Research Journal of Applied Sciences, Engineering and Technology 4(4): 342-349, 2012 ISSN: 2040-7467 © Maxwell Scientific Organization, 2012 Submitted: August 26, 2011 Accepted: October 07, 2011 Published: February 15, 2012 Design and Robustness Analysis of a PID Based Sliding Mode Controller for a dc-dc Converter 1 D.M. Mary Synthia Regis Prabha, 2S. Pushpa Kumar and 1G. Glan Devadhas 1 PRIST University, Thanjavur 2 Heera College of Engineering and Technology, Nedumancadu, Trivandrum Abstract: This study deals with the design and analysis of a dc-dc converter operating in continuous conduction mode with Proportional-Integral-Derivative control and PID based Sliding Mode Control (SMC). A small signal model is developed using Switching Flow Graph (SFG) from which the control coefficients for the PID controller is selected. PID based SMC uses a control law which constrains the weighted sum of the voltage error, its derivatives and the integral of the voltage error to zero. The equivalent control technique is used in its design which makes the converter more suitable for fixed frequency operation. Sensitivity of these controllers to supply voltage disturbances and load disturbances is studied and results are presented. Key words: dc-dc converter, PID based SMC, PID control, switching flow graph Modulation (PWM) for the controller. In this paper this technique has been demonstrated for a Buck Converter. PID control and PID based SMC are two different control techniques considered in this paper. PID control is a traditional linear control, while SMC is a type of nonlinear control. Linear PID controllers for dc-dc converters are usually designed by classical frequency response techniques applied to the small-signal models of converters (Liping et al., 2009). A bode plot is adjusted in the design to obtain the desired loop gain, cross-over frequency and phase margin. The transient response can be tuned using root locus type approaches (Prodic and Maksimovic, 2002). The stability of the system is guaranteed by an adequate phase margin. PID control is typically designed for one nominal operating point (Perry et al., 2007). For a buck converter, the magnitude of the frequency response depends on the duty cycle. Duty cycle variations do not change the shape of the magnitude plot of the transfer function, but only shifts the plot upward (Siew et al., 2008). Therefore, a PID controller may not respond well to significant changes in operating points. In this study, the problem of PID based quasi sliding mode control of a Buck Converter is discussed. The design of a PID controller for a dc-dc converter is discussed in detail. Finally, comparison of these two methods is performed in order to verify the dynamic and steady state responses and their robustness to sudden line variations. INTRODUCTION Switching mode dc-dc converters are widely used today in a variety of applications including power supplies for personal computers, mission critical space applications, laptop computers, dc motor drives, medical electronics as well as high power transmission. These converters are non-linear dynamical systems. The nonlinearities arise primarily due to switching, power devicesand passive components such as inductors and capacitors. A control technique suitable for dc-dc converters must cope with their intrinsic non-linearity and wide input voltage and load variations ensuring stability in any operating condition while providing fast transient