Sampled-data modeling and dynamical effect of output-capacitor time-constant for valley voltage-mode controlled buck-boost converter
Zhou Shu-Han, Zhou Guo-Hua, Zeng Shao-Huan, Leng Min-Rui, Xu Shun-Gang
Key Laboratory of Magnetic Suspension Technology and Maglev Vehicle of Ministry of Education, School of Electrical Engineering, Southwest Jiaotong University, Chengdu 610031, China

 

† Corresponding author. E-mail: ghzhou-swjtu@163.com

Project supported by the National Natural Science Foundation of China (Grant Nos. 61371033 and 51407054), the Foundation for the Author of National Excellent Doctoral Dissertation of China (Grant No. 201442), and the Fundamental Research Funds for the Central Universities of China (Grant No. 2682016CX035).

Abstract

By analyzing the output voltage ripple of a buck-boost converter with large equivalent series resistance (ESR) of output capacitor, one valley voltage-mode controller for buck-boost converter is proposed. Considering the fact that the increasing and decreasing slopes of the inductor current are assumed to be constant during each switching cycle, an especial sampled-data model of valley voltage-mode controlled buck-boost converter is established. Based on this model, the dynamical effect of an output-capacitor time-constant on the valley voltage-mode controlled buck-boost converter is revealed and analyzed via the bifurcation diagrams, the movements of eigenvalues, the Lyapunov exponent spectra, the boundary equations, and the operating-state regions. It is found that with gradual reduction of output-capacitor time-constant, the buck-boost converter in continuous conduction mode (CCM) shows the evolutive dynamic behavior from period-1 to period-2, period-4, period-8, chaos, and invalid state. The stability boundary and the invalidated boundary are derived theoretically by stability analysis, where the stable state of valley voltage-mode controlled buck-boost converter can enter into an unstable state, and the converter can shift from the operation region to a forbidden region. These results verified by time-domain waveforms and phase portraits of both simulation and experiment indicate that the sampled-data model is correct and the time constant of the output capacitor is a critical factor for valley voltage-mode controlled buck-boost converter, which has a significant effect on the dynamics as well as control stability.

1. Introduction

Voltage ripple-based control technique, using the output ripple voltage as the pulse width modulation (PWM) ramp, is commonly used to improve the transient response of the dc–dc converter,[1] such as hysteretic control,[2] constant-on-time control,[3] V2 control,[1,4] and so on.

The conventional V2 control is essentially a peak V2 control technique,[4] where an error amplifier exists between the reference voltage and the PWM comparator as shown in Fig. 1. When the output voltage of the error amplifier is varied very slowly, the output part can be represented by a fixed reference voltage. Under this assumption, the peak V2 control can be simplified into the peak voltage-mode control. The simple control architecture, which does not require an error amplifier but still contains the advantages of the ripple-based control technique, makes the peak voltage-mode control technique popular, especially in computer or portable electronic devices.[1] According to the duality, there would exist a valley voltage-mode control technique for switching converter. The output voltage is positive for the buck converter and boost converter. However, the output voltage of the buck-boost converter is negative. Thus, this leads to meaningfully investigating whether the peak voltage-mode control or valley voltage-mode control can be applied to the buck-boost converter.

Fig. 1. (a) Peak V2 controlled buck converter and (b) corresponding operation waveforms.

The ripple controllers, however, suffer some serious stability problems. All the voltage ripple-based controlled switching converter will become unstable and operate in a subharmonic oscillation state or chaos if the time constant (ESR times capacitance) of the output capacitor is smaller than a certain value.[37] Therefore, it is necessary to investigate the dynamical effect of the output-capacitor time-constant for the ripple-based controlled switching converter.

