AN9506: A 50W, 500kHz, Full-Bridge, Phase-Shift, ZVS Isolated DC to DC Converter Using the HIP4081A

No. AN9506
Intersil Intelligent Power
April 1995
A 50W, 500kHz, Full-Bridge, Phase-Shift, ZVS Isolated DC to DC
Converter Using the HIP4081A
Author: David J. Hamo
Introduction
ability to vary the turn-on delays of both upper and lower
MOSFET switches. This is an essential feature for realizing
zero voltage switching (ZVS). The net result is flexibility and
capacity to derive the control logic drive signals necessary
for phase-shift ZVS switching. The voltage rating of the
HIP4081A is 80V, which is ideal for telecom DC to DC converters. With the added overvoltage protection circuit which
turns on the lower MOSFETS and turns off the upper MOSFETS, further protection is supplied to the system. The block
diagram of the full-bridge phase-shift power supply
described in this application note is shown in Figure 1. The
circuit discussed here has an output power capability of
50W, but operation can be scaled upwards to the 500W
range with the appropriate power component changes.
Many articles and papers have been published recently promoting the performance and benefits of the Phase-Shift,
Full-Bridge Topology and rightly so. This topology productively utilizes the same elements that have been plaguing
power supply designers for decades, those infamous parasitic components. The topology enables designers to advantageously employ transformer leakage inductance, MOSFET
output capacitance and the MOSFET body diode, enabling
designers to easily move their designs upwards in frequency. The topology offers additional advantages like zerovoltage-switching at a constant switching frequency, which
substantially reduces switching losses. This can be significant enough to eliminate heatsinking of power MOSFETS
and/or enabling the use of less expensive power devices.
Reduced EMI and RFI are additional benefits, since the voltage and current switching waveforms are much “cleaner”
and waveform edges switch softly compared to conventional
pulse width modulation (PWM) techniques. The ability to
move upwards in frequency will ultimately reduce the overall
size and lower the cost of the supply. One megahertz operation and beyond is possible with this topology. This is truly a
major advancement in topological architecture. The requirements for this design are a full bridge configuration, an additional inductor to aid resonant operation and output structure
consisting of a dual diode rectifier and an LC filter. Special
thermal substrates may not be required. As a result, cost
savings can be realized by utilizing inexpensive FR4 printed
circuit board material in place of elaborate thermal designs.
What’s more, EMI/RFI filtering requirements and heatsinking
are less rigorous further reducing costs. Therefore, focusing
on the overall system cost, it can be demonstrated that
employing this topology does have merit.
HIP4081A Features
The HIP4081A is a member of the HIP408X family of high
frequency H-Bridge driver ICs. The HIP4081A H-Bridge
driver has the ability to operate from 8 to 80VDC for driving
N-channel MOSFET H-Bridges. The HIP4081A is packaged
in both 20 Lead DIP and 20 Lead SOIC, provide peak gate
current drive of 2.5A and can switch up to 1MHz. A combination of bootstrap and charge-pumping techniques is used to
power the circuitry which drives the upper halves of the HBridge. The bootstrap technique supplies high instantaneous current needed for turning on the power devices,
while the charge pump provides enough current to “maintain” bias voltage on the upper driver sections and MOSFETs. Since voltages on the upper bias supply pin “float”
along with the source terminals of the upper power switches,
the design of this family provides voltage capability for the
upper bias supply terminals of 95VDC. Two resistors tied to
pins HDEL and LDEL can provide precise delay matching of
upper and lower propagation delays. The programmable
delay range for this device is 10ns to 100ns. This variable
delay capability is imperative for zero voltage switching and
will be described shortly.
Presently, there are limited phase-shift controllers on the
market, and those that are available are still expensive. Having faced this problem, an investigation of designing a discrete solution was performed. What was discovered proved
to be encouraging. With any low cost single ended controller
and two low cost logic ICs, one can generate all the gating
and delay functions necessary to derive the phase-shift control waveforms. What makes this all possible is the Intersil
HIP4081A MOSFET driver. The HIP4081A is capable of
independently driving four MOSFETs directly, eliminating the
need for traditional drive transformers. The HIP4081A also
allows additional drive and control capabilities unavailable
with conventional gate drive transformers. This includes the
1-888-INTERSIL or 321-724-7143 | Copyright
Phase-Shift Control Overview
Due to the number of well written papers on the theory and
concepts of phase-shift ZVS conversion, we will limit the discussion to the implementation of the technology. Several
papers are listed in the reference section that address
phase-shift ZVS conversion topics further.
The full bridge drive control logic required for this topology is
not conceptually complex. Figure 2 shows the full bridge with
© Intersil Corporation 1999
1
Application Note 9506
CURRENT
SENSOR
EMI
FILTER
A
VIN
-36VDC
TO
-72VDC
C
+5V
START-UP
CIRCUITRY
D
B
VCC
PWM
CONTROLLER
PWM TO PHASE SHIFT
CONVERTER LOGIC
REFERENCE
AND
ISOLATION
HIP4081A
FIGURE 1. FULL-BRIDGE, PHASE-SHIFT POWER SUPPLY BLOCK DIAGRAM
POWER DELIVERY
INTERVAL
LEFT LEG
TRANSITION
INTERVAL
VIN
FREE WHEEL
INTERVAL
iPRI
RIGHT LEG
TRANSITION
INTERVAL
C
A
0
NS
LIK
VPRI
VPRI
LO
SLEW
INTERVAL
VO
NP
CO
iPRI
DRIVE
SIGNALS
NS
A
B
B
D
C
D
t0 t1
t3 t4
t2
t5
FIGURE 2.
FIGURE 3.
associated parasitic components essential for ZVS operation. Figure 3 shows the waveforms associated with the circuit of Figure 2. During phase-shift ZVS operation there are
five states or intervals of time, that take place per half cycle
of operation. These states will be briefly discussed and only
one half cycle of the bridge will be described due to the circuits symmetric operation. Refer to Figure 3 during the following descriptions and time interval identifications.
and to provide realistic transition delay times. The term resonant inductance will refer to the combination of transformer
leakage inductance and any additional inductance in the primary path.
