Low Power Very High Frequency Switch-Mode Power Supply With 50 V Input and 5 V Output
Low Power Very High Frequency Switch-Mode Power Supply With 50 V Input and 5 V Output
Low Power Very High Frequency Switch-Mode Power Supply With 50 V Input and 5 V Output
6569
I. INTRODUCTION
HEN designing power converters it is always a goal to
reduce the price and the physical size, i.e. increase the
power density. The development of switch mode power supplies
(SMPS) has made it possible to increase the power density
significantly, but it is limited by the size of the passive energy
storing components (inductors and capacitors). The value and
size of these are however dependent on the switching frequency.
By increasing the switching frequency it will hence be possible
to reduce the size of SMPSs further.
Traditional SMPS topologies like Buck and Boost are hard
switching, this means the MOSFET is switching while energy
is stored in the output capacitance. The result is that energy is
dissipated in the MOSFET every time it turns on. Although this
introduces losses in the converter, it is not critical for converters
switching at 50400 kHz. But when the frequency is increased
to the very high frequency (VHF) range (30300 MHz) the
dissipated power get almost 1000 times larger. This amount of
energy would ruin the efficiency and require extreme cooling
of the MOSFET. This leads to the development of resonant
converters.
In order to avoid switching losses and be able to increase
the frequency while keeping the efficiency high, new topologies
have to be used. For the last two decades (since 1988 [1]), research has been done in order to enable the use of resonant RF
Manuscript received August 1, 2013; revised December 29, 2013 and October 28, 2013; accepted January 31, 2014. Date of publication February 12,
2014; date of current version August 13, 2014. Recommended for publication
by Associate Editor J. A. Cobos.
The authors are with the Department of Electrical Engineering, Technical University of Denmark, 2800 Kongens Lyngby, Denmark (e-mail:
mpma@elektro.dtu.dk; akn@elektro.dtu.dk; ma@elektro.dtu.dk).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TPEL.2014.2305738
Fig. 1.
amplifiers (inverters) combined with a rectifier for dc/dc converters, see Fig. 1. With these type of converters, it is possible
to achieve zero voltage switching (ZVS) and/or zero current
switching (ZCS). In this case, the MOSFET turns ON when the
voltage and/or current across/through it is zero. Theoretically,
this should eliminate switching losses if the switching is done
instantaneously and at exactly the right time. This is not practically achievable, but even with slight deviations from the ideal
case very high efficiencies can be achieved.
As already mentioned the value of the passive components
depends on the switching frequency. Hence, an increase in frequency will lead to a reduction in size, as long as the size of
the passive scales with the value. This assumption generally
holds, but magnetic materials and packaging introduce some
challenges. When the frequency is pushed far into the meghertz
range, magnetic core losses increase rapidly and become unacceptably high for most core materials [2]. At this point, air core
and PCB embedded inductors become a viable solutions, as the
inductances needed at these frequencies can be made in a small
physical size and the core losses avoided [3], [4].
Increasing the switching frequency also leads to capacitors
with lower values. Electrolytic capacitors which often limit the
overall lifetime [5], [6] can hence be avoided. The reduction
in component values also leads to a cost reduction as smaller
components are generally cheaper. If the frequency is increased
enough, some of the components can even be left out as they can
be constituted by the parasitic parts of other components (this
will be explained further in Sections II and III). An increase
in switching frequency will also make it easier to comply with
EMI requirements, as switching harmonics can easily be filtered
out by small and cheap filters.
With a switching frequency in the VHF range, it will also
be possible to achieve very fast transient responses [7] which
are highly demanded, e.g., for envelope tracking [8]. However,
in order to fully benefit from this, an efficient and fast control
loop has to be implemented. This is a big challenge and while
some ways of achieving continuous regulation have been found
[9], [10], the best results are still achieved using burst mode
(or cell modulation) as in [11][13]. Due to the high switching
0885-8993 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
6570
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 12, DECEMBER 2014
TABLE I
RESULTS FROM PREVIOUS RESEARCH
Fig. 2.
