Re: [SI-LIST] : 20-H Rule for Power Planes

Larry Smith (ldsmith@lisboa.eng.sun.com)
Thu, 27 May 1999 16:05:12 -0700 (PDT)

John - You're right, this is a lot of fun.

I am a lot closer to the SI side of things than the EMI side of things,
but our disciplins are closely related. I agree that from an EMI
perspective, the voltage on the metal frame of a system WRT earth
ground must be very small or else we will have EMI radiation problems.
It must be at virtually "spice node 0" potential.

But spice node 0 is a node where you can dump an infinite amount of
current and get no measureable voltage. There is no node like that in
any of the systems I work on. :^) I wish there was!

A modern uP can draw a 20 amp transient in 10nSec from a 2 Volt power
distribution system. If there is just 1 mOhm in the "ground" circuit,
we get 20 mV ground bounce or 1% of the supply. One oz copper sheets
are about 0.5 mOhms per square, so it is easy to rack up 1 mOhm. If
there is just 1 nH of inductance in that circuit, we get 0.5 V of
noise, 25% of the power supply voltage!

The uP draws current from decoupling capacitors that are placed on very
distributed power planes. The power planes behave as if they are a
mesh of transmission lines with an impedance and delay. It may take
2nSec or more for for electromagnetic energy to get from one side of
the power planes to the other. In this environment, it is very
difficult to identify any node in the circuit that is worthy of being
called "spice node 0". If we did identify a spice node 0, none of the
circuits would pay any attention to it. All they care about is the
local Vdd and local Gnd at some position on the power planes at some
instant in time. So even if we do identify a spice node 0, voltages
WRT that node are not very meaningful to functioning circuits.