Power Delivery Interval (t2-t4)
The gate drive signals and timing diagram associated with
the full bridge are shown in Figure 4. The power delivery
interval of the phase shift topology is similar to the traditional
full bridge converter, in that two diagonal switches are on
(A&D or B&C). This applies the full input voltage across the
primary and results in power transfer to the load. The
amount of time these switches are on is directly proportional
to the phase shift between the two sets of waveforms AB
and CD. The phase between these sets of waveforms will
change as required, to regulate the output voltage. A 100%
phase shift will result in 100% duty cycle. Conversely, a 0%
phase shift will result in 0% duty cycle.
Slew Interval (t0-t2)
The slew interval is the time it takes for the primary current
to reverse directions. This time is established by the level of
input voltage across the primary, load current and the total
inductance in the primary path, referred to as the resonant
inductance. This inductance includes the leakage inductance of the transformer and any additional inductance in the
primary path. Additional inductance may be required to store
enough energy to displace the capacitive charge on COSS
2
Application Note 9506
The MOSFET output capacitance COSS is multiplied by 4/3
to approximate the average capacitance value during a varying drain-to-source voltage. The derivation of this equation is
listed in appendix A.
A
B
C
Right Leg Transition Interval(t3- t4)
D
The second ZVS delay is called the right leg (C&D) transition
time which terminates the power delivery interval. This is the
time required to displace the charge on the output capacitance of the C&D leg. The converters output inductor current
is reflected to the primary and therefore is the source of
energy which will displace this charge. The displacement of
this charge forces the voltage across MOSFET C to zero
(MOSFET D ZVS occurs during the cycles second half),
enabling zero voltage switching to take place. In this case
however, the mechanism for displacement of charge is not
resonant, but linear since this transition is modelled by a current source of reflected output current driving the output
capacitance. The time is given by:
A&D
ON TIME
B&C
ON TIME
FIGURE 4
Freewheel Interval (t4-t5)
Notice the following from Figure 3. There is a time when
A&C switches will conduct simultaneously and B&D
switches will conduct simultaneously. This state is called the
“freewheel” interval. This is how the phase-shift controller
can perform control without changing the frequency, unlike
other quasi-resonant topologies. As the load requirements
change, the freewheel time changes accordingly. The freewheel time increases with light loads and decreases with
heavy loads. In other words, the freewheel time is a way for
the controller to idle until the next appropriate state comes
along. During the freewheel time, reflected load current is
circulated through the FET switches A&C or B&D and the
voltage across the primary is zero. Figure 3 shows the droop
in primary current during this time. This is caused by conduction losses in the circulation path and output inductor ripple current.
C R × V IN
t RL = -----------------------IP
VIN = Voltage applied to full bridge
IP = Peak primary current
CR = Resonant capacitance
tRL = Transition time for the right leg interval
Both energy sources required to displace the charge on the
drain-to-source capacitances of the MOSFETs are load
dependant. This makes it difficult to maintain zero-voltageswitching at light loads. However, this fact does not pose a
serious problem as described later in the section titled ZVS
Design Considerations.
Transition Intervals (ZVS Delay)
Left Leg Transition Interval (t0- t1)
The method for ZVS involves the displacement of charge in
the drain-to-source capacitances of the MOSFETs and
occurs differently for the two legs of the bridge. The left leg
(A&B) transition interval begins after the freewheel state to
initiate the power delivery interval. This is the time required
to displace the charge on the output capacitance of the A&B
leg. For the left leg, the source of energy that displaces this
charge is stored in the transformer’s leakage inductance
plus any additional inductance in the primary path (the total
being the resonant inductance). The displacement of this
charge forces the voltage across MOSFET A to zero (MOSFET B ZVS occurs during the cycles second half), enabling
zero voltage switching to take place. Here the MOSFETs
output capacitances form a resonant circuit with the resonant inductance. The charge is displaced in a time equal to
one-fourth the resonant period. As a result, the left leg transition time is given by:
π
t LL = --- × L R C R
2
Phase-Shift Drive Derivation
Deriving the phase-shift drive control logic from a standard
off-the-shelf PWM controller is straight forward. The controller chosen is the Unitrode UC3823A. This was chosen for its
high speed operation and low start-up current. However, any
low-cost, high-speed controller could be used. Figure 5
shows the logic circuit used to derive the phase-shift control
logic. The timing diagram for this circuit is shown in Figure 6.
Notice that the clock output from the UC3823A is shown only
for reference purposes. The clock signal is not used in the
circuit, however, many of the following equations will include
the clock period tCLK in their composition. The clock period
tCLK is 2µs.
The PWM output from the controller clocks the flip-flop which
outputs waveforms at its Q and Q outputs, which become
a&b. At the same time the PWM logic signal is exclusive
ORed with the Q and Q outputs of the flip-flop. This generates the c&d waveforms as shown in the timing diagram.
Normally “slivers” are developed on the c&d waveforms.
They come about from the time delay caused by the PWM
clock signal propagation time through the flip-flop. Using
advanced CMOS logic the maximum sliver width will be
approximately 15ns. While it is possible to generate sliverless waveforms with additional circuitry, this does not
present a problem since small capacitors filter the slivers as
(EQ. 1)
tLL = Transition time for the left leg interval
LR = Transformer leakage inductance + additional inductance
CR = Resonant capacitance
The resonant capacitance is given by:
4
C R = --- × C OSS + C XFMR
3
(EQ. 3)
(EQ. 2)
COSS = MOSFET output capacitance
CXFMR = Transformer capacitance
3
Application Note 9506
shown in Figure 5. The Q and Q signals are passed through
an XOR gate which is configured as a non-inverting buffer
and become a&b. This is to match the timing of the a&b
drive outputs with the c&d drive outputs. These signals are
then input into the HIP4081A which in turn drives the MOSFET H-Bridge. The A&B drive signals drive the Left-Leg of
the bridge while the C&D drive signals drive the Right-Leg of
the bridge.
The ZVS timing delays are determined by Equation 1 and
Equation 3. Once the requirements are known the
HIP4081A turn on delay times can be set accordingly by the
resistor values on the HDEL and LDEL pins. Normally, the
left leg and right leg would like to be controlled independently. With the HIP4081A the upper and lower device delay
times are controlled independently. This causes a requirement that both delay times be identical. Typically the left leg
delay is slightly longer than the right leg delay. In this case
the HDEL and LDEL are set to the longer of the two. This
concept will be explained further in the section titled ZVS
Design Considerations.
