Buck Based LED Driver
Buck Based LED Driver
Buck Based LED Driver
The first switching LED driver that we will study is the buck converter. The buck
converter is the simplest of the switching drivers, and is a step-down converter for
applications where the load voltage is never more than about 85% of the supply
voltage. The limit of about 85% is due to switching delays in the control system. In a
buck converter circuit, a power MOSFET is usually used to switch the supply voltage
across an inductor and LED load connected in series. The inductor is used to store
energy when the MOSFET is turned on; this energy is then used to provide current
for the LED when the MOSFET is turned off. A diode across the LED and inductor
circuit provides a return path for the current during the MOSFET off time. A simple
schematic is shown in Figure 5.1.
CIRCULATING CURRENT
(MOSFET OFF)
V+ SUPPLY
350 mA LED
D1
L1
CURRENT
(MOSFET ON)
CONTROLLER
Q1
Buck converters are an attractive choice for LED drivers in offline and in low voltage
applications as they can produce a constant LED current at very high efficiencies.
A peak-current-controlled buck converter can give reasonable LED current variation
over a wide range of input and LED voltages and needs no design effort in feedback
control design. Coupled with the fact that these converters can be designed to operate
at above 90% efficiencies, the buck-based driver becomes an attractive solution to
drive high brightness LEDs.
OSC Rosc
VDD
250 mV
–
CM S
+
Q
R
LD –
GATE
CM
+
CS
PWM_D
100 k
HV9910
GND
The HV9910B has two current sense threshold voltages – an internally set 250 mV
and an external voltage at the LD pin. The actual threshold voltage used during
switching will be the lower of the two. The low value of sense voltage allows the use of
low resistor values for the current sense, which means high efficiency.
The HV9910B IC operates down to 8 V input, which is required for some automobile
applications, and can accept a maximum of 450 V input, which makes it ideal for
offline applications. The IC has an internal regulator that supplies 7.5 V to power to
the IC’s internal circuits from the input voltage, eliminating the need for an external
low voltage power supply. The IC is capable of driving the external MOSFET
directly, without the need for additional driver circuitry.
10–30 V DC
C1
4.7 µF
350 mA LED
D1
10BQ060
1 L1
470 µH
6 VIN
VDD
C3
2,2 µF 10 V HV9910 8
RT
R1
5 100K
PWM_D
7 4 Q1
LD GATE VN3205N8
2
CS
3
GND
R2
0,62R
capacitors in parallel are needed to absorb the high frequency ripple current. The
required high frequency capacitance can be computed as
Io TOFF
C1 ¼
ð0:05 VminÞ
In this design example, the high frequency capacitance required is about 4.7 mF 50 V.
This capacitor should be located close to the inductor L1 and MOSFET switch
Q1, to keep the high frequency loop current within a small area on the PCB. In
practice, two such capacitors with a small inductor between them (to make a PI filter)
are needed to limit EMI emissions.
Vo,max TOFF
L1 ¼
0:3 Io,max
In this example, L1 = 380 mH and the nearest standard value is 470 mH. Since this value is
a little higher than the calculated value, the ripple current will be less than 30%.
The peak current rating of the inductor will be 350 mA plus 15% ripple:
The RMS current through the inductor will be the same as the average current
(i.e. 350 mA).
The maximum RMS current through the MOSFET depends on the maximum duty
cycle, which is 80% in our example. Hence, the current rating of the MOSFET is
IFET Io,max 0:8 ¼ 0:28 A:
Typically a MOSFET with about three times the current is chosen to minimize the
resistive losses in the switch. For this application, choose a 50 V, >1 A MOSFET;
a suitable device is a Supertex part, VN3205N8, rated at 50 V 1.5 A.
The peak voltage rating of the diode is the same as the MOSFET. Hence,
Vdiode ¼ VFET ¼ 45 V
The average current through the diode under worst case conditions (minimum duty
cycle) is
Idiode ¼ 0:87 Io,max ¼ 0:305 A
0:25
R2 ¼
1:15 Io,max
This is true if the internal voltage threshold of 0.25 V is being used. Otherwise,
substitute the voltage at the LD pin instead of the 0.25 V into the equation. Note
that the current limit is set to 15% above the maximum required current, due to
the total 30% ripple specified.