response. Among the various control techniques available, sliding mode control offers several advantages, namely large signal stability, robustness, good dynamic response and simple implementation. The design of a SMC does not require an accurate model of the system. The ideal nature of the controller is to operate at an infinite switching. This nature enables the controlled variable to track a certain reference path to achieve the desired dynamic response and better steady state operation (Duan and Jin, 1999). This extreme high frequency operation results in excessive switch losses and Electromagnetic Interference (EMI) noise issues. In addition sliding mode control also exhibits steady state error for the output voltage. Hence, for sliding mode controllers to be applicable for dc-dc power converters, it becomes essential to constrict the switching frequency within a practical range. Moreover, variable switching frequency is undesirable for power converters. A method of ensuring constant switching frequency is to employ Pulse Width SYSTEM MODELING Linear controllers for dc-dc converters are often designed based on mathematical models. The most commonly used technique for modeling linear Corresponding Author: D.M. Mary, Synthia Regis Prabha, Associate Professor, Noor ul Islam University, Kumaracoil 342 Res. J. Appl. Sci. Eng. Technol., 4(4): 342-349, 2012 Fig. 1: Buck converter Fig. 2: On-state switching flow graph Fig. 3: Off-State switching flow graph Fig. 4: SFG of buck converter controllers is the traditional state space averaging method. The major drawback of this method is that the linearized models obtained through averaging process, do not predict the large-signal stability information, and are only sufficient to predict small-signal stability. A small signal model takes a circuit and based on an operating point (bias) and linearizes all the components. Nothing changes because the assumption is that the signal is so small that the operating point (gain, capacitance etc.) doesn't change. A large signal model on the other hand 343 Res. J. Appl. Sci. Eng. Technol., 4(4): 342-349, 2012 Table 1: Prototype Buck Converter circuit parameters Parameter Value 24 Input voltage, Vi Desired output voltage, Vod 12 Load resistance, Ro 6 Filter inductance, L 35 Filter capacitance, C 150 ESR of inductor, RL 0.12 ESR of capacitor, RC 0.03 takes into account the fact that the large signal actually affects the operating point and takes into account that elements are non-linear and that circuits can be limited by power supply values. A small signal model ignores supply values. Graphical Analysis Method (GAM): This is a non-linear modeling method. This is developed for pulse-width modulated converters. The GAM converts the switching converter, even the multistate converter systems, into a unified dynamic model. From unified model, it is possible to derive large, small signal and steady state models with mathematical manipulations. Each system can be represented by a flow graph. The switching flow graph is obtained by combining the flow graphs of the subsystems through the use of switching branches (Mummadi, 2004). The switching flow graph model is easy to derive, and it provides a visual representation of a switching converter system. Buck converter shown in Fig. 1 has two modes of operation: (i) Switch is on (ii) Switch is off. The circuit diagram of the Buck converter is shown below. The switching flow graph is drawn for the on-state