Inter national J our nal of Electrical and Computer Engineering (IJECE) V ol. 7, No. 5, October 2017, pp. 2374 2381 ISSN: 2088-8708 2374       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     Accurate Symbolic Steady State Modeling of Buck Con v erter Ek o Setiawan 1 , T akuya Hirata 2 , and Ichijo Hodaka 3 1 Interdisciplinary Graduate School of Agriculture and Engineering, Uni v ersity of Miyazaki, Japan 2 Department of Electronic Mechanical Engineering, National Institute of T echnology , Oshima Colle ge, Japan 3 Department of En vironmental of Robotics, Uni v ersity of Miyazaki, Japan Article Inf o Article history: Recei v ed: Apr 6, 2017 Re vised: Jun 9, 2017 Accepted: Jun 23, 2017 K eyw ord: Steady state analysis F ourier series Switching circuit Buck con v erter ABSTRA CT Steady state analysis is fundamental to an y electric and electronic circuit design. Buck con v erter is one of most popular po wer electronics circuit and has been analyzed in v arious situations. Although the beha vior of b uck con v erters can be understood approximately by the well-kno wn state space a v eraging method, little is kno wn in the sense of detailed beha vior or e xa ct solution to equations. In this paper a steady state analys is of b uck con v erter is proposed which allo ws the e xact cal culation of steady state response. Our e xact solution is e xpressed as a F ourier series. Our result is compared with numerical calculation to be v erified. Our method copes with more complicated problems such as describing a v erage po wer and root-mean-square po wer that are most critical issues in po wer electronics circuit. Copyright c 2017 Institute of Advanced Engineering and Science . All rights r eserved. Corresponding A uthor: Ichijo Hodaka Department of En vironmental Robotics, Uni v ersity of Miyazaki 1-1, Gakuen Kibanadai-nishi, Miyazaki, 889-2192, Japan Phone: +81 985 587 352 E-mail: hijhodaka[at]cc.miyazaki-u.ac.jp 1. INTR ODUCTION An analysis of steady-state re sponse of a system is important k e y in circuit design and control, included dc-dc con v erter . In con v ential m ethod, the steady state of dc-dc con v erter is assumed as the constant v alue. Man y researches based on state space a v eraging method and high switching frequenc y assumpt ion observ e steady state response [1, 2, 3, 4, 5]. The method gi v es simple w ay to analyze b ut some ripples are undescribed clearly . The po wer electronic handbook approximate linear ripple to analyze dc-dc con v erter more accurately [6]. The approximation may be correct if the switching frequenc y is high. Since there are some limitation in component, some high frequenc y is not al w ays reached. The importance of accurate steady-state analysis has already noticed in man y researc h e s [7, 8, 9, 10, 11, 12]. A significant part of the design of circuits requires the simulation of the steady-state response. P arameters such as the g ain, harmonic distortion and the input and output impedances are studied in the steady-state mode of operation [7]. Using con v entional time-stepping simulations and w aiting-time for possible steady state is often not practical because in most cases the time constants of the modes are much lar ger than the switching period [8]. In con v entional method of dc-dc con v erter analysis, steady state ripple v alues are ne gligible, compared to the steady state v alues themselv es. Switching po wer con v erters are inherently nonlinear and consequently it is v ery dif ficult to calculate the root-mean-square (RMS) v alues of the state v ariable ripple. These RMS v alues are important in order to calculate the current stresses of the dif ferent po wer con v erter de vices as well as to filter design in order to meet the gi v en specifications [9]. Though po wer electronic handbook [6] sho ws the RMS calculation using approximation linear ripple, the result is not absolutely correct due to linear approximation. In order to achie v e a high performance, proper design and control, it is necessary to ha v e an e xact model of con v erter [10, 11]. High accurac y is one of major features of a good modeling [11]. The dc -dc con v erter analysis can be classified into tw o cate gories, numeric and symbolic analysis. The numeric analysis is found in [7, 8, 10, 12, 13, 14, 15]. The numeric analysis observ es the system response by inputting para meter v alues i nto model. The numeric analysis needs some computation-time to sho w the output J ournal Homepage: http://iaesjournal.com/online/inde x.php/IJECE       I ns t it u t e  o f  A d v a nce d  Eng ine e r i ng  a nd  S cie nce   w     w     w       i                       l       c       m     DOI:  10.11591/ijece.v7i5.pp2374-2381 