IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 67, NO.

9, SEPTEMBER 2019 3603

A Generalized Approach for Multifrequency

Transmission Line Transformer With
Frequency-Dependent Complex
Source and Load
Liang Liu , Ronghong Jin , Fellow, IEEE, Xianling Liang , Senior Member, IEEE,
Haijun Fan, Member, IEEE, Xudong Bai, Member, IEEE, Han Zhou ,
Junping Geng , Senior Member, IEEE, and
Weiren Zhu , Senior Member, IEEE

Abstract— In this paper, a generalized approach to design a A quarter-wavelength impedance transformer is ubiquitous
multifrequency transformer with frequency-dependent complex in the design of microwave components with a single band.
source and load is proposed. By introducing an equivalent trans- Although multifrequency components can be realized based
formation between the perfectly matched complex-to-complex
(a complex source impedance transformed to another complex on it together with power divider/combiner [4] or radio
load impedance) and unmatched real-to-real transformers (a real frequency (RF) switch [5], it is still difficult to fulfill the
source impedance transformed to another real load impedance), increasing needs of multifrequency applications [6]. The rea-
the modified small reflection theory once limited to the design son for this is that they usually suffer from bulk size, high
of the real-to-real transformer is further extended to that of the cost, and low efficiency because of the insertion losses of
complex-to-complex transformer with more than two operating
bands. Moreover, it is very useful to the design of multifrequency the introduced power dividers/combiners or RF switches.
cascaded power amplifiers or low-noise amplifiers with compact In order to overcome the above drawbacks, a dual-frequency
size since its forward-stage output can be directly matched transformer with the frequency ratio of 2: 1 [7], [8] and
to the next-stage input. To validate the presented method, a dual-frequency [9], [10] and triple-frequency [11] trans-
several multifrequency transformers are designed, simulated, and former with arbitrary frequency ratio as well as impedance
measured. Good agreement among the calculated, simulated, and
measured results is achieved. transform ratio were reported. They are characterized by the
structure of multisection transmission lines and the analysis
Index Terms— Equivalent transformation, frequency- of transmission line theory or ABCD matrix method. As the
dependent complex impedance, modified small reflection
theory (MSRT), multifrequency transformer. corresponding dominated equations are usually several tran-
scendental equations, it is difficult for the aforementioned
I. I NTRODUCTION method to be further extended to a transformer with more

S INCE the last two decades, transformers matched simul-

taneously at multiple frequencies have been extensively
investigated, which are widely used in various multifre-
than triple frequencies and should resort to the method of
particle swarm optimization [12]. In order to provide a good
initial value for the optimization or obtain an approximate
quency microwave components to obtain a maximum power solution, an N-frequency transformer based on the structure
output [1]–[3]. of 2N-section transmission lines and the method of modified
small reflection theory (MSRT) was proposed [13]. Its domi-
Manuscript received December 18, 2018; revised April 19, 2019; accepted
June 23, 2019. Date of publication July 24, 2019; date of current version nated equations are reduced to a single linear matrix equation
September 4, 2019. This work was supported in part by the National Natural after variable substitution and proper selection of reference
Science Foundation under Grant 61571298, Grant 61571289, Grant 61671416, frequency. In order to realize the N-frequency transformer
and Grant 61701303, in part by the Natural Science Foundation of Shanghai
under Grant 17ZR1414300, in part by the Shanghai Pujiang Program under with an arbitrary frequency and real impedance transformer
Grant 17PJ1404100, and in part by the National Equipment Pre-Research ratio, m-section additional transmission lines are introduced to
Field Foundation of the 13th Five-Year Plan under Grant 6140135010103. fulfill the requirement of small reflection condition, and electri-
(Corresponding author: Xianling Liang.)
L. Liu, R. Jin, X. Liang, H. Zhou, J. Geng, and W. Zhu are with cal parameters can be directly derived from a mapping of small
the Department of Electronic Engineering, Shanghai Jiao Tong University, reflection coefficients between the 2N- and 2N + m-section
Shanghai 200240, China (e-mail: liangxl@[Link]). transmission line transformers [14].
H. Fan is with the Institute of Sensors Signals and Systems, Heriot-Watt
University, Edinburgh EH14 4AS, U.K. Recently, many researchers have increasingly been focus-
X. Bai is with the China Academy of Aerospace Electronics Technology, ing on the transformer with frequency-dependent complex
Shanghai Aerospace Electronics Co., Ltd., Shanghai 201821, China. impedance for applications, such as a power ampli-
Color versions of one or more of the figures in this article are available
online at [Link] fier [15]–[19] and an RF transceiver [20]. Dual-frequency
Digital Object Identifier 10.1109/TMTT.2019.2926250 transformers with fixed complex source and load were
0018-9480 © 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See [Link] for more information.

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
3604 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 67, NO. 9, SEPTEMBER 2019

proposed [21], [22], which are based on the structure of two-

section transmission lines. Later, a dual-frequency transformer
with fixed real source but frequency-dependent complex
load was also reported, which is realized by three- [23]
or two-section transmission lines together with several shunt
stubs [24], [25]. In [26] and [27], dual-frequency transform-
ers based on quad-section transmission lines were presented
further, and their source and load impedance both were
frequency-dependent complex impedances. One feature of the Fig. 1. (a) N -frequency perfectly matched real-to-complex transformer with
above dual-transformers with frequency-dependent complex the source of Z 0 and the frequency-dependent load of Z L (θ ). (b) Equivalent
unmatched real-to-real transformer with the virtual load of Z instead of the
load is that although their structures and analyses are sim- complex load of Z L (θ ).
ilar to those of the real-to-real (a real source impedance
transformed to another real load impedance) transformer with
fixed impedance transform ratio, their dominated equations A. Equivalent Transformation Between Real-to-Complex and
are much more complicated. The reason for this is that Real-to-Real Transformers
their impedance symmetry at two matching frequencies is 1) Virtual Load: As shown in Fig. 1(a), an N-frequency
no longer valid due to the frequency-dependent complex perfectly matched real-to-complex transformer with the fre-
load. Therefore, compared with the real-to-real transformer, quency ratio of f 0 : n 1 f 0 : · · · : n N−1 f 0 is composed
a method with simplified dominated equations extending to of 2N-section transmission lines with the electrical length of θ
the transformer with more operating frequencies as well as and the characteristic impedances of Z 1 ∼ Z 2N , respectively.
frequency-dependent complex source and load is in higher The 2N characteristic impedances are used to independently
desire. control the amplitude and phase of impedance transform ratio
In this paper, the method of MSRT, characterized by of a transformer at N matching frequencies. For analysis,
simple dominated equation but limited to the design of the source is assumed as a fixed real impedance of Z 0 , while
multifrequency real-to-real transformer, is utilized to fur- the load is supposed to be a frequency-dependent complex
ther design the multifrequency complex-to-complex (a com- impedance of Z L (θ ) [Z L (θ = 0) = Z L = R L + j X L at zero
plex source impedance transformed to another complex load frequency]. Based on the MSRT, the total reflection coefficient
impedance) transformer. In order to make all solved character- of (θ ) can be expressed in terms of the electrical length θ as
istic impedances of transmission lines real in MSRT, the per- well as the modified small reflection coefficients 
n [13], [14]
fectly matched complex-to-complex transformer is equivalent
2N 
 
to an unmatched real-to-real transformer with virtual load.
(θ ) = n · e− j 2nθ (1)
Moreover, in order to be applicable to the transformer
with arbitrary impedance transform ratio, cascaded sub- n=0

transformers based on an equal impedance transform ratio are 

n−1
 

0 = 0 
n = n · 1 − l2 , n = 1, 2, . . . , 2N (2)
introduced further.
l=0
II. E QUIVALENT T RANSFORMATIONS BASED ON MSRT where n is the small reflection coefficient and is defined by
It will be shown that the load of a transformer should ⎧
⎪ Z n+1 − Z L
be real to fulfill the requirement of real characteristic ⎪
⎪ , n=0
⎪
⎪ Z n+1 + Z L
impedances of transmission lines in MSRT, which, in practice, ⎨Z
n+1 − Z n
places demands on the equivalent transformation between the n = , n = 1, 2, . . . , 2N − 1 (3)
⎪
⎪ Z n+1 + Z n
complex-to-complex transformer and the transformer with real ⎪
⎪ Z − Zn
⎪
⎩ 0 , n = 2N.
load. On the other hand, as design approach for the real-to-
Z0 + Zn
real transformer with fixed impedance transform ratio has been
well developed [13], [14], it is expected that the complex- On the other hand, according to the equation for the conversion
to-complex transformer with frequency-dependent complex between S and Z parameters with complex source and load of
source and load can also be solved by using an equivalent a two-port network [28], one-port reflection coefficient of (θ )
transformation between the complex-to-complex and real-to- can also be expressed in terms of the input impedance and the
real transformers. To achieve this goal, a complex-to-complex load by letting Z 12 = Z 21 = 0, that is
transformer is first equivalent to a real-to-complex (a real Z in (θ ) − Z ∗L (θ )
source impedance transformed to a complex load impedance) (θ ) = (4)
transformer under the small reflection condition. Subsequently, Z in (θ ) + Z L (θ )
the obtained real-to-complex transformer is equivalent to an where the asterisk, *, denotes the complex conjugate operation.
unmatched real-to-real transformer. Since the equivalent trans- At zero frequency, as the electrical length of θ is equal to
formation between complex-to-complex and real-to-complex zero, the input impedance of Z in (θ = 0) is reduced to Z 0 .
transformers involves that between real-to-complex and real- It is also noted that the formula of (1) is valid for arbitrary θ .
to-real transformers, it will be discussed after the latter equiv- Then, letting θ = 0 in (1), substituting Z in (θ = 0) = Z 0
alent transformation. into (4), and equalizing (1) and (4), one obtains the following