In recent years, the complex dynamical behavior of the switching converter has become a hot research topic.[5,6,817] By establishing a one-dimensional or two-dimensional discrete iterative map model, various nonlinear phenomena and dynamical behaviors of switching converters, with the circuit parameters varied, are extensively investigated, such as period-doubling bifurcation,[5,6] multi-periodic behaviors,[8] border collision,[6,9] subharmonic oscillation,[10] Hopf bifurcation,[11,12] time bifurcation,[13] and chaos.[1417]

Assuming the output voltage is constant, a one-dimensional discrete iterative map model of the ripple-based controlled buck converter is obtained, which is simple but fails to be used to discuss the influence of output-capacitor time-constant.[5] The precise two-dimensional discrete iterative map models are obtained through solving the state equations of ripple-based controlled buck and boost converters with different operation modes.[6,18] However, the modeling process and the discrete iterative map models are rather complicated. Moreover, the corresponding eigenvalues of Jacobian and the maximal Lyapunov exponent are difficult to deduce.

The sampled-data modeling technique combines both the continuous form of the state-space averaged model and the high-frequency accuracy of the discrete-time model, which is usually applied to stability analysis and small-signal analysis.[1,7,19] Especially in describing the responses of inductor current(s) and capacitor voltage(s) at the switching instants, the sampled-data model fits naturally with the operation of the switching regulator. Since the output ripple voltage is much smaller than output voltage in almost all practical design, which has nothing to do with the slope of the inductor current, the rising and falling slopes of the inductor current can be assumed to be constant during every switching cycle.[1] Under this assumption, various transfer functions, including control-to-output voltage transfer function, control-to-inductor current transfer function, audio-susceptibility, and output impedance, can be easily obtained through performing a small-signal analysis by sampled-data modeling.[19] Using the sampled-data modeling, the critical ESR for the stability of the valley V2 controlled boost converter is investigated efficiently.[7] However, most of the studies focus on the small-signal analysis and stability analysis, no attention is paid to the dynamical modeling and bifurcation analysis of switching power converters.

In this paper, we will firstly investigate the possibility of a peak voltage-mode or valley voltage-mode controlled buck-boost converter and then propose its valley voltage-mode controller. In addition, a simple sampled-data modeling is performed, in which the bifurcation diagrams, the movements of eigenvalues, the Lyapunov exponent spectra, the boundary equations and the operating-state regions are used for revealing and analyzing the effect of the output-capacitor time-constant on the dynamics of the valley voltage-mode controlled buck-boost converter. Finally, circuit simulation and experimental results are provided to verify the correctness of the theoretical analysis and sampled-data model.

2. Sampled-data modeling of valley voltage-mode controlled buck-boost converter
2.1. Valley voltage-mode controlled buck-boost converter

Figure 2(a) shows a buck-boost converter, where vg is the input voltage, vo is the output voltage, RC is the equivalent series resistance (ESR) of output capacitor C, vC is the capacitor voltage, vESR is the voltage across the ESR, io, iC, and iL are load current, capacitor current, and inductor current, respectively. When the buck-boost converter operates in continuous conduction mode (CCM), its main steady-state waveforms are shown in Fig. 2(b) if the large ESR RC is used.

Fig. 2. (a) Buck-boost converter and (b) its steady-state waveforms with large ESR.

In Fig. 2(a), the output voltage is the sum of vC and vESR. If large ESR RC is used, the output voltage ripple is dominated by the ripple voltage across the ESR and thus looks piece-wise linearly. As shown in Fig. 2(b), when the switch S1 is ON, iL increases with the slope of m1 and C is discharged to supply the power to the load, which makes output voltage decrease. Due to the negative polarity of output voltage, output voltage in Fig. 2(b) exhibits negatively increasing. When the switch S1 is OFF, iL decreases linearly with increasing time t, C is charged, L and C supply the power to the load. In this duration, the output voltage also exhibits a negative increase due to a large ESR and the negative polarity of the output voltage. There exist two jump voltages in each switching cycle Ts because of the existence of ESR.

In the buck converter with peak voltage-mode control, the switch S1 is turned on at the beginning of each switching cycle and turned off after the positive output voltage is increased to a control signal. For the buck-boost converter, however, after the switch S1 is turned on, the negative output voltage increases, which means that the positive output voltage does not increase but decreases if the polarity of output voltage is inverted. Therefore, the peak voltage-mode control cannot be used for controlling the buck-boost converter. For the valley voltage-mode control, the switch S1 is turned off at the beginning of each switching cycle and turned on after the positive output voltage decreases to a control signal. It is just the case for the buck-boost converter during the switch S1 being in the OFF state (Fig. 2(b)) if the polarity of output voltage is inverted. Therefore, valley voltage-mode control can be applied to the buck-boost converter.