sub-circuit and the off-state sub-circuit which is shown in Fig. 2 and 3, respectively. These two sub-circuits are combined together which gives the switching flow graph of the Buck converter which is shown in Fig. 4. the filter capacitance C (Perry et al., 2007). The cut-off frequency of the second order low pass filter is TC =1/%LC Variations of D, varies the magnitude of the transfer function. Moreover, it does not change the shape of the frequency response, but shifts the magnitude plot upward or downward. Table 1 gives the circuit parameters of the prototype Buck Converter selected for experimentation. The nominal operating point of the prototype Buck converter is chosen as follows: Input Voltage, Vi = 24 V, Desired Output Voltage, Vod = 12 V, Duty ratio D = 0.5. Small signal control to output transfer function at this nominal operating point is given as: ⎤ V$0 ( s) ⎡ 24 + 10.8e − 4 s =⎢ −4 −9 2 ⎥ $ d ( s) ⎢⎣ 1 + 5817 . e s + 4.243 e s ⎥⎦ ⎤ V$0 ( s) ⎡ 1 + 45.39 e − 4 s = ⎢ 6 3 2⎥ $ Vin ( s) ⎢⎣ 235.65e + 137 e s + s ⎥⎦ V$0 ( s) V0 = D d$( s) (4) (1) Design of PID controller for a buck converter: The open loop transfer function G(S) H(s) of the Buck converter is given by Eq. (3). A PID compensator is designed with a phase margin of 45º at a cross-over frequency of 125.66 kHz. The steady state error for unit ramp input is considered to be 0.035%. The PID controller has a transfer function of GC(S) = KP+ Kis +Kds. Ki is decided by steady state requirements. Once we know KI , we can find out KP and Kd, Eq. (5) and (6). Small Signal Input to Output transfer function is derived as: V$0 ( s) DR0 = V$in ( s) ( R0 + RL ) ⎡ ⎤ ⎢ ⎥ 1 + sRC C ⎢ ⎥ ⎢ ⎛ ⎛ R0 + RC ⎞ ⎥ RL R0 L ⎞ 2 ⎢ 1 + s⎜ RC C + C+ ⎟ + S LC⎜ ⎟⎥ RL + R0 RL + R0 ⎠ ⎢⎣ ⎝ ⎝ R0 + RL ⎠ ⎥⎦ (3) Small Signal Input to Output transfer function at this nominal operating point is given as: Derivation of small-signal model using switching flow graph: Small signal control to output transfer function is derived as: ⎤ ⎡ ⎥ ⎢ 1 + sRC C ⎥ ⎢ ⎢ ⎛ ⎞ ⎛ RO + RC ⎞ ⎥ RL RO L 2 ⎢ 1 + s⎜ RC C + C+ ⎟ + s LC⎜ ⎟⎥ RL + RO RL + RO ⎠ ⎢⎣ ⎝ ⎝ RO + RL ⎠ ⎥⎦ Units V V S :H :F S S (2) In this transfer function, VO is the output voltage, D is the duty cycle, C is the output capacitance, L is the inductance and R is the load resistance. RL and RC are the Equivalent Series Resistance (ESR) of L and C, respectively. This transfer function is a second order low pass filter, with a left-half-plane introduced by the ESR of Kp = sin θ | G( jω ) H ( jω )| (5) Kd = Ki − ω cosθ ω 2 |G ( jω ) H ( jω )| (6) where, tan 2 = Kp T/Ki T2 Kd. The KP, Ki and Kd values are found to be 5.80125, 119.048 and 0.0869×10G4, respectively. 344 Res. J. Appl. Sci. Eng. Technol., 4(4): 342-349, 2012 Design of a sliding mode controller: C Step 1: Determination of the state variables: The output voltage error decay is essentially the main control objective in real time. The selected state variables are output voltage error, output voltage error dynamics and the integral of the error of the output capacitor (Utkin et al., 1999): C ⎤ ⎡V − βV O ⎥ ⎢ ref ⎥ ⎢d X = ⎢ (Vref − βVO ) ⎥ ⎥ ⎢ dt ⎥ ⎢ V − β V dt ( ) ref O ⎥⎦ ⎢⎣ C B 1 − ( L + CR0 RL ) LCRO 0 0⎤ ⎥ 0⎥ ⎥ 0⎥⎦ ⎡ x1 ⎤ ⎢ ⎥ ⎢ x2 ⎥ ⎢⎣ x3 ⎥⎦ B [V − β VO LC − ] + ( L + CRO RL ) β ic + CR0 <0 α1 β Lic α2 βV0 ( R0 − RL ) R0 (9) >0 For S ÷ 0G0 and Ð > 0 and âs ÷ 0G = 1, we get: CR0 ] (10) R0 The simplified existence condition in Eq. (10) is obtained by combining both Eq. (8) and (9): α3 α [V − βVO ]LC − 1 βLiC α 2 ref α2 ( L + CRO RL ) βV ( R − RL ) + < βVi βiC + 0 0 0< where ‘S’ is the instantaneous state variables trajectory. dt s→ 0 − <0 Step 5: Derivation of existence condition: For S ÷ 0+ and Ð < 0 and âs ÷ 0+ = 0, we get: [ ⎧1, when S > 0 u= ⎨ ⎩ 0, when S < 0 S = Kd = K T Ax + K T BU s→ 0 + α α β Vi > 3 Vref − βVo LC − 1 β LiC α2 α2 ( L + CRO RL ) β V ( R − RL ) + β ic + 0 0 Step 3: Define the switching status and sliding equation: d (Vref − β VO ) S& ref ⎡ ⎤ 0 ⎡ 0 ⎤ ⎢ ⎥ ⎢ − βV ⎥ β V ( R − RL ) ⎥ i⎥ +⎢ u+ ⎢ 0 O ⎢ ⎥ ⎢ LC ⎥ LCR0 ⎢ ⎥ ⎢ 0 ⎥ 0 ⎢⎣ ⎥⎦ ⎦ ⎣ C = K T Ax + K T BU s→ 0 − Step 2: Develop the state variables: ⎡ x&1 ⎤ ⎡ 0 ⎢ ⎥ ⎢ ⎢ x&2 ⎥ = ⎢ 0 ⎢⎣ x&3 ⎥⎦ ⎢ ⎢⎣ 1 S& s→ 0 + ∫ C has a form similar to a PID (Proportional, Integral and Derivative) controller. Third, it contained a form of state-feedback which provides a more flexible way to close the feedback loop to obtain any type of responses at will. Step 4: Ensure existence of sliding mode operation: The local reachability condition lims÷0 + s.Ñ < 0, must be satisfied. This is expressed as: CR0 + K p (Vref − β VO ) T ∫ R0 The condition given in Eq. (10) provides a range of employable sliding surface coefficients such that irrespective of the circuit parameters, the system trajectories near the surface are directed towards the sliding surface itself. (7) + Ki (Vref − Vo )dt O S = a1x1 + a2 x2 + a3 x3 = K T x (11) (8) Considering the design parameters, the existence condition is modified. From the left inequality of the existence condition, we get: where "1, "2 and "3 are sliding coefficients with KT = ["1 "2 "3]. This Eq. (7) has three significant implications. First, this control law states that not only the weighted sum of voltage error and its derivative needs to be constrained to be zero but also the integral of the voltage error must be included. Secondly, the Eq. (7) α1 < α2 LC α3 ( R − RL ) (V − βVo ) + βVo O α 2 ref RO βLic ⎡ 1 R ⎤ +⎢ + L⎥ CR L ⎦ ⎣ O 345 (12) Res. J. Appl. Sci. Eng. Technol., 4(4): 342-349, 2012 From the left inequality of the existence, we get: 1.00 a1 < a2 ( R − RL ) a3 (V − βVo ) − βVo o a2 ref Ro βLiC (13) ⎡ 1 RL ⎤ +⎢ + ⎥ L ⎦ ⎣ CRO 0.98 βVi − LC 0.96 0.94 0.92 0.90 0.88 Dividing Eq. (11) and (12): 0 ⎧ ⎡ ⎤⎫ R ⎤ α ⎡ Vref − Vo ⎥ ⎪ ⎪Vo ⎢1 − L ⎥ + 3 ⎢ Ro ⎦ α 2 ⎣ β ⎪ ⎦⎪ Vi = 2⎨ ⎣ ⎬ ⎪ ⎪ LC − LiC − i R C L ⎪ ⎪ CR o ⎭ ⎩ α1 < α2 LiC 25 30 35 17.6 17.4 17.2 17.0 16.8 16.6 16.4 16.2 (14) 16.0 15.8 15.6 0 ⎡ R ⎤ 1 +⎢ + L⎥ L ⎥⎦ ⎢⎣ CRo(max) 10 5 20 15 Load Ro ( Ω) 25 30 U eq = − [ K T Ax ]−1 K T [ Ax + D] ⎧⎪ ⎡ RL ⎤ Vi(min) ≥ 2⎨Vo ⎢1 − ⎥+ Ro(max) ⎥⎦ ⎩⎪ ⎢⎣ ⎫⎪ ⎤ Li c α3 ⎡ Vref − Vo ⎥ LC − − iC RL ⎬ ⎢ CRo(max) α2 ⎣ β ⎪⎭ ⎦ 35 Fig. 6: Plot between Ro and 81 for (17) â eq is continuous and hence 0< â eq < 1. Comparing Eq. (9) and (10) and mapping the equivalent control function onto the duty cycle function, we get: (15) Vcon = − λ1ic + λ2 (Vref − βVo ) + λ3βVo also α ( R − RL ) βVi − LC 3 (Vref − β Vo ) − βVo o α1 α2 Ro < α2 βLic Vramp = β Vi RL ⎤ ⎡ 1 +⎢ + CRO L ⎥⎦ ⎣ (18) (19) where ⎛ α1 λ1 = βL⎜ ⎝ α2 for ⎧ ⎡ ⎫ RL ⎤ α3 ⎪Vo ⎢1 − ⎪ ⎥+ ⎪⎪ ⎢⎣ Ro(max) ⎥⎦ α 2 ⎪⎪ Vi(min) ≥ 2 ⎨ ⎬ (16) ⎤ ⎪⎡ Vref ⎪ LiC − ic RL ⎪ − Vo ⎥ LC − ⎪⎢ CR β ⎪⎩⎣ ⎪⎭ o(max) ⎦ C 20 15 Load Ro (Ω) Fig. 5: Plot between Ro and 83 The range of input and loading conditions is considered and the modified existence condition is given as: ⎛ ⎞ RL ⎞ α ⎛ Vref ⎟ LC 3 ⎜ −Vo ⎟ +Vo ⎜⎜1− ⎟ R α2 ⎝ β ⎠ ⎝ o(max) ⎠ 10 5 and ⎛ − α R ⎞ 1 − L ⎟ , λ2 = 3 LC RoC L⎠ α2 λ3 = ⎜ 1 − ⎝ RL ⎞ ⎟ R0 ⎠ Variation of 81 and 83 with variations in the load is shown in Fig. 6 and 5 respectively. These figures show that the variation of 81 and 83 for 96.77% of load is only -11.15 and 11.65%, respectively. 