Dynamic Analysis of the Fixed-Frequency PWM LCC-Type Parallel Resonant Converter using Discrete Time Domain Modeling Vivek Agarwal Dept. of Electrical Engineering, Indian Institute of Technology, Bombay, India 400 076. A.K.S. Bhat, Senior Member, IEEE Dept. of Electrical and Computer Engineering, Univ. of Victoria, B.C., Canada V8W 3P6. Abstract - Dynamic analysis of the fixed-frequency pulse width modulated LCC-type parallel (or series-parallel) resonant converter is presented using discrete time domain modeling. First, the large-signal state-space model is derived and used to study the large signal behavior of the converter in the event of large, non-linear transient conditions. The results are verified using SPICE. The steady-state solution is obtained. Finally, the large signal equations are linearized about the steady-state equilibrium point to derive a linear small-signal model, which is used to study the shall signal dynamic behavior of the converter. I. INTRODUCTION Resonant Converters have been studied extensively by many authors for over one decade. Series-parallel resonant converter (SPRC)[Fig. I], has i n particular been more popular because of its advantages over the other configurations. Traditionally, the power control in resonant converters h a s been done by varying the operating frequency of the converter. Although this technique is very popular, it has many drawbacks [l]. To overcome these drawbacks, att,empt,s were made to propose alternate control schemes based on a “fixed” o p erating frequency. Thus came the idea of fixed frequency pulse-width modulated resonant, converters. Many techniques have been proposed for the fixed frequency operation. The most popular technique is based on applying phase shifted gating signals to the full bridge converter. Some authors [l, 21 have reported the fixed frequency operation and i t s analysis for the SPRC. Reference [l]describes the fixed-frequency PWM control of the SPRC. It is shown that high efficiency for large load variations is achievable along with a narrow range of duty-cycle ratio control and protection against load short circuit conditions. A simple analysis and design procedure based on complex ac circuit theory w a s presented in [l]. However, this analysis uses only fundamental components of voltages and currents. This type of analysis loses its accuracy at reduced pulse widths and can not predict the different modes 0-7803-3500-7196/$5.00 0 1996 IEEE of operation of the converter. In [2], state-plane technique has been used to analyse the third order SPRC, which has been reduced to a second order system (by transformation of variables) so that a two dimensional state-plane analysis can be applied. Although this technique is very useful, the transformation of the third order system t o a second order system, makes it difficult to understand how the various circuit variables are behaving physically. Dynamic analysis is very important. The large-signal analysis determines the response of the converter when its operating conditions undergo lorge variations in their steady-state values and is therefore useful for choosing appropriate component ratings. Similarly the small-signal analysis determines the converter’s response to small perturbations in its steady-state values and is required to design the closed loop around the converter. The large-signal analysis of the SPRC, operating in fixed frequency, pulse-width modulated manner is not available in the literature. The small-signal analysis, using an approximate extended describing function method was reported in [9]. Recently, a combination of extended describing function method and state-space approach has been used [8] to perform the small-signal analysis. In this paper, this analysis h a s been carried out using state-space technique. The main objectives of this paper are the following: (1) To present a large-signal analysis using a discrete time domain model. (2) To present a small-signal analysis by linearizing the large-signal equations about a steady-state operating point. 