microwave

6 network theory
and passive devices

INTRODUCTION 6.1
A microwave network is formed when several microwave devices and components such as sources,
attenuators, resonators, filters, amplifiers, etc., are coupled by transmission lines or waveguides for the
desired transmission of a microwave signal. The point of interconnection of two or more devices is called a
junction.
For a low-frequency network, a port is a pair of terminals whereas for a microwave network, a port is a
reference plane transverse to the length of the microwave transmission line or waveguide. At low frequencies,
the physical length of the network is much smaller than the wavelength of the signal transmitted. Therefore,
the measurable input and output variables are voltage and current which can be related in terms of the
impedance Z-parameters, or admittance Y-parameters, or hybrid h-parameters, or ABCD parameters. For a
two-port network as shown schematically in Fig. 6.1, these relationships are given by

È V1 ˘ È Z11 Z12 ˘ È I1 ˘
ÍV ˙ = Í Z ˙Í ˙ Impedance (Z) parameters (6.1)
Î 2˚ Î 21 Z 22 ˚ Î I 2 ˚
È I1 ˘ ÈY11 Y12 ˘ È V1 ˘
Í I ˙ = ÍY ˙Í ˙ Admittance (Y) parameters (6.2)
Î 2˚ Î 21 Y22 ˚ ÎV2 ˚
ÈV1 ˘ È h11 h12 ˘ È I1 ˘
Í I ˙ = Íh ˙Í ˙ Hybrid (h) parameters (6.3)
Î 2˚ Î 21 h22 ˚ ÎV2 ˚

ÈV1 ˘ È A + B ˘ ÈV2 ˘
ÍI ˙ = Í ˙Í ˙ Transmission (ABCD) parameters (6.4a)
Î 1˚ ÎC + D ˚ Î I 2 ˚
In Eq. (6.4a), +I2 for current at Port 2 is outward and –I2 taken for current at Port 2 is inward. Here, Zij,
Yij, and A, B, C and D are suitable constants that characterize the junction. A, B, C and D parameters are
convenient to represent each junction when a number of circuits are connected together in cascade. Here
the resultant matrix, which describes the complete cascade connection, can be obtained by multiplying the
matrices describing each junction as follows:
ÈA B˘ È A1 B1 ˘ È A2 B2 ˘ È An Bn ˘
ÍC D ˙ = ÍC D ˙ ÍC ˙ ... Í ˙ (6.4b)
Î ˚ Î 1 1˚ Î 2 D2 ˚ ÎCn Dn ˚
 Microwave Network Theory and Passive Devices 219

Fig. 6.1 A two-port network

These parameters can be measured under short- or open-circuit condition for use in the analysis of the
circuit.
At microwave frequencies, the physical length of the component or line is comparable to or much larger
than the wavelength. Furthermore, the voltage and current cannot be uniquely defined at a given point in
a single conductor waveguide. Furthermore, measurement of Z, Y, h and ABCD parameters is difficult at
microwave frequencies due to the following reasons:
1. Absence of unique definition of voltage and current as noted above and non-availability of terminal
voltage and current-measuring equipment, even in the cases of TEM lines (coaxial, strip and
microstrip lines) where such quantities can be uniquely defined.
2. Short-circuit, and especially open-circuit, are not easily achieved for a wide range of frequencies.
3. Presence of active devices makes the circuit unstable for short or open circuit.
Therefore, microwave circuits are analyzed using scattering or S-parameters which linearly relate the
amplitudes of scattered (reflected and transmitted) waves with those of incident waves. However, many
of the circuit-analysis techniques and circuit properties that are valid at low frequencies are also valid for
microwave circuits. Thus, for circuit analysis, S-parameters can be related to the Z or Y or ABCD parameters.
The properties of the parameters are described in the following sections.

PROPERTIES OF Z AND Y MATRICES FOR RECIPROCAL

NETWORKS 6.2
In a reciprocal network, the impedance and the admittance matrices are symmetrical and the junction media
are characterized by scalar electrical parameters m and e. For a multiport (N port) network, let the incident
wave amplitude Vn+ be so chosen that the total voltage Vn = Vn+ + Vn– at all ports n = 1, 2, ..., N, except the
ith port where the fields are Ei, Hi. Similarly, let Vn = 0 at all ports except jth one where the fields are Ej, Hj.
Then from the Lorentz reciprocity theorem,
Ú (Ei ¥ Hj – Ej ¥ Hi) ◊ dS = 0 (6.5)
S

where S is the closed surface area of the conducting walls enclosing the junction and N ports in the absence
of any source. Since the integral over the perfectly conducting walls vanishes, the only non-zero integrals are
those taken over the reference planes of the corresponding ports, so that
N
Â Ú (Ei ¥ Hj – Ej ¥ Hi) ◊ dS = 0 (6.6)
n = 1 tn

Since Vn except Vi and Vj are zero, Eti = n ¥ Ei and Etj = n ¥ Ej are zero on all reference planes at the
corresponding ports except ti and tj, respectively. Therefore, Eq. (6.6) reduces to

Ú (Ei ¥ H j )◊ dS = Ú (E j ¥ Hi ) ◊ dS (6.7)
ti tj
220 Microwave Engineering

or Pij = Pji (6.8)

where Pij represents the power at the reference plane i due to an input voltage at the plane j.
From the admittance matrix representation [I ] = [Y ] [V ] and power relation P = VI, Eq. (6.8) reduces
to
Vi Vj Yij = Vj Vi Yji or, Yij = Yji (6.9)
and Zij = Zji (6.10)
This proves that the impedance and admittance matrices are symmetrical for a reciprocal junction.

SCATTERING OR S-MATRIX REPRESENTATION OF

MULTIPORT NETWORK 6.3
As discussed in Section 6.1, the incident and reflected wave amplitudes of microwaves at any port are used
to characterize a microwave circuit. The amplitudes are normalized in such a way that the square of any of
these variables gives the average power in that wave in the following manner:
Input power at the nth port, Pin = 1/2 |an|2 (6.11)
Reflected power at the nth port Prn = 1/2 |bn|2 (6.12)
where an and bn represent the normalized incident wave peak amplitude and normalized reflected wave
peak amplitude at the nth port. The concept of scattering(s) parameters comes from the fact that rf and
microwave circuit may contain some discontinuity or discontinuities in the signal propagation path. At an
discontinuity the wave scattered in different directions as evanescent waves containing infinite number of
higher order modes. These modes attenuated very fast after a short distance from the point of discontinuity
within about a quarter wavelength. Then only the executed mode comes out from the different ports. All
these emerging waves are considered as reflected waves at the corresponding parts. The waves entering the
parts are considered the input or incident wave.
In a two-port network, we can express the normalized waves in terms of normalized voltages:
V1+ V1 – V1– , V2+ V2 – V2–
a1 = = a2 = = (6.13)
Z0 Z0 Z0 Z0

V1– V1 – V1+ , V2– V2 – V2+

b1 = = b2 = = (6.14)
Z0 Z0 Z0 Z0

where a’s represent normalized incident wave amplitude and b’s represent normalized reflected wave
amplitude at the corresponding ports. Here, the total voltage wave is the sum of incident and emergent
voltage waves V + and V – respectively:

V1 = V1+ + V1– I1 = I1+ - I1- (6.15)

V2 = V2+ + V2– I 2 = I 2+ - I 2- (6.16)

The numeric suffices represent the port number. The total or net power flow into any port is given by
r.m.s. value
1
P = Pi – Pr = (|a|2 – |b|2) (6.17)
2
1 Vpeak
Factor comes from the relation Vrms = or a and b are magnitor of peak values.
2 2
 Microwave Network Theory and Passive Devices 221

Therefore, in this normalization process, the characteristic impedance is normalized to unity. For a two-
port network (Fig. 6.1), the relation between incident and reflected waves are expressed in terms of scattering
parameters Sij’s:
b1 = S11 a1 + S12 a2 (6.18)
b2 = S21 a1 + S22 a2 (6.19)
The normalization process leads to a symmetrical scattering matrix for reciprocal structures. The physical
significance of S-parameters can be described as follows:
S11 = (b1 /a1 )a2 = 0 = Reflection coefficient G1 at Port 1 when Port 2 in terminated with a
matched load (a2 = 0)
S22 = (b2 /a2 )a1 = 0 = Reflection coefficient G2 at Port 2 when Port 1 in terminated with a
matched load (a1 = 0)
S12 = (b1 /a2 )a1 = 0 = Attenuation of wave travelling from Port 2 to Port 1 when a1 = 0

S21 = (b2 /a1 )a2 = 0 = Attenuation of wave travelling from Port 1 to Port 2 when a2 = 0
In general, since the incident and reflected waves have both amplitude and phase, the S-parameters are
complex numbers.
For multiport (N) networks or components, the S-parameter equations are expressed by
È b1 ˘ È S11 S12 ... S1N ˘ È a1 ˘
Í ˙ Í ˙Í ˙
b
Í 2˙ S S ... S2 N ˙ Í a2 ˙
Í 21 22
Í ˙ Í ˙Í ˙
Í ˙ = Í ˙Í ˙ (6.20)
Í ˙ Í ˙Í ˙
Í ˙ Í ˙Í ˙
Í ˙ Í ˙Í ˙
ÍÎbN ˙˚ ÎÍ SN 1 SN 2 ... SNN ˚˙ ÎÍ aN ˚˙

Figure 6.2 shows a linear multiport network having N ports. We

can write for any arbitrary n-th port of this junction the following:
an = Vn+ / Z on = I n+ Z on (6.20a)

bn = Vn- / Z on = I n- Z on (6.20b)
Net input power at port n
1 2 2
Pninp = ( an - bn ) (6.20c)
2
Power dissipated in a termination at port n is
1 2 2 1 2 2
Pndissp = ( bn - an ) = bn (1 - G n ) (6.20d) Fig. 6.2 A mul port (N) linear
2 2
network
bn
Gn = (6.20e)
an
is the magnitude of the load reflection coefficient at port n. Above relations hold for all the ports with proper
suffices. The total voltage and current at port n are
Vn = Vn+ + Vn- = Z on (an + bn ) (6.20f)
1
In = I n+ - I n- = (an - bn ) (6.20g)
Z on
222 Microwave Engineering

Solving the above equations for an and bn in terms of the terminal voltage and current yields
1Ê V ˆ
an = Á n + I n Z on ˜ (6.20h)
2 Ë Z on ¯

1 Ê Vn ˆ
bn = Á - I n Z on ˜ (6.20i)
2 Ë Z on ¯
In microwave devices or circuits, it is important to express several losses in terms of S-parameters when
the ports are match terminated. In a two-port network, if power fed at Port 1 is Pi, power reflected at the same
port is Pr and the output power at Port 2 is Po then following losses are defined in terms of S-parameters:

Pi | a |2
Insertion loss (dB) = 10 log = 10 log 1
Po | b2 |2
1 1
= 20 log = 20 log (6.21)
| S21 | | S12 |
Pi – Pr 1– | S11 |2
Transmission loss or attenuation (dB) = 10 log = 10 log (6.22)
Po | S12 |2
Pi 1
Reflection loss (dB) = 10 log = 10 log (6.23)
Pi – Pr 1 - | S11 |2

Pi
Return loss (dB) = 10 log
Pr
1 1
= 20 log = 20 log (6.24)
|G| | S11 |

Example 6.1 Find the S-matrix of a length l of a lossless transmission line terminated by matched
impedance.

