Matrix Integrated Magnetics (MIM) For Low Voltage, Interleaved DC-DC Converters
Matrix Integrated Magnetics (MIM) For Low Voltage, Interleaved DC-DC Converters
Matrix Integrated Magnetics (MIM) For Low Voltage, Interleaved DC-DC Converters
Abstract- In this paper, we present the application of the Matrix The low power supply voltages at very large currents pose
Integrated Magnetics (MIM) core to an interleaved current several design challenges for power supplies with efficiency
doubler rectifier (ICDR) controlled according to the symmetric and power density requirements of the order of 95% and 150-
modulation scheme. Detailed modeling and design equations that 200W/in3 respectively. Tight regulation specifications at these
describe the effect, due to magnetic coupling, of integrating the
low voltages, demand very low ripple magnitudes in output
magnetics of two CDRs on a single core structure on the currents
and fluxes are presented. Comparative analyses between an capacitor voltage and inductor current. This forces designers
ICDR using a MIM core and two EI-cores and a CDR using a to tackle the classic power electronics problem of balancing
single EI-core for a 48V-3.3V/100W DC-DC converter design increase in switching frequency to achieve low ripple, high
example show that the MIM core can be effectively used for the power density due to smaller passives, and improved transient
ICDR providing substantial benefits over a dual EI-18 core response with the detrimental effect of decreasing efficiency
solution or a single EI-22 core. Preliminary experimental results due to switching losses.
from a 100W, 48V-3.3V ICDR laboratory prototype
demonstrating the validity of the MIM core structure for this
Inductor current interleaving is a well-known technique to
application are presented.
reduce output voltage ripple without increasing switching
frequency and associated losses and without sacrificing
I. INTRODUCTION
transient performance [2,3]. The current doubler rectifier
Low voltage, high current DC-DC converters to be used as (CDR) [4,5,6] enables the interleaving of two currents while
point of load regulators for microprocessors and digital loads integrating the operation of a transformer and two inductors in
have attracted significant interest over the past decade. The a single EI-core. New techniques to achieve integrated
demand for increasing speed and integration density of magnetics for the CDR with improved performance have also
microelectronic circuits has led to continually decreasing been proposed [7,8]. The Matrix Integrated Magnetics (MIM)
supply voltages for microprocessors and digital devices core proposed in [9] provides a cellular, low profile structure
presenting more challenges for the design of high power amenable to the development of interleaved, multiphase DC-
density, high efficiency power converters. For example, the DC converters with or without transformer isolation.
microprocessor supply voltages in the range 1.5-3.3V with 40-
50A current ratings are commonplace. According to the In this paper, we present the application of the MIM core to
National Technology Roadmap for semiconductors, the supply an interleaved half-bridge, isolated current doubler rectifier
voltage for microprocessors would drop to 0.9-1.5V in the next (ICDR) controlled according to the symmetric modulation
three years with the power levels increasing well over the scheme. The effect of integrating the magnetics of two CDRs
100W level [1]. The reducing trend of power supply voltages on a single MIM core on the currents is described in detail.
and increasing supply currents is illustrated in the long-term The performance benefits enabled by the shorter air gap,
projection for microprocessor type applications in Fig. 1. cellular structure and lower profile through the use of the MIM
core are presented. Comparative analyses between an ICDR
1.2 800 using a MIM core and two EI-cores and a CDR using a single
Supply Voltage, Vs (V)
1.0
Vs Is 600 design. These analyses show that a MIM core can be
0.8
500 effectively used to implement integrated magnetics of the
0.6 400 ICDR providing substantial benefits over a dual EI-18 core
300 solution or a single EI-22 core for the chosen design example.
0.4
200 Preliminary experimental results from a 100W, 48V-3.3V
0.2
100 ICDR laboratory prototype demonstrating the validity of the
0.0 0 MIM core structure for this application are presented.
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
104
Applying volt-second balance to any of the outer leg fluxes io = is11 + is12 + is21 + is22 (7)
over a switching period yields the DC transfer ratio of the Np
ICDR given by = (i11 + i12 + i21 + i22 )
Ns
Vo N
=D s
(3)
=
(RLM + 4 RCM ) (φ + φ12 + φ 21 + φ22 )
11
Vdc Np Ns
III. MODELING AND ANALYSIS Eq. 7 shows that the load current is equal to a scaled version
of the sum of the interleaving fluxes. Hence, the integrated
The determination of magnetizing currents and hence, the
magnetics results in the load current to be interleaved although
primary and secondary currents in the two CDRs from the the magnetizing currents are different from their well known
outer leg fluxes is presented in this section. The MMF diagram triangular waveshapes expected in the discrete implementation.
of the MIM core is shown in Fig. 7.