11. CONVERTER OPERATION, DESIGN A N D TERMINOLOGY A. Operation Phase-shifted gating signals are used to generate a pulsewidth-modulated quasi-square wave across a and b. The gating signals for the switch pair S3 & S4 are phase shifted 272 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on November 5, 2008 at 00:09 from IEEE Xplore. Restrictions apply. + a combination of two events, called the kth and ( k events [3]. Each of the events is composed of three subevents or the intervals marked in Fig. 2 as intervals T2, T1 and A , respectively. These intervals are defined by the following relations. Figure 1: The full bridge version of the SPRC suitable for fixed frequency operation. Phase-shifted gating signals are used to generate a pulse-width-modulated quasi-square wave across a and b. with respect to S1 & S2, to obtain the desired zero voltage interval ( T a l t )in v,b. It should be noted that this zero voltage interval corresponds to the duration for which one switch and anti-parallel diode on the other arm in the upper limbs (or the lower limbs) are ON simultaneously. The output voltage regulation for load (or line) variations is obtained by adjusting Talt. Figure 2: Typical waveforms of the F F PWM SPRC circuit starting from the kth instant. The various intervals are marked. B. Design of the Converter I n order to identify various modes of operation, initially the design method given in [l] is used. Design is done for the worst case loading conditions i.e. for maximum load current with minimum input voltage. At the rated design conditions given below, the converter operates with full pulse width (i.e. Talt = 0) in lagging power factor mode and on the boundary of CCVM and DCVM. As the load current varies due to change in load resistance, the phase shift between gating signals is changed so as to maintain the constant, rated value of the output voltage. The SPRC designed has the following specifications: Input supply voltage, Vsmin(= 2 E ) = 50 V. Output voltage of the converter, Vd = 24 V. Output voltage ripple, V&p-p)= ? 0.025 V. Output current ripple, ZL-p = t 0.010 A. Maximum output power, Po = 100 W a t t s . Switching frequency, ft = 200 k H z . The design values obtained are: L = 17.74pH ; C, = 0.047pF ; Ct = n2Ci = 0.047pF. The rated load resistance is, Ri = 5.76 $2. It should be noted that Yk,the half period of the operating cycle, is a constant and hence the subscript "k" will not be used. A careful inspection reveals that the case is analogous to the variable frequency case with (& replacing Y k as the controlling parameter [6]. The input supply to the tank circuit is assumed constant either at its clamped value of zero or +_ Ek, for a given event k. The ripple i n its value is considered negligible as compared to the large step changes it makes at iiistants t 2 ( k ) and t O ( k + l ) (low ripple approximation [3]) The same approximation is applied for the load current also. 111. DISCRETESTATESPACEMODEL The first step in arriving at the model is to write the large signal equations representing the converter, during the various intervals of the first half of the operating cycle (called kth event). The subscript "k" is added to give a discrete time domain interpretation. C . Terminology 1) kth Eoent: Interval T2; t O ( k ) < f < fl(k) Fig. 2 shows typical waveforms for one of the predominant modes of the FF PWM SPRC in discrete time domain. The converter operation over one cycle has been represented as 273 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on November 5, 2008 at 00:09 from IEEE Xplore. Restrictions apply. + +cflzk A 3 ~ 2 * (1 ~ = uct(t') (6) uc8(t0(k)) + ~ ~ ( w o t 'B)~) T z ~ . s ~ ~ ( w o ~ ' ) +C3T2r (wOt') + (7) uct ( t O ( k ) ) Equations for the other two intervals, Ti and A , are of the same form as interval Tz. For example, equation for the parallel capacitor voltage corresponding to interval Ti is given below: = vct(t") A3Tik ( 1 - cos(w0t")) + &TIk +C3Tlr, .sin(wot") +uct(tl(k)) (8) where t" = t - t l ( k ) . All the remaining equations and coefficients A 1 ~ ,2B1Tzk ~ etc. are defined in [lo]. 