iv. linear n-port networks
TRANSCRIPT
1
Figure 1: Basic 2-port network.
IV. Linear N-Port Networks
ECE 420 — Fall 2001 — Prof. Frey
Thus far, we have studied the fundamental characterization of linear circuits. Such an
approach is of particular value in understanding the intrinsic properties of circuits. However, it is
common in Electrical Engineering to consider networks as having inputs and outputs that are applied
and measured at ports, respectively. As a result, an entire theory of electrical networks has been
developed for the characterization of electrical networks viewed as N-ports. We will now explore
some of this theory and use it to say some interesting things about electronics.
2-Ports--Modeling
By far the most common N-port is that
where N=2--namely, the 2-port. Figure 1 depicts a
general 2-port network. Each port of the network
comprises 2 wires across which we may measure a
voltage as indicated by V and V , corresponding1 2
to ports 1 and 2, respectively. We also define
corresponding port currents, I and I , flowing into1 2
the ports at the positive voltage reference. This is,
of course, the natural extension of a 1-port where only one pair of wires is considered. It is worth
noting that in general the wires shown explicitly to describe the ports correspond to nodes within the
linear 2-port network and, therefore, the ports could share the same internal nodes. For example,
ground could be common to both ports.
The characterization of 2-ports is a natural generalization of that for 1-ports. While a 1-port
network can be described as a single branch in some electrical network, a 2-port can be represented
as a pair of branches. The interesting difference, however, is that the pair of branches associated
with a 2-port may be coupled. Recall that Thevenin’s Theorem tells us that there is in general an
affine relationship between voltage and current is a one port--that is, the port voltage (current) is a
linear function of the port current (voltage) plus a source term due solely to internal sources. The
V ' Z I
V ' /0000/0000
V1
V2
; I ' /0000/0000
I1
I2
; Z ' /0000/0000
Z11 Z12
Z21 Z22
2
(1)
port voltage and current are not intrinsically functions of any other voltages and currents outside of
this 1-port. On the other hand, in a 2-port, a port voltage or current may be a function of not only
that port’s current or voltage, but also the other port’s voltage or current, in addition to any internal
sources.
Another natural extension of 1-port ideas to 2-ports is that regarding power. The port current
and voltage references in 1-ports is chosen so that the product of the port voltage and current yields
the power dissipated by, or within, the 1-port. By choosing the referencing similarly for 2-ports we
obtain a natural generalization. Specifically, the sum of the products of voltage and current at each
of the ports equals the power dissipated by, or within, the 2-port. Hence, V I + V I = P, where P1 1 2 2
denotes the power dissipated in the 2-port. Notice that, unlike in 1-ports, it is possible for the power
to be zero, despite the fact that the port voltages and currents may all be nonzero. This allows for
some very useful possibilities.
Before continuing, let us note that it is customary in the context of the following discussion
to assume that N-ports contain no internal independent sources. Such a restriction is not necessary,
but allows us to focus on the intrinsic coupling properties of 2-ports without the clutter due to
possible internal sources. Besides, due to the superposition principle, we can always turn off internal
sources for the present analysis and turn them on again at a later point in the analysis if we want to
determine their effect on the system. This will be done later, but for now let us assume that we can
characterize 2-ports by purely linear (as opposed to affine) relations, making the port voltages and
currents dependent upon only one another.
There are many choices for the possible relations between port variables. One popular choice
is the so-called z-parameter model. In this case the port voltages are expressed as functions of the
port currents. These relations are typically expressed in matrix form as shown below:
The elements of the Z matrix are referred to as the z-parameters for the 2-port, and they explicitly
I ' Y V ; Y ' /0000/0000
Y11 Y12
Y21 Y22
V1 ' (R1%R2) I1 % R2 I2 ; V2 ' R2 I1 % R2 I2
Y Z ' /0000/0000
(R1%R2) R2
R2 R2
3
(2)Figure 2: Example 2-port network.
(3)
tell how the port voltages are controlled by the port currents. Clearly, the z-parameters each have
units of impedance, which explains the choice of letters. Note that if Z and Z are both zero then12 21
the 2-port reduces to a pair of uncoupled 1-ports, which are simply impedance elements since
internal sources are assumed to be zero. It is the cross coupling expressed through nonzero values
for the off diagonal elements, Z and Z , that give a 2-port its character, as we shall see later.12 21
An alternative characterization for a
2-port is given by its so-called y-parameter
representation. In this case the port currents
are given as linear functions of the port
voltages. Such a characterization is given
mathematically as,
In this case, each y-parameter has units of admittance,
explaining the choice of letters. Again, it is the nonzero off diagonal elements that give a 2-port its
character. Looking at (1) and (2), it seems clear that there must be a simple relationship between the
z-parameter matrix, Z, and the y-parameter matrix, Y. Specifically, Y must be the inverse of Z. This
is indeed the case. Anytime that both z- and y-parameter descriptions exist for a 2-port, the Z and
Y matrices will be inverses of one another.
Let us consider an example to help clarify the discussion. Consider the 2-port network of
Figure 2, where current sources have been attached to the ports for the purpose of determining the
network’s z-parameters. It is a simple matter to solve for the port voltages in terms of the port
currents which are necessarily equal to the applied current sources. The result of these calculations
is given below.
