RD.MIDDLEBROOK AND SLOBODAN CUK ABSTRACT A method f o r model l i ng swi t chi ng-convert er power s t ages i s devel oped, whose s t a r t i n g poi nt i s t he uni f i ed st at e- space r epr es ent at i on of t h e switched networks and whose end r e s u l t i s e i t h e r a complete st at e-space des cr i pt i on or i t s equi val ent smal l -si gnal low-frequency l i ne a r c i r c u i t model. A new canoni cal c i r c u i t model i s proposed, whose fi xed t opol ogy cont ai ns a l l t he e s s e n t i a l i nput -out put and c ont r ol pr oper t i es of any dc-to- dc swi t chi ng conver t er , r egar dl es s of i t s de t a i l e d conf i gur at i on, and by which d i f f e r e n t conver t er s can be char act er i zed i n t h e form of a t a b l e con- veni ent l y s t or ed i n a computer d a t a bank t o pro- vi de a us e f ul t o o l f o r computer ai ded desi gn and opt i mi zat i on. The new canoni cal c i r c u i t model pr edi ct s t h a t , i n genera1, swi t chi ng a c t i on i nt r o- duces bot h zer os and pol es i n t o t he dut y r a t i o t o out put t r a ns f e r f unct i on i n addi t i on t o t hose from t he e f f e c t i ve f i l t e r network. 1. INTRODUCTION 1.1 Bri ef Review of Exi st i ng Modelling Techniques I n model l i ng of swi t chi ng conver t er s i n gener al , and power s t ages i n p a r t i c u l a r , two main approaches - one based on st at e- space model l i ng and t he ot her usi ng an averagi ng t echni que - have been developed ext ensi vel y, but t her e has been l i t t l e c or r e l a t i on between them. The f i r s t approach remains s t r i c t l y i n t he domain of equat i on mani pul at i ons, and hence r e l i e s heavi l y on numeri cal methods and computerized Implementationa. Its pri mary advantage i s i n t h e uni f i ed des cr i pt i on of a l l Pover s t ages r egar dl es s of t h e t ype (buck. boost . buck-boost o r any ot he r va r i a t i on) t hrough u t i l i z a t i o n of t h e exact st at e-space equat i ons of t he two swi t ched models. -On t h e ot her hand, Procesei ng Systems .'I based on equi val ent c i r c u i t mani pul at i ons, r e s ul t i ng i n a s i ngl e equi val ent l i n e a r c i r c u i t model of t he par er st age. Thi s has t he d i s t i n c t advantage of provi di ng t he c i r c u i t desi gner wi t h physi cal i ns i ght i n t o t he behavi our of t he o r i g i n a l swi t ched c i r c u i t , and of al l owi ng t he powerful t ool s of l i n e a r c i r c u i t a na l ys i s and s ynt hes i s t o be used t o t he f u l l e s t ext ent i n desi gn of r egul at or s i ncor por at i ng swi t chi ng conver t er s. 1.2 Proposed New St at e-Space Averaging Approach The method proposed i n t h i s paper bri -dges t he gap e a r l i e r consi dered t o e x i s t between t he s t a t e - space t echni que and t h e averagi ng t echni que of model l i ng power s t a ge s by i nt r oduct i on of s t a t e - space averaged model l i ng. At t h e same t i me i t o f f e r s t he advant ages of bot h e x i s t i n g methods - t h e gener al uni f i ed t reat ment of t he st at e- space approach, a s wel l a s an equi val ent l i n e a r c i r c u i t model a s i t s f i n a l r e s u l t . Furt hermore, it makes c e r t a i n gener al i zat i ons pos s i bl e, which ot herwi se coul d not be achi eved. The proposed st at e- space averagi ng method, out l i ned i n t h e Flowchart of Fi g. 1, al l ows a uni f i ed t r eat ment of a l a r g e v a r i e t y of power s t ages c ur r e nt l y used, s i nc e t h e averagi ng s t e p i n t he st at e- space domain i s very si mpl e and c l e a r l y defi ned (compare bl ocks l a and 2a) . It merely c ons i s t s of averagi ng t he two exact st at e- space de s c r i pt i ons of t he swi t ched models over a s i ngl e cycl e T, where f s = 1/ T i s t he swi t chi ng frequency (bl ock 2a) . Hence t he r e i s no need f o r s pe c i a l "know-howf' i n massaging t he two swi t ched c i r c u i t models i n t o t opol ogi cal l y equi val ent forms i n or der t o appl y ci r cui t - or i ent ed procedure d i r e c t l y , a s r equi r ed i n [ l ] (bl ock l c ) . Nevert hel ess, t hrough a hybri d model l i ng t echni que (bl ock 2c) , t he c i r - c u i t s t r u c t u r e of t h e averaged c i r c u i t model (bl ock 2b) can be r e a d i l y recogni zed from t he averaged st at e-space model (bl ock 2a) . Hence a l l t he be ne f i t s of t h e previ ous averagi ng t echni que a r e r et ai ned. Even though t h i s out - I n e i t h e r case, a per t ur bat i on and l i ne a r i z a t i on @ 1976 IEEE. Reprinted, vith permission, f r w Proceedings of the IEEE Pwer El ect r~ni cs Specialists Conference, J~~~ * - 10, 1976, Cleveland, OH. stolc - spore C ~ Y O ~ I O ~ S rlrody slotc ldct modd d - o: d, d' ~' - 2, x- x+; , ax+dv, - o - x - - ~ - h v , ; r-C? y . ~ + i , v, -Lj . i dynom,c ioc ~ m o / / s 1 9 ~ / 1 mo t / . Fi g. 1, Flowchart of averaged modelling approaches process r equi r ed t o i ncl ude t he dut y r a t i o modulation e f f e c t proceeds i n a very s t r a i ght f or - ward and formal manner, t hus emphasizing t he corner-stone char act er of bl ocks 2a and 2b. At t h i s s t age (block 2a or 2b) t he st eady- st at e (dc) and l i n e t o out put t r a ns f e r f unct i ons a r e al r eady avai l abl e, a s i ndi cat ed by bl ocks 6a and 6b r es pect i vel y, whi l e t he dut y r a t i o t o out put t r a ns f e r f unct i on i s a va i l a bl e a t t he f i nal - st age model (4a or 4b) a s i ndi cat ed by bl ocks 7a and 7b. The two f i n a l s t age models (4a and 4b) then gi ve t he complete des cr i pt i on of t he swi t chi ng convert er by i ncl us i on of bot h independent con- t r o l s , t he l i n e vol t age va r i a t i on and t he dut y r a t i o modulation. Even though t he c i r c u i t t ransformat i on pat h b might be pr ef er r ed from t he pr a c t i c a l desi gn st andpoi nt , t he st at e- space averagi ng pat h a i s i nval uabl e i n reachi ng some gener al concl usi ons about t he smal l -si gnal low-frequency models of any dc-to-dc swi t chi ng conver t er (even t hose ye t t o be i nvent ed) . Whereas, f o r pat h b, one has t o be prekented wi t h t he pa r t i c ul a r c i r c u i t i n or der t o proceed wi t h modelling, f o r pat h a t he f i na l st at e- space averaged equat i ons (block 48) gi ve t he complete model des cr i pt i on through gener al mat r i ces A1, A2 and vect or s bl, b2' c T, and c2T of t h e two s t a r t i n g switched models ( h o c k l a ) . Thi s i s a l s o why al ong pat h b i n t he Flowchart a pa r t i c ul a r example of a boost power s t a ge wi t h p a r a s i t i c e f f e c t s was chosen, whi l e al ong pat h a gener al equat i ons have been r et ai ned. Spe c i f i c a l l y, f o r t he boost power s t age bl = b2 - b. Thi s example w i l l be l a t e r pursued i n d e t a i l al ong both pat hs. I n addi t i on t he st at e- space averagi ng approach o f f e r s a c l e a r i ns i ght i n t o t he qua nt i t a t i ve na t ur e of t he ba s i c averagi ng approximation, which becomes be t t e r t he f ur t he r t he e f f e c t i ve low-pass f i l t e r cor ner frequency f is below t he swi t chi ng frequency f, , t ha t i s, f z / f s << 1. Thi s i s, however, shown t o be equi val ent t o t he requirement f o r smal l out put vol t age r i ppl e , and hence does not pose any s e r i ous r e s t r i c t i o n or l i mi t a t i on on modelling of prac r i c a l dc-to-dc conver t er s. Fi na l l y, t he st at e- space averagi ng approach s er ves a s a ba s i s f or der i vat i on of a us e f ul gener al c i r c u i t model t ha t des cr i bes t he i nput - out put and c ont r ol pr oper t i es of any dc-to-dc conver t er . 