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: GENERALIZED APPROACH FOR MULTIFREQUENCY TRANSMISSION LINE TRANSFORMER 3605

zero-frequency condition:

2N
Z 0 − Z ∗L

n = . (5)
Z0 + Z L
n=0

The zero-frequency condition in MSRT acting as a bound-

ary or initial condition must be satisfied since it determines the Fig. 2. (a) Single-section perfectly matched real-to-complex transformer.
(b) Equivalent single-section unmatched real-to-real transformer.
impedance transform ratio of a transformer. For simplification,
n−1
one can neglect the penetration coefficient of l=0 (1 − l2 ) TABLE I
in 
n since the small reflection coefficient of l is usually E LECTRICAL PARAMETERS OF R EAL - TO -C OMPLEX AND
small. Namely,  n is reduced to n . Then, substituting it and R EAL - TO -R EAL T RANSFORMERS C OMPOSED OF AN
I DENTICAL S INGLE -S ECTION T RANSMISSION L INE
Z L = R L + j X L into (5), and incorporating (3) with (5), one
has

2N
Z 0 − RL + j X L Z 1 − RL − j X L
n = − (6)
Z 0 + RL + j X L Z 1 + RL + j X L
n=1
its small reflection coefficients are real due to the virtual
where R L , X L , Z 0 , and Z 1 are real. In addition, according load of Z .
to (3), n (n ∈ 1 ∼ 2N) is also real when all the solved For the generalized case of the N-frequency real-to-complex
characteristic impedances of transmission lines are real. Then, and equivalent unmatched real-to-real transformers shown
separating (6) into real and imaginary parts, after some alge- in Fig. 1, as the transformer shown in Fig. 1(a) is perfectly
braic computations, one gets matched, one obtains Z in (θi ) = Z ∗L (θi ) (θi is the electrical
length at the matching frequency of n i−1 f 0 ). Subsequently,

2N
Z 02 − R 2L + X 2L Z 12 − R 2L − X 2L
n = − (7a) the input impedance of Z in (θi ) shown in Fig. 1(b) is also equal
n=1
(Z 0 + R L )2 + X 2L (Z 1 + R L )2 + X 2L to Z ∗L (θi ), and the reason for this is that both two transformers
  shown in Fig. 1(a) and (b) as well as their source are the
RL Z1
0 = 2X L + . (7b) same. Then, the total reflection coefficient of (θi ) shown
(Z 0 + R L ) + X L (Z 1 + R L )2 + X 2L
2 2
in Fig. 2(b) is given by
It can be easily concluded that the formula of (7b) is only Z in (θi ) − Z
valid when X L is equal to zero (R L > 0, Z 1 > 0 ) for the case (θi ) =
Z in (θi ) + Z
of real n (n ∈ 1 ∼ 2N). Namely, the zero-frequency condition Z ∗ (θi ) − Z
with real small reflection coefficients of 1 ∼ 2N can only = ∗L
Z (θi ) + Z
satisfy that the load impedance is real. Therefore, in order L 
to utilize MSRT in a design of transformer with fixed real Z L (θi ) − Z ∗
=
source and frequency-dependent complex load, a virtual load Z L (θi ) + Z
defined by virtual fixed real impedance should be introduced = |γ (θi )|e− j ϕ(θi ) (8)
instead of its actual complex load to obtain real characteristic
impedances of transmission lines. where γ (θi ) = |γ (θi )|e j ϕ(θi ) = [R L (θi ) + j X L (θi ) −
2) Dominated Condition of Equivalent Unmatched Real-to- Z ]/[R L (θi ) + j X L (θi ) + Z ]. Incorporating (1) with (8), one
Real Transformer: For a given transformer with fixed source, obtains the following reflection coefficient dominated condi-
it is known that its reflection coefficient varies with its load. tion of an equivalent unmatched real-to-real transformer:
Therefore, when a perfectly matched real-to-complex trans- 
2N
Z ∗ (θi ) − Z
former shown in Fig. 1(a) is given, there exists an equivalent (
n · e− j 2nθi ) = ∗L = |γ (θi )|e− j ϕ(θi ) (9)
unmatched real-to-real transformer as shown in Fig. 1(b) with Z L (θi ) + Z
n=0
an arbitrary selected virtual load of Z and an accordingly where the involved n of an equivalent unmatched real-to-real
changed reflection coefficient. transformer with virtual load of Z is defined by
To better understand the equivalent transformation between ⎧
the perfectly matched real-to-complex and unmatched real- ⎪ Z n+1 − Z
⎪
⎪ , n=0
to-real transformers, two single-section transformers shown ⎪ Z n+1 + Z
⎪
⎨Z
n+1 − Z n
in Fig. 2 are taken as an example. Although the loads n = , n = 1, 2, . . . , 2N − 1 (10)
and reflection coefficients of the two transformers listed ⎪ Z n+1 + Z n
⎪
⎪
⎪ Z − Zn
in Table I are different, an identical single-section trans- ⎪
⎩ 0 , n = 2N.
mission line with the characteristic impedance of Z c and Z0 + Zn
electrical length of θ listed in the last two columns can be It is noted that (9) only involves the virtual load of Z and
obtained according to the transmission line theory. There- the electrical parameters of the real-to-complex transformer
fore, in order to design a perfectly matched real-to-complex at the i th matching frequency, and thus, it can be extended
transformer shown in Fig. 2(a), an equivalent unmatched real- to N matching frequencies and N dominated equations for
to-real transformer shown in Fig. 2(b) can be utilized, and the unmatched real-to-real transformer can then be obtained.

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
3606 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 67, NO. 9, SEPTEMBER 2019

Fig. 3. (a) Single-section perfectly matched complex-to-complex transformer.

(b) Equivalent single-section perfectly matched real-to-complex transformer.

TABLE II
E LECTRICAL PARAMETERS OF P ERFECTLY M ATCHED C OMPLEX - TO -
C OMPLEX AND R EAL - TO -C OMPLEX T RANSFORMERS

In addition, the zero-frequency condition of the equivalent

unmatched real-to-real transformer is altered to

2N
Z0 − Z

n = γ (0) = . (11) Fig. 4. (a) Perfectly matched complex-to-complex transformer. (b) Equivalent
Z0 + Z unmatched real-to-real transformer of a perfectly matched real-to-complex
n=0
transformer. (c) Transformer with the source of Z 0 α(θi )e j φ(θi ) and the virtual
In summary, in order to design an N-frequency per- load of Z . (d) Equivalent transformer of Fig. 4(c).
fectly matched real-to-complex transformer based on MSRT,
an equivalent unmatched N-frequency real-to-real transformer Z c and the electrical length of θ . Therefore, in a design of
with dominated equations of (9) and (11) can be utilized, complex-to-complex transformer shown in Fig. 3(a), an equiv-
where the small reflection coefficients and characteristic alent real-to-complex transformer shown in Fig. 3(b) can be
impedances of transmission lines are real. It should be pointed utilized to simplify its design.
out that either Z 0 or arbitrary real impedance among the load For a generalized N-frequency perfectly matched complex-
norms of all matching frequencies can be selected as the virtual to-complex transformer with the source of Z 0 α(θi )e j φ(θi ) and
load of Z , which has a little impact on the derived electrical the load of Z l (θi ), as shown in Fig. 4(a), we suppose that
parameters of the transformer. The reason for this is that the source of its equivalent perfectly matched real-to-complex
identical frequency-dependent complex loads Z L (θi ) of the transformer is an arbitrarily selected real impedance of Z 0 ,
real-to-complex transformer at the N matching frequencies can while the load of that is a frequency-dependent complex
be obtained from (8) although the selected virtual loads and impedance of  Z l (θi ) to be solved. In order to represent the
reflection coefficients are different. In practice, the identical reflection coefficient utilizing MSRT, it can be further equiva-
N complex equations (i.e., 2N real equations) of frequency- lent to an unmatched real-to-real transformer with the source
dependent complex load Z L (θi ) fully determine an identical of Z 0 and the virtual load of Z , as shown in Fig. 4(b). The
transformer with 2N unsolved parameters of Z 1 ∼ Z 2N . following task is to derive  Z l (θi ) according to the dominated
In this paper, Z 0 or geometrical-mean norm of loads at all equations of two transformers shown in Fig. 4(a) and (b).
matching frequencies is selected as Z . To achieve this goal, one considers another transformer with
the source of Z 0 α(θi )e j φ(θi ) and the virtual load of Z ,
B. Equivalent Transformation Between the Complex-to- as shown in Fig. 4(c).
Complex and Real-to-Complex Transformers First, one obtains the reflection coefficient of   (θi ) accord-
ing to the dominated condition of the complex-to-complex
For a given transformer, it is known that its load impedance
transformer shown in Fig. 4(a). As it is perfectly matched,
varies with its source impedance. Therefore, when a perfectly  (θ ) = Z ∗ (θ ) is valid. It is also noted that as both sources
Z ina i l i
matched complex-to-complex transformer is given, generally,
and transformers shown in Fig. 4(a) and (c) are identical, one
there exists an equivalent real-to-complex transformer with an  (θ ) =Z  (θ ). Consequently, it is followed that:
gets Z inc i ina i
arbitrarily selected real source and an accordingly changed