According to the polarity of the reference voltage, one valley voltage-mode controller for the buck-boost converter can be proposed. Figure 3(a) shows a valley voltage-mode controller with negative reference voltage, where the controller consists of an error amplifier, a comparator, a latch and a clock clk, Kv is the output voltage sensing coefficient (Kv > 0), and −Vref is the reference voltage (Vref > 0). It is noted that if the reference voltage is positive, i.e., Vref < 0, the output voltage sensing coefficient must be negative, i.e., Kv < 0, and the two inputs of the comparator should be interchanged. The corresponding operation waveforms are shown in Fig. 3(b). It can be seen from Figs. 3(a) and 3(b) that the output voltage is used to generate the ramp signal vs. The control objective of the valley voltage-mode controlled buck-boost converter is to make the negative peak value of vs follow Vref.

Fig. 3. (a) Valley voltage-mode controller for buck-boost converter and (b) corresponding operation waveforms.

At the beginning of every switching cycle, the clock resets the latch and makes the driver signal vQ be low level, the switch S1 is turned off, which causes iL to decrease linearly from the initial value. The inductor current ripple ΔiL will flow through C when S1 is OFF. Under the condition of large ESR, the voltage ripple of vs is approximately equal to KvΔiLRC. When vs with negative polarity increases to Vref, the output of the comparator sets the latch and makes the driver signal vQ be high level, S1 is turned on, which makes iL increase and vs (with negative polarity) increase till the end of the present switching cycle.

2.2. Sampled-data modeling

To investigate the dynamics of a valley voltage-mode controlled buck-boost converter, an especial sampled-data modeling is performed by assuming that the increasing and decreasing slopes of the inductor current are constant during each switching cycle. This condition is reasonable in almost all practical designs because the output voltage ripple has a trivial influence on the slope of the inductor current. The capacitor voltage vC and inductor current iL at the beginning of the nth switching cycle are denoted as vn and in. When the switch S1 is turned on, inductor current in+k can be obtained easily from Fig. 3(b) as follows: where toff is the off time duration of switch S1 in the nth switching cycle and m2 = −vo/L is the decreasing slope of iL. From Fig. 2, during the off time duration of switch S1, the capacitor current iC is equal to −(iL + io). Therefore, the capacitor voltage vn+k can be derived as Similarly, iL and vC at the end of the nth switching cycle can be expressed as where m1 = vg/L is the increasing slope of iL.

From Fig. 3(b), the sensed output voltage vo(n + k) at the turn-on instant of the switch S1 is equal to the reference voltage Vref, i.e., Kvvo(n + k) = Vref. Therefore, the control constraint can be expressed as follows:

3. Dynamical analysis
3.1. Bifurcation behaviors

Significantly, it is found that instability exists in the valley voltage-mode controlled buck-boost converter when duty ratio D is less than 0.5 (D < 0.5), which is the same as that in the valley V2 controlled boost converter.[6] Therefore, the dynamical analysis in this paper mainly focuses on the case of D > 0.5. Equations (1)–(5) constitute a complete two-dimensional sampled-data model of the valley voltage-mode controlled buck-boost converter, based on which, when input voltage vg = 4 V, Vref = 0.602 V, Kv = 0.1, Vc = −0.6 V, Ts = 20 μs, L = 20 μH, C = 940 μF, and R = 3 Ω keep constant while RC is varied from 20 mΩ to 10 mΩ, bifurcation diagrams of inductor current and output voltage of the valley voltage-mode controlled buck-boost converter are obtained, as shown in Fig. 4.

Fig. 4. (color online) Bifurcation diagrams of (a) iL versus RC and (b) vo versus RC.