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2375 response. Relation between each parameter is not describe in the equation. The relation between parameter and output is al w ays observ ed by comparison between parameter change to output change. The [7] construct impedance or admittance matrix of dc-dc con v erter . The output response of the con v erter is calculated by Ne wton-Raphson. The steady-state output is solv ed per fix ed time-step (fix ed sampling interv al). The accurac y of analysis is dependent on time-step. Fe wer sampling point causes less accurac y . On the other side, more sampling points need more calculation time. The [13] substitute original circuit with periodically switched linear (PSL) circuit. The PSL can be observ ed by F ourier series. The paper use 110 as sequence number in F ourier series summation. The [14] also use F ourier series to simulate steady-state response. Comparing with [13], the [14] only uses 21 sequences. The [13] applies F ourier series of current switching part. The F ourier series of switching part substitutes original part to be analyzed by Kirchhof f V oltage and Current La w . The paper uses 180 sequence number to dra w a steady state response of system. The [12] analyzes b uck con v erter in frequenc y domain due to high accurac y comparing with con v entional time-domain. The paper simulate in three dif ferent sequences number that 0, 10 and 100. The paper sho ws that 10th sequences order is enough to describe steady state response. In F ourier series based method, greater number send the accurate steady-state model b ut it need more calculation. High order sequence number doesnt bring significant accurac y . Determination of the proper sequence number is another problem beside the main steady-state analysis problem. Contrary with numerical analysis, symbolic analysis describes relation between parameter and output in the equation. The relation between parameter and output can be observ ed roughly by the equation. The symbolic analysis is found in [9] and [11]. Symbolic analysis is complicated to be done although it sho w relation between parameter and output in equation. The [9] perform steady state symbolic analysis to calculate rms v alue. The paper sho w solution in a matrix form that is more complicated than o r dinary equation. The a v erage and rms calculation include term which is obtained from zero deri v ati v e state assumption. Since the ripple is e xist and cannot be ne glected, this assumption is contrast with early definition. The [11] solv e s teady-state equation by Laplace transform and re v ert back i nto time-domain by in v erse of Laplace transform. The solution need the kno wn initial v alue. The Z-transform is applied due to similarity between initial v alue and last v alue in one period. Though the paper sho w symbolic analysis, the proses is long enough due to calculation of three transformations (Laplace transform, Z-transform and in v erse of Laplace transform). In this paper , an alternati v e method is proposed that accurately predict and analyze the steady state of switching po wer con v erter . The proposed method based on F ourier series since man y references sho w the accurac y [12, 13, 14]. The reco v ery function is also proposed to generate analysis without dependent in number of sequences orders. The paper is subdi vided in se v eral sections to present clearly e xplanation. Section 2 sho ws the basic idea of proposed method. Section 3 discusses about implementation of proposed m ethod in b uck con v erter and the reco v ery function. Complete steady state output function of dc-dc con v erter is sho wn in this section. Section 4 v erify proposed steady state function by circuit simulation. Finally , the conclusion of this paper is declared in Section 5 2. PR OPOSED FOURIER SERIES METHOD Man y phenomena studied in electric circuit are periodic in nature when time goes to infinit y . These periodic functions in time t can be e xpressed as F ourier series as in the follo wing equation [16]. f ( t ) = 0 + X n 6 =0 n e j n! t = ::: + 2 e j 2 ! t + 1 e j ! t + 0 + 1 e j ! t + 2 e j 2 ! t + ::: (1) with F ourier coef ficients 0 ( f ) = 1 T Z T 0 f ( t ) dt; n ( f ) = 1 T Z T 0 f ( t ) e j n! t dt (2) where T is the period and ! is the angular frequenc y defined by ! = 2 T . In order to help understanding ho w F ourier series w orks, let us gi v e an input square-w a v e function as follo ws. u ( t ) = ( g (0 t < T d ) 0 ( T d t < T ) u ( t ) = u ( t + T ) (3) F ourier coef ficients of (3) are calculated as follo ws. 0 ( u ) = g d; n ( u ) = g j n! T 1 e j n 2 d (4) Accur ate Symbolic Steady State Modeling of Buc k Con verter (Ek o Setiawan) Evaluation Warning : The document was created with Spire.PDF for Python.
