We propose an efficient high voltage level shifter for a high side Buck converter driving a lightemitting diode (LED) lamp. The proposed circuit is comprised of a low voltage pulse width modulation (PWM) signal driver, a coupling capacitor, a resistor, and a diode. The proposed method uses a property of a PWM signal. The property is that the signal repeatedly transits between a low and high level at a certain frequency. A low voltage PWM signal is boosted to a high voltage PWM signal through a coupling capacitor using the property of the PWM signal, and the boosted high voltage PWM signal drives a pchannel metal oxide semiconductor (PMOS) transistor on the high side Buck converter. Experimental results show that the proposed level shifter boosts a low voltage (0 to 20 V) PWM signal at 125 kHz to a high voltage (370 to 380 V) PWM signal with a duty ratio of up to 0.9941.
1. Introduction
Direct current (DC) power distribution is an effective solution considering the properties of DCpreferred loads. Many household appliances and office machines such as digital TVs, computers, facsimiles, copy machines, and solidstate lighting are DCpreferred loads. DC distribution has no alternating current (AC) losses and no reactive power issues
[1
,
2]
. DC distribution also helps eliminate the power factor correction (PFC) stages in the appliances and machines
[2]
.
Energy management in smart buildings with DC distribution that supplies power to DC electrical appliances such as lighting has recently attracted increasing interest from the research community
[3]
. Integrated lighting controls for demandside energy management in a building can significantly improve energy efficiency, and enhance occupant comfort and satisfaction with the built environment
[3]
. In
[4]
, a DClevel dimmer supplies variable DC voltage to ballasts powering light emitting diode (LED) lamps, as shown in
Fig. 1(a)
. The DClevel dimmer sets its output voltage according to a dimming level, and ballasts adjust the brightness of their LED lamps according to the level of the DC voltage. The simplicity of this configuration, which requires no communication lines between the dimmer and ballasts, is a strong point. But this system is not appropriate for implementing a smart lighting system due to the lack of a bidirectional communication channel between the dimmer and the ballasts. In
[3
,
5]
, wireless or wired communication is applied to an intelligent smart lighting system.
Fig. 1(b)
shows a communication channel between the main controller and ballasts. The main controller can control the dimming level of the lamps through a communication channel, and supply optimal DC voltage to the ballasts using the information collected from the ballasts. The system shown in
Fig. 1(b)
was used in this study.
(a) DClevel dimming system [4], (b) smart lighting system with variable DC voltage and dedicated communication channel, and (c) high side Buck converter.
A conventional 20 W LED tube that is retrofittable for a fluorescent tube uses 132 LEDs when 0.2 W LEDs are used. In the LED tube, 12 LED strings are connected in parallel, with 11 LEDs connected in series in an LED string
[6
,
7]
. There are deviations of the currents among the 12 LED strings connected in parallel, which gives deviations of the light intensities of the strings
[7]
. In this study, we adopted a high side Buck converter as shown in
Fig. 1(c)
. The converter drives an LED string where 100 LEDs are connected in series. In this case, the turnon voltage of the LED string is about 300 V. As the LEDs are connected in series, the current flowing into each LED is the same, and the intensity of each LED is the same.
There are two options to implement the switch in
Fig. 1(c)
. One is to use an NMOS transistor, and the other is to use a PMOS transistor. An NMOS transistor is appropriate for a highpower and highefficiency Buck converter because the onresistance of the transistor is lower than that of a PMOS transistor. However, the driving circuit for the NMOS transistor is more complex than that for the PMOS transistor because the source voltage of the NMOS transistor is not fixed
[8
,
9]
. The onresistance of a PMOS transistor is relatively high compared with that of an NMOS transistor, but the driving circuit for the PMOS transistor is relatively simple compared with that for the NMOS transistor because the source of the PMOS transistor is connected to a constant voltage. Thus, a PMOS transistor is appropriate for a lowpower and lowcost Buck converter, especially when the output current of the converter is low. In this study, we chose the PMOS transistor because the Buck converter drives a 20 W LED lamp, and the output current of the converter is less than 100 mA.
In a Buck converter, the current flowing into an LED string can be controlled to reach a target by increasing the duty ratio of a pulse width modulation (PWM) control signal while appropriately decreasing the input voltage of the converter if the input voltage is controllable. In this case, the efficiency of the Buck converter can be increased. The high side Buck converter shown in
Fig. 1(c)
can exactly measure and control the current flowing into an LED string even if the input voltage is widely changed
[10]
. However, this structure requires a high side switch. Therefore, we must convert a low voltage PWM signal into a high voltage PWM signal. Many high voltage level shifter circuits have been proposed. The previous works were usually implemented using high voltage semiconductor process technology. The manufacturing cost of a level shifter usually increases as the target voltage of the level shifter is increased. Thus, the previous works are not appropriate for application in a low cost ballast circuit driving an LED sting when the target voltage of the level shifter is high. The previous works are described in Section 2.