I1 '1R1
V1 &1R1
V2 ; I2 ' &1R1
V1 % (1R1
%1R2
)V2
Y Y ' /0000/0000
G1 &G1
&G1 (G1%G2)
/0000/0000
V1
I2
' H /0000/0000
I1
V2
' /0000/0000
h11 h12
h21 h22
/0000/0000
I1
V2
4
(4)
(5)
The application of voltage sources instead of current sources to the 2-port of Figure 2 allows us to
solve for the port currents in terms of the applied port voltages, yielding y-parameters as follows:
It is a simple matter to see that Z and Y in (3) and (4) are inverses of one another. The only time this
property will not technically hold is when the 2-port in question fails to possess both z- and y-
parameter characterizations. This could happen, for example, if R were replaced by a short in the1
circuit of Figure 2. In this case the z-parameters exist and are given by the result in (3) with R = 0.1
However, the z-parameter matrix is now singular and, hence, its inverse fails to exist. But inspection
of the new circuit reveals that with R replaced by a short the port voltages are now forced to be1
equal so that a pair of independent port voltages may not be specified. This means that y-parameters
may not be found for such a network, which is suggested by the lack of an inverse for the z-
parameter matrix. The dual situation occurs if the resistor, R , is replaced by an open circuit. Now,2
y-parameters may be found, but z-parameters may not, since the port currents are no longer
independent. This fact is anticipated by the fact that the y-parameter matrix is singular this time.
Another popular characterization for 2-ports is the so-called h-parameter model, where now
one port voltage, V , and one port current, I , are assumed to be dependent upon the remaining port1 2
variables--namely, I and V . In this case, the general form of the 2-port equations is as given below.1 2
Equation (5) defines the so-called h-parameters for a 2-port. These may be found for the network
of Figure 2 by attaching a current source at port 1, forcing the current, I , and a voltage source at port1
2, forcing the voltage, V . The complementary variables may be easily computed with the result,2
V1 ' R1 I1 % V2 ; I2 ' & I1 %1R2
V2
Y H ' /0000/0000
R1 1
&1 G2
/0000/0000
I1
V2
' G /0000/0000
V1
I2
' /0000/0000
g11 g12
g21 g22
/0000/0000
V1
I2
5
(6)
(7)
As might be expected, the h-parameter matrix, H, may be computed using either the z- or y-
parameter matrices; however, the calculations are a little more messy.
The last of this group of 2-port characterizations is given by the so-called g-parameter model,
which is the complement to the h-parameter characterization. Specifically, for g-parameters, we
have,
Observe that the g-parameters must each have different units, since they do not all relate one type
of quantity--for example, a current--to another type--for example, a voltage. This is the case for h-
parameters as well. Comparing the h- and g-parameter definitions of (5) and (7), it becomes clear
that the g-parameter matrix, G, must be the inverse of the h-parameter matrix, H. As a result, we can
find the g-parameters by either attaching a voltage source to port 1 and a current source to port 2, and
computing the complementary variables, or by inverting H. While g-parameters have been defined
for completeness, they find little use in Electrical Engineering. Historically, z-, y-, and h-parameters
have been used almost exclusively for practical circuits.
Another benefit of the 2-port characterizations introduced thus far is in their ability to suggest
equivalent circuits for 2-ports. In particular, just as the affine relation between port current, port
voltage, and internal sources suggest equivalent circuits for 1-ports--namely, Thevenin and Norton
equivalents--so do the different 2-port characterizations given above suggest equivalent circuits. To
see this, consider the z-parameter characterization for a 2-port given in (1). Writing out the z-
parameter equations individually yields,
V1 ' Z11 I1 % Z12 I2 ; V2 ' Z21 I1 % Z22 I2
6
(8)
Figure 3: Z-parameter 2-port model.
Figure 4: Y-parameter model for a 2-port.
A reasonable interpretation for the equations in (8) is that each port voltage may be found by adding
two voltages. These two voltages are given by an impedance times the respective port current and
a transimpedance times the other port current. Electrically, this is equivalent to an impedance
element carrying the port current in series with a current controlled dependent voltage source. The
circuit of Figure 3 shows this idea. This is
an equivalent circuit that may be used to
model any 2-port that possesses a z-
parameter characterization. For example,
by using the z-parameters given in (3) in the
2-port model of Figure 3, we obtain the
equivalent z-parameter model for the circuit
of Figure 2.
Following similar logic to that
above in finding the z-parameter model, we may determine the y-parameter model for a 2-port.
Specifically, by writing out the y-parameter characterization given in (2) in a way analogous to that
shown in (8), we may observe that each of the port currents is given by the sum of two separate
currents. These two currents are given by an admittance times the respective port voltage and a
transadmittance times the other port
voltage. Because of this interpretation, we
may derive an equivalent circuit at each
port of a 2-port to be the parallel
combination of an admittance and a voltage
controlled dependent current source.
Putting these ideas together yields the y-
parameter model for a 2-port shown in
Figure 4.
It is interesting to compare the
/0000/0000
V1
I1
' /000 /000A B
C D/0000
/0000V2
& I2
; T / /000 /000A B
C D
7
Figure 5: h-parameter model for a 2-port.
(9)
models of Figures 3 and 4. Notice that the z-parameter model resembles a pair of coupled Thevenin
equivalent circuits. On the other hand, the y-parameter model of Figure 4 resembles a pair of
coupled Norton equivalent circuits. It is
then a simple matter to understand that an
h-parameter model is a hybrid of the z- and
y-parameter models had by using a
combination of Thevenin and Norton type
circuits at the ports. In particular, one can
easily show that the h-parameter model for
a 2-port is given by the circuit in Figure 5.
Clearly, the g-parameter model is that
circuit with the Thevenin type circuit at port 2 and the Norton type circuit at port 1. Having found
the various 2-port equivalent circuits, one may freely employ them in replacing any linear 2-port
(with internal sources inactive) for the purpose, for example, of simplifying some larger system.
There are other types of models which have been used to characterize 2-ports. One in
particular that has found use in cascading networks is the so-called transmission matrix, which is
also referred to as the ABCD parameter, characterization. Unlike the z-, y-, h-, and g-parameter
models, the transmission matrix characterizes the 2-port as a relationship between the ports-that is,
the port 1 variables are consided as dependent upon the port 2 variables. Specifically, we have,
T is defined to be the transmission matrix, composed of ABCD-parameters. Notice that, in this
characterization, the output (port 2) controls the input (port 1), and that the output port current (-I )2
is measured leaving port 2. With this slight change in perspective it is easy to calculate the
composite T matrix of a cascade of 2-ports by multiplying the T matrices for each of the individual
networks.
