Mixed mode control for dimmable fluorescent lamp

[0001] This application is a continuation of U.S. application Ser. No. 10/883,342 filed Jul. 1, 2004, which claims the benefit of Provisional Application No. 60/540,222, filed Jan. 29, 2004. This application incorporates both of these applications by reference.

[0002] 1. Field of the Invention

[0003] The present invention relates to control of a dimmable discharge lamp, and more specifically to generation of a dimming current for a fluorescent lamp.

[0004] 2. Description of the Related Art

[0005] The electronic ballast for fluorescent lamp dimming control could use either a series LC resonant series loaded circuit, a series resonance parallel loaded circuit, or a series parallel resonance circuit, controlled by either the frequency or the duty cycle of input voltage pulses. The existing duty cycle control employs symmetrically chopped pulses. The series LC resonant series loaded and series parallel resonance circuits are not commonly used for electronic ballast because they behave like a band-pass filter, and so cannot satisfy the high gain required at high impedance for ignition and low load dimming.

[0006] The most common type of conventional electronic ballast uses a series resonance parallel loaded circuit, the structure of which is shown in FIG. 1A. They behave like a low-pass filter and show a high gain at high impedance that is required during ignition and low dimming by the fluorescent lamp. The input of the ballast comes from a DC source that could be a pre-stage power factor correction (PFC) universal boost unit. Switching elements 51 and S2 turn on and off in response to a signal from a controller 10 to convert the DC voltage into an AC voltage. The controller 10 controls states of the switching elements 51 and S2, and thus the waveform of the AC voltage, in accordance with a desired dimming level from a dimmer 11. That is, by adjusting the dimmer 11, the current I_lamp flowing through a fluorescent lamp 12 can be changed, and the light output of the fluorescent lamp 12 can be varied. A resonance circuit, comprising an inductor L and a capacitor C1, is formed between the switching stage, including switching elements S1 and S2, and the fluorescent lamp 12. A capacitor C2 blocks DC voltage to the fluorescent lamp 12.

[0007] The main relations among the signals in the circuit are as follows:

[0008] $\begin{array}{cc}{i}_L=i_{\mathrm{lamp}}+C1\left(d{v}_{C1}/dt\right)& \left(1\right)\\ v_{c1}=V_{\mathrm{in}}-L\left(d{i}_L/dt\right)=\frac{1}{C1}\int \left(i_L-i_{\mathrm{lamp}}\right)dt+V_{c1}\left(IC\right)& \left(2\right)\\ v_{c2}=V_{\mathrm{in}}-L\left(d{i}_L/dt\right)-R_{\mathrm{lamp}}·i_{\mathrm{lamp}}=\frac{1}{C2}\int i_{\mathrm{lamp}}dt+V_{c2}\left(IC\right)& \left(3\right)\end{array}$

[0009] wherein V_c1(IC) is the Initial Condition of voltage across C1, and V_C2(IC) is the Initial Condition of voltage across C2.

[0010] As shown in FIG. 1B, the series resonance parallel loaded circuit behaves as a low-pass filter. The fundamental frequency of the square input pulse would be in the pass band of the network and higher harmonics mainly would be attenuated. The transfer function of the series resonance parallel loaded circuit is:

[0011] $Gp\left(jw\right)=\frac{Vo\left(\mathrm{j\omega }\right)}{Vi\left(\mathrm{j\omega }\right)}=\frac{1}{\sqrt{{\left(1-{\left(\frac{\omega }{\omega o}\right)}^2\right)}^2+{\left(\frac{\omega }{\omega oQp}\right)}^2}}$ $\mathrm{wherein},{\omega }_o=1/\surd LC1$ $Q_o=R/L{\omega }_o=RC{\omega }_o$

[0012] The series resonant parallel loaded ballast with double switch choppers at the DC output of the PFC boost is preferred over other conventional ballasts, because it is adjustable with high voltage requirement at high impedance of ignition, is short circuit proof, and its voltage increases in high impedance and low load during dimming.

[0013] According to one of the conventional approaches, the controller 10 changes the current I_lamp by controlling the frequency f_sw at which the switching elements S1 and S2 turn on and off. The frequency control is used with a fixed duty cycle D=50%. Square pulses of V_in to the ballast are assumed to be DC modulated with a sine wave of switching frequency. The DC component shifts the AC voltage across C1 and is blocked by C2. The average DC voltage, V_av=V_dc/2, remains constant in all loads and a uniform resonance sine wave is assumed over the whole period.