Solution For a length l of a transmission line, there is no discontinuity at the two ends, so that S11 = S22
= 0. The output signals arise due to input:
b1 = a2 e–jbl
b2 = a1 e–jbl

È b1 ˘ È 0 e – j b l ˘ È a1 ˘
or, Íb ˙ = Í – jb l ˙ Í ˙ = [ S ] [ a]
Î 2˚ ÍÎe 0 ˙˚ Î a2 ˚ Fig. 6.3 Transmission lines of
Example 6.1
È 0 e– jbl ˘
\ [S] = Í ˙
ÍÎe – j b l 0 ˙˚

Properties of S-Parameters for a Junction of Ports having

Common Characteristic Impedance 6.3.1
In general, the scattering parameters are complex quantities having the following properties for different
characteristics of the microwave network or junction where characteristic impedance is same at all ports.
 Microwave Network Theory and Passive Devices 223

1. S-matrix is always a square matrix of order N ¥ N for an N-port network and its elements are
complex quantities (real and imaginary parts).
2. Phase-shift property of S-Matrix
At any given frequency and for a given positions of the reference planes, the elements of S-matrix, Sij, have
definite values.
(i) If the frequency is changed, these elements change values also.
(ii) At a given frequency, the complex S-parameters of a network are defined with respect to the
positions of the reference plan (ports). If these positions change, S-parameters of the network also
change. For an outward shift of reference plane (Fig. 6.4) by a distance lj, the phase shift occurs is
fj = bjlj and correspondingly new elements of the S-matrix become
S¢ij = Sij e- jf j ; i π j
- j 2f j
S¢ij = Sij e ;i = j
Power of exponent is doubles for i = j, since reflected wave travels 2lj distance.
Let the original S-matrix equation for the two-port network with reference planes 1, 2 be
[b] = [S][a] (6.25a)
Equation (6.25a) is modified to
[b] = [S¢][a] (6.25b)
where [S] is changed to [S¢] for outward shift in reference planes to 1¢, 2¢ but the incident and reflected wave
voltage can be still denoted by [a] and [b], respectively:
S¢11 = S11 e- j 2f1 , S22
¢ = S22 e- j 2f2 , S12
¢ = S12 e- j (f1 + f2 ) , S21
¢ = S21 e- j (f1 + f2 ) .
These can be proved as follows. For the above shift in reference planes, the new wave variables
becomes
- jf - jf
a1¢ = a1 e 1 , a2¢ = a2 e 2
b1¢ = b1 e- jf1 , b2¢ = a2 e- jf2 (6.25c)
where f = b1l1, f2 = b2l2 and –ve sign in the power of exponent arises due to outward shift of ports. For
inward shift, this sign would be +ve. Hence, we can write

È b1¢ ˘ È b1 e - jf1 ˘ È - jf1 0 ˘ È b1 ˘

Í ˙ = Í ˙ = Íe
- jf2 ˙
Í ˙ (6.25d)
Îb2¢ ˚ ÍÎb2 e - jf2 ˙˚ Î 0 e ˚ Îb2 ˚
Similarly,

È a1¢ ˘ Èe - jf1 0 ˘ È a1 ˘
Í ˙ = Í - jf2 ˙
Í ˙ (6.25e)
Îa2¢ ˚ Î 0 e ˚ Î a2 ˚
From (6.25a), (6.25b), (6.25d) and (6.25e), the new S-matrix becomes

Èe - jf1 0 ˘0 ˘ È S11 e - j 2f1

Èe - jf1 S12 e - j (f1 + f2 ) ˘
[S¢] = Í ˙ =
[S] Í
˙ Í ˙ (6.26)
Î 0 e - jf2 ˚
e - jf2 ˚ ÍÎ S21 e - j (f1 + f2 )
Î 0 S22 e - j 2f2 ˙˚
This property valid for any number of ports and is called the phase shift property.
224 Microwave Engineering

Fig. 6.4 Phase-shi property of S

3. Diagonal elements are zero for perfect matched network.

For an ideal N-port matched network with matched termination at all the ports, Sii = 0 since there is no
reflection from any port. Therefore, under perfect matched conditions, the diagonal elements of [S] are
zero.
4. S-matrix is symmetric for a reciprocal network
A reciprocal device has the same transmission characteristics in either direction of a pair of ports and is
characterized by a symmetric scattering matrix,
Sij = Sji (i π j) (6.27)
which results in transpose of S-matrix:
[S]t = [S] (6.28)
This condition can be proved in the following manner.
� Proof
For a reciprocal network with the assumed normalization
Vn = V +n + V –n In = I +n – I n–
1
Z on I n = an - bn ; I n = (an - bn )
Z on
Vn
= an + bn ; Vn = Z on (an + bn ) (6.28a)
Z on
The z-matrix equation is given by
[V] = [z] [I] (6.28b)
With the assumed normalization (6.28b) can be written using (6.28a) as
1
Z 0 = ([ a ] + [b]) = [ z ] ([ a ] - [b])
Z0
or Z0[b] + [z][b] = [z][a] – z0[a]
or [b](z0[u] + [z]) = [a] ([z] – z0[U]) (6.29c)
where [U] is the unit matrix having unity diagonal elements and zero off-diagonal elements. (6.28c) can now
becomes
[b] = ([z] – z0[U]) (z0[U] + [z])–1 [a] (6.29d)
The S-matrix equation for the network is
[b] = [S] [a] (6.30)
Comparing Eqs (6.29) and (6.30), we have
[S] = ([Z] + [U])–1 ([Z] – [U]) (6.31)
For reciprocal network, the Z-matrix is symmetric. Hence, from the symmetric of Z-matrix
([Z] – [U])t = [Z] – [U] (6.32)
 Microwave Network Theory and Passive Devices 225

([Z] + [U])t = [Z] + [U] (6.33)

Therefore, from (6.31), (6.32), and (6.33)
[S] t = ([Z] – [U])t ([Z] + [U])t–1 = ([z] – [U]) ([z] + [U])–1 = [S] (6.34)
Thus, it is proved that [S] t = [S] for a symmetrical junction.
5. S-matrix has unitary property for a lossless network
For a lossless N-port network S-matrix satisfies the following Unitary Property:
(i) The sum of the products of each term of any one row or any column of the S-matrix multiplied by
its complex conjugate (dot product) is unity
N
 Sni ◊ Snj* = 1; i= j
n =1

(ii) The sum of the products of each term of any one row or any column of the S-matrix multiplied
by the complex conjugate of a different row or column, respectively, is zero (columns are ortho-
normal)
N
 Sni ◊ Snj* = 0; iπ j
n =1

� Proof
Let us consider a 4-port lossless Network (N = 4).
The S-matrix equation can be written as
[b] = [S][a]
or, b1 = S11a1 + S12a2 + S13a3 + S14a4
b2 = S21a2 + S22a2 + S23a3 + S24a4

b4 = S41a1 + S42a2 + S43a3 + S44a4 (6.35)

For a loss-less network, the total power leaving N-ports must be equal to the total power input to these
ports, so that
N =4 N =4
 Â
2 2
bn = an
n =1 n =1

N =4 N =4
 Â
2
or, an = Sn1 a1 + Sn2 a2 + + Snn an
n =1 n =1

N =4 2
or, 2
a1 + a2 + a3 + a4
2 2 2
= Â Sn1 a1 + Sn2 a2 + Sn3 a3 + Sn4 a4 (6.36)
n =1

If only one port (ith say) is excited and all other ports are match terminated, an = 0, n π i.
N 2 N 2
 Sn  Sn
2 2
In this case, ai = i
ai = ai i
n =1 n =1
N 2
or, Â Sn i
=1 (6.37)
n =1
226 Microwave Engineering

2
*
If Sni is complex, Sn
i
= Sni ◊ Sni (6.38)

From Eqs (6.37) and (6.38),

N
 Sni ◊ Sni* = 1 (6.39a)
n =1

or, S1i ◊ S*1i + S2i ◊ S*2i + + Sni ◊ S*ni = 1; (6.39b)

For N = 4, i = 1
S11 S*11 + S21 ◊ S*21 + + S41 ◊ S*41 = 1
For N = 4, i = 2
S12 S*12 + S22 ◊ S*22 + + S42 ◊ S*42 = 1
For N = 4, i = 4
S14 S*14 + S24 ◊ S*24 + + S44 ◊ S*44 = 1 (6.39c)
Equations (6.39a)–(6.39c) prove the statement 3(i).
If two ports i and j are excited and all others are match terminated, an = 0 when n π i, n π j. Therefore,
from Eq. (6.36), we can write
N =4 2
Â
2
2
ai + a j = Sni ai + Sn j a j
n =1

2 2 2 2
= S1i ai + S1 j a j + S2i ai + S2 j a j + S3i ai + S3 j a j + S4i ai + S4 j a j (6.40a)

Since for any complex quantity |S|2 = S ◊ S*, the above expression may be multiplied out to give the
following results:
2
ai + a j
2
= [S + S + S + S \ a
11
2
21
2
31
2
41
2
i
2

2
+ [S + S \a
2 2 2 2
+ S + S
12 22 32 42 j

* * * *
+ {S11 S12 + S21 S22 + S31 S32 + S41 S42 } ai a*j
* * * *
+ {S12 S11 + S22 S21 + S32 S31 + S42 S41} a j ai*

Using Eq. (6.39a), the above equation reduces to

2 2 2 2 * * * *
ai + a j = ai + a j + {S11S12 + S21 S22 + S31S32 + S41S42 } ai a*j
* * * *
+ {S12 S11 + S22 S21 + S32 S31 + S42 S41} a j ai* (6.40b)
Equating LHS with RHS of Eq. (6.40b) yields
* *
{S11S12 + + S41S42 } ai a*j + {S12 S11
*
+ *
+ S42 S41} a j ai* = 0 (6.40c)

Since ai and aj are independent signals, they may be chosen in any convenient manner, let ai = aj. Then
from (6.40c),
* * * *
{S11S12 + + S41S42 } + {S12 S11 + + S42 S41}= 0 (6.40d)
 Microwave Network Theory and Passive Devices 227

Let ai = jaj where multiplier j = -1 , then

* * * *
{S11S12 + + S41S42 } - {S12 S11 + + S42 S41}= 0 (6.40e)

Since a*i = (jaj)* = –ja*j

Combining Eqs (6.40d) and (6.40e) results in
* * * *
S11S12 + S21S22 + S31S32 + S41S42 =0
¸
* * * * ˝ (6.40f)
or, S12 S11 + S22 S21 + S32 S31 + S42 S41 =0 ˛
In general, for an N-port network, above equations yield
N
 Sni ◊ Snj* = 0 ; iπj (6.40g)
n =1

Equation (6.40g) proves the statement 3(ii).

