Special family of PWM-based sliding-mode voltage controllers for basic DC–DC converters in discontinuous conduction mode S.-C. Tan, Y.M. Lai, C.K. Tse and L. Martinez-Salamero Abstract: A family of fixed-frequency pulsewidth-modulation-based sliding-mode voltage controllers for DC–DC converters operating in the discontinuous conduction mode is proposed. The proposed topology is developed for buck, boost and buck–boost converters. Preliminary verification and evaluation of these controllers are performed through computer simulations using precise models of the systems. 1 Introduction DC–DC converters can operate in either continuous conduction mode (CCM) or discontinuous conduction mode (DCM), depending on the choice of the switching frequency, and the relative sizes of the load and the inductive storage. In practice, DCM operation enjoys a faster transient response at the expense of higher device stresses. It is still a popular operating mode for low power applications and its practical importance should not be overlooked. Similar to CCM converters, the regulation of DCM converters relies mainly on conventional PWM current or voltage controllers with compensation networks designed from the small-signal linearised models of the converters. Naturally, being small-signal based, they inherit the drawback of being optimal only in regions within the specified operating condition. In operating conditions that deviate greatly from the specified point, poor control performance is experienced. In the case of load deviation, such degradation in performance is typically more prominent in DCM converters than in their CCM counterparts. This is because output load has a greater influence on the system’s characteristics when the converter is operated in DCM. Hence, unlike CCM converters whereby a relatively good load regulation can be obtained for a wide range of operating conditions using a fixed compensation network in the controller, the same cannot be achieved in DCM converters. In fact, the conventional practice of using a fixed compensation network for the control of DCM converters is, strictly speaking, inaccurate. It is necessary to consider the output load into the computations of the controller if good load regulation is desired. This can be achieved through adaptive control means whereby the compensation network of the controller changes with the variation of the output load. However, discounting the drawback of requiring r The Institution of Engineering and Technology 2007 doi:10.1049/iet-epa:20060067 Paper first received 20th February and in final revised form 24th May 2006 S.-C. Tan, Y.M. Lai and C.K. Tse are with the Department of Electronic and Information Engineering, The Hong Kong Polytechnic University, Hong Kong, People’s Republic of China L. Martinez-Salamero is with the Department of Electrical, Electronic and Automatic Control Engineering, Rovira i Virgili University, Taragona, Spain E-mail: ensctan@polyu.edu.hk 64 additional current sensing, the implementation of such schemes is still too computationally exhaustive to be feasible. Hence, to truly benefit from the advantages of employing converters operating in DCM without sacrificing the regulation property, it is necessary to search for better means of controlling these converters. One such alternative is to adopt a nonlinear controller: the sliding-mode controller. The sliding-mode (SM) controller is well known for its robustness, stability and good regulation properties in a wide range of operating conditions. It is also deemed to be a better candidate than other nonlinear controllers for its relative ease of implementation [1–4]. In particular, the fixedfrequency pulsewidth-modulation (PWM)-based SM controllers, which are basically pulsewidth modulators that employ control signals derived from SM control technique, are found to be more suited for practical implementation in power converters [1, 5–10]. However, the results presented in these papers are only valid for converters in CCM operation. As with conventional hysteresis modulation (HM)-based SM controllers, the proposed PWM-based SM controllers are not applicable to DC–DC converters operating in DCM because of the fundamental difference in the dynamic properties between the two operations. Thus, if PWM-based SM controllers are to be adopted for DC–DC converters in DCM operation, the system models and control laws have to be redeveloped with consideration of these properties. In this paper, a family of fixed-frequency pulsewidthmodulation-based sliding-mode voltage controllers is developed for basic DC–DC converters operating in discontinuous conduction mode. The methods of deriving the system models and control laws are illustrated. Computer simulations of the precise systems model have been used for evaluation and verification purposes. Furthermore, it is also worth mentioning that the proposed approach to deriving these controllers can be used for the design of multi-switches/ multi-structured converter systems, e.g. power factor correction (PFC) and parallel-connected converters in either CCM or DCM operation. 