Relation between
V I2N f S
PO U T
TABLE II
DESIGN SPECIFICATION FOR THE CONVERTER
frequency the converter will reach steady state after just a few
microseconds, this makes it possible to use an array of small
converters and switch them on and off as needed. In this way,
each converter is designed to operate with a defined load/output.
This makes the design much easier as resonant inverters are
generally very load dependent.
The fact that resonant inverters are load dependent, makes it
very hard to achieve good performance at varying loads. Furthermore, resonant inverters need a large load impedance to
operate in the ideal situation (having both ZVS and ZCS). This
makes them well suited for boost-type converters, but making a
buck type is a bit more challenging. The most commonly used
way to overcome this challenge is to add an autotransformer at
the output, in that way the load impedance seen by the inverter
is increased [11], [14], [15]. Another way to achieve low output
voltage is to use an array of converters with the input in series
and the output in parallel in [16].
The most commonly used inverter is the class E, however
several other topologies exist. Some of the research results are
summed up in Table I. From the table it is seen that very high
efficiencies are achievable for the inverters up to 97%. However, the efficiency drops around 10% for the complete dc/dc
converters, i.e., when a rectifier is added.
From Table I, it is also seen that the converters have limited
gains, with a step down of 6.6 times and a step up of 2 being the
largest. As already stated the large reductions are not produced
solely by the converter, but with different ways to make the load
impedance appear larger. By further inspection, the connection
given by (1) can be seen, this relation is shown in Fig. 2
2
fS
1
VIN
.
POUT
(1)
Equation (1) shows that it is problematic to have a high input voltage and switching frequency while having a low output
power and still keeping the efficiency high. The input voltage
sets (together with COSS ) the energy stored in the output capacitance of the MOSFET each switching period and fS sets
MADSEN et al.: LOW POWER VERY HIGH FREQUENCY SWITCH-MODE POWER SUPPLY WITH 50 V INPUT AND 5 V OUTPUT
Fig. 3.
Fig. 4.
6571
A. Class E
The class E rectifier is a rather simple circuit, consisting of a
diode, two capacitors, and an inductor as shown in Fig. 3. Together these components constitutes a resonant rectifier capable
of rectifying the ac input current to a dc output.
For now, it will be assumed that the output capacitance is
infinite, so the output voltage is constant, and the diode is assumed to be ideal, i.e., no forward voltage drop, no junction
capacitance, and no reverse current.
In this case, the rectifier will appear resistive at the switching
frequency, if the resonance frequency of LR and CR are set to
this frequency. This will simplify the design of the inverter as
most design formulas are for a resistive load.
The scaling of the two components will determine the duty
cycle of the diode, DD . As the forward voltage drop of a diode
increases with the current running through it, it is desirable to
keep DD as high as possible. However, as the diode is connected
to the output through an inductor, the average voltage across it
has to be VOUT . Hence, a high DD will lead to a high peak
voltage across the diode.
In order to select DD , and thereby the scaling of the resonant
components, the values of real components have to be considered. All the considered inverters have a capacitor at the output
ensuring a pure ac path, without this there would be a direct dc
path from input to output and it would be impossible to reduce
the voltage. The average current through the diode will therefore
be the same as the output current.
With the forward voltage drop of a standard Schottky diode
being around 0.5 V (10% of the output voltage), it is crucial to use a diode with low forward voltage drop. Fairchilds
MBR0520L has one of the lowest forward voltages available,
max 385 mV, and can handle a reverse voltage of 20 V. With
this diode, the diode loss will be up to 77 mW or 7.7% of the
output power. This clearly limits the maximum achievable efficiency and fits very well with the 10% efficiency drop, seen in
Section I when going from an inverter to a complete converter.
If DD = 50% is chosen, the peak diode voltage will be 3.6
VOUT = 17.8 V leaving a little margin up to the maximum [29].
With this DD the value of CR should be [29]
1
= 67.5 pF.