complex load. Z inc (θi ) = Z l∗ (θi ). (12)
Similarly, in order to clearly demonstrate the basic idea
of equivalent transformation between the complex-to-complex Then, the reflection coefficient of   (θi ) in Fig. 4(c) can be
and real-to-complex transformers, two single-section trans- expressed in terms of the virtual load of Z and the derived
 (θ ), namely
input impedance of Z inc
formers shown in Fig. 3 are taken as an example, whose i
electrical parameters are listed in Table II. According to the  (θ ) − Z
Z inc Z l∗ (θi ) − Z
  (θi ) =
i
transmission line theory, it is readily verified that although  (θ ) + Z = . (13)
Z inc i Z l∗ (θi ) + Z
the sources and loads of the two transformers are different,
both of them are perfectly matched by an identical single- On the other hand, one can also get the reflection coef-
section transmission line with the characteristic impedance of ficient of   (θi ) according to the dominated condition of

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: GENERALIZED APPROACH FOR MULTIFREQUENCY TRANSMISSION LINE TRANSFORMER 3607

the unmatched real-to-real transformer shown in Fig. 4(b). Especially, when the imaginary part of Z l (θi ) is not large
As it is equivalent to the perfectly matched real-to-complex compared with its real part, the virtual load of Z can be
transformer with the source of Z 0 and the load of  Z l (θi ), approximately equal to Z l∗ (θi ), and then, the formula of (20)
according to the reflection coefficient dominated condition is reduced to
of (9), one has e j φ(θi ) − j α(θi ) · tan(2Nθi )

Z l (θi ) = Z l (θi ) · . (21)

2N−1
Z 0 − Z 2N − j 4Nθi 
Z ∗ (θi ) − Z α(θi ) − j e j φ(θi ) · tan(2Nθi )
(
n e− j 2nθi ) + ·e = l∗ .
Z 0 + Z 2N 
Z l (θi ) + Z It should be pointed out that both (20) and (21) merely
n=0
(14) involve the electrical parameters of the i th matching frequency.
Therefore, they can be independently utilized at all matching
It is noted that the transformer shown in Fig. 4(c) is not a frequencies of the N-frequency complex-to-complex trans-
real-to-real transformer, and its equivalent transformer shown former. In order to obtain an equivalent N-frequency real-
in Fig. 4(d) is also considered to represent reflection coefficient to-complex transformer with fixed source of Z 0 , the source
using MSRT. It utilizes a virtual real impedance of  Z and of equivalent transformation at every matching frequency is
an extended electrical length θi of the 2Nth transmission line selected to be an arbitrarily selected identical impedance of Z 0 .
instead of Z 0 α(θi )e j φ(θi ) shown in Fig. 4(c). Then, one obtains In this paper, it is selected to be the geometric-mean norm of
the following identical reflection coefficient of (θi ) shown sources at all matching frequencies, namely,
in Fig. 4(c) and (d): 
N
 
Z − Z 2N − j 2θ  Z 0 α(θi )e j φ(θi ) − Z 2N Z0 =  N
|Z s (θi )|
(θi ) = e i = . (15) (22a)

Z + Z 2N Z 0 α(θi )e j φ(θi ) + Z 2N i=1
As the transformer shown in Fig. 4(d) is a real-to-real trans- |Z s (θi )|
α(θi ) = (22b)
former and the two transformers shown in Fig. 4(c) and (d) are Z0 
equivalent, one gets   (θi ) shown in Fig. 4(c) and (d) based φ(θi ) = arg Z s (θi ) . (22c)
on the MSRT
In (22), Z s (θi ) is the source of a complex-to-complex

2N−1 
Z − Z 2N − j (4Nθi +2θ  )
  (θi ) = (
n e− j 2nθi ) + e i . (16) transformer at the i th matching frequencies, which is a given

Z + Z 2N initial parameter. Then, α(θi ) and φ(θi ) can be derived. It is
n=0
also noted that both Z l (θi ) and N are given initial parameters,
Substituting (15) into (16), it is followed that:
and θi can be got when the reference frequency is selected

2N−1
Z 0 α(θi )e j φ(θi ) − Z 2N − j 4Nθi (discussed in Section III-B). Therefore, both the load and
  (θi ) = (
n e− j 2nθi ) + ·e . source of the equivalent N-frequency real-to-complex trans-
Z 0 α(θi )e j φ(θi ) + Z 2N
n=0 former can be fully determined according to (20) or (21)
(17) and (22).
From (14) and (17), one has
  III. D ESIGN OF N -F REQUENCY C OMPLEX - TO -
2Z 0 Z 2N · e− j 4Nθi · α(θi )e j φ(θi ) − 1 
Z ∗ (θi ) − Z
   + l∗
 (θi ) = C OMPLEX T RANSFORMER BASED ON
(Z 0 + Z 2N ) · Z 0 α(θi )e j φ(θi ) + Z 2N 
Z l (θi ) + Z E QUIVALENT T RANSFORMATIONS
α(θ )e j φ(θi ) − 1 
Z ∗ (θ i ) − Z
≈ e− j 4Nθi ·
i In Section II, two equivalent transformations under the
+ l∗ (18)
α(θi )e j φ(θi ) + 1 
Z l (θi ) + Z small reflection condition have been introduced. Based
where Z 2N ≈ Z 0 in the denominator is utilized, which is on them, an N-frequency complex-to-complex transformer
usually well approximated due to the small refection condition. with arbitrary frequency-dependent source and load will be
Incorporating (13) and (18), one has the following equivalent discussed.
transformation between the complex-to-complex and real-to-
complex transformers: A. Cascaded Sub-Transformers Based on Equal Impedance
 
2Z · Z l∗ (θi ) − 
Z l∗ (θi ) − j 4Nθi α(θi )e
j φ(θi ) − 1 Transform Ratio
 ∗    = e · .
Z l (θi ) + Z · Z +  Z l∗ (θi ) α(θi )e j φ(θi ) + 1 In order to apply the introduced equivalent transformations
(19) to a design of complex-to-complex transformer, the small
reflection condition should be satisfied, which indicates that
After some algebraic manipulations, the formula of (19) can the impedance transform ratio of the transformer under con-
be rewritten to sideration should be sufficiently small. To achieve this goal,
 2Z l (θi )(θi ) + [Z l (θi ) + Z ] · e j 4Nθi the complex-to-complex transformer with arbitrary frequency-
Z l (θi ) = Z · (20) dependent source of Z S (θi ) and load of Z L (θi ) can be realized
2Z (θi ) − [Z l (θi ) + Z ] · e j 4Nθi
by m + 1 (m = 0, 1, . . .) cascaded sub-transformers, as shown
where in Fig. 5(a). These sub-transformers are characterized by the
1 + α(θi )e− j φ(θi ) equal impedance transform ratio of R(θ ), and consist of
(θi ) = .
1 − α(θi )e− j φ(θi ) 2N-section transmission lines shown in Fig. 4(a).