[0014] As shown in FIG. 2, at higher loads the current I_lamp increases with the decrease of the frequency f_sw. However, in some threshold of low dim range, the curve becomes flat, and the light output of the fluorescent lamp 12 cannot be effectively adjusted by changing the frequency f_sw. This threshold depends on the lamp characteristic, input/output voltage, as well as the optimized component selection of C1 and L.

[0015] Another disadvantage of conventional frequency control dimming is that in this flat area of low load control the ballast is too sensitive to the frequency changes. When the frequency f_sw is raised quickly, the response of the circuit is so fast that the ballast becomes unstable. Thus, conventionally, only gradual dimming could be used.

[0016] According to another conventional approach, the controller 10 changes the current I_lamp by controlling the duty cycle D_sw of the switching elements S1 and S2. As shown in FIG. 3, dimming is achieved by reducing pulse width of both switches symmetrically, and symmetric charge/discharge time is used to avoid DC voltage drop. However, there is a gap between the turn on (or close) time of the two switching elements, which may cause a discontinuous conduction mode in a resonant tank circuit at low dimming, and high peak current that lowers the efficiency.

[0017] Therefore, it would be advantageous to provide a method and apparatus for effective and efficient control of the dimming of the fluorescent lamp.

[0018] It is an object of the present invention to provide smooth and continuous control of the output of a discharge lamp. A load threshold, below which the output of the discharge lamp could not be effectively adjusted by the conventional frequency control, is determined. During the dimming of the discharge lamp, when the load is not lower than the load threshold, a conventional frequency control is employed. However, when the load is lower than the load threshold, a complementary duty cycle control is used. The duty cycle of input pulses to a resonance circuit of a ballast is reduced to lower the output of the lamp.

[0019] The present invention uses the general structure of the conventional series resonant parallel loaded ballast with double switch choppers. In low dim light, when one of the switching elements turns off, the other one complementarily turns on. There is no gap between the turn on time of the two switching elements, except for a short delay to prevent short circuit. There is no overlap between the turn on time of the two switching elements, either.

[0020] In the conventional duty cycle control, the dimming is achieved by reducing pulse width of both switches. However, in the present invention, the turn on time of one of the switching elements is reduced, but the turn on time of the other switching element is complementarily increased.

[0035] Objects and advantages of the present invention will become apparent from the following detailed description.

[0036] The present invention employs the general structure of the conventional series resonant parallel loaded ballast with double switch choppers. FIG. 4 shows a mixed mode ballast for controlling a dimmable fluorescent lamp according to one embodiment of the present invention. To provide a smooth and continuous control range of 100% to 10% or lower light control, the present invention provides a mixed mode controller 100 to control the switching elements S1 and S2. The present invention also optimizes the values of L, C1 and C2 of the ballast. The switching elements could be transistors, specifically, FETs.

[0037] In one embodiment, when the load is adjusted from 100% to near threshold of losing the sensitivity to frequency control, the method of the present invention uses the conventional frequency control, and makes use of its symmetrical waveforms, i.e., D=0.5. FIG. 5A shows input pulses to the switching elements S1 and S2 shown in FIG. 4 at high load according to one embodiment of the present invention. The frequency f_sw can be increased to reduce the current I_lamp. As shown in FIG. 2, the I_lamp−f_sw curve has a good slope in this range.

[0038] For dimming in low light, e.g., when the load is less than the threshold, the conventional frequency control does not respond accurately and becomes too sensitive and hard to adjust. The present invention uses pulse width control, or PWM, with complementary gating of switches. The frequency f_sw is fixed at the last adjusted f_sw value, the turn on time of the switching element S1 is reduced, and the turn on time of the switching element S2 is complementarily increased to adjust the current I_lamp. As shown in FIG. 5B, during a working cycle T_sw of the two switching elements, when one of the switching elements turns off, the other turns on. There is no gap between turn on (or close) times of the two switching elements, except for a short adaptive delay (based on the turn-on, turn-off time of switches) to prevent short-circuiting the output from the PFC boost. There is no overlap between the turn on time of the two switching elements either. The complementary gating of S1 and S2 could be implemented as a part of a semiconductor chip, with enough delay to avoid overlap. Also, as shown in FIG. 5C, the complementary gating and consequent complementary adjustment of duty cycles can vary over any given sequence of operation.