We propose an efficient high voltage level shifter circuit for a high side Buck converter that drives an LED string. The proposed method uses the property of a PWM signal that repeatedly transits between a low and high level at a certain frequency. The circuit boosts a low voltage PWM signal to a high voltage PWM signal through a coupling capacitor using the property of the PWM signal. The high voltage PWM signal drives a pchannel metal oxide semiconductor (PMOS) transistor on the high side Buck converter. Although the proposed circuit consists of a low voltage PWM signal driver, a coupling capacitor, a resistor, and a diode, the circuit can provide enough high voltage PWM signal to drive the PMOS transistor. Experimental results show that the proposed level shifter was successfully applied to the high side Buck converter with an input voltage of 380 V, and boosts the low voltage (0 to 20 V) PWM signal at a frequency of 125 kHz to a high voltage (370 to 380 V) PWM signal with a duty ratio of up to 0.9941.
The remainder of this paper is organized as follows. Section 2 describes previous works. Section 3 explains the proposed circuit and its behavior. Section 4 describes our experimental results. Our conclusions are given in Section 5.
2. Previous Works
Fig. 2(a)
shows a simple level shifter whose output is determined by the ratio of the two resistors and
V_{DDH}
. The level shifter consumes static power when the input is high level, and the output driving capability is low due to the two resistors. This type of level shifter is usually used in low cost and low
V_{DDH}
applications.
(a) Simple level shifter; (b) latchtype CMOS level shifter; (c) DMOS cascade high voltage level shifter; (d) high voltage level shifter using coupling capacitors, and (e) pulse transformer coupled gate driver with DC restore circuit.
Fig. 2(b)
shows a latchtype complementary MOS (CMOS) level shifter. The circuit has a crosscoupled PMOS latch on the top, and converts a low voltage input signal to a high voltage output signal by driving two nchannel MOS (NMOS) transistors on the bottom
[11]
. The level shifter has no static power consumption and a high driving capability, and can be integrated into a microcontroller unit (MCU) and other controllers based on CMOS semiconductor process technology. In a CMOS process technology, the maximum allowed source to gate voltage and source to drain voltage of PMOS are less than two times of
V_{DDL}
, respectively. This level shifter circuit is limited in its ability to increase voltage
V_{DDH}
because the voltage between the source and the gate of the PMOS can reach to
V_{DDH}
, and the voltage between the source and the drain of the PMOS can also reach to
V_{DDH}
. Improved versions of this type of level shifter have been reported, wherein
V_{DDH}
is about two to five times higher than
V_{DDL}
[12
,
13]
.
Fig. 2(c)
shows a high voltage version of
Fig. 2(b)
that uses a drainextended MOS (DMOS). The structure of a DMOS is similar to that of a MOS except the drain of the DMOS is extended. The maximum allowed voltage between the gate and the source of the DMOS is the same as that of a conventional MOS, but high voltage can be applied across the drain and the source of the DMOS. In
Fig. 2(c)
, two nchannel DMOS (NDMOS) devices protect NMOS M
_{3}
and M
_{4}
. In the same way, two pchannel DMOS (PDMOS) devices protect PMOS M
_{1}
and M
_{2}
, thus M
_{1}
and M
_{2}
are placed between
V_{SSH}
and
V_{DDH}
[11
,
14]
. Many improved versions of the level shifter shown in
Fig. 2(b)
have been reported
[11
,
15
,
16]
. However, the manufacturing cost of a level shifter using DMOS semiconductor process technology is higher than that of the level shifter using CMOS semiconductor process technology. Furthermore, the level shifter based on DMOS technology is not appropriate for integration into an MCU and other controllers based on CMOS technology. Thus, this type of level shifter is used as a standalone offchip device.
Fig. 2(d)
shows a level shifter circuit using coupling capacitors, wherein the initialization circuit for the drain nodes of the two PMOS transistors is omitted. The level shifter is an appropriate way to increase
V_{DDH}
, because most of the value of
V_{DDH}
exists across the coupling capacitors. However, the level shifter requires an initialization circuit and a crosscoupled PMOS latch
[8
,
17]
. Thus, the whole circuit of the level shifter in
[8]
was implemented on a single chip including two power switches (not shown in
Fig. 2(d)
), the initialization circuit, the crosscoupled PMOS transistors, and the coupling capacitors (several pF). In many semiconductor processes, high voltage capacitors can be constructed only with normal routing metals, requiring large silicon area to obtain reasonable capacitance values and increasing the manufacturing cost
[11]
.
