25. Estimating the Ou Votage Ripe in Comerers Containing To Pte Low-PassFters 3 238) a c Equation (2.51) was used to substitute forthe values of, Yaad jy during the second subinterval During the first subinterval, the quantities (0, 4(9). and v( change by 2Ai,, ~ 28%, and ~ 2An, respectively. These changes are equal to the slopes given in Eq. (2.54), multiplied by the sub- interval length DT,, yielding 256) ‘The de relationships, Eg. (2.53), can now be used to simplify these expressions and eliminate V;, V3, and Jj leading to esp ‘These expressions can be used (0 select values of L, Ly and Cy, such that desired valves of switching ripple magnitudes are obtained, Similar arguments cannot be used to estimate the switching ripple magnitude in the output capacitor voltage vs(). According to Fig. 2.22(0), the curenti() is continuous: unlike v,, Vj, nd ici ‘the capacitor curent i() is nonpulsatng. Ifthe switching ripple of i) is neglected, then the capacitor ceurtent (0 does not contain ana component. The smal-ripple approximation then leads tothe conclu- sion thatthe output switching ripple Av, is zero, (Of course, the output voltage switching ripple isnot zero. To estimate the magnitude ofthe out- put voltage ripple in this converter, we must not neglect the switching ripple present inthe inductor eur rent i,(), since this current ripple isthe only source of ac current driving the output capacitor CA simple way of doing this inthe Cuk converter and in other similar converters i discussed inthe next sec- tion, 25 ESTIMATING THE OUTPUT VOLTAGE RIPPLE IN CONVERTERS CONTAINING ‘TWO-POLE LOW-PASS FILTERS ‘A case where the small ripple approximation i not useful is in converters containing two-pole low-pass filters, suchas inthe output ofthe Cuk converter (Fig. .20) or the buck converter (Fig, 2.25) For these32 Principles of Seay State Comer Analyt Fig. 225 The buck converte contains a two-poe opt files converters, the smal-ripple approximation predicts zero output voltage ripple, regardless of the value of ‘the outpu filter capacitance. The problem is that the only component of output capacitor current in these «eases is that arising from the inductor curren ripple, Hence, inductor current ripple cannot be neglected ‘when calculating the output capacitor voltage ripple, and a more accurate approximation is needed, ‘An improved approach that is useful for this case is to estimate the capacitor current waveform {{0) more accurately, accounting forthe inductor eurent ripple. The capacitor voltage ripple can then be related to the total charge contained in the positive portion of the it) waveform, Consider the buck converter of Fig. 2.25. The inductor current waveform i,() contains a de component 1 ang linear ripple of peak magnitude Ai, as shown in Fig. 2.10. The de component J must flow entirely through the load resistance R (why?), while the ac switching ripple divides between the load resistance R and the filter capacitor C, Ina well-designed converter, in which the capacitor provides significant filtering ofthe switching ripple, the capacitance Cis chosen large enough that its impedance at the switching frequency is much smaller than the load impedance R. Hence nearly all ofthe inductor curtent ripple flows through the capacitor, and very litle flows through the load. As shown in Fig, 2.26, the capacitor current waveform i,() is then equal tothe inductor current waveform withthe de compo: nent removed. The current ripple i linear, with peak value Ai. ‘When the capacitor current (i positive, charge is deposited on the capacitor plates and the capacitor voltage v(t) increases, Therefore, between the two zero-crossings of the capacitor current ‘waveform, the capacitor voltage changes between its minimum and maximum extrema, The waveform is symmetrical, and the total change in vi the peak-to-peak output voltage ripple, of 2Av. ‘This change in capacitor vollage ean be related tothe total charge q contained in the positive igo Total charge 4 Fe 226 ouput cipconvatage y fe and current waveforms, for the buck = Comererin Fig 225 —u— F- 7, 4 or, i vo $ nf gay Tas225. Estimating he Oupu Vag Ripe Comers Containing Two-PleLow-Pse Fars 38 Ot trap flux linkage x ¥ fav ‘ Fig. 227. Estimating inductor curent i ripple when the inductor voltage wave i formnis continuous, 72 -— pr,—++—_—— 7, 0 {ai portion of the capacitor