One other popular 2-port characterization is that involving the so-called S-parameters. This
8
Figure 6: 2-port driven by a source at port 1.
model for characterizing networks is best suited to high frequency networks where the propagation
and reflection of signals must be taken into account. Specifically, S-parameters relate the reflected
signals at the ports of the network to the incident signals. All incident and reflected signals will be
of the same type--i.e., voltages or currents--in this type of characterization. When considering
lumped networks (as we have been doing), the S-parameter characterization becomes unwieldy and
a bit contrived. The best appreciation of S-parameters, especially in the context of lumped networks,
is had by thinking in terms of the power delivered to the various ports of a network by external
sources, where each source has some nonzero source impedance.
Reciprocity
Having discussed the basics of 2-port representations, it is of interest to note an important
generic classification typically given to them. Namely, 2-ports are typically referred to as being
either reciprocal or non-reciprocal. The idea of reciprocity is an old one and arises from a very
simple, and sometimes very useful,
property that 2-ports may possess. In order
to understand this, consider the 2-port
network of Figure 1, where we have
attached a current source at port 1, as
shown in Figure 6. We may calculate the
response, V , due to the applied current2
source that specifies I . Given the z-1
parameters for the 2-port, and recognizing that I = 0 (by inspection), it is a simple matter to verify2
that V = Z I . Hence, Z is the transfer function from the source, I , to the response, V . Now2 21 1 21 1 2
suppose that we were to remove the current source from port 1 and connect it to port 2, thereby
specifying I . This time port 1 would be open, and if we were to find V in response to this source,2 1
we would find that V = Z I . Now Z is the transfer function from the source, I , to the response,1 12 2 12 2
V . Clearly, if Z = Z , then the network response voltage to the current source will be the same1 12 21
for both cases. This result seems quite unusual to most people who see it for the first time, because
networks that are quite asymmetrical looking often have the property that Z = Z . For example,12 21
9
Figure 7: 2-port driven by a voltage source atport 1.
the circuit of Figure 2 possesses this property as shown in (3), despite the fact that it looks different
at the two ports. Networks possessing this property--namely that, Z = Z --are called reciprocal12 21
networks.
Considering the fact that the different 2-port characterizations given earlier must all be
related, reciprocity must specify more than just the properties above. For example, we have already
observed that the y-parameter matrix is the inverse of the z-parameter matrix. Whenever a 2-port
is reciprocal, its z-parameter matrix must be symmetric, since the off-diagonal elements, Z and Z ,12 21
are equal. A basic property of matrices is
that the inverse of a symmetric matrix is
symmetric. Hence, the off-diagonal
elements of the y-parameter matrix, Y and12
Y , must also be equal. (We assume, of21
course, that both the z- and y-parameter
matrices exist.) Just as the equality of Z12
and Z implies a circuit property, so does21
the equality of Y and Y . To see this,12 21
consider the network of Figure 7, where a 2-port is being driven by a voltage at port 1, and a short
has been placed across port 2. Observe that the port 1 voltage is set by the source and the port 2
voltage is zero. We may look at the response, I , to the source making V . It is a simple matter to2 1
verify that I = Y V , making Y the transfer function from input, V , to output, I . Now suppose2 21 1 21 1 2
that we replace the source at port 1 with the short and the short at port 2 with the voltage source.
This time V is equal to the voltage source and V is equal to zero. Now the transfer function from2 1
the input, V , to the response, I , is given by Y . Therefore, if the 2-port is reciprocal, then these2 1 12
transfer functions must be equal and the response measured in the two cases is the same.
We can say something about the h- and g-parameters regarding reciprocity as well. As
suggested above, the h- and g-parameters may be derived from the z- or y-parameters (assuming they
exist.). In particular, it can be easily proven that,
h12 'z12
z22
; h21 ' &z21
z22
; g12 'y12
y22
; g21 ' &y21
y22
10
(10)
Figure 8: 2-port driven by current sources.
From these relations it is clear that a reciprocal network will have h- and g-parameters obeying the
constraints, h = -h and g = -g .21 12 21 12
It is reasonable at this point to ask what kind of circuits are going to be reciprocal. With the
theoretical tools developed so far, we are in a position to answer this question. As suggested earlier,
let us view a linear 2-port as being a pair of branches possessing a joint constitutive law given by one
of the 2-port characterizations discussed above. To make things less abstract, assume that the pair
of branches are characterized by a known z-
parameter matrix. Let us assume that the 2-
port being considered is driven by a pair of
current sources, as shown in Figure 2, for
example. We have specified a network
having 4 branches, two of which are current
sources, and two of which comprise the
linear 2-port. The situation is shown
schematically in Figure 8.
Now let us consider the response of this network to the source at port 1, labeled I , measured3
as the voltage across the current source at port 2, labeled I . Despite the orientation of I , however,4 4
let us measure this voltage from top to bottom, making it equal to the voltage labeled V in Figure2
8. We assume that I is set to zero during this measurement. As usual in a linear system, this4
response will be some transfer function times the input source, I . Now let us recall the adjoint3
network concept discussed earlier, and apply it to this system. Specifically, the adjoint to the circuit
of Figure 8 is that circuit having the same topology but with the input and output reversed. That is,
we now assume the input to be applied via I and the output to be measusred as the voltage across4
I from top to bottom, equal to V , as can be seen from Figure 8. It is a straightforward matter to3 1
verify that the adjoint network for this very simple circuit is that network where the coupled branches
have a joint constitutive law that is the transpose of the original. This follows directly from the fact
/0000/0000
V1
V2
' /0000/0000
sL1 sM
sM sL2
/0000/0000
I1
I2
; M ' k L1 L2
11
Figure 9: Ideal transformer representation.