This effect is true even for integrated magnetics for a single
CDR implemented on an EI-core. The average values Φ11, Φ12,
Φ21 and Φ22 of the outer leg fluxes are equal and can be
F21
_
F11
_ _
F12 F22
_
+ + + + determined from Eq. 7 to be given as
FW
FW
RCM and RLM in Fig. 7 are the center leg and outer leg
FW
FW
reluctances of the MIM core given by i11
1 lGM (4) I11,av
RCM ≈
µ o ACM i12
1 l LM
R LM = I12,av,
µ r µ o A LM
i21
where,
lLM – magnetic path length of outer leg (5)
ALM – Area of cross section of the outer leg I21, av
ACM – Area of cross section of the center leg
lGM – height of air gap i22
µo, µr – permeability of air and core material I22, av
respectively
The MMF sources in Fig. 7 are given by is11
F11 i11 φ11 (6)
0
F12 i
=N =R
12 φ12
p MIM
F21 i21 φ21 is12
F22 i22 φ22
RLM + RCM RCM RCM RCM 0
R R + R R R
RMIM = CM LM CM CM CM
is21
RCM RCM RLM + RCM RCM
RCM RCM RCM RLM + RCM
0
From Eq.6, it is clear that the MIM core results in the is22
interleaving of the magnetizing currents, weighted by its outer
leg and center leg reluctance. Representative waveforms of the 0
magnetizing currents and secondary currents to illustrate this Fig. 8. Magnetizing and secondary currents on ICDR with the MIM core
effect of the MIM core are shown in Fig. 8.
Air gap Length
It can be also seen from Table 1 and Eq. 2, that the load
current io, which is equal to the sum of the secondary currents Let us consider the general case when n CDRs are
any stage of operation is equal to the sum of the magnetizing interleaved using a MIM core. The MIM core would then have
currents scaled by the turns ratio and is given by 2n outer legs each with one primary and secondary winding.
105
Extending Eq. 8 to this general case, the average flux density higher than that of the EI-core by ACM/ACE, where ACE is the
in the outer legs and center leg is given by area of cross section of the center leg of the EI-core.
Φ CM NsIo 1 (9)
BCM = = IV. COMPARATIVE ANALYSIS
ACM (RLM + 2nRCM ) ACM
1 Φ CM In this section, a 48V-3.3V/100W DC-DC converter is
B LM = chosen as a design example and the performance of the ICDR
2n A LM
with a MIM core is compared against a single CDR with an EI-
From Eq. 9, it is clear that the flux density in the center leg 22 core and an ICDR with 2 EI-18 cores. The relevant
is equal to that in the outer legs if ACM = 2nALM. Using the specifications of the converter are presented in Table 2.
definitions of the reluctances from Eqs. 4 and 5 Table 2. Specifications of 3.3V/100W half-bridge CDR
Variable Value
NsIo 1 (10) Input Voltage, 2Vdc 48 V
BCM =
2nRCM + RLM ACM Output voltage, Vo 3.3 V
NsIo µ N I Load current, Io 30A
= ≅ o s o Primary turns, Np 2
2n lLM 2nlGM Secondary turns, Ns
lGM + 1
µ o µ r Switching Frequency, fs 500kHz
Primary switch duty cycle, D 0.275
It is clear that setting n=1 and 2 renders Eqs. 9 and 10 Max flux density of 3F3 core material 250mT
applicable to a single CDR with an EI-core and the ICDR with
An EI-core and a MIM core are shown in Fig. 9 with their
the MIM core respectively. The required height of the air gap
relevant dimensions specified.
for a given load current and maximum allowable flux density
can be determined from Eq. 10. It is well known that eddy
currents are induced in the planar windings by the fringing flux
from the air gap. The amount of fringing flux is determined by
xM
the reluctance in the magnetic path across the air gap. In order
yE
to minimize eddy current losses, the planar windings and the
air gap are separated by a sufficient height. This prevents the
reduction of the window height below a lower bound thereby
being detrimental to achieving low profile magnetics. It can be tPE tPM
seen from Eq. 10, that the length of the air gap for the MIM lLE lLM
core can be reduced by a factor n when compared to the EI- WLE WWE WCE WWE WLE WLM WWM WCM WWM WLM
core. As will be shown later, the cross sectional area of the xE xM
center leg of the MIM core can be increased resulting in further (a) (b)
reduction of the magnetic path reluctance across the air gap. Fig. 9. (a) EI-core (b) MIM core
Hence, the height of the magnetics can be reduced to achieve The physical dimensions of EI-22 and EI-18 cores are
lower profile without sacrificing efficiency of the converter presented in Table 3.
and maintaining the same flux density as in the EI-core.
