Professional Documents
Culture Documents
I. INTRODUCTION
UTOMOTIVE electronics are one field of power electronics that has been growing rapidly in recent years.
Some good examples of this are the electronics involved in the
dual battery system vehicles that use 14-V and 42-V batteries.
One of the specific converters for these vehicles is the bidirectional module placed in between those batteries that is in
charge of the power flow. Typically, the power of this converter
ranges from 500 to 1000 W. Due to the relative high current
of this application, some approaches use the interleaving technique [1]. The main advantages of using this technique in this
application are the filters reduction and efficiency. State of the
art engineering for this application proposes the use of three
to five paralleled buck stages (phases) to build the converter
[2][5]. A comparison between this multiphase converter with a
single buck converter is carried out in [2], where the advantage
of this technique for this application can be seen. Reference
[3] proposes a CAD tool to calculate the number of phases to
optimize cost, size, and weight. A similar analysis, but more
oriented to calculate power losses, can be found in [4]. A
magnetic component to couple all the phases is introduced in
[5], obtaining a size reduction compared with inductors for
the same power losses. A quite different solution is presented
in [6], where the authors propose a multilevel converter to
Manuscript received March 9, 2005; revised October 26, 2005. This work
was presented in part at PESC04 and APEC05. Recommended by Associate
Editor J. Shen.
O. Garca, A. de Castro, and J. A. Cobos are with the Divisin de Ingeniera Electrnica, Universidad Politcnica de Madrid, Madrid 28006, Madrid.
(e-mail: o.garcia@upm.es).
P. Zumel is with the Departamento de Tecnologa Electrnica, Universidad
Carlos III de Madrid, Madrid, Spain.
Digital Object Identifier 10.1109/TPEL.2006.872379
579
TABLE I
BRIEF OVERVIEW OF THE TWO PROPOSED DESIGNS
can be said about the capacitive losses because each transistor has a smaller capacitance but the addition of all of
them is constant. Switching losses are also constant if rise
and fall times are considered independent of the number of
phases.
In the case of using discrete semiconductors, the calculations should be done for each particular design since the
technology is different for each manufacturer. At least, the
advantage of using a high number of phases is that due to
ripple cancellation the switching frequency can be reduced
and, therefore, the switching and capacitive losses are reduced.
Inductors: the optimum number of phases from the point
of view of the magnetic component is hard to determine.
From the point of view of inductors, high current ripple is
preferred for the same averaged current, since it implies
lower losses in the inductor. However, several phases are
required in order to obtain a high cancellation of phase
current ripple and then a small output capacitance.
Each time a phase is added, the average current of it is
divided; if inductance is increased in the same way, the
total energy in the inductors is kept constant. Thus, size is
not drastically affected. However, a size reduction can be
obtained if inductance is kept constant when a new phase is
added and then the phase current ripple ratio is increased.
But finally, for a high number of phases, inductance should
be increased to avoid a very high current ripple with a very
small dc current. This will produce a poor efficiency.
Anyway, a high number of phases allows the use of small
magnetic components, and packaging may be improved
quite a bit, especially the height. In this case, the use of
16 phases allows an easy implementation of the inductor
using a RM7 core. Six turns are required and windings are
embedded in the printed circuit board (PCB) as shown in
Fig. 2.
Capacitors: a high degree of interleaving produces a ripple
cancellation that reduces the filter needs. As a result, the
designer can take advantage of ceramic capacitors. Both
capacitors are reduced although, in practice, the minimum
input capacitance is limited by the pulsating input current
of each phase. For a high number of phases, nonideal effects must be taken into account. Inductor tolerances produce different values in the current ripple per phase, and
then the output capacitance for the same output voltage
ripple is higher than in the ideal case.
580
TABLE II
COMPARISON OF SEVERAL DC/DC CONVERTERS FOR THE SAME APPLICATION BUT WITH DIFFERENT NUMBER OF INTERLEAVED PHASES
581
(1)
being the phase number, the number of phases, and
the resolution of the duty cycle, which is equal to the range of
the counter.
Other equivalent hardware structure would be substituting the
additions by counters, assuring that these counters were delayed
according to expression (1).
B. Shift-Register Phase-Shifter
The second possibility is to introduce the driving signal of
the first phase into a shift-register [see Fig. 5(a)]. In this way,
a delay is obtained which is equal to the length of the shiftregister multiplied by the clock cycle. The total length of the
, while each driving signal
shift-register is at most equal to
is extracted from the position obtained by expression (1). These
582
C. Phase-Shifters Comparison
Three comparison criteria can be used to distinguish between
these two solutions.
Duty cycle resolution. The shift-register method leads to
shorter critical paths. Therefore, higher clock frequency
can be achieved and, as a consequence, duty cycle resolution can be increased.
Closed-loop dynamics. Using addition and comparison phase-shifters, duty cycle changes start affecting
all driving signals immediately, while for shift-register
phase-shifters these changes only affect the first phase
immediately. The rest of the phases are affected only after
a time equal to their delay. Therefore, somewhat higher
closed-loop dynamics is achieved with the addition and
comparison method.