Live
V+
230 VAC
+ C1 C2
33uF 330 nF
350 mA LED
NTC
D1
1 UF4005 L1
VIN 4,7 mH
6
VDD
C3
2,2uF 10 V HV9910
4 Q1
7 GATE STD2NM60T4
LD 2
CS
5
PWM_D R2
0,62R
R1
3 Rosc 8
GND
470K
voltage during the cusps between each half-cycle of the input. Applying high voltage
across high value capacitors creates a large inrush current that can cause damage, so
an inrush limiter (negative temperature coefficient thermistor) is required.
The current rating will depend on the highest average current drawn by the
converter, which is at minimum input voltage (DC level, allowing for a ‘droop’
across the input capacitor between the AC line voltage peaks) and at maximum
output power. The minimum input voltage must be more than half the maximum
LED string voltage, to make sure that the duty cycle stays below 50% and thus
remains stable. For this example, the minimum rectified voltage should be
Vmin,dc ¼ 2 Vo,max ¼ 80 V:
Vo,max Io,max 14
Ibridge ¼ ¼ ¼ 0:194 A
Vmin,dc 72
For this design, using a 230 V AC supply, choose a 600 V 1 A diode bridge.
The thermistor should limit the inrush current to not more than five times the
steady state current, assuming maximum voltage is applied. The required cold
resistance is:
pffiffiffi
2 Vmax,ac
Rcold ¼
5 Ibridge
Vmin,dc ¼ 2 Vo,max ¼ 80 V
The hold-up capacitor required at the output of the diode bridge will have to be
calculated at the minimum AC input voltage. The capacitor can be calculated as
Vo,max Io,max
C1
2 V2min,ac V2min,dc freq
In this example,
C1 26:45 mF
The voltage rating of the capacitor should be more than the peak input voltage.
pffiffiffi
Vmax,cap 2 Vmax,ac
) Vmax,cap 375 V
Io,max 0:25
C2 ¼
fs ð0:05 Vmin,dc Þ
1 ffiV
p o,max
2 Vac,nom
Here, di is the ripple current = 0.3 Io,max and dt is the off-time dt ¼ fs .
ð1 DÞ
Note, a buck circuit duty cycle is given by D ¼ VVout
in
, so the off-time is dt ¼ fs .
Then, the inductor L1 can be computed at the rectified value of the nominal input
voltage as
V
Vo,max 1 pffiffi o,max
2 Vac,nom
L1 ¼
0:3 Io,max fs
In this example, L1 = 4.2 mH. The nearest standard value is 4.7 mH. Since this value
is a little higher than the calculated value, the ripple current will be less than 30%.
The peak current rating of the inductor will be 350 mA plus 15% ripple:
The RMS current through the inductor will be the same as the average current
(i.e. 350 mA).
Note that with a large inductance value, the parasitic capacitance across the coil
could be significant and will affect switching losses.
The maximum RMS current through the MOSFET depends on the maximum duty
cycle, which is 50% by design. Hence, the current rating of the MOSFET is
pffiffiffiffiffiffiffi
IFET Io,max 0:5 ¼ 0:247 A
Typically a MOSFET with about three times the current is chosen to minimize the
resistive losses in the switch. For this application, choose a 600 V, >1 A MOSFET; a
suitable device is an ST part, STD2NM60, rated at 600 V 2 A. This MOSFET has
2.8 on-resistance. With 350 mA being passed up to 50% of the time, the conduction
losses will be 171 mW.
Although a MOSFET with lower on-resistance could be used to reduce the
conduction losses, the switching losses, which are caused by parasitic capacitance
and diode reverse recovery current, will then be higher. The diode D2 passes current
in the reverse direction for a short period: imagine a mechanical value that is passing
a fluid – when the pressure reverses it takes a short time for the valve to close and
shut off the reverse flow. The analogy can be applied to diodes, because they have free
electrons in their conduction band that have to be swept out by the reverse potential
before current flow stops. Each time the MOSFET turns on, a current spike passes
through the MOSFET, but the current is limited by the MOSFET current rating, so a
lower current rating can reduce the switching losses.