74ACT86
a
VREF
74ACT86
PR
D
FROM
PWM
CONTROLLER
Q
74ACT74
b
TO HIP4081A
74ACT86
INPUT
The Design Process
Now that the proper control signals for the phase-shift topology have been realized its time to begin working through the
design of the power supply of Figures 7A and 7B.
c
Q
CL
74ACT86
VREF
d
To begin, the overall power supply requirements have been
defined by:
Input Voltage = -36V to -72V
FIGURE 5.
Output Voltage = +5V
The flip-flop and XOR gates receive their power from the
controllers VREF terminal of the UC3823A, which outputs
+5V. The power requirements for the logic devices are well
within the reference output current capabilities. However, the
reference should be properly by-passed.
Output Current = 10A
Switching Frequency = 500kHz
tCLK
CONTROLLER
CLOCK
PWM
OUTPUT
74ACT74
OUTPUT
Q
a
Q
b
PWM
+
Q=c
PWM
+
Q=d
ZVS
DELAY
(EQ. 1)
ZVS
DELAY
(EQ. 1)
HIP4081A OUTPUT
A
B
ZVS
DELAY
(EQ. 3)
ZVS
DELAY
(EQ. 3)
C
D
FIGURE 6. PHASE SHIFT TIMING DIAGRAM
4
POWERRTN
R33
33K
R34
10
Q5
BF720T1
R38
47K
T2
T37-8
C15
0.1µF
100V
+
BOOT
D2
75V
BZX84C75LT1
ILIM
R42
20K
D18
1N4148
C11
4µF
100V
VREF
U4
VREF
15
13
R23
10K
D15
BZX84C12LT1
12V
2
1
3
8
5
R22
10K
10
C20
0.1µF
VC
OUTA
VREF
OUTB
ILIM
NI
INV
CT
EAO
RAMP
RT
SS
GND
PGND
4
11
14
9
6
1
2
7
5
4
5
C32
470pF
VREF
4
C17
0.1µF
2
C22
0.22µF
C34
1µF
3
R24
6.49K
C21
47µF
10K
6
B
U5C
8
C
13
74ACT74
U5D
11
D
74ACT86
C26
2200pF
R45
4.7K
C25
2200pF
VREF
POWERAIL
R21
R5
619K
12
R43
1K
Q7
MMBT3904LT1
-36V TO -72V
6
CLK Q
1
U5B
74ACT86
CL
C18
1µF
+
9
10
U6A
D PR Q 5
+
+
A
74ACT86
_
C19
470pF
R39
2.2K
3
74ACT86
12
UC3823A
U5A
R7
10K
C29
0.1µF
FDBK
FIGURE 7A.
Application Note 9506
C14
0.1µF
100V
16
CLK
VCC
-36V to -72V
+5.0V
J2
ILIM
Q6
MMBT5401LT1
0.470Ω 2W
POWERRTN
R44
BOOT
D3
MBRS1100T3
IRFR120
Q1
R11
D4
U1
1
10
6
B
2
C
5
D
3
6
8
9
4
AHO
AHB
BHO
AHI
AHS
ALI
BHS
BHI
VDD
BLI
VCC
DIS
BLO
HDEL
ALO
LDEL
BLS
VSS
ALS
C7
0.1µF
IRFR120
Q3
R12
10
20
C4
4µF
100V
2.0µH
12
10T
19
16
R13
IRFR120
Q4
10
13
R14
17
14
C23
0.1µF
+
2T
2T
R35
100 1W
+5V/10A
4µH
+
IRFR120
Q2
D1
BAV70LT1
C5
0.47µF
C3
100µF
C30
2200pF
5T
5T
15
18
L1
T50-8
T1
EPC-19
L2
T44-6
D8
MBRB2535CTL
R4
4.99K
R2
C8
15K 0.1µF
U2
UC39432
2
10
C28
10µF
CMP
HIP4081A
5
4
6
R40
100K
MOLEX 22-59-1310
R41
100K
3
VCC
REF
EA+
COLL
SEN
GND ISET
R3
5.11K
POWERAIL
L3
+
C16
33µF
470µH
DT1608-474
VREF
FDBK
FIGURE 7B.
ISO1
PS2701-1
7
8
R1
39
1
Application Note 9506
7
A
BHB
11
MBRS1100T3
1
2
3
4
5
6
7
8
9
10
MOLEX 22-59-1310
10
C6
0.1µF
J1
1
2
3
4
5
6
7
8
9
10
Application Note 9506
For an EPC-19 transformer operating at 250kHz, a 25oC rise
will occur operating with a peak flux density of 1200G. If we
let the maximum primary duty cycle reach 80% the maximum on time will be 0.8t, then the maximum flux density
change within the time span of 0.8t/2 will be ∆B = 2400G.
The voltage across the transformer has been reduced by 2V
due to an anticipated voltage drop caused by the two MOSFET switches rDS(ON). The minimum primary turns from
Equation 5 is:
Power Semiconductor Selection
One of the overall goals of this supply is to maintain the lowest profile and size possible. This is important to module
manufacturers, as well as board level OEMs. We selected a
form factor close to the standard 3” x 3”, along with a minimum profile. The voltage and current requirements along
with size constraints has lead to the choice of four Intersil
IRFR120s for the full-bridge switches. They have a rDSON of
0.27Ω, a breakdown voltage of 100V and are available in the
surface mount TO-261(D-PAK)™ package. The output rectifier diode chosen is the Motorola MBRB2535CTL available
in a (D2-PAK)™. This rectifier was chosen because of its
very low forward voltage drop, slightly over 0.3V at 10A. This
is a important consideration since most of the power loss is
in the output rectifier.
 0.8 × 4 × 10 – 6 
36 – 2 ×  -------------------------------------  × 10 8
2
N P = ---------------------------------------------------------------------------------- ≥ 10Turns
0.227 × 2400
Now in the most ideal case:
NS
V O = V IN × -------- × D
N
Transformer Design
P
Maintaining the form factor requirements mentioned earlier
has lead to the choice of an EPC-19 ferrite core from TDK.
In this application the transformer size is limited by core loss.