Fig. 2(e)
shows a pulse transformer coupled gate driver with DC restore circuit. The advantages of the method are that it does not require isolated power supplies to drive the secondaryside NMOS transistor and maintains electrical isolation between the control and gate drive electronics. However, a problem can occur when a large transient drive current flowing in the inductive coils causes ringing. This can switch the gate on and off when not intended and damage the NMOS transistor. The transformer can deliver only AC signals with maintaining a voltsecond balance, thus the transformer coupled gate driver needs a DC restore circuit to increase duty cycle capability.
Fig. 2(e)
shows a commonly used DC restore circuit that consists of
C
_{2}
and
D
_{2}
. The effective duty cycle range of the transformer coupled gate driver with the circuit is from 5 to 95%
[18]
.
3. Proposed Level Shifter
Fig. 3
shows the proposed level shifter circuit that consists of a low voltage pulse width modulation (PWM) signal driver, a coupling capacitor
C_{c}
, a resistor
R
_{1}
, and a diode
D
_{1}
. Capacitance
C_{g}
is the equivalent gate capacitance of the PMOS transistor. During the poweron period, the gate voltage of the PMOS is charged to
V_{DDH}
through resistor
R
_{1}
. If the voltage of the PWM control signal
v_{c}
(
t
) changes from
V_{DDL}
to zero, the gate voltage changes from
V_{DDH}
to
V_{DDH}
– (
C_{c}
/
C_{A}
)
V_{DDL}
and is then continuously increased due to the current flowing through
R
_{1}
, where
C_{A}
=
C_{c}
+
C_{g}
. Then, if the voltage of
v_{c}
(
t
) changes from zero to
V_{DDL}
, the gate voltage changes to
V_{DDH}
+
V_{D}
and then slightly decreases toward
V_{DDH}
due to the voltage difference across resistor
R
_{1}
, where
V_{D}
is the forward bias voltage of the diode.
Proposed level shifter circuit.
Fig. 4(a)
shows the timing of the control signal
v_{c}
(
t
), whose duration is
T
and duty ratio is
T
_{1}
/
T
. When the control signal shown in
Fig. 4(a)
is applied to the proposed circuit in
Fig. 3
, the voltage of the gate and the voltage difference between the gate and the source of the PMOS are as shown in
Figs. 4(b)
and
(c)
, respectively. As 0 ≤
T
_{1}
<
T
, the maximum value of
T
_{1}
goes toward
T
. Eq. (1) can be obtained by applying Kirchhoff’s current law (KCL) to the gate node, where
v_{G}
is the gate voltage of the PMOS.
(a) Low voltage PWM control signal v_{c}(t), (b) the gate voltage of the PMOS, and (c) the voltage difference between the gate and the source of the PMOS, where k is defined as the deviation ratio.
When the control signal
v_{c}
(
t
) transits from
V_{DDL}
to zero at
t
= 0 as shown in
Fig. 4(a)
(i.e.,
v_{c}
(
t
) =
V_{DDL}
u
(–
t
)), the gate voltage at
t
= 0
^{+}
becomes
v_{G}
(
t
= 0
^{+}
) =
V_{DDH}
+
V_{D}
 (
C_{c}
/
C_{A}
)
V_{DDL}
and the solution of Eq. (1) is given by Eq. (2) for 0 ≤
t
<
T
_{1}
, where
τ
=
R
_{1}
C_{A}
and
C_{A}
=
C_{c}
+
C_{g}
.
Thus, the voltages of
V
_{1}
,
V
_{2}
,
V_{A}
, and
V_{B}
in
Fig. 4
are expressed as Eqs. (3), (4), (5), and (6), respectively:
We define the deviation ratio
k
of the gate voltage as Eq. (7), which is depicted in
Fig. 4(c)
. By applying Eqs. (5) and (6) to Eq. (7), we obtain Eqs. (8) and (9). Eq. (9) shows that the time constant of the proposed circuit depends on the deviation ratio
k
.
In a PMOS transistor,
V_{T}
means the threshold voltage at which the channel of the transistor starts to be created.
V_{GS}
_{(TYP)}
means a typical turnon voltage of the transistor, at which the transistor is in a sufficiently turnon state. The manufacturer of the transistor usually uses
V_{GS}
_{(TYP)}
to represent the performance of the transistor on the data sheet.
