Finding return current paths via synchronized
measurements in a multiphase DCDC buck
converters
Peng Shao
#1
, Frank Chang
*2
, Chris Reade
*3
,
Ponniah Ilavarasan
*4
, David Pommerenke
#5
Department of Electrical and Computer Engineering, Missouri University of Science & Technology, Rolla, MO, USA 65409
1
psrg8@mst.edu
5
davidjp@mst.edu
NVIDIA Corporation, 15520 SW Jay St .Beaverton, OR 97006
2
frchang@nvidia.com
3
creade@nvidia.com
4
pilavarasan@nvidia.com
Abstract—Evaluation of a product in terms of radiated
emissions involves identifying the current return path. In a
complex system multiple sources can contribute to the current at
one place and frequency. Identifying the source of the current
can be achieved by correlating the current to different sources. A
multiphase buck converter uses multiple parallel buck
converters having different switching times to source a large
mount of current. In a multiphase buck converter currents in the
capacitors are caused by all phases of the converter, however, the
phases do not switch at the same time. Thus, synchronizing to a
specific phase allows evaluating how the current of this specific
phase spreads throughout the board. With the objective of
localizing current one can evaluate if the capacitor placement is
optimal and find improved layout and placement solution for a
multi-phase buck converter.
I. INTRODUCTION
Multi-phase buck DCDC converters (figure 1) are used to
provide very large currents at low voltages. Typically, they
provide the core voltage of processors, so values of 100A at
1.3V are not uncommon. In a multi-phase configuration
typically 4-6 converters step the voltage down from 12V to
1.3V. Each converter may run at e.g., 300 kHz, however their
switching is distributed over the 3 us cycle time. This is done
to reduce the size of the output capacitors. Some
randomization maybe introduced to reduce the spectral
density at the 300 kHz and its harmonics.
Figure 1. Typical Synchronous Buck Converter[1] [2] Schematic
Figure 1 shows the schematic of a typical synchronous buck
converter. The three MOSFETs (high-side and two low-sides)
switch alternatively. The switching of the MOSFETs is
controlled by a controller IC which generates a PWM signal
and drives the gates. The node where the three MOSFETs and
the output inductor share is called ‘phase node’. The voltage
waveform of the phase node is expected to be a rectangular
pulse train with a certain duty cycle, D. The duty cycle of the
phase voltage (fig. 2) waveform determines the output voltage
of the converter: V
out
= D * V
in
. The output L-C filter averages
the phase voltage waveform to a DC output, V
out
. [3]
53 978-1-4244-4267-6/09/$25.00 ©2009 IEEE