2 State-space model of DC–DC converters under DCM Naturally, the state-space model of the converters required in the controller development in this work must be that of IET Electr. Power Appl., Vol. 1, No. 1, January 2007 iL iL uL = 1 uL = 1 diLf diLr dt uB = 1 dt 0 Fig. 1 uB = 0 diLf diLr dt u =1 uL = 0 dt uB = 1 dT u=0 t T t u=0 u =1 0 dT (d+d2 )T T Typical inductor current behaviour of DC–DC converter in CCM (left) and DCM operations (right) Table 1: State-space model of buck, boost, and buck–boost converters in DCM Type of converter diLr dt diLf dt Buck vi vo L vo L 1 L Boost vi L vi vo L 1 L Buck–boost vi L vo L 1 L iL a DCM converter. The difference between this model and the model of a CCM converter is the addition of the zeroinductor-current stage. In this work, we develop the DCM converter models by introducing additional terms, known as the virtual switching components, into the model. It should be noted that such an analogy is only a theoretical representation. No additional physical switch is added to the converter circuit. For the cases of buck, boost and buck–boost converters, the virtual switching components uL and uB, on top of the actual physical switching component u where logic 1 and 0 represent the ‘ON’ and ‘OFF’ stages of the actual power switch, are introduced into the models: uL ¼ 1 ¼ ‘ON’ when iL 40 1 ¼ ‘OFF’ when iL ¼ 0 ð1Þ and uB ¼ 1 ¼ ‘ON’ when iL 40 and u ¼ 0 1 ¼ ‘OFF’ when iL ¼ 0 and u ¼ 0 ð2Þ Here, condition (1) infers that uL inherits a logic state 1 whenever inductor current iL is conducting, and condition (2) infers that uB inherits a logic state 1 only when inductor current is conducting and the power switch is off (u ¼ 0). Figure 1 shows the typical inductor current behaviour of a DC–DC converter in both CCM and DCM operation. The respective rate of change of inductor currents (i.e. diLr/dt and diLf/dt) and state space inductor current iL descriptions for the buck, boost and buck–boost converters are given in Table 1. It can be seen that, in CCM operation, u. Thus, a model uB is always 1 whenever u ¼ 0, i.e. uB ¼ derived for the DCM operation, which consists of uB, can be easily transformed to a model for the CCM operation, u. through the substitution of the condition uB ¼ The relationship between the converters’ state-space output voltage vo, inductor current iL, switching signal u/ u, and load current ir are also given in Table 1. IET Electr. Power Appl., Vol. 1, No. 1, January 2007 3 vo Z Z Z ½vi u vo uL dt 1 C ½vi u þ ðvi vo ÞuB dt 1 C ½vi u vo uB dt 1 C Z Z Z ½iL ir dt ½iL u ir dt ½iL u ir dt Proposed controllers This Section outlines the modelling method and describes the detailed procedures for deriving the proposed family of PWM-based SM controllers for DC–DC converters in DCM operation. 3.1 System modelling Figure 2 shows the schematic diagrams of the three PID sliding-mode voltage-controlled (SMVC) DC–DC converters to be discussed in this paper in the conventional HM configuration. Here, C, L and rL denote the capacitance, inductance and instantaneous load resistance of the converters, respectively, iC, iL and ir denote the instantaneous capacitor, inductor and load currents, respectively; Vref, vi and vo denote the reference, instantaneous input and instantaneous output voltages, respectively; b denotes the feedback network ratio; and u ¼ 0 or 1 is the switching state of power switch SW and u ¼ 1 u is the inverse logic of u. In the case of PID SMVC converters, the control variables x may be expressed in the form: 2 3 Vref bvo 2 3 x1 6 dðV bv Þ 7 6 7 ref o ð3Þ x ¼ 4 x2 5 ¼ 6 7 4 5 dt x3 R ðVref bvo Þdt where x1, x2 and x3 represents the voltage error, the voltage error dynamics (or the rate of change of voltage error), and the integral of voltage error, respectively. Substitution of the converters’ behavioural models under DCM (in Table 1) into (3) produces the following control variable descriptions: xbuck, xboost and xbuck–boost for buck, boost and buck–boost converters, respectively. 3 2 x1 ¼ Vref bvo Z 7 6 6 x ¼ bvo þ bðvo uL vi uÞ dt 7 7 6 2 ð4Þ xbuck ¼ 6 LC rL C 7 7 6 Z 5 4 x3 ¼ x1 dt 65 iL L SW ir iC Vi + D R1 Vo C rL R2 α 3x 3 S u α 3 ⋅ ∫ dt β Vo + α1x1 α1 + + α 2 x2 d α2⋅ Sliding Mode Controller − x1 + For the boost converter: 2 3 2 3 0 1 0 2 3 2 3 0 x x_ 1 1 6 7 6 7 1 76 7 6 7 6 7 6 u 4 x_ 2 5 ¼ 6 0 0 74 x2 5 þ 6 bvo uB bvi uB 7 4 5 4 rL C LC LC 5 x_ 3 x3 0 1 0 0 ð8Þ Vref dt a iL L D iD ir iC + SW Vi R1 Vo C rL R2 α 3x 3 βVo + α1x 1 α1 + + α 2x 2 d S u α 3 ⋅ ∫ dt α 2⋅ Sliding Mode Controller − x1 + For the buck converter: 2 3 2 3 2 3 2 3 0 0 0 1 0 2 3 x_ 1 6 7 x1 6 7 6 7 74 x2 5 þ 6 bvi 7u þ 6 bvo u 7 4 x_ 2 5 ¼ 6 0 1 0 L5 4 5 4 5 4 rL C LC LC x3 x_ 3 1 0 0 0 0 ð7Þ For the buck–boost converter: 2 3 2 3 0 1 0 2 3 2 3 0 x_ 1 6 7 x1 6 7 74 x2 5 þ 6 bvo u 7 4 x_ 2 5 ¼ 6 0 1 0 B 5u 4 5 4 rL C LC x3 x_ 3 0 1 0 0 Vref Rearrangement of the state-space descriptions (7), (8) and (9) into the standard form gives dt x_ ¼ Ax þ Bv þ D b D SW iD ir iL iC L Vi R2 C + Vo rL R1 α 3x 3 S u α 3 ⋅ ∫ dt + α1x1 α1 + + α 2x 2 d α2 ⋅ Sliding Mode Controller β Vo x1 − + Vref Schematic diagrams of PID SMVC converters xboost where S is the instantaneous state variable’s trajectory, and is described as 3 x1 ¼ Vref bvo Z 6 u 7 6 x ¼ bvo þ bðvo uB vi uB Þ dt 7 7 6 2 ¼6 LC rL C 7 7 6 Z 5 4 x3 ¼ x1 dt 3 x1 ¼ Vref bvo Z 7 6 bvo uB u 7 6 x2 ¼ bvo dt 7 þ 6 ¼6 7 LC rL C 7 6 Z 5 4 x3 ¼ x1 dt Controller derivation and design Having obtained the state-space descriptions, the next stage is the derivation and design of the controller. For these systems, it is appropriate to have a general SM control law that adopts a switching function such as 1 when S40 u¼ ð11Þ 0 when So0 a SMVC buck converter b SMVC boost–converter c SMVC buck–boost converter 2 ð10Þ where v ¼ u or u (depending on topology). Results are summarised in Table 2. An inspection of the equations verified the presence of the three state-space structures (trilinear structure) in each converter. In the case of the buck converter, one structure exists when u ¼ 1 and uL ¼ 1, another exists when u ¼ 0 and uL ¼ 1, and the third exists when u ¼ 0 and uL ¼ 0. For the boost and buck–boost converters, one structure exists when u ¼ 1 and uB ¼ 0, another exists when u ¼ 0 and uB ¼ 1, and the third exists when u ¼ 0 and uB ¼ 0. 3.2 dt c Fig. 2 ð9Þ S ¼ a1 x1 þ a2 x2 þ a3 x3 ¼ J T x ð5Þ 2 ð12Þ T with J ¼ [a1 a2 a3] and a1, a2, and a3 representing the control parameters, termed as sliding coefficients. Equations (11) and (12) describe the control equations needed for implementing the SM controller in conventional HM form. For implementation in the PWM form, the control equations must be translated. ð6Þ 3.2.1 Derivation of control equations for PWM-based controller: The implementation of Next, the time differentiation of (4), (5) and (6) produces the state-space descriptions required for the controller design of the respective converter. PWM-based controllers is possible by first translating the SM control law to obtain its equivalent control function. This function is then mapped onto the duty cycle function of the pulse-width modulator [10]. The equivalent control signal ueq can be formulated using the invariance conditions by setting the time differentiation of (12) as dS/dt ¼ 0 [2]. xbuckboost 66 IET Electr. Power Appl., Vol. 1, No. 1, January 2007 Table 2: Matrix description of SMVC buck, boost and buck–boost converters in DCM Type of converter Buck A B 2 2 0 1 6 60 1 4 rL C 1 0 2 Boost 0 1 6 60 1 4 rL C 1 0 2 Buck–boost 0 1 6 60 1 4 rL C 1 0 0 3 0 v 3 0 6 7 6 bvo 7 6 7 4 LC uL 5 0 2 3 0 405 0 0 2 3 0 6 bvo 7 bvi 6 7 4 LC uB LC uB 5 3 7 07 5 u u 0 0 0 2 0 6 7 6 bvi 7 6 7 4 LC 5 7 07 5 0 D 3 2 3 0 2 3 0 405 0 3 6 bvo 7 6 7 4 LC uB 5 7 07 5 u 0 0 (i) Buck converter and ramp signal ^vramp : T T T Equating dS/dt ¼ J Ax+J Bueq+J D ¼ 0 equivalent control function yields a1 1 a3 vc ¼ ueq ¼ bL iC þ a2 r L C a2 the ueq ¼ ½J T B1 J T ½Ax þ D bL a1 1 a3 LC vo ðVref bvo Þ þ uLeq ¼ iC þ bvi a2 rL C a2 bvi vi LCðVref bvo Þ þ bvo uLeq ð18Þ and ^vramp ¼ bvi ð19Þ ð13Þ where both ueq and uLeq are continuous and bounded by 0 and 1. Specifically, the equivalent control function ueq is a smooth function of the discrete input function u, uLeq is the result equivalent component of the discrete virtual switching component uL. We first derive the equivalent control function by substituting (13) into 0 o ueq o 1 and multiplying by bvi to give a1 1 0oueq ¼ bL iC a2 r L C a3 ð14Þ þ LC ðVref bvo Þ þ bvo uLeq obvi a2 Next, the function uLeq can be approximated from wellestablished mathematical models of the buck converter. First, for the buck converter in DCM, the duty cycle is described as sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi v2o 2L ð15Þ DDCM ¼ vi ðvi vo Þ rL T Additionally, the duty cycle for the turn-off period in DCM operation is rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vi vo vi vo 2L D2 ¼ DDCM ¼ ð16Þ vo vi r L T Since the maximum time duration that uL ¼ 1 can exist is T (during CCM operation), and in DCM operation uL ¼ 1 is applied for the time duration (DDCM+D2)T, the equivalent virtual switching component applied to the system in DCM operation can be described as sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðDDCM þ D2 ÞT vi 2L ¼ ð17Þ uLeq ¼ T vi vo r L T Finally, the mapping of the equivalent control function (14) onto the duty ratio control D, where 0 o D ¼ vc =^vramp o 1, gives the following relationships for the control signal vc IET Electr. Power Appl., Vol. 1, No. 1, January 2007 (ii) Boost converter Equating dS/dt ¼ JTAx+JTB ueq ¼ 0 yields the equivalent control function ueq ¼ ½J T B1 J T Ax bL a1 1 ¼ iC bðvo vi ÞuBeq a2 rL C a3 LC ðVref bvo Þ a2 bðvo vi ÞuBeq ð20Þ where both ueq and uBeq are continuous and bounded by 0 and 1. Here, the equivalent control function ueq is a smooth function of the discrete input function u, and uBeq is the resulted equivalent component of the discrete virtual switching component uB. Using the same approach, the equivalent control function can be derived as a1 1 a3 0oueq ¼ bL iC þ LC ðVref bvo Þ a2 r L C a2 þ bðvo vi ÞuBeq oðbðvo vi ÞuBeq ð21Þ Next, the function uBeq can be approximated from wellestablished mathematical models of the boost converter. First, for the boost converter in CCM, the duty cycle is described as vi ð22Þ DCCM ¼ 1 vo and in DCM, the duty cycle is described as sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vo vo 2L ð23Þ 1 DDCM ¼ rL T vi vi Additionally, the duty cycle for the turn-off period in DCM operation is given as sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vi vo 2L