From Fig. 4, it is shown that the valley voltage-mode controlled buck-boost converter has complex dynamical behaviors. With the decrease of RC, the orbit of the converter is transformed from steady period-1 state to subharmonic oscillation (period-2 state), owing to the first period-doubling bifurcation at point A where RC = 19.2 mΩ. As RC gradually decreases, the converter goes from the period-2 state to the period-4 state due to the second period-doubling bifurcation at point B where RC = 12.25 mΩ. From Fig. 4, it can also be known that a period-8 state region exists in the bifurcation diagram during RC = 11.69 mΩ ∼ 11.6 mΩ, owing to the third period-doubling bifurcation taking place at point C where RC = 11.69 mΩ. As RC decreases further, the period-8 state disappears and a chaotic orbit appears at RC = 11.6 mΩ. When RC decreases to 10.65 mΩ finally, the converter with valley voltage-mode control is invalid.

The following is analyzed in a similar way when C is taken as a bifurcation parameter. By fixing RC = 19 mΩ, the inductor current bifurcation diagram can be obtained and shown in Fig. 5(a) with C changing from 500 μF to 1000 μF, as well as its output voltage bifurcation diagram drawn in Fig. 5(b).

Fig. 5. (color online) Bifurcation diagrams with the variation of C: (a) iL versus C; (b) vo versus C.

It is found from Fig. 5 that as the output capacitor C decreases from 1000 μF to 573.9 μF, the orbit of the converter has a transition from steady period-1 state to multi-period state due to period-doubling bifurcation. With C decreasing to 573.9 μF, the orbit enters into a chaotic state. When C ≤ 526.7 μF, the chaotic state disappears, and the voltage-mode controlled buck-boost converter is invalid. In the ranges of C = 950.2 μF–605.2 μF, 605.2 μF–578.3 μF, and 578.3 μF–573.9 μF, the system operates in period-2, period-4, and period-8 due to the first, the second, and the third period-doubling bifurcation, respectively, at the points D, E, and F shown in Fig. 5, where C = 950.2 μF, 605.2 μF, and 578.3 μF.

From the analysis above, it is shown that the valley voltage-mode controlled buck-boost converter with the variations of output-capacitor time-constant RC and C has the same routes from stable period-1 to period-2, period-4, and period-8 via the first, the second, and the third period-doubling bifurcation, entering into chaos and ultimately being an invalid state.

3.2. Jacobian and eigenvalues

By investigating the eigenvalues of the Jacobian around the fixed point, the stability of the equilibrium state can be determined. Furthermore, by studying the movements of the eigenvalues of the Jacobian under circuit parameter variations, the stability status such as the occurrence of bifurcations and boundaries of operating regimes can be identified.

The Newton–Raphson method or other numerical algorithms[20,21] can be used to obtain the eigenvalues of Jacobian at the fixed point XQ = [iL vo]T by putting xn+1 = xn = XQ. The Jacobian of the sampled-data model around the fixed point XQ can be derived as where J11 = ∂in+1/∂in, J12 = ∂in+1/∂vn, J21 = ∂vn+1/∂in, and J22 = ∂vn+1/∂vn.

From Eqs. (1)–(5), Jij in Eq. (6) are given as where Δ = in + iom2(CRC + toff).

The eigenvalues can be obtained by solving the characteristic equation in λ

Based on the sampled-data model described by Eqs. (1)–(5) and the Jacobian given by Eq. (6), the bifurcation and stability characteristics of the valley voltage-mode controlled buck-boost converter can be obtained by examining the movements of the eigenvalues under the variations of the specific parameters. The movements of the eigenvalues reveal the bifurcation phenomena and the way parameter variations affect the operation of the buck-boost converter with valley voltage-mode control. If all eigenvalues of the Jacobian evaluated at the fixed point are inside a unit circle, the buck-boost converter is stable. Any movements of the eigenvalues crossing from the interior of the unit circle to the exterior indicate a loss of stability of the crossing point.

Here, special attention should be paid to the loci under RC and C variation. Listed in Tables 1 and 2 are the eigenvalues for different values of RC and C respectively. Figure 6 illustrates the loci movements of the eigenvalues when (a) C = 940 μF and RC varies from 19.5 mΩ to 18.9 mΩ and (b) RC = 19 mΩ, and C varies from 960 μF to 930 μF. The arrows of the locus movements indicate the directions of movements of the eigenvalues with decreasing RC and C.

