Pakistan Journal of Engineering and Technology, PakJET

Multidisciplinary | Peer Reviewed | Open Access

Volume: 1, Number: 1, Pages: 21-28, Year: 2018

Estimating Dynamics of Switching Converters Using

System Identification Technique
Nasir Abbas 1 and Ghulam Abbas *,1
1
Electrical Engineering Department, The University of Lahore, Lahore 54000, Pakistan
*
Corresponding author: Ghulam Abbas (e-mail: [Link]@[Link])

Abstract—The switching converters are widely used in power electronics. These converters convert low input DC voltage to higher DC
output voltage by switching the source voltage. These converters are nonlinear in nature. This paper deals with the estimation of
dynamics of converters. Applying system identification technique, one can estimate the transfer function of switching converters which
can be used for the controller design. Although many identification techniques such as cross correlation and circular correlation-based
and state space averaging have been used to estimate converter dynamics but they require heavy mathematical computations. In the
proposed research, frequency response-based data from duty cycle to output voltage estimation have been employed to estimate
converter dynamics. In this method, frequency response data (FRD) are collected by injecting the sine stream signal having frequency
from few hertz to many kilo hertz at duty cycle input which excite the system at different frequencies and varying output response is
obtained. The transfer functions of converters such as buck-boost, boost and buck are estimated using such a system identification
technique. Well- estimation of the converter transfer functions is validated through MATLAB/Simulink based simulation results.

Index Terms—Bode plot, frequency response data, identified transfer function, switching converters, system identification,
MATLAB/Simulink.

I. INTRODUCTION buck, boost and buck-boost are considered and have been
In modern era, all the services depend upon continuous identified. MATLAB/Simulink will be used for simulation
functioning of different electronics devices and systems which purposes.
require energy for their reliable operation. The energy is Besides, it is possible to create arbitrary excitation
treated by switching converters for reliable operation which is waveform with a broad band spectrum to collect all spectral
prime importance for their quality. These are used to convert information in one measurement. There are ten signals used in
one DC voltage level to other one by switching the source [2] to estimate transfer function of switching converters by
voltage. Due to their switching behavior, they are nonlinear in frequency response data. It used Maximum length binary
nature. Determining their dynamics through well-established sequence (MLBS) to estimate the transfer function. The
techniques is a difficult task. Because the lengthy derivation MLBS is used to evaluate transfer function of any linear
and complicated calculation is compulsory to obtain desired system and now also being used in field of acoustic [3],
results. Firstly, one has to linearize, then average the circuit impedance spectroscopy of single living cells [4], sensors for
over a complete cycle to get approximated dynamics. To gas, odor and or aroma analysis [5], sonar system for
overcome past complications, we will use frequency response zooplankton survey [6]. The method is equally employed for
technique to achieve the intended objectives. By using this power converters [7]-[11]. But this method is not feasible to
technique, we will be able to find the transfer function of any measure all transfer functions of power converters. So, we will
system as it does not require haddock mathematical derivation. use sine sweep method to estimate transfer function of
The dynamics of switching converters can be characterized purposed switching converters. Linear systems can also be
by transfer function. In this paper, the transfer function of the identified by extensively studied method called cross
switching converter is estimated using the system correlation in system identification tools [12]. But this
identification techniques. In one of the techniques, the transfer method, working on the basis of cross-correlating the system
function is determined from frequency response data with the input and output, causes errors in results when we deal with
help of Sine Sweep Method (SSM) [1]. In this method, finite signal length. Another familiar method Autoregressive
frequency response is obtained by injecting the sinusoidal with Exogenous Terms (ARX) is used in system identification
signal having frequency from few hertz to many kilo hertz into to estimate the frequency response of model. This method
the input of the converter. The estimated transfer function will creates disturbances which are also part of the system
be helpful to find the controlling parameters of the switching dynamics.
converter to regulate the output. Three converters, namely In the proposed method named Sine Sweep Method (SSM),
Nasir Abbas et al. PakJET

a change in amplitude and phase of sinusoidal output with Switching Circuit 2nd Order Low-Pass Filter Load
respect to input signal, on the application of a sinusoidal signal Q1 L Vout
of known frequency to a linear system input, is observed. The
frequency response data estimation is best technique to find
the dynamics of a switching converters.
The paper is constructed in the following way. Operation of Vin
D C R
switching converters is described in Section II. Dynamics of
the converters using system identification technique are
estimated in Section III. In Section IV, additional simulation
results are presented to further validate the estimated transfer
functions. Section V contains conclusion. Fig. 2. Basic buck converter circuit.