82 remains constant irrespective of the variations in the load. Thus, the control signal does not undergo noticeable changes for variations in the load. Step 6: Derivation of control equations using equivalent control method: In the invariance condition, Ö = 0, substituting â- â eq, we get: 346 Res. J. Appl. Sci. Eng. Technol., 4(4): 342-349, 2012 20 26 18V 33V 30V 24V 13V 24 15 22 25 20 10 18 5 16 5.5 6.0 0 0.4 0.2 X10-4sec 0.6 0.8 (a) Fig. 7: Output response of a PID controlled buck converter for various inputs 11.86 Inductor current Capacitor current 2.0 7.0 6.5 0 11.85 1184 11.83 11.82 1.5 1.0 11.81 0.5 11.79 11.8 11.78 0.0 5.5 6.5 6.0 7.0 X10-4sec -0.5 0 22 24 26 -4 (b) 28 Fig. 9: Dynamic response of the PID controlled buck converter to step input change from 24 to 18 V (a) Input voltage (b) Output response Fig. 8: Inductor and capacitor current waveforms RESULTS AND DISCUSSION 11.94 PID controller: Figure 7-10 illustrates the results provided by the computer simulation of a PID controlled Buck converter using Matlab-Simulink. Figure 7 shows the output responses of the converter when it subjected to various input voltages. It can be noticed that as the input voltage decreases from 33 to 13 V, the peak overshoot as well as the settling time of the converter decreases. The inductor and capacitor current waveforms of the converter under continuous conduction mode are shown in Fig. 8. Figure 9 shows that when the input voltage is changed from 24 to 18V at 0.6 ms, we can infer that the system takes 0.09 ms to settle at a new steady state value which is less than the initial steady state value by 0.04 V. Figure 10 shows the dynamic response of the converter for a load step change from 6 to 25S. It is clear that the system takes 0.058 ms to settle at a new steady state point which is 0.008 V greater than its steady state value. 11.92 11.90 11.88 11.86 11.84 11.82 5.8 6.0 6.2 6.4 6.6 6.8 -5 Fig. 10: Dynamic response of the PID controlled buck converter to load step change from 6 to 25S waveforms of the converter under continuous conduction mode are shown in Fig. 12 which is similar to that of the waveforms of a PID controlled converter. When the input voltage is changed from 24 to 18 V at 0.4 ms, (Fig. 13), we can infer that the system takes 0.02 ms to settle at a new steady state value which is less than the initial steady state value by 0.6 :V (Fig. 13) which is a very negligible value when compared to that of the PID controller. SMC waveforms: Figure 11-15 illustrates the results provided by the computer simulation of a PID based SM controlled Buck converter using Matlab - Simulink. 347 Res. J. Appl. Sci. Eng. Technol., 4(4): 342-349, 2012 15V 33V 30V 24V 20 13V 12.05 12.04 12.03 15 12.02 25 12.01 10 12.00 11.99 5 11.98 0 0 2 1 3 2.5 X10-4sec 3.5 X10-4 sec Fig. 14: Dynamic response of the PID based SMC controlled buck converter to load step change from 6 to 25S Fig. 11: Output response of a PID based SMC controlled buck converter for various inputs PID controller Inductor current Capacitor current 2.5 3 20 2.0 SM controller 15 1.5 25 1.0 10 0.5 5 0.0 0 -0.5 0 2.4 2.8 2.6 3.0 0.2 0.4 0.6 X10-4sec 0.8 X10-4sec Fig. 15: Output response of PID based SMC versus PID controllers Fig. 12: Inductor and capacitor current waveforms 26 SMC PID 12.05 24 O