The transformer will be designed for a temperature rise of
50oC. That coupled with a maximum ambient temperature of
50oC, the transformer can reach a maximum temperature of
100oC. PC40 material was chosen since it has a curie temperature in excess of 215oC, and low core loss at the switching frequency. Design curves given with the core material
(TDK catalog #BAE-030D) show temperature rise for a given
core loss. The design curves indicate that the core temperature will rise 50oC with the core dissipating 800mW. This
value is for core loss only and excludes any copper losses. If
the core losses and copper losses were equally distributed
in the transformer the core loss will be 400mW. Therefore,
we can determine the maximum core loss limitation for this
design:
P CORE
P CLOSS = -------------------Ve
For the non-ideal case:
NS


V O =  ( V IN – V MOSDROP ) × -------- – V RECT  × D
NP


Rearranging and solving for the secondary turns ratio:
VO


-------- + V RECT 

D
N S =  -------------------------------------------------  × N P
 V IN – V MOSDROP 


5
 ------- + 0.3 
 0.8

N S =  -----------------------  × 10 = 2Turns
 36 – 2 
The transformer turns ratio has been designed to yield a
given flux density excursion, thereby maintaining the limits of
temperature range. Next, the wire size must be determined
and the copper losses must not exceed 400mW to limit the
rise in core temperature to 50oC. For the primary windings a
current density of 500 circular mils per RMS ampere was
used. For the secondary a current density of approximately
200 circular mils per RMS ampere was used because of the
higher current and the fact that there are only two windings
per side required for the secondary. Using these current
densities, wire size requirements become 20AWG for the
primary and 19AWG for the secondary. For the secondary,
eight strands of 28AWG magnet wire will be used, yielding
an equivalent circular mil area of 19AWG. With wire
selected, the copper loss including skin effects and
increased wire resistance at 100oC, approach 300mW.
Therefore we can expect a maximum temperature rise of
slightly less than 50oC. The transformer was wound by interleaving the secondary halves between the primary. The first
half of the secondary will be wound first and the primary will
be wound next. The remaining secondary half will be wound
and finally, the auxiliary will be wound.
(EQ. 4)
mW
400mW
P CLOSS = --------------------------- = 382 ----------cm 3
1.047cm 3
Using the curves once again for the PC40 material, the core
loss vs flux density curves indicate that the peak flux density
for a core loss of 382 mW/cm3 is approximately 1200 gauss.
The switching frequency is 500kHz but with the full-bridge
topology the core flux swings at half the switching frequency.
Therefore the transformer switching frequency will be
250kHz while operating in the first and third quadrants of its
hysteresis curve. The remaining transformer design procedure is now straight forward.
From Faraday’s Law:
E × ∆t × 10 8
N P = -------------------------------Ae × ∆B
(EQ. 6)
(EQ. 5)
Where E = voltage across the transformer windings (Volts)
NP = number of primary turns
ZVS Design Considerations
Ae = Iron area of the core (cm2)
Now that the transformer has been designed it is a good
time to determine the ZVS transition times. This will allow
the turn on delays to be set properly for zero-voltage-switch-
∆B = Core flux density (Gauss)
∆t = time in which the flux is changed
D-PAK™ and D2-PAK™ is a registered trademark of Motorola
7
Application Note 9506
ing to take place. However, choosing the left leg transition
time requires much thought because this time is a function of
many variables. Therefore, before going any further some
concepts need to be clarified and an equation will be derived
that will allow this value to be chosen quickly and correctly.
From Figure 8:
∆D × t CLK = t 2 – t 1
Performing the substitution and multiplying by two since
there are two such transitions per period, the loss of duty
cycle on the secondary is then equal to:
In Figure 8, primary bridge voltage, primary current and secondary voltage waveforms are shown. Notice that the primary and secondary duty cycles are different. From these
waveforms the following relationship can be determined:
∆D = D – D
2 × N S × L R × I LOAD
∆D = ------------------------------------------------------t CLK × N P × V i
Solving for LR:
(EQ. 7)
e
∆D × t CLK × V i × N P
L R = ----------------------------------------------------2 × I LOAD × N S
∆D = Loss of duty cycle on the secondary side.
D = Primary voltage duty cycle.
We now have an expression for the total resonant inductance in terms of loss of duty cycle so that its value can be
easily determined. From the beginning of the design the
maximum secondary duty cycle has been chosen to be 80%.
Using this value and selecting a duty cycle loss of 15%, will
yield a maximum primary duty cycle of 95%. The leakage
inductance of the transformer is approximately 500nH and
the total resonant inductance calculation becomes:
De = Secondary voltage duty cycle or effective duty cycle.
The loss of duty cycle on the secondary side is a key concept. This loss is caused by the time it takes to change the
direction of the primary current (t2-t1). Therefore it is imperative that the resonant inductance value not be excessive.
Otherwise, this might require a larger turns ratio, since the
primary duty cycle could reach its maximum value while the
secondary duty cycle is incapable of sustaining the appropriate output voltage. This concept could be a stumbling block
for the unsuspecting. This leads to the importance of deriving an equation for LR in terms of this loss of duty cycle ∆D
which enables LR to be bounded properly. This in turn, will
lead to the correct value chosen for the left leg transition tLL.
0.15 × 2 × 10 – 6 × ( 36 – 2 ) × 10
L R = ---------------------------------------------------------------------------------- = 2.55µH
2 × 10 × 2
L RINDUCTOR = 2.05 µH – 0.5µH = 2.05µH
Before determining the left leg transition time the resonant
capacitance must be calculated. The IRFR120 MOSFET
switches have a COSS capacitance equal to 130pF and the
transformer primary capacitance is approximately equal to
10pF. Using Equation 2 the resonant capacitance is calculated:
Deriving an equation for ∆D:
Let t2-t1 equal the time it takes for the primary current to
slew. Since we know the slope of the sluing primary current,
the following equation can be determined from Figure 8.
4
C R = --- × 130 × 10 – 12 + 10 – 12 = 183pF
3
NS
-------- × I LOAD
NP
=
------------------------------t2 – t1
Vi
------LR
The left leg transition is then calculated using Equation1:
π
t LL = --- × 2.55 × 10 – 6 × 183 × 10 – 12
2
i
0
(EQ. 8)
PRI
I LOAD
= ----------------n
t LL = 34ns
In the previous calculation an alternative expression for tLL
could have been used:
DtCLK
VPRI
∆D × t CLK × V i × C R × N P
π
t LL = --- × -------------------------------------------------------------------2
2 × I LOAD × N S
V
i
SLOPE = -------L
R
tCLK
So making the left leg transition 34ns will cause the maximum primary duty cycle to be approximately 95% at full load
with the minimum input voltage applied. This allows 5% margin for variations in CR and LR, assuming nearly 100% duty
cycle is possible. These numbers can be adjusted easily by
the previous equations for your particular needs.
