For the uninitiated, like me, these two points
cannot be be reinforced enough!
It's simply amazing what problems are solved
(due to unclean power) by using a 0.1mF cap
across Vcc and ground near the chip.
I amazed myself anyway... The noise from the
chip itself is enough to call for this filter
cap, forget about unclean supply! Your own
chip will defy you -- suck the noise out for
only 0.1mF, what a deal!
And if you haven't heard it already - put a
pullup resistor on RESET if you don't have
any other electronics controlling the reset pin.
Better yet - Atmel says it's OK to tie reset
DIRECTLY to Vcc, can't hurt.
(I found that probing the reset pin with o'scope
was enough to cause reset when using a 10K pullup,
but no way does it reset when tied straight to Vcc.)
Without *any* tie up, using the I/O line NEXT TO
the reset pin can be enough to trigger a reset when
transitioning that I/O from hi to lo. It's true!
Just my two bits (semi-nibble?) worth.