Table 3. Dimensions of EI-22 and EI-18 core
Variable Value (mm)
Inductance EI-22 core EI-18 core
From Eq. 8, the flux linkage of each secondary winding xE 22 18
yE 16 10
during the freewheeling period is given by WCE 5 4
(11) WLE 2.5 2
1 N s2 I o
λ LM = N sφ LM = . WWE 6 5
2n 2nRCM + RLM lLE 5.7 4
tPE 2.5 2
Since there are 2n outer legs, the current flowing through
each secondary winding during the freewheeling period is Core Volume
given by Io/2n. Hence, the inductance of each secondary Contour plots of the volume and area of cross section ACM,
winding is determined as of the center leg of the MIM core as a function of the width
λ LM N s2 µ N2A (12) WCM of the center leg and the length xM of a side of the core are
LM = = ≅ o s CM shown in Fig. 10. As the width WCM, of the center leg
I o 2n 2nRCM + RLM 2nlGM
increases, the width WLM, of the outer legs also has to increase
Substituting n=1, the expression for the inductance reduces in order to maintain ACM=4ALM for constant average flux
to that of the EI-core. Since the air gap length can be reduced density (Eq. 9). However, this results in the reduction of the
by a factor n, the inductance of each secondary winding is allowable window width for the planar windings (Fig. 9b).
Contour plots of the maximum allowable width of the planar
106
windings (dotted line) are also shown in Fig. 10a to illustrate The load current ripple for the 2 EI-18 core solution was
this effect. determined to be 0.85A and that for the single EI-22 core
24 24
ACM (mm2)
solution was 4.63A. This result is relatively obvious due to the
23.5
23
4 VE-22
3200
23.5
23
160 significant improvement allowed by using two interleaving
22.5
3
2887
22.5
140 CDRs. The parameter space of interest identified in Fig. 10 is
3
now reduced to the crosshatched region in Fig. 11 in order to
xM (mm)
xM (mm)
22 22
21.5 VE-18 21.5
21
2 2500
21
100
120
achieve a lower current ripple using the MIM core while
20.5
20 1800
2160
1
20.5
maintaining the reduced volume and winding width.
20 80
19.5 19.5
19
2 2.5 3
WCM (mm)
3.5 4 19
2 2.5 3
WCM (mm)
3.5 4 Flux density
(a) (b)
One of the most important advantages provided by the MIM
Fig. 10. Contour plots of (a) volume (solid) in mm3 and allowable width in
mm of planar winding (dotted) within window of MIM core and (b) Area of core over the 2 EI-18 core solution is the interleaving of the
cross section ACM in mm2 fluxes in the center leg. This can be seen in Table 4, where the
peak flux density and the flux density ripple in the center leg
The volume of the MIM core is calculated according to for each of the three cases is given
2
V MIM = x M (lLM + t PM ) (13) Table 4. Peak and ripple magnitude of center leg flux density
BL,pk (mT) BC,pk (mT) ∆BC (mT)
The thickness of the top plate tPM is determined such that the MIM Fig. 12a 244 Fig. 12b
maximum allowable flux density in the plate is not exceeded EI-18* 361.2 278.7 74.25
and is given by EI-22 303 261.8 37.12
ACM (14) Contour plots of the peak and ripple magnitude of the flux
t PM =
4 (x M − WCM ) density in the outer leg and center leg respectively of the MIM
core are shown in Fig. 12. The parameter space of interest
The height lLM of the MIM core window is determined to (smaller shaded area) is now reduced to that shown in Fig. 12
satisfy the condition that the PC board consisting of the planar to achieve a peak flux density lower than that in an EI-18 core.
windings be accommodated in the window and that there is a The larger colored area in Fig.12a represents the region where
prescribed distance from the air gap to the topmost layer of the the peak outer leg flux density is comparable to that in the EI-
winding to minimize induced eddy currents due to the fringing 22 core. This region can be plotted on the other contour plots
flux. Due to the reduction in the air gap length in the MIM shown in Figs. 10, 11 and 12b to determine how the other
core, the windings can be closer to the air gap thereby utilizing performance indices of the MIM core compare with the EI-22
the window more effectively. VE-18 and VE-22 is Fig. 10 are core. The significant reduction in the magnitude of the center
determined from the definitions in Table 3 as leg flux density ripple in the MIM core can be seen from Fig.
12b.
V E −18 = 2(18 × 10 × 6 ) = 2160mm 3 (15)
24 22
BLM,pk (mT) ∆BCM (mT)
3 4.5
V E −22 = 22 × 16 × 8.2 = 2887mm
23.5
21.5
23
22.5
21
xM (mm)
22
309 5
21.5
23.5
∆io (A) The form factor (Irms/Iav) of the secondary currents for the
23
MIM core, EI-22 core and dual EI-18 core solutions are
22.5
presented in Table 5.
xM (mm)
22
107
losses, even though the secondary current waveforms appear to
be distorted (Fig. 8) due to the magnetic coupling between the
outer legs of the MIM core.
s11
vp1
vp2
(a) (b)
Fig. 13. (a) MIM core and PC board with planar windings (b) 48V-
3.3V/100W MIM ICDR power board
108