DetCLK
VSEC
Vi
n
∆DtCLK
0
t0 t1
t2
(EQ. 9)
t3 t4
Now that the resonant inductor and left leg transition time
have been selected, the right leg transition time needs to be
determined. It turns out that the maximum right leg transition
time occurs during the maximum input voltage and at a load
boundary called the ZVS operational limit. The ZVS operational limit is the point at which the power supply no longer
maintains zero-voltage-switching. This is a normal function of
this topology. As mentioned earlier, the two energy sources
(resonant inductance and output inductance) required to dis-
t5
FIGURE 8.
Rearranging:
N S × L R × I LOAD
t2 – t1 = --------------------------------------------NP × Vi
8
Application Note 9506
place the charge on the drain-to-source capacitances of the
MOSFETs are load dependant. Therefore at some load
value less than maximum, the energy stored in these
sources will be less than adequate to displace this charge.
This is the point at which the converter will no longer operate
in the ZVS mode. This is best exemplified by Figure 9. Figures 9 A, B and C show how the left leg transition is formed.
Figure 9A shows that the energy in LR is larger than the
energy required to displace the capacitance charge. Figure
9B shows the ZVS operational limit where the energy in LR
is equal to the energy required to displace the capacitance
charge. Figure 9C shows the energy in LR is not capable of
fully displacing the capacitance charge. The same scenario
is true for the right leg transitions but they will ramp at a linear rate and their energy source is the output inductance as
stated earlier.
VIN
VIN
tLL
A
So the power at which the supply stops zero-voltage-switching is 16.6W, well below 1/2 Po(max), which is within the
design goal requirements.
The right leg transition can now be determined using
Equation 3:
183 × 10 – 12 × 72
t RL = -------------------------------------------- = 20ns
0.662
The left-leg transition takes 34ns and the right-leg transition
takes 20ns. These values can now be programmed as turn
on delays via the HDEL and LDEL resistor values on the
HIP4081A. As previously mentioned, since the HIP4081A
controls the upper and lower delay times, both HDEL and
LDEL should be set equal to the longest delay time. This
time will be 34ns.
Resonant Component Selection
VIN
The core material selected for the resonant inductor was
chosen for its high Q and low core loss characteristics.
Micrometals powdered iron RF core T44-6 has a Q of 200 at
the resonant frequency of interest, 7MHz (FRES=1/(4xtLL)).
tLL
tLL
The output filter inductor was another Micrometals powdered
iron RF core, T50-8 chosen for its low core loss at 500kHz.
This material has the lowest core loss of any other material
they offer. As a result, it is also the most expensive. Another
core which may be used is the T50-52. The 52 material has
slightly higher core loss and is less expensive.
C
B
FIGURE 9.
It is more important to maintain ZVS at higher loads for two
reasons. The first and more obvious is that the switching
losses are greatly reduced. The second reason is because
the free-wheel time is at a minimum during full load. Therefore the circulation of the reflected load current during the
free-wheel time is shortened thereby reducing the I2R losses
in the free-wheel circulation path. During lighter loads the
power dissipation in the MOSFET switches should not be
significant if switching losses begin to manifest due to the
smaller primary currents during lighter load values. For this
reason, the design goal was to maintain ZVS operation
down to half the maximum output power (25W). It should be
noted that at higher converter power levels (>200W) it may
be necessary to place saturable inductors in series with the
anode leads of the output rectifier. This extends the range of
zero-voltage-switching to very low power levels further
improving efficiency. This concept is discussed in great
detail in several of the listed references.
Effects of Variations in MOSFET Parasitic
Output Capacitance
The MOSFETs parasitic output capacitance, COSS can vary
from part to part, and from manufacturer to manufacturer.
However, this variability does not cause any severe aberration in the operation of the converter. The only noticeable
effect can be a slight decrease in efficiency. This loss in efficiency is due to two factors.
The first and most significant is capacitive turn-on losses.
This is caused by the energy required to displace the charge
on the output capacitance when there is insufficient energy
in the resonant inductor to perform this task. This is indicated by the first term in Equation 11. The other losses are
simply turn-on and turn-off losses which normally occur if
zero voltage switching is not in effect. These turn-on and
turn-off switching losses are caused by the slight movement
of the resonant peak with respect to the programmed ZVS
delay time as COSS varies from its nominal value. This may
cause drain-to-source voltage to occur simultaneously with
drain current during a portion of the switching time, thereby
decreasing efficiency. This is indicated by the second and
third terms in Equation 11.
The equation which determines the minimum current
required for ZVS operation is:
1
--2
3
--2
I PRI ( critical ) =
2 × C R × V OSS × V IN ( MAX )
--------------------------------------------------------------------------LR
I PRI ( critical ) =
2 × 183 × 10
× 25 × 72
-------------------------------------------------------------------------- = 0.662A
2.55 × 10 – 6
– 12
1
--2
(EQ. 10)
3
--2
Figure 10 shows the curve of output capacitance as a
function of drain-to-source voltage. COSS is measured at a
drain-to-source voltage = 25V. This voltage point is
designated as VOSS. At this voltage point COSS is equal to
162pF as indicated by the curve.
The minimum primary current to maintain ZVS switching is
0.662A. To see what this means in terms of output power the
following calculation is performed:
NP
I O ( Critical ) = -------- × 0.662 = 5 × 0.662 = 3.31A
NS
Maximum variations of COSS for Intersil IRFR120 is ±20% of
the measured value as shown by Figure 10. However,
changes in COSS can be much more significant when select-
P O ( Critical ) = 3.31A × 5V = 16.6W
9
Application Note 9506
ing the same MOSFET from a different manufacturer. This is
because each MOSFET manufacturer has different processes, resulting in changes from the typical value. This
value can be significantly different from the typical value
shown in the data sheets. Therefore, it is a good idea to
work closely with your supplier to obtain the measured value
for COSS. The measured value for COSS at VOSS = 25V is
162pF as shown in Figure 10. Therefore, one can expect a
worst case variation of ±20% from this value as indicated by
Figure 10. It will be shown that even for large changes in COSS,
the total power dissipation in the converter will not be impacted
greatly. It should be pointed out that the initial design proceeded with COSS = 130pF as stated on the data sheet for the
typical value. It was later determined that the measured value
was 162pF. This difference had no noticeable effect on converter performance confirming that moderate variations in
COSS have little or no effect on converter performance.
efficiency will be slightly degraded only if the COSS value
turns out larger than the target design. To counter this effect
of possibly larger COSS, one could simply program the
delays slightly longer than the calculated values. It should be
pointed out that tSW was chosen to be 10ns in Equation 11,
to generate the curves of Figure 11.