CR =
2 2 fS RL
(2)
1
= 417 nH.
(2 fS )2 CR
(3)
6572
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 12, DECEMBER 2014
TABLE III
PROS AND CONS OF THE INVESTIGATED RECTIFIER TOPOLOGIES
C. Selection of Rectifier
Fig. 5.
Fig. 6.
IIN,p eak RL
+ CR RL
cos() = 1
2 CR VOUT
IIN,p eak
2
(1 cos( 2 DD ))
CR =
.
RL (1 + cos( 2 DD ))
DD =
(4)
(5)
(6)
(7)
MADSEN et al.: LOW POWER VERY HIGH FREQUENCY SWITCH-MODE POWER SUPPLY WITH 50 V INPUT AND 5 V OUTPUT
6573
VDS,p eak = VIN
Fig. 7.
.
2 (1 D)
(11)
(12)
(13)
2
VIN
8
2 + 4 POUT
(8)
POUT
2 .
2 CS VIN
(9)
It is desirable to keep the duty cycle low in order to reduce the peak voltage across the MOSFET. However, due to
turn on and off times and delays, it is decided to keep it close
to 50%. From (11), it is found that a duty cycle of 45% will
give a peak voltage of 142.8 V, leaving a little headroom if a
150 V MOSFET is used. Using this value along with the previous results, the needed reactance is found to be 326 . If a
capacitor of 680 pF is used, the value of the inductor can be
calculated according to
LR =
CR XRC S + 1
= 1.77 H.
CR S2
(16)
fS
.
2 (1 D)
(17)
1D
.
2 fR XRC
(18)
6574
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 12, DECEMBER 2014
TABLE IV
CURRENTS (RMS) IN THE CLASS E INVERTER
COSS , which are able to handle 150 V, have an output capacitance of 20 pF at 50 V. It is therefore necessary to reduce
the input inductor, in order to increase CS while keeping the
resonance frequency at the switch node equal to fR . The output
capacitance of the MOSFET is only contributing to the resonance in the part of the period where the MOSFET is OFF,
hence it has to be scaled by 1D in order to find the effective
capacitance. The effective capacitance of the output capacitor
CS
= 36.4 pF, hence the total inductance of the
is CS,e = 1D
resonance circuit and the input inductor should be
Ltotal =
1
= 936 nH.
R2 CS,e
Fig. 9.
(19)
1
LIN
1
+
R
XR C
LIN =
1
Ltotal
R
XR C
= 2.91H.
(20)
From a simulation, the rms current through the inductors and
the MOSFET is found (see Fig. 8 and Table IV). The IRF5802
MOSFET has the lowest available output capacitance for a
150 V power MOSFET capable of switching in the VHF range.
It has 20 pF output capacitance at 50 V and an on-resistance of
1.2 , this will give a conduction loss in the MOSFET of up
to 33 mW. The ac resistance of the inductors is estimated to be
100 m, thus they will have a combined loss of 5.37 mW. As
for the rectifiers, the losses in the capacitors are assumed to be
negligible. Thus, the total loss in the class E inverter is estimated
to 38 mW.
B. Class 2
As written in Section I, the large voltage peak across the
MOSFET is a big problem when the input voltage is large.
The class 2 (or EF2 ) inverter, which is a hybrid between the
class E and F2 inverters, was developed in order to make the
voltage across the MOSFET closer to a square wave. The voltage across the MOSFET should thereby become significantly
smaller (ideally 2 VIN for D = 50%). This is done by inserting
a LC circuit in parallel with the MOSFET as shown in Fig. 9.
This circuit is designed to have a resonance frequency at the
second harmonic, which causes the voltage across the MOSFET to become a trapezoidal wave consisting of the first and
third harmonics. The same benefits can be achieved with the
flat-top class-E amplifier described in [35].