Area. The addition and comparison phase-shifter is very
sensitive to the number of phases, as each phase needs its
own adder and comparator. The shift-register phase-shifter
is not sensitive to the number of phases, as including more
phases needs is just extracting more driving signals from
the already available shift-register. However, it is sensitive
to the duty cycle resolution, as the length of the shift-register is proportional to it.
IV. CURRENT SHARING
One of the concerns of the interleaved converters is current
sharing. Commercial integrated circuits solve this problem by
including an additional current loop [11], [12]. As a consequence, the cost of the IC is quite high. Also, the additional
circuitry grows, increasing size and decreasing reliability.
Therefore, although the aforementioned IC controllers have
been designed with the capability of paralleling some of them,
in practice, a high number of phases is not feasible.
The purpose of this paper is to use a high number of phases
but without any current loop. The dc current depends strongly
on the conduction mode of the converter.
A. Continuous Conduction Mode (CCM)
Fig. 6 shows the equivalent dc circuit of a multiphase buck
converter when it operates in CCM. Each phase is characterized
; the voltage applied to this reby a dc parasitic resistance
(2)
(3)
The worst-case for a single phase takes place when this phase
has the maximum duty cycle and the minimum resistance while
the rest have minimum duty cycle and maximum resistance.
In such a case, the phase current is maximum while the other
phases will handle a current below the average value
. In
order to determine which of both factors (differences in duty
cycle or in resistance) is the most important, we can analyze
each one independently. This analysis can be found in detail in
[13], but the main results are the following ones. The differences
caused by resistance unbalance when only one resistance is different from the others can be calculated as shown in
(4)
being the common resistance for the rest of the phases and
the difference in the unbalanced resistance. On the other
hand, the differences caused by duty cycle unbalance when only
one duty cycle is different from the others can be calculated as
shown in
(5)
being the common duty cycle for the rest of the phases,
the difference in the unbalanced duty cycle, and the power
efficiency due to losses on the resistance exclusively.
583
TABLE III
DC CURRENT VALUES IN DCM AND CCM
FOR 1% AND 5% DUTY CYCLE INEQUALITIES
In order to compare both factors, a numerical example is analyzed. For a 16-phases converter with 98% efficiency due to
resistance (2% losses in the resistance), a 10% difference in one
of the resistances causes less than a 10% difference in the current of that phase. However, for the same converter a minimum
1% difference in one of the duty cycles causes a 47% unbalance
in the current of the unbalanced phase. As it can be seen, duty
cycle is responsible for the main current unbalance unless the
resistance causes very high losses (over 10%), which is avoided
by design. Regarding the inductor value, its differences cause
only unbalanced current ripples (peak to peak), but the dc current per phase is unaffected in CCM. However, it affects dc current in discontinuous conduction mode (DCM), as explained in
the next point.
Therefore, it can be stated that duty cycle is the main cause of
current unbalance in CCM. However, the use of digital control
drastically reduces unbalances caused by duty cycle, because
the driving signal is generated with great accuracy (differences
below 1 ns). Differences in duty cycle of the phases will be
produced by drivers and MOSFETs variations (so they should
be chosen taking this into account).
Thus, in many cases, it is possible to eliminate current sensing
circuits, current loops, and all the associated circuitry. In conclusion, the control stage is composed of a single voltage loop
and driving signals generator, making it feasible to build a multiphase converter with many phases (more than the classical three
or four) at a reasonable cost.
In the experimental results section, the converter has been
designed without current loops trusting in the digital control for
current equalization. It will be seen that it is not necessary to
include this current loop.
B. Discontinuous Conduction Mode (DCM)
DCM is a very interesting option for multiphase converters
because the equalization of the currents is much better. Inductor
current of a single phase in a switching cycle is shown in Fig. 7.
The average value of the inductor current (output current of a
phase) can be calculated from Fig. 7 and is defined in
(6)
and even a 5% difference in duty cycle causes just a 10% current unbalance. Regarding the inductor value, a 10% difference
causes a 9% current unbalance, and even a 20% difference in
inductance causes just a 17% current unbalance.
C. Comparison Between CCM and DCM
Table III shows a numerical comparison between these two
conduction modes in terms of current unbalance. The calculations have been carried out for the following specifications and
data.
Sixteen-phases synchronous buck converter (to test DCM,
free-wheeling MOSFETs is turned-off when current
reaches 0).
One phase has 1% (or 5%) higher duty cycle than the other
15 phases.
.
Input voltage: 42
.
Output voltage: 14
Output power: 1000 W.
Inductance: 5.4 H.
Parasitic resistance: 35 m MOSFET inductor .
Switching frequency: 120 kHz.
With a relative small duty cycle deviation such as 1%, CCM
shows an unacceptable current unbalance (current is 1.84
the nominal) while DCM current unbalance is kept below 2%.
