top of page
Search

# Why EM Is Replacing SPICE For Simulation of Board-Level Power Delivery

Updated: Dec 27, 2021

How2Power.com - November 2021

by Heidi Barnes, Keysight Technologies, Santa Rosa, Calif. and Steve Sandler, Picotest, Phoenix, Ariz.

There is nothing magical about power delivery for high-speed digital loads. It follows the laws of physics and travels in waves according to Maxwell’s equations. Now, if you have ever tried to solve Maxwell’s equations by hand, you will completely understand why electronics engineers spend so much time with lumped-element circuit theory.

Lumped-element circuit theory makes use of the much simpler Ohm’s law and Kirchoff’s laws for nodal analysis and is the basis for SPICE-based simulators. The question that is left unanswered is “When do the assumptions that the circuit is the lumped equivalent of a schematic break down and drive the need to include a higher fidelity electromagnetic (EM) model of the physical interconnects between the components in a schematic?”

In the power electronics world of voltage regulators there is such an entrenched history of SPICE models that few engineers look further than running a freeware LTspice simulation with vendor-supplied models to predict behavior. This “lumped-SPICE” type simulation assumes perfect conductors connecting between the bill-of-material components in the schematic.

However, signal integrity engineers designing power delivery to high-speed digital loads are finding out the hard way that lumped-SPICE simulations are leaving out critical time delays and parasitic behaviors that only an EM simulated model of the printed circuit board (PCB) interconnect can get right. Getting it wrong can result in incorrect prediction of power rail resonances and lead to voltage regulator designs that are on the edge of instability. Ultimately, it can lead to a complete failure of the high-speed circuitry.

In this article, we will discuss and demonstrate the differences in simulation results obtained with lumped-SPICE modeling of circuits lacking PCB parasitics versus simulation results obtained with EM modeling that includes the board effects. We’ll illustrate these differences through simulations of voltage regulator output impedance and capacitor network impedance, noting the implications of the discrepancies.

Hopefully, these examples will drive home the need for EM modeling of both the voltage regulator and the power distribution networks (PDNs) in which they are used. However, in the last section we’ll explore how the lack of clarity surrounding vendor-supplied models of decoupling capacitors presents a barrier to accurate power delivery simulations with EM models.