V_{GS}
_{(MIN)}
is the minimum allowed voltage of
V_{GS}
that can be applied to the transistor. The relation among the three voltages is given by
V_{GS}
_{(MIN)}
<
V_{GS}
_{(TYP)}
<
V_{T}
< 0. In Eq. (5), if we set
V_{A}
=
V_{GS}
_{(TYP)}
, then we can obtain Eq. (10):
By using Eqs. (8) and (10), we can obtain the capacitance of
C_{c}
as shown by Eq. (11). We can also obtain the resistance of
R
_{1}
as given by Eq. (12) by using Eqs. (9) and (11). Thus,
Now, we calculate the power consumption of
R
_{1}
and
D
_{1}
. From Eqs. (2) and (6), the voltage difference between the source and the gate of the PMOS transistor is expressed as
v_{SG}
(
t
) = −
V_{B}e
^{−t/τ}
in 0 ≤
t
<
T
_{1}
. Thus, the average power consumption
P
_{R1}
of resistor
R
_{1}
is expressed as Eq. (13).
The diode
D
_{1}
can be approximated as a piecewiselinear model, as shown in
Fig. 5
. As the diode is turned on during
T
_{1}
≤
t
<
T
, the average power consumption
P
_{D1}
of the diode is expressed as Eq. (14).
Approximation of diode D_{1} with piecewiselinear model.
By applying KCL to the gate node, we can express the gate node voltage as Eq. (15), where
τ_{d}
=
C_{A}
r_{d}
. We assume that the current flowing through
R
_{1}
is negligible because the voltage difference across the resistor is about
V_{D}
when the diode is turned on.
Thus, the diode current
i_{d}
(
t
+
T
_{1}
) is expressed as Eq. (16).
The voltage difference
V_{A}

V_{B}
is due to the current flowing through
R
_{1}
during 0 ≤
t
<
T
_{1}
, and is expressed as Eq. (17).
Now, we can solve Eq. (14) by applying Eqs. (16) and (17) to Eq. (14). The solution is expressed as Eq. (18).
The PWM signal driver also consumes power because it drives a capacitive load
C_{L}
, where
C_{L}
≈
C_{c}
C_{g}
/ (
C_{c}
+
C_{g}
) if
R
_{1}
is large enough for its effect on the circuit to be negligible. If (
V_{A}
−
V_{B}
) / (−
V_{B}
) << 1 is satisfied in
Fig. 4(c)
, we can say the resistor gives little effect on the circuit. Because (
V_{A}
−
V_{B}
) is expressed as Eq. (17), (
V_{A}
−
V_{B}
) / (−
V_{B}
) <
T
/
τ
is satisfied, and we can use
T
/
τ
<< 1 instead of (
V_{A}
−
V_{B}
) / (−
V_{B}
) << 1. Thus the power consumption of the driver is expressed as Eq. (19) if
T
/
τ
<< 1. On the other hand,
C_{L}
is expressed as
C_{L}
≈
C_{c}
if
T
/
τ
>> 1.
During the poweron period, the PMOS transistor should be turned off. To satisfy this condition, the voltage between the gate and the source of the PMOS transistor should be larger than
V_{T}
; i.e.,
v_{GS}
(
t
) >
V_{T}
. In
Fig. 1(b)
, if the output voltage of the DCDC converter is programmable, the converter can also control the rise time of the output voltage. In this case, we can turn the PMOS off during the poweron period. Now, we will explain a necessary condition to turn off the PMOS during the poweron period if the output voltage of the DCDC converter is not programmable.
Fig. 6
shows the timing diagram of supply voltage
v_{DDH}
(
t
) during the poweron period where
V_{DDH}
is the steadystate value of the supply voltage and
t_{r}
is the rising time of the supply voltage. In this analysis, we assumed that the voltage of the control signal
v_{c}
(
t
) is zero during the poweron period. By applying KCL to the gate node of the PMOS, we obtain Eq. (20):
Supply voltage during the poweron period.
In the poweron period (i.e.,
t
≤
t_{r}
), the supply voltage is expressed as
v_{DDH}
(
t
) = (
V_{DDH}
/
t_{r}
)
t
. Thus, the voltage of the gate node during the poweron period is expressed as Eq. (21) by solving Eq. (20):
Thus,
v_{GS}
(
t
) is expressed as Eq. (22) using Eq. (21):
Eq. (22) has minimum value when
t
=
t_{r}
for
t
≤
t_{r}
. Thus,
v_{GS}
(
t_{r}
) should be larger than
V_{T}
; i.e., Eq. (23) should be satisfied in order to turn off the PMOS transistor during the poweron period.
