9
"One advance in technology could have a big impact
on PDN design: the incorporation of nonlinear features
into DC-to-DC converter control loops."
Ethan Koether, Sr Hardware
Engineer, Oracle Corporation
Ethan Koether is a senior hardware
engineer at Oracle Corporation, where he
focuses on power integrity for cloud-based
server boards. He holds a master's degree
in electrical engineering and computer
science from the Massachusetts Institute
of Technology. Ethan's work and interests
include power integrity research and
methodologies and power distribution
network design, simulation, measurement,
and analysis.
The trend in power distribution network (PDN) design has been to support tighter
voltage tolerances and higher currents. Market demands are forcing design engineers
to create designs that push tighter specifications. Meeting these specifications means
creating PDNs that service greater device density with higher power efficiencies and,
in many cases, at lower cost. All these requirements make the whole design process
more difficult.
The standard approach in PDN design is to use a target impedance methodology
in which we design to a target impedance calculated by the worst-case voltage
fluctuation we can tolerate divided by the worst-case current excitation we can
expect. The goal is to have a PDN that operates below the target impedance over the
bandwidth of the load the converter is supporting. As current demands rise and voltage
fluctuation tolerances fall, we must lower impedance further with higher-density
capacitance. This shift requires the use of more expensive parts, such as capacitors
capable of surviving the higher heat that will be created by the higher currents in a
denser design. It's necessary to balance all these factors when designing a PDN.
Nonlinear Components Could Make PDNs Smaller
and Power Supplies More Efficient