2376 ISSN: 2088-8708 Then the F ourier series of the equation (3) is written by equation (5). u ( t ) = g d + X n 6 =0 g j n 2 1 e j n 2 d e j n! t (5) An electric system can be described in a sim ple block diagram as sho wn in Figure 1. It has been e xplained that a periodic input has possibility to be analyzed with F ourier series. A periodic input for a transfer function G ( s ) will generate a periodic output. By kno wing its transfer function, the output al so can be analyzed into F ourier series. This paper analyzes steady state output in the equation (6). Buck con v erter can be modelled as a transfer function. Then as we e xplained the abo v e, b uck con v erter can be e xpressed as the equation (6). x ss ( t ) = G (0) 0 ( u ) + X n 6 =0 G ( j n! ) n ( u ) e j n! t (6) Figure 1. Block diagram Figure 2. Buck con v erter circuit 3. B UCK CONVER TER AN AL YSIS Buck con v erter circuit is sho wn in Figure 2. Principally , b uck con v erter is dri v en by tw o contrary switching ON and OFF . In continuous conduction mode (CCM), b uck con v e rter has only tw o modes. Each mode can be arranged as non-switching circuit by assuming switching part connected (ON) or disconnected (OFF). Figure 3 sho ws the tw o modes. Let us assume v oltage at switching parts is equal to input function u ( t ) as sho wn in Figure 4. Input function u ( t ) can be described as in the equation (3). (a) Mode-1 (b) Mode-2 Figure 3. Modes of Buck con v erter Based on Figure 4, let G ( s ) be a transfer function from the input v oltage u ( t ) to the output v oltage x ( t ) as in the follo wing equation. G ( s ) = 1 LC s 2 + 1 R C s + 1 LC (7) The transfer function (7) has tw o poles. Here we assume the poles are comple x-conjug ate pairs ( s = ! j ! ), that is, 1 R 2 C 2 4 LC < 0 : (8) W ith s = j ! , the frequenc y transfer function can be written as in the follo wing equation. IJECE V ol. 7, No. 5, October 2017: 2374 2381 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2377 Figure 4. Equi v alent circuit G ( j n! ) = 1 LC ( j n! ! ) 2 + ( ! ) 2 (9) Moreo v er , the transfer function at s = 0 is called a DC-g ain as follo ws. G (0) = 1 LC 0 2 + 1 R C 0 + 1 LC = 1 (10) In the pre vious section, F ourier series of square-w a v e function w as e xplained by the equation (5). Then, F ourier coef ficients of the input function (3) were calculated in (4). By assuming that q = j n and substituting (2), (9), and (10) into (6), the proposed equation of b uck con v erter can be e xpressed as follo ws. x ss ( t ) = g d + X n 6 =0   1 ! 2 LC ( q ) 2 + 2 ! g 2 q (1 e 2 q d ) e q ! t = g d + X n 6 =0 g 2 ! 2 LC  1 ( q ) 2 + 2  1 q (1 e 2 q d ) e q ! t (11) 3.1. Reco v ery function The infinite series of equation (11) can be represented as follo ws. f saw ( t ) = ! t; (0 t < T ) ; f saw ( t ) = f saw ( t + T ) f c ( t ) = 2 + 2 + e ( ! t 2 ) cos( ! t ) e ! t cos( ( ! t 2 )) cosh(2 ) cos(2 ) ; (0 t < T ) ; f c ( t ) = f c ( t + T ) f s ( t ) = e ( ! t 2 ) sin( ! t ) e ! t sin( ( ! t 2 )) cosh(2 ) cos(2 ) 2 + 2 ; (0 t < T ) ; f s ( t ) = f s ( t + T ) (12) F ourier coef ficients of reco v ery functions are calculated by (2). By assuming q = j n , F ourier series of proposed reco v ery function is described as follo ws. f saw ( t ) = X n 6 =0 n ( f saw ) e q ! t f c ( t ) = X n 6 =0 n ( f c ) e q ! t f s ( t ) = X n 6 =0 n ( f s ) e q ! t (13) where n ( f saw ) = 1 q , 0 ( f saw ) = 0 n ( f c ) = q ( q ) 2 + 2 , 0 ( f c ) = 0 n ( f s ) = ( q ) 2 + 2 , 0 ( f s ) = 0 (14) Accur ate Symbolic Steady State Modeling of Buc k Con verter (Ek o Setiawan) Evaluation Warning : The document was created with Spire.PDF for Python.