We can obtain the approximate rise time
t_{r}
that satisfies Eq. (23). If
t_{r}
/
τ
has large value,
becomes very small compared to 1, and Eq. (23) is arranged as follows:
Eq. (24) can be arranged as
and
if
V_{DDH}
=380 V and
V_{T}
= –3 V. Thus, the assumption that
t_{r}
/
τ
has a large value and
becomes significantly small compared to 1 is valid if
V_{DDH}
is relatively high compared to –
V_{T}
. Eq. (24) is arranged as Eq. (25) using Eq. (9). During the poweron period, Eq. (25) should be satisfied in order to turn off the PMOS transistor:
In the analysis of the proposed circuit, we used parameter
T
_{1}
. The range of
T
_{1}
is 0 ≤
T
_{1}
<
T
, and the maximum value of
T
_{1}
goes toward
T
. Thus
T
is used instead of
T
_{1}
when designing the proposed circuit, i.e.,
T
is used instead of
T
_{1}
when determining the values of resistor
R
_{1}
and the coupling capacitor
C_{c}
. If the frequency of the PWM signal is 125 kHz (
T
= 8 µs),
V_{DDH}
is 380 V,
V_{T}
is –3 V,
C_{g}
is 0.55 nF,
k
is 10%,
V
_{GS(TYP)}
is –10 V,
V_{DDL}
is 20 V, and
V_{D}
is 0.7 V, then we can obtain
C_{c}
= 0.78 nF and
R
_{1}
= 63.3 KΩ by using Eqs. (11) and (12). From Eq. (13) and (18), the power consumption of the resistor
R
_{1}
and that of the diode
D
_{1}
are expressed as
P
_{R1}
< 1.92 mW and
P
_{D1}
< 0.21 mW, respectively. The time constant (
τ
= 84.2 us) is about 10 times larger than the duration (
T
= 8 us), thus we can use Eq. (19) to calculate the power consumption of the PWM signal driver as
P_{PWM}
≈ 16.1 mW.
P_{PWM}
is a major part of the total power consumption of the proposed circuit, and the sum of
P
_{R1}
and
P
_{D1}
represents only about 12 % of the total power consumption. The rise time of the high voltage
v_{DDH}
(
t
) during the poweron period should be greater than 10.6 ms based on Eq. (25).
Fig. 7
shows the simulation results of the proposed circuit using the parameters selected above when the duty ratio is 0.5. The wave form of the voltage of the gate node is quite similar to the wave form shown in
Fig. 4(b)
.
Simulation results of the voltage of the gate node.
Until now, we assumed that a resistor and capacitor have fixed values, but the values vary with the temperature and tolerance. We set the error rates of the resistance and capacitance as
p
and
q
, respectively. When the resistance
R
_{1}
and capacitance
C_{c}
are respectively changed to
R
_{1}
(1+
p
) and
C_{c}
(1+
q
), the deviation ratio
k
is also changed according to Eq. (26) that is derived from Eq. (9), where
τ
' = (1 +
p
)
R
_{1}
(
C_{g}
+ (1 +
q
)
C_{c}
) = (1 +
p
) (
τ
+
qR
_{1}
C_{c}
).
Table 1
shows the variation of
k
when
p
and
q
are within ±20% using
T
= 8 μs,
R
_{1}
= 63.3 KΩ,
C_{g}
= 0.55 nF, and
C_{c}
= 0.78 nF. The values of the parameters were used in the previous design example. When there are no errors in the resistance and capacitance, (i.e.,
p
=
q
= 0%), the deviation ratio
k
is 10% that is the same as the value set in the previous example. In this case,
V_{GS}
(
t
) <
V
_{GS(TYP)}
is satisfied in 0 ≤
t
<
T
. When
k
is less than 10%, (i.e., τ' > τ), the voltage of the gate node is slowly charged compared with the case of
k
=10%. In this case,
V_{GS}
(
t
) <
V
_{GS(TYP)}
is also satisfied in 0 ≤
t
<
T
. However, the voltage of the gate node is rapidly charged when
k
> 10% compared with
k
= 10%, because τ' < τ. In this case, there are some ranges where
V_{GS}
(
t
) <
V
_{GS(TYP)}
is not satisfied and the onresistance of the PMOS transistor is increased. We need a new design guideline that ensures the proposed circuit works correctly even though
k
is varied, i.e., we have to choose
R
_{1}
and
C_{c}
such that the time constant is larger than τ in the worst case.
Variation ofkwith change ofpandq
Variation of k with change of p and q
Now, we set
as the adjusted resistance of
R
_{1}
. If
p
is within ±
p_{MAX}
, then
(1 –
p_{MAX}
) =
R
_{1}
should be satisfied to ensure
R
_{1}
' ≥
R
_{1}
. Thus,
R
_{1}
' is expressed as Eq. (27).