II. SWITCHING CONVERTERS FUNDAMENTALS The duty ratio k relating input voltage Vin and output voltage
Switching converters are electronic systems which alter one
Vout of the of buck converter is given in (1)
DC level to another DC level by switching action. Switching
converters are DC to DC voltage converters which provide Vout
high efficiency as compared to linear power supplies. The k (1)
Vin
main trend of using switching converters is their smaller size,
flexibility, not having transformer and provides efficiency of
75% to 98%. Therefore, they are used extensively in solar B. Boost Converter
battery charging system, UPS system, medical electronics, Boost converter is the DC to DC converter which converts
communication, computer peripherals and personal computers low DC voltage level to another DC voltage level higher than
[13]. Modern electronic instruments need light weight, high input. The boost converter consists of a MOSFET switch,
quality and efficient power supplies. The operation of buck energy storing elements inductor and capacitor, diode, power
converter, boost converter and busk boost converter is supply and load as shown in Fig 3.
discussed here. The input and output of switching converters The input voltage Vin and the output voltage Vout , in case of
are shown in Fig. 1. Here duty cycle is input to switching boost converter, are related as [14]
converters and voltage is the output.
Vin
Vout  (2)
Dynamics by Averaging and Linearization 1  k 
Technique (Tedious)
L D Vout

Duty Cycle Output Voltage

Switching Converters
(Input) (Output)
Q1 C R
Vin

Dynamics by System Identification Technique

(Suggested)

Fig. 1. Input-output representation of switching converter. Fig. 3. Basic boost converter circuit.

A. Buck Converter C. Buck-Boost Converter

The buck converter circuit consists of a switching transistor Buck-boost converter (see Fig. 4) like other switched mode
along with flywheel circuit containing capacitor, inductor and power supplies converts one DC level to another DC level and
diode. When transistor is on, the current starts to flow through is the combination of principles of both buck and boost
the energy storing element called inductor and stores energy in converters in single circuit due to switching action of
the form of magnetic field. The inductor opposes the abrupt MOSFET.
change in current. In this way the output of switching When switch is turned on, the input energy is supplied to
transistor output is prevented from increasing suddenly to its inductor and it gets magnetized. The current flows from
peek value because inductor stores energy taken from the positive polarity to negative polarity via inductor. The diode is
increasing output. The stored energy in the form of magnetic reverse biased and current passes through load and charges the
field is delivered to the circuit when transistor is switched off. capacitor. The load is supplied by the capacitor that was
The basic buck converter circuit is shown in Fig. 2. charged in previous cycle. When MOSFET is turned off, the
inductor and input is disconnected. During off state the

22
Nasir Abbas et al. PakJET

inductor is the main source of energy. Now diode is forward

biased and circuit current flows through load and capacitor is
discharged. In this way the output voltage is at least equal to
or greater than source voltage.
The output voltage Vout and the input voltage Vin are related as

Vout k
 (3)
Vin 1 k

where k is duty cycle of buck-boost converter.

Q1 VD D Vout

Fig. 6. Sinestream injected at point of linearization.

The complete algorithm is illustrated in flow chart shown in

Vin L C
R Fig 7.

Start

 
Taking Switching Converter
Fig. 4. Basic buck-boost converter circuit.