D2 ¼ ð24Þ DDCM ¼ vo vi vo vi rL T 67 Since the maximum time duration that uB ¼ 1 can exist is D2maxT ¼ (1 DCCM)T (during CCM operation), and in DCM operation, uB ¼ 1 is applied for the time duration D2 T, the equivalent virtual switching component applied to the system in DCM operation can be described as sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi D2 T vo vo 2L ¼ uBeq ¼ ð25Þ D2 max T vi vo vi rL T Since the maximum time duration that uB ¼ 1 can exist is D2maxT ¼ (1 DCCM)T (during CCM operation), and in DCM operation, uB ¼ 1 is applied for the time duration D2 T, the equivalent virtual switching component applied to the system in DCM operation can be described as rffiffiffiffiffiffiffiffi D2 T vo 2L ¼ 1þ uBeq ¼ ð33Þ D2 max T rL T vi Finally, the mapping of the equivalent control function (21) onto the duty ratio control D, where 0oD ¼ ðvc =^vramp Þo1, gives the following relationships for the control signal vc and ramp signal ^vramp : a1 1 vc ¼ ueq ¼ bL iC a2 r L C a3 þ LC ðVref bvo Þ þ bðvo vi ÞuBeq ð26Þ a2 and ^vramp ¼ bðvo vi ÞuBeq ð27Þ Finally, the mapping of the equivalent control function (29) onto the duty ratio control D, where 0oD ¼ ðvL =^vramp Þo1, gives the following relationships for the control signal vc and ramp signal ^vramp : a1 1 vc ¼ ueq ¼ bL iC a2 r L C a3 þ LC ðVref bvo Þ þ bvo uBeq ð34Þ a2 (iii) Buck–boost converter Equating dS/dt ¼ JTAx+JTB ueq ¼ 0 yields the equivalent control function ueq ¼ ½J T B1 J T Ax bL a1 1 a3 LC ðVref bvo Þ ¼ iC bvo uBeq a2 rL C a2 bvo uBeq ð28Þ where both ueq and uBeq are continuous and bounded by 0 and 1. The equivalent control function is derived as a1 1 0oueq ¼ bL iC a2 r L C a3 þ LC ðVref bvo Þ þ bvo uBeq obvo uBeq ð29Þ a2 Similarly, the function uBeq can be approximated from wellestablished mathematical models of the buck–boost converter. First, for the buck–boost converter in CCM, the duty cycle is described as vo ð30Þ DCCM ¼ vo þ vi and in DCM, the duty cycle is described as rffiffiffiffiffiffiffiffi vo 2L DDCM ¼ ð31Þ vi rL T Additionally, the duty cycle for the turn-off period in DCM operation is given as rffiffiffiffiffiffiffiffi vi 2L ð32Þ D2 ¼ DDCM ¼ rL T vo and ^vramp ¼ bvo uBeq ð35Þ The simplified control equations required for the implementation of the respective PWM-based SMVC converter operating in DCM are given in Table 3. The representation takes into consideration that rL ¼ vo/ir. Figure 3 shows the schematic diagrams of the respectivePWM-based SMVC converters for DCM operation. The controllers are modelled from the equations illustrated in Table 3. It is interesting to note that these controllers are highly nonlinear, requiring various square root, division, multiplier, addition and subtraction operators in their computations. This reflects the true nature of the DCM converter system, showing the large amount of nonlinearity actually required in processing the information needed for the robust control of such converters. It also revealed how overly simplified the existing PWM controllers apparently are, in their applications in DCM converters. This explains the inevitable poor performances resulting from DCM converters using such controllers. In essence, the proposed controllers accurately represent the sort of nonlinear controllers truly required for providing good control to DCM converters operating over a wide variation of input/load range. Despite its architectural complexity and requirement for more state-variable sensing than in conventional PWM controllers, the application of this special family of PWM-based SM voltage controllers to DCM converters requiring tight regulation is well justified in circumstance whereby conventional PWM controllers fail for the purpose. Finally, it should also be pointed out that the implementation of the proposed controllers in Fig. 3 does not automatically qualify them as SM controllers. They need to comply with the three necessary conditions of SM operation: the hitting condition, existence condition, and stability condition, to make them SM controllers. The hitting condition has been satisfied by the control law in Table 3: Control equation of PWM-based SMVC buck, boost and buck–boost converters operating in DCM Type of converter Buck Boost Buck–boost vc qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vi vo ir vi vo vo pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Kp1 iC þ Kp2 ðVref bvo Þ þ Kp3 ir ðvo vi Þ vi ! vo pffiffiffiffiffiffiffiffi vo ir Kp1 iC þ Kp2 ðVref bvo Þ þ Kp3 1 þ vi Kp1 iC þ Kp2 ðVref bvo Þ þ Kp3 v^ramp bvi vo pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ir ðvo vi Þ vi vo pffiffiffiffi v o ir Kp3 1 þ vi Kp3 pffiffiffiffiffiffiffiffiffiffiffi *Kp1, Kp2, and Kp3 are fixed parameters calculated using Kp1 ¼ bLðða1 =a2 Þ ð1=rLðminÞ CÞÞ; Kp2 ¼ ða3 =a2 ÞLC; and Kp3 ¼ b 2L=T 68 IET Electr. Power Appl., Vol. 1, No. 1, January 2007 (11) [11]. This ensures that regardless the initial condition of the converter, the controller will always make control decisions that bring the state trajectory to a point within the local vicinity of the sliding surface. Once the trajectory comes within such vicinity, the pre-satisfaction of the existence condition will take over to ensure that the state trajectory is