ECE 463/ECE 663

Spring 2024

Mehrdad Kazerani
Electrical & Computer Engineering
University of Waterloo
©Mehrdad Kazerani, 2024
 LECTURE 6-2
Non-Isolated DC/DC Converter
Topologies and Control Techniques

DC/DC Converter Modeling

4-May-24 2
 DC/DC Converter Modeling-1
Introduction
 The objective in converter modeling is to represent the converter
operation in steady-state and during transients by a number of
mathematical equations. Mathematical models are usually used for
controller design.
 Converter models are used in transient analysis of power electronic
systems when exact simulation or implementation of the converter is
time-consuming or is not necessary due to the nature of the study.
 One of the most common modeling techniques in dc-to-dc
converters is the averaging technique.
 In average modeling, one is usually interested in the behavior of the
system on the average-per-switching-period basis. The instantaneous
values of voltages across capacitors and currents through inductors
in the converter are not of interest in average modeling. However, a
proper average model should not miss the variations of capacitor
voltage and inductor current from one switching period to the next.
 The average models derived here do not account for converter
losses, for the sake of simplicity.
 Average model of a converter can be equation-based or circuit-
based. In this course both models will be introduced.

4-May-24 3
 DC/DC Converter Modeling-2
Introduction
 The equation-based average modeling starts with writing first-order
differential equations for inductor currents and capacitor voltages in
different topological modes of the converter, based on circuit
analysis laws (e.g., KVL and KCL). The differential equations
obtained in this way are then averaged over a switching period by
multiplying right-hand-side of each equation by the corresponding
portion of a switching period, adding them up and dividing the sum
by the switching period.
 The average model derived in this way consists of a number of
nonlinear first order differential equations. Linearization around an
operating point can be performed to derive state-space equations and
transfer functions representing the small signal-model of the
converter.
 A controller designed based on a small-signal linear model is
effective around the same operating point used for model
development. For example, the proportional and integral gains of a
PI controller designed based on a small-signal linear model for the
converter are only guaranteed to result in the desired performance
around one operating point. For other operating points, the gains
have to be adjusted or scheduled (gain scheduling).

4-May-24 4
 DC/DC Converter Modeling-3
Introduction
 In circuit-based average modeling, the first-order differential
equations for inductor currents and capacitor voltages do not need to
be considered, as capacitors and inductors are included in circuit
simulations. Instead, the current through or the voltage across the
switching devices (diodes and switches) in different topological
modes are averaged over a switching period. Each average function
obtained is then used to represent the current or voltage of a diode or
switch in the form of a controlled current or voltage source.
 The time step of simulation must be at least one order of magnitude
smaller than the smallest time constant of the circuit. In average
modeling, where the details within one switching period are not of
interest, the time step is not dictated by the switching frequency;
instead, the time step is determined by the values of circuit
parameters (e.g., L, C and R). Therefore, simulation runtime can be
significantly reduced when performing simulations based on average
models.

4-May-24 5
Buck Converter Equation-Based Average Modeling-1
Continuous-Conduction Mode of Operation

 Topological Mode I: Switch ON

 diL 1 1
 dt = − L vC + L Vd
 dv 1 1
(I )
 C = iL − vC
 dt C RC
 This system of first-order differential equations is valid during
ON-state of the switch, ton.

4-May-24 6
Buck Converter Equation-Based Average Modeling-2

 Topological Mode II: Switch OFF

 diL 1
 = − vC
dt L
 dv 1 1
( II )
 C
= iL − vC
 dt C RC

 This system of first-order differential equations is valid during

OFF-state of the switch, toff.

4-May-24 7
Buck Converter Equation-Based Average Modeling-3
 In order to get the average model for buck converter, the first-order
differential equations written for the ON- and OFF-periods of the
switch are averaged over one switching period.
diL 1 di di 
, ave. = ton  L (on − state ) + toff  L (off − state )
dt Ts  dt dt 
di di
= d  L (on − state) + (1 − d )  L (off − state )
dt dt
 di di  di
= d   L (on − state ) − L (off − state) + L (off − state )
 dt dt  dt
1 1
= d  Vd − vC
L L
dvC  dv dv  dv
, ave. = d   C (on − state ) − C (off − state ) + C (off − state )
dt  dt dt  dt
1 1
= iL − vC
C RC

4-May-24 8
Buck Converter Equation-Based Average Modeling-4
 For simplicity of results, let
 diL diL
 dt = dt , ave.
 dv dv
 C = C , ave.
 dt dt
Therefore,  diL 1 1
 dt = − vC + d Vd
L L
 dv 1 1
 C = iL − vC
 dt C RC
This is a system of first-order nonlinear differential equations of the
general form:
d x
 = f ( x, u )  iL  Vd 
 dt where x= ,u = 
 y = g ( x, u ) vC  d 

