I would like to add few comments about the guard trace.
* The effectiveness of the guard trace depends upon the
electrical length and the type of termination at the other ends
of the traces.
Electrical length of the guard traces should be "many times" less than
3.0e+08/(f*e_r). Here e_r is the dielectric constant and f is the
operating frequency {in case of digital applications generally
f=3*clock frequency is taken: ``third harmonic'}.
The effectiveness of guard trace depends upon the termination
at the other ends of the lines (impedance of drivers and receivers).
If the impedance of drivers and receivers is very low then
shorted guard trace would not be effective as the coupling
between two lines via current in guard trace (inductive coupling :
due to large current flow in the guard trace) will be large. But in the case
of large impedance of drivers and receivers, the current flowing
in primary lines is less, therefore the associated
coupling will be less. Similarly in the case
of low impedance of drivers and receivers, open circuit guard trace
will give less coupling (capacitive coupling). It is to be noted
that capacitive and inductive coupling both are present in the case
of distributed systems but due to small electrical length of the
guard traces one coupling dominates another depending upon the
termination at the end of lines and the guard trace (short/open).
For a good design it is better to calculate [L] and [C] matrix for the
3 three coupled lines and simulate with the spice this three coupled lines
system with the lumped elements (corresponding to the L & C ([L]*length of the
trace, [C]*length of the trace) matrices).
For the present digital and RF systems (f << 2.0GHz), guard traces (open/short:
depending upon the drivers and receivers impedances, having length
of about 1 or 1.5cm, e_r ~ 4 ) can significantly reduce coupling.
With Regards,
Alok Tripathi
Dept. of Elect. and Comp. Engn.
Oregon State University,
Corvallis, OR97331