maintained within this vicinity while being concurrently directed towards the equilibrium. Clearly, if the controller design satisfies the stability condition, which checks if the equilibrium point on the sliding surface is stable, a stable steady-state regulation will be expected from the converter. The following Section discusses the derivation vi ir iL L SW vo ir iC vi + D vo C rL iC 1 − K p1iC − K p1 X β vo iC 0 u Kp2 x1 + K p2 + + vC + PWM − vramp β vi − x1 + Vref Sliding Mode Controller 4 Existence and stability conditions 4.1 Complying to existence conditions The method of ensuring the existence of SM operation for the DCM converter system is basically similar to the case of the CCM converter system, despite the structural difference in the composition of their state variable trajectory S. Figure 4 show the difference between the state variable trajectory behaviour in CCM and DCM operations. Typically, for SM controlled CCM converters, all the structures of the state variable trajectory must comply with the local reachability condition limS-0 S (dS/dt) o 0 under their respective configurations to ensure that the trajectory is moving within a small vicinity from the sliding plane towards the equilibrium, i.e. complying with the existence condition [2]. However, for DCM converters, which have a trilinear structure trajectory, the abidance of the local reachability condition by the first two structures of the trajectory provides a sufficient condition for the compliance of the existence condition. It is unnecessary for the third structure also to meet the local reachability condition. This is graphically illustrated in Figs. 5a and b. Here, it is shown that as long as structure 1 and structure 2 meet the local vo vi ir vi voir vi − vo K p3 of the existence and the stability conditions necessary for completing the requirement. x3 a vi ir iL L iD D vo sliding plane ir iC + SW vi converging the origin vo C rL x1 trajectory within small vicinity of sliding plane iC 1 − K p1iC − K p1 X Kp2x1 + Kp2 + + vC − PWM + u vramp v K p3 o vi Sliding Mode Controller βvo iC 0 − x1 + bilinear structure trajectory Vref a vo vi ir ir (vo − vi ) x3 b vi sliding plane ir D SW iD ir iL vi C + 1 − K p1iC 0 − + u vC PWM − Kp1 K p2x1 + K p2 + + vramp K p3 Sliding Mode Controller vo vo iC X iC x1 v 1+ o vi rL β vo − + Vref voir vo vi ir trilinear structure trajectory O x2 b Fig. 3 Schematic diagrams of PWM-based PID SMVC converters for DCM operation IET Electr. Power Appl., Vol. 1, No. 1, January 2007 x1 diverging structure c a SMVC buck converter b SMVC boost converter c SMVC buck–boost-converter converging the origin trajectory within small vicinity of sliding plane iC L x2 O Fig. 4 Graphical representations of state variable trajectory behaviour in SM control process a Trajectory behaviour in CCM operationFillustrating how bilinear structure trajectory is maintained within a small vicinity from the sliding plane b Trajectory behaviour in DCM operationFillustrating how trilinear structure trajectory is maintained within a small vicinity from the sliding plane, even though one of its structures is not directing the trajectory towards the sliding plane 69 Structure 3 (i L = 0, converging) Structure 1 (u = 1) O sliding plane Structure 2 (u = 0) converging the origin a Structure 3 (i L = 0, diverging) Structure 1 (u = 1) O sliding plane Structure 2 (u = 0) converging the origin b Fig. 5 Graphical representations of the two possible scenarios of state variables’ trajectory’s behaviour during DCM operation a Structure 3 converges to sliding plane b Structure 3 diverges from sliding plane reachability condition, the direction in which structure 3 of the state trajectory travels will not affect the overall existence of the SM operation, whereby the trajectory is maintained within the vicinity of the sliding plane and is moving towards the equilibrium. Intuitively, the local reachability conditions of structure 1 and structure 2 in the DCM converter systems are identical to the local reachability conditions derived for both the bilinear structures of the CCM converter systems [10], which in the proposed system can be expressed as 9 dS > > ¼ J T Ax þ J T BvS!0þ þ J T Do0 > = dt S!0þ ð36Þ > > dS T T T > ¼ J Ax þ J BvS!0 þ J D40 ; dt S!0 Hence, the overall existence conditions of the SMVC buck, boost and buck–boost converters in DCM operations can be recalled from the previous case of CCM operations [10]. The results are given in Table 4. For the detail of the mathematical derivation, readers are referred to [10]. The selection of sliding coefficients for the controller of each converter must comply with its stated inequalities. We have taken into account the complete ranges of operating conditions (minimum and maximum input voltages, i.e. vi(min) and vi(max), and minimum and maximum load resistances, i.e. rL(min) and rL(max). This ensures the compliance of the existence condition for the full operating ranges of the converters. Additionally, for SM controller with a static sliding surface, a practical approach is to design the sliding coefficients to meet Table 4: Existence condition of buck, boost and buck–boost converters operating in DCM Buck Boost Buck–boost 70 0o Kp1 iCðSSÞ þ Kp2 Vref