[0039] In the complementary pulse width control mode, the duty cycle D is smaller than 0.5. A resonance wave is assumed with different amplitude and phase for two states:
V_in=V_dc, when 0<t<DT_sw; and  State I:
V_in=0, when DT_sw<t<T.  State I:

[0040] Thus, the mixed mode control of the present invention could achieve a smooth continuous I_lamp, or output light, control range from 100% to 10% or lower.

[0041] From the above relations (1)-(3), though a third order differential equation could be defined, since C2>>C1, the role of C2 in the main resonance response could be ignored and is merely intended to block the average DC component of the input pulses. A second order equation of resonance between L and C1 is introduced by the general form:
S²+ξS+ω_o²=0, wherein
ε=R_lamp/L; and
ω_o²=1/LC1

[0042] The general solution for the inductor current with complex roots (τ±ω_r) of the above equation would give:

[0043] $\begin{array}{cc}{i}_L=e^{\tau t}\left(A_1\mathrm{Sin}{\omega }_{r}t+A_2\mathrm{Cos}{\omega }_{r}t\right)\tau =-R_{\mathrm{lamp}}/2L;\mathrm{and}{\omega }_r=\sqrt{\left(\frac{1}{LC1}\right)-\left(\frac{1}{4R_{\mathrm{lamp}}^{2}C{1}^2}\right)}& \left(4\right)\end{array}$

[0044] The effect of load resistance on resonance frequency is usually expressed by the quality factor:
Q_L=R_lamp/√L/C1=R_lamp/Lω_o=R_lampCω_o; ω_r²=ω_o²(1¼Q_L²)

[0045] At steady state, i.e., e^τt decayed, the general form of resonance current through the inductor is a resonating sine wave with the DC component of forced response. The DC forced response of i_L would be V_in−V_av/R_lamp:
i_L(t)=[I_m sin(ω_rt−Φ)]+(V_in−V_av)/R_lamp  (5)

[0046] wherein the constants I_m and Φ could be derived from the Initial Conditions of i_L and V_C1 at the switching instants or as derived below by the boundary solution, I_m represents the peak of the sinusoidal current flowing through the inductor L. The general form of the lamp current is:
i_lamp(t)=[v_c1(t)−V_av]/R_lamp  (6)

[0047] The voltage across C1 is:
v_c1(t)=V_in−L[di_L(t)/dt]=V_in−Lω_rI_m·cos(ω_rt−Φ)  (7)

[0048] The voltage across C2 is the average of the input pulses that is blocked from the lamp plus a small AC oscillation of charging/discharging around this DC component which represents the current through the load, i_c2=i_lamp=C₂(dv_c2/dt). As C₂ is much bigger than C₁ this AC component of voltage across C₂ is very small and can be ignored.
v_c2=V_av=DV_dc  (8)

[0049] FIG. 6A shows two distinct states of input voltage and four intervals of conduction with their equivalent circuits. State I includes intervals 1 and 2, during which the V_in is V_dc. State II includes intervals 3 and 4, during which the V_in is 0V.

[0050] As the above resonance circuit operates with zero voltage switching (ZVS), in FIG. 4, diode D1 is conducting when switch S1 turns ON).

[0051] During interval 1, as shown in FIG. 6B, D1 is conducting, and V_in=V_dc. The current direction in the resonant tank circuit is negative, feeding back the output of the PFC Boost.

[0052] During interval 2, as shown in FIG. 6C, S1 is conducting, and V_in=V_dc. The current direction in the resonant tank circuit is positive.

[0053] During interval 3, as shown in FIG. 6D, D2 is conducting, and V_in=0V. The current direction in the resonant tank circuit is changed from positive to negative.

[0054] During interval 4, as shown in FIG. 6E, S2 is conducting, and V_in=0V. The current direction in the resonant tank circuit is negative.

[0055] The above differential equations derived for the resonant tank are solved for each of the two states. As a result, the resonating inductor current for each state is as follows:
when 0<t<DT_sw,V_in=V_dc,i_L(t)=I_m1 sin(ω_rt−Φ1)+((1−D)V_dc)/R_lamp; and  State I:
when DT_sw<t<T_sw,V_in=0,i_L(t)=I_m2·sin [ω_r(t−DT_sw)−Φ2]−D_dc/R_lamp  State II:

[0056] Equalizing the inductor currents and capacitor voltages at the boundary of these states gives the boundary values of inductor current, capacitor voltage, and the relation between the current of lamp I_lamp at dimming condition and the pulse width, or duty cycle D, could be derived.