Table 1.

Eigenvalues for different ESRs.

.
Table 2.

Eigenvalues for different output capacitors.

.
Fig. 6. (color online) Movements of eigenvalues for valley voltage-mode controlled buck-boost converter: (a) loci of eigenvalues when C = 940 μF and RC varies from 19.5 mΩ to 18.9 mΩ; (b) loci of eigenvalues when RC = 19 mΩ and C varies from 960 μF to 930 μF. Arrows indicate the directions of movements of the eigenvalues with decreasing RC and C.
3.3. Maximal Lyapunov exponent

The two Lyapunov exponents of the two-dimensional system are represented as[22,23] and the maximal Lyapunov exponent is given by Here, eig(Jn Jn–1J1) is the function to obtain the eigenvalues of (Jn Jn–1J1) and max(λm1, λm2) is the function to acquire the maximal value between λm1 and λm2.

By using Eqs. (6), (8), and (9), the maximal Lyapunov exponents corresponding to Figs. 4 and 5 are shown in Figs. 7(a) and 7(b), respectively. From Fig. 7(a), the maximal Lyapunov exponent vanishes as RC ≤ 10.65 mΩ, the valley voltage-mode controlled buck-boost converter is invalid. As RC increases, the maximal Lyapunov exponent becomes positive from zero when 10.65 mΩ < RC ≤ 11.6 mΩ, the converter is in a chaotic state. Compared with the scenario in the bifurcation diagram, the maximal Lyapunov exponent has dramatic changes at some points where chaotic orbits jump to the periodic orbits, which indicates that multiple periodic windows exist in the chaotic region. At the points A, B, and C shown in Fig. 7(a), where RC = 19.2 mΩ, 12.25 mΩ, and 11.69 mΩ, respectively, the maximal Lyapunov exponent reaches zero from the negative, which means that the converter undergoes the period-doubling bifurcation, corresponding to the points A, B, and C of the bifurcation diagram shown in Fig. 4.

Fig. 7. (color online) Maximal Lyapunov exponents for valley voltage-mode controlled buck-boost converter when (a) RC varies from 10 mΩ to 20 mΩ and (b) C varies from 500 μF to 1000 μF.

Similarly, in Fig. 7(b), when C > 950.2 μF, 605.2 μF < C ≤ 950.2 μF and 578.3 μF < C ≤ 605.2 μF, the converter is in a periodic state with CCM operation, the maximal Lyapunov exponent becomes negative. When C = 950.2 μF, C = 605.2 μF, and C = 578.3 μF, the maximal Lyapunov exponent equals zero, which means that the converter exhibits a period-doubling bifurcation, corresponding to the points D, E, and F of the bifurcation diagram shown in Fig. 5. In the range of 526.7 μF–573.9 μF, the maximal Lyapunov exponent changes from negative to positive, which indicates the converter is in a chaotic state. When C ≤ 526.7 μF, the maximal Lyapunov exponent equals zero, the corresponding dynamical behavior of the buck-boost converter with valley voltage-mode control is invalid as shown in Fig. 4.

4. Boundary equations and operating-state regions
4.1. Boundary equations

According to the above discussion of bifurcation behavior in Subsection 3.1, it is found that the valley voltage-mode controlled buck-boost converter with the variations of RC and C presents the same dynamical behavior, including the transition from period-doubling bifurcation to chaos, and to invalid control finally.

In Subsection 3.2, it is mentioned that the variations of eigenvalues can evaluate the parameter range of the converter working in a stable state. When the eigenvalue of first iteration map xn+1 = f(xn) is equal to −1, the first period-doubling bifurcation occurs and drives the system entering into period-2 from period-1.

From Eq. (7), the eigenvalues λ1,2 of the characteristic equation for the sampled-data modeling can be expressed as Hence, by putting min(λ1, λ2) = −1 and substituting Eq. (6) into Eq. (10), the first period-doubling bifurcation boundary as well as the stability boundary τ1 can be yielded as Equation (11) indicates that the first period-doubling bifurcation of valley voltage-mode controlled buck-boost converter is influenced by output-capacitor time-constant (the product of RC and C), input voltage vg, output voltage vo, inductance L, load resistance R, and the switching period Ts. The system will operate in the stable period-1 region when the left part of Eq. (11) is greater than the right one, i.e., the stable region is located above the boundary τ1.