1. 3 New Canoni cal Ci r cui t Model The cul mi nat i on of any of t hes e der i va- tions al ong e i t h e r pat h a or pat h b i n t he Flowchart of Fi g. 1 i s an equi val ent c i r c u i t (bl ock 5) , va l i d f or smal l - si gnal low-f requency var i at i ons superimposed upon a dc oper at i ng poi nt , t h a t r epr es ent s t h e two t r a ns f e r f unct i ons of i n t e r e s t f o r a swi t chi ng conver t er . These .re t he l i n e vol t age t o out put and dut y r a t i o to out put t r a n s f e r f unct i ons . The equi val ent c i r c u i t i s a canoni cal model t h a t cont ai ns t h e e s s e n t i a l pr ope r t i e s of dc-to-dc swi t chi ng conver t er , r egar dl es s of t he det ai l ed conf i gur at i on. As seen i n bl ock 5 f o r the general case, t h e model i ncl udes an i d e a l t ransformer t h a t des cr i bes t he ba s i c d c - t e d c t ransformat i on r a t i o from l i n e t o out put ; a low-pass f i l t e r whose el ement val ues depend upon t he dc dut y r a t i o ; and a vol t age and a cur r ent generat or pr opor t i onal t o t h e dut y r a t i o modula- t i on i nput . The canoni cal model i n bl ock 5 of t he Flow- char t can be obt ai ned fol l owi ng e i t h e r pat h a or pat h b, namely from bl ock 4a or 4b, a s w i l l be &own l a t e r . However, fol l owi ng t h e gener al des cr i pt i on of t he f i n a l averaged model i n bl ock ha, c e r t a i n gener al i zat i ons about t he canoni cal model a r e made pos s i bl e, which a r e ot her wi se not achi evabl e. Namely, even though f or a l l c ur r e nt l y known swi t chi ng dc-to-dc conver t er s (such a s t he buck, boost , buck-boost, Venable 131, Weinberg [ 4 ] and a number of ot her s ) t h e frequency dependence appears onl y i n t h e dut y- r at i o dependent vol t age generat or but not i n t h e cur r ent gener at or , and the ~n bs,a@,a,fkaLs.f d e ~ . L~i , r l ~l e ~z c r _o~. . e oI ~omi a l i n pol es of t h e e f f e c t i v e f i l t e r network which e s s e nt i a l l y c o n s t i t u t e t h e l i n e vol t age t o out put t r a ns f e r f unct i on. Moreover, i n gener al , both dut y- r at i o dependent gener at or s , vol t age and cur- r e nt , a r e frequency dependent ( a ddi t i ona l zer os and pol es ) . That i n t h e p a r t i c u l a r cas es of t he boost or buck-boost conver t er s t h i s dependence reduces t o a f i r s t or der pol ynomi al r e s u l t s from t he f a c t t h a t - t h e or der of t h e syst em which i s involved i n t h e swi t chi ng a c t i on i s onl y two. Hence from t he gener al r e s u l t , t h e or der of t he Polynomial i s a t most one, though it coul d reduce t o a pure cons t ant , a s i n t h e buck or t h e Venable convert er [ 3] . The s i gni f i c a nc e of t h e new c i r c u i t model i s t h a t any swi t chi ng dc-to-dc conver t er can be reduced t o t h i s canoni cal f i xed t opol ogy form, a t l e a s t a s f a r a s i t s i nput -out put and c ont r ol a;nt"y";;;f; 5-7 i;-g;;,;j, ;~;;er~f~~c:'iv:a for el ement s depend on dut y r a t i o D) , and t h e conf i - gur at i on chosen which opt i mi zes t he s i z e and wei ght . Al so, comparison of t he frequency depen- dence of t he two dut y- r at i o dependent gener at or s provi des i ns i ght i n t o t he quest i on of s t a b i l i t y once a r egul at or feedback l oop i s cl os ed. 1.4 Ext ensi on t o Complete Regul at or Treat ment Fi na l l y, a l l t h e r e s u l t s obt ai ned i n model l i ng t he conver t er o r , more accur at el y, t he network which e f f e c t i ve l y t akes pa r t i n swi t chi ng a c t i on, can e a s i l y be i ncor por at ed i n t o more compl i cat ed syst ems cont ai ni ng dc-to-dc conver t er s. For example, by model l i ng t he modul at or s t a ge al ong t he same l i n e s , one can obt ai n a l i n e a r c i r c u i t model of a cl osed-l oop swi t chi ng r egul at or . St andard l i near feedback theory ran t hen be used f o r bot h a na l ys i s and s ynt hes i s , s t a b i l i t y cons i der at i ons , and proper desi gn of feedback compensating net - works f o r mul t i pl e l oop a s wel l a s si ngl e-l oop r e gul a t or conf i gur at i ons . 2. STATE-SPACE AVERAGING I n t h i s s e c t i on t he st at e- space averagi ng method i s developed f i r s t i n gener al f o r any dc- to-dc swi t chi ng conver t er , and t hen demonst rat ed i n d e t a i l f o r t he p a r t i c u l a r cas e of t he boost power s t age i n which p a r a s i t i c e f f e c t s ( e s r of t he capaci t or and s e r i e s r e s i s t a nc e of t he i n- duct or ) a r e i ncl uded. General equat i ons f or bot h st eady- st at e (dc) and dynamic performance ( ac) a r e obt ai ned, from which i mport ant t r a n s f e r in f unct i ons a r e der i ved and a l s o appl i ed t o t he s p e c i a l cas e of t he boost power s t age. swi t chi ng between two l i n e a r net works cons i s t i ng of i d e a l l y l o s s l e s s s t or a ge el ement s, i nduct ances and capaci t ances. I n p r a c t i c e , t h i s f unct i on may be obt ai ned by use of t r a n s i s t o r s and di odes which oper at e a s synchronous swi t ches. On t he assumpt i on t h a t t h e c i r c u i t oper at es i n t he so- c a l l e d "continuous conduct i on" mode i n which t he i nst ant aneous i nduct or c ur r e nt does not f a l l t o zer o a t any poi nt i n t he cycl e, t h e r e a r e onl y two d i f f e r e n t " s t at es " of t h e c i r c u i t . Each s t a t e , however, can be r epr esent ed by a l i n e a r c i r c u i t model ( a s shown i n bl ock l b of Fi g. 1 ) or by a correspondi ng s e t of s t at e- s pace equat i ons (bl ock l a ) . Even though any s e t of l i n e a r l y i ndependent va r i a bl e s can be chosen a s t he s t a t e va r i a bl e s , i t i s customary and conveni ent i n e l e c t r i c a l net works t o adopt t h e i nduct or c ur r e nt s and capa- c i t o r vol t ages. The t o t a l number of s t or age el ement s t hus det ermi nes t h e or der of t he system. k e , E , , ~ s , " d , ~ :~,~h,,B,,:h,~,i,c,~ e:f LaLve_c,to,E equat i ons f o r t h e two swi t ched models: (i) i nt er val Td : ( 11) i nt er val Td' : where Td denotes t he i nt er val when t he switch i s i n t he on s t at e a n d T( l - d) 5 ~ d ' i s t he i nt erval f or which it i s i n t he o f f s t at e , as shown i n Fig. 2 . The s t a t i c equations yl = clTx and y = c2Tx are necessary i n order t o account f or t i e case when t h e output quant i t y does not s w ~ t c h dr i v e o f f 7 k r- 1 Td Td' 4 Fig. 2. Def i ni t i on o f t he two switched i nt er val s Td and Td' . coincide wi t h any of t he s t at e var i abl es , but i s rather a cert ai n l i near combination o f t he s t at e vari abl e$. Our obj ect i ve now i s t o replace t he st at e- space descri pt i on o f t he two l i near c i r c ui t s emanating from t he two successive phases o f t he switching cycl e T by a si ngl e state-space des- cri pt i on which represents approximately t he beha- viour o f t he c i r c ui t across t he whole period T. We t heref ore propose t he following simple avera- ging s t ep: t ake t he average o f bot h dynamic and s t at i c equations f or t he two switched i nt er val s ( I ) , by summing t he equations f or i nt er val Td mul t i pl i ed by d and t he equations f or i nt er val Td' mul t i pl i ed by d' . The following l i near continuous system r e s ul t s : Af t e r rearranging ( 2 ) i nt o t he standard l i near continuous system state-space descri pt i on, we obt ai n t he basi c averaged state-space descrip- t i on (over a si ngl e period T ) : Thi s model i s t he basi c averaged model which i s the s t ar t i ng model f or a l l ot her deri vat i ons (bot h state-space and ci r cui t or i ent ed) . Note t hat i n t he above equations t he dut y r at i o d i s considered const ant ; it i s not a time dependent vari abl e ( y e t ) , and part i cul arl y not a switched discontinuous vari abl e which changes between 0 and 1 as i n [ l ] and [ 2 ] , but i s merely a f i xed number f or each cycl e. This i s evi dent from t he model deri vat i on i n Appendix A. In part i cul ar, when d = 1 ( swi t ch const ant l y on) t he averaged model ( 3 ) reduces t o switched model ( 11) , and when d = 0 ( swi t ch o f f ) i t reduces t o switched model ( 111) . In essence, comparison between ( 3) and ( 1 ) shows t hat t he system matrix of t he averaged model i s obtained by taking t he average o f two switched model matrices A and A2, i t s cont rol i s t he average o f two cont rol vect ors bl and b 2 , and i t s output i s t he average o f two outputs yl and y over a period T. 2 The j us t i f i c at i on and t he nature o f the approximation i n s ubs t i t ut i on f or t he two switched models o f (1) by averaged model ( 3) i s indicated i n Appendix A and given i n more de t ai l i n [b]. The basi c approximation made, however, i s t hat o f approximation o f t he fundamental matrix eAt = I + At + by i t s f i rst - order l i near term. Thi s i s , i n turn,shown i n Appendix B t o be t he same approximation necessary t o obt ai n t he dc condi t i on independent o f t he storage element val ues (L,C) and dependent on t he dc dut y r at i o onl y. I t al s o coincides wi t h t he requirement f or low output vol t age r i ppl e , which i s shown i n Appendix C t o be equi val ent t o f c / f << 1 , namely t he e f f e c t i v e f i l t e r corner frequency much lower than t he switching frequency. The model represented by ( 3 ) is an averaged model over a si ngl e period T . I f we now assume t hat t he dut y r at i o d i s constant from cycl e t o c y c l e , namely, d = D (st eady s t at e dc duty r a t i o ) , we get : where Since ( 4 ) i s a l i near system, superposition holds and it can be per t ~r bed by i nt roduct i on o f l i ne vol t age vari at i ons v as v = V + C , where V i s t he dc l i ne i nput v81t agef cauging 8 c8rrespo:ding perturbation i n t he s t at e vect or x = X + X , where !gain X i s t he dc value of t he s t at e vect or and x t he superFposed ac pertur- bat i on. Si mi l ar l y, y = Y + y , and Separation o f the steady-state ( dc) part from the dynamic ( ac) part then r es ul t s i n t he steady s t at e ( dc) model and t he dynamic ( ac) model I t i s i nt erest i ng t o note t hat i n ( 7 ) t he steady s t at e ( dc) vect or X wi l l i n general only depend on the dc duty r at i o D and resi st ances i n t he ori gi nal model, but not on t he storage element values ( L' s and C' s ) . Thi s i s so because X i s t he sol ut i on o f t he l i near system of equations i n which L' s and C' s are proportionality con- st ant s. Thi s i s i n complete agreement wi t h t he f i rst -order approximation of t he exact dc conditions shown i n Appendix B , which coincides wi t h expression ( 7 ) . From t he dynamic ( ac) model, t he l i ne voltage t o state-vector t r ans f er functions can be easi l y derived as: Hence at t h i s stage both steady-state ( dc) and l i ne t r ans f er functions are avai l abl e, as shown by block 6a i n t he Flowchart of F i g . 1. We now undertake t o include t he duty r at i o modulation e f f e c t i nt o t he basic averaged model ( 3 ) . 2.2 Perturbation Suppose now t hat the duty r at i o changes from cycle t o cycl e, t hat i s , d ( t ) = D + a where D i_s the steady-state ( dc) duty r at i o as bef ore and d i s a superimposed ( ac) vari at i on. With t he A corres~ondi na perturbatipn def i ni t i on x = X + x , dc term l i ne duty r at i o vari at i on vari at i on nonlinear second-order term dc ac ac term nonlinear term term term The perturbed state-space descri pt i on i s nonlinear owing t o t he presence o f the_ prodyct o f the two time dependent quant i t i es x and d . 2 . 3 Linearization and Final State-Space Averaged Model Let us now make t he small-signal approxima- t i on, namely t hat departures from t he steady st at e values are negl i gi bl e compared t o t he steady st at e values themselves: Then, using approximations (12) we negl ect a l l nonlinear terms such as t he second-order terms i n (11) and obtain once again a line_ar system, but including duty-ratio modulation d . Af t er sepa- rat i ng steady-state ( dc) and dynamic ( ac) parts o f t hi s linearized system we arri ve at t he follow- ing resul t s for t he f i nal state-space averaged model. Steady-state ( dc) model: Dynamic (ac small-signal) model: T In these r es ul t s , A, b and c are given as bef ore by ( 5 ) . Equations (13) and ( 14) represent t he small- signal low-frequency model o f any No- st at e switching dc-to-dc converter working i n t he con- tinuous conduction mode. I t i s important t o note t hat by negl ect of t he nonlinear term i n (11) t he source o f harmonics i s e f f e c t i v e l y removed. Theref ore, t he l i near description (14) i s act ual l y a l i neari zed describing function r es ul t t hat i s the l i m i t o f t he de s ~r i bi ng~f unc t i on 2s t he amplitude of t he input signals ve andlor d becomes vanishingly explained i n [ I ] , 121, or la]' i n h i c h smail- signal assumption (12) i s preserved. Very good agreement up t o close t o hal f t he switching frequency has been demonstrated repeatedly (111, [21, 131, 171). 2.4 Example: Boost Power St age wi t h Pa r a s i t i c s We now i l l u s t r a t e t he method f o r t h e boost power s t a ge shown i n Fi g. 3. i n which I i s t he d c i nduct or c ur r e nt , V is t h e dc capaci t or vol t age, and Y i s t he dc out put vol t age . Dynamic ( a c s mal l s i gna l ) model: Fi g. 3. Example f o r t he s t at e- s pace averaged model l i ng: boost power s t a ge wi t h para- s i t i c ~ i ncl uded. Fi g. 4. Two swi t ched c i r c u i t models of t he c i r c u i t i n Pi g. 3 wi t h assumpt i on of i d e a l swi t ches. A l l el ement s i n t h e f i n a l s t a t e - space averaged mode4 (13) and (14) a r e obt ai ned: A b , c from a ) f o r i n t e r v a l Td, and ~ ~ , b i : ~ k homb) f o r i n t e r v a l Td' . With assumption of i d e a l swi t ches, t h e two swi t ched models a r e a s shown i n Fig. 4. For choi ce of st at e- space vect or xT = ( 1 v ) , t h e s t a t e space equat i ons become: .--I where Note t h a t (15) i s t he s p e c i a l case of (1) i n which bl = b2 = b = [ I / L b]T. Using (16) and (5) i n t h e gener al r e s u l t (13) and ( 14) , we obt ai n t he fol l owi ng f i n a l st at e- space averaged model. St eady- st at e (dc) model: 2 i n which R' (1-D) R + R~ + D ( ~ - D ) ( R ~ ~ ~ R ; . We now l ook more c l os e l y a t t he dc vol t age t r ansf or mat i on r a t i o i n (17): v Y 1 2 _ r - = - (1-D) R (19) V V 1-D (I-D)~R + Rp + D(1-D) (Rc /b) W L T i d e a l c or r e c t i on f a c t o r dc gai n Thi s shows t h a t t h e i d e a l d c vol t a ge gai n i s 1 / D ' when a l l p a r a s i t i c s a r e zer o (R = 0, Rc = 0) and t h a t i n t h e i r presence i t is s l f g h t l y reduced by a c or r e c t i on f a c t o r l e s s t han 1. Also we observe ..L-.. ------- --- ..I.