4-May-24 9
Buck Converter Equation-Based Average Modeling-5
 Let’s linearize the system of nonlinear differential equations
around a steady-state operating point. The perturbed system is
represented by:
d f f
x = x + u
dt x * u *
f f
where and are the Jacobean matrices evaluated at
x *
u *

the steady-state operating point and are given as:

 f1 f1   f1 f1 
 f  iL vC   f  Vd d 
=  , = 
 x  f 2 f 2  u  f 2 f 2 
 iL vC   Vd d 

4-May-24 10
Buck Converter Equation-Based Average Modeling-6
 The elements of the Jacobean matrices can be found as follows:

f1 f1 1
=0 = d
iL Vd L
f1 1 f1 1
=− = Vd
vC L d L
f 2 1 f 2
= =0
iL C Vd
f 2 1 f 2
=− =0
vC RC d

4-May-24 11
Buck Converter Equation-Based Average Modeling-7
Steady-State Operating point
 At steady-state, the average per switching period of capacitor
voltage and inductor current are constant. Therefore, the average
per switching period of rates of change with respect to time of
capacitor voltage and inductor current are equal to zero. Thus,
one can write:
 diL  1 * 1 * *
 dt = 0 − L VC + L d Vd = 0
 
 dvC = 0  1 I* − 1 V * = 0
 dt  C L R*C C
where the superscript ‘*’ denotes steady-state value.
 * VC*
 d = *
 Vd
 Given Vd , VC and R , d and I L can be found as: 
* * * * *

I * = 1 V *
 L R* C
Note that the steady-state operating point could have been obtained by looking
at steady-state equations relating average output to input voltage (VC = dVd ),
and inductor current to output current ( I L = I o = VC R ).

4-May-24 12
Buck Converter Equation-Based Average Modeling-8
 The general form of the state-space representation for the
linearized system is:
d f f
x = x + u
dt x * u *
The state-space representation of the linearized buck converter
model therefore is:
 1 
 0 −  1 * 1 *
d  i  L  i   d Vd   Vd 

L
 = 
L
 + L
  d 
L
dt  vC   1 1   vC  
− *  0 0 
 C 
R C
or:
 1   1 VC* 1 *
0 −
d  iL   L   iL   Vd   Vd 
  =   +  L Vd* L  
dt  vC   1 v   
− *   C   0
1 d
 C 0 
R C 

4-May-24 13
Buck Converter Equation-Based Average Modeling-9
 Let’s now define the matrices in the standard-form state-space
representation of the linearized system.
d x
 = Ax + Bu
 dt
 y = C x + Du

 1   1 VC* 1 *
 0 −
 iL  L   Vd 
x=  ; A=  ; B =  L Vd* L ;
 vC  1 − 1   0
 C 0 
R*C 
 V  0 1  0 0
u =  d;C =   ; and D = 0 0
 d  1 0   

 With the duty ratio-to-output voltage transfer function in mind,

the output of interest is vC and the input of interest is d .

4-May-24 14
Buck Converter Equation-Based Average Modeling-10
Duty Ratio to Output Voltage Transfer Function
 The goal is now to find the desired input-to-output transfer
function of the converter vC / d .
 From linear control theory,
Transfer Function H ( s) = C ( sI − A) −1 B + D
 Using
 1 *
V
B ' =  L d  , C =  0 1 , and D = 0,
' '

 
 0 
−1
 1 
 s   1 *
vC  L Vd 
( s ) = C ( sI − A) B + D =  0 1 
−1 ' L
H (s) =
' ' '

d − 1 1   
s+ *  0 
 C R C 

4-May-24 15
Buck Converter Equation-Based Average Modeling-11
 Inverse of a 22 Matrix: If a b 
M =  ,
c d 
then,
−1 1  d −b 
M =   , where det .( M ) = a  d − b  c
det .( M )  −c a 
Therefore,
−1
 1   1 1
 s  s +
 R*C − 
L 1 L
  =  
1 1  1
− 1 s+ *  s 
1 s(s + * ) +
 C R C  R C LC  C 
and
1  1 *
 C s   1 * Vd
vC V
L  = LC
H ( s) =
'
( s) = d

d 1 1   1 1
s2 + * s +  0  s 2
+ s +
RC LC R*C LC
 This transfer function shows that buck converter, in continuous
conduction mode, is a stable system with two poles in the left-half
s-plane.