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
3608 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 67, NO. 9, SEPTEMBER 2019

sake of analysis, the characteristic impedances of involved

2N(k − 1)-section transmission lines are renumbered as Z 1 ∼

Z 2N(k−1) , as shown in Fig. 5(b). Meanwhile, the corresponding
 (θ ) ∼ Z 
input impedances are assumed to be Z l1 i l[2N(k−1)] (θi ).
Then, the load of the kth sub-transformers Z l(k−1) (θi ) is given
by the following recursive formulas:

Z l0 (θi ) = Z L (θi ) (26a)

Z l(h−1) (θi ) + j Z h · tan θi

Z lh (θi ) = Z h ·   ,
Z h + j Z l(h−1) (θi ) · tan θi
h ∈ [1, 2N(k − 1)] (26b)

Z l(k−1) (θi ) = Z l[2N(k−1)] (θi ). (26c)

Fig. 5. (a) Transformer comprising m + 1 cascaded sub-transformers. As the source of the last sub-transformer is Z S (θi ), it is
(b) Renumbered transmission lines with the characteristic impedance of
 
an accurate and given initial parameter. On the other hand,
Z 1 ∼ Z 2N (k−1) .
the load of the last sub-transformer derived from (26) is
also accurate. Therefore, the approximation error of the
For analysis, the source of the kth sub-transformer is sup- whole transformer, in practice, is only resulted from the last
posed to be Z lk (θi ) (k ∈1∼ m + 1). Then, the load of the kth sub-transformer.
sub-transformer is Z l(k−1) (θi ) due to the cascaded structure. It should be pointed out that one sub-transformer is usually
Especially, the source of the m+1th sub-transformer is Z S (θi ), sufficient for the design of most N-frequency complex-to-
while the load of the first sub-transformer is Z L (θi ). According complex transformers. Nevertheless, for the transformer with
to the impedance transform ratio of every sub-transformer, one large impedance transform ratio, more sub-transformers should
obtains be introduced. Fortunately, due to the limited sub-transformers,
Z lm (θi ) Z l(k−1) (θi ) Z L (θi ) their number can be readily obtained by trial-and-error.
R(θi ) = = ··· = = ··· = .
Z S (θi ) Z lk (θi ) Z l1 (θi )
(23) B. Design of Complex-to-Complex Sub-Transformer
1) Selection of Reference Frequency: For analysis, we sup-
From (23), one has pose that the frequency ratio of the N-frequency transformer
R(θi ) = |R(θi )|e j (θi ) is f 0 : n 1 f0 : · · · : n N−1 f 0 and define r f 0 as a reference
 m  frequency corresponding to the electrical length of π/2. Then,

 Z lm (θi )  Z l(k−1) (θi ) Z L (θi )

= m+1 · similar to [13] and [14], in order to avoid the introduction
Z S (θi ) Z lk (θi ) Z l1 (θi ) of additional undesired matching frequencies in the operating
k=2
 frequency range of f 0 ∼ n N−1 f 0 , r f 0 should be larger than
Z L (θi ) the highest matching frequency of n N−1 f0 . On the other
= m+1 . (24)
Z S (θi ) hand, the upper limit of reference frequency can be evaluated
by the electrical length of a dual-frequency transformer with
According to (24), when m is large, the impedance transform
its lowest two matching frequencies. Then, r is given by [26]
ratio of every sub-transformer can be sufficiently small to ful-
fill the small reflection condition. Incorporating (23) with (24), N(1 + n 1 )π
n N−1 < r ≤ (27)
the source of each sub-transformer is given by θ A + θ D + 2π
k m+1−k where θ A , and θ D are defined in (28), as shown at the bottom
Z lk (θi ) = Z S (θi ) m+1 · Z L (θi ) m+1 . (25)
of the next page. In (28), n A and n D are positive integers to
In theory, the load of each sub-transformer can also be achieve compact size. R S (θ1 )+ j X S (θ1 ) and R S (θ2 )+ j X S (θ2 )
obtained from (25) due to the cascaded structure of the are the source impedances at f 0 and n 1 f0 , while R L (θ1 ) +
transformer. However, considering that MSRT in the design of j X L (θ1 ) and R L (θ2 ) + j X L (θ2 ) are the load impedances at
a sub-transformer is an approximate method, generally, minor f 0 and n 1 f 0 .
approximation error will be introduced, which results in a 2) Dominated Equations of Complex-to-Complex Sub-
mismatch of each designed sub-transformer. It is also noted Transformer Based on Equivalent Transformations: In
that the mismatch of the forward-stage sub-transformer has Section III-A, the N-frequency complex-to-complex sub-
an impact on the next-stage sub-transformer. Consequently, transformers under the small reflection condition have been
the matching of whole transformer may be not very good. constructed, where source and load are dominated by (25)
In order to achieve good impedance matching of the and (26), respectively. Based on the two equivalent transforma-
whole transformer, the actual load of each sub-transformer tions of Section II, every complex-to-complex sub-transformer
should be adopted in its design. The actual load of the can be equivalent to the perfectly matched real-to-complex
first sub-transformer is Z L (θi ), while that of the kth sub- sub-transformer as well as the unmatched real-to-real sub-
transformer can be obtained from the electrical parameters transformer shown in Fig. 4(b). Then, the reflection coeffi-
of the forward-stage k − 1 derived sub-transformers. For the cient dominated condition of (9) should be satisfied at all

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: GENERALIZED APPROACH FOR MULTIFREQUENCY TRANSMISSION LINE TRANSFORMER 3609

matching frequencies of the N-frequency complex-to-complex

sub-transformer.
According to the relationship between frequency and elec-
trical length, the electrical length at the i th matching frequency
is given by
1 π n i−1 π
θi = · , i =1 θi = · , i = 2, 3, . . . , N.
r 2 r 2
(29)
Substituting (29) into (9) and separating the derived N for-
mulas into real and imaginary parts, 2N equations can be
obtained. In addition, the zero-frequency condition of (11)
should also be satisfied, and thus, one has the following linear
matrix dominated equation:
Fig. 6. Design flowchart of the N -frequency complex-to-complex transformer
=B
A· (30) with arbitrary frequency-dependent source and load.

where A, ,  and B are defined by

and (2), respectively. Finally, the characteristic impedance of
⎛ ⎞
1 cos(2θ1 ) cos(4θ1 ) · · · cos[(4N −2)θ1 ] cos(4Nθ1 ) all 2N-section transmission lines is given by
⎜0 sin(2θ1 ) sin(4θ1 ) · · · sin[(4N −2)θ1 ] sin(4Nθ1 ) ⎟ ⎧
⎜ ⎟ ⎪ 1 + n
⎜· · · ··· ··· ··· ··· ··· ⎟ ⎪
⎨Z · , n=0
A=⎜⎜1 cos(2θ N ) cos(4θ N ) · · · cos[(4N −2)θ N ]
⎟ 1 − n
⎜ cos(4Nθ N )⎟
⎟ Z n+1 =
1 + n
(31)
⎝0 sin(2θ N ) sin(4θ N ) · · · sin[(4N −2)θ N ] ⎪
⎪
sin(4Nθ N )⎠ ⎩Zn · , n = 1, 2, . . . , 2N − 1.
1 1 1 ··· 1 1 1 − n
⎛ ⎞

0 When all sub-transformers are designed similarly and cas-
⎜  ⎟ caded in order, the desired transformer is achieved. In order
⎜ 1 ⎟
⎜ ··· ⎟ to clearly demonstrate the design process of an N-frequency
 ⎜
= ⎟
⎜ ⎟ complex-to-complex transformer with arbitrary frequency-
⎜ 2N−2 ⎟
⎝ 2N−1 ⎠ dependent source and load, Fig. 6 shows its design flowchart.

2N 3) Parameters Analyses of Sub-Transformer: According
⎛ ⎞ to (29)–(31), it is readily concluded that the parameters of θi ,
|γ (θ1 )| cos[ϕ(θ1 )]  and Z n vary with r , which indicates that the final circuit
A, ,
⎜ |γ (θ1 )| sin[ϕ(θ1)] ⎟
⎜ ⎟ parameters, including the electrical length and characteristic
⎜ ··· ⎟
B=⎜⎜ ⎟. impedances, are affected by the selection of r . However, for
|γ (θ )| cos[ϕ(θ )]) ⎟
⎜ N N ⎟ different selections of r in (27), a well-matching of the trans-
⎝ |γ (θ N )| sin[ϕ(θ N )] ⎠
formers based on the corresponding final circuit parameters
(Z 0 − Z )/(Z 0 + Z ) can be achieved at the designated matching frequencies since
In (30), as N is a given initial parameter and θi (i = 1 ∼ N) an identical matching condition represented by B of (30) is
is derived from (29), A can be obtained. On the other hand, applied. It can be interpreted as nonuniqueness of the solved
the load and source of the equivalent real-to-complex sub- transformers.
transformer can be derived from (20) or (21) and (22) accord- On the other hand, according to (9), the rapid variation of
ing to those of the complex-to-complex sub-transformer, which source and load impedances is fully reflected by that of
are determined by (25) and (26). Thus, both |γ (θi )| and ϕ(θi ) the total reflection coefficient. Considering that the norm
can be got. Subsequently, B can also be obtained. Then, both of derivative of every reflection coefficient term  n e− j 2nθi

n and n can be derived according to the formulas of (30) is 4πn(μ) |1/2 n |L (μ and  are the permeability and the

   
Z A R S (θ1 ) − R S (θ2 )
θA = n A π + arctan (28a)
R S (θ1 )X S (θ2 ) − R S (θ2 )X S (θ1 )
   
Z D R L (θ1 ) − R L (θ2 )
θD = n D π + arctan (28b)
R L (θ1 )X L (θ2 ) − R L (θ2 )X L (θ1 )