utput voltage (VO) 22 20 18 16 3.2 3.4 3.6 3.8 4.0 4.2 4.4 X10-4sec 12.00 11.95 11.90 11.85 11.80 11.75 (a) 12 11.992 14 16 18 20 22 24 Input voltage (Vt) 26 28 30 Fig. 16: Variation of output with variation of input 11.900 Table 2: Comparison between PID and PID based SMC Buck Converter Parameters PID control PID based SMC Settling time 4.7×10G4s 1.84×10G4s Maximum peak overshoot 6.22 V 5.253 V Steady state error 0.16 V 0.009 V 11.988 11.966 11.964 11.962 3.2 3.4 3.6 3.8 4.0 4.2 When there is a load step change from 6 to 25S, the system takes 0.035 ms to settle with negligible steady state error (Fig. 14). Figure 15 shows the output response of a PID based SMC versus PID controller. It can be inferred from the figure that the SMC response has much less settling time than that of the PID controlled Buck converter response. 4.4 X10-1sec (b) Fig. 13: Dynamic response of the PID controlled buck converter to step input change from 24 to 18 V at 0.4 ms (a) Input voltage (b) Output response 348 Res. J. Appl. Sci. Eng. Technol., 4(4): 342-349, 2012 Moreover, the maximum peak overshoot and steady state error is more for the latter when compared with the former. When subjected to sudden line and load variation, the PID based SMC settles at a much less time as compared to that of a PID controller. Table 2 gives a comparison of the performance parameters of the PID controlled Buck converter and PID based SMC controlled Buck Converter. Figure 16 shows the variation of output with variation in input for both converters. It can be inferred that when there is 56.67% variation in the input, the PID controller shows a variation of 1.072%, whereas a PID based SMC shows a variation of only 0.192%. REFERENCES Duan, Y. and H. Jin, 1999. Digital Controller Design for Switch mode power converters. Proceedings in 14th Annual Power Electronics Conference Exposition, Dallas, Tx, 2: 14-18, 967-973. Liping, G., Y.H. John and R.M. Nelms, 2009. Evaluation of DSP based PID and fuzzy controllers for DC-DC Converters. IEEE T. Indus Electr, 56(6): 2237-2248. Mummadi, V., 2004. Signal flow graph modeling and analysis of dc-dc converters. IEEE T. Aerospace Electr. Syst., 40(1): 259-271. Perry, G., G. Feng, Y.F. Liu and P.C. Sen, 2007. Design method for PI-like fuzzy logic controllers for dc-dc converters. IEEE T. Indus. Electr, 54(5): 2688-2695. Prodic, A.M., 2002. Design of a digital PID regulator based on look-up Tables for control of high frequency dc-dc converters. Proceedings in IEEE Workshop on Computing and Power Electronics, pp: 18-22. Siew, C.T., Y.M. Lai and K.T. Chi, 2008. Practical issues in the design of sliding mode controllers. IEEE T. Indus. Elect., 55(3): 1160-1174. Utkin, V., J. Guldner and J.X. Shi, 1999. Sliding mode in Electromechanical Systems. Taylor and Francis, London, UK. CONCLUSION The design and analysis of a dc-dc Buck converter operating in continuous conduction mode with Proportional-Integral-Derivative control and PID based Sliding Mode Control (SMC) is studied in detail in this paper. The converter employing these two controllers is subjected to sudden line variation as well as load variation. Analysis shows that the PID controller relies more on the operating point whereas the PID based SMC is much robust to small-signal and large variation from the operating point. Moreover the PID based SMC produces negligible steady state error and settles with a much less settling time than the conventional linear PID controller. 349