- a.. ---- /u di s c ont i nui t y of out put vol t a ge was not in?iu$.&$+L i n [Z], but was c or r e c t l y account ed f o r i n 111. From t h e dynamic model (18) one can f i nd t he dut y r a t i o t o out put and l i n e vol t age t o out put t r a n s f e r f unct i ons , which agr ee exact l y wi t h t hose obt ai ned i n [ l ] by fol l owi ng a d i f f e r e n t method of averaged model de r i va t i on based on t he equi val ence of c i r c u i t t opol ogi es of two swi t ched net works. The fundament al r e s u l t of t h i s s e c t i on i s t he development of t h e gener al st at e- space averaged model r epr esent ed by (13) and ( 14) , which can be e a s i l y used t o f i nd t he smal l - si gnal low-frequency model of any swi t chi ng dc-to-dc conver t er . Thi s was demonst rat ed f o r a boost power s t a g e wi t h p a r a s i t i c s r e s u l t i n g i n t h e averaged model (17) and (18). It i s i mport ant t o emphasize t h a t , unl i ke t he t r a n s f e r f unct i on de s c r i pt i on, t he st at e- space des cr i pt i on (13) and a 4 ) gi ves t he compl et e syst em behavi our. Thi s i s very us ef ul i n implementing two-loop and mul t i -l oop feedback when two o r more s t a t e s a r e used-i n a feedback pat h t o modul at e t h e dut y r a t i o d. For example. bot h out put vol t a ge and i nduct or c ur r e nt may be r et ur ned i n a feedback l oop. 3. HYBRID MODELLING In t h i s sect i on it wi l l be shown t hat f or any s pe c i f i c converter a us ef ul c i r c ui t r eal i z at i on o f t he basi c averaged model given by ( 3 ) can always be found. Then, i n t he following s ect i on, t he perturbation and l i neari zat i on st eps wi l l be carried out on t he c i r c ui t model f i nal l y t o arri ve a t t he c i r c ui t model equivalent o f ( 13) and ( 14) + The c i r c ui t r eal i z at i on wi l l be demonstrated f or t he same boost power stage example,for which t he basi c state-space averaged model ( 3 ) becomes: In order t o "connect" t he c i r c ui t , we express t he capacitor vol t age v i n terms of t he desired output quant i t y y as : R+Rc v = - y - ( 1 - d ) Rc i R or , i n mat ri x form Subst i t ut i on o f (21) i nt o ( 20) gives i deal From ( 22) one can e as i l y reconstruct t he c i r c ui t represent at i on shown i n F i g . 5 . The basi c model ( 22) i s val i d f or t he dc regime, and t he two dependent generators can be modeled as an i deal d ' : l transformer whose range extends down t o dc, as shown i n F i g . 6 . F i g . 5 . Ci rcui t r eal i z at i on o f t he basi c st at e- space averaged model ( 20) through hybrid modelling. d': I Fig. 6 . Basic c i r c ui t averaged model f or t he boost c i r c ui t example i n F i g . 3. Both dc- to-dc conversion and l i n e vari at i on are modelled when d(t)-D. As be f or e , we f i nd t hat t he c i r c ui t model i n Fig. 6 reduces f or d = 1 t o switched model i n F i g . 4a, and f or d = 0 t o switched model i n F i g . 4b. In both cases t he addi t i onal resi st ance R = dd' ( R I I R) disappears, as it should. 1 I f t he duty r at i o i s constant so d = D , t he dc regime can be found e as i l y by considering inductance L t o be short and capacitance C t o be open f or dc, and t he transformer t o have a D' :l r at i o. Hence t he dc vol t age gain ( 19) can be di r e c t l y seen from F i g . 6 . Si mi l arl y, a l l l i ne t r ans f e r f unct i ons corresponding t o ( 10) can be e as i l y found from Fig. 6 . I t i s i nt er es t i ng now t o compare t h i s i deal d ' : l transformer wi t h the usual ac transformer. While i n t he l a t t e r t he t urns r at i o i s f i x e d, t he one employed i n our model has a dynamic t urns r at i o d ' : l which changes when t he dut y r at i o i s a func- t i on o f t i me, d ( t ) . I t i s through t hi s i deal transformer t hat t he actual cont rol l i ng f unct i on i s achieved when t he feedback loop i s cl osed. In addi t i on t he i deal transformer has a dc t rans- formation r at i o d ' : l , while a real transformer works f or ac si gnal s onl y. Nevert hel ess, t he concept o f t he i deal transformer i n Fig. 6 wi t h such properties i s a very us e f ul one, si nce a f t e r a l l t he switching converter has t he overal l property o f a dc-to-dc transformer whose t urns r at i o can be dynamically adjusted b y dut y r at i o modulation t o achieve t he cont rol l i ng f unct i on. We wi l l , however, see i n t he next sect i on how t hi s can be more e x pl i c i t l y modelled i n terms of dut y- rat i o dependent generators onl y. Following t he procedure out l i ned i n t h i s sect i on one can e as i l y obt ai n t he basi c averaged c i r c ui t models o f t hree common converter power stages, as shown i n t he summary o f Fig. 7 . (a) buck power St e qe : (b) boos t Power s t a a r : buck (c) buck b o o s t power s boost R Fig. 7. Summary of basi c ci r cui t averaged models f or t hree common power stages: buck, boost , and buck-boost. The two switched ci r cui t state-space models f or t he power stages i n Fig. 7 are such t hat t he general equations ( 1) reduce t o t he speci al cases A1 - A2 - A, bl # b2 = 0 (zero vect or) f or t he buck parer st age, and A1 # A2, b - b2 = b f or t he boost power st age, whereas f or t he buck-boost power stage A + A2 and bl # b2 = 0 so t hat t he general case 3s retained. 4. CIRCUIT AVERAGING As indicated i n t he Introduction, in t h i s section t he al t ernat i ve path b i n t he Flowchart o f F i g . 1 wi l l be followed, and equivalence wi t h t he previously developed path a f i rml y established. The f i nal ci r cui t averaged model f or t he same example o f t he boost power stage wi l l be arrived a t , which i s equivalent t o i t s corresponding state-space deacription given by (17) and (18) . The averaged ci r cui t models shown i n ' Fi g . 7 could have been obtained as i n [ 2 ] by di r ect l y averaging t he corresponding components of t he two switched models. However, even f or some simple cases such as t he buck-boost or tapped inductor boost [l ] t hi s presents some d i f f i c u l t y owing t o t he requirement o f having two switched ci r cui t models t opol ogi cal l y equi val ent , while t here i s no such requirement i n t he outlined procedure. 4.1 Perturbation I f t he averaged model i n F i g . 7b i s perturbed according t o v g - v +G iA- I+$, d - m-a, d' ,- D' - a, v - ~ 4 , ~ ~ H'Y+~ t he nonlinear model i n F i g . 8 r es ul t s . F i g . 8. Perturbation o f t he basi c averaged ci r cui t model i n F i g . 6 includes t he duty r at i o modulation e f f e c t 2, but r es ul t s i n t h i s nonlinear ci r cui t model. 4.2 Linearization Under t he small-signal approximation ( 12) , t he following l i near approximations are obtained: and t he f i nal averaged ci r cui t model of F i g . 9 r es ul t s . In t h i s ci r cui t model we have f i nal l y obtained t he cont rol l i ng f unct i on separated i n terms o f duty r at i o a dependent generators e, and j , , while t he transformer t urns r at i o i s dependent on t he dc duty r at i o D onl y. The ci r cui t model obtained i n F i g . 