I guess if there were any easy standard grounding rules, a lot of us on
this net would be looking for work. A few general observations.
I have never had problems with a design where I was allowed to implement
a grounding plan/design before the mechanical and/or PCB design was
started. Designs that I have had the most trouble with (and therefore
make me the most money because like the oil filter commercial, "You can
pay me now or pay me later") were driven only by the designer (drafter
type), followed by the mechanical engineer, and PCB designer. This is
not a shot at these people who are by in large talented people in their
own fields, it's simply a statement that grounding is a difficult
endeavor that most of these people have never been trained to implement.
However, I will take my shots at the throw the design over the cube wall
circuit design engineers. At least the circuit designer should make
sure the grounding is adequate for the circuit to function. He just
might be able to help out in the other aspects of ground too.
Besides the C I R C U I T, here are the other factors I consider in
ground design.
1) Radiated EMI
2) EMI susceptibility
3) System noise
4) ESD
5) Safety
While there are some synergistic answers such as radial grounding, they
don't work for everything and should never be blindly thrown at a
problem.
For instance radial grounding is an absolute disaster in RF design
because ground impedance inverts every quarter wave. This means at 300
MHz a ground point 25 centimeters away is like no ground at all.
Good ground design goes way beyond the simple ground and requires
intimate working knowledge of every circuit and the fields generated by
those circuits. In other words... no simple answers.
Fred Townsend
DC to Light Consulting Services
Silicon Valley
Mark McKee wrote:
>
> Doug,
>
> I have a question regarding your comment:
>
> >Since all grounds are (usually) at the same nominal potential (zero volts)
> >they are usually tied together. But if tied at more than one point, there is
> >a possibility for a (current)(ground)(noise) LOOP to exist, causing (you
> >guessed it) noise. So they are tied together at ONE point --- to prevent loops.
>
> I have heard in the industry of people who believe very strongly that tieing
> the grounds together to prevent loops means tie ALL ground together at a single
> point and furthermore, make that point back at the power supply. This includes
> the chassis ground as well. However, if you have parts/traces radiating energy
> onto (for instance) a chassis ground, don't you want to minimize the return
> path for those currents to minimize EMI.
>
> In other words, tieing grounds at a single point, but not ALL at the same
> point, and tieing them together as soon as possible. (This is actually the
> question - when is comes to implementation, we find we have several different
> grounds, what is the theory and practive behind tieing them together?)
>
> Any other si-list who cares to add their theories or experience in this area
> please do so.
>
> Mark