In the same way, we can obtain the adjusted capacitance
C_{c}
' as Eq. (28), where
q
is within ±
q_{MAX}
.
The adjusted resistance
R
_{1}
' and capacitance
C_{c}
' are used instead of
R
_{1}
and
C_{c}
.
4. Experiments
The proposed level shifter was applied to the high side Buck converter as shown in
Fig. 3
, where an inductor
L
of 82 mH and a capacitor
C
_{1}
of 0.5 μF are used, and a Fairchild FQD2P40 PMOS transistor is selected. The specification of the PMOS is as follows: the maximum gate capacitance
C
_{g(max)}
= 0.55 nF, the maximum threshold voltage
V
_{T(max)}
= −3 V, the minimum threshold voltage
V
_{T(min)}
= −5 V, the typical turnon voltage
V
_{GS(TYP)}
= −10 V, and the minimum allowed gatetosource voltage
V
_{GS(MIN)}
= −30 V. In this design, we set the parameter
k
= 10%,
V_{DDH}
= 380 V, and
V_{DDL}
= 20 V. With these parameters, the capacitance
C_{c}
and the resistance
R
_{1}
were determined to be
C_{c}
= 0.78 nF and
R
_{1}
= 63.3 kΩ as described in Section 3. The resistor used in this experiment has a tolerance of ±1% and temperature coefficient of ±100 ppm/℃, so we set the maximum error rate of the resistor as
p_{MAX}
= 2%. The capacitor used in this experiment (ECWU type from Panasonic) has a tolerance of ±5% and capacitance variation of ±1% in the range of −30 ℃ − 80 ℃, so we set the maximum error rate of the capacitor as
q_{MAX}
= 6%. The adjusted capacitance and resistance from Eq. (28) and (27) are
C_{c}
' = 0.83 nF and
R
_{1}
' = 64.6 KΩ, respectively. In this experiment, we selected available capacitor and resistor as
C_{c}
' = 1 nF and
R
_{1}
' = 68 kΩ. In this case, the voltages of
V_{A}
and
V_{B}
were calculated to be
V_{A}
= −11.3 V and
V_{B}
= −12.2 V based on Eqs. (5) and (6), respectively.
Fig. 8
shows the experimental configuration. An N5752A high voltage power supply (Agilent Technology) was used to generate 380 V, and IWSS5552 LEDs (Itswell) were used in the LED string where 100 LEDs are connected in series.
Fig. 9
shows the implemented circuit board whose size is 90 mm × 58 mm.
Experimental configuration.
Implementation of the proposed circuit.
In the first experiment, we measured the levelshifted high voltage PWM signal without connecting the LED string; i.e., the current flowing into the LED string was zero, and the drain node was floated.
Fig. 10
shows the voltage of the PWM control signal
v_{c}
(
t
) and the voltage of the gate node signal that is the levelshifted high voltage PWM signal. The frequency of the control signal is 125 kHz and the duty ratio of the signal is 0.762.
Low voltage PWM signal and levelshifted high voltage PWM signal without connecting LED string.
The theoretical voltage of
V
_{1}
is 368.5 V from Eq. (3). Our experimental results show that
V
_{1}
is 367.4 V (
Fig. 10
). There is a voltage difference of 1.1 V between the theoretical voltage and the experimental voltage. When a MOS transistor operates in linear mode,
C_{gs}
and
C_{gd}
have the same capacitance of
C
_{g(max)}
/ 2 if the source and the drain are connected at the same potential
[19]
. In this experiment, the PMOS transistor was operated in linear mode when the transistor was in the turnon state, but
C_{gd}
was less than
C
_{g(max)}
/ 2 because the drain node was floated. Thus, the gate capacitance was less than
C
_{g(max)}
, which gave a lower value of
V
_{1}
than the theoretical voltage.
In the second experiment, we measured the levelshifted PWM signal with the connecting LED string.
Fig. 11
shows the voltage of the gate node that was measured under the same conditions as the first experiment, except for the connecting LED string. In
Fig. 11
,
V
_{1}
is 368.5 V, which is the same as the theoretical value calculated by using Eq. (3). The voltage
V_{A}
= −11.5 V (
V_{A}
=
V
_{1}
−
V_{DDH}
) was lower than
V
_{GS(TYP)}
. Thus, the PMOS transistor was in a sufficiently turnon state at low level of the gate node voltage.
Low voltage PWM signal and levelshifted high voltage PWM signal with connecting LED string.