Equation (6.39a) and (6.40g) can be written in a single form:

[ S* ][ S ]t = [U ] or [ S* ] = [ S ]t-1 (6.41)

Therefore, (6.39a) and (6.40g) or (6.41) is the necessary condition for the network to be lossless and is
called unitary property of S-matrix. [S] is called a unitary matrix.

S-parameters to a Two-port Network with Mismatched Load 6.3.2

A two-port network or junction is formed when there is a discontinuity between the input and output ports
of a transmission line. Many configurations of such junctions practically exist, some of which are shown in
Fig. 6.5.
During propagation of a microwave through the junction from one port to another, evanescent modes are
excited at each discontinuity which contains reactive energy. Evanescent modes decay very fast away from
the junction and become negligible after a distance of the order of one wavelength. The terminal reference
planes 1 and 2 are chosen beyond this distance so that the equivalent voltage and currents at these positions
are proportional to the total transverse electric and magnetic fields, respectively, for the propagating mode
only. These circuits are analyzed using S-matrix formulation.
Consider a two-port network of Fig. 6.5(c) terminated by normalized load and generator impedances
ZL Zg
and . Then the load reflection coefficient
Z0 Z0
a2 Z L / Z 0 – 1
G2 = = (6.42)
b2 Z L / Z 0 + 1
Now, b1 = S11 a1 + S12 a2 = S11 a1 + S12 b2 G2 (6.43)
b2 = S21 a1 + S22 a2 = S21 a1 + S22 b2 G2 (6.44)

Solving for the input reflection coefficient,

S12 S21 G 2
G1 = b1/a1 = S11 + (6.45)
1 – S22 G 2
228 Microwave Engineering

Fig. 6.5 Two-port junc ons: (a) waveguide step junc on, (b) coaxial to
waveguide transi on, and (c) a two-port network model

Therefore, for a mismatch load, input reflection coefficient G1 π S11. For a reciprocal network, S12 = S21
so that
2
S12 G2
G1 = S11 + (6.46)
1 – S22 G 2
Further, if the junction is lossless, from Eqs (6.39) and (6.40),
S11 S11* + S12 S12* = 1 (6.47)
S22 S22* + S12 S12* = 1 (6.48)
S11 S12* + S12 S22* = 0 (6.49)
Therefore, for a lossless, reciprocal two-port network, terminated by a mismatch load, Eqs 6.47) and
(6.48) give
|S11| = |S22| (6.50)
From Eqs (6.49) and (6.50),
|S12| = (1–| S11 |2 ) (6.51)
 Microwave Network Theory and Passive Devices 229

and the input reflection coefficient

2
S12 G2
G1 = S11 + (6.52)
1 – S22 G 2
The last equation is the working equation for the computation of the S-parameters. By measuring G1
for known values of G2 (0, –1 and 1) a set of simultaneous equations are obtained which will give the
S-parameters of a reciprocal junction.

Comparison between [S], [Z] and [Y] Matrices 6.3.3

From the analysis of S, Z and Y parameters, it is known that [S], [Z] and [Y] represent unique intrinsic
properties of the device, supplying all the circuit characteristics of the device at a given frequency. Therefore,
terminal conditions may be changed for convenience without affecting the matrices themselves. The [S] can
be expressed in terms of [Z] and [Y] as given below.

Relation between [S] and [Z] Matrices

For a multiport network, [V] = [Z] [I]. The steady-state total voltage and the current at the kth port are
Vk = Vk+ + Vk– = Z k1 I1 + Z k2 I 2 + + Z kn I n

n
= Â Z kj I j (6.53a)
j =1

+ 1
and
–
Ik = I k – I k = (Vk+ – Vk- ) (6.53b)
Z0

or, Z0Ik = (Vk+ – Vk– )

or, Vk+ = Vk– + Z 0 I k = Vk – Vk+ + Z 0 I k

1 1Ê n ˆ
or, Vk- =
2
(Vk + Z 0 I k )= Á Â Z kj I j + Z 0 I k ˜ ;
2 ÁË j =1 ˜¯
(6.53c)

1Ê n ˆ
Similarly, Vk- = Á Â Z kj I j – Z 0 I k ˜ ; (6.53d)
2 ÁË j =1 ˜¯

È Z11 + Z 0 Z12 .... Z1n ˘

Í ˙
Z 21 Z 22 + Z 0 .... Z2n
[V+] = Í ˙ [I ]
1
Therefore,
2 Í ˙
Í ˙
ÎÍ Z n1 Z n2 .... Z nn + Z 0 ˚˙

1
= ([ Z ] + Z 0 [U ]) [ I ] (6.53e)
2
or, [I] = 2 ([ Z ] + Z 0 [U ])-1 [V + ] (6.53f)

1
Similarly, [V–] = ([ Z ] - Z 0 [U ]) [ I ]
2
244 Microwave Engineering

The shielding effectiveness of the outer conductor is expressed in terms of transfer impedance ZT of the
cable, which is defined as
Longitudinal voltage Vi induced per unit
length on one side of the shield (outside) V
ZT = = i (6.60)
The leakage current I s flowing on the other side Is
(inside) of the shield
If both ends of a short cable are terminated into
matched loads, as shown in Fig. 6.19, the load
voltage is
V IZ
V0 = i = s T (6.61)
2 2
Therefore,
ZT = 2V0/Is (6.62)
For a thin-walled tubular shield, the transfer Fig. 6.19 Transfer impedance of a coaxial cable
impedance is given by
1 (1 + j )t /d
ZT = for t << b << d (6.63)
2p bs t sinh [(1 + j )t /d ]
where b is the inner radius of the shield, t is the shield wall thickness, s is the conductivity of the shield, d is
the skin depth in the shield. For a better shielding, ZT should be smaller.
Some standard coaxial cables are given in Table 6.1 with their radio guide (RG) and universal (U)
numbers and specifications.

Table 6.1 Some standard coaxial cables

Number RG Z0 ohms Inner and outer conductor

9/U 50 Silver-plated copper
11/U 75 Tinned copper
58A/U 50 Tinned copper
59/U 75 Copper
141A/U 50 Silver-plated copper
179B/U 75 Silver-plated copper
214/U 50 Silver-plated copper

Coaxial Connectors and Adapters 6.4.2

Coaxial cables are terminated or connected to other cables and components by means of shielded standard
connectors. The outer shield makes a 360 degree, extremely low-impedance joint to maintain shielding
integrity. These connectors are of various types depending on the frequency range and the cable diameter.
Commonly used microwave connectors are type N (male/female), BNC (male/female), TNC (male/female),
APC (sexless), etc. Adapters, having different connectors at the two ends, are also made for interconnection
between two different ports in a microwave system. The basic schematic diagrams of these connectors and
adapters are shown in Fig. 6.20. The type N (Navy) connector is a 50 and 75-ohm connector which was
designed for military system applications during World War II. This is suitable for flexible or rigid cables in
the frequency range of 1–18 GHz. The BNC (Bayonet Navy Connector) is suitable for 0.25 inch, 50-ohm or
 Microwave Network Theory and Passive Devices 245

75-ohm flexible cables used up to 1 GHz. The TNC (Threaded Navy Connector) is like BNC, except that,
the outer conductor has a thread to make firm contact in the mating surface to minimise radiation leakage at
higher frequencies. These connectors are used up to 12 GHz.

(a)

Fig. 6.20 Coaxial connectors and adapters

The SMA (Sub-Miniature A) connectors are used for thin flexible or semirigid cables. The higher
frequency is limited to 24 GHz because of generation of higher-order modes beyond this limit. All the above
connectors can be of male or female configurations except the APC-7(Amphenol Precision Connector-7
mm) which provides coupling without male or female configurations. The APC-7 is a very accurate 50-ohm,
low VSWR connector which can operate up to 18 GHz.
Another APC-3.5 connector is a high-precision 50-ohm, low VSWR connector which can be either the
male or female and can operate up to 34 GHz. It can mate with the oppositely sexed SM connector. Table 6.2
shows the type, dielectric in mating space and impedance of some of the above standard connectors.
246 Microwave Engineering

Table 6.2 Ma ng connectors

Type Sex Dielectric in mating space Impedance (ohms)

N M/F Air 50/75
BNC M/F Solid 50/75
TNC M/F Solid 50/75
SMA M/F Solid 50
APC-7 Sexless Air 50
APC-3.5 Sexed Air 50

Waveguide Sections 6.4.3

Waveguide sections are used for low-loss transmission, change of direction and polarization of the signal.
Sections may be straight with rectangular or circular cross sections, rectangular twisted sections in either E
or H planes, rectangular bends, circular flexible section, rectangular or circular tapered section, and transition
from circular to rectangular cross sections. Some of them are as shown in Fig. 6.21. The standard waveguides
are made of brass, bronze or aluminium. For higher frequency bands, the surface is finished with silver
coatings to minimise ohmic loss. The S-matrix for ideal sections are
È0 1 ˘
[S] = Í ˙ (6.64)
Î1 0 ˚

Fig. 6.21 Waveguide sec ons: (a) rectangular (b) circular (c) rectangular taper

The general theory of waveguides is given in Chapter 3. This section describes some practical aspects of
rectangular and circular waveguides for the construction of waveguide components.

Rectangular Waveguide Sections

It has been shown that attenuation due to finite conductivity of waveguide walls decreases with increase of
frequency at f > fc to a minimum and increases again for a given b/a ratio. The attenuation also increases when
b/a reduces. The design requirements of a rectangular waveguide for the dominant mode are as follows:
1. Energy transmission through a single mode TE10 (dominant)
2. The breakdown power should be higher than the transmitted power
3. Low attenuation constant
For a waveguide of dimension b/a < 0.5, the successive ascending-order modes are TE10, TE20, TE01,
etc. Therefore, the dimensions of the cross section which ensure only the dominant mode propagation are
determined as follows.
lmax < lc = 2a of TE10 dominant mode
lmin > lc = a of TE20 mode
lmin > lc = 2b of TE01 mode (6.65)
260 Microwave Engineering

l0
sin (pd/a) = , there will be zero power at Port 4 in the upper guide and input power at Port 1 will be
6a
coupled into Ports 2 and 3 only. The 4-port device is then called a directional coupler or Bethe hole coupler

which is described in this chapter later.