III. SYSTEM IDENTIFICATION TECHNIQUE

Input: Defining the Input and Output of Output:
System identification refers to making the mathematical Duty Cycle Converter Output Voltage
models of dynamic system with the help of measured data.
System identification is divided into parametric and
nonparametric techniques [15], [16]. In parametric Generating Sinestream to Inject it
at Input of Converter
identification a system model is assumed and model
parameters are estimated. But in nonparametric no assumption
is made about the model under test and identification is used
Collecting Frequency Response
to directly compute the frequency response of system. Data and Drawing Bode Plot
Nonparametric method includes correlation analysis, transient
response analysis, frequency response, Fourier, and spectral
analysis. In this research, we focus on nonparametric Estimating the Transfer Function
from Frequency Response Data,
identification. We have used frequency response estimation to thus Bode Plot
find the dynamics of converter.
Frequency response estimation requires input signal at input
linearization point which excites the model at desired Validating the Estimated Transfer
Function
frequencies, like sinestream or chirp signal. We have used
sinestream signal as shown in Fig. 5, which is a series of
sinusoids. Output response is achieved at output linearization End
point (see Fig. 6).
Fig. 7. Flow chart describing identification technique mechanism.

The input and output, duty ratio and switching frequencies

for the three converters, namely buck, boost and buck-boost
have been summarized in Table I.
TABLE I
CONVERTER SPECIFICATIONS
Switching Input Output Duty
Converter
Frequency Voltage Voltage Ratio
Buck 10 kHz 200 V 150 V 75 %
Boost 100 kHz 24 V 48 V 50 %
Buck-Boost as 10 kHz 12 V -5 V 29.4 %
Fig. 5. Sinestream having different frequencies. Inverting Buck
Buck-Boost as 10 kHz 12 V -28 V 70 %
Inverting Boost

23
Nasir Abbas et al. PakJET

A. Buck Converter Transfer Function Estimation Once the Bode plot is achieved from the frequency response
Buck converter (stepping down 200 V to 150 V by duty data, transfer function of the converter can be easily extracted
ratio of 0.75) to be estimated is illustrated in Fig. 8. The open from the Bode plot. The estimated transfer function of the
loop response is obtained by introducing the sinusoidal signal buck converter, thus, is
having frequency variation from few 100 kHz to 20 kHz at the
‘duty cycle input’. Resultantly, a plot of gain and phase 2.198  109
GBuck ( s )  (4)
variation with respect to frequency sweep is shown in the form s 2  2975 s  1.105  107
of Bode plot in Fig. 9.
The Bode plot of the estimated transfer function of the buck
converter matches exactly to the one obtained through
frequency response data (see Fig. 11). This confirms the well-
mapping of FRD into the transfer function.

Fig. 8. Buck converter Simulink model.

Fig. 11. Bode plot of FRD (blue line) and estimated transfer function (asterick
red line).

B. Boost Converter Transfer Function Estimation

The boost converter steps up 24 V to 48 V. The sine stream
Fig. 9. Frequency response Bode plot of buck converter Simulink model.
input signal is a series of sinusoids with frequency range 100
kHz to 20 kHz. The proposed model configuration is shown in
The filtered steady state time response along with its FFT of Fig. 12. The frequency response data is collected and then
the circuit is shown in Fig. 10. used to get the Bode plot. The Frequency Response Bode Plot
(FRBP) is illustrated in Fig. 13.

Fig. 12. Boost converter Simulink model.

Fig. 10. Output response of buck converter against sinestream.

24
Nasir Abbas et al. PakJET

1) Buck-Boost as An Inverting Buck

The Simulink model is illustrated in Fig. 15. The frequency
response Bode plot of buck-boost converter is shown in Fig.
16 against the sinestream applied at the duty cycle input.

Fig. 15. Buck-boost Simulink model.

Magnitude (dB)
Fig. 13. Frequency response Bode plot of boost converter Simulink model.

The transfer function of boost converter stepping up voltage

from 24 V to 48 V is estimated using frequency response data
and is expressed as
3.151 105 s  2.637  108
GBoost ( s )  2 (5)
s  3282 s  2.789  106
The Bode plot of the estimated transfer function of the
Phase (deg)

boost converter matches exactly to the one obtained through

frequency response data (see Fig. 14). This confirms the well-
mapping of the frequency response data into the transfer
function.

Fig. 16. Frequency response Bode plot of buck-boost Simulink model.

When buck-boost converter acts as buck converter then its

duty cycle should be less than 50%. The considered buck-
boost converter converts 12 V to -5V using duty cycle of
29.4%. The output displayed by the converter against the
sinestream is shown in Fig. 17.

Fig. 14. Bode plot of the FRD (blue line) and estimated transfer function
(asterick red line).

