SINGLE-STAGE HIGH-POWER-FACTOR SELF-OSCILLATING ELECTRONIC BALLAST FOR FLUORESCENT
LAMPS WITH SOFT START
Abstract: In this paper a new solution to implement and control a single-stage electronic ballast based on the integration of a
buck-boost power factor correction stage and a half bridge resonant inverter is presented. The control signals are obtained
using the inverter resonant current by means of a saturable transformer. Core saturation is used to control the required dead
time between the control pulses on both switches. The turn-on time of one of the inverter switches is controlled to provide
proper cathode preheating during the lamp ignition process. No special integrated circuits are required to control the ballast
and the total number of components is minimized. Analysis and basic design guidelines are presented in the paper together with
experimental results obtained from a laboratory prototype.
1.- Introduction
Most of the electronic ballasts use a double AC-DC-AC power conversion. The traditional low cost solution for the AC-DC
conversion using a full wave rectifier and a filter capacitor can not be used due to the restrictions of the harmonic content in the
input current imposed by regulations such as IEC 1000-3-2. Most of the commercially available electronic ballasts comprise an
AC-DC power factor correction stage followed by an inverter. Recently, several topologies have been proposed to lower the cost
of this kind of solutions by reducing the total number of components [1],[2],[3]. One of these topologies is a single-stage
double-switch converter derived from the integration of a buck-boost based power factor correction stage and a resonant inverter
[4]. In present paper, a new solution to control an electronic ballast based in this topology is proposed and analyzed. A low-cost
self-oscillating control system is used, making unnecessary any kind of control IC’s. The basic design guidelines are also given
in this paper. Finally, the experimental results obtained from a laboratory prototype are presented.
2.- Power stage characteristics and implementation
The power stage is based on the buck-boost inverter [4]. This topology comes from the integration into a single stage of a buckboost DC-DC converter working in DCM followed by a half bridge resonant inverter (See Fig. 1). The integration of these two
stages is made using one switch that simultaneously plays the roles of the buck-boost switch and one of the half bridge switches.
The most relevant waveforms of this topology are shown in Fig. 2. The input characteristics of this converter correspond to that
of a buck-boost stage working in DCM. Operating at a constant frequency and duty cycle, the mean averaged value of the input
current follows the rectified line voltage in capacitor CF. Using a suitable filter capacitor CF and including some additional
filtering to eliminate switching noise, an input current almost proportional to the line voltage is obtained. The power delivered to
the load can be calculated using the following expression:
2
⋅ δ2 ⋅ T ⋅ η
V
(1)
PLAMP = LINE( peak )
4L
Where δ is the duty cycle of switch Q2, T is the switching period, η is the converter efficiency and L is the buck-boost inductor
value (see Fig. 1).
From the load standpoint, the converter behaves as a half bridge resonant inverter. Fig. 2 shows the resonant tank input voltage
and current. For a given current phase φ, the death-angle between the control pulses on both switches φd can be adjusted between
zero and φ without changing the operating point of the resonant tank. Therefore, the duty cycle of the shared switch Q2 could be
used to control the operating point of the buck-boost stage independently of the inverter stage. The duty cycle could be varied
from a minimum value of:
(2)
δ min =
π−φ
2π
to a maximum of 0.5. This range could be used to control the power delivered to the lamp or as an additional design parameter.
3.- Operation of the self-oscillating control circuit
One of the control circuits most commonly used in electronic ballasts to drive the switches of a resonant inverter is made using a
saturable transformer that generates the control signals from the resonant current [5]. Once the circuit oscillation has been
started, the resonant current alternately drives the inverter switches in such a manner that oscillation is self-sustained. Usually a
high permeability saturable toroidal core is used for the control transformer. The death-time between the control pulses on both
switches is provided when the core reaches saturation. This control strategy could also be used on a single stage ballast as the
one presented in this paper. The driver circuit used in the laboratory prototype is a variation of the circuit typically used in selfoscillating resonant inverters. This variation has been made in order to provide a more precise control of the switching times.