(11)
that the modified nodal matrix of an adjoint network must be the transpose of that of the original.
Therefore, if the z-parameter matrix of the 2-port in the circuit of Figure 8 is symmetric, then it is
self-adjoint, and the response measured using I as the source will match that when using I as the4 3
source in the above scenario. In summary, we have proven that a self-adjoint network is reciprocal,
at least for the case of a z-parameter model. Following the above reasoning, you can prove with a
little extra effort that self-adjoint networks are reciprocal in general in the context of 2-port analysis.
Moreover, this idea of self-adjoint networks shows how to extend the idea of reciprocity to N-port
networks in general.
Special 2-port Networks
Having discussed 2-ports in the abstract, we are now ready to consider the more well known
specific types of 2-ports. We begin with the most widely used reciprocal 2-port--namely, the
transformer. A transformer is simply a pair
of coupled inductors and must, as a dierct
consequence of the physics, be a reciprocal
network. Figure 9 shows the basic idea of
a transformer. Of course, the transformer is
only that part of the circuit within the
dotterd lines. In a transformer, we assume
that the inductors, L and L , are coupled1 2
via some form of flux linkage so that there
is a mutual inductance that can be measured
between them, given by M. The usual dot notation has been used to indicate the orientation of the
flux linkage between the inductors. The equations relating voltage and current in the ideal
transformer of Figure 9 are given by,
V1 ' NV2 ; I2 ' &NI1 Y /0000/0000
V1
I2
' /000 /0000 N
&N 0/0000
/0000I1
V2
12
(12)
where k is defined as the coupling coefficient which may take on values from 0 to 1. Typically, k
is a number quite close to one in practice--for example, in the range from 0.95 to 0.999--and is
assumed equal to 1 for an ideal transformer. Notice that the transformer has been characterized with
a z-parameter representation. This is standard practice and quite convenient in light of the
constitutive laws specified by the physics of a transformer.
It is common, however, to use an even simpler description for an ideal transformer. This
description takes into account the so-called turns ratio, N, of the transformer. Specifically, suppose
that for the transformer of Figure 9, the primary winding--that is, the coil of wire comprising L --has1
N times as many turns as the secondary of the transformer--that is, the coil comprising L . Then we2
have the common idealized description for a transformer given in (12) below.
This characterization of a transformer is certainly idealized, since it predicts no frequency
dependence whatsoever and, in particular, that DC signals may be coupled between the ports, which
is, of course, impossible using a physical transformer. Nevertheless, the characterization given in
(12) is quite useful is understanding the operation of real transformers as an approximation. Note
that the representation given in (12) is an h-parameter representation, unlike that of (11). This is
because the ideal relations between current and voltage using the turns ratio do not permit either a
z-parameter or a y-parameter representation. Also notice that the h-parameter matrix is skew-
symmetric, which implies that an ideal transformer is a reciprocal 2-port. Finally, it is important to
observe that an ideal transformer is a lossless network--that is, the total power dissipated in the 2-
port is identically zero. This is easily determined by using the relations in (12) to compute P = V1
I + V I = V I + (V /N) (-NI ) = 0. 1 2 2 1 1 1 1
It is an interesting exercise to find the range of inductance and frequency that one must
assume in the actual transformer representation of (11) to get the approximation of (12). The
problem is not well posed unless one considers the impedance at the primary and the secondary of
the transformer that would result from attaching a source, having a nonzero source impedance, and
I2 ' gV1 ; I1 ' &gV2
I ' /0000/0000
I1
I2
' /000 /0000 &g
g 0/0000
/0000V1
V2
' YV
V ' /0000/0000
V1
V2
' /000 /0000 &r
r 0/0000
/0000I1
I2
' Z I ; r ' 1/g
13
(13)
(14)
(15)
a load. This would make a good homework problem. Another interesting perspective on the ideal
transformer described by (12) is had by considering whether a circuit can be created to implement
these 2-port relations. Even though coupled inductors could never truly implement the equations of
(12), there is no fundamental reason why a circuit could not be designed to do the job. Perhaps you
could think of how to do this (Another good homework problem.). The discussion below will offer
some possibilities.
Gyrators
We now turn to another special 2-port, called a gyrator, that has received much attention over
the years. Unlike the transformer, this 2-port is non-reciprocal, although it is anti-reciprocal, which
is a special property in its own right. In the context of the above, a gyrator is simply a special case
of a linear 2-port. While there are many possible 2-port descriptions, the most common way of
writing the basic equations relating the port parameters in a gyrator is as follows:
Using these equations it is simple to write the y-parameter 2-port description for a gyrator as,
This suggests that a gyrator can be implemented with voltage controlled current sources, having
gains of g and -g, respectively. By inverting the relations in equation (14), one obtains a gyrator
formulation based upon current controlled voltage sources. Specifically,
P ' V T I ' V1 V2/0000
/0000I1
I2
' V1 V2 /000 /0000 &g
g 0/0000
/0000V1
V2
' V2 V1 & V1 V2 ' 0
14
(16)
Figure 10: The gyrator circuitsymbol.
Figure 11: Gyrator with a load.
where the z-parameter matrix, Z, is just the inverse of the y-parameter matrix, Y. Since it is most
convenient to realize practical voltage controlled current source networks, as opposed to current
controlled voltage source networks, the formulation in equation (14) is generally preferred. For
theoretical purposes, of course, both formulations are useful. Using equation (14) it is simple to
show that a gyrator is a lossless electrical network. Specifically,
The fact that gyrators are, in theory, lossless makes them
attractive in filter synthesis. Recall that ideal transformers
are also lossless.
