Do tell.... Earth Management on big systems is a
hard lesson to learn.
[/QUOTE]
Now thats something I'm interested in hearing more about...
You've already brushed on the subject. It seems
to be a question of understanding that all devices
generate earth currents, the obvious and the not-so
obvious, (eg, power transistors on a heatsink, or
a switcher, or the field coils of a motor that you
mentioned). It is important to plot a path to ensure
that all such earth currents are explicitly returned
to where they came from, not allowed to find their
way home via some other device. That is an obvious
thing to say, but sometimes not so easy to see and
implement in practice.
Bitter experience results in automatic design habits
that help to avoid earthing troubles in the completed
system. Stupid little things, such as.......
Make twisted pairs, (triples, quads), your baseline
wiring habit. This forces you to consider and make
the explicit return between source and load, and
twisted pairs reduces differential pickup or radiation.
Keep the twisted pairs going, even through such things
as fuses or CT-assemblies.
Screen each functional set if possible, with the
screen normally Earthed at one end only.
With heavyweight source-loads, Earth at the source,
and have a low impedance Earth-Earth connection between
the source and the load.... already mentioned is the
circumstance where this may be via the screen, although
I would put in a separate Earth wire.
Low level devices which have an 'earthy' 0v and which are
to be connected to an earthed device are always a source
of ambush, because of the possibility of A.N Other's earth
currents coming back through them. This would be things
like scopes, counter-timers, comms from PC's, etc.
Isolation would be nice, but often not practical. The
secondary solution is low-value resistors in the 0v lines,
use differential signals if possible and/or try to have some
sort of CMRR capability.
eg. Use RS485 not RS232 if possible, and try to check the
CMV at the receiver end.
eg. DC to moderate freqency signals can be transmitted as
currents, with a burden resistor at the receiving end.
The c-c source does the CMRR and the transmitted signal
is converted to a voltage at the receiver, with respect
to the receiver's 0v. This can be applied to dc control
signals, RS232 comms (if you are stuck with it) and even
scope or counter-timer pickups.
Note: Run RS232 comms as two single-screened cables.
This avoids Txd/Rxd crosstalk over long distances.
Simple tricks really.................