The risk of causing signal integrity, power integrity, and EMI problems with a split in the ground plane strongly outweighs the potential benefit; there is only one case when there might be a benefit for a split ground plane. It is explained here. Once connectivity is established, the only thing an interconnect is going to do is add noise. When an engineer designs interconnects, they must design them to reduce the noise they might generate. When making design choices, we are always asking the question, what is the noise problem we are trying to fix, and how do we engineer the interconnects to reduce this source of noise? If you are considering using a split ground plane, one must first answer the question, what is the problem a split ground plane is trying to fix? And what are the potential other problems that might be created, the law of unintended consequences?

**Why Continuous Return Path Planes**

**Where Return Currents Flow**

In this equation, Z is the loop impedance of the current loop path, R is the series resistance of the loop, and L is the loop inductance of the path.

Imagine the signal-return path currents as composed of continuous current filaments taking any path they can down the interconnect. The filaments that have the most current are those with the lowest loop impedance. The more current flowing down one of these filaments, the higher the voltage drop across this series impedance. This pushes more current into adjacent higher impedance filaments until the cross section of current distribution, balanced by the impedance of each filament and the amount of current in each filament, create an equipotential across the direction of propagation.

There will always be a frequency, above which the ωL term dominates, and the current paths are driven by the path of lowest loop inductance. This is the region we refer to as the skin depth region. The current redistribution for lowest loop inductance is what drives the skin depth effect.

The lowest impedance path is when currents within the same conductor are farthest apart, to reduce the partial self-inductances, but are closest together between the signal and return paths, to increase the partial mutual inductances. This is illustrated in ** Figure 1**, showing the current distributions in a simple microstrip at 1, 10, and 100 MHz; simulated with Ansys Q2D.

As the signal conductor meanders over the surface of the ground plane, the return currents will follow right along under the signal path. *Figure 2** *shows an example of the return current distribution in a plane when the signal conductor changes direction, simulated for 1 MHz frequency components.

**Return Current at Low Frequency**

**.**

*Figure 3*

*Figure 4**shows the measured current amplitude in the signal-return loop at the far end, flat with frequency, and the measured current amplitude in the shunt, which drops off with a 1-pole response, above about 10 kHz.*

**.**

*Figure 5***Inductively Coupled Noise**

*shows the measurement set up of the two configurations and the measured inductively coupled crosstalk on the two victim traces.*

**Figure 6**