2 ) (IC + l)'h Event: The (IC + l ) t h interval circuit equations are identical to those of ICth interval with a change in the signs of all the variables. 3) Selection of Discrete State Variables: The following discrete state variables are chosen (corresponding t o the storage elements in the circuit) for the ICth and (IC l)'* events. + 21(k) ZZ(k) = -vct(to(k)); Zl(k+l) = v c s ( t O ( k + l ) ) ; 23(k+l) Z3(k) Z2(k+l) = -iL(tO(k)); = -"cs(tO(k)) = iL(tO(k+l)) Vct(tO(k+l)) (9) (10) (11) For the output section, two additional discrete state variables are introduced for the ICt'' event. 4) Formulation of the Model: By using the final values of an interval as the initial values of the immediately occurring interval, it is possible t o express tlie initial values ofthe ( k + l ) t hevent in terms of the initial values of the ICth event. Inserting the state variables defined by (9) - ( l a ) , in the resulting equations, the following discrete state-space model is obtained; Iv. RESULTSOF LARGES I G N A L ANALYsI s THE The discrete time domain model described in section 111 was used in predicting the transient behavior of the designed converter operating under open loop conditions. PRO-MATLAB was used t o solve the discrete time domain equations on tlie computer. These results are also verified using SPICE software. Results obtained for different transient conditions are explained next. 1) Sudden Switching ON of the S u p p l y Voltage: In this section, the effect of sudden switching ON of the supply voltage is studied. An operating condition is considered, corresponding to half the rated load current and rated output voltage. The corresponding duty ratio is x 70%. Figs. 3(a) and (b) show the plots of the tank current and the parallel capacitor voltage obtained for this step change using the large-signal model. Figs. 3(c) and (d) sliow the corresponding SPICE plots. Fig. 4(a) shows the peak component stresses. Figs. 4(b) and (c) show the plots of tank state variables and output state variables (current and voltage) as obtained with the model. Fig. 4(c) also shows the corresponding SPICE plot for the switch ON transient conditions. 2) Step Change in Load: In this section the effect of a sudden step change in load is considered. The converter is operating at half the rated load condition with 70 % duty cycle, when the load changes to quarter of rated load in a step manner. For the open loop condition, the duty cycle remains at 70 %. These results are explained next. Fig. 5 shows the plots of the tank current and tlie parallel capacitor voltage obtained for this step change using the large-signal model and SPICE. Fig. G(a) shows the peak component stresses. Figs. 6(b) and (c) show the plots of 274 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on November 5, 2008 at 00:09 from IEEE Xplore. Restrictions apply. tank state variables and output state variables (current and voltage), as obtained with the model. Fig. 6(c) also shows the corresponding SPICE plot for the switch ON transient conditions. 3) Closed Loop Example: In this section, a typical example of closed loop operation is considered. The converter is operating at half the rated load with rated output voltage value and 70 % duty cycle, when the load suddenly changes to quarter the rated load in a step mahner. However, unlike in the previous section, where the open loop operation was considered (and so the duty ratio was not altered), the duty ratio is altered to regulate the output voltage (as a result of closed loop operation) to x 63 %. The results obtained for this case are discussed next. It is assumed that the closed loop action is instantaneous. Fig. 7 shows the plots of the tank current and the parallel capacitor voltage obtained using the large-signal model and SPICE. Fig. 8(a) shows the peak component stresses. Figs. 8(b) and (1‘ show the Plots Of tank state and Output state variables (current and voltage) as obtained with the model. Fig. 8(c) also shows the corresponding SPICE plots. V. SMALL-SIGNAL ANALYSIS OF FF PWM SPRC In this