current waveform. By the capacitor relation Q= CK, 4=C(28) (238) As illustrated in Fig. 2.26, the charge qs the integral of the cutent waveform between its zero crossings For this example, the integral can be expressed asthe area ofthe shaded wrangle, having a height Ai, Owing othe symmetry of the eurent waveform, the 270 crossings occur atthe centerpoint ofthe DT, and D7, subintervals, Hence, the base dimension of the tingle is 7/2. So the oa charge qi given by 239) a= hai Substitution of Eg, (2.58) into Eq, (2.59), and solution for the voltage ripple peak magnitude Av yields ait, 2.60 aye th (2.60) ‘This expression can be used to select a value forthe capacitance C such that a given voltage ripple Av is jobiained. In practice, the addtional voltage ripple caused by the capacitor equivalent series resistance (est) must also be include. Similar arguments can be applied to inductors. An example is considered in Problem 2.9, in ‘hich a two-pole input filter is added to a buck converter as in Fig. 2.32. The capacitor voltage ripple cannot be neglected; doing so would ead to the conclusion that no ac voltage is applied across the input filter inductor, resulting in 270 input current ripple. The actual inductor voltage waveform is identical to the ac portion ofthe input filter capacitor voltage, with linear ripple and with peak Value Av as illustrated in Fig. 2.27. By use ofthe inductor relation A= Li, a result similar to Eq, (2.60) can be derived, The der- ivation i left asa problem for the student26 Principles of Seay State Comerter Anal SUMMARY OF KEY POINTS ‘The de component ofa converter waveform i given by is average value, or te integral over one switeh- ing period, divided by the switching peiod. Solution of a de-de converter 10 find is de, of steady-state, voltages and curens therefore involves averaging the waveforms. ‘The linea. (or small) ripple approximation greatly simplifies the analysis, Ina well-designed converter, the switching ripples inthe inductor curent and capacitor voltage are small compared tothe respective <é components, and canbe neglected, “The principle of inductor voltsecond balance allows determination ofthe de voltage components in any switching converte. In steady tat, the average voltage applied to an inductor must be er, ‘The principle of capacitor charge balance allows determination ofthe de components of the inductor eur ‘ents in a switching converte. In steady state, the average current applied oa capacitor must be 22. By knowledge ofthe slopes ofthe inductor curent and capacitor vltage waveforms, the ac switching rip ple magnitudes may be computed, Inductance and capacitance values can then be chosen to obtain desired ripple magnitudes. In converters containing mutipl-poe fers, continuous (aonpusating) voltages and currents ae applied {o one oe mor ofthe inductors o capacitors. Computation ofthe ac switching ripple in these element can ‘be done using capacitor charge andor inductor Mux-linkage arguments, without use ofthe smalpple approximation, (Converters capable of increasing (boost), dereasng (buck), and inverting the voltage polarity (buck boost and Cub) have been deseribed. Converter eres are explored more fall inthe problems and ina Tater chapter REFERENCES uy ey B) ™ i to nm 8) 8. CuK, “Basis of Switchad-Mode Power Conversion: Topologies, Magnetics, and Control," n Advances In Switched-Mode Power Conversion, No.2, pp. 279-310, lve, CA: Tesaeo, 198] 1N, MowaX, T, UNDELAND, and W. ROBBINS, ower Electronics: Converters, Applications, and Design, 2° edi, New York: John Wiley & Sons, 1995, {LKASSAKIAN, M, SCHLECHT, and G, VEROESE, Principles of Power Electronics, Reading, MA: Addison- Wesley, 1991 R.SEVERNS and GE. BLOOM, Modem De-tode Switch Mode Power Converter Circuits, New York: Van Nostrand Reinhold 1985, D.HxRr,Innoducton so Power Electronics, New York: Preatice Hal, 197, 1M, RastD, Power Electonics: Cireuls; Devices, and Applications, 2 edit, New York: Prentice Hal 1983, P. KREIN, Elements of Power Electronics, New York: Oxford University Press, 198, K, Kir SUM, Switch Mode Power Conersion-Basie Theory and Design, New York: Marcel Dekker, 198,PROBLEMS a 22 Proslems = 38 ‘Analysis and design of buck-boost converter A buck-boost converters ilstratd in practical implementation using wansisior and diode is shown i Fig. 2.280). i. 2.28), and ” es + cc Fig. 228 Buck-boost comer, cm rey ‘of Problem 2.1: (a ideal converter L circuit, (0) implementation using [MOSFET and dio, - co) o >, (a) Find the dependence of the equiibium outpt voltage V and inductor cutent onthe duty rato _. input voltage V, and load resistance R. You may assume tat the inductor current ripple ad capacitor voltage ripple are smal (©) Plot your esuls of part (a over the range 0S Ds (©) Dedesign: forthe speciieations ¥,=30 v=-20v a f=40 ee (Find Dand (i) Caleta te vale of that will make the peak inductor current ripple A equal to ten pr cent ofthe average inductor curt (Gi) Choose C suc that he peak output voltage ripple Av is. V, (@) Sketch the transistor drain current waveform i) for your design of pas (ence the effects ‘of inductor curent iple. What she peak vue off? Also sketch i) forthe case when Lis decreased such that Ii 50% off What happens tothe peak value off this ease? (©) Sketch the diode current waveform i) for the two cases of pat) Ina certsin application, an unregulated de input voltage can vary between 18 and 36, Is desired to produce a regulated outpt of 28 V to supply a2 load. Hence, a converters neoded that i capable of both increasing and decreasing the voltage. Since the inpat and output voltages are both psi, com ‘verters that invert the voltage polarity suchas te basi bak: boost converter) are not site for this application. One converter thai eapabl of performing the required function isthe nonisolated SEPIC (sn sle-ended primary inductance converter) shown in Fig, 229, This converter asa conversion ratio M(D) ‘hat can both buck ad bost te voltage, but he voltage polarity not inverted, Inthe normal converter operating mode, the transistor conducts ding the fist subinterval (<1 < DT.) andthe diode conducts during the second subinterval (D7, <1 Ww Fla, 321 Bauivalent ccc of the bck ¥ convert, ncn aD de tansforner Ade sole scr windngsesisane yy k sod Dh, 326 ‘The integral off) is equal othe area under the i) curve, or DT, according to Fig. 3.18. The de com- ponent /, is therefore (DT,,T, = Dl, Equation (3.26) states that, the de component of current drawn, by the converter out ofthe V, source, is equal to DI, An equivalent circuit is given in Fig. 3.19. ‘A complete model for the buck converter can now be obtained by combining Figs, 3.17 and 3.19 to obsain Fig. 3.20, The dependent current and voltage sources can be combined into a de trans- ‘ormer, since the DV, dependent voltage source has value D times the voltage V, across the dependent current source, and the curent source is the same constant D times the current /, through the dependent voltage source. So, according to Fig, 3.2, the sources are equivalent to a de transformer with turns ratio 1:D, as shown in Fig. 3.21 In general, 10 obtain a complete de equivalent circuit that models the converter input port, i is necessary 10 write an equation forthe de component of the converter input current. An equivalent circuit$2 Steady Sta Egulont Cit Moding, Lane ad Eiceney corresponding to this equation is then constructed. In the case ofthe buck converter, as well as in other ‘converters having pulsating input currents, this equivalent circuit contains a dependent current source hich becomes the primary of a de transformer model. In the boost converter example of Section 3.3 it vas unnecessary to explicitly write this equation, because the input current (2) coincided with the induetor current i, and hence a complete equivalent circuit could be derived ising only the inductor voltage and capacitor current equation. 3 EXAMPLE: INCLUSION OF SEMICONDUCTOR CONDUCTION LOSSES IN THE. BOOST CONVERTER MODEL Asa final example, let us consider modeling semiconductor conduction loses in the boost converter of Fig, 322. Another major source of power loss is the conduction loss due to semiconductor device for- ward voltage drops. The forward voltage of a motal oxide semiconductor feld-ffect transistor (MOS- FET) of bipolae junction transistor (BIT) can be modeled with reasonable accuracy as an on-tesistance 2... Inthe case ofa diode, insulated-gate bipolar transistor (IGBT), or thyristor, a voltage source plus an on-resistance yields a model of good accuracy; the on-esistance may be omitted ifthe converter is being ‘modeled ata single operating point ‘When the gate drive signal i high, the MOSFET turns on and the diode