bvoðSSÞ þbvoðSSÞ obviðminÞ 0oKp1 iCðSSÞ Kp2 Vref bvoðSSÞ ob voðSSÞ viðmaxÞ 0oKp1 iCðSSÞ Kp2 Vref bvoðSSÞ obvoðSSÞ the existence conditions for steady-state operation [10, 12]. Under such consideration, the state variables iC and vo can be substituted with their expected steady-state parameters, i.e. iC(SS) and vo(SS), which can be derived from the design specification. This ensures the compliance of the existence condition at least in the small region of the origin. Remarks: Note that by inspecting the reachability condition of structure 3 for the respective converters, it was found that the SMVC buck converter can be designed to have both converging and diverging structure 3, as shown in Figs. 5a and b, respectively. However, structure 3 of the SMVC boost and buck–boost converters can only be of diverging type, as shown in Fig. 5b. This is because structure 3 in both these converters are of opposing directions to structure 1. Hence, the necessary compliance of the reachability condition for structure 1 leads to the breaching of the reachability condition for structure 3. Yet, as emphasised, the direction in which structure 3 travels will not affect the overall existence of the SM operation. However, having one structure of its trajectory moving in an opposing direction to the origin will significantly prolong the overall settling time of the converters as compared to the CCM converters, which have both the structures of the trajectory moving towards the origin. This is evidenced in the simulation results illustrated in the later part of the paper. 4.2 Complying to stability conditions The method of assuring the stability of the SMVC DCM converters is the same as in the case of CCM converters. For the proposed family of controllers, the same method of selecting the sliding coefficients based on the desired dynamic properties to automatically arrive at a stable equilibrium as in [9, 10, 13], is adopted. This is a direct approach to assuring stability, whereby the same objective of making the eigenvalues of the Jacobian matrix of the system in SM operation contain negative real parts is achieved. In our proposed system, the stability condition is easily satisfied by selecting the sliding coefficients for a specific type of dynamic response using one of the following equations: (i) Under-damped response: a1 10 ¼ a2 Ts and a3 25 ¼ 2 a2 z Ts2 ð37Þ where Ts denotes the settling time and the damping ratio is vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u 2 u Mp u ln u 100 ð38Þ z¼u 2 o1 u t 2 Mp p þ ln 100 where Mp is the percentage of the peak overshoot. (ii) Critically-damped response: a1 10 ¼ a2 T s and a3 25 ¼ a2 Ts2 ð39Þ (iii) Over-damped response: a1 ¼ a2 10 sffiffiffiffiffiffiffiffiffiffiffiffiffi! 1 1 1 2 Ts z and a3 25 pffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ a2 ðz z2 1ÞT 2 s ð40Þ where z 4 1. IET Electr. Power Appl., Vol. 1, No. 1, January 2007 The mathematical derivations for (38), (39), and (40) can be found in [9, 10]. Thus, the design of the sliding coefficients is now dependent on the desired settling time and the type and amount of damping required in the response of the ideal sliding operation, in conjunction with the existence condition of the respective PWM-based controllers. 5 Simulation results and discussions The derived PWM-based controller’s equations in Table 3 have been verified through computer simulation. The simulation was performed using Matlab/Simulink. The step size taken for all simulations is 10 ns. 5.1 simulation are vc ¼ 1:15iC þ 52:1103ðVref bvo Þ rffiffiffiffiffiffiffiffiffiffiffiffiffiffi vi vo ir þ 0:6179 and ^vramp ¼ bvi vi vo ð41Þ Figure 6 shows the steady-state behaviour of the SMVC buck converter. Figure 7a shows a plot of the DC output voltage against the different operating load resistances. The results show good load regulation property for the load range Buck converter Table 5 shows the specification of the buck converter used in the simulation. The PWM-based SM voltage controller is designed to give a critically-damped response with a bandwidth of fBW ¼ 20 kHz (i.e. first-order response time constant t20 kHz ¼ 7.956 ms settling time Ts(20 kHz) ¼ 39.780 ms). Assuming that the controller is designed for maximum load current (i.e. minimum load resistance rL(min), the reference voltage is chosen as Vref ¼ 2.5 V, and the ratio of the voltage divider network b ¼ 0.208; the sliding coefficients are calculated as a1/a2 ¼ 251327 and a3/a2 ¼ 15791367040. Finally, the control parameters are determined as Kp1 ¼ bL ((a1/a2) (1/r pL(min) ffiffiffiffiffiffiffiffiffiffiffiC)) ¼ 1.15, Kp2 ¼ (a3/a2) LC ¼ 52.1103, and Kp3 ¼ b 2L=T ¼ 0.6179. Therefore, the control equations implemented in the Table 5: Specification of buck converter Description Parameter Nominal value Input voltage vi 24 V Capacitance C 150 mF Capacitor ESR Cr 21 mO Inductance L 22 mH Inductor resistance lr 0.12 O Switching frequency fS 200 kHz Minimum load resistance rL(min) 17.5 O Maximum load resistance rL(max) 120 O Desired output voltage Vod 12 V Fig. 7 Steady-state and dynamic performances of SMVC buck converter operating in DCM Fig. 6 Waveforms of control signal vc, input ramp vramp, generated gate pulse u (left), and inductor current