Attenuators 6.4.14
Attenuators are passive devices used to control power levels in a microwave system by partially absorbing
the transmitted signal wave. Both fixed and variable attenuators are designed using resistive films (aquadag
coated dielectric sheet).
A coaxial fixed attenuator uses a film with losses on the centre conductor to absorb some of the power as
shown in Fig. 6.39(a). The fixed waveguide type [Fig. 6.39(b)] consists of a thin dielectric strip coated with
resistive film and placed at the centre of the waveguide parallel to the maximum E field. Induced current on
the resistive film due to the incident wave results in power dissipation, leading to attenuation of microwave
energy. The dielectric strip is tapered at both ends up to a length of more than half wavelength to reduce
reflections. The resistive vane is supported by two dielectric rods separated by an odd multiple of quarter
wavelength and perpendicular to the electric field [Fig. 6.39(b)].

Fig. 6.39 Microwave a enuator: (a) coaxial line fixed a enuator (b) and (c) waveguide a enuators

A variable-type attenuator can be constructed by moving the resistive vane by means of micrometer screw
from one side of the narrow wall to the centre where the E field is maximum [Fig. 6.39(b)] or by changing
the depth of insertion of a resistive vane at an E field maximum through a longitudinal slot at the middle of
the broad wall as shown in Fig. 6.39(c). A maximum of 90 dB attenuation is possible with VSWR of 1.05.
The resistance card can be shaped to give a linear variation of attenuation with the depth of insertion.

Precision Variable Attenuator

A precision-type variable attenuator makes use of a circular waveguide section (C) containing a very
thin tapered resistive card (R2), to both sides of which are connected axisymmetric sections of circular-
to-rectangular waveguide tapered transitions (RC1 and RC2) as shown in Fig. 6.40(a). The centre circular
section with the resistive card can be precisely rotated by 360° with respect to the two fixed sections of
circular to rectangular waveguide transitions. The induced current on the resistive card R2 due to the incident
signal is dissipated as heat and produces attenuation of the transmitted signal. The incident TE10 dominant
wave in the rectangular waveguide is converted into a dominant TE11 mode in the circular waveguide. A very
thin tapered resistive card is placed perpendicular to the E field at the circular end of each transition section
 Microwave Network Theory and Passive Devices 261

so that it has a negligible effect on the field perpendicular to it but absorbs any component parallel to it.
Therefore, a pure TE11 mode is excited in the middle section.
With reference to Fig. 6.40(b), if the resistive card in the centre section is kept at an angle q relative to
the E field direction of the TE11 mode, the component E cos q parallel to the card gets absorbed while the
component E sin q is transmitted without attenuation. This later component finally appears as electric field
component E sin2 q in the rectangular output guide. Therefore, the attenuation of the incident wave is
E 1 1
a= = =
E sin 2 q sin 2 q | S21 |
or, a (dB) = –40 log (sin q) = –20 log |S21| (6.79)
Therefore, the precision rotary attenuator produces attenuation which depends only on the angle of
rotation q of the resistive card with respect to the incident wave polarization. Attenuators are normally
matched reciprocal devices, so that
|S21| = |S12| (6.80)

Fig. 6.40 Precision-type variable a enuator

R1, R2, R3 : Tapered resis ve cards
RC1 and RC2: Rectangular-to-circular waveguide transi ons
C: Circular waveguide sec on
262 Microwave Engineering

VSWR–1
and |S11| or |S22| = << 0.1 (6.81)
VSWR + 1
where the VSWR is measured at the input or output port concerned. The S-matrix of an ideal precision rotary
attenuator is
È 0 sin 2 q ˘
[S] = Í ˙ (6.82)
ÍÎsin 2 q 0 ˙˚

Phase Shifters 6.4.15

A phase shifter is a device which provides variable insertion phase in a microwave signal path without
altering the physical path length. Most phase shifters are two-port devices characterized by low insertion
loss and low VSWR. The phase of a microwave signal can be changed by inserting a low loss dielectric
material or a ferrite material in the wave propagation path. Accordingly, phase shifters are called dielectric
phase shifters and ferrite phase shifters, respectively. Dielectric phase shifters are reciprocal phase shifters.
On the other hand, ferrite phase shifters can be reciprocal as well as non-reciprocal depending upon whether
the variable differential phase shift through the device is a function of the direction of propagation. These
are described below.

Dielectric Phase Shifters

A dielectric phase shifter can be realized by placing a lossless dielectric slab within a waveguide parallel to
and at the position of maximum E field. A differential phase change is produced due to the change of wave
1 1
velocity up = through the dielectric slab compared to that through an empty
m 0e 0e r Ê fc ˆ
2
1- Á ˜
Ë f ¯
waveguide in which er = 1. Two ports are
matched by reducing the reflections of the wave
from the dielectric slab tapered at both ends as
shown in Fig. 6.41.
The propagation constant through a length l Fig. 6.41 Phase shi ers
of a dielectric slab and of an empty guide excited
in dominant mode TE10 are, respectively,
l0
2p er
b1 = , l g1 = (6.83)
l g1 2
Ê l ˆ
1- Á 0
˜
Ë 2 a er ¯

2p l0
b2 = , l g0 = (6.84)
l g0 Êl ˆ
2
1- Á 0˜
Ë 2a ¯

Thus, the differential phase shift produced by the phase shifter is Df = (b1 ~ b0)l.
By adjusting the length l, different phase shifts can be produced.
 Microwave Network Theory and Passive Devices 263

The S-matrix of an ideal phase shifter can be expressed by

È 0 e - j Df ˘
[S] = Í ˙ (6.85)
ÍÎe - j Df 0 ˙˚

Precision Dielectric Rotary Phase Shifter

A precision phase shifter can be designed as a rotary type as shown in Fig. 6.42. This uses a section of
lg
circular waveguide c containing a lossless dielectric plate of length 2l = called half-wave (180°) section.
2
This section can be rotated over 360° precisely between two sections of circular to rectangular waveguide
lg
transitions A and B, each containing lossless dielectric plates of length l = called quarter-wave (90°)
4
sections oriented at an angle of 45° with respect to the broad wall of the rectangular waveguide ports at the
input and output. The incident TE10 wave in the rectangular guide becomes a TE11 wave in the circular guide
A. The half-wave section C produces a phase shift equal to twice its rotation angle q with respect to the
quarter-wave section. The dielectric plates are tapered through a length of quarter wavelength at both ends
for reducing reflection due to discontinuity.

Fig. 6.42 Precision dielectric rotary phase shi er

264 Microwave Engineering

The principle of operation of the rotary phase shifter can be explained as follows. The TE11 mode incident
field Ei in the input quarter-wave section can be decomposed into two transverse components, one E1,
polarized parallel and other, E2 perpendicular to quarter-wave plate. After propagation through the quarter-
wave plate, these components becomes
- jb l - jb l 1
E1 = Ei cos 45° e 1 = E0 e 1 cos 45° = sin 45° = (6.86)
2
- j b2 l
E2 = Ei sin 45° e = E0 e - j b 2 l (6.87)
where, E0 = Ei / 2 . The length l is adjusted such that these two components will have equal magnitude but
a differential phase change of (b1 – b2) l = 90°. Therefore, after propagation through the quarter-wave plate,
these field components become
- jb l
E1 = E0 e 1 (6.88)
- jb1l jp /2 - j ( b1 l - p /2)
E2 = E0 e ◊e = E0 e (6.89)
Thus, the quarter-wave sections convert a linearly polarized TE11 wave to a circularly polarized wave and
vice versa. Equations (6.88) and (6.89) are input components to half-wave section C.
After emergence from the half-wave section, the field components parallel and perpendicular to the half-
wave plate can be represented as
- j 2 b2 l - j 3 b1l
E3 = (E1 cos q – E2 sin q ) e = E0 e–jq e (6.90)
- j 2 b2 l –jq - j 3 b1l –jp /2
E4 = (E2 cos q + E1 sin q) e = E0 e e e (6.91)
since, 2 (b1 – b2)l = p or –2 b2 l = p – 2 b1l (6.92)
After emergence from the half-wave section, the field components E3 and E4 may again be decomposed
into two TE11 modes, polarized parallel and perpendicular to the output quarter-wave plate B. At the output
end of this quarter-wave plate, the field components parallel and perpendicular to the quarter-wave plate can
be written as
- jb l - j 4 b1l
E5 = (E3 cos q + E4 sin q) e 1 = E0 e–j2q e ; (6.93)
- j b2 l - j 4 b1l
E6 = (E4 cos q – E3 sin q) e = E0 e–j2q e ; (6.94)
Therefore, the parallel component E5 and perpendicular component E6 at the output end of the quarter-
wave plate are equal in magnitude and in phase to produce a resultant field which is a linearly polarized TE11
wave
- j 4 b1l
Eout = 2 E0 e–i2q e
- j 4 b1l
= Ei e–j2q e = Ei e - j (2q + 4 b1l ) (6.95)
having the same direction of polarization as the incident field Ei with a phase change of 2q + 4b1l. Since
q can be varied and 4b1l is fixed at a given frequency and structure, a phase shift of 2q can be obtained by
rotating the half-wave plate precisely through an angle of q with respect to the quarter-wave plates. The
rotation angle can be calibrated to obtain phase shift.

Ferrite Phase Shifters

It is known that most of the ferrite materials have relative dielectric constants in the range 9 to 16, and
dielectric loss tangents are normally small (< 0.001). Unlike a dielectric substrate, the permeability of the
ferrite depends on the RF magnetic field configuration in the medium relative to the static magnetization state.
Consequently, the propagation constant in the ferrite media changes with permeability since b = w me .
 Microwave Network Theory and Passive Devices 265

Phase shift in a ferrite device is generated by magnetizing the ferrite inside the wave guide by RF current
such that the phase constant in the propagating line alters due to change in permeability with magnetization
(Ref: para 2.18.3 in Chapter 2).
Ferrite phase shifters may be latching or non-latching, depending upon whether continuous holding
current must be supplied to sustain the magnetic bias field. Figure 6.43 shows a schematic diagram of a
twin-toroid phase shifter. The twin toroid is a latching, nonreciprocal device. It uses either a closed magnetic
circuit or a magnetic circuit with very small air gaps. The relative permeability of the ferrite is controlled by
adjusting the magnetic flux level existing in the closed magnetic circuit.
The dielectric spacer is used to concentrate the rf energy in the center of the waveguide. The walls of
the ferrite toroids, which contact the dielectric spacer, are located in those regions of the waveguide which
support a circularly polarized magnetic field. If b + is the propagation constant when a positive bias field
saturates the ferrite and b – be the same for negative bias field, the maximum amount of phase shift per unit
length is (b + – b –). The variable phase shift less than this can be achieved by reducing the bias field level.
When the the direction of propagation changes, the values of b – and b + interchange.