Therefore, if CCM mode is preferred the duty cycle should be
very precise or the designer is forced to include one current
loop per phase. Table II shows the benefits of operating in DCM
even if the duty cycle is not very accurate. With a 5% deviation,
the current is only 10% over the average.
Therefore, a conservative criterion is to select DCM as an
operation mode. Thus, current loops can be removed obtaining a
very good current balance. However, due to the accuracy of the
digital control, the dc currents are very similar in CCM as will
be shown in the next section. This operation mode is interesting
because root-mean-square (RMS) currents are smaller at full
load, and therefore the converter efficiency is higher.
V. EXPERIMENTAL RESULTS
584
Fig. 10. Prototype made with 16 phases: (a) single phase power stage and (b)
whole power stage.
585
Fig. 11. (a) Measured current waveforms and (b) dc value in the 36-phases prototype at half-load; same in (c) and (d) but at full load.
Fig. 13. Efficiency of the converter with 36 phases as a function of the output
current.
Fig. 12. Evolution of four phase currents of the converter during a transient.
These converters have been designed with automotive specifications. The main feature is that they have been implemented
using surface mounting devices (SMD), keeping a very good
efficiency (94%95%), and avoiding the use of heatsinks. Thus,
the look of these dc/dc converters is rather different than others
of the state of the art in the same power range, mainly because
their low profile.
The two main problems associated with many power stages
have been overcome in this proposal.
586
Fig. 14. Prototype made with 36 phases: (a) MOSFETs and drivers PCB and
(b) inductors and capacitors PCB.
REFERENCES
[1] B. A. Miwa, D. M. Otten, M. E. Schlecht, and , High efficiency power
factor correction using interleaving techniques, in Proc. IEEE Appl.
Power Electron. Conf. Expo (APEC92), 1992, pp. 55756.
[2] A. Consoli, G. Scarcella, G. Giannetto, and A. Testa, A multiphase
DC/DC converter for automotive dual voltage power systems, IEEE
Ind. Applicat. Mag., vol. , pp. 3542, Nov./Dec. 2004.
[3] T. C. Neugebauer and D. J. Perrault, Computer aided optimization of
DC/DC converters for automotive applications, in Proc. IEEE Power
Electron. Spec. Conf. (PESC00), 2000, vol. 2, pp. 689695.
[4] M. Gerber, J. A. Ferreira, I. W. Hofsaer, and N. Seliger, Interleaving
optimization in synchronous rectified DC/DC converters, in Proc.
IEEE Power Electron. Spec. Conf. (PESC04), 2004, pp. 46554661.
[5] J. Czogalla, J. Li, and C. R. Sullivan, Automotive application of
multi-phase coupled-inductor DC-DC converter, in Proc. Ind. Applicat. Conf., 2003, vol. 3, pp. 15241529.
[6] F. Z. Peng, F. Zhang, and Z. Quian, A magnetic-less dc-dc converter
for dual voltage automotive systems, IEEE Trans. Ind. Applicat., vol.
39, no. 2, pp. 511518, Mar./Apr. 2003.
[7] L. Jourdan, J. L. Schanen, J. Roudet, M. Bensaeid, and K. Segueni,
Design methodology for non insulated DC-DC converter: application
to 42 V14 V Powernet, in Proc. IEEE Power Electron. Spec. Conf.
(PESC02), 2002, vol. 4, pp. 16791684.
[8] J. A. Oliver, P. Zumel, O. Garca, J. A. Cobos, and J. Uceda, Passive component analysis in interleaved buck converters, in Proc. IEEE
Appl. Power Electron. Conf. (APEC04), 2004, vol. 1, pp. 623628.
[9] High Efficiency High Density Polyphase Converters for High Current
Applications, Application note 77, Linear Technology Inc., Sep. 1999.
[10] A. de Castro, T. Riesgo, O. Garcia, and J. Uceda, A methodology to
design custom hardware digital controllers for switching power converters, in Proc. IEEE Power Electron. Spec. Conf. (PESC04), 2004,
vol. 6, pp. 46764681.
[11] High-frequency multiphase controller, Tech. Rep. TPS40090, Texas
Instrument Datasheet, Oct. 2003.
[12] Polyphase, high efficiency, synchronous step-down switching regulators, Tech. Rep. LTC1629, Linear Technology Datasheet, 1999.
[13] A. V. Peterchev, J. Xiao, and S. R. Sanders, Architecture and IC implementation of a digital VRM controller, IEEE Trans. Power Electron.,
vol. 18, no. 1, pp. 356364, Jan. 2003.
[14] O. Garcia, P. Zumel, A. de Castro, and J. A. Cobos, High current
dcdc converter with SMT components, in Proc. IEEE Appl. Power
Electron. Conf. (APEC05), Mar. 2005, vol. , pp. 14011406.
[15] O. Garcia, P. Zumel, A. de Castro, J. A. Cobos, and J. Uceda, An automotive 16 phases DC/DC converter, in Proc. IEEE Power Electron.
Spec. Conf. (PESC04), 2004, vol. 1, pp. 350355.