In conclusion of this topic, once the measured value for
COSS is known, one can expect a worst case change of
±20%. This change is insignificant to the converters performance as shown by Figure 11. Once a vendor is selected
and transition delays are determined, the delays can be set
and forgotten about.
1
405pF 72V
PSWLOSS, (COSS, VIN) (W)
0.8
COSS , CAPACITANCE (pF)
1000
800
600
405pF 54V
0.6
COSS
∆COSS
0%
54V 195pF
+20%
72V 405pF +150%
195pF 72V
0.4
405pF 36V
162pF 72V
195pF 54V
0.2
400
162pF 54V
195pF 36V
+20%
200
0
-20%
0
0
VIN
36 V 162pF
5
10
15
20
VDS , DRAIN-TO-SOURCE VOLTAGE (V)
25
4
6
OUTPUT CURRENT (A)
8
10
Performance
Figure 11 shows the effects of increased switching losses by
varying COSS and the input voltage (essentially VDS). Equation 11 was used to determine these curves and is derived in
appendix A. From the curves of Figure11, it can be seen that
even for large variations the impact in switching losses is not
that great, especially if the converter is operating near full
load. Again, one needs to understand their application to
determine how much variation in COSS can be tolerated.
Hopefully Figure11 will help give understanding to the impact
of COSS variation on converter switching losses.
1
2
FIGURE 11. NON-ZVS OPERATION
FIGURE 10. COSS CURVE FOR THE INTERSIL IRFR120
Figures 12 and 13 show the primary voltage and current
waveforms of the converter taken at a load current of 10A
and an input voltage of 48V. As indicated below, the frequency of the primary side is 250kHz while the secondary
side frequency will be twice the primary or 500kHz. The primary current scale is 0.5A per division yielding a 3AP-P
waveform. Notice how clean these waveforms are. This is
one of the significant benefits of zero voltage switching.
Additionally, the primary current and rectified secondary voltage is shown in Figure 14 to illustrate the erosion of the secondary duty cycle as mentioned earlier. The erosion is
approximately 200ns, yielding a 10% reduction in the secondary side on time. The calculated value for a 48V input
was 11% for a difference of 1%. This exemplifies the importance of using Equation 8 to calculate the erosion of the secondary. On the secondary side we have the same problem
as with more classic topologies. A moderate snubber was
added to the rectifier to reduce the ringing. The two graphs
of Figures 15 and 16 show percent regulation and percent
efficiency. Percent regulation was excellent. Over the full
input voltage and load range, the maximum change in output
voltage was 54mV. During moderate to high output current
the efficiency was predominately in the low 80% range. The
actual measurements correlated very closely with the loss
analysis performed on this converter prior to its design.
While this topology offers lossless switching, the conduction
3
----2
4
2
2
--- × COSS × VOSS × VIN – 1
--- L R × I PRI
3
2
= --------------------------------------------------------------------------------------------------- +
P
SWLOSS
tCLK
162pF 36V
0
(EQ. 11)
I PRI × V SWLL × t SWLL I PRI × V SWRL × t SWRL
--------------------------------------------------------------------- + ----------------------------------------------------------------------tCLK
tCLK
From Equation 11 it can be seen that the switching loss is
≤ zero if the resonant inductance has enough energy to displace the output capacitance and if the switching losses are
zero due to ZVS operation. In the first term of the Equation,
decreasing values of output capacitance are an advantage.
Here the resulting energy in the resonant inductance
exceeds the energy needed to displace the capacitance
charge. Decreasing capacitance also means that the delay
time set for the ZVS transition will be longer than necessary.
This has no significant effect on converter efficiency. The
10
Application Note 9506
data is given in Table 1. From these figures one can target
specific areas to optimize the converter for greater efficiency.
For instance, the current sense resistor can be replaced with
a current transformer for improved efficiency.
losses do increase. What this topology offers for this configuration of input and output voltage is the ability to increase
the switching frequency, while at the same time, providing
much cleaner waveforms. The breakdown of the power loss
TEK RUN: 50.0ms/s AVERAGE
TEK RUN: 100ms/s AVERAGE
C1 FREQ
251.52kHz
M2 FREQ
250.26kHz
1
M2
MATH2
20.0V
CH1 10.0mV
0.5A/DIV
500ns PER DIV
FIGURE 12.
500ns PER DIV
FIGURE 13.
TEK RUN: 100ms/s AVERAGE
REG AT 48V
1.5
C1 FREQ
252.403kHz
REG AT 72V
% REGULATION
C2 FREQ
506.033kHz
1
1
0.5
REG AT 36V
2
0
CH2
10.0V
0
500ns PER DIV
5
OUTPUT CURRENT (A)
FIGURE 14.
FIGURE 15.
EFF AT 36V
80
EFF AT 72V
% EFFICIENCY
CH1 10.0 mV
0.5A/DIV
EFF AT 48V
60
40
0
5
OUTPUT CURRENT (A)
FIGURE 16.
11
10
10
Application Note 9506
The plots shown in Figure 17 illustrate the progression of
zero voltage switching. The first and second plots show that
the delay time is skewed from the nominal value. This is
because the HIP4081A delay was made excessively large to
illustrate the resonant half sinusoid. Ideally you want to set
the delay to 1/4 the period of the resonant frequency. This
delay time is determined by Equation 1. The final plot shows
the delay set to the optimum point for proper zero voltage
switching. Notice the absence of ringing on VDS compared to
when non-zvs operation is taking place!