Hrf = magnetization field

Fig. 6.43 Ferrite twin-toroid phase shi er

The device geometry and the saturation magnetization of the ferrite influence the frequencies for practical
realization of these phase shifters. The conductive and dielectric losses are directly proportional to the length
or inversely proportional to the saturation magnetization while the magnetic loss varies approximately
directly with the saturation magnetization. The sum of these losses shows a minimum value in the frequency
range given by
w
0.2 £ m £ 0.6
w
where wm = 2pg (4pMs), g = 2.8 MHz/Oe, 4pMs is the saturation magnetization and w is the microwave
radian frequency. The latching phase shifters have been realized from about 2 GHz to 94 GHz.

MIC Ferrite Phase Shifter

Passive MIC phase shifters are constructed with ferrite substrate using the property that the propagation
constant of microwaves along the microstrip line on ferrite substrate can be changed by varying the magnitude
and/or the direction of the static magnetic field H0. Thus, the phase shift, for a fixed length of the microstrip
line can be controlled electrically leading to low cost, small and highly reliable components.
266 Microwave Engineering

The following tree shows the versions of ferrite MIC phase-shifters—analog and digital.

� Analog Control In analog phase shifter, the phase is controlled by varying either the magnitude or the
direction of the magnetization vector M using applied static magnetic field H0.
� Digital Control In digital phase shifter, the phase is controlled by using high remanence ferrite materials
and latching the substrate by a current pulse to change direction (±) of dc magnetic field digitally. The
remanence of magnetic materials is defined in terms of its residual magnetic induction when an externally
applied magnetizing current is reduced to zero as explained in the B–H curve in Fig. 6.44. A pulse current
of sufficient amplitude, through a single turn wire threading the ferrite, is used to magnetize the ferrite
substrate.
If n is the number of bits, then the least significant bit will have a f-shift of 360°/2n. A digital phase shifter
360∞ 360∞ 360∞
of n-bits will consist of n separately actuated sections giving 1
, 2 , º, n of f-shift. A series of
2 2 2
different Df = 180°, 90° and 45° can be cascaded in a single housing to form a 3-bit digital f-shifter.

Fig. 6.44 Ferrite remanence

Reciprocal Phase Shifters

When a ferrite substrate is magnetized to saturation in a direction perpendicular to the microstrip line,
as depicted by Hdc in Fig. 6.45, there is a very small interaction between the ferrite and the RF magnetic
field. As a result, the effective RF relative permeability is approximately unity mreff ª 1. When the ferrite
sample is magnetized to saturation in its plane parallel to the strip line (i.e., ^r to the RF magnetic field),
maximum interaction is obtained between the ferrite substrate and the RF magnetic field. Then the RF
effective permeability becomes,
 Microwave Network Theory and Passive Devices 267

m2 – K 2
mreff = (6.96a)
m
where m and K are the component of the permeability tensor (m = B/H, K = g 4pM/w, M = magnetic dipole
moment per unit volume)
When the applied magnetic field is small,
mreff ª 1– (wm/w)2 for w > g 4p Ms = wm (6.96b)
where 4pMs = ferrite magnetization, g = gyromagnetic ratio (2.8 MHz/oersted). When w Æ wm, the
magnetic loss cannot be neglected and mreff expression gets modified. The propagation constant of the TEM
mode on the ferrite substrate is given by

b = w e 0 m0 e reff m reff (6.96c)

The different values of mreff for the two mutually perpendicular directions of magnetizations yield
two different values of the phase constant for wave propagation along the microstrip line on the ferrite
substrate.
Writing in terms of a filling factor q as proposed by wheeler for dielectric substrate.
eeff = 1 + qe (er – 1); qe ª 0.6 (6.96d)
1
meff = (6.96e)
Ê 1 ˆ
1 + qm Á – 1˜
Ë mr ¯
–1
meff –1
qm = (6.96f)
m –1 – 1
If magnetic filling factor qm = 0.5, then for parallel magnetization
b11 @ w m0 e 0 e reff [{2 – 2(w m /w )2} {2 – (w m / w )2}]1 2 (6.96g)
and for perpendicular magnetization
b^ ª w m0 e 0 e reff (6.96h)
When ferrite magnetization direction changed between two states, the fractional change in phase constant
is
Df Db 1/2
= = 1 – [{2 – 2(w m /w )2} {2 – (w m / w )2}] (6.96i)
f b

Fig. 6.45 Parallel and perpendicular magne za on

268 Microwave Engineering

Both analog and digital controlled reciprocal phase shifters can be

constructed using a loosely meander line microstrip conductor to minimise
size as shown in Fig. 6.46. The adjacent lines of the meander scheme are
spaced by distance > 5 times substrate height to minimize RF coupling and
consequently to avoid non-reciprocal effects. DC current through
perpendicular wires penetrating holes 1–1¢ and 2–2¢ generates magnetic
fields for magnetization which can be varied, and also the direction can be
changed as per requirement of phase change. In digital phase shifter two
states of magnetization corresponding to two values of f-shifts are obtained
by sending dc currents through 1–1¢ and 2–2¢, respectively (2–2¢ for parallel Fig. 6.46 Reciprocal f-shi er
and 1–1¢ for perpendicular magnetizations). schema c

Non-reciprocal Phase Shifter

In a non-reciprocal f-shifters, a circularly polarized magnetic field region is obtained by using a meander
microstrip configuration with light coupling between the lines as shown in Fig. 6.47. This is achieved by
closely spacing the lines of length l/4 at the centre frequency. Because of tight coupling, the two magnetic
field components at a centre point P between two adjacent lines become spatially orthogonal. Since lines
are l/4 long, the RF current flowing at the centre of line 2 is delayed by 90° relative to the current flowing
at the centre of line 1. Thus, the resulting RF magnetic field at P is circulary polarized. However, at points
away from the centre of the lines, the polarization changes from circular to elliptical and then to linear at
the ends.

Fig. 6.47 Non-reciprocal phase shi er: (a) Top view of strip (b) Side view from the plane of the strip

By magnetizing the ferrite along the direction of the lines and reversing the direction of the magnetization,
a non-reciprocal differential interaction is obtained resulting in a non-reciprocal differential phase shift.
Digital/latching phase shifter can be obtained by using sufficiently thin ferrite and small wires through holes
in the substrate for dc currents I1 and I2. The magnitude and direction of magnetization can be altered by
current pulses. Typical parameters for such a phase shifter are the following:
Phase shift—1 bit (360° + 10°)
 Microwave Network Theory and Passive Devices 269

Substrate size—1¢¢ ¥ 1¢¢

Substrate height = 50 mil; 1 mil = 0.254 mm
Insertion loss = 2.4 dB over 10% bandwidth
Centre frequency = 5.75 GHz

Waveguide Tees 6.4.16

Waveguide tees are three-port components. They are used to connect a branch or section of the waveguide
in series or parallel with the main waveguide transmission line for providing means of splitting, and also of
combining power in a waveguide system. The two basic types, E-plane (series) T and H-plane (shunt) T, are
constructed as shown in Fig. 6.48. These are named according to the axis of the side arm which is parallel to
the E field or the H field in the collinear arms, respectively.
Because of the junction, waveguide tees are poorly matched devices. Adjustable matching reactance can
be introduced by means of a tuning screw at the centre. Because of symmetry and absence of non-linear
elements in the junction, the S-matrix is symmetric: Sij = Sji; i = 1, 2; j = 1, 2. The general S-matrix for a tee
junction is

Fig. 6.48 Waveguide tees: (a) E-Tee or Series-T (b) H-Tee or Shunt-T

È S11 S12 S13 ˘

Í ˙
[S] = Í S12 S11 S23 ˙ (6.97)
ÍÎ S13 S23 S33 ˙˚

E-Plane Tee
From considerations of symmetry and the phase relationship of the waves in each of the arms, it can be seen
that a wave incident at Port 3 will result in waves at Ports 1 and 2, which are equal in magnitude but opposite
in phase, i.e., S31 = S13 = – S23 = – S32, S12 = S21. If two in-phase input waves are fed into Ports 1 and 2 of the
collinear arm, the output waves at Port 3 will be opposite in phase and subtractive. Sometimes, this third port
is called the difference arm. By analogy with the voltage relationship in the series circuit, E-plane junction is
also called a series junction. All diagonal elements of the S-matrix of an E-plane T-junction cannot be zero
simultaneously since the tee junction cannot be matched to all the three arms simultaneously. Considering
the Port 3 as matched, the S-matrix of a E-plane T can be derived as follows.
270 Microwave Engineering

Denoting the incident and outgoing signal variables as the ith port by ai and bi, respectively, for an input
1 1
power at Port 3, the net input power to Port 3 is (|a3|2 – |b3|2) = |a3|2 (1 – |S33|2), and the output power
2 2
1
is (|b1|2 + |b2|2) = |a3|2 ◊ |S13|2, since |S31| = |S32| by symmetry. For lossless junction, input power must be
2
equal to the output power, i.e.,
1
|a3|2 (1 – |S33|2) = |a3|2 ◊ |S13|2 or, (1 – |S33|2) = 2 |S13|2
2
By suitable matching elements, we can make S33 = 0, so that |S13| = 1/ 2 . From the symmetry
characteristics described above,
S13 = S31 = 1/ 2 , S23 = S32 = – 1/ 2
After matching Port 3, if one attempts to match either Port 1 or 2 by a similar method, the matching
elements, such as irises or tuning screws will interact with each other and matching at Port 3 would be
disturbed since no port is isolated. Based on power consideration, it can also be shown that S11 = S22 = 1/2
and S12 = S21 = 1/2 for S33 = 0. Therefore, with matching at Port 3, the S-matrix of an E-plane T can be
expressed by real values with proper choice of reference plane:
È 1/2 1/2 1/ 2 ˘ È 1 1 2 ˘
Í ˙ 1Í ˙
[S] = Í 1/2 1/2 1/ 2 ˙ = Í 1 1 - 2˙ (6.98)
Í ˙ 2Í ˙
ÍÎ1/ 2 -1/ 2 0 ˙˚ ÍÎ 2 - 2 0 ˙˚
There are three cases to describe below:
� Case 1 Input at port 3 and no inputs at port 1 and 2, a3 π 0, a1 = a2 = 0.
1
b1 = a3
2
1
b2 = – a3
2
b3 = 0
� Case 2 Input at port 1 and Port 2, and no input at port 3. a3 = 0, a1 = a2 = a.
a a
b1 = + = a
2 2
a a
b2 = + =a
2 2
b3 = 0
� Case 3 Input at port 1 and Port 2, and no input at port 3. a1 π 0, a3 = a2 = 0.
a1
b1 =
2
a
b2 = 1
2
a1
b3 = -
2
Similarly, we have all combinations of input and output.
 Microwave Network Theory and Passive Devices 271