4-May-24 16
Buck Converter Equation-Based Average Modeling-12
 The block diagram of the closed-loop system will be as follows:
d*
+
VC ,ref Error d d Buck vC
+
 Controller (e.g.,
PI Controller)  Converter
+
-
VC , fbk
Low-Pass
Filter
K

 d* biases the system at the desired operating point and the loop
takes care of the small variations in d around the steady-state value
to realize the desired output voltage in presence of disturbances,
component value drifts and non-idealities.
 The objective in control loop design (based on a PI controller) will
be finding the proportional and integral gains of the PI controller,
given a subset of performance specifications such as: steady-state
error, settling time, band-width, gain margin, and phase margin. In
order to find the proportional and integral gains of the PI controller
(2 unknowns), two design specifications are required.
 Controller design is based on small-signal linear model developed.

4-May-24 17
Buck Converter Equation-Based Average Modeling-13
Example
 Simulate a buck converter with the following parameters in the
following two ways: (i) using PSIM and (ii) based on the
nonlinear average model developed.
– Vi = 24V
– d = 0.4
– L = 1mH
– C = 400F
– fs = 10kHz
– R = 4

4-May-24 18
Buck Converter Equation-Based Average Modeling-15
Solution
 The circuit setup in PSIM is shown below.

4-May-24 19
Buck Converter Equation-Based Average Modeling-16
Solution (Cont.)
 The simulation results for inductor current and capacitor voltage
are shown below.

4-May-24 20
Buck Converter Equation-Based Average Modeling-17
Solution (Cont.)
 The simulation results obtained from MATLAB based on the
 diL 1 1
 dt = − vC + d Vd
nonlinear average model L L are shown below.
 dv 1 1
 C = iL − vC
 dt C RC (a)
8

6


iL


0
0 0.005 0.01 0.015 0.02 0.025
time (s)
(b)
15


vC 10
❖

0
0 0.005 0.01 0.015 0.02 0.025
time (s)
Note that the results obtained from the average model follow the average per switching
period of the results obtained from detailed simulation.
4-May-24 21
 Buck Converter Circuit-Based Average Modeling-1
 Note that the objective here is to derive the circuit-based average model, not
the equation-based average model of buck converter. As a result, the
relations developed to derive the average model will not include the first-
order differential equations for the inductor current and capacitor voltage, as
these components are integrated with the circuit-based average model for
simulation. The model focuses on the switch current and diode voltage
(indicated by the dashed rectangle) only.
 Given that S has two states (ON and OFF), two topological modes can be
identified for the buck converter.
– Topological Mode 1: In this case, S is ON and D is OFF, resulting in the following
relations: iS = iL iL
iS L io
S
vD = −vi
+ − +
– Topological Mode 2: In this case, S is
vi vD C the vo
OFF and D is ON, resulting in
D
following relations: i = 0 − + −
S

vD = 0
Buck Converter Circuit Diagram

4-May-24 22
Buck Converter Circuit-Based Average Modeling-2
 By averaging the equations for the switch current and diode voltage over
one switching period, one can derive the following relations, where d
is the switch duty cycle, and all variables are average-per-switching-
period quantities: iS S L iL io
iS = diL + (1 − d )( 0 ) = diL + − +
vD = d ( −vi ) + (1 − d )( 0 ) = − dvi vi vD
D
C vo
− + −

Buck Converter Circuit Diagram

iS iL L io

+ +
vi vo
− C −
Buck Converter Circuit-Based Average Model
Note that instead of switch current and diode voltage, switch voltage and diode current could
have been use to derive the average mode. The average model would be different in appearance,
but the results would be the same.
4-May-24 23
 Example
 Simulate a buck converter with the following parameters in the
following two ways: (i) using PSIM, and (iii) based on the circuit-
based average model.
– Vi = 24V
– d = 0.4
– L = 1mH
– C = 400F
– fs = 10kHz
– R = 4

4-May-24 24
 Solution
 The circuit setup in PSIM is shown below.

4-May-24 25
 Solution (Cont.)
 The PSIM simulation results for inductor current and capacitor voltage
are shown below. (Detailed Simulation)

Simulation time step: 1

4-May-24 26
 Simulation Results (Cont.)
 The PSIM simulation results for inductor current and capacitor voltage,
based on the buck converter circuit-based average model and the
parameters used for detailed simulation, are shown below.

Simulation time step: 100

- Note that the circuit-based average model provides the same information as obtained
from PSIM detailed simulation on an average-per-switching–period basis.
- A larger simulation time step (100 times larger than that used in detailed simulation)
implies a much shorter simulation time, which is the main advantage for average
modeling.
4-May-24 27