C. Buck-Boost Converter Transfer Function Estimation

The buck-boost converter is constructed by amalgamating
the principle of both buck and boost converter. The proposed
technique is very simple and can be used to linearize all types Fig. 17. Output response of buck-boost converter acting as inverting buck.
of plant.

25
Nasir Abbas et al. PakJET

The estimated transfer function of the buck-boost converter,

in case of inverting buck, from FRD is expressed by

1.019  105 s  3.64  109

GBuck  Boost ( s )  (6)
s 2  5.304  10 4 s  2  108
The Bode plot of the estimated transfer function of the
buck-boost converter matches exactly to the one obtained
through frequency response data (see Fig. 18). This confirms
the well-mapping of the frequency response data into the
transfer function.

Fig. 20. Output response of buck-boost converter acting as inverting boost.

The estimated transfer function of the buck-boost converter,

in case of inverting boost, from FRD is expressed by

1.851 105 s  1.081 109

GBuck  Boost ( s )  (7)
s 2 1985 s  8.747  106

The Bode plot of the estimated transfer function of the

buck-boost converter matches exactly to the one obtained
through frequency response data (see Fig. 21). This confirms
the well-mapping of the frequency response data into the
Fig. 18. Bode plot of FRD (blue line) and estimated transfer function (asterick
red line).
transfer function.

2) Buck-Boost as An Inverting Boost

The buck-boost converter acts as an inverting boost
converter when duty cycle lies between 0.5 and unity (in our
case it is 0.7). The Bode plot obtained by FRD is shown in
Fig. 19 whereas the output against the sinestream is shown in
Fig. 20.
Magnitude (dB)
Phase (deg)

Fig. 21. Bode plot of FRD (blue line) and estimated transfer function (asterick
red line).

In summary, the estimated transfer functions of three

converters named buck, boost and buck-boost are tabulated in
Fig. 19. Bode plot of frequency response data.
Table II.

26
Nasir Abbas et al. PakJET

TABLE II
ESTIMATED TRANSFER FUNCTIONS OF SWITCHING CONVERTERS
Converter Estimated Transfer Function
Buck 2.198  109
GBuck ( s )  2
s  2975 s  1.105  107
Boost 3.151  105 s  2.637  108
GBoost ( s )  2
s  3282 s  2.789  106
Buck-Boost 1.019  105 s  3.64  109
as Inverting Buck GBuck  Boost ( s )  2
s  5.303  104 s  2  108
Buck-Boost as Inverting 1.851  105 s  1.081  109
Boost GBuck  Boost ( s ) 
s 2 1985 s  8.747  106 Fig. 24. Simulink model of boost converter.

IV. ADDITIONAL SIMULATION RESULTS

For the sake of validating the transfer functions of the
converters estimated through proposed identification
technique, additional MATLAB/Simulink based simulation
results are presented. Open-loop response for the circuits and
their estimated transfer functions is observed. From Figs. 22 to
27, it can be observed that the estimated transfer functions
show the same step response as that of their circuits from
which they are derived. This further validates that the
estimated transfer functions well-model the circuits.

Fig. 25. Response shown by boost converter.

Fig. 22. Simulink model containing buck converter circuit and its estimated
transfer function.

Fig. 26. Simulink model of buck-boost converter.

Vout (t)

Fig. 23. Response shown by buck converter circuit and its estimated transfer
function.

Fig. 27. Response shown by buck-boost converter.