X S (θ1 ) + X S (θ2 )  
ZA = R S (θ1 )R S (θ2 ) + X S (θ1 )X S (θ2 ) + · R S (θ1 )X S (θ2 ) − R S (θ2 )X S (θ1 ) (28c)
R S (θ2 ) − R S (θ1 )

X L (θ1 ) + X L (θ2 )  
ZD = R L (θ1 )R L (θ2 ) + X L (θ1 )X L (θ2 ) + · R L (θ1 )X L (θ2 ) − R L (θ2 )X L (θ1 ) (28d)
R L (θ2 ) − R L (θ1 )

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
3610 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 67, NO. 9, SEPTEMBER 2019

permittivity, respectively, while L is the physical length of From (35), it is seen that the upper limit of the introduced
every section transmission line), it is readily concluded that the approximation error increases with the norm of the small
rapid variation of source and load impedances mainly results reflection coefficient |n |. Particularly, when |n | approaches
in the characteristic impedance variation of the transmission to zero, the approximation error of small reflection coefficient
lines with large order number, but it does not mean that their is close to zero, which indicates that both the derived A and
variations are large due to the coefficient of 4πn. solution of (30) are accurate.
It should be pointed out that as the attention of the proposed 2) Relationship Between Small Reflection Coefficients and
method based on the perfect matching is focused on the match- Impedance Transform Ratio: Under the small reflection condi-
ing frequencies rather than the bandwidth, its bandwidth of tion, it is easily shown that the impedance transform ratio of a
every designed frequency may be narrow. However, the desired real-to-complex transformer is approximately equal to that of
bandwidth of a transformer at a designated center frequency its equivalent complex-to-complex sub-transformer according
can be readily achieved by introducing additional matching to (21). For the sake of analysis, the source and impedance
frequencies adjacent to the center frequency. transform ratio of a real-to-complex transformer is assumed
to be a real impedance of Z 0 and R(θ ) = |R(θ )| · e j ψ(θ),
C. Error Analysis respectively. Then, an equivalent unmatched real-to-real trans-
former with the virtual load of Z = Z 0 can be obtained
For a generalized complex-to-complex transformer shown further, where reflection coefficients are dominated by (30).
in Fig. 5(a), it has been shown in Section III-A that its approx- Due to the selected virtual load of Z = Z 0 , the element of
imation error is only determined by the last sub-transformer (Z 0 − Z )/(Z 0 + Z ) in B is reduced to 0, and the remainder is
since those of the remainder sub-transformers are eliminated
by the recursive formulas of (26). Then, the following attention |R(θi )|e j ψ(θi ) − 1
γ (θi ) = |γ (θi )|e j ϕ(θi ) = . (36)
will be paid to the error analysis of single sub-transformer |R(θi )|e j ψ(θi ) + 1
under small reflection condition. Letting
Considering the dominated equation (30) of a sub-
N $
 %
transformer, B is derived from the equivalent transformation
 =
  2i−1 + |γ (θi )| sin ϕ(θi )
|γ (θi )| cos ϕ(θi )  2i
between perfectly matched real-to-complex and unmatched
real-to-real transformers, which is accurate. Therefore, i=1
the only error of solution is resulted from A, which is derived (37a)
from the total reflection coefficient represented in terms of 
N
 
small reflection coefficients. B= |γ (θi )| cos ϕ(θi )B2i−1 + |γ (θi )| sin ϕ(θi )B2i
i=1
1) Approximation Error of Small Reflection Coefficients:
(37b)
According to [13], the recursive formula of two successive
T
transmission lines is given by Bm = (0, 0, . . . , 1, 0, . . . , 0) , m = 1, 2, . . . , 2N
 (37c)
n + n+1 (θ ) · e− j 2θ
n (θ ) = 
1 + n · n+1 (θ ) · e− j 2θ where the superscript of T denotes the transpose operation of
  1 , 
a matrix,  2 , · · · , 
2 N , and Bm are column matrices with
= n + 1 − n2
2N + 1 elements, and the only nonzero element of 1 in Bm
∞
   (m+1) − j 2(m+1)θ ! is in the mth row. Substituting (37) into (30), one has
· (−n )m · n+1 (θ ) ·e (32)
m=0  m = Bm , m = 1, 2,. . ., 2N.
A· (38)
where n is the small reflection coefficient, and and n (θ )  m is also invariable.
 In (38), as A and B are constant, 
n+1 (θ ) are the total reflection coefficients of interfaces
Applying the inequality of a + b ≤ a + b to (37a),
between n- and n+1-section transmission lines and between
then, it is followed that:
n+1- and n+2-section transmission lines, respectively. Based
on the MSRT, the formula of (32) is approximated by N $
  %
    ≤
 2i−1 + 
|γ (θi )| ·  2i . (39)
n (θ ) = n + 1 − n2 · n+1 (θ ) · e− j 2θ . (33) i=1

Then, one obtains the introduced error of Error(θ ) due to the From (39), it is seen that the upper limit of small reflec-
approximation of (33) instead of (32)  is merely subjected to |γ (θi )| since
tion coefficient 
" " 
m is constant. According to (36), after some algebraic
" ∞
  m+1 − j 2(m+1)θ ""
" m  computations, one has
|Error(θ )| = " 1−n · (−n ) n+1 (θ )
2
·e ".
" " 
m=1
2[|R(θi )|2 + 1]
(34) |γ (θi )| = −1 + . (40)
#∞ #∞ |R(θi )| + 2|R(θi )| cos ψ(θi ) + 1
2
Using the inequalities of | m=1 an | ≤ m=1 |an | (an is an

arbitrary complex number) and |n+1 (θ )| ≤1, one has As Re[Z L (θi )/Z 0 ] > 0, one has ψ(θi ) ∈(−π/2, π/2).
Then, from (40), it can be easily concluded that |γ (θi )|
|Error(θ )| ≤ (1 + |n |) · |n |. (35) increases with |ψ(θi )| in the range of 0 to π/2 when |R(θi )|

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: GENERALIZED APPROACH FOR MULTIFREQUENCY TRANSMISSION LINE TRANSFORMER 3611

TABLE III
E LECTRICAL PARAMETERS OF THE P ENTA -F REQUENCY T RANSFORMER
W ITH D IFFERENT V IRTUAL L OADS (U NIT: )

Fig. 8. Frequency responses of the transformer with the frequency ratio of

f 0 : 1.8 f 0 : 2.9 f 0 : 3.9 f 0 : 5.8 f 0 .

TABLE IV
E LECTRICAL PARAMETERS OF A P ENTA -F REQUENCY
T RANSFORMER W ITH THE F REQUENCY R ATIO
OF f 0 : 1.8 f 0 : 2.9 f 0 : 3.9 f 0 : 5.8 f 0 (U NIT: )

Fig. 7. Frequency responses of the penta-frequency transformer with different

virtual loads.

is invariable. In order to further obtain the relationship between

|γ (θi )| and |R(θi )|, the formula of (40) is rewritten as

4 cos ψ(θi ) calculated frequency responses of the designed transformer.
|γ (θi )| = 1 − . (41)
|R(θi )| + |R(θ
1
i )|
+ 2 cos ψ(θi ) It is seen that a return loss better than 30 dB can be achieved
at the designated matching frequencies. Furthermore, the fre-
It can be readily shown that |γ (θi )| increases with |R(θi )|
quency responses of the transformer with the virtual loads
in the range of [1, ∞) when ψ(θi ) is invariable. Therefore, of 50 or 80 agree well, which indicates that the selection of
when the impedance transform ratio is close to 1, both |γ (θi )| virtual load has a little impact on the computation accuracy.
and   approach to 0. Then, according to (35), the error
Fig. 8 further shows the frequency responses of two
of A is close to 0. Consequently, the solution of dominated transformers with another frequency ratio of f 0 : 1.8 f 0 :
equation (30) is accurate. 2.9 f 0 : 3.9 f 0 : 5.8 f 0 . Their only difference lies in the fourth
In summary, the error of a complex-to-complex transformer
band. One denoted by #1 is perfectly matched at 3.5 GHz,
is attributed to the approximation error of small reflection coef- whereas the other one denoted by #2 is perfectly matched at
ficients, which can be sufficiently small by introducing enough 3.3 and 3.7 GHz to achieve a wider bandwidth at 3.5 GHz.
sub-transformers based on an equal impedance transform ratio. Their design parameters, including the matching frequen-
cies (bands of applications), the source of Z S (θi ), and the load
IV. S IMULATIONS AND E XPERIMENTS of Z L (θi ), are listed in the first three rows of Table IV. When
To validate the feasibility of the proposed method, several the reference frequencies are selected as 5.35 and 5.8 GHz for
multifrequency transformers with frequency-dependent com- transformers #1 and #2, respectively (the different reference
plex source and load are designed. The first example is a frequencies resulted from their different targeted matching
penta-frequency transformer with the frequency ratio of f 0 : frequencies), their electrical parameters can be obtained as
2.1 f 0 : 2.7 f 0 : 3.9 f 0 : 5.4 f 0 , where its design parameters, listed in Table IV. Note that for the identical complex-to-
including the matching frequencies (bands of applications), complex transformers #1 and #2, their equivalent real-to-
source impedance of Z S (θi ), and load impedance of Z L (θi ), complex transformers are different since that their targeted
are listed in the first three rows of Table III. By selecting matching frequencies are five and six, respectively. On the
the reference frequency of 5.2 GHz and the virtual load other hand, at the fourth band of transformer #2, the equivalent
of 50 or 80 and following the steps of design flowchart real-to-complex transformer has two different loads due to its
shown in Fig. 6, the equivalent real-to-complex transformer as two matching frequencies of 3.3 and 3.7 GHz. From Fig. 8,
well as the characteristic impedances of all transmission lines it is seen that the return loss at the designated matching
can be readily obtained, as listed in Table III. Fig. 7 shows the frequencies is better than 30 dB, and the frequency deviations

