Re: +3.3,5-board stackup problem / return currents...

Larry Smith (ldsmith@lisboa)
Mon, 7 Jul 1997 15:53:37 -0700

> This point brings to (my) mind related questions:
> If a signal crosses supply domain, is there any possiblity
> of return currents trying to flow on power planes?
> I would think that 5v > 3.3v and the complement would have
> an overwhelming proportion of return current forced to the
> ground layer alone, as long as these signals are routed such
> that a reference ground layer is available.
> Can the power-vs-ground as return mechanism be predicted,
> if so, is there any simplistic rule of thumb (for physical
> design folks with many thumbs, such as myself)?
> Many thanks in advance,
> --
> Jeff Seeger Applied CAD Knowledge Inc
> Chief Technical Officer Tyngsboro, MA 01879
> jseeger "at" appliedcad "dot" com 508 649 9800

Jeff - return current always travels on the nearest reference plane.
That could be ground, 3.3 volts, 5.0 volts or even a floating piece of
metal if that is nearby. If there are two power planes nearby (ie one
over and another under the trace) current will travel on both planes
with the majority of current traveling on the closest plane.

This is true for most of the length of a transmission line (two
dimensional analysis applies). There is a 3 dimensional discontinuity
at the end of the trace (via, package pin, connector, etc.). At that
point, return current must jump from the old reference plane to some
other reference conductor. This often happens by displacement current
in a decoupling capacitor.

One possibility is the natural decoupling capacitance between Vdd and
Gnd planes. If the printed circuit board stackup has sig/Vdd/Gnd/sig
and signal current is forced to go through a via from one side of the
pcb to the other, return current will flow on Vdd for one portion of the
net and on Gnd for the another portion of the net. Charge will build up
on Vdd and Gnd near the via and cause the power planes to bounce a
little, as return current becomes displacement current in the
dielectric of the Vdd/Gnd plane capacitor.

Larry D Smith
Sun Microsystems