When the control signal was increased from 0 to 20 V, the voltage of the gate node became 381 V, which is about
V_{DDH}
plus
V_{D}
. Then, the voltage of the gate node decreased by more than the expected value and became 378.6 V. The phenomenon is due to the voltage change of the PMOS drain node from 380 to 0 V with the small capacitance of several pF between the gate and the drain when the PMOS transistor is turned off. However, the lowest voltage of the gate node is 378.6 V at high level, and this gives
V_{GS}
= −1.4 V, which is higher than the worstcase threshold voltage
V
_{T(max)}
= −3 V. Thus, the PMOS transistor kept the turnoff state at high level of the gate node voltage. This phenomenon can be reduced by connecting one or two diodes with diode
D
_{1}
in series, or by increasing the capacitance of
C_{c}
.
Fig. 12
shows the current of the LED string, which is about 80 mA.
Current flowing into the LED string.
The proposed method uses the property of a PWM signal, which transits between a low and high level repeatedly at a frequency. Theoretically, the proposed circuit can work in 0 ≤
T
_{1}
<
T
. In the third experiment, we measured the maximum duty ratio wherein the proposed circuit works correctly. We then define that
V
_{A(with)}
=
V_{A}
when the LED string is connected, and
V
_{A(without)}
=
V_{A}
when the LED string is not connected. If
V
_{A(with)}
≤
V
_{GS(TYP)}
is satisfied, the circuit works correctly at the duty ratio because the PMOS transistor is sufficiently switched on for 0 ≤
t
<
T
_{1}
and off for
T
_{1}
≤
t
<
T
. However, it is quite difficult to increase the duty ratio in the high side Buck converter when the LED string is connected, because the current flowing into the LED string is exponentially increased as the duty ratio is increased.
In this experiment, we used
V
_{A(without)}
instead of
V
_{A(with)}
. However, there is a voltage difference between
V
_{A(with)}
and VA(without), and the difference is 1.1 V at a duty ratio of 0.762 from the first and second experiments shown in
Figs. 10
and
11
; i.e.,
V
_{A(with)}
=
V
_{A(without)}
+1.1 V. By considering the voltage difference, we approximately determined that the proposed circuit works correctly at the given duty ratio if
V
_{A(without)}
+1.1 ≤
V
_{GS(TYP)}
is satisfied.
Fig. 13
shows the high voltage PWM signal at a duty ratio of 0.9941.
V
_{A(without)}
is − 11.6 V (=368.4−380 V) that satisfies
V
_{A(without)}
+ 1.1 ≤
V
_{GS(TYP)}
. Thus, the proposed circuit works correctly at a duty ratio of 0.9941. However,
V
_{A(without)}
is −10.45 at a duty ratio of 0.9961, thus the proposed circuit does not work correctly at a duty ratio of 0.9961 because
V
_{A(without)}
+ 1.1 ≤
V
_{GS(TYP)}
is not satisfied. Thus, the proposed circuit works correctly up to a duty ratio of 0.9941.
High voltage PWM signal without the connecting LED string at a duty ratio of 0.9941.
Now we will compare the proposed method with the gate driver using pulse transformer in
Fig. 2(e)
. The gate driver using pulse transformer addressed in Section 2 has the advantages that the driver can drive an NMOS transistor and it does not require isolated power supplies to drive the secondaryside NMOS transistor. However, the driver usually takes more components and requires the design of the transformer or at least the understanding of its operation and specification. The proposed method is very simple and easy to design, and works correctly up to a duty ratio of 99.5%, but the method is not applicable to an NMOS transistor. Another limitation of the proposed circuit is that the circuit may not work correctly if the frequency of the PWM signal applied to the circuit is lower than the target frequency.
5. Conclusion
We propose a high voltage level shifter using a coupling capacitor. Although the proposed circuit is a simple structure consisting of a low voltage PWM signal driver, a coupling capacitor, a resistor, and a diode, the proposed circuit can level a low voltage PWM signal up to high voltage PWM signal. The behavior of the level shifter was mathematically analyzed and a method to determine the capacitance of the coupling capacitor
C_{c}
and the resistance of the resistor
R
_{1}
is proposed. The proposed level shifter was applied to a high side Buck converter to drive the LED string. In the experiment, the input voltage of the converter was 380 V and the LED string consists of 100 LEDs connected in series. Our experimental results show that the proposed circuit converts a low voltage (0 to 20 V) PWM signal to a high voltage (370 to 380 V) PWM signal with a duty ratio of up to 0.9941.
Acknowledgements
This work was supported by a 2014 Yeungnam University research grant.