[0057] Specifically, lamp resistance at each dimming condition is defined based on the linearized approximation of the VI characteristics of the fluorescent lamp. An approximate graph of a fluorescent VI characteristic is shown in FIG. 7.

[0058] In the general approximation of the VI characteristic, the relation for the rms values could be written as follows:

V_lamp—rms=V_nl−R_neg−I_lamp—rms wherein V_nl represents the cross point of the approximate line with V-axis and R_neg represents the negative slope of the line.

[0059] Consequently, the equivalent resistance of lamp at each operating point is defined by:
R_lamp=V_lamp—rms/I_lamp—rms=(V_lamp—rms/I_lamp—rms)−R_neg  (9)
When 0<t<DT_sw, V_in=V_dc, and
i_L(t)=I_m1 sin(ω_rt−Φ1)+((1−D)V_dc)/R_lamp;  (10)
v_c1(t)=V_dc−Lω_rI_m1 cos(ω_rt−Φ1);  (11)
i_lamp(t)=(1/R_lamp)[(1−D)V_dc−Lω_rI_m cos(ω_rt−Φ1)];  (12)
When DT_sw≦t≦T_sw, V_in=0; time shift of DT_sw
i_L(t)=I_m2·sin [ω_r(t−DT_sw)−Φ2]−DV_dc/R_lamp;  (13)
v_c1(t)=−Lω_rI_m2·cos [ω_r(t−DT_sw)−Φ2]; v_c2≈DV_dc  (14)
i_lamp(t)=(1/R_lamp){−DV_dc−Lω_rI_m1·cos [ω_r(t−DT_sw)−Φ2]};  (15)

[0060] From the boundary conditions for the inductor current and capacitor voltage, the following equations could be derived (in literature usually the frequency ratio y=ω_sw/ω_r=F_sw/F_r is used in relations):

[0061] when t=0, or T_sw:
−I_m1 sin Φ1+((1−D)V_dc)/R_lamp=I_m2·sin [ω_r(1−D)T_sw−Φ2]−(DV_dc)/R_lamp;  (16)
V_dc−Lω_rI_m1 cos Φ1=−Lω_rI_m2 cos [ω_r(1−D)T−Φ2];  (17)
when t=DT_sw:
I_m1 sin(ω_oDT_sw−Φ1)+((1−D)V_dc)/R_lamp; =−I_m2·sin(Φ2)−DV_dc)/R_lamp  (18)
V_dc−Lω_rI_m1 cos(ω_rDT_sw−Φ1)=−Lω_rI_m2 cos(Φ2)  (19)

[0062] An extra relation between the adjusted dimming current and the required control of the duty cycle is obtained from calculation of the lamp RMS current from the i_lamp waveform as below:

[0063] $\begin{array}{cc}{I}_{\mathrm{lamp}\_\mathrm{rms}}=\sqrt{\frac{1}{T}\left[{\int }_0^{DT}{i}_{\mathrm{lamp}}^2·dt+{\int }_{DT}^T{i}_{\mathrm{lamp}}^2·dt\right]}{I}_{\mathrm{lamp}\_\mathrm{rms}}^2=\frac{1}{T}\langle {\int }_0^{DT}\frac{{{\mathrm{Vdc}}^2\left(1-D\right)}^2}{R_{\mathrm{lamp}}^2}+\frac{L^2{\omega }_r^{2}Im{1}^2}{2·R_{\mathrm{lamp}}^2}\left[1+\cos \left(2{\omega }_{r}t-2\mathrm{\varphi 1}\right)\right]-\frac{2\mathrm{Vdc}\left(1-D\right)L{\omega }_{r}Im1}{R_{\mathrm{lamp}}^2}\cos \left({\omega }_{r}t-\mathrm{\varphi 1}\right)+\dots +{\int }_{DT}^T\frac{{\mathrm{Vdc}}^2{D}^2}{R_{\mathrm{lamp}}^2}+\frac{L^2{\omega }_r^{2}Im{2}^2}{2·R_{\mathrm{lamp}}^2}\left[1+\cos \left(2{\omega }_{r}t-2\mathrm{\varphi 2}\right)\right]+\frac{2VdcDL{\omega }_{r}Im2}{R_{\mathrm{lamp}}^2}·\cos ({\omega }_{r}t-\mathrm{\varphi 2}\rangle & \left(20\right)\end{array}$

[0064] Using the boundary values calculated by equations (16)-(20), the relation between the total RMS current of I_lamp and D is derived. By sensing of circuit parameters, based on the derived relations, the controller would adjust the Frequency/Duty Cycle for the required dimming.