Let us take a closer look at the properties which
gyrators possess that make these circuits interesting for use
in electronics. A first property is that these 2-ports are not
reciprocal networks, since their y-parameter matrices are
not symmetric. In fact, these matrices are skew symmetric--
that is, they equal the negative of their transpose. It is well
known to circuit theorists that non-reciprocal networks cannot be realized with only passive
components--that is resistors, capacitors, and
inductors. This means that gyrators are strictly
active networks that must, therefore, be realized
with active components, such as transistors or
operational amplifiers. 2-port gyrators have been
given their own circuit symbol which is shown in
Figure 10. The gyration constant, g, is built into
the symbol.
Perhaps the most important property of a gyrator is its ability to transform admittances into
impedances. Specifically, when an admittance is connected to one port of a gyrator, the impedance
Zload 'V2
&I2
; Zinput 'V1
I1
'I2 /g
&gV2
'1
g 2
1Zload
'1
g 2Yload
Zinput '1
g 2sC ' sLeq ; Leq '
C
g 2
15
(17)
(18)
Figure 12: RLC bandpass filter.
Figure 13: Bandpass filter using a gyrator.
looking into the other port is exactly a scaled version of that admittance. The derivation can be
accomplished with the help of Figure 11, where Z is the impedance attached to port 2, Y is itsload load
reciprocal--that is, the admittance attached to port 2--and Z is the impedance seen looking intoinput
port 1. We have,
The gyration constant, g, determines the scale factor, but the nature of the input impedance is
determined by the admittance attached to port 2. Therefore, if a capacitor, C, is attached to port 2,
then we have,
This simple relation explains the vast majority of
the gyrator’s popularity in electronic design. It
shows that a capacitor can be used to replace an
inductor in a circuit with the help of a gyrator.
Since inductors are rarely desirable in electronic
circuits operating below about 1GHz, this idea is
quite appealing. Capacitors and gyrators
are convenient to realize as part of
integrated circuits.
Filter synthesis based on the inductor
simulation described above is usually done
by starting with an RLC prototype, and
replacing the inductors with
capacitor/gyrator combinations. Consider
the following example of a very simple second order bandpass filter shown in Figure 12. After
replacing the inductor with a gyrator/capacitor combination, the filter is realized solely using
Yload ' jN
k'1Yk Y Zinput '
Yload
g 2' j
N
k'1
Yk
g 2
Zload ' jN
k'1Zk Y Yinput '
1Zinput
'1
Yload /g 2' g 2 Zload ' j
N
k'1g 2 Zk
16
(19)
Figure 14: Gyrator realization usingtransconductance amplifiers.
capacitors as the reactive elements, as shown in Figure 13. Furthermore, the filter can be tuned
electronically if the gyration constant can be varied electronically.
A byproduct of the property discussed above is that series and parallel circuits may be
interchanged with the help of a gyrator. Suppose one port, say port 2, of a gyrator is loaded with a
parallel combination of elements. The admittance of this combination is the sum of the admittances
of each of the elements. At the other port, port 1, the input impedance will be a scaled version of
this admittance; hence, a sum of impedances. Since the composite input impedance seen at port 1
is given by a sum of impedances, it must be equivalent to a series combination of elements.
Therefore, the gyrator converts a parallel network into a series network. Using similar logic, it
becomes clear that a series network connected to port 2 will be reflected as a parallel network
looking into port 1. These results are summarized below.
The realization of gyrators in
electronic form is quite simple; however, as
usual, different circuit realizations are
preferable to others, depending on the
application. To begin, consider the
simplest generic realization comprised of a
pair of transconductance amplifiers, as
shown in Figure 14. Each transconductance
amplifier is assumed to have infinite input
and output impedance, with an output
current equal to the transconductance, G = g, times the input voltage applied between the + and -m
terminals. The circuit shown in the figure satisfies the basic 2-port relations for a gyrator, given by
equation (14).
Z 'V2
&I2
; &Z 'V1
I1
V1 ' kV2 ; I2 ' kI1 YV1
I1
'kV2
I2 /k' &k 2
V2
& I2
' &k 2 Z
17
(20)
(21)
The circuit of Figure 14 does not implement the most general form of a gyrator, since both
ports of the gyrator realization in the figure have ground in common. Therefore, only ground
referenced impedances may be transformed as described above. This limitation stems from the fact
that the transconductors in Figure 14 have single-ended outputs. If differential input/differential
output transconductors are used then a general “floating” gyrator realization is created--that is, a
gyrator whose ports need not be referenced in any way to ground. On the other hand, if you get
clever, you can realize a floating gyrator with a pair of grounded gyrators. Can you see how this
could be done?
Negative Impedance Converters
Now that we have introduced the gyrator, it is interesting to see what other creative 2-port
networks we might find. The gyrator, as shown above, can be used to make impedances at one port
appear differently at the other port. Specifically, a gyrator can make a capacitor look like an
inductor. A negative impedance converter (NIC) is a 2-port, as its name suggests, that can reflect
the negative of a given impedance at its opposite port. Rather than just give the definition of this 2-
port, let’s see if we can invent it. Suppose we start with the assumption that an impedance, Z,
connected to port 2 of this network will reflect an impedance of -Z at port 1. Then we have,
Suppose, for the sake of generality, that we let the port 2 current follow the port 1 current with a gain
of k, and that the port 1 voltage follows the port 2 voltage also with a gain of k. Then we have,
Clearly, if k = 1, then the impedance seen looking into port 1 is the negative of that attached to port
2. Hence, a network following the relations in (21) is an NIC. Putting this into the matrix form that
we have been using gives us this general representation for a negative impedance converter.
/0000/0000
V1
I2
' /000 /0000 k
k 0/0000
/0000I1
V2
18
(22)
Note that the above matrix representation gives the h-parameter description of this 2-port.