According to [14], the rms voltage across the MOSFET can
be estimated by VDS,rm s = VIN 42 ; thus, the needed reactance
of the resonant circuit is different (see (14)). The new values
can be calculated to LR = 1.2 H and CR = 522 pF. No exact
equations for the calculations of the added LC circuit or the input
inductance are given in the literature; however, the following
Fig. 10. Tuned class 2 inverter waveforms simulated with PLECS (the peak
drain source voltage is reduced to 132 V).
TABLE V
CURRENTS (RMS) IN THE CLASS 2 INVERTER
1
= 625 nH
fS2 CS
(21)
LM R =
1
= 375 nH
15 2 fS2 CS
(22)
CM R =
15
CS = 18.8 pF.
16
(23)
MADSEN et al.: LOW POWER VERY HIGH FREQUENCY SWITCH-MODE POWER SUPPLY WITH 50 V INPUT AND 5 V OUTPUT
Fig. 11.
CS =
Fig. 12.
1
= 21.9 pF.
2 (1 D) 2 fS XR
(26)
C. Class DE
The class DE inverter has the same ZVS properties as the class
E inverter and the low voltage stresses of the class D inverter. It is
the counterpart of the class DE rectifier considered in Section II
and, as seen in Fig. 11, the two circuits are very alike. As the DE
rectifier, the DE inverter has two switches connected directly to
the dc voltage, in this case MOSFETs connected to the input
voltage. Both MOSFETs have capacitors across them which can
be tuned to achieve ZVS. The only additional components are
a resonant circuit at the output, just as seen for the previous
inverters.
As for the class E inverter, ZVS and ZDS switching can be
achieved in very specific situations. However, the values needed
are different [32]
RL =
2
VIN
2 POUT
(24)
fS,m ax =
POUT
2 .
2 CS VIN
(25)
6575
D. Selection of Inverter
During the analysis of the inverters, the values for all the
passive components where found and they are summarized in
Table VII.
All the inverters had some pros and cons, thus the same inverter will not be best for all applications. Some of the pros and
cons are listed in Table VIII.
The class E inverter consists of only one MOSFET, two inductors, and a capacitor (if CS is composed by the output capacitance of the MOSFET). It has however the largest voltage peak
across the MOSFET which will limit the input voltage for a
given MOSFET. Furthermore, the two inductors are both larger
than any of the inductors used for the two other converters.
This might limit the minimum size of the inverter as inductors
are assumed to be the largest components. The total loss was
estimated to be 38 mW or 4% of the output power.
6576
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 12, DECEMBER 2014
TABLE VII
COMPONENT VALUES FOR THE CLASS E, 2 AND DE INVERTERS
TABLE VIII
PROS AND CONS OF THE INVESTIGATED INVERTER TOPOLOGIES
TABLE IX
COMPARISON OF MOSFET CHARACTERISTICS
The class 2 inverter was a lot like the class E, the only difference being the added LC circuit put in to reduce the voltage
across the MOSFET. While this is a good way of keeping the
voltage down, the steep voltage curves require larger currents
making the loss larger than seen for the class E inverter. Although it has two extra components, compared to the class E
inverter, the physical size is expected to be more or less the
same as the values (and thereby the size) of the inductors are
smaller. The total loss was estimated to 62.8 mW which is a
65% increase compared to the class E inverter.
The class DE inverter was the only inverter with two switches.
However, as CS 1 and CS 2 are composed by the parasitic capacitance of the MOSFETs, the total component count of the class
DE inverter are the same as for the class E inverter. The peak
voltage across the MOSFETs were by far the lowest seen in
any of the inverters and the currents were also the lowest. These
things combined gave the smallest output inductor (which also
is the only inductor) and the lowest losses (13.2 mW).
From this analysis, the class DE inverter seems to be the best
solution and the class E inverter comes in second. However,
during this analysis the gate drive has not been considered. A
good high side gate drive which is capable of operating in the
VHF range has yet to be developed whereas a low side gate
drive can be made with few components [24]. The complexity,
price, and losses associated with the added high side gate drive
will, at least, reduce the benefits of the DE inverter.
With the above considerations in mind, the class E inverter
was chosen for the final design.