9 i s equivalent t o the state-space descri pt i on given by (17) and ( 18) . F i g . 9 . Under small-signal assumption (121, the model i n F i g . 8 i s l i neari zed and t h i s f i nal averaged ci r cui t model of t he boost stage i n F i g . 3 i s obtained. 5 . THE CANONICAL CIRCUIT MODEL Even though t he general f i nal state-space In t hi s sect i on we proceed with t he perturba- averaged model i n (13) and (14) gives t he complete t i on and l i neari zat i on steps applied t o t he ci r - descri pt i on o f t he system behaviour, one might s t i l l cui t model, continuing wi t h t he boost power stage wish t o derive a ci r cui t model describing i t s as an example i n order t o include expl i ci t l y t he input-output and cont rol properties as i l l ust rat ed duty r at i o modulation e f f e c t . i n F i g . 10. (a) state - space i n p u t output bas is '0' cont r ol -7G cont rol F i g . 10. Def i ni t i on o f t he modelling obj ect i ve: ci r cui t averaged model describing input- output and control properties. In going from t he model o f F i g . 10a t o t hat o f F i g . l ob some information about t he i nt ernal behaviour o f some o f t he s t at es wi l l cert ai nl y be l o s b b u t , on t he other hand, important advantages wi l l be gained as were br i e f l y outlined i n t he Int roduct i on, and as t hi s sect i on wi l l i l l us t r at e. We propose t he following f i xed topology ci r cui t model, shown i n F i g . 11, as a real i zat i on control f u ~ t i o n basic dc-to-dc effective low-pass vi a d transforrrration f I l ter network F i g . 11. Canonical ci r cui t model real i zat i on o f t he "black box" i n F i g . l ob, modelling t he three essent i al functions o f any dc-to-dc converter: cont rol , basic dc conversion, and l w- pass f i l t e r i ng. o f t he "black box" i n F i g . l ob. We cal l t h i s model t he canonical ci r cui t model, because any switching converter input-output model, regardless o f i t s det ai l ed conf i gurat i on, could be represented i n t h i s form. Di f f er ent converters are represented simply by an appropriate s et o f formulas f or t he four elements e ( s ) , j ( s ) , p . He(s) i n t he general equivalent ci r cui t . The pol ari t y o f t he i deal p:1 transformer i s determined by.whether or not t he power stage i s pol ari t y i nvert i ng. I t s t ur ns r at i o p i s dependent on t he dc duty r at i o D , and since f or modelling purposes t he transformer i s assumed t o operate down t o dc, it provides t he basi c dc-to-dc l evel conversion. The single-sec- t i on low-pass LeC f i l t e r i s shown i n F i g . 11 only f or i l l us t r at i on purposes, because t he actual number and conf i gurat i on of t he L' s and C' s i n t he e f f e c t i v e f i l t e r t r ans f er function real i zat i on depends on t he number o f storage elements i n t he ori gi nal converter. The resi st ance Re i s included i n t he model o f F i g . 11 t o represent t he damping properties o f t he e f f e c t i v e low-pass f i l t e r . I t i s an " ef f ect i ve" resi st ance t hat accounts f or various seri es ohmic resi st ances i n t he actual ci r cui t (such as RE i n t he boost ci r cui t example), t he additional "switching" resi st ances due t o di s - cont i nui t y o f t he output vol t age (such as Rl = DD'(R 1 1 R) i n t he boost ci r cui t example), and al so a "modulation" resi st ance t hat ari ses from a modulation of t he svitching t ransi st or storage time 111. 5.1 Derivation of t he Canonical Model through State-Space From t he general state-space averaged model (13) and ( 14) , we obtain di r ect l y using t he Laplace transform: Now, from t he complete set o f t ransf er functions we si ngl e out those which describe t he converter input-output properties, namely ,. A i n which t he G's are known expl i ci t l y i n terms o f t he matrix and vector elements i n ( 23) . Equations (24) are analogous t o t he two-port network representation of t he terminal properties o f t he n:twork (output voltage y ( s ) and input current i ( s ) ) . The subscripts designate t he corresponding t ransf er f unst i ons. For example , Gv i s t he source voltage v t o output voltage y t r %ns f er funct;on, Gid i s tfFe duty r at i o 8 t o input current i ( s ) t r ans f er f unct i on, and so on. For t he proposed canonical ci r cui t model i n F i g . 11, we di r ect l y get : or , af t e r rearrangement i nt o t he form o f (24) : Direct cornoarison o f ( 24) and (26) ~r o v i d e s t he . - - - . sol ut i ons f or He( s) , e ( s ) , and j ( s ) i n terms of t he known t ransf er f unct i ons Gvg, Gvd, Gig and Gid as: Note t hat i n (27) t he parameter 1/ p represents t he i deal dc voltage gain when a l l t he parasi t i cs are zero. For t he previous boost power stage example, from (19) we get u - 1-D and t he correc- t i on factor i n (19) i s then associated wi t h t he e f f e c t i v e f i l t e r network He(s). However, l~ could be found from 1 ?- = - cT~- l b - - X (correction f act or) (28) v 8 u by set t i ng a l l parasi t i cs t o zero and reducing t he correction f act or t o 1. The physical si gni f i cance of t he i deal dc gain p i s t hat it ari ses as a consequence of t he switching act i on, so it cannot be associated with t he e f f e c t i v e f i l t e r network which at dc has a gain ( act ual l y attenuation) equal t o t he cor- rect i on f act or. The procedure f or finding t he four elements i n the canonical model of F i g . 11 i s now b r i e f l y reviewed. Fi r s t , from (28) t he basi c dc-to-dc conversion f act or p i s found as a f unct i on o f dc duty r at i o D. Next, from t he set o f a l l t r ans f er functions (23) only those defined by (24) are actually calculated. Then, by use of t hese four t r ans f er functions G vd' ' v " d' ' i f (27) the frequency dependent geherators e s and j ( s ) as, wel l as t he low-pass f i l t e r t ransf er function H ( s ) are obtained. The two generators could be f urt her put i nt o t he form where f l ( 0) - f 2 ( 0 ) = 1 , such t hat t he parameters E and J could be i dent i f i ed as dc gains o f t he frequency dependent functions e ( s ) and j ( 6) . Fi nal l y, a general synt hesi s procedure 1101 f or real i zat i on of L,C t r ans f er f unct i ons terminated i n a si ngl e load R could be used t o obtain a low-pass ladder-network ci r cui t real i zat i on o f t he e f f e c t i v e low-pass network He(s). Though f or t he second-order example of He(s) t h i s step i s t r i v i a l and could be done by i nspect i on, f or higher-order t r ans f er functions the orderly procedure o f t he synt hesi s [ l o ] i s almost mandatory. 5.2 Example: Ideal Buck-boost Power Stage For t he buck-boost ci r cui t shorn i n Fin. 7c ,-. i n which the output voltage y coinc!des wi t h the state-variable capacitance voltage v . From (28) and ( 29) one obtains p = D ' / D . Wi t h use o f (29) t o derive t ransf er f unct i ons, and upon subst i t ut i on i nt o ( 27) . there r es ul t s i n which V i s t he dc output voltage. The e f f e c t i v e f i l t e r t r ans f er f unct i on i s eas i l y seen as a low-pass LC f i l t e r wi t h Le = LID'^ and wi t h load R. The two generators i n t he canonical model of F i g . 11 are i dent i f i ed by We now derive the same model but t hi s time using t he equivalent ci r cui t transformations and path b i n t he Flowchart of F i g . 1. Af t er perturbation and l i neari zat i on o f the ci r cui t averaged model i n F i g . 