27
Nasir Abbas et al. PakJET

V. CONCLUSIONS in IEEE Transactions on Power Electronics, vol. 20, no. 5, pp. 1093-
1099, Sept. 2005.
In past the non-parametric identification of switching [11] B. Miao, R. Zane and D. Maksimovic, "A modified cross-correlation
converter has not been discussed in detail. It is due to lack of method for system identification of power converters with digital
knowledge about potential of fast frequency response control," 2004 IEEE 35th Annual Power Electronics Specialists
Conference (IEEE Cat. No.04CH37551), Aachen, Germany, 2004, pp.
measurement and analysis. Besides this, frequency response 3728-3733 Vol.5.
measurement requires modern knowledge regarding [12] B. Miao, R. Zane and D. Maksimovic, "Practical on-line identification
identification and restriction of practical implementation. Here of power converter dynamic responses," Twentieth Annual IEEE Applied
Power Electronics Conference and Exposition, 2005. APEC 2005.,
we have used Sine Sweep Method (SSM) by means of which Austin, TX, 2005, pp. 57-62 Vol. 1.
proposed converters from frequency response data has been [13] M. K. Kazimierczuk, Pulse-Width Modulated DC-DC Power
identified accurately. Using frequency response estimation Converters. John Wiley & Sons, 2015.
technique, the transfer functions of all converters have been [14] Ljung, Lennart, "System identification," In Signal analysis and
identified successfully. The proposed technique is numerically prediction, pp. 163-173. Birkhäuser, Boston, MA, 1998.
verified and validated in MATLAB Simulink Environment. [15] L. Ljung, System Identification: Theory for the User. Prentice Hall PTR,
1999.
Cross Correlation method has been used in past to identify the
[16] G. F. Franklin, J. D. Powell, and M. L. Workman, Digital Control of
switching converters but it causes error to identification results Dynamic Systems. Addison-Wesley, 1998.
especially when considering finite length of signal. Now Sine
Sweep Method is introduced in this research to nullify the
inaccuracies on frequency response measurement. The
proposed method can be applied both in development phase
and production of switching converters. Here the main critical
thing is to select the suitable range of amplitude an applying
too much greater value can cause non-linear distortions. In
future the controller can be designed to find suitable excitation
amplitude about the converter under experiment.

REFERENCES
[1] M. Vilkko and T. Roinila, "Designing maximum length sequence signal
for frequency response measurement of switched mode converters,"
In Nordic Workshop on Power and Industrial Electronics
(NORPIE/2008), June 9-11, 2008, Espoo, Finland. Helsinki University
of Technology, 2008.
[2] R. M. H. M. Rathnayake, R. Dutta, J. E. Fletcher and D. Xiao, "Torque
ripple minimization of an IPMSG with fractional-slot, concentrated-
winding using uncontrolled rectifier-connected boost converter
control," 2017 20th International Conference on Electrical Machines
and Systems (ICEMS), Sydney, NSW, 2017, pp. 1-6.
[3] J. Schoukens, R. Pintelon, E. van der Ouderaa and J. Renneboog,
"Survey of excitation signals for FFT based signal analyzers," in IEEE
Transactions on Instrumentation and Measurement, vol. 37, no. 3, pp.
342-352, Sept. 1988.
[4] B. Miao, R. Zane and D. Maksimovic, "Practical on-line identification
of power converter dynamic responses," Twentieth Annual IEEE Applied
Power Electronics Conference and Exposition, 2005. APEC 2005.,
Austin, TX, 2005, pp. 57-62 Vol. 1.
[5] J. Vanderkooy, “Aspects of MLS Measuring Systems,” JAES, vol. 42,
no. 4, pp. 219–231, Apr. 1994.
[6] S. Gawad, T. Sun, N. G. Green, and H. Morgan, “Impedance
spectroscopy using maximum length sequences: Application to single
cell analysis,” Review of Scientific Instruments, vol. 78, no. 5, p. 054301,
May 2007.
[7] Ning Xiang and Dezhang Chu, "Fast m-sequence transform for quasi-
backscatter sonar in fisheries and zooplankton survey
applications," Proceedings 7th International Conference on Signal
Processing, 2004. Proceedings. ICSP '04. 2004., Beijing, 2004, pp.
2433-2436 vol.3.
[8] M. E. H. Amrani, R. M. Dowdeswell, P. A. Payne, and K. C. Persaud,
“Pseudo-random binary sequence interrogation technique for gas
sensors,” Sensors and Actuators B: Chemical, vol. 47, no. 1, pp. 118–
124, Apr. 1998.
[9] A. Barkley and E. Santi, "Improved online identification of switching
converters using digital network analyzer techniques," 2008 IEEE
Power Electronics Specialists Conference, Rhodes, 2008, pp. 891-896.
[10] B. Miao, R. Zane and D. Maksimovic, "System identification of power
converters with digital control through cross-correlation methods,"

28