Page 107 - Embedded Microprocessor Systems Real World Design

differential interfaces are good at noise rejection, they must be built from real parts,
and those parts usually have a maximum common-mode offset voltage that differ-
ential receivers can tolerate. An RS485 interface with correct d~fmential voltages
but with a 20V ground offset between the two systems is not going to work. And
remember that two grounds that are the same most of the time are not necessarily
the same all of the time. I have seen interface drivers and receivers actually
destroyed when an air conditioner switched on, yanking the ground on an embed-
ded system many volts away from the ground of the system it was communicat-
ing with.

RadlaW Susceptibility Interference from external RF sources such as cellular
telephones and walkie-talkies can affect an embedded system. The interference
may directly affect the processor circuitry or cause problems through secondary
effects.
To minimize the possibility of susceptibility, use small value pullups on unter-
minated lines. A lOOK pullup on a CMOS input makes the input impedance about
lOOK and potentially capable of picking up strong RF signals. Use 10K or add an
R/C terminator to the signal.
A sensor such as a strain gauge may pick up the RF and produce erroneous
outputs. Protect against this by performing sanity checks in the software. An input
that is picking up an external RF signal may go berserk and produce random inter-
rupts. Protect against these by checking for a continuous string of faster-than-
normal interrupts.
Since embedded systems usually control something in the real world, and those
things often involve motors, be sure you do not build a selfdestroying system. A
rotating, insulated pulley driving an insulated belt can become a fairly good static-
electricity generator. I saw a system once with an insulated plastic drum running
against a Mylar strip that could create a half-inch arc-not good for the micro-
processor that was controlling everything.
In a system with DC motors, it is a good idea to have a separate path for the
motor voltage to return to the power supply. This prevents the startup and braking
surge currents in the motor from yanking the ground on the electronics boards
around.
Finally, in designing for EMC, keep in mind what a certification engineer I know
used to say, “First it has to work.” In other words, it is not a good idea to make
changes or design compromises that let the system pass the EMC tests but degrade
performance in the real application. I have seen so much inductance added to a
clock line in an attempt to reduce emissions that the circuitry receiving the signal
could not operate reliably. I have seen video cables in a high-speed imaging system
with huge ferrite beads added. This fixed the emissions problem but killed the
video signal. First it has to work.

Hardware Design I 89