While the converter is transferred from steady period-1 state to period-2 state, the second iteration map xn+2 = f(f(xn)) = f(2) (xn) is existent. Similarly, making xn+2 = xn = XQ2, the corresponding fixed point XQ2, the Jacobian and the eigenvalues are obtained. The second period-doubling bifurcation boundary τ2 can be obtained by letting , equal −1. Using the same method can obtain the third period-doubling bifurcation boundary τ3.

As shown in Fig. 3(b), when the valley voltage-mode controlled buck-boost converter operates in OFF state, the variation of vs is positive for the duration of nTs ∼ (nTs + toff). If the variation of vs is negative for the duration of nTs ∼ (nTs + toff), the comparator cannot set the latch and the driver signal vQ is always at a low level, S1 is turned off in the whole switching period. In this condition, the convertor cannot work and the control technology is invalid. According to vs = Kvvo = Kv(vC + vESR), we have

Supposing that the peak inductor current is Imax during the n-th switching cycle, the inductor current at t1 can be found to be

The output capacitor current iC = −(iL + io) when S1 is turned off, therefore the variations of vC and vESR during nTs ∼ (nTs + t1) are

For the buck-boost converter in CCM, the peak inductor current can be expressed as Combining Eqs. (12) and (14)–(16), equation (12) can be rewritten as Because Kv is positive and vo is negative, by considering t1 = toff in the whole OFF state, equation (17) can be simplified into From Eq. (18), we know if the valley voltage-mode controlled buck-boost converter can work normally during one switching cycle, the output-capacitor time-constant must satisfy Eq. (18). Otherwise, the valley voltage-mode control of the buck-boost converter will be invalid. Thus, we can obtain the invalidated boundary τ4 The valley voltage-mode control for the buck-boost converter is invalid when the left part of Eq. (19) is less than the right one. The region located below the boundary τ4 which should be avoided in practical design, is called a forbidden region.

4.2. Operating-state regions

Considering the output-capacitor time-constant parameters with RC = 10–20 mΩ and C = 500–1000 μF, and the other circuit parameters are chosen to be the same as those in Fig. 4, the operating-state regions of RC and C can be obtained from Fig. 8. Figure 8(a) shows the operation spaces divided by boundary equations. The two-parameter dynamic behavior distribution divided by the sampled-data model as shown in Fig. 8(b), where the higher periodicities are depicted in grey color, the white area means low period, and the darkest shade refers to chaos.

Fig. 8. (color online) Operation spaces varying with output-capacitor time-constant CRC: (a) regions divided by boundary equations; (b) regions divided by sampled-data model.

Comparing Fig. 8(a) with Fig. 8(b), it is easily noted that the operation spaces divided by boundary equations are consistent with the regions divided by the sample data model. There exist five operation state regions, i.e., CCM period-1, CCM period-2, CCM period-4, CCM period-8 and chaos, and the forbidden region, of the valley voltage-mode controlled buck-boost converter over the parameter space which can be divided by four stability boundaries τ1τ4. The CCM period-8 and chaotic region contains a CCM period-8 region which is so weak that the change of grey level is obscure.

By choosing a horizontal line with C = 940 μF in Fig. 8(b), the bifurcation process will be obtained when the RC varies among 10 mΩ–20 mΩ, which is shown in Fig. 4. In the same way, the bifurcation process while the C varies among 500 μF–1000 μF will be obtained via a vertical line with RC = 19 mΩ in Fig. 8(b), the corresponding bifurcation diagrams are observed from Fig. 5. The period bifurcation points AC in Fig. 4 and DF in Fig. 5 are in complete agreement with the results in Fig. 8, which verifies the correctness and feasibility of operating spaces.

As shown in Fig. 8, the different operation state regions of the converter can be demonstrated clearly. It is found that the system is stable when the output-capacitor time-constant CRC is relatively large, unstable and even invalid when CRC is small. Using such operation spaces the designer can place the nominal operating point away from behaviour boundaries, especially from the forbidden region.