iL and output voltage ripple vo (right) for PWM-based SMVC buck converter operating in DCM at vi ¼ 24 V and rL ¼ 30 O IET Electr. Power Appl., Vol. 1, No. 1, January 2007 71 5.2 Assuming that the controller is designed for maximum load current (i.e. minimum load resistance rL(min), the reference voltage is chosen as Vref ¼ 8 V, and the ratio of the voltage divider network b ¼ 0.1667, the sliding coefficients are calculated as a1/a2 ¼ 25132.7 and a3/a2 ¼ 157913670. Finally, the control parameters are determined as 48.5 48.4 Discontinous Conduction Mode (DCM) 48.3 Output Voltage Vo (V) 17.5 O r rL r 120 O with only a 0.036 V deviation in vo (i.e. about 0.3% of vo(17.5 O)). Figure 7b shows the output voltage ripple waveforms of the system at dynamic load condition. The step change responses are critically damped, which conform to the designed behaviour. The settling times of the system are 400 ms (load resistance changes from 30 to 120 O) and 260 ms (load resistance changes from 120 to 30 O). These settling times are much longer than the ideal SM settling time of around Ts(20 kHz) ¼ 40 ms. There are two reasons for this. First, the presence of excessive excursion of the state trajectory within the vicinity of the sliding surface (created by the finiteness of the switching frequency) lengthens the actual settling time of the system as compared to the ideal case where the state trajectory travels exactly on the sliding surface. Secondly, as explained earlier, structure 3 of the trajectory in DCM converters, which moves on an opposing direction to the origin, will significantly prolong the overall settling time of the converters. Yet, it must be emphasised that even though (38), (39) and (40) do not accurately reflect the true nature of the overall response time of the physical converter, they do provide the general design guidelines for a stable system with the desired type (damping) of secondorder dynamical response in a most simplified form. Critical resistance at RL = 430 Ω 48.2 48.1 48.0 47.9 47.8 400 450 500 550 600 650 700 750 800 850 Load Resistance RL (Ω) a Boost converter Table 6 shows the specification of the boost converter used in the simulation. The PWM-based SM voltage controller is designed to give a critically-damped response with a bandwidth of fBW ¼ 2 kHz (i.e. first-order response time constant t2 kHz ¼ 79.56 ms settling time Ts(2 kHz) ¼ 397.8 ms). 48.45 48.40 48.35 Output Voltage Vo (V) Table 6: Specification of boost converter 48.50 48.30 48.25 48.20 Description Parameter Nominal value Input voltage vi 12 V Capacitance C 1000 mF Capacitor ESR Cr 36 mO 48.05 Inductance L 50 mH 48.00 Inductor resistance lr 0.14 O Switching frequency fS 200 kHz Minimum load resistance rL(min) 430 O Maximum load resistance rL(max) 800 O Desired output voltage Vod 48 V 48.15 48.10 83 ms 48 ms 0 10 20 30 40 50 60 70 80 90 100 110120130140150160170180190200210 Time (ms) b Fig. 9 Steady-state and dynamic performances of the SMVC boost converter operating in DCM a Output voltage vo against load resistance rL b vo at step load changes alternating between 480 and 800 O Fig. 8 Waveforms of control signal vc, input ramp vramp, generated gate pulse u (left) and inductor current iL and output voltage ripple vo (right) for the PWM-based SMVC boost converter operating in DCM at vi ¼ 12 V and rL ¼ 480 O 72 IET Electr. Power Appl., Vol. 1, No. 1, January 2007 Kp1 ¼ 0.209, Kp2 ¼ 7.896 and Kp3 ¼ 0.745. Therefore, the control equations implemented in the simulation are vo pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i r ð vo vi Þ vc ¼ 0:209iC þ 7:896ðVref bvo Þ þ 0:745 vi and ^vramp ¼ 0:745 vo pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi i r ð vo vi Þ vi ð42Þ conversion. The PWM-based SM voltage controller is designed to give a critically-damped response with a bandwidth of fBW ¼ 2.5 kHz (i.e. first-order response time constant t2.5 kHz ¼ 63.62 ms, settling time Ts(2.5 kHz) ¼ 318.3 ms). Assuming that the controller is designed for maximum load current (i.e. minimum load resistance rL(min)) the reference voltage is chosen as Vref ¼ 6 V, and the ratio of the voltage divider network b ¼ 0.1667; the sliding coefficients are calculated as a1/a2 ¼ 31416 and Figure 8 shows the steady-state behaviour of the SMVC boost converter. It should be noted that the peak magnitude of the ramp signal will change adaptively with the changes in load resistance. This is due to the equivalent virtual component term in the ^vramp computation (see (27)). Figure 9a shows a plot of the DC output voltage against the different operating load resistances. The results show how good the load regulation property is for the load range 430 O r rL r 800 O with only a 0.376 V deviation in vo (i.e. about 0.78% of vo(430 O)). Figure 9b show the output voltage ripple waveforms of the system at dynamic load condition. The step change responses are critically damped, which conform to the designed behaviour. The settling times of the system are 83 ms (load resistance changes from 480 to 800 O) and 48 ms (load resistance changes from 800 to 480 O). Again, these settling times are much longer than the ideal SM settling time of around Ts(2 kHz) ¼ 0.4 ms. 5.3 Buck–boost converter Table 7 shows the