[0065] Alternatively, the relation of the real power and energy transferred from ballast input to output lamp could be used. Ignoring the parasitic losses in the circuit, the average of input power (that is only during state I, 0<t<DT_sw and V_in=V_dc) to the output power consumed in the fluorescent lamp is:

[0066] $\begin{array}{cc}\mathrm{Pin}=\mathrm{Pout}=V_{\mathrm{lamp}\_\mathrm{rms}}·I_{\mathrm{lamp}\_\mathrm{rms}};P_{\mathrm{in}}=\frac{1}{T_{\mathrm{sw}}}{\int }_0^{DTsw}{V}_{dc}[I_{m1}\sin \left({\omega }_{r}t-\mathrm{\varphi 1}\right)+\frac{\left(1-D\right)\mathrm{Vdc}}{R_{\mathrm{lamp}}}dt=-\frac{V_{dc}·I_{m1}}{T_{\mathrm{sw}}·{\omega }_r}[{\cos \left({\omega }_{r}t-\mathrm{\varphi 1}\right]}_0^{DTsw}+{\left[\frac{\left(1-D\right){\mathrm{Vdc}}^2}{R_{\mathrm{lamp}}}·t\right]}_0^{D·Tsw}=\frac{V_{dc}{I}_{m1}{F}_{sw}}{{\omega }_r}\left[\cos \mathrm{\varphi 1}-\cos \left(\frac{{\omega }_{r}D}{F_{sw}}-\mathrm{\varphi 1}\right)\right]+\frac{D\left(1-D\right){\mathrm{Vdc}}^2}{R_{\mathrm{lamp}}}& \left({21}^{\prime }\right)\\ V_{\mathrm{lamp}\_\mathrm{rms}}{I}_{\mathrm{lamp}\_\mathrm{rms}}=\frac{V_{dc}{I}_{m1}{F}_{sw}}{{\omega }_r}\left[\cos \mathrm{\varphi 1}-\cos \left(\frac{{\omega }_{r}D}{F_{sw}}-\mathrm{\varphi 1}\right)\right]+\frac{D\left(1-D\right){\mathrm{Vdc}}^2}{R_{\mathrm{lamp}}}& \left(21\right)\end{array}$

[0067] According to another aspect of the present invention, the elements of the ballast of the present invention, L, C1 and C2, could be optimized for high efficiency and high performance at full load.

[0068] FIGS. 8A-8D illustrate operational simulated waveforms of the mixed mode control, with different values of dimming current, frequency and duty cycle, according to one numerical example for an embodiment of the present invention.

[0069] FIGS. 9A and 9B show the two portions of dimming control characteristic illustrating the relation between dimming current and frequency (at higher loads)/duty cycle (at lower loads) during mixed mode dimming control according to one embodiment of the present invention.

[0070] FIG. 10 shows a basic block diagram of the mixed mode controller 100 shown in FIG. 4 according to one embodiment of the present invention. As shown, a dimming control 1001 provides a processor, e.g., a DSP core 1000, with dimming signals and a load detector 1002 informs the DSP core 1000 whether the load of the circuit is below a threshold. When the load is not below the threshold, the DSP core 1000 outputs signals to an upper driver 1003, which drives the switches, e.g., S1 and S2 shown in FIG. 4, to operate in a frequency control mode. When the load is below the threshold, the DSP core 1000 outputs signals to a lower driver 1004, which drives the switches to operate in a complementary duty cycle control mode. The mixed mode controller 100 receives voltage signal from a Vcc source 1005. The mixed mode controller 100 also has: an under voltage lock out module 1006 and an over voltage protection module 1007 for protecting the DSP core 1000 from under voltage an over voltage; a power program block 1005 for control the power of the DSP core 100; a phase detect block 1009 for detecting a phase of the input signal; and a ΔF/ΔI program block 1010 for calculating ΔF/ΔI.

[0071] While the invention has been described in detail above with reference to some embodiments, variations within the scope and spirit of the invention will be apparent to those of ordinary skill in the art. For example, the lamp is not limited to a fluorescent lamp, but could be another type of discharge lamp. In addition, the load threshold could go higher or lower based on the optimization point of the design and input/output voltages required. Thus, the invention should be considered as limited only by the scope of the appended claims, and not by the described embodiments.