H-Plane Tee
In an H-plane tee, if two in-phase input waves are fed into ports 1 and 2 of the collinear arm, the output waves
at Port 3 will be in-phase and additive. Because of this, the third port is called the sum arm. Conversly, an
input wave at Port 3 will be equally divided into ports 1 and 2 in phase. Because the magnetic field loops
get divided into two arms 1 and 2 in a manner similar to currents between branches in the parallel circuit,
an H-plane junction is also called a shunt junction. For a symmetrical and lossless junction, in absence of
non-linear elements at the H-plane junction, the S-parameters are obtained in a similar manner as in the case
of E-plane junction:
Since it is a three-port junction, the scattering matrix can be derived as follows:
[S] Matrix of order 3 ¥ 3.
È S11 S12 S13 ˘
[S] = ÍS21 S22 S23 ˙ (6.99a)
Í ˙
ÍÎ S31 S32 S33 ˙˚

Because of plane of symmetry of the junction, the Scattering coefficients are

S23 = S13
As the waves coming out of port 1 and 2 of the collinear arm will be opposite in phase and in same
magnitude. The negative sign indicates phase difference.
If Port 3 is perfectly matched to the junction, S33 = 0
For symmetric property, Sij = Sji
\ S12 = S21: S13 = S31: S23 = S32 = S13 (6.99b)
With the above properties. [S] becomes,
È S11 S12 S13 ˘
Í ˙
[S] = Í S12 S22 S13 ˙ (6.99c)
ÍÎ S13 S13 0 ˙˚
For the symmetry and lossless properties,
[S]t = [S] and [S]* [S]t = [S]* [S] = [U] (6.99d)

or, |S11|2 + |S12|2 + S13|2 = 1

|S12|2 + |S22|2 + S13|2 = 1
0 + |S13|2 + |S13|2 = 1
S13 ◊ S*11 + S13 ◊ S*12 = 0 (6.99e)
From first two equations of (6.99e), we get
S11 = S22 (6.99f)
From third equation in (6.99c),
1
S13 = = S23 = S32 (6.99g)
2
From last equations of (6.99e), we get
* *
S13(S11 + S12 )=0

but S13 π 0 \ (S*11 + S12

*
)=0
272 Microwave Engineering

\ S12 = –S11

Substituting these values in the first equation of (6.99e),

2 2 1
S11 + S11 + =1
2
1
or, S11 = = S22
2
1
S12 = -
2
Substituting these values in (6.99c), we get S-matrix of the H-plane Tee

È 1 -1 2˘
1Í ˙
[S] = Í -1 1 2˙ ((6.99h)
2
Í 2 0 ˙˚
Î 2

when Port 3 is match terminated. Two cases can be described:

� Case 1 Input is given at Port 3 and no inputs at Ports 1 and 2, a3 π 0, a1 = a2 = 0
1
b1 = a3
2
1
b2 = a3
2
b3 = 0
Hence, the power coming out of ports 1 or port 2 is 3 dB down with respect to input power at Port 3, hence
the H-plane Tee is called 3 dB splitter.

� Case 2 Input is gives at Port 1 and Port 2, and no input at Port 3, a3 = 0, a1 = a2 = a.

a a
b1 = - =0
2 2
a a
b2 = - =0
2 2
a a
b3 = + = 2a
2 2

Output at Port 3 is the addition of the two inputs at Port 1 and Port 2 and these are added in phase.

Example 6.17 Consider a lossless two-port network. (a) If the network is reciprocal, show that
|S21|2 = 1 – |S11|2. (b) If the network is nonreciprocal, show that it is impossible to
have unidirectional transmission, where S12 = 0 and S21 π 0.

È S11 S21 ˘
(a) [S] = Í ˙ S12 = S21 since reciprocal
ÎS21 S22 ˚
 Microwave Network Theory and Passive Devices 273

[S] is unitary if lossless, so

First row: |S11|2 + |S21|2 = 1 (or 1st col)
|S21|2 = 1 – |S11|2

ÈS S ˘
(b) [S] = Í 11 21 ˙ S12 π S21 since nonreciprocal
Î 0 S22 ˚
First row: |S11|2 + |S21|2 = 1
First Column: |S11|2 = 1
\ |S21| = 0

Example 6.18 A 20 mW signal is fed into one of the collinear (port 1) of a lossless H-plane
T-junction. Calculate the power delivered through each port when other ports are
terminated in matched load.
Solution Since Ports 2 and 3 are matched terminated, a2 = a3 = 0, |S11| = 1/2. The total effective power
input to Port 1 is
1
P1 = |a1|2 (1 – |S11|2)
2
= 20 (1 – 0.52) = 15 mW
The power transmitted to Port 3 is
1
P3 = |a1|2 |S31|2
2
= 20 ¥ (1/ 2 )2 = 10 mW
The power transmitted to Port 2 is
1
P2 = |a1|2 |S21|2
2
= 20 ¥ (1/2)2 = 5 mW.
Therefore, P1 = P3 + P2

Example 6.19 In an H-plane T-junction, compute power delivered to the loads of 40 ohms and 60
ohms connected to arms 1 and 2 when a 10 mW power is delivered to the matched
port 3.
Solution With Port 3 matched, the scattering matrix for H-plane T is
È 1 -1 2˘
Í ˙
[S] = 1/2 Í -1 1 2˙
Í ˙
ÍÎ 2 2 0 ˙˚

Therefore, input power at Port 3 is equally divided in arms 1 and 2. Since input at Port 3 = 10 mW =
0.01 W, power towards ports 1 and 2 = 0.005 W = (1/2) |b1|2 = (1/2) |b2|2. Considering first-order reflection,
reflected power from ports 1 and 2 are
(1/2) |G1b1|2 and (1/2) |G2b2|2
274 Microwave Engineering

Therefore, power delivered to load Z1 = 40 ohms and Z2 = 60 ohms are

P1 = (1/2) |b1|2 – (1/2) |G1b1|2 = (1/2) |b1|2 (1 – |G1|2)
and P2 = (1/2) |b2|2 – (1/2) |G2b2|2 = (1/2) |b2|2 (1 – |G2|2)
Now taking the characteristic impedance of the line = 50 ohms
|G1| = |40 – 50| / |40 + 50| = 1/9 ; |G1|2 = 0.01234
|G1| = |60 – 50| / |60 + 50| = 1/11 ; |G2|2 = 8.2694 ¥ 10–3
Therefore, P1 = 0.005 (1 – 0.01234) = 4.938 ¥ 10–3 = 4.9383 mW
P2 = 0.005 (1 – 8.2694 ¥ 10–3) = 4.9586 ¥ 10–3 W
= 4.9586 mW

Hybrid or Magic-T
A hybrid tee is formed with the combination of the E-plane and H-plane tees and is called a magic-T. It has
four ports as shown in Fig. 6.49(a) and 6.49(b).

Fig. 6.49 Magic-T s with ports 3 and 4 interchanged

The magic-T has the following characteristics when all the ports are terminated with a matched loads. Let
us consider the structure (a) with the ports as indicated.
1. If two waves of equal magnitude and equal phase are fed into ports 1 and 2, the output at Port 3 is
subtractive and becomes zero and total output will appear additively at the port 4. Hence, Port 3 is
called the difference or E-arm and 4, the sum or H-arm.
2. A wave incident at Port 3 (E-arm) divides equally between ports 1 and 2 but is opposite in phase
with no coupling to Port 4 (H-arm). Thus,
S13 = –S23, S43= 0 (6.100)
3. A wave incident at Port 4 (H-arm) divides equally between ports 1 and 2 in phase with no coupling
to port 3 (E-arm). Thus,
S14 = S41 = 1/ 2 = S24 = S42 and S34 = 0 (6.101a)
4.A wave fed into one collinear port, 1 or 2, will not appear in the other collinear Ports, 2 or 1,
respectively. Hence, two collinear ports 1 and 2 are isolated from each other, making
S12 = S21 = 0 (6.101b)
A magic-T can be matched by putting tuning screws suitably in the E and H-arms without destroying the
symmetry of the junctions. Therefore, for an ideal lossless magic-T matched at ports 3 and 4, S33 = S44 = 0.
The procedure of derivation of the S-matrix considers the symmetry property at the junction for which S14
 Microwave Network Theory and Passive Devices 275

= S41 = S24 = S42, S31 = S13 = –S23 = –S32, S34 = S43 = 0, S12 = S21 = 0. Therefore, the S-matrix for a magic-T,
matched at ports 3 and 4 given by
È S11 S12 S13 S14 ˘
Í ˙
S S22 - S13 S14 ˙
[S] = Í 12 (6.102)
Í S13 - S13 0 0 ˙
Í ˙
Î S14 S14 0 0 ˚
From the unitary property applied to rows 1 and 2, we get
|S11|2 + |S12|2 + |S13|2 + |S14|2 = 1 (6.103)
2 2 2 2
|S12| + |S22| + |S13| + |S14| = 1 (6.104)
Subtracting these two equations:
|S11|2 – |S22|2 = 0 or, |S11| = |S22| (6.105)
Form the unitary property applied to rows 3 and 4,
2 |S13|2 = 1, or |S13| = 1/ 2 (6.106)
2
2 |S14| = 1, or |S14| = 1/ 2 (6.107)
Substituting these values in Eq. (6.103),
|S11|2 + |S12|2 + 1/2 + 1/2 = 1 or, |S11|2 + |S12|2 = 0 (6.108)
which is valid if S11 = S12 = 0 (6.109)
From Eqs (6.105) and (6.109), S22 = 0 (6.110)
È 0 0 S13 S13 ˘
Í ˙
0 0 - S13 S13 ˙
Therefore, [S] = Í (6.111)
Í S13 - S13 0 0 ˙
Í ˙
ÍÎ S13 S13 0 0 ˙˚
where |S13| = 1/ 2 = |S14|
By proper choice of reference planes in arms 3 and 4, it is possible to make both S13 and S14 real, resulting
in the final form of S-matrix of magic-T.
È0 0 1 1˘
Í ˙
0 0 -1 1˙
[S] = 1/ 2 Í (6.112a)
Í 1 -1 0 0˙
Í ˙
Î1 1 0 0˚
For the structure (b) where ports 3 and 4 are interchanged, the S-matrix becomes
È0 0 1 1˘
Í ˙
0 0 1 -1˙
[S] = 1/ 2 Í (6.112b)
Í1 1 0 0˙
Í ˙
Î 1 -1 0 0˚