7c (wi t h Re-0) the seri es o f equivalent ci r cui t s %-;ip. 12 i s obtained. p i g . 12. Equivalent ci r cui t transformations of t he f i nal ci r cui t averaged model ( a ) , leading t o i t s canonical ci r cui t real i zat i on (c) demonstrated on t he buck-boost example of Pig. 7c ( wi t h Rf-0 , Rc-0 ). The obj e c t i ve of t he t r ans f ormat i ons i s t o reduce t h e o r i g i n a l f our dut y- r at i o dependent gen- e r a t o r s i n Fi g. 12a t o j u s t two gener at or s ( vol t - age and cur r ent ) i n Fi g. 1 2 ~ which a r e a t t he i n- put por t of t he model. As t hes e c i r c u i t t r ans - format i ons unfol d, one s e e s .how t h e frequency de- pendence i n t he gener at or s a r i e e s n a t u r a l l y , as i n ~ i g . 12b. Al so, by t r a n s f e r of t he two gen- e r a t o r s i n Fi g. 32b from t he secondary t o t he pri mary of t he l : D t r ansf or mer , and t he i nduct ance L t o t he secondary of t he D' : l t r ansf or mer , t he cascade of two i d e a l t ransformers is reduced t o t he s i n g l e t ransformer wi t h equi val ent t ur ns r a t i o D' :D. A t t he same t i me t he e f f e c t i v e f i l t e r network Le, C, R is generat ed. Expresei one f o r t he el ement s i n t he canoni cal equi val ent c i r c u i t can be found i n a s i mi l a r way f o r any conver t er conf i gur at i on. Resul t s f o r t he t hr e e f a mi l i a r conver t er s , t he buck, boost , and buck-boost power s t a g e s a r e summarized i n Tabl e I. Tabl e I Def i ni t i on of t h e el ement s i n t h e canoni cal c i r c u i t model of Fi g. 11 f o r t h e t h r e e common power s t a ge s of Fi g. 7. It may be not ed i n Tabl e I t h a t , f o r t h e buck- boost power s t a ge , paramet ers E and J have negat i ve s i gns , namely E = - v / D~ and J = - v/ ( D' ~R) . However, as seen from t he p o l a r i t y of t h e i d e a l D' :D t r ansf or mer In Fi g. 12c t h i s s t a ge is an i nve r t i ng one. Hence, f o r p o s i t i v e i nput dc vol t a ge V , t h e out put d c vol t a ge V is negat i ve (V < 0 ) s f nce V/Vg - -D/D8. Therefore E > 0, J > 0 and consequent l y t he p o l a r i t y of t he vol t age and c ur r e nt dut y- r at i o dependent gener at or s i s not changed but i s as shown i n Fi g. 12c. Hore- over , t h i s i s t r u e i n general : r e ga r dl e s s of any i nver s i on pr oper t y of t h e power s t a ge , t he p o l a r i t y of two gener at or s s t a y s t he same a s In Fi g. 11. 5.3 Si gni f i cance of t he Canoni cal Ci r c ui t Model and Rel at ed Gener al i zat i ons = . - . - - - . - - . pass f i l t e r i n g ( r epr esent ed by t h e e f f e c t i v e low- pass f i l t e r net vor k H,(s)). Note a l s o t h a t t he c ur r e nt gener at or j ( s ) a i n t he canoni cal c i r c u i t model, eyen though euperfl uoue when t he sour ce vol t age v (a) i e i d e a l , i s necessar y t o r e f l e c t t he i n f l u k c e of a mn i d e a l sour ce gener at or (wi t h some i n t e r n a l impedance) o r of an i nput f i l t e r [ 7] upon t he behavi our of t he conver t er . Its presence enabl es one e a s i l y t o i ncl ude t he l i n e a r i z e d c i r - c u i t model of a swi t chi ng conver t er power s t a ge i n ot her l i n e a r c i r c u i t s , aa t he next s e c t i o n w i l l i l l u s t r a t e . Another s i g n i f i c a n t f e a t ur e of t he canon- i c a l c i r c u i t model i s t h a t any s wi t chi ng dc-to-dc conver t er can be reduced by use of ( 23) , (24). (27) and (28) t o t h i s f i xed t opol ogy form, a t l e a s t a s f a r a s i t s i nput -out put and c ont r ol prop- e r t i e s a r e concerned. Hence t he p o s s i b i l i t y a r i s e s f o r use of t h i s model t o compare i n an easy and uni que way var i ous performance c h a r a c t e r i s t i c s of d i f f e r e n t conver t er s. Some examples of such comparisons a r e gi ven below. 1. The f i l t e r net works can be compared wi t h r es pect t o t h e i r e f f e c t i ve ne s s t hroughout t he dynamic dut y c yc l e D range, because i n gener al t he e f f e c t i v e f i l t e r el ement s depend on t he st eady s t a t e dut y r a t i o D. Thus, one has t he oppor t uni t y t o choose t he conf i gur at i on and t o opt i mi ze t he s i z e and wei ght . 2. Basi c dc-to-dc conversi on f a c t o r s pl(D) and p2(D) can be compared as t o t h e i r e f f e c t i v e range. For some conver t er s, t r a v e r s a l of t he range of dut y r a t i o D from 0 t o 1 gener at es any conversi on r a t i o ( a s i n t he i d e a l buck- boost conver t er ) , whi l e i n ot he r s t h e conver- s i on r a t i o might be r e s t r i c t e d ( a s i n t he Weinberg conver t er [ 4 ] , f o r which kp- 3) . 2 3. I n t h e c ont r ol s e c t i on of t h e canoni cal model one can compare t h e frequency dependences of t he gener at or s e ( s ) and j ( s ) f o r d i f f e r e n t conver t er s and s e l e c t t he conf i gur at i on t h a t be s t f a c i l i t a t e s s t a b i l i z a t i o n of a feedback r egul at or . For example, i n t he buck-boost con- v e r t e r e ( ~ ) i s a pol ynomi al , cont ai ni ng a c t u a l l y a r e a l zer o i n t he r i g h t hal f - pl ane, which undoubt edl y causes some s t a b i l i t y problems and need f o r proper compensation. 4. Fi nal l y, t h e canoni cal model af f or ds a very conveni ent means t o s t o r e and f i l e i nf or - mat i on on var i ous dc-to-dc conver t er s i n a com- put er memory i n a form comparable t o Tabl e I. Then, t hanks t o t he f i xed t opol ogy of t he canoni cal c i r c u i t model, a s i n g l e computer pro- gram can be used t o c a l c ul a t e and p l o t var i ous q u a n t i t i e s as f unct i ons of frequency ( i nput and out put impedance, audi o s u s c e p t i b i l i t y , dut y r a t i o t o out put t r a n s f e r response, and s o on). Al so, var i ous i nput f i l t e r s and/ or a ddi t i ona l out put f i l t e r networks can e a s i l y be added i f desi r ed. f unct i ons of complex frequency s. Hence, ge ne r a l bot h some new zer os and pol es a r e i nt r o- duced i n t o t he dut y r a t i o t o out put t r a n s f e r f unct i on owing t o t he swi t chi ng a c t i on, i n a ddi t i on t o t h e pol es and zer os of t he e f f e c t i v e f i l t e r network ( or l i n e t o out put t r a n s f e r fun- c t i on) . However, i n s p e c i a l c a s e s , a s i n a l l t hose shown i n Tabl e I, t h e frequency dependence mi ght reduce si mpl y t o pol ynomi al s, and even f ur - t he r it might show up onl y i n t h e vol t age dependent gener at or s ( a s i n t h e boost , o r buck- boost ) and reduce t o a const ant ( f ( s ) i 1) f o r t h e c ur r e nt gener at or . ~ e v e r t g e l e s s , t h i s does not pr event us from modi fyi ng any of t he s e Ci r c ui t s i n a vay t h a t would e x h i b i t t h e gener al r e s u l t -- i nt r oduct i on of bot h a ddi t i ona l zer os a s v e l l as pol es. Let ue now i l l u s t r a t e t h i s gener al r e s u l t on a climple modi f i cat i on of t h e f a mi l i a r boost c i r - c u i t , wi t h a r esonant L1,C1 c i r c u i t i n s e r i e s wi t h t he i nput i nduct ance L, a s shown i n Fi g. 13. Fi g. 13. Modified boost c i r c u i t a s an i l l u s t r a t i o n of gener al frequency behavi our of t h e gener at or s i n t h e canoni cal c i r c u i t model of Fi g. 11. By i nt r oduct i on of t h e canoni cal c i r c u i t model f o r t h e boost power s t a ge ( f o