MADSEN et al.: LOW POWER VERY HIGH FREQUENCY SWITCH-MODE POWER SUPPLY WITH 50 V INPUT AND 5 V OUTPUT
Fig. 13.
6577
Fig. 14. Temperature measurements of the MOSFET and the diodes in the prototype with low C O S S (W b o a rd = 40 mm). (a) Placement of the components.
(b) The MOSFET is 55.3 C. (c) The diodes are 52.2 C.
TABLE X
MEASUREMENTS ON POWER STAGES
6578
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 12, DECEMBER 2014
Fig. 18.
The achieved
PO U T
V I N f S
by previous researchers. Furthermore, the PION U TS -factor explained in Section I is much smaller than for any of the converters
shown in Table I.
Fig. 15. Temperature measurements of the MOSFET and the diodes in the
prototype with low R D S (O N ) (W b o a rd = 32 mm). (a) Placement of the components. (b) The MOSFET is 65.1 C. The diodes are 53.2 C.
Fig. 16.
V. CONCLUSION
Fig. 17. Temperature measurements of the MOSFET and the diodes in the
prototype with large input inductor. (a) Placement of the components. (b) The
MOSFET is 46.2 C. (c) The diodes are 50.5 C.
The theoretical design of the resonant converter was considered in Sections II and III. Several different topologies were
considered and based on complexity and efficiency estimates a
class E inverter and rectifier were chosen.
The class E inverter was chosen based on complexity, efficiency, and the fact that it did not require a high side switch.
With a simple and efficient high side gate drive the DE inverter
is theoretically better, especially for converters with even higher
input voltages. Such a gate drive was however yet to be invented
and this topology was therefore not used for the practical implementation.
For the rectifier part it was again the class E topology that
were chosen, this time due to the forward voltage drop of the
diodes. With a low-voltage output, the forward voltage drop of
the diode becomes a significant percentage of the output voltage
and a single diode rectifier was found to be the best choice. For
higher output voltages, the DE rectifier might be better as the loss
due to forward voltage drop in the diodes becomes insignificant
and the voltages stress of the devices the major concern.
Three different power stages were made; one with a MOSFET with the lowest available output capacitance, one with a
MADSEN et al.: LOW POWER VERY HIGH FREQUENCY SWITCH-MODE POWER SUPPLY WITH 50 V INPUT AND 5 V OUTPUT
MOSFET with low on-resistance, and one with increased output power allowing a large input inductor. All the converters
had 50 V input and 5 V output and the achieved efficiencies
were between 60.7% and 82.9%. This shows that it is possible
to make low power very high frequency converters with high
step down ratio running at subnominal condition as long as the
components are chosen carefully.
REFERENCES
[1] W. C. Bowman, J. F. Balicki, F. T. Dickens, R. M. Honeycutt, W. A.
Nitz, W. Strauss, W. B. Suiter, and N. G. Ziesse, A resonant DC-to-DC
converter operating at 22 megahertz, in Proc. IEEE Third Annu. Appl.
Power Electron. Conf. Expo., Feb. 15, 1988, pp. 311.
[2] Y. Han, G. Cheung, A. Li, C. R. Sullivan, and D. J. Perreault, Evaluation
of magnetic materials for very high frequency power applications, IEEE
Trans. Power Electron., vol. 27, no. 1, pp. 425435, Jan. 2012.
[3] J. Qiu and C. R. Sullivan, Design and fabrication of VHF tapped power
inductors using nanogranular magnetic films, IEEE Trans. Power Electron., vol. 27, no. 12, pp. 49654975, Dec. 2012.
[4] H. B. Kotte, R. Ambatipudi, and K. Bertilsson, High-speed (MHz) series resonant converter (SRC) using multilayered coreless printed circuit
board (PCB) step-down power transformer, IEEE Trans. Power Electron.,
vol. 28, no. 3, pp. 12531264, Mar. 2013.
