144 IEEE POWER ELECTRONICS LETTERS, VOL. 3, NO. 4, DECEMBER 2005 Interpreting Small Signal Behavior of the Synchronous Buck Converter at Light Load Julian Yan Zhu Abstract—The pulse-width-modulation (PWM) buck converter with synchronous rectifiers operating at light load is usually modeled by its continuous conduction mode (CCM) model. However, the actual power-stage small-signal control-to-output response shows a different behavior from what the traditional CCM model predicts, specifically, more damping around the double-pole frequency, instead of more resonance. This paper presents a modified small-signal light-load model for a synchronous buck converter. The developed model accurately predicts the actual small-signal behavior of a PWM converter at light load. The derived averaged switch model for light load can also be used for the small-signal models of the other basic PWM converters operating in CCM at light load. Fig. 1. Bode plots of the power-stage transfer function of a synchronous buck converter. Index Terms—Light load, PWM converter, small-signal model. I. INTRODUCTION ULSE-WIDTH-MODULATION (PWM) buck converters are usually modeled by the well-known small-signal averaging models that have existed for many years [1], [2]. The models were developed for continuous and discontinuous conduction operation at heavy load and light load respectively. Among them, the continuous conduction mode (CCM) model is more often used for its wide applications and easy control design. The model predicts a control-to-output voltage transfer function with a double-pole resonance. Its dc gain changes with the input voltage, and its resonance become stronger with an increased phase lag around the resonant poles at light load. Therefore, the feedback control loop should be designed carefully to ensure stability over the entire range of input voltage and load. In recent years, the synchronous buck converter has been widely adopted by industry by replacing the rectifier diode with a complementary switching MOSFET. It can achieve high efficiency by taking advantage of the low conduction loss of the MOSFET. Moreover, it is able to operate in CCM even at light load, and its operation allows the inductor current to reverse, making it possible to turn on the top MOSFET with zero voltage. It was called a quasi-square-wave converter in the early 1990s [3]–[5]. The CCM model described in Middlebrook’s paper [1] is usually used to design the control loop of the synchronous buck converter. However, the actual measurement of the power-stage transfer function shows a result different from what people expect from the model. The resonance is smoothed out when the load current decreases, as Fig. 1 shows. This is against the theory P Manuscript received May 31, 2005; revised November 17, 2005. Recommended by Associate Editor J. Sun. The author is with the Power Business Unit, Linear Technology Corporation, Milpitas, CA 95035 USA (e-mail: jzhu@linear.com). Digital Object Identifier 10.1109/LPEL.2005.863605 that says the resonance should be stronger at light load. We tend to believe that the model could not be wrong, as it has been published in textbooks [6], [7] for more than twenty years. But our measurement also seems to be correct. Thus, the reasons that cause this discrepancy have to be ascertained so that a proper control loop can be designed based on the right model. So far, few papers have addressed this issue. In past years, some papers discussed the models of quasi-square-wave converters [8]–[10]. However, reference [9] addresses the variable-frequency control model, which is not applicable to our constant-frequency PWM control. Reference [10] uses sampled-data modeling techniques to develop a small-signal model for the boost converter. It involves complicated mathematical modeling and the calculation of discrete-time small-signal model matrices. Therefore, it is not convenient for engineering design. Although its results also show the damping effect at light load, it does not explain the reason. Reference [8] derives the steady state and small-signal ac model for various nonisolated dc-dc converters. Its models include the switching transition times, which are longer than the traditional PWM hard-switching converters. However, its ac model still predicts a more severe resonance at light load. So the above phenomenon still cannot be explained. As such, it is necessary to derive a light-load small-signal model for the real situation. This paper derives a new model for the synchronous buck converter at light-load operation. The model predicts the lightload small-signal behavior that the experimental result exhibits. It includes the nonideal device parameters that are usually ignored in the model. II. SMALL SIGNAL MODEL OF SYNCHRONOUS BUCK CONVERTER The CCM model for the traditional PWM buck converter is shown in Fig. 2. The small-signal control-to-output transfer function is described in (1). It can be seen that the resonance is affected by load resistor R and the inductor dc resistor .A 1540-7985/$20.00 © 2005 IEEE Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:034 UTC from IE Xplore. Restricon aply. ZHU: INTERPRETING SMALL SIGNAL BEHAVIOR Fig. 2. 