section the small-signal modeling of FF PWM SPRC is performed using the large-signal model obtained in section 111. As was stated earlier, the small-signal analysis is concerned with the response of the converter to small perturbations in its steady state operating conditions. A. Linearization of SPRC About the Equilibrium Point First the equilibrium point of operation is obtained by applying the symmetry condition q ( k ) = z i ( k + l ) to (13) (17) [lo]. The large-signal state-space model represented by (13) - (17) has the following general form: where “f,” represents some non-linear function of the relevant independent specified on the right hand side of the above equation and i = 1 . . .5. A a i^\ -7 -_ -- - - J P . * P - I ~ D L I Y ~ . ’$ __;O do W 7%” --_-- 1 “ d ;1 -Y E IEi -- 2x J /--l i T mn 7 2 7igure 3: (a) Resonating inductor current and (b) parallel .apacitor voltage, for step change in input voltage supply iom OV to 25V at half the rated load conditions. In all the ,lots the step change occurs at the origin. (c),(d): SPICE dots corresponding to (a) and (b). j _ _ k 2 w *Marus r------- ““‘WTUBI ---~---_-__ I $1 a m x o CI 8 0 7 ~ T-nUmrrm Figure 4: (a) Peak component stresses (P.u.), (b) Tank discrete state variables and (c) Output discrete state variables, for step change in input supply. In (c), SPICE results are also plotted for the sake of comparison. Note that all these plots have been obtained with discrete set of poillts obtained one per half cycle, and connected to give an over all continuous picture. 275 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on November 5, 2008 at 00:09 from IEEE Xplore. Restrictions apply. These equations can be linearized using the Taylor's expansion about the equilibrium point resulting in five linearized small-signal equations represented in the matrix form as below: 01.1 = m.cos(y) +[" ". I + r2.( 1 - 2cos(P + q ) ) 8"4(k) a21 aak (25) eq + [-- = -Z.rl.sin(y) a"l(&)a a k The elements aijs and 6,js of matrices A and B can be determined using the following: a22 = -ri.cos(y) + r1 - 1 + . eo (26) [*.e] (27) az2(k) a a k eo As an example, some of these elements are given below. Others have also been derived and are given in [IO]. a11 = -cos(-/) + .s 5/ , . , , , , , , , . . . . . . , , Figure 5: (a) Resonating inductor current and (b) parallel capacitor voltage, for step change in load from half the rated load to quarter the rated load. (c),(d): SPICE plots corresponding to (a) and (b). Figure 6: (a) Peak component stresses (P.u.); (b) Tank ais-. Crete State Variables and (c) Output discrete state variables for step change in load from half the rated load to quarter the rated load. 276 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on November 5, 2008 at 00:09 from IEEE Xplore. Restrictions apply. + 2 . r l . a ) .dok -dak - - -(r2.Z.sin(a) D ' -dEk =o ax4(k) where D = ( z 2 + z 3 ) s i n ( a )+ Z.(-zl + +r~.+4)cos(rr) Z.rl.x4 = a25 = a35 = agl = a52 = a53 = 0 b52 = 0 b13 = b23 = b33 = b43 = b51 The partial derivatives involved in the above equations are evaluated next. For this, the parallel capacitor voltage is equated to zero at the end of interval B, which leads to the following equation: rZ.(XZ(k) (44) dfz 8ak af3 + x 3 ( k ) ) . ( l - c o s ( a k ) )+ Z.%'.(-Xl(k) + Z.rl . r Z . z q ( k ) . a k - z 3 ( k ) = @g) Then, using ( 3 9 ) , the partial derivatives (evaluated at the equilibrium point) obtained are given by: I \ , ...b -2 m s sa io Tmn- ,m A ( b) ,;o (40) ,a A +iF--G (46) (41) r2.D ,w + +2 . ~ 1 = Z.rz.z4[2.r2.cos(q p) (45) Equation ( 1 9 ) , with coefficients of matrices A and B given by ai, and b,, , respectively, represents the small signal model of the FF PWM SPRC in matrix form. This model can be used to predict the small-signal behavior of the F F PWM SPRC. I dx3(k) = Z.rl.r2.x4[2cos(P+ q ) - 21 aak +r2.24(k))sin(ak) 80, - --Z.Sin(CY). dak - -(I - COS(&)) -axl(k\ D 'azZfk, D 8ak 1 - r z . ( l - cos(a)) -- (43) Other partial derivatives are obtained from ( 1 3 ) - (17). For example: (37) (38) a15 (42) sa *, hh- ,* A *A -M A k (d) Figure 7: (a) Resonating inductor and (b) Parallel capacitor voltage for closed loop operation example where the duty cycle changes in response to a step change in load from half the rated load to quarter the rated load. (c),(d): SPICE plots corresponding to (a) and (b). Figure 8: (a)Peak component stresses (P.u.); (b) T a n k discrete state variables and (c) output discrete state variables, for closed loop operation example where the duty cycle changes in response to a step change in load from half the rated load to quarter the rated load. In (c), SPICE results are also plotted for the sake of comparison. 