TABLE 1. LOSS ANALYSIS AT VIN = 48V, IOUT = 10A
MOSFET Full-Bridge Conduction Losses
4W
Output Diode Losses
3W
Switching Losses
0W
Power Transformer Losses
0.8W
Output Inductor Losses
0.75W
Resonant Inductor Losses
0.57W
Current Sense Resistor Losses
1.2W
Snubber Losses
0.38W
Miscellaneous Losses
1.24W
Total Power Loss
11.94W
Efficiency
The plots of Figures 18 and 19 illustrate zero voltage switching of switch A and switch B. In the first plot of Figure 18, the
voltage across A is zero during turn-on of switch A. Here you
can see the HIP4081A driving the high side FET. Notice the
12V step on this waveform. This 12V step voltage is being
supplied by the HIP4081A bootstrap capacitor which turns
on switch A. The second plot shows the gate drive of switch
B, along with the same phase node, VDS of switch B. Here,
the voltage across switch B is zero during the turn-on of
switch B. In the remaining plots of Figure 19, the mechanism
is the same, but here the MOSFETS are being turned off. In
all of these waveforms you can see that the delay time is
nearly 34ns as calculated in the application note.
81%
Operating the converter with a load greater than 5A will
require a heatsink attached to the output rectifier. In addition,
the full bridge may require a larger copper area or small
heatsink for heat removal.
TEK RUN: 1.00GS/S AVERAGE
TEK RUN: 1.00GS/S AVERAGE
1
1
CH1
20.0V
50.0ns
CH1
FIGURE 17A. HIP4081A DELAY SET FOR LARGE SKEW
20.0V
50.0ns
FIGURE 17B. HIP4081A DELAY SET FOR MODERATE SKEW
TEK RUN: 1.00GS/S AVERAGE
1
CH1
20.0V
50.0ns
FIGURE 17C. HIP4081A DELAY SET FOR PROPER ZVS OPERATION
FIGURE 17. NON-ZVS TO ZVS PROGRESSION, VDS OF SWITCH B, VIN = 48V, IOUT = 3A
12
Application Note 9506
TEK RUN: 2.00GS/S AVERAGE
TEK RUN: 2.00GS/S AVERAGE
VOLTAGE OF
VOLTAGE OF
A AND B PHASE NODE
A AND B PHASE NODE
1
1
VGATE TO SOURCE
VGATE TO GND
SWITCH A
SWITCH B
2
2
CH1
20.0V
CH2
20.0V
CH1
25.0ns PER DIV
FIGURE 18A. VOLTAGE ACROSS SWITCH A = 0
DURING TURN-ON OF SWITCH A
20.0V
CH2
5.00V 25.0ns PER DIV
FIGURE 18B. VOLTAGE ACROSS SWITCH B = ZERO
DURING TURN-ON OF SWITCH B
FIGURE 18. ZERO VOLTAGE SWITCHING OF SWITCH A AND B, ZVS TURN-ON
TEK RUN: 2.00GS/S AVERAGE
TEK RUN: 2.00GS/S AVERAGE
VOLTAGE OF
A AND B PHASE NODE
VOLTAGE OF
A AND B PHASE NODE
2
2
VGATE TO SOURCE
SWITCH B
1
VGATE TO GND
SWITCH A
CH1
20.0V
CH2
20.0V
1
CH1
25.0ns PER DIV
FIGURE 19A. VOLTAGE ACROSS SWITCH A = ZERO
DURING TURN-OFF OF SWITCH A
5.00V
CH2
20.0V
25.0ns PER DIV
FIGURE 19B. VOLTAGE ACROSS SWITCH B = ZERO
DURING TURN-OFF OF SWITCH B
FIGURE 19. ZERO VOLTAGE SWITCHING OF SWITCH A AND B, ZVS TURN-OFF
Conclusion
References
This topology was exciting and surprisingly simple to implement. It has been shown that the HIP4081A can be used
successfully to realize the phase shift ZVS full-bridge topology. Not only does the HIP4081A drive the H-bridge but it
also is capable of delivering the needed ZVS transition delay
times required by this topology. In addition, a simple logic
block was used to convert a single ended PWM output into
the required phase shift logic drive signals.
Guicho Hua, Fred Lee, Milan Jovanovic, An Improved FullBridge Zero-Voltage-Switched PWM Controller Using a Saturating Inductor, IEEE Transactions on Power Electronics,
October 1993.
J.A. Sabate, V.Valtkovic, R.B. Ridley, F.C. Lee, B.I. Cho, Design
Considerations For High-Voltage High-Power Full-Bridge ZeroVoltage-Switched PWM Converter, IEEE APEC 1990
Dhaval B. Dalal, A 500kHz Multi-Output Converter with Zero
Voltage Switching, IEEE 1990
Abraham Pressman, Switching Power Supply Design,
McGraw Hill, 1991
M.M. Walters, W.M. Polivka, Extending The Range of SoftSwitching In Resonant-Transition DC-DC Converters, International Telecommunications Energy Conference, October 1992.
Bill Andreycak, Designing a Phase Shifted Zero Voltage Transition Power Converter, Unitrode SEM-900 Power Supply
Design Seminar Handbook.
Edwin Oxner, Power FETS And Their Applications, PrenticeHall, 1982
George Danz, HIP4081, 80V High Frequency H-Bridge Driver,
Intersil Application Note, Publication # AN9325
A design process was developed to enable designers to
accomplish their own designs. This was achieved by deriving essential equations and exposing key concepts. Following these procedures should allow designers to obtain
success when incorporating this topology. What’s more,
much of the mystery of this topology has been removed,
especially in the area of parasitic functionality within the
design. It was shown that the output capacitance variation
does not have a great impact on overall performance. It was
shown that the effects of this variation can be determined
easily with graphical methods.
13
Application Note 9506
HIP4081A Converter
14
Application Note 9506
Material List
1
2
3
4
1
2
1
7
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
2
1
2
2
1
1
3
1
1
1
2
1
1
1
1
1
1
1
1
1
4
1
1
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
2
1
1
1
1
1
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
5
C3
C11, C4
C5
C6, C7, C8, C17,
C20, C23, C29
C15, C14
C16
C18, C34
C19, C32
C21
C22
C25, C26, C30
C28
D1
D2
D3, D4
D8
D15
D18
IS01
J1
J2
L1
L2
L3
Q1, Q2, Q3, Q4
Q5
Q6
Q7
R1
R2
R3
R4
R5
R21, R7
R11, R12, R13,
R14, R34
R23, R22
R24
R33
R35
R38
R39
R40, R41
R42
R43
R44
R45
T1
T2
U1
U2
U4
U5
U6
Appendix A
100µF
4µF
0.47µF
0.1µF
Derivation of Equation 1
10V
100V
10V
50V
The equation for resonant frequency:
1
Fr = ------------------2π LC
0.1µF
100V
33µF
20V
1µF
20V
470pF
50V
47µF
20V
0.22µF
50V
2200pF
50V
10µF
20V
BAV70LT1
BZX84C75LT1
MBRS1100T3
MBRB2535CTL
BZX84C12LT1
1N4148
NEC PS2701-1
MOLEX 22-59-1310
MOLEX 22-59-1310
4µH
MICROMETALS T50-8
2.0µH
MICROMETALS T50-6
470µH
COILCRAFT DT1608
IRFR120 INTERSIL
BP720T1
MMBT5401LT1
MMBT3904LT1
39
15K
1%
5.11K
1%
4.99K
1%
619K
1%
10K
10
The left leg transition takes place within a period of 1/4 the
resonant period.