The peak voltage rating of the diode is the same as the MOSFET. Hence,
Choose a 600 V, 1 A ultra-fast diode. The UF4005 is a low cost ultra-fast type, but for
greatest efficiency a faster diode like STTH1R06 should be used. If we assume a
forward voltage drop of 1 V at 350 mA, the conduction loss will be less than 350 mW
at low duty cycles. The switching loss could be higher that this value, but is less of a
problem in faster diodes because the reverse conduction is for a shorter time period.
0:25
R2 ¼
1:15 Io,max
This is true if the internal voltage threshold of 0.25 V is being used. Otherwise,
substitute the voltage at the LD pin instead of the 0.25 V into the equation. A lower
voltage could be applied to the LD pin to enable a convenient value of R2 to be used,
as described earlier.
For this design, R2 = 0.625 . The nearest standard value is R2 = 0.62 .
Note that capacitor C3 is a bypass capacitor for holding up the HV9910B internal
supply VDD during MOSFET switching, when high frequency current pulses are
required for charging the gate. A typical value for C3 of 2.2 mF, 16 V is recommended,
although for AC applications smaller capacitors as low as 0.1 mF have been used
successfully. The switching frequency tends to be lower and so the MOSFET gate
current requirements are low. Also with a higher voltage on the input supply pin, the
voltage drop across the internal regulator during MOSFET switching is unlikely to
cause under-voltage drop out.
62 nF
47K 220K
BR100
0.1 mH
TRIAC
47 nF
The input of an inactive LED driver is high impedance, with a large capacitor on
the DC side of the bridge rectifier. The capacitor across the triac allows a small
current to flow through the bridge rectifier and the smoothing capacitor starts to
charge. When the voltage builds up, the LED driver will try to operate. The result
is an occasional flicker of the LED.
What is required is a discharge circuit, to keep the smoothing capacitor voltage
below that required to start the LED driver. A 390 resistor was found to keep
the smoothing capacitor voltage below 5 V. To prevent high power loss when
the circuit is active, a simple voltage detector can be used to disconnect the
390 resistor when a voltage above about 8 V is detected. This circuit is shown
in Figure 5.6.
DC RAIL
AC INPUT
+
1N4006 1N4006
+9 V
390R
220K
47K
VN2460N3
VN0106N3
7V5
7V5 100K
The triac needs to see a load. Once a triac is triggered, it is the load current that
keeps it switched on; the triac is a self-sustaining switch. However, an LED driver
provides no load until the input voltage has risen above the LED voltage, and it takes
a little time for this current to be stable at a sufficiently high level to keep the triac
turned on. For this reason, an additional load must be switched across the LED
driver input at low voltages.
Tests have shown that a 2K2 resistor works as a triac load and that it should
remain in circuit until the supply voltage has risen to about 100 V, but should then be
switched off until the rising edge of the next half-wave. A latching circuit to provide
this function is shown in Figure 5.7.
These circuits can be combined. The voltage detector for the smoothing capacitor
discharge circuit can also be used to provide an enable signal for the LED driver
DC RAIL
AC INPUT
+
1N4006 1N4006
2K2
1W
1N4148 220K
330K
47K
+9 V
VN2460N3 100 V
VN0106N3
7V5
7V5 100K
(PWM input). Thus when the triac is off, the LED driver is also off. The combined
circuit is shown in Figure 5.8.
