment where the units
> are not powered....

The common thing is a wrist strap connected to ground through a 1 meg resis-
tor.  This will let through enough current to prevent electrostatic charge
from building up, but will limit current to well under a miliamp in case of
accidental line contact (even with 250VAC, the RMS current is only 1/4mA).

> There are chips that protect IO pins. Check MAXIM MAX336 and co.

Such devices can be useful in some cases; in many cases they can be avoided
with careful design.  I personally like optoes for protecting I/O, but they
aren't free; pretty oops-proof, though.

> One thing ppl. keep forgetting when using PICs with external connections,
> is the fact that the same bulk and protection diodes that allow a PIC to
> be powered by its IO pins (featured elsewhere in this list), still work
> when there is power applied to the Vcc and GND pins as it should. Since
> most regulators cannot sink current, this can easily lead to a PIC being
> powered by its (incorrectly designed) load, which can bring the VCC RAIL
> TO FAR MORE THAN +5v.

This can be a problem; even in cases where the PIC doesn't mind the extra
voltage, other devices may squawk and cause bad things to happen.  Also, an
overly-high VDD is one of the few ways I've seen to destroy PICs while run-
ning close to their specification (I accidentally ran a PIC 16F84 off 7.5
volts; oddly, it now works at higher voltages but won't run or program at
5V any more).

>                        The cure is a zener connected between Vcc and GND
> after the regulator, directly at the PIC power pins (Z5V6 or such), AND
> the usual limiting diodes/networks at every concerned IO pin... else the
> limiting diodes only supply power to the PIC...

Depending upon the application and the number of I/O pins, I sorta like using
a 5.1V zener on the I/O pins.  That one component provides both ground and
VDD clamping while avoiding the overvoltage-VDD problem.  Otherwise, I just
use a series resistor on inputs and make sure VDD goesn't get pulled too high
(ex. power enough stuff from VDD that a minimal current draw is assured; on
at least one occasion I used a dumping resistor from VDD to ground).

By the way, does anyone have any idea how much current it takes to damage a
PIC with excess VDD?  For example, if a PIC were connected to a +24 supply
through a 1meg resistor, would it fry or would it safely clamp the voltage
to something reasonable?

> I've seen many such designs in mags, and it's scary. You have this PIC
> running at 5 V from a 78L05, and an input or output going through a
> resistor to +12V or whatever. The resistor is dimensioned to limit the
> current into the PIC input pin. However, if the PIC sleeps then the design
> current flowing through the resistor raises the VCC to almost the level of
> the higher voltage, due to the low current drain of the part.... the
> regulator opposes absolutely no resistance to this.

Some regulators do offer some sinking capability; others don't.  Any design
which relies upon that behavior, however, may fail when a "drop-in replace-
ment" regulator is substituted.