2378 ISSN: 2088-8708 3.2. T ime-delay function Let time-delay function be described as in the follo wing. ~ f ( t ) = f ( t T h ) (15) where T is periodic t ime and h is delay constant. F ourier coef ficients of time-delay function are simply described as n ( ~ f ) = e q 2 h n ( f ) : (16) 3.3. Pr oposed steady-state function W e can reco v er function of time e xpressed by the e xponential and trigonometric functions by using our reco v ery functions (13). The proposed steady state b uck con v erter equation can be also solv ed by the reco v ery functions. The summation part of equation (11) can be partially decomposed as follo ws. 1 ( q ) 2 + 2  1 q = 1 ( 2 + 2 ) 1 q + q + 2 ( q ) 2 + 2 = 1 2 + 2 1 q q ( q ) 2 + 2 + ( q ) 2 + 2 (17) In the ne xt step, we substitute (17) into (11). x ss ( t ) = g d + g 2 ! 2 LC ( 2 2 ) X n 6 =0 " 1 q q ( q ) 2 + 2 + ( q ) 2 + 2 (1 e 2 q d ) e q ! t # (18) Let us assume that m = g 2 ! 2 LC ( 2 2 ) ; = : (19) By assuming q = j n , the equation (18) may be written as in the follo wing. x ss ( t ) = g d + m X n 6 =0 1 q (1 e 2 q d ) e q ! t m X n 6 =0 q ( q ) 2 + 2 (1 e 2 q d ) e q ! t + m X n 6 =0 ( q ) 2 + 2 (1 e 2 q d ) e q ! t = g d + m 0 B B B B B @ X n 6 =0 n ( f saw ) e q ! t | {z } f saw ( t ) X n 6 =0 n ( f saw ) e 2 q d e q ! t | {z } f saw ( t T d ) 1 C C C C C A m 0 B B B B B @ X n 6 =0 n ( f c ) e q ! t | {z } f c ( t ) X n 6 =0 n f c e 2 q d e q ! t | {z } f c ( t T d ) 1 C C C C C A + m 0 B B B B B @ X n 6 =0 n ( f s ) e q ! t | {z } f s ( t ) X n 6 =0 n ( f s ) e 2 q d e q ! t | {z } f s ( t T d ) 1 C C C C C A (20) The equation (20) is easy to be understood if we use the reco v ery functions as (21). x ss ( t ) = g d + m f saw ( t ) f saw ( t T d ) m f c ( t ) f c ( t T d ) + m f s ( t ) f s ( t T d ) (21) IJECE V ol. 7, No. 5, October 2017: 2374 2381 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2379 The proposed equation (21) co v ers infinite series with the reco v ery functions. Furthermore, the calculation of the a v erage and RMS po wer are traceable by the reco v ery functions as follo ws accurately . P av g = 1 T Z T 0 ( x ss ( t )) 2 R dt P r ms = s 1 T Z T 0 ( x ss ( t )) 4 R 2 dt (22) (a) Complete response (b) Steady state response Figure 5. Comparison between SPICE and proposed analysis of parameter -1 4. SIMULA TION RESUL T This section v alidates our proposed function (21) result by comparing with SPICE (Simulation Program with Inte grated Circuit Emphasis). Numerical parameter of dc-dc con v erter is determined as sho wn in T able 1. Numerical parameter of resistance ( R ), inductance ( L ) and capacitance ( C ) gi v e comple x-conjug ate poles. Numerical calculation utilized mathematical softw are to plot the proposed steady state response. SPICE simulates the actual circuit responses. Complete response of capacitor v oltage by parameter -1 is sho wn in Figure 5a. Numer - ical calculation of proposed method is plotted in solid-line while SPICE result in dashed-line. The magnification of steady state response is sho wn in Figure 5b. Comparison between numerical calculation of proposed function and steady state of SPICE result has similarity in shape and v alue. The other complete response of circuit simulation using parameter -2 is sho wn in Figure 6a and steady state of SPICE and proposed result is descr ibed Figure 6b. Figure 5b and 6b describe that numerical calculation of proposed methods gi v e consistent result with SPICE result. The proposed method has adv antage in obtaining steady state response without w aiting transient time. Accur ate Symbolic Steady State Modeling of Buc k Con verter (Ek o Setiawan) Evaluation Warning : The document was created with Spire.PDF for Python.