[5] W. Chen and S. Y. Ron Hui, Elimination of an electrolytic capacitor in
AC/DC light-emitting diode (LED) driver with high input power factor
and constant output current, IEEE Trans. Power Electron., vol. 27, no. 3,
pp. 15981607, Mar. 2012.
[6] H. Ma, J.-S. Lai, Q. Feng, W. Yu, C. Zheng, and Z. Zhao, A novel
valley-fill SEPIC-derived power supply without electrolytic capacitor for
LED lighting application, IEEE Trans. Power Electron., vol. 27, no. 6,
pp. 30573071, Jun. 2012.
[7] P. Shamsi and B. Fahimi, Design and development of very high frequency
resonant DC-DC boost converters, IEEE Trans. Power Electron., vol. 27,
no. 8, pp. 37253733, Aug. 2012.
[8] O. Garcia, M. Vasic, P. Alou, J. Oliver, and J. A. Cobos, An overview
of fast DC-DC converters for envelope, IEEE Trans. Power Electron.,
vol. 28, no. 10, pp. 47124722, Oct. 2013.
[9] J. J. Alonso, M. S. Perdigao, D. Vaquero, A. J. Calleja, and E. Saraiva,
Analysis, design, and experimentation on constant-frequency DC-DC
resonant converters with magnetic control, IEEE Trans. Power Electron.,
vol. 27, no. 3, pp. 13691382, Mar. 2012.
[10] K. Fukui and H. Koizumi, Class E rectifier with controlled shunt capacitor, IEEE Trans. Power Electron., vol. 27, no. 8, pp. 37043713, Aug.
2012.
[11] J. Rivas, Radio Frequency DC-DC Power Conversion, Ph.D. dissertation, Massachusetts Inst. of Technol., Cambridge, MA, USA, 2006.
[12] J. R. Warren, III, K. A. Rosowski, and D. J. Perreault, Transistor selection and design of a VHF DC-DC power converter, IEEE Trans. Power
Electron., vol. 23, no. 1, pp. 2737, Jan. 2008.
[13] J. Hu, A. D. Sagneri, J. M. Rivas, Y. Han, S. M. Davis, and D. J. Perreault,
High-frequency resonant SEPIC converter with wide input and output
voltage ranges, IEEE Trans. Power Electron., vol. 27, no. 1, pp. 189
200, Jan. 2012.
[14] J. M. Rivas, O. Leitermann, Y. Han, and D. J. Perreault, A very high
frequency DC-DC converter based on a class 2 resonant inverter, IEEE
Trans. Power Electron., vol. 26, no. 10, pp. 29802992, Oct. 2011.
[15] D. J. Perreault, J. Hu, J. M. Rivas, Y. Han, O. Leitermann, R. C. N. PilawaPodgurski, A. Sagneri, and C. R. Sullivan, Opportunities and challenges
in very high frequency power conversion, in Proc. IEEE 24th Annu. Appl.
Power Electron. Conf. Expo., Feb. 1519, 2009, pp. 114.
[16] J. S. Glaser, J. Nasadoski, and R. Heinrich, A 900W, 300V to 50V DC-DC
power converter with a 30 MHz switching frequency, in Proc. IEEE 24th
Annu. Appl. Power Electron. Conf. Expo., Feb. 1519, 2009, pp. 1121
1128.
[17] I. Douglas de Vries, High power and high frequency class-DE inverters,
Ph.D. dissertation, Univ. Cape Town, Cape Town, South Africa, 1999.
[18] Z. Kaczmarczyk, High-Efficiency Class E, EF2, and E/F3 Inverters,
IEEE Trans. Ind. Electron., vol. IE-53, no. 5, pp. 15841593, Oct. 2006.
[19] J. M. Rivas, Y. Han, O. Leitermann, A. D. Sagneri, and D. J. Perreault, A
high-frequency resonant inverter topology with low voltage stress, IEEE
Trans. Power Electron., vol. 23, no. 4, pp. 17591771, Jul. 2008.
6579
6580
IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 12, DECEMBER 2014