The simplified schematics of the control circuit is shown in Fig. 3. The primary winding of the control transformer is placed in
series with the resonant tank. The resonant current is divided between the equivalent magnetizing inductance of the transformer
and one of the secondary windings. While current keeps flowing through one of the diodes placed in series with the secondary
windings the voltage is clamped by the zener diode. When the magnetizing current reaches its saturation value, the voltage VS
drops to zero and all the resonant current flows through the saturated magnetizing inductance. The VS signal obtained this way is
suitable to control the switches of the power stage. If the secondary winding voltage is fixed by the zener diode during all nonsaturation intervals, the turn on time of both switches can be approximately calculated using:
2Φ max
(3)
t ON = N 2·
Vzener
The driver circuit used in the laboratory prototypes is shown in Fig. 4. This circuit is a variation of the previously described and
provides an enhanced rise time. Transistor Q1 applies voltage to the current amplifier made by Q2 and Q3 only when the
secondary voltage grows higher than the zener diode voltage. This circuit reduces the rise time due to the gradual saturation
turn-out of the transformer core.
4.- Design considerations
The use of a self-oscillating control scheme in a single stage converter as the one analyzed in this paper, makes the operating
point of the circuit highly dependant on the lamp characteristics. So the operating frequency and the DC bus voltage are
influenced by the changes in the lamp equivalent resistance that take place during the warm-up phase. The design of the
converter has to be made in order to avoid a DC bus overvoltage, that should damage the converter switches, and also to avoid a
sustained DC bus undervoltage that will cause the converter to enter in continuous conduction mode.
The converter design can be greatly simplified if the saturation current of the transformer core is neglected (i.e. isat≈0 and φd≈φ,
see Fig. 3), in that case, the control pulses of both switches begin at the zero crossing of the resonant current and the following
expression could be derived:
t ON
1 φ
(4)
= δ = 1 − 
T
2
π
Combining expressions (1) and (4), it can be found:
2
1 − φ 
π
2
(5)
⋅ VLINE( peak )
PLAMP ∝ 
Ω
Where Ω is the normalized operating frequency. The relation between φ and Ω is given by the resonant tank input impedance.
Using the fundamental approach this impedance can be calculated by:
1− α
1
and
(6)
φ = arg(Z n )
Z n = jΩ +
+
jΩ 1
jΩ
+
α Q
The base values and circuit parameters used are summarized in table 1. Figure 5 depicts expression (5) as a function of the
normalized tON time for a VLINE(peak) = 1. As it can be seen, the output power increases slowly with the lamp equivalent resistance
thus providing a stable operating point.
Figures 6 and 7 show the normalized frequency versus load and phase angle versus load characteristics respectively. The dotted
line in figures 5, 6 and 7 represent the boundary between DCM and CCM for the laboratory prototype that will be described
later in this paper. This boundary can be calculated using the following condition:
δ
(7)
VBUS = VLINE ( peak ) ⋅
1− δ
and the expression relating the lamp power to the DC bus voltage in the resonant tank:
2
(8)

 2

1
PLAMP = 
⋅ VBUS  ⋅ Re
⋅
π
Z
Z
 n BASE





To assure a high input current power factor the converter operating point should never reach the CCM area due to changes in the
lamp characteristic.
Figures 5, 6 and 7 have been plotted using the fundamental approach. The shadowed area corresponds to operating points where
the third harmonic of the resonant current is greater than 10% of the fundamental. This has been chosen as a practical limit for
using the fundamental approach.
From expression (1) it is deduced that the power delivered to the lamp depends highly on the input voltage. In case of having a
fixed lamp equivalent resistance, the output power will be proportional to the square of the line input voltage. However, the
negative incremental impedance presented by fluorescent lamps provides a lower dependency on the input voltage. Figure 8
shows the ratio between the lamp power variation percentage and the line input voltage variation percentage as a function of the
normalized load Q and tON time for small changes in the line input voltage. This characteristic has been obtained supposing that
lamp voltage is unaffected by changes in the power delivered to the lamp. It can be demonstrated that the actual dependency
between the output power and the line voltage is slightly lower due to the existing lamp voltage to lamp power relationship on
fluorescent lamps.