BIO
KwangSu Seong He received the B.S. degree in electronic engineering from Hanyang University, Seoul, Korea in 1990, and the M.S. and Ph.D. degrees in electrical engineering from the Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 1992 and 1997, respectively. He is currently an Associate Professor with the Department of Electronic Engineering, Yengnam University, Gyeongbuk, Korea. His current research interests include capacitive sensors, MCU design, and wireless power transmission for mobile devices.
Seo G. S.
,
Lee K. C.
,
Cho B. H.
2013
“A new DC antiislanding technique of electrolytic capacitorless photovoltaic interface in DC distribution systems,”
IEEE Trans. Power Electron
28
(4)
1632 
1641
DOI : 10.1109/TPEL.2012.2208226
Dong D.
,
Cvetkovic I.
,
Boroyevich D.
,
Zhang W.
,
Wang R.
,
Mattavelli P.
2013
“Gridinterface bidirectional converter for residential DC distribution systems –part one: highdensity twostate topology,”
IEEE Trans. Power Electron
28
(4)
1655 
1666
DOI : 10.1109/TPEL.2012.2212462
Tan Y. K.
,
Huynh T. P.
,
Wang Z.
2013
“Smart personal sensor network control for energy saving in DC grid powered LED lighting system,”
IEEE Trans. Smart Grid
4
(2)
669 
676
DOI : 10.1109/TSG.2012.2219887
Seo G. S.
,
Cho B. H.
,
Lee K. C.
“DC level dimmable LED driver using DC distribution,”
Proc. ECCE
1519 Sep. 2013
4650 
4654
Matta S.
,
Mahmud S. M.
“An intelligent light control system for power saving,”
Proc. IECON
710 Nov. 2010
3316 
3321
Jeong H. S.
,
Park C. K.
2011
“A study on optical losses for tubular LED lamp’s component,”
J Korean Inst Illum Electr Install Eng
25
(7)
1 
8
Yoon C. G.
,
Choi H. K.
,
Park J. Y.
,
Song Y. J.
,
Lee K. W.
,
Jung I. H.
“Analysis tubular LED lamps location of the difference in brightness,”
Proc. Autumn Annual Conference of the Korean Institute of Illuminating and Electrical Installation Engineers
23 Nov. 2011
98 
101
Liu Z.
,
Lee H.
“A 100V gate driver with subnanosecond delay capacitivecoupled level shifting and dynamic timing control for ZVSbased synchronous power converters,”
Proc. CICC
2225 Sep. 2013
Javadian V.
,
Kaboli S.
“Reliability assessment of some high side MOSFET drivers for Buck con verter,”
Proc. EPECS
24 Oct. 2013
1 
6
Oh I. H.
“An analysis of current accuracies in peak and hysteretic current controlled power LED drivers,”
Proc. APEC
2428 Feb. 2008
572 
577
Moghe Y.
,
Lehmann T.
,
Piessens T.
2011
“Nanosecond delay floating high voltage level shifters in a 0.35 um HVCMOS technology,”
IEEE J. SolidState Circuits
46
(2)
485 
497
DOI : 10.1109/JSSC.2010.2091322
Liu J.
,
Deng L.
,
Zaho J.
,
Liao H.
“A 1.8 V to 10 V CMOS level shifter for RFID transponders,”
Proc. ICSICT
14 Nov. 2010
491 
493
Serneels B.
,
Steyaert M.
,
Dehaene W.
“A high speed, low voltage to high voltage level shifter in standard 1.2V 0.13um CMOS,”
Proc. ICECS
1013 Dec. 2006
668 
671
Choi B.
2009
“Enhancement of current driving capability in data driver ICs for plasma display panels,”
IEEE Trans. Consumer Electron.
55
(3)
992 
997
DOI : 10.1109/TCE.2009.5278054
Wendt M.
,
Thoma L.
,
Wicht B.
,
SchmittLandsiedel D.
2008
“A configurable highside/lowside driver with fast and equalized switching delay,”
IEEE J. SolidState Circuits
43
(7)
1617 
1625
DOI : 10.1109/JSSC.2008.923734
Lehmann T.
2014
“Design of fast lowpower floating highvoltage levelshifters,”
Electronics letters
50
(3)
202 
204
DOI : 10.1049/el.2013.2270
Tan S. C.
,
Sun X. W.
2002
“Low power CMOS level shifters by bootstrapping technique,”
Electronics letters
38
(16)
876 
878
DOI : 10.1049/el:20020627
Rashid M. H.
2011
Power electronics handbook: devices, circuits, and applications handbook
3rd ed.
ButterworthHeinemann
Burlington, MA, USA
557 
558
Weste N.
,
Harris D.
2005
CMOS VLSI Design: A circuits and systems perspective
3rd ed.
Pearson Education
Boston, MA, USA
75 
80