5. Simulation and experimental results
5.1. Simulation waveforms

Time-domain waveforms and phase portraits are usually used to visualize nonlinear phenomena, which can also be used to validate the correctness of the sampled-data model. According to Figs. 2 and 3(a), a circuit simulation model is built by using PSIM software. The circuit parameters are chosen to be the same as those in Fig. 4. Figure 9 shows simulation results of the time-domain waveforms (steady-state output voltage, sensed output voltage, and inductor current) and the phase portraits (inductor current versus output voltage) of the valley voltage-mode controlled buck-boost converter with different ESRs.

Fig. 9. (color online) Simulation results of valley voltage-mode controlled buck-boost converter with different ESRs: (a) RC = 20 mΩ; (b) RC = 16 mΩ; (c) RC = 11 mΩ.

By substituting the circuit simulation parameters into Eqs. (11) and (19), two critical output capacitor time-constants of the converter state transiting can be gained. The converter will move from steady period-1 state to period-2 state at RC = 19.15 mΩ calculated from Eq. (11). The valley voltage-mode controlled buck-boost converter is invalid at RC = 10.64 mΩ calculated from Eq. (19). When RC = 20 mΩ > 19.15 mΩ, which means that the output capacitor time constant is greater than the right part of Eq. (11), the valley voltage-mode controlled buck-boost converter operates in CCM and stable period-1 state as shown in Fig. 9(a1), the output voltage of the buck-boost converter is negative and looks piece-wise linear. From Figs. 9(b1) and 9(c1), it can be seen clearly that the converter is in period-2 state with RC = 16 mΩ, and in chaos with RC = 11 mΩ separately. Their corresponding phase portraits are shown in Figs. 9(a2), 9(b2), and 9(c2). It can be seen that the stable state is transformed into an unstable state by using a relatively small output capacitor ESR, which is consistent with the bifurcation diagram in Fig. 4 and the operation space distribution in Fig. 8.

To validate the invalid region mentioned in bifurcation diagrams, valley voltage-mode controlled buck-boost converters with different ESRs, which operate in CCM period-1 and invalid state, are further investigated. The time-domain simulation waveforms of output voltage and inductor current under ESR variation are shown in Fig. 10. The system operates in CCM period-1 with RC = 20 mΩ for the durations of 20 ms–30 ms. At time t = 30 ms, through changing RC = 10.5 mΩ (which is smaller than 10.65 mΩ in Fig. 4, and also located in the forbidden region in Fig. 8(a)), the inductor current descends to zero quickly while the output voltage is gradually raised to zero, which means that the valley voltage-mode controlled buck-boost converter is invalid and thus the converter does not work.

Fig. 10. Simulation waveforms of output voltage and inductor current under ESR variation.
5.2. Experimental verification

To verify the theoretical analysis and simulation results given earlier and to illustrate the effect of output capacitor time-constant, the experimental studies of the voltage-mode controlled buck-boost converter are performed with the same circuit parameters: vg = 4 V, Vref = 0.602 V, Kv = 0.1, L = 40 μH, C = 940 μF, R = 6 Ω, and Ts = 20 μs. A photograph of the experimental set-up is shown in Fig. 11, where switch S1 is the N-channel power MOSFET (IRF740), switch S2 is the diode (ES4J), the comparator is high-speed LM319, the flip-flop latch consists of 74LS02, and the control signals are generated by the driver A3120, respectively.

Fig. 11. (color online) Photograph of the experimental set-up.

Figure 12(a) shows that for RC = 20 mΩ, the converter operates in the period-1 state. For RC = 16 mΩ and RC = 11 mΩ, the converter operates in the period-2 state and chaos, as shown in Figs. 12(b) and 12(c). These experimental results are consistent with the theoretical analysis and simulation results. In hardware experiment, at the moment when the switch turns on or off, there are some spikes existing in the output voltage due to many parasitic parameters in circuit components.