5.- Cathode preheating circuitry
Figure 9 shows the variations made to the control circuit of the shared switch Q2 in order to obtain cathode preheating. Initially,
transistor QC5 is in the off-state and so does QC4. The voltage in the secondary winding VSW will be approximately equal to the
zener diode voltage VDZ plus the voltage drop in resistor RPR. When transistor QC5 is turned on, the resistor RPR is bypassed by
transistor QC4 and the voltage VSW becomes lower. The width of the control pulse depends mainly on the voltage applied to the
secondary windings, so this circuit can be used to drive Q2 switch initially with a low duty cycle and increase it to the normal
operation value thus providing a programmable starting scenario. Another interesting characteristic of this circuit is that
provides soft-start of the buck-boost semi-stage thus reducing the risk of entering in discontinuous conduction mode during the
starting transient.
6.- Experimental results
Figure 10 shows the complete electrical diagram of the laboratory prototype. The circuit also includes the additional circuitry
required to start oscillations using a thyristor. The prototype has been adjusted to deliver 40W to a TLE-40 fluorescent lamp
from Philips using a 220V/50Hz AC source. Figure 11 shows the converter input current and voltage. The harmonic content of
these signals is presented in table 2. Figure 12 shows the control signal of the shared switch together with the resonant current
and the secondary windings voltage of the control transformer. Figure 13 shows the lamp voltage and current during the
cathode preheating process and ignition transient.
7.- Conclusions
A new low-cost high power factor electronic ballast for fluorescent lamps has been presented and analyzed. The total number of
components has been reduced using a single stage power topology and a self-oscillating control system that requires no extra
control IC’s and that could be adjusted to provide a suitable starting scenario. The basic design guidelines has been given and
experimentally verified using a laboratory prototype. The experimental results obtained with this prototype have been presented
showing a good overall performance.
References
[1] J.M. Alonso, A.J. Calleja et al.; “Analysis and experimental results of a single-stage high-power-factor electronic ballast
based on flyback converter”, IEEE APEC’98 proceedings, pp. 1142-1148.
[2] T.F. Wu and T.H. Yu; “Off-Line Applications with Single-Stage Converters”, IEEE Transactions on Industrial Electronics,
Vol. 44,No. 4, October 1997, pp. 638-647.
[3] J. Quian, F.C. Lee and T. Yamauchi; “Current-source charge-pump power-factor-correction electronic ballast”, IEEE
Trans. on Power Electronics, Vol. 13, May 1998, pp. 564-572.
[4] J.M. Alonso, A.J. Calleja et al.; “Single-stage constant-wattage high-power-factor electronic ballast with dimming
capability”, IEEE PESC’98, pp. 1330-1336, Fukuoka, Japan, 1998.
[5] Y.R. Yang and C.L. Chen; “Steady-state analysis and simulation of a BJT self-oscillating ZVS-CV Ballast driven by a
saturable transformer”, IEEE Trans. Industrial Electronics, Vol. 46, No. 2, April 1999.
Figure 1: Schematics of the power stage
Figure 2: Basic waveforms
Figure 3: Operation of the control circuit
PARAMETERS
α
δ
CS
CS + C P
t ON
T
BASE VALUES
ZBASE
ωBASE
1
Figure 4: Schematics of the control circuit
 C ⋅C 
L ⋅  S P 
 CS + CP 
L ⋅ CS ⋅ C P
CS + C P
tBASE
1
ω BASE
Table 1: Base values and definitions
1 − φ 
π

Ω
2
Figure 5: Lamp power – load characteristic
Figure 6: Normalized frequency – load characteristic
dPLAMP VLINE
⋅
dVLINE PLAMP
Figure 7: Phase angle – normalized load characteristic
Figure 8: Lamp power sensibility to line voltage variations
Figure 9: Driver circuit with programmed start
Figure 10: Schematics of the laboratory prototype
Figure 11: Input current and voltage
Figure 12: Resonant current (200mA/div), Q2 gate signal (5V/div)
and secondary winding voltage (5V/div).
Figure 13: Lamp voltage and current during ignition
Efficiency
Power factor
THD
1st Har.
3rd Har.
5th Har.
7th Har.
9th Har.
83%
0.99
6.8%
0.213
0.005
0.002
0.000
0.005
Table 2: Harmonic content of the input current
11th Har. 13th Har.
0.005
0.004
15th and above
<0.004