Furthermore, since the off-daigonal terms are equal, this is a non-reciprocal network; hence, it cannot
be realized with only passive elements. Also notice that this description is identical to the ideal
transformer except that the off-diagonal terms have the same sign. Recall that an ideal transformer
can be used to reflect different impendances as well; however, a transformer is a positive impedance
converter, in light of the current discussion. The idea of a network that creates negative impedances
is intriguing. Now the only challenge is in finding a use for such an unusual network. Certainly one
use might be to cancel the effects of a positive impedance, which is exactly what is necessary in an
oscillator circuit, for example.
As a final thought on transformers, gyrators, and NICs, notice that by cascading a pair of
gyrators or a pair of NICs, one obtains an ideal transformer. Does this give you any ideas about
possible circuit realizations for an ideal transformer?
Ideal Op Amps
Operational amplifiers certainly represent an important component in modern electronics.
While they possess many characteristics that are interesting, it usually requires an extensive look at
the nonideal properties of op amps to be able to use them effectively in practical designs.
Nevertheless, the idea of ideal op amps still is used to introduce students to the concept, and then
to help model the behavior of circuits incorporating op amps. This is because nonideal op amps can
always be modeled as being ideal op amps with surrounding components. Therefore, the study of
ideal op amps remains of interest even to the expert in the field, as well as to the circuit theorist.
One of the most intriguing features of ideal op amps is that they are 2-ports with unusually
degenerate constitutive laws. Recall that for an ideal op amp we assume that the input differential
voltage is zero and that the output voltage is finite. Actually, we assume that the output voltage is
an infinite gain times the differential input, but that the output is finite. This leads immediately to
V1 ' 0 ; I1 ' 0 ; V2 , I2 are unknown
19
(23)
Figure 15: Circuit symbols for thenullator and norator.
the assumption of zero input voltage. On the other hand, we typically assume that an ideal op amp
is a 2-port having an open circuit at port 1 and an ideal voltage controlled voltage source (VCVS)
at port 2. Thus, its input current is zero. This leaves us with the curious characterization that both
the input voltage and current for the ideal op amp are zero. Furthermore, this complete knowledge
of the input port variables is accompanied by a complete lack of knowledge of the output port
variables. Specifically, we cannot say what the output voltage or current are for an ideal op amp
solely from the knowledge of the input voltage and current (which are both zero!). The 2-port
equations for an ideal op amp are summarized below.
Despite the unusual nature of these relations they qualify as a perfectly acceptable set of constitutive
laws for a 2-port. That is, they impose exactly 2 constraints on the four network variables, which
is exactly what every other 2-port constitutive law does. Therefore, larger networks incorporating
ideal op amps admit to all of the formalism we have
developed thus far to characterize general circuits. In
addition, we can expect that linear networks including ideal
op amps will possess unique solutions, and that such
networks will have linear transfer functions and have
topological duals and adjoints.
Unlike the other 2-ports we have considered, ideal
op amps cannot be given 2-port models that use known
branches. Specifically, what branch has identically zero
voltage and current? It looks like a short and an open
simultaneously! Also, what branch has absolutely no constraints on voltage and current? Because
no such branches exist in our experience, they have been created by circuit theorists. In particular,
a branch having identically zero voltage and current is called a nullator. A branch with absolutely
no constraint on its voltage and current is called a norator. The circuit symbols for these fictitious
branches have been assigned as well and are shown in Figure 15. Using these symbols, we can
V ' (V1 ,V2 , @@@ ,VN )T ; I ' ( I1 , I2 , @@@ , IN )T
20
Figure 16: General N-port network.
(24)
describe an ideal operational amplifier as being the 2-port having a nullator at port 1 and a norator
at port 2. Such a characterization can be used to do some interesting transformations on circuits
including operational amplifiers. We discussed some of these in class.
General N-port networks
Having described 2-ports in detail, we are
now in a position to generalize the concept of a 2-
port to that of the N-port. As the name suggests, N-
ports are networks having N ports defined by pairs
of wires, across which we may measure port
voltages, and into which we may measure port
currents. The basic idea is given in Figure 16. As
in the case of 2-ports, the wires defining each port
of this network may not all connect to independent nodes within the network. For example, all of
the ports may have ground in common. Notice that the currents are all defined as flowing into the
wire attached to the positive voltage reference. This makes the definition of power delivered to the
N-port a natural extension of that for 2-ports. Specifically, the sum of the products of voltage and
current measured at all ports equals the power delivered to the N-port. To give a more concrete
mathematical basis for the discussion, let us define the port voltage and port current vectors as
follows:
The power, P, delivered to the N-port is now simply the inner product of these vectors.
We may generalize the characterizations for 2-ports given earlier is a natural way.
Specifically, the generalization of the z-parameter matrix is the open circuit impedance matrix, Z .OC
This N x N matrix of elements each having units of impedance yields the port voltage vector in terms
of the port current vector. The short circuit admittance matrix, Y , generalizes the y-parameterSC
matrix by allowing us to express the port current vector in terms of the port voltage vector.
V ' ZOC I ; I ' YSC V
21
(25)
Specifically,
We can define a hybrid matrix, H, along the lines of h-parameters by defining port vectors having
a mix of voltages and current and using H to form a relation between them. We will see this idea
below. Finally, note that we may generalize the idea of reciprocity. An N-port is reciprocal if its
open circuit impedance matrix and (assuming both matrices exist)/or its short circuit admittance
matrix are symmetric. An interesting consequence of this generalization is that all 1-ports must be
reciprocal. Another interesting result is that every N-port composed of exclusively reciprocal
elements--that is, 1-ports, reciprocal 2-ports, etc.--is itself reciprocal. Can you see how to prove this
result?
Memoryless and reactive N-ports
Let us now consider a special subset of N-port networks--namely, memoryless N-ports.