1
π
t LL = --- ( 2π LC ) = --- LC
4
2
Derivation of Equation 2
The output capacity COSS is a depletion-dependant capacity
whose value depends upon the impressed drain-to-source
voltage. Therefore the drain-to-source capacitance value over
varying drain-to-source voltages can be approximated as:
 V OSS  n
C DS ( V DS, n ) = C OSS  ---------------- 
 V DS 
Where n is between 1/2 and 1/3 for most MOSFETs, COSS is
the measured output capacitance at a drain to source voltage
VOSS. Most manufacturers measure this value at VOSS = 25V.
Derive Energy and substitute for current:
E =
∫ ( v × i ) dt
dQ
i = -------dt
E =
∫ v dQ
Capacitance is a function of drain-to-source voltage:
C ( V DS ) =
dQ
d V DS
E ( V DS, n ) =
∫ VDS C ( VDS, n) dVDS
n
∫
1–n
E ( V DS, n ) = C OSS V OSS V DS dV DS
Integrating:
10K
1%
6.49K
1%
33K
100
1W
47K
2.2K
100K
20K
1K
0.470
2W
4.7K
EPC-19
MICROMETALS T37-8
HIP4081A
INTERSIL
UC39432
UC3823A
74ACT86
74ACT74
n
2–n
C OSS × V OSS × V DS
E ( V DS, n ) = -----------------------------------------------------------2–n
Evaluate at n = 1/2 for the INTERSIL IRFR120:
 1--  3--- 
 2
 2 
2


E =  --- C OSS   V OSS   V DS 
3



Notice that the output capacitance is multiplied by the factor
of 2/3. For the phase-shift ZVS topology there are two output
capacitances (COSS) in parallel during each resonant transition. Therefore the effective energy and resonant capacitance is multiplied by 2 becoming:
 1--  3--- 
 2
 2 
4
E =  --- C OSS   V OSS   V DS 
3



4
C R = --- C OSS
3
15
Application Note 9506
Derivation of Equation 10
Determine switched VDS for tLL Skew:
∆tLL tSW
When the energy in the resonant inductor is equal to the
energy in the resonant capacitance the critical point of ZVS
operation is reached. When energy in the resonant inductor
is below the stored MOSFET capacitance energy there is
Non-ZVS operation. When it is above there is ZVS operation. Therefore:
VIN
1
3
----2
2
2
1
C R × V OSS × V DS = --- × L R × I PRI
2
VMIN
Where tSW is the MOSFET switching time during the nonzvs portion of the waveform. This can be caused by a delay
too long as shown or by a delay too short.
VDS is essentially VIN,
where VIN is the maximum input voltage.
I PRI ( Critical ) =
1
3
----2
2
2 × C R × V OSS × V IN ( MAX )
 V MIN 
1
∆t LL = ----------------------- × arc sin  -------------- 
2πF RES
 V IN 
--------------------------------------------------------------------------LR
Derivation of Equation 11
V MIN
sin ( ω RES × ∆t LL ) = -------------V IN
The energy of the resonant capacitance CR can be displaced by the energy in the resonant inductance LR. When
this happens the capacitive turn on loss is equal to zero:
V SWLL ( C OSS, V IN ) = V IN × sin ( ω RES × ∆t LL ( C OSS ) )
1
3
----2
2
2
1
C R × V OSS × V DS – --- × L R × I PRI = 0
2
Determine switched VDS for tRL Skew:
From Equation 3:
Since VDS is essentially VIN, the power required to displace
the resonant capacitance is:
∆t RL ( C OSS ) × I PRICRIT
V SWRL ( C OSS ) = --------------------------------------------------------------------
1
3


----
2
2
2 
1
P SWLOSS =  C R × V OSS × V IN – --- × L R × I PRI  × F XFMR
2


C
Combining:
1
3
----2
2 1
2
V
×
– --- × L × I
C ×V
OSS
IN
R
R
PRI
2
P
= ---------------------------------------------------------------------------------------------------

SWLOSS C
,V
t CLK
 0SS IN 
1
F XFMR = ---------------2t CLK
Substituting and multiplying by two since there are two such
transitions per period:
P SWLOSS =
R
I PRI × V SWLL  C OSS, V IN  × t SWLL
+ -------------------------------------------------------------------------------------------------------------t CLK
1
3
----2
2
2
1
C R × V OSS × V IN – --- × L R × I PRI
2
---------------------------------------------------------------------------------------- × 2
2t CLK
I PRI × V SWRL  C OSS  × t SWRL
+ -------------------------------------------------------------------------------------------------t CLK
1
3
----2
2
2
1
C R × V OSS × V IN – --- × L R × I PRI
2
=
---------------------------------------------------------------------------------------t CLK
From the ZVS Transition times:
π
∆t LL ( C OSS ) =  --- × C R ( C OSS ) × L R  – t LL
2
 C R ( C OSS ) × 72 
∆t RL ( C OSS ) =  -------------------------------------------------  – t RL
 I PRICRIT ( C OSS ) 
All Intersil semiconductor products are manufactured, assembled and tested under ISO9000 quality systems certification.
Intersil products are sold by description only. Intersil Corporation reserves the right to make changes in circuit design and/or specifications at any time without
notice. Accordingly, the reader is cautioned to verify that data sheets are current before placing orders. Information furnished by Intersil is believed to be accurate
and reliable. However, no responsibility is assumed by Intersil or its subsidiaries for its use; nor for any infringements of patents or other rights of third parties which
may result from its use. No license is granted by implication or otherwise under any patent or patent rights of Intersil or its subsidiaries.
For information regarding Intersil Corporation and its products, see web site http://www.intersil.com
16