specification of the buck–boost converter used in the simulation. The converter operates for step-up Table 7: Specification of buck–boost converter Description Parameter Nominal value Input voltage vi 24 V Capacitance C 1000 mF Capacitor ESR Cr 36 mO Inductance L 25 mH Inductor resistance lr 0.12 O Switching frequency fS 200 kHz Minimum load resistance rL(min) 150 O Maximum load resistance rL(max) 700 O Desired output voltage Vod 36 V Fig. 11 Steady-state and dynamic performances of the SMVC buck–boost converter operating in DCM a Output voltage vo against load resistance rL b vo at step load changes alternating between 150 and 600 O Fig. 10 Waveforms of control signal vc, input ramp vramp, generated gate pulse u (left), and inductor current iL and output voltage ripple vo (right) for PWM-based SMVC buck–boost converter operating in DCM at vi ¼ 24 V and rL ¼ 180 O IET Electr. Power Appl., Vol. 1, No. 1, January 2007 73 a3/a2 ¼ 246740000. Finally, the control parameters are determined as Kp1 ¼ 0.131, Kp2 ¼ 6.169 and Kp3 ¼ 0.527. Therefore, the control equations implemented in the simulation are 7 vc ¼ 0:131iC þ 6:169 ðVref bvo Þ vo pffiffiffiffiffiffiffiffi þ 0:527 1 þ vo i r vi and ^vramp vo pffiffiffiffiffiffiffiffi ¼ 0:527 1 þ vo i r vi ð43Þ Figure 10 shows the steady-state behaviour of the SMVC buck–boost converter. Similar to that of boost converter, the peak magnitude of the ramp signal also changes with load resistance (35). Figure 11a shows DC output voltage against different operating load resistances. The results show good load regulation property for the load range 63 O r rL r 700 O with only a 0.615 V deviation in vo (i.e. about 1.6% of vo(63 O)). Figure 11b shows the output voltage ripple waveforms of the system at dynamic load condition. The settling times of the system are 13.5 ms (load resistance changes from 150 to 600 O) and 7 ms (load resistance changes from 600 to 150 O). Again, these settling times are much longer than the ideal SM settling time of around Ts(2.5 kHz) ¼ 0.32 ms. 6 Conclusions This paper has proposed a family of fixed-frequency PWMbased SM voltage controllers for basic DC–DC converters operating in discontinuous conduction mode. A method of developing the state-space models of these converters in DCM operation required for the controller design is described. Simple control equations for implementation of the PWM based SM voltage controllers for the different converters are derived and verified through computer 74 simulations. The results show that the proposed controllers are capable of offering good control performance for the respective converters. References ! 1 Venkataramanan, R., Sabanovic, A., and Cuk, S.: ‘Sliding mode control of DC-to-DC converters’. Proc. IEEE Conf. on Industrial Electronics, Control and Instrumentations (IECON), 1985, pp. 251– 258 2 Utkin, V., Guldner, J., and Shi, J.X.: ‘Sliding mode control in electromechanical systems’ (Taylor and Francis, London, UK, 1999) 3 Mattavelli, P., Rossetto, L., Spiazzi, G., and Tenti, P.: ‘Generalpurpose sliding-mode controller for DC/DC converter applications’. IEEE Power Electronics Specialists Conf. Record (PESC), June 1993, pp. 609–615 4 Alarcon, E., Romero, A., Poveda, A., Porta, S., and MartinezSalamero, L.: ‘Current-mode analogue integrated circuit for slidingmode control of switching power converters’, Electron. Lett., 2002, 38, (3), pp. 104–106 5 Nguyen, V.M., and Lee, C.Q.: ‘Indirect implementations of slidingmode control law in buck-type converters’. Proc. IEEE Applied Power Electronics Conf. and Exp. (APEC), March 1996, Vol. 1, pp. 111–115 6 Mahdavi, J., Emadi, A., and Toliyat, H.A.: ‘Application of state space averaging method to sliding mode control of PWM DC/DC converters’. Proc. IEEE Conf. on Industry Applications (IAS), October 1997, Vol. 2, pp. 820–827 7 Wu, C.C., and Young, C.M.: ‘A new PWM control strategy for the buck converter’. Proc. IEEE Conf. on Industrial Electronics Society (IECON), 1999, pp. 157–162 8 Mazumder, S.K., and Kamisetty, S.L.: ‘Experimental validation of a novel multiphase nonlinear VRM controller’. IEEE Power Electronics Specialists Conf. Record (PESC), June 2004, pp. 2114–2120 9 Tan, S.C., Lai, Y.M., Tse, C.K., and Cheung, M.K.H.: ‘A fixedfrequency pulse-width-modulation based quasi-sliding mode controller for buck converters’, IEEE Trans. Power Electron., 2005, 20, (6), pp. 1379–1392 10 Tan, S.C., Lai, Y.M., Tse, C.K.: ‘A unified approach to the design of PWM based sliding mode voltage controller for basic DC–DC converters in continuous conduction mode’, IEEE Trans. Circuits Syst. I: Fundam. Theory Appl. 2006, 53,(8), pp 1816–1827 11 Tan, S.C., Lai, Y.M., Tse, C.K.: ‘General design issues of sliding mode controllers in DC–DC converters’, IEEE Trans. Ind. Electron. (to be published 2007) 12 Martinez-Salamero, L., Calvente, J., Giral, R., Poveda, A., and ! Fossas, E.: ‘Analysis of a bidirectional coupled-inductor Cuk converter operating in sliding mode’, IEEE Trans. Circuits Syst. I: Fundam. Theory Appl., 1998, 45, (4), pp. 355–363 13 Ackermann, J., and Utkin, V.: ‘Sliding mode control design based on Ackermanns formula’, IEEE Trans. Autom. Control, 1998, 43, (2), pp. 234–237 IET Electr. Power Appl., Vol. 1, No. 1, January 2007