r t h e c i r c u i t t o t he r i g h t of c r os s s e c t i on AA') and use of d a t a from Tabl e I, t h e equi val ent averaged c i r c u i t model of Fi g. 14a i s obt ai ned. Then, by appl i ca- t i on of t h e equi val ent c i r c u i t t r ansf or mat i on a s out l i ned pr evi ousl y, t h e averaged model i n t he canoni cal c i r c u i t form i s obt ai ned i n Fi g. 14b. As can be seen from Fi g. 14b, t h e vol t age generat or ha s . a doubl e pol e a t t he r esonant f r e - quency U r = 1/m of t h e p a r a l l e l L1,C, net - work. However, t h e e f f e c t i v e f i l t e r t r a k f e r f unct i on has a doubl e zer o ( n u l l i n magnitude) a t Pr eci s el y t h e same l oc a t i on such t h a t t h e two Fi g. 14. Equi val ent c i r c u i t t r ansf or mat i on l eadi ng t o t h e canoni cal c i r c u i t model (b) of t he c i r c u i t i n Fi g. 13. p a i r s e f f e c t i v e l y cancel . Hence, t h e r esonant n u l l i n t he magnitude r esponse, whi l e pr esent i n t he l i n e vol t age t o out put t r a n s f e r f unct i on, i s not seen i n t h e dut y r a t i o- t o out put t r a n s f e r func- t i on. Therefore, t he pos i t i ve e f f e c t of r e j e c t i on of c e r t a i n i nput f r equenci es around t he r esonant frequency w i s not accompanied by a det r i ment al e f f e c t on t ge l oop gai n, which wi l l not con- t a i n a n u l l i n t he magnitude response. Thi s example demonst rat es y e t anot her impor- t a n t as pect of model l i ng wi t h use of t he averagi ng t echni que. I ns t ead of appl yi ng i t d i r e c t l y t o t he whole c i r c u i t i n Fi g. 13, we have i ns t ead imple- mented i t onl y wi t h r es pect t o t he s t or age el ement network which e f f e c t i v e l y t akes p a r t i n t he swi t ch- i n g a c t i on, namely L, C, and R. Upon s u b s t i t u t i o n of t he swi t ched p a r t of t h e net work by t h e averaged c i r c u i t model, a l l ot he r l i n e a r c i r c u i t s of t h e complete model a r e r e t a i ne d a s t hey appear i n t he o r i g i n a l c i r c u i t (such as L1,C1 i n Fi g. 14a) . Again, t h e cur r ent gener at or i n Fi g. 14a i s t h e one which r e f l e c t s t he e f f e c t of t he i nput r esonant c i r c u i t . I n t he next s e c t i on, t he same pr oper t y i s c l e a r l y di spl ayed f o r a cl osed-l oop r egul at or - conver t er wi t h o r wi t hout t he i nput f i l t e r . 6. SWITCHING MODE REGULATOR MODELLING Thi s s e c t i on demonst rat es t h e eas e wi t h which t h e d i f f e r e n t conver t er c i r c u i t models developed i n previ ous s e c t i ons can be i ncor por at ed i n t o more compl i cat ed syst ems such a s a swi t chi ng- mode r egul at or . I n a ddi t i on, a b r i e f di s cus s i on of model l i ng of modul at or s t a ge s in gener al i s i ncl uded, and a compl et e gener al switching-mode r egul at or c i r c u i t model i s gi ven. A gener al r epr es ent at i on .of a swi t chi ng-mode r e gul a t or i s shown i n Fi g. 15. For concr et eness, t h e swi t chi ng-mode conver t er i s r epr esent ed by a buck-boost power s t age, and t h e i nput and pos s i bl e a ddi t i ona l out put f i l t e r a r e r epr esent ed by a ,unrroulated nnout rrquf at ed output 7 i nput and out put f i l t e r s . The bl ock gram i s gener al , and si ngl e- sect i on dia- LC f i l t e r s and a buck-boost conver t er a r e shown a s t y p i c a l r e a l i z a t i ons . si ngl e-sect i on low-pass LC conf i gur at i on, but t h e di scussi on appl i es t o any convert er and any f i l t e r conf i gur at i on. The main d i f f i c u l t y i n anal ysi ng t h e swi t ch- ing r egul at or l i e s . i n t he model l i ng of i t s non- l i n e a r Pa r t , t he switching-mode conver t er . HOW- ever, we have succeeded i n previ ous s e c t i ons i n obt ai ni ng t h e smal l -si gnal low-frequency c i r c u i t model of any "two-state" swi t chi ng dc-t o-dc con- ver t er , oper at i ng i n t he cont i nuous conduct i on mode, i n t h e Canonical Ci r c ui t form. The out put f i l t e r i s shown s epar at el y, t o emphasize t he f a c t t h a t i n averaged model l i ng of t he switching-mode conver t er onl y t he st or age el ement s which a r e act ual l y involved i n t he swi t chi ng act i on need be t aken i n t o account , t hus minimizing t h e e f f o r t i n i t s model l i ng. The next s t e p i n development of t he regul a- tor equi val ent c i r c u i t is t o obt ai n a model f o r t he modulator. Thi s i s e a s i l y done by wr i t i ng an expressi on f or t he e s s e n t i a l f unct i on of t he modu- l a t o r , which i s t o Convert an (analog) cont r ol vol t age Vc t o t he swi t ch dut y r a t i o D. Thi s ex- pr essi on can be wr i t t e n D = V /Vm i n which, by de f i ni t i on, Vm i s t he range oh cont r ol s i g n a l requi red t o sweep t he dut y r a t i o over i t s f u l l range from 0 t o 1. A smal l va r i a t i on vc super- imposed upon Vc t her ef or e produces a correspon- di ng var i at i on a = Gc/vm i n D, which can be general i zed t o account f or a nonuniform frequency response as 'm C i n which fm(0) 1. Thus, t he cont r ol vol t age t o dut y r a t i o smal l -si gnal transmission char act er - i s t i c of t he modulator can be represent ed i n gen- e r a l by t he two paramet ers Vm and f m( s ) , regard- l e s s of t he det ai l ed mechanism by which t he modu- l a t i o n i s achieved. Hence, by s ubs t i t ut i on f o r a from (32) t he two gener at or s i n t he canoni cal c i r c u i t model of t he swi t chi ng conver t er can be expressed i n terms of t h e a c cont r ol vol t age cc, sad t he r e s ul t i ng model is t hen a l i n e a r a c equi - val ent c i r c u i t t h a t r epr esent s t he smal l -si gnal t r a ns f e r pr oper t i es of t he nonl i near processes i n t he modulator and convert er. It remains si mpl y t o add t he l i n e a r ampli- f i e r and t he i nput and out put f i l t e r s t o obt ai n t he a c equi val ent c i r c u i t of t he complete cl osed- loop r egul at or as shown i n Fi g. 16. The modulator t r a ns f e r f unct i on has been i n- corporat ed i n t he gener at or desi gnat i ons, and t he gener at or symbol has been changed from a c i r c l e t o a square t o emphasize t he f a c t t ha t , i n t h e closed-loop r egul at or , t he gener at or s no l onger a r e independent but a r e dependent on anot her s i g- n a l i n t he same system. The connect i on from poi nt Y t o t he e r r o r ampl i f i er , v i a t he r ef er ence vol t age summing node, r epr esent s t he b a s i c v01- t age feedback necessary t o e s t a bl i s h t h e syst em as a vol t age r egul at or . The dashed connect i on from Poi nt Z i ndi cat es a poes i bl e a ddi t i ona l feedback sensi ng ; t h i s second feedback s i g n a l may V, + cg converter and modulator model E e,(s)= - f , ( ~ ) f ~ ( s ) ~ ~ ------- Vm dc ref Pig. 16. General smal l -si gnal a c equi val ent c i r c u i t f o r t h e swi t chi ng-mode r egul at or of Fi g. 15. be deri ved, f o r example, from t he i nduct or f l ux, i nduct or cur r ent , o r capaci t or c ur r e nt , a s i n var i ous "two-loop" conf i gur at i ons t h a t a r e i n use t91. Once agai n t he c ur r e nt gener at or i n Fi g. 16 i s r esponsi bl e f o r t he i nt e r a c t i on between t he switching-mode r egul at or - conver t er and t he i nput f i l t e r , t hus causi ng performance degradat i on and/ o r s t a b i l i t y problems when an a r b i t r a r y i nput f i l t e r is added. The problem of how pr oper l y t o desi gn t he i nput f i l t e r is t r e a t e d i n d e t a i l i n (71 As shown i n Fi g. 16 we have succeeded i n ob- t a i ni ng t he l i n e a r c i r c u i t model of t he complete swi t chi ng mode-regulator. Hence t he we l l - k n m body of l i n e a r feedback t heory can be we d f or bot h anal ys i s and desi gn of t h i s t ype of regul a- t or . 