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
3612 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 67, NO. 9, SEPTEMBER 2019

TABLE VI
E LECTRICAL PARAMETERS OF THE Q UAD -F REQUENCY T RANSFORMER
W ITH H IGH M ATCHING F REQUENCIES (U NIT: )

Fig. 9. Frequency responses of the quad-frequency transformer with large

impedance transform ratio.

TABLE V
E LECTRICAL PARAMETERS OF THE Q UAD -F REQUENCY T RANSFORMER
W ITH L ARGE I MPEDANCE T RANSFORM R ATIO (U NIT: )

Fig. 10. Frequency responses of the quad-frequency transformer with high

matching frequencies.

sub-transformers, it is well matched at the desired match-

ing frequencies. The reason for this is that the impedance
transform ratio and maximum small reflection coefficient
of a transformer with two sub-transformers are reduced to
2.0 and 0.11, respectively. Therefore, for a transformer with
between the obtained and desired matching frequencies are arbitrary impedance transform ratio, its frequency deviation
about 2%. Moreover, the bandwidth (defined by a return loss from the desired matching frequencies can be sufficiently small
of 20 dB) of transformer #2 at the fourth band is 5.4 times of by introducing more sub-transformers.
that of transformer #1 due to an additional matching frequency Another quad-frequency transformer with matching fre-
adjacent to 3.5 GHz. Therefore, the desired bandwidth of a quencies (bands of applications) listed in the first row of
transformer can be achieved by introducing more matching Table VI is also taken as an example to validate that the
frequencies. It should be pointed out that the total electrical proposed method is not limited to applications with low
lengths of transformers #1 and #2 are 151.4◦ and 167.58◦ at operating frequencies. Given the source of Z S (θi ) and the load
0.9 GHz, respectively, which indicates that the total electrical of Z L (θi ) listed in the second and third rows of Table VI and
length slowly increases with the number of targeted matching selected the reference frequency of 12.6 GHz, its electrical
frequencies and may result in a slight increase of insertion parameters can be readily obtained as listed in the last four
loss. rows of Table VI. Subsequently, its frequency responses are
In order to validate that the proposed method is applicable shown in Fig. 10, and it is seen that the frequency devia-
to a transformer with arbitrary frequency-dependent complex tions between the obtained and desired matching frequencies
source and load, Fig. 9 shows the frequency responses of are less than 2%. It is noted that the rapid variation of
a quad-frequency transformer with the frequency ratio of source and load impedances between 3.5 and 5.2 GHz mainly
f 0 : 2.7 f 0 : 3.9 f 0 : 5.8 f 0 and impedance transform ratio results in the variation of characteristic impedances among
larger than 4. Its matching frequencies (bands of applica- Z 6 , Z 7 , and Z 8 .
tions), source impedance of Z S (θi ), and load impedance Fig. 11 shows two equivalent circuits mimicking the source
of Z L (θi ) are listed in the first three rows of Table V. and load of cascaded power amplifiers or low-noise ampli-
In order to study the impact of impedance transform ratio fiers [29]. It is assumed that the electrical parameters of
on its design accuracy, the transformer is carried out through the equivalent circuits for the realization of a wide range of
one or two cascaded sub-transformers, whose electrical para- complex impedance are selected, as listed in Table VII. Note
meters are all listed in Table V in addition to θ = 13.97◦ that the impact of bias network and blocking capacitors on the
at 0.9 GHz. As seen in Fig. 9, the transformer composed of source and load impedances also should be taken into account
only one sub-transformer is poorly matched with a maximum in a practical application. Namely, the final source/load
frequency deviation of 17.8% from the desired matching impedance of power amplifiers or low-noise amplifiers based
frequencies. However, for the transformer composed of two on the proposed method should be the combination of

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: GENERALIZED APPROACH FOR MULTIFREQUENCY TRANSMISSION LINE TRANSFORMER 3613

Fig. 11. Equivalent circuits and frequency-dependent complex impedances

mimicking impedances of cascaded power amplifiers or low-noise amplifiers.
(a) Source of Z S (θi ) and its equivalent circuit. (b) Load of Z L (θi ) and its
equivalent circuit. Fig. 12. Fixture for TRL calibration.

TABLE VII
E QUIVALENT-C IRCUIT PARAMETERS OF S OURCE AND L OAD

TABLE VIII
E LECTRICAL PARAMETERS OF THE P ENTA -F REQUENCY T RANSFORMER
W ITH F REQUENCY-D EPENDENT C OMPLEX S OURCE
AND L OAD (U NIT: )

independent source/load impedance, bias network, and block-

ing capacitors. Alternatively, the ceramic capacitors used for
dc blocking are selected to be large enough, and the input
impedance of multifrequency bias network [30], [31] con- Fig. 13. Calculated, simulated, and measured frequency responses of
nected to the matching network approaches to infinity. Then, fabricated transformers. (a) Transformer #1. (b) Transformer #2.
given the matching frequencies (bands of applications) of
transformers #1 and #2 as listed in the first and sixth rows source and load of 50 , which is fully dominated by the
of Table VIII, their sources and loads listed in Table VIII can transformer and independent of its source and load. Then,
be obtained according to the equivalent circuits. Following the the frequency responses of the transformer shown in Fig. 13
steps of design flowchart shown in Fig. 6, one can easily obtain can be obtained on the basis of post-processing, the derived
their electrical parameters listed in Table VIII in addition to impedance matrix, and the given frequency-dependent com-
θ = 16.7◦ at 0.5 GHz for transformer #1 and θ = 15.8◦ plex source and load [28]. It is seen that the calculated results
at 0.5 GHz for transformer #2. The different selections of (MATLAB 2016a) agree well with the simulated (HFSS 13.0)
θ are due to their differences of matching frequencies listed and measured results (Agilent E8361C), and their slight dis-
in Table VIII. crepancy is mainly attributed to the manufacture tolerance
The two transformers as well as a fixture for the thru-reflect- and meander line. Furthermore, both transformers are well
line (TRL) calibration are simulated and implemented on a matched at the desired frequencies with a minimum return
Rogers 4350B substrate with the relative dielectric constant loss of 25 dB. Thus, a good initial value for a design can be
of 3.48 and the thickness of 0.508 mm. The fixture as shown provided by using the proposed mathematical approach. Even
in Fig. 12 is utilized to shift the measurement reference when it is used for an ADS-based optimization of the complex-
planes from the vector analyzer to a transformer [32]–[34]. to-complex transformer with more bands, both goal-realization
Based on it, the impedance matrix of a transformer can and considerable time-reduction are expected. In addition,
be derived from the S-parameters of the transformer with the bandwidths of transformer #1 at 1.2 and 2.6 GHz, defined

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
3614 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 67, NO. 9, SEPTEMBER 2019

and load is proposed in this paper, which is based on MSRT

combining with an equivalent transformation between the
perfectly matched complex-to-complex and unmatched real-
to-real transformers. Moreover, the transformer with arbitrary
impedance transform ratio can be achieved by introducing
sufficient sub-transformers based on equal impedance trans-
form ratio. The proposed method can be utilized to design
multifrequency or wideband cascaded power amplifiers and
Fig. 14. Photographs of fabricated transformers. (a) Transformer #1.
(b) Transformer #2.
low-noise amplifiers with compact size.