Fig. 12. (color online) Experimental results of valley voltage-mode controlled buck-boost converter with different ESRs: (a) RC = 20 mΩ; (b) RC = 16 mΩ; (c) RC = 11 mΩ.
6. Conclusions

The valley voltage-mode controller for the buck-boost converter is proposed in this paper. A simple sampled-data model is established to investigate the effect of output-capacitor time-constant on its dynamics through dynamical analysis. Based on the sampled-data model and eigenvalue of the corresponding characteristic equation, the bifurcation diagrams, the movement of eigenvalues, the Lyapunov exponent spectrum, the boundary equations, and the operating-state regions of output-capacitor time-constant CRC are obtained. With the RC and C gradually decreasing, the buck-boost converter in CCM has the same route from stable period-1 to period-2, period-4, and period-8 via the first, the second, and the third period-doubling bifurcation, respectively, entering into chaotic state and ultimately being an invalid state. The stability boundary and the invalidated boundary are derived theoretically through stability analysis, where the stable state of valley voltage-mode controlled buck-boost converter can transit to an unstable state, and the converter can shift from operation region to forbidden region. Research results show that the time constant of the output capacitor is a critical factor for the voltage-mode controlled buck-boost converter, which has an important influence on the dynamics as well as control stability of voltage-mode controlled buck-boost converter. These results verified by time-domain waveforms and phase portraits from both simulation and experiment indicate that the sampled-data model is correct. The results in this paper can provide useful guidance for designing the buck-boost converter with valley voltage-mode control.

Reference
[1] Redl R Sun J 2009 IEEE Trans. Power Electron. 24 2669
[2] Schweitzer B P Rosenstein A B 1964 IEEE Trans. Aerospace and Electron. Syst. 2 1171
[3] Wang J P Bao B C Xu J P Zhou G H Wen H 2013 IEEE Trans. Ind. Electron. 60 1759
[4] Zhou G H Xu J P Wang J P 2014 IEEE Trans. Ind. Electron. 61 1280
[5] He S Z Zhou G H Xu J P Bao B C Yang P 2013 Acta Phys. Sin. 62 110503 in Chinese
[6] He S Z Zhou G H Xu J P Wu S R Chen L 2014 Acta Phys. Sin. 63 130501 in Chinese
[7] Zhou G H He S Z Chen X Cui H F 2014 Electron. Lett. 50 627
[8] Shi G D Zhang H M Bao B C Feng F Dong W 2015 Acta Phys. Sin. 64 010501 in Chinese
[9] Xie F Yang R Zhang B 2011 IEEE Trans. Circuits Syst. I, Reg. Papers 58 2269
[10] Bao B C Zhou G H Xu J P Liu Z 2011 IEEE Trans. Power Electron 26 1968
[11] Deivasundari P S Uma G Poovizhi R 2013 IET Power Electron. 6 763
[12] Wang F Q Ma X K 2013 Chin. Phys. 22 120504
[13] Zhou Y F Chen J N Iu H H C Tse C K 2008 Int. J. Bifurc. Chaos 18 121
[14] Zhao Y B Zhang D Y Zhang C J 2007 Acta Chin. Phys. 16 933
[15] Wang F Q Zhang H Ma X K 2012 Chin. Phys. 21 020505
[16] Zhou G H Xu J P Bao B C Zhang F Liu X S 2010 Chin. Phys. Lett. 27 090504
[17] Banerjee S Parui S Gupta A 2004 IEEE Trans. Circuits Syst. II 51 649
[18] He S Z Zhou G H Xu J P Wu S R Yan T S Zhang X 2014 Acta Phys. Sin. 63 170503 in Chinese
[19] Fang C C 2001 IEEE Trans. Power Electron. 16 345
[20] Banerjee S Chakrabarty K 1998 IEEE Trans. Power Electron. 13 252
[21] Xie F Yang R Zhang B 2010 Acta Phys. Sin. 59 8393 in Chinese
[22] Chen Y F Tse C K Qiu S S Lindenmuller L Schwarz W 2008 IEEE Trans. Circuits Syst. I, Reg. Papers 55 3335
[23] Bao B C 2013 An Introduction to Chaotic Circuits Beijing Science Press in Chinese 164