These are N-ports containing no reactive elements. Any circuit made up of only resistors, for
example, would be memoryless. In addition, any circuit containing only resistors, ideal transformers,
ideal op amps, gyrators, and NICs would also be memoryless. It should also be noted that the idea
of a memoryless N-port is not restricted to linear networks. Any nonlinear network not containing
reactive elements is also considered memoryless. (For that matter, the idea of reciprocity is not
limited to linear networks either.) Note that in general we must exclude any subnetwork with
memory, meaning that memoryless N-ports may not contain transmission lines, or any other circuitry
with delay, such as a sample and hold. Such components have not been considered thus far, so we
will not worry about them in our present discussion.
You might ask why the idea of a memoryless N-port is introduced. The answer lies in the
fact that such a network relates all port variables through purely instantaneous relations. As a result,
the N-port description, such as the open circuit impedance matrix, may be used to relate the port
variables in either the time or the frequency domain. Furthermore, a memoryless N-port has no
dynamics associated with it. In particular, it has no internal states and its complete response,
V ' Zoc I % VS ; I ' Ysc V % I S
22
(26)
measured at any of its ports, to an input applied to any of its ports will contain no transient. Note,
however, that since a memoryless N-port may contain memoryless active components, it may be
inherently unstable. A simple example would be that of a negative resistor, which constitutes a
memoryless 1-port which adds power to any other network to which it is connected. Note that
memoryless N-ports dissipate or contribute exclusively “real” power at each port.
The complementary network to a memoryless N-port is that of a purely reactive N-port. Such
an N-port contains purely reactive components. As opposed to memoryless N-ports, reactive N-ports
dissipate or contribute zero “real” power to a network. While power associated with a memoryless
N-port would be measured in Watts, power associated with a reactive N-port would be measured in
VARs. A capacitor or an inductor provides the simplest example of a reactive N-port (N=1). A pair
of coupled inductors (assuming no winding resistance or core losses) represents an interesting
reactive 2-port. It is also interesting to note that a capacitive 2-port--that is, a reactive 2-port
comprising only capacitors--may look like a pair of coupled capacitors if the internal capacitive
branches form a loop. Coupling between the ports of an N-port is evidenced by off-diagonal terms
in its open circuit impedance or short circuit admittance matrix.
As a final matter, it is sometimes useful to allow N-ports to include internal independent
sources. Up until now, we have been assuming that all internal independent sources were absent,
or at least turned off in the spirit of the superposition principle. The inclusion of sources introduces
an affine relation between port variables of the type considered in discussing Thevenin and Norton
Equivalents. To make things clearer in the discussion below, let us formally write down the N-port
voltage/current relations discussed thus far assuming the presence of internal independent sources.
We have,
In this formulation, the source vectors, V and I , are analogous to the Thevenin and Norton sourcesS S
in 1-ports. In fact, for the special case where N=1, the source vectors, V and I , are scalars exactlyS S
equal to the Thevenin and Norton sources, respectively. The hybrid formulation of these equations
will be given below.
23
Figure 17: RLC circuit.
Figure 18: Decomposed version of thecircuit of Figure 17.
State Equation Formulation
An elegant formulation for state equations in general circuits can be written using the ideas
introduced above. Suppose that we have a circuit composed of arbitrary components which includes
a number of capacitors and inductors. It is always possible to partition this network into a
memoryless (often referred to as “resistive”) N-port attached to a pair of reactive N-ports,
specifically one capacitive N-port and one inductive N-port, where the “N” will be different in
general for these reactive networks than the “N” for the main resistive network. It is further possible
to define the resistive N-port in such a way that the subset of ports attached to the capacitive N-port
is completely exclusive of the remaining ports attached to the inductive N-port. To be more specific,
let there be exactly N ports associated with the capacitiveC
N-port, and exactly N ports associated with the inductiveL
N-port. Then it follows from the above statements that it
is always possible to characterize the resistive N-port with
a description incorporating N = N + N ports.C L
Since this decomposition may seem a bit confusing,
let us consider a simple example. Figure 17 shows an RLC
circuit that we would like to decompose along the
lines of the above discussion. We will include the
capacitor in a capacitive N-port with N = N = 1.C
We will include the inductor in an inductive N-port
with N = N = 1. Finally, the resistive N-port willL
comprise the voltage source and the resistor and
will have N = 2 ports. The RLC circuit has been
redrawn in Figure 18 to reflect these definitions.
Note that the two ports of the resistive N-port are in
parallel. This is allowed since one of these ports
has a capacitor attached, and the other has an
inductor attached. Note that while the port voltages
of the resistive 2-port will be equal, the port
/0000/0000
V1
I2
' H /0000/0000
I1
V2
% VS ; H ' /000 /0000 1
&1 G; VS ' /000 /000
0
&GVS
V1 ' VL ' Lddt
IL ' &Lddt
I1
I2 ' & IC ' &Cddt
VC ' &Cddt
V2
/0000/0000
V1
I2
' & /000 /000L 0
0 Cddt
/0000/0000
I1
V2
' H /0000/0000
I1
V2
% VS
ddt
/0000/0000
IL
VC
' /000 /0001 /L 0
0 &1/C/000 /000
0 1
1 G/0000
/0000IL
VC
% /000 /0001 /L 0
0 &1/CVS
24
(27)
(28)
(29)
(30)
currents will not in general be equal. Continuing with the example, let us characterize the various
N-ports defined. Suppose we choose to characterize the resistive N-port by attaching a current
source to the port (call it port “1") where the inductor is connected and by attaching a voltage source
to the port (call it port “2") where the capacitor is connected. Defining these sources to be I and V1 2
, respectively, we may easily write an h-parameter characterization for the resistive 2-port, where it
will be augmented by a term due to the internal independent source. For this case we have,
Now let us continue by writing the characterization for the reactive 1-ports. Consulting Figure 18
to make sure we get the right orientation for currents, we get,
Putting together the above results yields the following set of differential equations:
These equations may be easily solved for the derivatives of I and V , yielding state equations whose1 2
state variables are I and V . Note that these variables are just the negative of the inductor current1 2
and exactly the capacitor voltage, respectively. Alternatively, we may write these equations directly
in terms of inductor current and capacitor voltage, with the result,
VL '
/000000000000000
/000000000000000
VL1
VL2
!