145 CCM ac small-signal circuit model for a buck converter. Fig. 4. Small-signal model of the synchronous buck converter in Fig. 3. Fig. 5. Inductor current at very light load. Fig. 3. Synchronous buck converter including parasitic components. high R incurs strong resonance. The experimental measurement shows that the control-to-output small-signal behavior of the synchronous buck converter under heavy load can be predicted well by this model. It is understandable, since the synchronous buck converter works exactly as a traditional buck converter. However, the light-load behavior is different, as shown in Fig. 1. The following sections will derive a modified small-signal lightload model based on the PWM switch-averaging model: amount of valley current is shifted into the Schottky diode. After a short deadtime, , the top MOSFET M1 is turned on again, the diode stops conducting, and the inductor current increases. During the conducting time of the diode, a relatively high resistance appears in the current path. After averaging this resistance can be written in (3). over the whole switching cycle, (3) (1) In order to evaluate the influence from the circuit parasitic components at light load, the synchronous buck converter is is the drawn in Fig. 3, including parasitic components. of the top MOSFET M1 and trace resistance, is the equivis the alent series resistance in the rectifier current path. ESR of the input capacitors, is the dc resistance of the inductor, and is the equivalent series resistance of the output capacitors. The small-signal model including these parasitic components is derived in Fig. 4. Its control-to-output transfer function is modified as in (2), shown at the bottom of the page. It can be , and all contribute to an equivalent seseen that , which adds to the damping of the ries resistor together with model. Assuming that the device temperatures are around 25 at all the load conditions, , and do not change significantly with the output current, while does. This is because is combined with of the bottom MOSFET M2 and the ESR of the Schottky diode, where the ESR of the Schottky diode increases considerably when the forward current is small. It can be found from most diode datasheets. At light load, the valley current is close to zero before it goes to negative. When the MOSFET M2 is turned off, this small is normally very small compared to the switching Although cycle (from tens of nanoseconds to a hundred nanoseconds), the can easily increase from tens of to hundreds averaged of , because the ESR of the diode can go up to a few ohms when the forward current drops below 1A. The increased damps the resonance at light load. When the load current reduces further, the inductor current goes negative, as shown in Fig. 5. At the turn-off points of the top MOSFET and the bottom MOSFET, the inductor current charge and discharge the junction capacitors of the MOSFETs ramps up and down and diodes. The switch node voltage linearly during the switching intervals, which become longer than those under heavy load. The small-signal behavior changes accordingly. In reality, the inductor current is not always able to up to , The resultant will have a rising edge charge in Fig. 5. Based on the waveforms in Fig. 5, a new model is derived as follows. Assume the transition intervals are very small and ., The inductor current can then be treated as a constant current during the intervals. According to the three-terminal-switch average model [2], the averaged terminal current and voltage can be written in (4) and (5): (4) (5) (2) Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:034 UTC from IE Xplore. Restricon aply. 146 IEEE POWER ELECTRONICS LETTERS, VOL. 3, NO. 4, DECEMBER 2005 TABLE I PARASITIC COMPONENTS OF THE DEVICES AND CIRCUIT Fig. 6. Small-signal model of the synchronous buck converter at light load. On the other hand, the input current is also affected by the input voltage disturbance with an additional current component. When the switching intervals approach zero, the model is reduced to the original without transition intervals in Fig. 4. The complete model parameters are derived and given in the Appendix. In (4) and (5): III. EXPERIMENTAL VERIFICATION In the above formulas, , , is the junction capacitance , respectively, and of M1, M2 and the Schottky diode during , , is the junction capacitance of M1, M2 and the Schottky diode during , respectively. The small-signal model can be derived in (6) and (7) by applying a small perturbation around the steady-state operating points to (4) and (5) as follows: (6) (7) Letting and , the equivalent small small-signal ac model is shown in Fig. 6. It can be seen that this light-load operation includes an additional lossless resistance in the small-signal ac equivalent circuit. This resistance causes more damping to the resonance at light load. The value of is derived in (8), and the control-to-output voltage transfer function can be modified as (9), shown at the bottom of the page, where : (8) The developed model was established for an 8 24 V input, 3.3 V/15 A output synchronous buck converter. The converter operates at 525 kHz. The output inductor and capacitor are and , respectively. The top and chosen as 1 bottom MOSFETs are HAT2168H and HAT2165H, respectively. A Schottky diode (B340) is also added in parallel with the synchronous bottom FET. Other parameters of the converter are summarized in Table I. Based on the above model, a MathCAD program is written to calculate the transfer functions of the power and control circuits. The calculated control-to-output transfer functions for different load currents are plotted in Fig. 7, in comparison with the measured ones (using a Venable frequency analyzer). It can be seen that the calculated transfer functions correlate well with the measured results both at light load and heavy load. The transfer functions based on the original model are also plotted in Fig. 8, which shows significant discrepancies to the measured results at light load. It is noted that the light-load small-signal behavior observed above actually causes less phase lag, which is good for the stability of the system fortunately. This explains the fact that the compensation is commonly designed at heavy load for the worst case consideration, and thus helps us better understand the light-load small-signal model of PWM converters. Also, the developed model is helpful to the appropriate compensation loop design for some special converters operating at critical conduction mode or quasi-square-wave mode. IV. CONCLUSIONS This letter presents a modified light-load small-signal ac model for a synchronous buck converter. The developed model accurately predicts the actual small-signal behavior of a PWM converter at light load: more damping effect, contrary to strong resonance, as the ideal model predicts. The derived averaged-switch model for the light load can also be used to model other synchronous PWM converters operating at light load, such as boost, buck/boost, sepic and Cuk converters. The (9) Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:034 UTC from IE Xplore. Restricon aply. ZHU: INTERPRETING SMALL SIGNAL BEHAVIOR 147 Fig. 7. Measured power-stage small-signal response and the calculated ones with the modified model (solid line: measured; dotted line: simulated). Fig. 8. Measured power-stage small-signal response and the calculated ones with the original model (solid line: measured; dotted line: simulated). general modeling idea is also helpful for accurate modeling of some converters operating in the critical conduction mode. APPENDIX Note: is assumed at light load. REFERENCES [1] R. D. Middlebrook and S. Cuk, “A general unified approach to modeling switching converter power stages,” in Proc. IEEE Power Electronics Specialists Conf. (PESC), 1976, pp. 18–34. [2] V. Vorperian, “Simplified analysis of PWM converters using the model of the PWM switch Part 1: continuous conduction mode,” IEEE Trans. Aerosp. Electron. Syst., vol. 26, pp. 490–496, May 1990. [3] V. Vorperian, “Quasi-Square-Wave Converters: Topologies and Analysis,” IEEE Trans. Power Electron., vol. 3, no. 2, pp. 183–192, Apr. 1988. [4] C. P. Henze, H. C. Martin, and D. W. Parsley, “Zero-voltage switching in high frequency power converters using pulse-width modulation,” in Proc. IEEE Applied Power Electronics Conf. (APEC), 1988, pp. 33–40. [5] D. Marksimov and S. Cuk, “Constant–frequency control of quasi-resonant converters,” IEEE Trans. Power Electron., vol. 6, no. 1, pp. 141–150, Jan. 1991. [6] R. D. Middlebrook and S. Cuk, Advances in Switched-Mode Power Conversion. Pasadena, CA: TESLAco, 1981, vol. I & II. [7] G. Chrysiss, High Frequency Switching Power Supplies: Theory and Design. New York: MCGraw-Hill, 1984. [8] W. Moussa and J. E. Morris, “DC and AC characteristics of zero voltage switching PWM converters,” in Proc. IEEE PESC, vol. 1, 1992, pp. 236–242. [9] J. M. F. Dores Costa and M. Medeiros Silva, “Small signal models and dynamic performance of quasisquare-wave ZVS converters with voltage-mode and current-mode control,” in Proc. IEEE Midwest Symp. Circuits and Systems, vol. 2, 1995, pp. 1183–1188. [10] A. J. Forsyth and R. I. Gregory, “Small-signal modeling and control of the quasi-square-wave boost converter,” IEEE Trans. Power Electronics, vol. 13, no. 1, pp. 36–46, Jan. 1998. Authorized licensd use limted to: IE Xplore. Downlade on May 10,2 at 19:034 UTC from IE Xplore. Restricon aply.