277 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on November 5, 2008 at 00:09 from IEEE Xplore. Restrictions apply. VI. DYNAMIC PERFORMANCE PARAMETERS The dynamic performance parameters like the control t o output gain, audio susceptibility and output impedance, which are useful in evaluating the dynamic behavior of the converter are determined in this section. These parameters can be used to design suitable cofipensation for the closed loop system. Fig. 9(a) shows the plot of control to output transfer function as obtained with the help of the proposed model. Fig. 9(b) shows the corresponding plot for the audio-susceptibility transfer function. The output impedance transfer function plot is shown in Fig. 9(c). 50 VII. CONCLUSIONS A discrete state-space large signal model has been presented and used to predict the response of the converter t o large signal disturbances. The large-signal analysis provided the following results. During switch ON conditions, the converter enters leading p.f mode of operation before reaching the steady state where it is designed to o g erate in lagging p.f. mode. All the tank components are over stressed during the switch ON transients. The closed loop operation during the step change in load, causes reduced component stresses. All the results obtained with the proposed model, have been verified using SPICE. A small-signal model has also been obtained by linearizing the large signal state space equations. More results will be presented in a future paper. I REFERENCES A.K.S. Bhat, Fized Frequency P WM Series-Pamllel Resonant Converter,” IEEE IAS, Oct. 1989, pp. 1115-1121. C.Q. Lee, S. Sooksatra and R. Liu, Constant Frequency Controlled Full Bridge LCC Type Resonant Converter, IEEE APEC-91 pp. 587-600. 0 R.J.King and T.A. Stuart, “A large-signal Dynamic Simulation for the Series Resonant Converter, ” IEEE Trans. on Aerospace and Electronic Systems, Vol. AES-19, No.6, Nov. 1983. V. Vorperian and S. Cuk, “Small Signal Analysis of Resonant Converters, ” IEEE Power Electromcs Specialist Conference, PESG83, pp. 269-282. M.G. Kim, J.H. Lee, J.H. Koand X.J. Y o u , “ A Discrete m . ’-1 . Y Time Domain Modeling and Analysis of Controlled Pamlle1 Resonant Converter, ” IEEE Power Electronics Specialist Conference, PESC-91, pp. 730-736. Vivek Agarwd and A.K.S. Bhat, “Small Signal Analysis of the LCC- Type Parallel Resonant Converter using Discrete Time Domain Modeling,” IEEE PESC-199-1, Vol. 2, pp. 805-813. Vivek Agarwal and A.K.S. Bhat, “Large Sigrid Analysis of LCC- Type parellel resonant converter using discrete trmc domain analysis,” IEEE Trans. Power Electron., Vol. 10. no. 2, pp. 222-238, Mar. 1995. IO‘ Id -- J.L. Duarte and J.M.A. Willaert, “A fully digitized phaseshifl modulated LCC resonant converter,” IEEE PESC 1994, Vol. 2, pp. 1303-1308. E.X. Yang, F.C. Lee and M.M. Jovanovic, uSmall-signal Modeling of LCC Resonant Converter,” IEEE PESC-1992, pp. 941-948. Id cc 1 J Figure 9: Results of the small-signal analysis: (a) The plot of control t o output transfer function. (b) Plot of audiosusceptibility. (c) Plot of output impedance transfer function as obtained with the small-signal model obtained in this paper. Vivek Agarwal, “Steady-State and Dynamic Analysis of the LCC-Type Pamllel Resonant Converter, .,PhD dissertation, University of Victoria, Canada, Oct. 1994. 278 Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on November 5, 2008 at 00:09 from IEEE Xplore. Restrictions apply.