VLNL
' Lddt
/000000000000000
/000000000000000
IL1
IL2
!
ILNL
' Lddt
I L ; I C '
/000000000000000
/000000000000000
IC1
IC2
!
ICNC
' Cddt
/000000000000000
/000000000000000
VC1
VC2
!
VCNC
' Cddt
VC
25
(31)
It is a simple matter to verify that these are the same equations that would result from our earlier
approach to getting the state equations.
Having seen this example, it is a fairly straightforward matter to decide how to extend the
approach to the general case. Using the preliminaries introduced above, let us suppose that a
collection of inductors are connected to the first N ports of a resistive N-port. This collection ofL
inductors may include couplings and inductor cutsets within it. In any event, it constitutes a reactive
(inductive) N-port, where N = N . Let us further suppose that a collection of capacitors, some ofL
which may form loops, are connected to the last N ports of the resistive N-port. This collection ofC
capacitors constitutes a reactive (capacitive) N-port, where N = N . (Please note: In class, IC
attached the capacitive N-port to the first set of ports. There is no problem in doing this, of
course, but it will cause some confusion in comparing to your notes if you miss this point. I
chose to do it this way when I originally wrote this section of the notes because I thought it
flowed better.) Let us define the port relations associated with these reactive multi-port networks
in the following way, where all port variables are defined with the usual relative reference
orientations:
In this case, L and C are N x N and N x N matrices, respectively. Off-diagonal terms in theseL L C C
matrices will be due to internal coupling, inductor cutsets, and capacitor loops. We now continue
by looking at the resistive N-port to which these reactive multi-ports are connected. Without loss
of generality, we may assume that each of the port voltages of the resistive N-port is oriented to
match the voltage orientation of the respective ports of the reactive multi-port networks. Then, each
of the port currents of the resistive N-port will be equal to the negative of the respective port currents
of the reactive multi-port networks. Assuming that we characterize the resistive N-port using a
hybrid model where each of the first N ports are driven by current sources and each of theL
/0000000000000000000000000000000000000
/0000000000000000000000000000000000000
V1
V2
!
VNL
INL%1
INL%2
!
IN
'/000000
/000000VL
& I C
' H
/0000000000000000000000000000000000000
/0000000000000000000000000000000000000
I1
I2
!
INL
VNL%1
VNL%2
!
VN
% VS ' H/000000
/000000& I L
VC
% VS
/000000/000000
VL
& I C
' /0000/0000
HLL HLC
HCL HCC
/000000/000000
& I L
VC
% VS
/000000/000000
VL
I C
' /0000/0000
&HLL HLC
HCL &HCC
/000000/000000
I L
VC
% /000 /0001 0
0 &1VS ; /000000
/000000VL
I C
' /000 /000L 0
0 Cddt/000000
/000000I L
VC
26
(32)
(33)
(34)
remaining N ports are driven by voltage sources, we have the general description given below.C
In order to deal with the minus signs more carefully, let us rewrite the above hybrid characterization
showing a rather obvious partitioning of the H matrix in consideration of the partitioning of the port
voltage/current vectors. Specifically, we have,
Now let us rewrite and put together all of the multi-port equations above in a way that allows them
to be easily combined in a final step. This is done below.
In the formulation above, the “0"s and the “1"s appearing in the matrices represent appropriately
dimensioned blocks of zeros or identity matrices, respectively. Also notice that the capacitive and
inductive multi-port networks have been combined into a single reactive N-port, characterized by
a block diagonal N x N matrix. Combining the resistive and reactive N-port equations above easily
yields this general state space representation for the overall system:
ddt/000000
/000000I L
VC
' /0000/0000
L &1 0
0 C &1/0000
/0000&HLL HLC
HCL &HCC
/000000/000000
I L
VC
% /0000/0000
L &1 0
0 &C &1VS
27
(35)
This formulation is elegant for several reasons. First, it shows how the state equations for
a general circuit are related to its resistive and reactive N-port characterizations. Second, it naturally
handles the problem of determining the correct number of state variables, regardless of the
topological nature of the network. This is because the definitions of the capacitive and inductive
multi-ports can only be made in ways that specify the right number of independent ports. This is
easily seen by observing that there must always be enough ports to fully characterize the connectivity
of the reactive and resistive multi-ports. However, there may never be too many reactive ports
defined without precluding a proper characterization of the reactive multi-ports.
Another benefit of the formulation above is that it provides excellent insight into the structure
of the state matrix for general circuits. Notice that if the resistive N-port is reciprocal, its H matrix
must be symmetric or skew symmetric on a block basis--that is, H and H are symmetric and HLL CC LC
= -H . This is the natural generalization of our earlier result, and may be easily proven from theCLT
fact that the open circuit impedance matrix and the short circuit admittance matrix of a reciprocal
N-port must be symmetric. All purely reactive N-ports are reciprocal, which follows directly from
the fact that they are self-adjoint. Therefore, the matrices, C and L, defined above are symmetric.
It now follows immediately that the state matrix of an arbitrary passive RC or RL network must be
symmetric and, coupled with some basic passivity arguments--that is, that Z and/or Y must beOC SC
“positive definite”-- its eigenvalues must be real. This proves that no passive RC or RL filter may
have complex poles. Either there must be active elements in the network, yielding an H matrix with
asymmetric blocks on the diagonal, or the network must have both capacitors and inductors. Until
now, we could not easily prove this fundamental circuit design result.