7. CONCLUSIONS A gener al method f o r model l i ng power s t ages of any swi t chi ng dc-to-dc conver t er ha s been developed through t h e st at e- space approach. The fundamental s t e p is i n repl acement of t he s t a t e - space des cr i pt i ons of t he two swi t ched networks by t h e i r average over t he s i n g l e swi t chi ng per i od T, which r e s u l t s i n a s i n g l e cont i nuous s t a t e - space equat i on des cr i pt i on (3) desi gnat ed t he bas i c averaged s t at e-space model. The e s s e n t i a l approxi mat i ons made a r e i ndi cat ed i n t he Appen- di ces , and a r e shown t o be j u s t i f i e d f o r any p r a c t i c a l dc-to-dc swi t chi ng conver t er . The subsequen t per t ur bat i on and l i n e a r i - zat i on step under t he smal l -si gnal aesumption (12) l eads t o t he f i n a l st at e- space averaged model gi ven by (13) and (14). These equat i ons t hen s e r ve as t he b a s i s f o r development of t he most i mport ant q u a l i t a t i v e r e s u l t of t h i s vork, t he canoni cal c i r c u i t model of Fi g. 11. Di f f er ent conver t er s a r e r epr esent ed si mpl y by an appropri - a t e s e t of formulas ((27) and (28)) f o r f our el ement s i n t h i s gener al equi val ent c i r c u i t . Be- s i de s i t s uni f i ed des cr i pt i on, of which s ever al examples a r e gi ven i n Tabl e I, one of t he sdvan- t ages of t he canoni cal c i r c u i t model is t h a t vari ous performance c ha r a c t e r i s t i c s of di f f e r e nt swi t chi ng conver t er s can be compared i n a qui ck and easy manner. Although t he st at e- space model l i ng approach has been developed i n t h i s paper f o r t wo-st at e swi t chi ng conver t er s, t he method can be extended t o mul t i pl e- st at e conver t er s. Examples of t hree- s t a t e conver t er s a r e t he f ami l i ar buck, boost , and buck-boost power s t ages operat ed i n t he dl s - cont i nuous conduct i on mode, and dc-to-ac swi t ch- i ng i nve r t e r s i n which a s p e c i f i c out put wave- form i s "assembled" from d i s c r e t e segments a r e examples of mul t i pl e- s t at e conver t er s. I n c ont r a s t wi t h t he st at e- space model l i ng approach, f o r any p a r t i c u l a r convert er an a l t e r - nat i ve pat h v i a hybri d model l i ng and c i r c u i t t ransformat i on could be fol l owed, which a l s o ar - r i ves f i r s t a t t he f i n a l c i r c u i t averaged model equi val ent of (13) and (14) and f i n a l l y , a f t e r equi val ent c i r c u i t t ransformat i ons, agai n a r r i ve s a t t he canoni cal c i r c u i t model. Regardl ess of t he der i vat i on pat h, t he canoni cal c i r c u i t model can e a s i l y be i ncorpora- t ed i n t o an equi val ent c i r c u i t model of a com- pl e t e swi t chi ng r egul at or , a s i l l u s t r a t e d i n Fi g. 16. Perhaps t he most i mport ant consequence of t he canoni cal c i r c u i t model der i vat i on v i a t he gener al st at eyapace averaged model (13). ( 14) , (23) and (24) i s i t s pr edi ct i on through (27) of addi t i onal zeros a s we l l a s pol es i n t he dut y r a t i o t o out put t r a ns f e r f unct i on. I n addi t i on frequency dependence i s ant i ci pat ed i n t he dut y r a t i o dependent cur r ent gener at or of Fi g. 11, even though f o r p a r t i c u l a r conver t er s consi dered i n Tabl e I, i t reduces merely t o a const ant . Furthermore f o r some awi t chi ng networks vhi ch would e f f e c t i ve l y i nvol ve more t han two s t or age el ement s, hi gher or der polynomials shoul d be ex- pect ed i n f l ( s ) and/ or f 2( s ) of Fi g. 11. The i ns i ght s t h a t have emerged from t he gener al st at e- space model l i ng approach suggest t h a t t her e i s a whole f i e l d of new swi t chi ng dc- to-dc convert er pover s t ages yet t o be conceived. Thi s encourages a renewed sear ch f o r i nnovat i ve c i r c u i t desi gns i n a f i e l d which is y e t young, and promises t o yi e l d a s i g n i f i c a n t number of i n- vent i ons i n t he st ream of i t s f u l l development. This progress wi l l na t ur a l l y be f u l l y support ed by new t echnol ogi es coming a t an ever i ncr eas i ng pace. However, even though t h e ef f i ci ency and performance of cur r ent l y e xi s t i ng conver t er s wi l l i ncr ease through bet t er , , f a s t e r t r a n s i s t o r s , more i d e a l capaci t or s (wi t h lower ear ) and s o on, it wi l l be pr i mar i l y t he r e s pons i bi l i t y of t he c i r - c u i t desi gner and i nvent or t o put t hese components t o bes t use i n an o p t i ~ l topology. Search f or new c i r c u i t conf i gur at i ons, and h w be s t t o use pr esent and f ut ur e t echnol ogi es, w i l l be of prime importance i n achi evi ng t he ul t i mat e goal of near- i de a l gener al swi t chi ng dc-to-dc convert ers. REFERENCES R. D. Middlebrook, "A Continuous Model f o r t he Tapped-Inductor Boost Convert er, " IEEE Power El ect r oni cs Sp e c i a l i s t s Conference, 1975 Record, pp. 63-79 (IEEE Publ i c a don 75 CHO 965-4-AES). G. W. Wester and R. D. Middlebrook, "Low- Frequency Char act er i zat i on of Switched dc- dc Convert ers, " IEEE Tr ans. ' on Aerospace and El ect r oni c Systems, Vol. AES-9, No. 3 , May 1973, pp. 376-385. R. Haynes, T. K. Phel ps. J. A. Col l i ns , and R. D. Middlebrook, "The Venable Convert er: A New Approach t o Power Processi ng, " IEEE Power El ect r oni cs Sp e c i a l i s t s Conference, NASA Lewis Research Cent er, Cl evel and, Ohio, June 8-10, 1976. A. H. Weinberg, "A Boost Regul at or wi t h a New Energy-Transfer Pr i nci pl e, " Proceedi ngs of Spacecr af t Power Condi t i oni ng Seminar, pp. 115-122 (ESRO Publ i cat i on SP-103, Sept . 1974). H. F. Baker, "Q t he I nt e gr a t i on of Li near Di f f e r e n t i a l Equations,' ' Proc. Lcndon Math. SOC., 34, 347-360, 1902; 35, 333-374, 1903; second s e r i e s , 2, 293-296, 1904. R. D. Middlebrook and S. Cuk, Fi na l Report , "Modelling and Anal ysi s of Power Processi ng Systems,'' NASA Cont ract NAS3- 19690. R. D. Middlebrook, "Input F i l t e r Considera- t i ons i n Design and Appl i cat i on of Swi t chi ng Regul at ors, " IEEE I ndust r y Appl i cat i ons Soci et y Annual Meeting, Chicago, Oct. 11- 14, 1976. R. D. Middlebrook, "Descri bi ng Funct i on Pr oper t i es of a Magnetic Pul sewi dt h Modulator," IEEE Trans. on Aerospace and El ect r oni c Systems, Vol. AES-9, No. 3, May, 1973, pp. 386-398. Y. Yu, J. J. Bi ess, A. D. Schoenfel d and V. R. Lal l i , "The Appl i cat i on of St andard- i zed Cont r ol and I nt e r f a c e Ci r c ui t s t o Three DC t o DC Power Converters,' ' IEBE Power El ect r oni cs Spe c i a l i s t s Conference, 1973 Record, pp. 237-248 (IEEE Publ i cat i on 73 CHO 787-2 AES). F. F. Kuo, "Network Anal ysi s and Synt hesi s, " John Wiley and Sons, I nc.
Mental Math: How to Develop a Mind for Numbers, Rapid Calculations and Creative Math Tricks (Including Special Speed Math for SAT, GMAT and GRE Students)