Example 6.20 A magic-T is terminated at collinear ports 1 and 2 and difference port 4 by
impedances of reflection coefficients G1 = 0.5, G2 = 0.6 and G4 = 0.8, respectively.
If 1W power is fed at the sum port 3, calculate the power reflected at Port 3 and power transmitted to the
other three ports.
276 Microwave Engineering

Solution S-matrix for a matched magic-T with collinear ports 1 and 2 and sum and difference ports 3
and 4, respectively, is given by
È0 0 1 1˘
Í ˙
0 0 1 -1˙
[S] = 1/ 2 ÍÍ
1 1 0 0˙
Í ˙
Î 1 -1 0 0˚
If a1, a2, a3 and a4 be the normalized input voltages and b1, b2, b3, and b4 are the corresponding output
voltage at ports 1, 2, 3, and 4 respectively, then
a1 = G1b1, a2 = G2b2, a3 = input applied voltage at port 3, and
a4 = G4b4.
Now, Pi = |a3|2/2 = 1W, or, a3 = 2V
Therefore,
È b1 ˘ È0 0 1 1˘ È .5b1 ˘
Í ˙ Í ˙Í ˙
Í b2 ˙ = 1/ 2 Í0 0 1 -1˙ Í.6b2 ˙
Í b3 ˙ Í1 1 0 0˙ Í 2 ˙
Í ˙ Í ˙Í ˙
Îb4 ˚ Î 1 -1 0 0 ˚ ÍÎ.8b4 ˙˚

or, b1 + 0 + 0 – 0.8 b4/ 2 = 1

0 + b2 + 0 + 0.8 b4/ 2 = 1
– 0.5b1/ 2 – 0.6 b2/ 2 + b3 + 0 = 0
– 0.5b1/ 2 + 0.6 b2/ 2 + 0 + b4 = 0
The unknown quantities b’s may be solved by Cramer’s rule.

È 2 0 0 - 0.8 ˘
Í ˙
1 Í 2 2 0 0.8 ˙
Í ˙
2 Í 0 - 0.6 2 0 ˙
Í ˙
Î0 0.6 0 2 ˚ 2.0782
b1 = = = 0.928 V
È 2 0 0 - 0.8 ˘ 2.2381
Í ˙
1 Í 0 2 0 0.8 ˙
Í ˙
2 Í- 0.5 - 0.6 2 0 ˙
Í ˙
Î- 0.5 0.6 0 2 ˚

2.3981
Similarly, b2 = = 1.07 V
2.2381
1.7524
b3 = = 0.78 V
2.2381
–0.2827
b4 = = – 0.126 V
2.2381
 Microwave Network Theory and Passive Devices 277

Therefore,
1
Power transmitted at Port 1 = |b1|2 = 0.4309 W
2
1
Power transmitted at Port 2 = |b2|2 = 0.5738 W
2
1
Power transmitted at Port 4 = |b4|2 = 0.00797 W
2
1
Power reflected at Port 3 = |b3|2 = 0.3065 W
2
Note: Power absorbed at Port i = 1/2 (|bi|2 – |ai|2); i = 1, 2, 4. Total power absorbed by the system
1
= (|a3|2 – |b3|2) .
2

Application of Magic-T
The magic-T has a number of applications in various microwave circuits, such as the E-H tuner for impedance
matching, balanced mixer in a microwave superheterodyne to balance out the local oscillator noise at the IF
amplifier input, power combiner and duplexer.
� E-H Tuner In an E-H tuner (Fig. 6.50), both the E and
H-arms are terminated by movable shorts which act as E-plane
and H-plane stubs. The position of the shorts can be adjusted so
that a wide range of load impedance may be matched to reduce
the VSWR of a waveguide system connected through the collinear
arms.
� Balanced Mixer In a balanced microwave mixer
configuration, an incoming signal is fed to the E-arm and a local
oscillator signal is fed to the H-arm as shown in Fig. 6.51. When Fig. 6.50 E-H tuners
these two signals enter the collinear arms, the crystal diodes
placed in these arms produce the IF signal or difference signal in
the following manner.

Fig. 6.51 Microwave mixer

278 Microwave Engineering

The local oscillator signal from the H-arm will arrive at the diodes in collinear arms in-phase, whereas
the incoming signal from the E-arm will arrive at the diodes out-of-phase. These signals are mixed in the
non-linear diodes and produce IF signals in the collinear arms which are out-of-phase by 180°. Since local
oscillator noise will be in-phase at the diodes, this gets cancelled at the balanced IF input to IF amplifier
whereas, the IF signals are added up in phase for amplification in IF amplifier. LO and RF signals are
uncoupled due to magic-T properties of E and H-arms. For equal power inputs at isolated ports 3 and 4, P3
1
= P4 = P = |a|2, where a3 = a4 = a. If Port 1 is matched, a1 = 0. Therefore,
2
b2 = (0 + 0 – a + a)/ 2 = 0.
and hence a2 = 0. Consequently b3 = b4 = 0. Thus
b1 = S13 a + S14 a = 2 a
1
Therefore, output power at Port 1 is P1 = |b1|2 = |a|2 = 2P = sum of two equal input powers.
2
In a duplexer circuit of a radar system, a common antenna is connected to Port 1 while the transmitter and
receiver are connected to isolated E and H-arms and a dummy load is connected at Port 2. Half of the power
transmitted is coupled to the antenna and 50 % of the received power gets into the receiver, remaining 50 %
power is absorbed in dummy load.

Ferromagnetic Insert and Component 6.4.17

Ferromagnetic materials (ferrite: Mg + Mn, Ni + Zn alloys) when placed in dc magnetic field, electromagnetic
wave propagation becomes non-reciprocal. This property is used for construction of circulators and
isolators.

Circulators
A circulator is a multiport junction in which
the wave can travel from one port to the next
immediate port in one direction only as shown
in Fig. 6.52(a). Commonly used circulators are
three-port or four-port passive devices although
more number of ports is possible.
� Four-port Circulator A four-port circulator
can be constructed from two magic-T’s and a non-
reciprocal 180° phase shifter or a combination of
two 3 dB side hole directional couplers with two
non-reciprocal phase shifters as shown in Figs
6.52 (b) and (c).
In Fig. 6.52 (b), an input signal at Port 1 is split
into two in-phase and equal amplitude waves in Fig. 6.52(a) Schema c diagram of a four-port circulator
the collinear arms b and d of the magic-tee, T1 and
added up to emerge from Port 2 in the magic tee, T2. On the other hand a signal at Port 2 will split into two
equal amplitude and equiphase waves in the collinear arms of the magic-tee, T2 and appears at point b and d
out of phase due to presence of the non-reciprocal 180° phase shifter. These out-of-phase waves add up and
appear from Port 3 in the magic-tee, T1. In a similar manner, an input signal at Port 3 will emerge from 4, an
input at Port 4 will appear at Port 1. Thus, the circulator property is exhibited.
 Microwave Network Theory and Passive Devices 279

In Fig. 6.52(c), each of the two 3 dB couplers introduces a 90° phase shift. An input signal at Port 1 is
split into two components by the coupler 1 and the coupled signals are again split into two components by
the coupler 2 with a 90° phase shift in each. Each of the two phase shifters produces additional phase shift
so that the signal components at Port 2 are in phase, and at Port 4, they are out of phase. Since Port 3 is the
decoupled port for the directional coupler, the input signal at Port 1 appears in Port 2. Similarly, signals from
Port 2 to Port 3, from Port 3 to Port 4 and from Port 4 to Port 1.
A perfectly matched, lossless, and non-reciprocal four-port circulator has S-matrix:
È0 0 0 1˘
Í ˙
1 0 0 0˙
[S] = Í (6.113)
Í0 1 0 0˙
Í ˙
Î0 0 1 0˚

Fig. 6.52(b–c) Four-port circulators

� Three-port Circulator A three-port circulator is formed by a 120° H-plane waveguide or strip line
symmetrical Y-junction with a central ferrite post or disc. A steady magnetic field H0 is applied along the
280 Microwave Engineering

axis of the disc as shown in Fig. 6.53. Depending on the polarization of the incident wave and the direction
of H0, the microwave signal travels from one port to the immediate next one only.
For a perfectly matched, lossless, non-reciprocal three-port circulator, the S-matrix is
È 0 0 S13 ˘
Í ˙
[S] = Í S21 0 0 ˙ (6.114)
ÍÎ 0 S32 0 ˙˚
If the terminal planes are properly chosen to make the phase angles of S13, S21 and S32 zero,
S13 = S21 = S32 = 1
so that
È0 0 1 ˘
Í ˙
[S] = Í1 0 0 ˙ (6.115)
ÍÎ0 1 0 ˙˚

Fig. 6.53 Three-port circulator: (a) Waveguide type (b) Stripline type

The matching of the junction can be achieved by placing suitable tuning elements in each arm.
Since, in practice, losses are always present, the performance is limited by finite isolation and non-zero
insertion loss. Typical characteristics can be represented by
Insertion loss < 1 dB
Isolation ª 30–40 dB
VSWR < 1.5

Example 6.21 Prove that it is impossible to construct a perfectly matched, lossless, reciprocal
three-port junction.
Solution A perfectly matched three-port junction has a symmetric scattering matrix with zero diagonal
elements:
È 0 S12 S13 ˘
Í ˙
[S] = Í S12 0 S23 ˙
ÍÎ S13 S23 0 ˙˚
 Microwave Network Theory and Passive Devices 281

For a lossless junction, S-matrix is unitary, so that

S12 S12* + S13 S13* = 1 (1st row)
S12 S12* + S23 S23* = 1 (2nd row)
* *
S13 S13 + S23 S23 = 1 (3rd row)
* *
and S13 S23 = S12 S23 = S12 S13* = 0
If S12 is not equal to zero, the fourth equation from above gives S13 = 0 = S23. But this does not satisfy the
third equation. Therefore, it is not possible to construct a perfectly matched, lossless, reciprocal three-port
network.