2380 ISSN: 2088-8708 T able 1. Numerical parameter V ariable P arameter -1 P arameter -2 Resistance (R) 6.35 1.81 Inductance(L) 100 H 285 H Capacitance (C) 62.7 F 21.9 F T ime-period (T) 50 s 20 s V oltage source (g) 10 V 15 V Duty-ratio (d) 0.5 0.5 Pole (s) -0.0101 j0.1 -0.0402 j0.0034 5. CONCLUSION W e ha v e sho wn that the proposed method describes steady-s tate response directly without calculating transient response. It gi v es e xact calculation and symbolic complete solution of steady state output. T ransition between each mode is desc ribed clearly by proposed method. Reco v ery function gi v es an accurate solution of F ourier series without depending on numerical calculation of summation. Proposed steady state analysis of b uck con v erter has been clarifie d by comparing with SPICE. Moreo v er , proposed method mak es calculation of po wer traceable. REFERENCES [1] R. D. Middlebrook, S. ´ Cuk, A General Unified Approach to Modelling Switching Po wer Con v erter Stages, IEEE PESC Record , pp. 18-34, 1976. [2] R. W . Erickson, S. ´ Cuk, R. D. Middlebrook, “Lar ge Signal Modeling and Analysis of Switching Re gulators, IEEE PESC Record , pp. 240-250, 1982. [3] V . T ran, M. MahD, “Modeling and Analysis of T ransformerless High Gain Buck-boost DC-DC Con v erters, Po wer Electronics and Dri v e Systems (IJPEDS), V ol. 4, No. 4 , pp. 528-535, Dec 2014. [4] J.S. Renius A, V . K umar K, A. Fredderics, R. Guru, S.L. Na ir , “Modelling of V ariable Frequenc y Synchronous Buck Con v erter , nternational Journal of Po wer Electronics and Dri v e System (IJPEDS) V ol. 5, No. 2 , pp. 237-243, October 2014. [5] P . Szcze ´ sniak, A Comparison Between T w o A v erage Modelling T echniques of A C-A C Po wer Con v erters, International Journal of Po wer Electronics and Dri v e System (IJPEDS), V ol. 6, No. 1 , pp. 32-44, March 2015. [6] R. W . Erickson, D. Maksimo vi ´ c, Fundamental of Po wer Electronics , Kluwer Academic Publishers, Netherland, 2001. [7] S. R. Naidu, D. A. Fernandes, “T echnique for Simulating the Steady-State Response of Po wer Electronic Con- v erter , IET Po wer Electronics , V ol. 4, pp. 269-277, 2011. [8] L. Iannelli, F . V asca, G. Angelone, “Computation of Steady-State Oscillations in Po wer Con v erter Through Complementary , IEEE T ransactions on Circuits and SystemsI: Re gular P apers, V ol. 58, No. 6 , pp. 1421-1432. June 2011. [9] G. T . K ostakis, S. N. Manias, N. I. Mar g aris, A Generalized Method for Calculati ng the RMS V alues of Switching Po wer Con v erters, IEEE T ransactions on Po wer Electronics , V ol. 15 No 4, pp. 616-625, July 2000. [10] M. Daryaei, M. Ebrahimi, S. A. Khajehoddin, Accurate P arametric Steady State Analysis and Des ign T ool for DC-DC Po wer Con v erters, IEEE Applied Po wer Electronics Conference and Exposition (APEC) , pp. 2579- 2586, 2016. [11] H. M. Mahery , E. Babaei , “Mathematical Modeling of Buck-boost DC-DC Con v erter and In v estig ation of Con v erter Element on T rans ient and Steady State Responses, Electrical Po wer and Ener gy System 44 , pp. 949-963, 2013. [12] B. Tsai, O. T rescases, B. Francis, An In v estig ation of Steady-State A v eraging for the Single-Inductor Dual- Output Buck Con v erter using F ourier Analysis, 2010 IEEE 12th W orkshop on Control and Modeling for Po wer Electronics (COMPEL) , June 2010. [13] R. T rinchero, I. S. Stie v ano, F . G. Cana v ero, “Steady-State Analysis of Switching Po wer Con v erter via Aug- mented T ime-In v ariant Equi v alents, IEEE T ransaction on Po wer Electronics , V ol. 29 No 11, pp. 5657-5661, No v 2014. [14] F . Mis ¸ oc, M. M. Morcos, J. Lookadoo, “F ourier -Series Models of DC-DC Con v erters, IEEE 38th North American Po wer Symposium , pp. 193-199, 2006. [15] R. T rinchero, P . Manfredi, I.S.Stie v ano, F .G. Cana v ero, “Steady-State Analysis of Switching Con v erter via IJECE V ol. 7, No. 5, October 2017: 2374 2381 Evaluation Warning : The document was created with Spire.PDF for Python.
IJECE ISSN: 2088-8708 2381 (a) Complete response (b) Steady state response Figure 6. Comparison between SPICE and proposed analysis of parameter -2 Frequenc y-Domain Circuit Equi v alents, IEEE T ransactions on Circuits and SystemsII: Express Briefs, V ol. 63, No. 8 , pp. 748-752, August 2016. [16] G. Strang, Computational Science and Engineering , W ellesle y-Cambridge Press, USA, 2007. Accur ate Symbolic Steady State Modeling of Buc k Con verter (Ek o Setiawan) Evaluation Warning : The document was created with Spire.PDF for Python.