NTC
100 nF 100 nF V+
X2 X2
+ C1 C2
220K 33uF 330 nF
350 mA LED
+9 V
9V1 D1
1 UF4005 L1
VIN 4,7 mH
6
C3 VDD
1N4006 1N4006
HV9910
Q1
GATE 4 STD2NM60T4
7
2K2 LD 2
CS
1W 1N4148
330K 220K +9 V 5 R2
PWM_D
0,62R
47K 390R 220K 8 R1
+9 V 47K 3 Rosc
GND 470K
100V VN2460N3
VN2460N3 VN0106N3 VN0106N3
7V5 7V5
7V5 100K 7V5
circuit to react; it needs to be at least 300 ns. Operating at 20 kHz will give an on-time
of 466 ns, which is close to the limit for accurate control. A double buck may be
needed (see next section).
Another error is not taking into account the parasitic capacitance of the inductor
windings and the reverse current in the flywheel diode. These factors can be ignored
in low voltage DC applications, but not in AC applications where the rectified supply
is high voltage. The current peak through the MOSFET can be high enough to trip
the current sense circuit, resulting in erratic switching. An RC filter between the
current sense resistor and the current sense input of the integrated circuit may be
necessary. A 2.2 k series resistor followed by a 100 pF shunt capacitor to ground
should be sufficient.
Live
L1 L2
0.1 mH 2 mH
V+
230 VAC
C1 C2
47 nF 47 nF D2 C3
D1 BYD57J 470 nF L3
36 V
0.33 mH
D3 D4
BYD57J BYD57J
1 350 mA LED
6 VIN
+ VDD
C3
2,2uF 10 V HV9910
4 Q1
7 GATE STD2NM60T4
LD 2
CS
5
PWM_D R1 R2
470K 0,56R
3 Rosc 8
GND
The double buck is used when the output voltage is very low and the input voltage is
high. An example is driving a single power LED from an AC supply line. A single
buck stage cannot work easily because the on-time of the buck converter is too small,
unless a very low switching frequency is used.
Assume the maximum duty cycle, Dmax, is less than 0.5; also assume that the first
stage (L2) is in boundary conduction mode (BCM) at Dmax. Boundary conduction
mode means that the current through the inductor only just falls to zero and the next
switching cycle begins.
Vo
Vin min ¼
Dmax2
The storage capacitor voltage at Vin min and Dmax is given by the equation:
The peak current through the input stage inductor, at Vin min equals:
The transfer ratio for a DCM buck converter (where R is load resistor seen by the
converter) is given by:
Vc 2
¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Vin 1 þ 1 þ 8 L2
R Ts D2
The resistor R seen by the first stage (and assuming second stage is in CCM) is
given by:
Vc 2
R¼
Po
2 ð Vc DÞ2 Vo 2
)R D ¼ ¼
Po Po
Combining the previous two equations (which turn out to be a constant):
Vc 2
¼K¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Vin 1 þ 1 þ 8 L2 Po Ts Vo 2
And we can now show that the peak inductor current through L2 is a constant
over the operating input voltage:
ðVin Vc Þ D Ts
iL2,pk ¼
L2
K0
Vin ð1 KÞ Vin Ts
¼
L1
ð1 KÞ K0 T s
¼
L2
We
pffiffiffi can now define the average input voltage as the maximum input voltage
ð 2Vac max Þ and the minimum operating input voltage:
The storage capacitor value is computed based on 10% voltage ripple on the
capacitor at Vin min and Dmax:
The voltage across the storage capacitor, with average voltage input, is given by:
We can now compute the average duty cycle (at average input voltage):
Vo
Davg ¼
Vc avg
Computing the value of L3:
Flywheel Diode
Comparator
–
MOSFET
+ Switch
The MOSFET is turned on when the current level is at or below a minimum reference
voltage. The MOSFET is turned off when the current is at or above a maximum
reference voltage. This is shown in Figure 5.11. By this method, the average LED
current remains constant, regardless of changes in the supply voltage or LED
forward voltage.
Vcs (high)
Vcs (Average)
Vcs
Vcs (Low)
1 VCSðhighÞ þ VCSðlowÞ
RSENSE ¼
2 ILED
In words, the average current sense voltage (midway between the high and low levels)
divided by the average LED current required. The datasheet of the hysteretic
controller being used will give the upper and lower current sense voltage levels that
the comparator uses.