TABLE IX R EFERENCES
C OMPARISON OF VARIOUS M ULTIFREQUENCY T RANSFORMERS
[1] M. Chongcheawchamnan, S. Patisang, M. Krairiksh, and
I. D. Robertson, “Tri-band Wilkinson power divider using a three-
section transmission-line transformer,” IEEE Microw. Wireless Compon.
Lett., vol. 16, no. 8, pp. 452–454, Aug. 2006.
[2] K.-K. M. Cheng and C. Law, “A novel approach to the design and
implementation of dual-band power divider,” IEEE Trans. Microw.
Theory Techn., vol. 56, no. 2, pp. 487–492, Feb. 2008.
[3] Y. Wu, S. Y. Zheng, S. W. Leung, Y. Liu, and Q. Xue, “An analytical
design method for a novel dual-band unequal coupler with four arbitrary
terminated resistances,” IEEE Trans. Ind. Electron., vol. 61, no. 10,
pp. 5509–5516, Oct. 2014.
[4] A. Cidronali, I. Magrini, N. Giovannelli, M. Mercanti, and G. Manes,
“Experimental system level analysis of a concurrent dual-band power
amplifier for WiMAX and WCDMA application,” Int. J. Microw. Wire-
less Technol., vol. 1, no. 2, pp. 99–107, Feb. 2009.
[5] A. Fukuda, T. Furuta, H. Okazaki, and S. Narahashi, “A 0.9-5-GHz wide-
range 1W-class reconfigurable power amplifier employing RF-MEMS
switches,” in IEEE MTT-S Int. Microw. Symp. Dig., Jun. 2006,
pp. 1859–1862.
[6] W. Webb, The Complete Wireless Communications Professional:
A Guide for Engineers and Managers. Boston, MA, USA: Artech House,
2003.
by a return loss of 20 dB, are 4.2% and 3.5%, while those [7] Y. L. Chow and K. L. Wan, “A transformer of one-third wave-
of transformer #2 are up to 22.3% and 9.2%. The reason length in two sections—For a frequency and its first harmonic,”
IEEE Microw. Wireless Compon. Lett., vol. 12, no. 1, pp. 22–23,
for this is that for transformer #2, there are two matching Jan. 2002.
frequencies of 1.25 and 1.45 GHz adjacent to their center [8] C. Monzon, “Analytical derivation of a two-section impedance trans-
frequency of 1.35 GHz, and 2.32 and 2.48 GHz adjacent to former for a frequency and its first harmonic,” IEEE Microw. Wireless
Compon. Lett., vol. 12, no. 10, pp. 381–382, Oct. 2002.
their center frequency of 2.4 GHz. [9] C. Monzon, “A small dual-frequency transformer in two sections,” IEEE
The photographs of fabricated transformers are shown Trans. Microw. Theory Techn., vol. 54, no. 4, pp. 1157–1161, Apr. 2003.
in Fig. 14. It is seen that corner cuts are utilized to compensate [10] S. J. Orfanidis, “A two-section dual-band Chebyshev impedance
transformer,” IEEE Microw. Wireless Compon. Lett., vol. 13, no. 9,
the parasitic capacitor due to the meandering transmission pp. 382–384, Sep. 2003.
lines [35], and the overall dimensions of the penta- and [11] M. Chongcheawchamnan, S. Patisang, S. Srisathit, R. Phromloungsri,
triple-band transformers excluding connectors are 0.0145λ2g and S. Bunnjaweht, “Analysis and design of a three-section transmission-
line transformer,” IEEE Trans. Microw. Theory Techn., vol. 53, no. 7,
and 0.0130λ2g (λg is the guided wavelength at 0.5 GHz), pp. 2458–2462, Jul. 2005.
respectively. It is noted that their overall sizes are mainly sub- [12] M. Khodier, N. Dib, and J. Ababneh, “Design of multi-
jected to the electrical parameters of the lowest two matching band multi-section transmission line transformer using particle
swarm optimization,” Elect. Eng., vol. 90, no. 4, pp. 293–300,
frequencies, and those of the remainder matching frequencies Apr. 2008.
have less impact on the overall size due to their much higher [13] L. Liu et al., “A novel analytical method for multi-frequency transmis-
frequencies. In order to achieve compact size of a transformer, sion line transformer,” IEEE Microw. Wireless Compon. Lett., vol. 26,
no. 8, pp. 556–558, Aug. 2016.
the reference frequency of r in formula (27) can be selected as [14] L. Liu et al., “Multifrequency transformer with arbitrary frequency and
high as possible to reduce the electrical length of every section real impedance transform ratio,” IEEE Microw. Wireless Compon. Lett.,
transmission line. Furthermore, more meandering transmission vol. 27, no. 9, pp. 785–787, Sep. 2017.
[15] P. Colantonio, F. Giannini, and L. Scucchia, “Matching network design
lines can be introduced. In addition, other configuration based criteria for wideband high-frequency amplifiers,” Int. J. RF Microw.
on the proposed method will be developed in the future work. Comput.-Aided Eng., vol. 15, no. 5, pp. 423–433, Sep. 2005.
Various transmission line transformers listed in Table IX [16] C. F. Jou, K.-H. Cheng, C.-M. Lin, and J.-L. Chen, “Dual band CMOS
power amplifier for WLAN applications,” in Proc. Midwest Symp.
are compared. It is seen that the proposed method can be Circuits Syst., vol. 3, Dec. 2003, pp. 1227–1230.
utilized to the design of transformer with arbitrary impedance [17] X. Fu, D. T. Bespalko, and S. Boumaiza, “Novel dual-band matching
transform ratio and matching frequencies. network for effective design of concurrent dual-band power amplifiers,”
IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 61, no. 1, pp. 293–301,
Jan. 2014.
V. C ONCLUSION [18] Y. L. Wu, L. X. Jiao, and Y. A. Liu, “Comments on novel dual-band
matching network for effective design of concurrent dual-band power
A generalized approach for the multifrequency transmission amplifiers,” IEEE Trans. Circuit Syst. I, Reg. Papers, vol. 62, no. 9,
line transformer with frequency-dependent complex source pp. 2361–2365, Sep. 2015.

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
LIU et al.: GENERALIZED APPROACH FOR MULTIFREQUENCY TRANSMISSION LINE TRANSFORMER 3615