is reverse-biased. The circuit then reduces to Fig. 3.23(a). Inthe conducting state, the MOSFET is modeled by the on-esistance R.y- The inductor winding resistance is again represented as in Fig. 35. The inductor voltage and capaci- tof current are given by = Ri YI am idn=—fo- ‘The inductor current and capacitor voltage have again been approximated by their de components. ‘When the gate drive signal is low, the MOSFET tumns off. The diode becomes forward-biased by the inductor current, and the circuit reduces to Fig. 3.23(b). In the conducting state, the diode is mod- ced in this example by voltage source Vand resistance R,- The inductor winding resistance is again ‘modeled by resistance R The inductor voltage and capacitor current for this subinterval are V,=IR, Vo ER G28) Fig. 322 Boost converter example.5.5. Example Inlon of Semiconductor Conaction Lowe in the Bout Comenter Model o © Fig. 323 Boost converter circuits (a) when MOSFET conducts, 6) when diode conduct The de component ofthe inductor voltage is given by (02) =0[v, 18,1) +0 (Vj-1RsVp~Ry-V) 62 By collecting terms and noting that D+ D’= 1, one obtains Vy-1R,=1DRa.~DVq~IDRp~ DV =0 ex ‘This equation describes the de components ofthe voltages around a loop containing the inductor, with Joop current equal to the de inductor current /.An equivalent cccutis given in Fig. 3.25, Wo |-— pr, —| Fig. 324 Inductor vitae »(9 and ; apacitor event ic) waveforms, for the converter of Fig. 3.22, V, “IR, ~Vp— IRV io Ive vm54 Stoady Stat Eure Cit Maelng Lones, and Eiioncy Fig. 328 Equivalent circuit ‘oesponding 10 Bg, (30). wR Fig, 326 Equivalent circuit comresponding (0 Eq, (3:32). oi ft v R ‘The de component of the capacitor current is 63 ‘Upon collecting terms, one obtains (332) This equation describes the de components of the curents flowing into a node connected fo the capacitor, with de capacitor voltage equal to V. An equivalent circuits given in Fig. 3.26. “The two circuits are drawn together in 3.27. The dependent sources ate combined into an ideal ‘D/:1 transformer in Fig. 3.28, yielding the complete de equivalent circuit model Solution of Fig. 3.28 for the output voltage V yields 633) Fig. 327 The circuits of Figs. 325 and 326, drawn together45. Example: nln of Semiconductor Coacion Loses in th Bost Cometer Model $8 Fig. 328 Equivalent circuit model ofthe boost converter of Fig. 3.22 including ideal dc transformer, inductor ‘winding sistance, and MOSFET and diode conduction losses Dividing by V, gives the voltage conversion ratio: 03 ean be seen thatthe effect of the loss elements Voy Ry, Ray ralio below the ideal value (1/D’). ‘The eicency is given by 9 and R, isto decrease the voltage conversion VD'L Hence, » JP, From Fig. 328, P,,= Vand P.,, 635) For high efficiency, we require Wo te 039 t= Dh Iemay seem strange thatthe equivalent circuit model of Fig. 3.28 contains effective resistances DR, and DR, whase values vary with duty eycle, The reason for this dependence is thatthe semiconductor on- resistances are connected in the circuit nly when their respective semiconductor devices conduct, For example, at D = 0, the MOSFET never conducts, andthe effective resistance DR,, disappears from the model. These effective resistances correctly model the average power losses in the elements. For instance, the equivalent circuit predicts thatthe power loss in the MOSFET on-resistance is FDR... In ‘the actual circuit, the MOSFET conduction loss is PR,, while the MOSFET conducts, and 2era while the MOSFET is off. Since the MOSFET conducts with duty eycle D, the average conduction los is DPR,,, ‘which coincides with the prediction ofthe model In general, to predict the power loss ina resistor R, we must calculate the root-mean-square cur ‘et Ip through the resistor, rather than the average current. The average power loss is then given by Ing2R. Nonetheless, the average model of Fig. 3.28 correctly predicts average power loss, provided that ‘the inductor current ripple is small. For example, consider the MOSFET conduction loss in the buck con- verter. The actual transistor current waveform is sketched in Fig. 3.29, for several values of inductor cur- rent ripple Ai. Case (a) corresponds to use of an infinite inductance F, leading to zero inductor current ripple. As shown in Table 3.1, the MOSFET conduction loss is then given bY Ipg?Ry = DI?R, Which