Example 6.22 A three-port circulator has an insertion loss of 1 dB, isolation-30 dB and VSWR =
1.5. Find the S-matrix.
Solution The S-matrix of a three-port circulator is
È S11 S12 S13 ˘
Í ˙
[S] = Í S21 S22 S23 ˙
ÍÎ S31 S32 S33 ˙˚
Insertion loss = 1 dB = – 20 log |S21| Fig. 6.54 Three-port circulator for
–1/20
Example 6.22
or, |S21| = 10 = 0.89
For the same insertion loss between ports 1 and 2, 2 and 3, 3 and 1, |S21| = |S32| = |S13| = 0.89.
The isolation between the ports is 30 dB = – 20 log |S31|
or, |S31| = 10–30/20 = 10–1.5 = 0.032
= |S23| = |S12|
Since VSWR S = 1.5, reflection coefficient
S - 1 1.5 - 1
|G| = = = 0.2 = |S11|
S + 1 1.5 + 1
= |S22| = |S33|
By placing reference planes suitably to make the phase of S-parameters zero, the S-matrix becomes
È0.200 0.032 0.890 ˘
[S] = Í0.890 0.200 0.032 ˙
Í ˙
ÎÍ0.032 0.890 0.200 ˙˚

� Microstrip Circulator A microstrip circulator is

constructed in the same form as strip line one without having
the top ground plane. The entire circulator can be made by
using the whole ferrite substrate. The basic design criteria
are
∑ Selection of radius R of ferrite disc
∑ Calculation of radius R of magnet cylinder for the
whole ferrite substrate.
The value of R is given by
1.84
R= (6.116a)
w 0 (e 0e r m0 meff )
Fig. 6.55 Microstrip circulator
282 Microwave Engineering

where w0 = Centre angular frequency

e0 = Free-space permittivity
er = Relative permittivity of the ferrite
m0 = Free-space permeability
meff= (m2 – K2)/m = scalar effective permeability
m = 1 – ps/(1 – s2)
p
K =
1–s2
p = |g |Ms/w = Normalized saturation magnetization
= Ms/H0
s = |g |H0/w = Normalized biasing magnetic field
= Hdc/H0

w
H0 = = Field required for gyromagnetic resonance in the infinite ferrite medium
|g |
For low magnetic loss, ferrite material selection is done such that
g 4pMs/w = 0.6 (6.116b)
The insertion loss between coupled ports is determined by
1. Copper loss of the strip and ground plane
2. Dielectric loss of the input/output strips
3. Magnetic loss of the ferrite disc
The isolation between the ports is primarily dependent on the mismatching between the ports and the
junction. Tuning screws, l/4 transformer or tapered strips can be used for matching and this can be the parts
of the mask used in photo-etching process. The magnetic disc can be located above or below the substrate.
Using YIG substrate of thickness h = 0.055¢¢ one can get Isolation > 20 dB, VSWR < 1.2, insertion loss
1 L < 0.8 dB, power handling capability of 60 W over the frequency range 8.5–9.9 GHz.
The power-handling capability of such a device can be increased by
1. Lowering the impedance or increasing intrinsic line width
2. Increasing substrate thickness h
3. Decreasing 4pMs of the material by substituting Al ions in the YIG material.

Isolators
An isolator is a two-port, non-reciprocal device which produces a minimum attenuation to wave propagation
in one direction and very high attenuation in the opposite direction. Thus, when inserted between a signal
source and load, almost all the signal power can be transmitted to the load and any reflected power from
the load is not fed back to the generator output port. This eliminates variations of source power output and
frequency pulling due to changing loads.
An isolator can be constructed in a rectangular waveguide (a ¥ b) operating in dominant mode as shown
in Fig. 6.56(a). The non-reciprocal characteristics are obtained by establishing a steady magnetic field H0 in
the y direction and placing a ferrite slab at any of the longitudinal planes x = x1 near and parallel to the narrow
waveguide wall, where the magnetic field exhibits circular polarization. This occurs at x1 = a/4 or, 3a/4.
For the propagation of waves in +z direction, direction of rotation of H in the planes at x1 = a/4 and 3a/4
are opposite to each other. The non-reciprocal characteristic is achieved by placing a ferrite slab at any one
of these two planes. The required steady state magnetic field H0 in the y-direction is established by placing
permanent magnetic poles between the two broad walls.
 Microwave Network Theory and Passive Devices 283

Fig. 6.56(a) Waveguide isolator

It is known that the attenuation in ferrite for negative/clockwise circular polarization is very small whereas
for positive/counter clockwise circular polarization is very large at and near the resonance frequency f ª f0.
Therefore, the ferrite slab is placed in such a way that while transmission it encounters negative circular
polarization in the reverse direction. The steady magnetic field is set to be equal to the resonant value. The
isolation of the order of 20 – 30 dB in the backward direction and a transmission loss of 0.5 dB in the forward
direction can be achieved with a VSWR of the order of 1.1.
Since the reverse power is absorbed in the ferrite and dissipated as heat, the maximum power handling
capability of an isolator is limited. To increase the capacity of heat dissipation, two ferrite slabs of smaller
heights are used instead of one with a larger height.
The requirement of very high steady magnetic field (10,000 oersteds at l = 1 cm) is the main drawback
of waveguide resonance isolator at higher frequencies.
� Faraday Rotation Isolator A Faraday rotation
isolator is a circular waveguide section axially loaded
with a ferrite rod of smaller diameter as shown in
Fig. 6.56(b).
The ferrite rod is subjected to a steady axial magnetic
field H0 of strength much smaller than the resonant
intensity so that dissipative loss in the ferrite is neglected.
The dominant TE11 mode in the circular section can
be decomposed into two oppositely rotating circularly
polarized waves of equal magnitude. These waves
encounter different permeabilities m¢+ and m¢– for the
clockwise and anticlockwise directions of field rotation
and exhibit changes in the phase velocities. This will
result in a change in the plane of polarization of the main
mode TE11 which will experience gradual rotation q
during propagation to the other end. The rotation angle q
is proportional to the length of the ferrite rod. In this case
for the reverse wave the direction of rotation remains the Fig. 6.56(b) Faraday rota on isolator
same confirming the non-reciprocal characteristics of the
ferrite.
The isolator input is a 45° twist where a tapered resistive card is mounted parallel to the broad wall of
the rectangular waveguide part. The dominant TE10 mode does not get attenuated while transmission and is
284 Microwave Engineering

rotated at 45° at the twist output and enters the circular waveguide through rectangular to circular waveguide
transition as the TE11 mode. The length of the ferrite rod is selected so as to obtain Faraday rotation q = 45°
at the output and regain its original polarization. The plane of polarization of the reflected wave from the
load is again rotated by the same angle q = 45° and at the emergence through the 45° twist becomes aligned
with the surface of the absorbing plate and gets absorbed. Thus, non-reciprocal isolation action lakes place.
Typical insertion loss and isolation are approximately 1dB and 20~30 dB, respectively, for these isolators.
Isolators are also available in the coaxial and microstrip forms.
For an ideal lossless, matched isolator
|S21| = 1, |S12| = |S11| = |S22| = 0 (6.117a)

È0 0 ˘
i.e., [S] = Í ˙ (6.117b)
Î1 0 ˚

Example 6.23 A matched isolator has insertion loss of 0.5 dB and an isolation of 25 dB. Find the
scattering coefficients.

Solution
Insertion loss = 0.5 dB = –20 log |S21|
or, S21 = 10–.5/20 = 10–0.025 = 0.9441
The isolation is 25 dB = –20 log |S12|
or, S12 = 10–25/20 = 10–1.2 = 0.0631
Since there is no reflection, S11 = S22 = 0. Therefore, the S-matrix for the isolator is
È 0 10 -1.2 ˘ È 0 0.0631˘
[S] = Í - 0.025 ˙=Í
ÍÎ10 0 ˙˚ Î0.9441 0 ˙˚

� Microstrip Isolator A microstrip isolator can be formed by match terminating one port of a 3-port
circulator. Several termination methods are described with the help of Fig. 6.57(a–d).
(a) Method 1 A chromium layer is deposited on ferrite initially in all circuits for adherence. A tapered
Cr film of 0.6¢¢ long with a surface impedance of 270-ohm termination gives VSWR < 1.2 over the
frequency range 5.5–11.0 GHz.
(b) Method 2 Initial Cr adherence is done as Method 1. Then the Cr film acts as a lumped 50-ohm
resistor which is terminated in a quarter wavelength open circuit impedance. This arrangement
provides VSWR < 1.2 over the frequency range 4.5–6.0 GHz.
(c) Method 3 A 50-ohm lumped resistor shunt mounted across the strip line and terminated by a short
at the end.
(d) Method 4 An isolator in MIC form on ferrite substrate can be designed with a triangular patch
having a narrow transverse slot and corner cuts as shown in Fig. 6.57(d). The dc magnetic field
is applied perpendicular to the substrate plane. For X-band operation, the approximate size of the
isolator may be of the order of 10 mm × 8 mm × 1/.8” and slot size of approximately 6 mm ¥ 0.8
mm for best matching at 11 GHz. The insertion loss of nearly 2 dB and isolation of 18 dB could be
achieved. The metallization on the Ni–Zn ferrite is deposited by sputtering a seed layer of titanium
tungsten (TiW) prior to the final metallization.
 Microwave Network Theory and Passive Devices 285

(a) (b)

(c) (d)

Fig. 6.57 Microstrip isolator

YIG Filters and Oscillators

Yttrium Iron Garnet (YIG) is a complex solid having ferromagnetic properties and is represented by
Me3Fe5O12 or 5Fe2O3 3Me2O3 or Me3Fe2(FeO4)3, where Me is a rare-earth metal. It has much smaller losses
than ordinary ferrites. A small sphere (40 mil diameter) of YIG usually has a very small loss and can be used
as a resonator in microwave filters and oscillators with resonant frequency at f0 = 2.8 H0, where f0 is in MHz
and H0 is steady external field in oersteds. Thus, the resonant frequency can be controlled electronically by
changing H0. The low frequency limit (about 3.5 GHz) of the YIG resonator is controlled by the minimum
m0 applied for saturation of magnetization. This frequency can be brought down by doping pure YIG with
gadolinium (Gd).
A small YIG sphere is placed at the centre of two mutually perpendicular loops in yz and xz planes
(Fig. 6.58). A dc magnetic field H0 along z-axis is created perpendicular to the two loops. Input RF current
is fed to the input loop and the output loop is
connected to the load. In the steady H0 field, the
YIG has a net magnetization in the H0 direction.
The resulting magnetic field component Hx makes
the magnetization vector precess about the z-axis
resulting in a y-component of magnetization which
in turn induces a voltage at the output loop. The off-
frequency components of RF input couples a small
power to the output loop. Thus, the system acts like
a tuned band-pass filter at frequency f0.
In a waveguide, the YIG sphere is placed in the
coupling hole of two rectangular waveguides having
dominant mode magnetic fields perpendicular to Fig. 6.58 YIG filters and oscillators
each other for filtering action.