[19] R. Fagotti, A. Cidronali, and G. Manes, “Concurrent hex-band GaN Ronghong Jin (M’09–SM’13–F’17) received the
power amplifier for wireless communication systems,” IEEE Microw. B.S. degree in electronic engineering, the M.S.
Wireless Compon. Lett., vol. 21, no. 2, pp. 89–91, Feb. 2011. degree in electromagnetic and microwave technol-
[20] S. F. R. Chang et al., “A dual-band RF transceiver for multistandard ogy, and the Ph.D. degree in communication and
WLAN applications,” IEEE Trans. Microw. Theory Techn., vol. 53, no. 3, electronic systems from Shanghai Jiao Tong Uni-
pp. 1048–1055, Mar. 2005. versity (SJTU), Shanghai, China, in 1983, 1986, and
[21] Y. Wu, Y. Liu, and S. Li, “A dual-frequency transformer for complex 1993, respectively.
impedances with two unequal sections,” IEEE Microw. Wireless Com- In 1986, he joined the Department of Electronic
pon. Lett., vol. 19, no. 2, pp. 77–79, Feb. 2009. Engineering, SJTU, where he was an Assistant Pro-
[22] S. C. D. Roy, “Comment on ‘A dual-frequency transformer for com- fessor, a Lecturer, an Associate Professor, and is
plex impedances with two unequal sections,”’ IEEE Microw. Wireless currently a Professor. From 1997 to 1999, he was a
Compon. Lett., vol. 19, no. 9, p. 602, Sep. 2009. Visiting Scholar with the Department of Electrical and Electronic Engineering,
[23] X. Liu, Y. Liu, S. Li, F. Wu, and Y. Wu, “A three-section dual- Tokyo Institute of Technology, Meguro, Japan. From 2001 to 2002, he
band transformer for frequency-dependent complex load impedance,” was a Special Invited Research Fellow with the Communication Research
IEEE Microw. Wireless Compon. Lett., vol. 19, no. 10, pp. 611–613, Laboratory, Tokyo, Japan. From 2006 to 2009, he was a Guest Professor with
Oct. 2009. the University of Wollongong, Wollongong, NSW, Australia. He is also a
[24] M.-L. Chuang, “Dual-band impedance transformer using two-section Distinguish Guest Scientist with the Commonwealth Scientific and Industrial
shunt stubs,” IEEE Trans. Microw. Theory Techn., vol. 58, no. 5, Research Organization, Sydney, NSW, Australia. He has authored or coau-
pp. 1257–1263, May 2010. thored over 400 papers in refereed journals and conference proceedings
[25] M. A. Nikravan and Z. Atlasbaf, “T-section dual-band impedance trans- and has coauthored 7 books. He holds more than 70 patents in antenna
former for frequency-dependent complex impedance loads,” Electron. and wireless technologies. His current research interests include antennas,
Lett., vol. 47, no. 9, pp. 551–553, Apr. 2011. electromagnetic theory, numerical techniques of solving field problems, and
[26] Y. Wu, Y. Liu, S. Li, C. Yu, and X. Liu, “A generalized dual- wireless communications.
frequency transformer for two arbitrary complex frequency-dependent Dr. Jin is a Committee member of the Antenna Branch of the Chinese
impedances,” IEEE Microw. Wireless Compon. Lett., vol. 19, no. 12, Institute of Electronics, Beijing, China. He was a recipient of the
pp. 792–794, Dec. 2009. National Technology Innovation Award, the National Nature Science Award,
[27] K.-A. Hsieh, H.-S. Wu, K.-H. Tsai, and C.-K. C. Tzuang, “A dual- the 2012 Nomination of National Excellent Doctoral Dissertation (Supervisor),
band 10/24-GHz amplifier design incorporating dual-frequency complex the 2017 Excellent Doctoral Dissertation (Supervisor) of the China Institute
load matching,” IEEE Trans. Microw. Theory Techn., vol. 60, no. 6, of Communications, the Shanghai Nature Science Award, and the Shanghai
pp. 1649–1657, Jun. 2012. Science and Technology Progress Award.
[28] D. A. Frickey, “Conversions between S, Z, Y, H, ABCD, and
T parameters which are valid for complex source and load impedances,” Xianling Liang (M’11–SM’17) received the B.S.
IEEE Trans. Microw. Theory Techn., vol. 42, no. 2, pp. 205–211, degree in electronic engineering from Xidian Uni-
Feb. 1994. versity, Xi’an, China, in 2002, and the Ph.D. degree
[29] Cree Corporation. CGH40010 Data Sheet. [Online]. Available: in electric engineering from Shanghai University,
[Link] Shanghai, China, in 2007.
[30] R. Negra, A. Sadeve, S. Bensmida, and F. M. Ghannouchi, “Concurrent From 2007 to 2008, he was a Post-Doctoral
dual-band class-F load coupling network for applications at 1.7 and Research Fellow with the Institute National de
2.14 GHz,” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 55, no. 3, la Recherche Scientfique, University of Quebec,
pp. 259–263, Mar. 2008. Montreal, QC, Canada. In 2008, he joined the
[31] S. Gao, Z. Wang, and C.-W. Park, “Concurrent dual-band power ampli- Department of Electronic Engineering, Shanghai
fier with second harmonic controlled by gate and drain bias circuit,” Jiao Tong University (SJTU), Shanghai, as a
in Proc. IEEE Int. Conf. Microw. Technol. Comput. Electromagn., Lecturer and then became an Associate Professor in 2012. He has
May 2011, pp. 309–312. authored or coauthored over 250 papers, including 141 journal papers and
[32] G. F. Engen and C. A. Hoer, “Thru-reflect-line: An improved technique 113 conference papers and coauthored 1 book and 3 chapters in microwave
for calibrating the dual six-port automatic network analyzer,” IEEE and antenna fields. He holds 15 patents in antenna and wireless technologies.
Trans. Microw. Theory Techn., vol. MTT-27, no. 12, pp. 987–993, His current research interests include OAM-EM wave propagation and antenna
Dec. 1979. design, time-modulated/4-D array and applications, anti-interference antenna
[33] R. B. Marks, “A multiline method of network analyzer calibration,” and array, integrated active antenna and array, and ultra-wideband wide-angle
IEEE Trans. Microw. Theory Techn., vol. 39, no. 7, pp. 1205–1215, scanning phased array.
Jul. 1991. Dr. Liang was a recipient of the Award of Shanghai Municipal Excellent
[34] H. Fan et al., “A method of accurate co-simulation by considering Doctoral Dissertation in 2008, the Nomination of the National Excellent
lumped port setting in EM simulator,” in Proc. IEEE 4th Asia–Pacific Doctoral Dissertation in 2009, the Best Paper Award presented at the
Conf. Antennas Propag., Jun./Jul. 2015, pp. 112–116. International Workshop on Antenna Technology: Small Antennas, Innovative
[35] D. M. Pozar, Microwave Engineer, 4th ed. New York, NY, USA: Wiley, Structures, and Materials in 2010, the SMC Excellent Young Faculty and the
2011. Excellent Teacher Award of SJTU in 2012, the Shanghai Natural Science
Award in 2013, the Best Paper Award presented at the IEEE International
Symposium on Microwave, Antenna, Propagation, and EMC Technologies
in 2015, the 4th China Publishing Government Book Award, and the Okawa
Foundation Research Grant in 2017.

Haijun Fan (M’17) received the B.S. degree in

applied physics from the University of Electronic
Liang Liu received the B.S. degree in electronics Science and Technology of China, Chengdu, China,
science and technology from Southwest Jiaotong in 2009, and the M.S. degree in electromagnetic
University, Chengdu, China, in 2005, and the M.S. and microwave technology and Ph.D. degree in
degree in electronics science and technology from electronic science and technology from Shanghai
Zhejiang University, Hangzhou, China, in 2008. Jiao Tong University, Shanghai, China, in 2011 and
He is currently pursuing the Ph.D. degree at 2016, respectively.
Shanghai Jiao Tong University, Shanghai, China. From 2016 to 2017, he was a Post-Doctoral Fellow
From 2008 to 2009, he was an RF Engineer with with the Department of Electronic Engineering, The
Winhap Communications Inc., Shenzhen, China. Chinese University of Hong Kong, Hong Kong.
From 2009 to 2010, he was a Research Assis- Since 2017, he has been a Post-Doctoral Fellow with the Institute of Sensors
tant with the City University of Hong Kong, Signals and Systems, Heriot-Watt University, Edinburgh, U.K. His current
Hong Kong. From 2010 to 2013, he was an RF Engineer with SED Wire- research interests include microwave reconfigurable circuits, power dividers,
less Communications Inc., Shenzhen. His current research interests include filter synthesis, microwave switches, beam-forming networks, antenna arrays,
multifrequency/wideband power dividers, filters, and power amplifiers. and antenna diversity.

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.
3616 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 67, NO. 9, SEPTEMBER 2019

Xudong Bai (M’18) received the B.S. degree in Junping Geng (M’08–SM’17) received the B.S.
electronic engineering, the M.S. degree in electro- degree in plastic working of metals, the M.S.
magnetic and microwave technology, and the Ph.D. degree in corrosion and protection of equipment,
degree in electronic science and technology from and the Ph.D. degree in circuit and system from
Shanghai Jiao Tong University (SJTU), Shanghai, Northwestern Polytechnic University, Xi’an, China,
China, in 2009, 2012, and 2016, respectively. in 1996, 1999, and 2003, respectively.
In 2016, he joined the RD Center, Shanghai From 2003 to 2005, he was a Post-Doctoral
Aerospace Electronics Co., Ltd., Shanghai, where Researcher with Shanghai Jiao Tong Univer-
he is currently an Advanced Engineer. His cur- sity (SJTU), Shanghai, China. In 2005, he joined
rent research interests include phased arrays, elec- the Electronic Engineering Department, SJTU, as a
tromagnetic metamaterials, and OAM-EM wave faculty member, where he is currently an Associate
propagation and antenna design. Professor. He has been involved in electromagnetic compatibility for high-
altitude platform stations, multiantennas for terminals, smart antennas, and
nanoantennas. He has authored or coauthored over 260 refereed journal and
conference papers, 2 book chapters, and 1 book. He holds 50 patents with over
50 pending. His current research interests include antennas, electromagnetic
theory, and computational techniques of electromagnetic and nanoantennas.
Dr. Geng is a member of the Chinese Institute of Electronics, Beijing,
China. He was a recipient of the Technology Innovation Award of the Chinese
Ministry of Education in 2007 and the Technology Innovation Award of the
Chinese Government in 2008.

Weiren Zhu (M’16–SM’18) received the B.S. and

Ph.D. degrees in physics from Northwestern Poly-
technical University, Xi’an, China, in 2006 and
2011, respectively.
From 2011 to 2012, he was a Post-Doctoral Fellow
with the Nonlinear Physics Centre, Australian
National University, Canberra, ACT, Australia. From
2012 to 2016, he was a Research Fellow with
the Advanced Computing and Simulation Labora-
tory, Department of Electrical and Computer Sys-
Han Zhou received the B.S. degree from Shanghai tems Engineering, Monash University, Clayton, VIC,
Jiao Tong University, Shanghai, China, in 2014, Australia. Since 2016, he has been with the Department of Electronic
where he is currently pursuing the Ph.D. degree in Engineering, Shanghai Jiao Tong University, Shanghai, China, as an Associate
electromagnetic and microwave technology. Professor. He has authored or coauthored over 100 refereed journal papers
His current research interests include multi- and over 40 conference proceedings. His current research interests include
in multi-out antenna, metamaterials, microwave electromagnetic metamaterials, antennas and RF devices, and surface plasmon
devices, and phased arrays. polaritons.
Dr. Zhu is a Senior Member of the Optical Society of America (OSA).
He was a recipient of the Shanghai Pujiang Talent Program of the Shanghai
Science and Technology Commission in 2017. He currently serves as an
Associate Editor for the IEEE P HOTONICS J OURNAL and IEEE A CCESS .

Authorized licensed use limited to: Heriot-Watt University. Downloaded on January 14,2021 at 23:13:33 UTC from IEEE Xplore. Restrictions apply.