A somewhat related situation, is the opposite case; where the planes are
solid but the signal trace jogs back and forth, like a clock trace with
a tuned length. Of course the velocity is much slower than the
point-to-point distance and the dielectric constant would indicate. In
this case the quasi-TEM wave tends to follow the signal trace (even at
low frequencies??), so the electrical length is nearly the same as the
total trace physical length. But if you make the jogs smaller so that
coupling from one to the next increases, the apparent velocity might get
faster because of this coupling (from a previous SI-list topic, but
ignoring the effects of the corners themselves). And of course, in the
limit as the jog size approaches zero, the velocity becomes that of a
straight wire because the wave would become TEM over the entire length
instead of just each little section.
Is there any parallel that we can draw from this to the mesh plane
situation? Are the propagating waves within the dielectric at all
similar? It does seem like the quasi-TEM waves might follow some sort
of path between the signal trace and the angled return traces, which
would be at about a 22.5 degree angle from the signal trace. If this
holds even at low frequencies (as it seems to for the serpentine signal
trace), then maybe it explains something.
>The reasoning seems to suggest that whenever a return current is unable
>to follow the signal path closely, then the propagation delay will
>increase.
It's not the velocity that the closeness of the return current affects
(in a homogeneous medium). It's the impedance. In situations like the
power-plane split or poorly placed bypass cap, what you get is a large
discontinuity that affects the integrity of the signal.
For the mesh plane, perhaps you get lots of little discontinuities.
Regards,
Andy