CH 2

Class Notes, 31415 RF-Communication Circuits
Chapter II
RF-CIRCUITS
Jens Vidkjær
NB230
ii
Contents
II RF-Circuits, Concepts and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

II-1 Parallel Resonance Circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Frequency Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Poles and Zeros . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Transient Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
II-2 Series Resonance Circuit Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 11
II-3 Narrowband Approximations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
II-4 Series-to-Parallel Conversions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Example II-4-1 ( impedance matching ) . . . . . . . . . . . . . . . 20
Conversions in Narrowband Applications . . . . . . . . . . . . . . . . . . . . 23
Conversions in Broadband Modeling . . . . . . . . . . . . . . . . . . . . . . . 26
Example II-4-2 ( spiral inductor ) . . . . . . . . . . . . . . . . . . . 26
II-5 Tuned Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Synchronously Tuned Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Example II-5-1 ( synchronous tuning ) . . . . . . . . . . . . . . . . 33
Butterworth Stagger Tuned Amplifiers . . . . . . . . . . . . . . . . . . . . . 36
Example II-5-2 ( Butterworth amplifier ) . . . . . . . . . . . . . . 41
II-6 Transformers and Transformerlike Couplings . . . . . . . . . . . . . . . . . . . . 45
Review of Mutual Inductances . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Equivalent Circuits for Two Coupled Inductors . . . . . . . . . . . . . . . . 49
RF-Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Example II-6-1 ( practical RF transformers ) . . . . . . . . . . . 55
Tuned Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Example II-6-2 ( transformer-coupled tuned amplifier ) . . . . 58
Autotransformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Example II-6-3 ( autotransformers in a tuned amplifier ) . . . 62
Transformerlike Couplings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Example II-6-4 ( uncoupled reactance transformers in
tuned amplifiers ) . . . . . . . . . . . . . . . . . . . . . . . . 64
Three-Winding Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Transformer Hybrids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Example II-6-5 ( diode-ring mixer preliminaries ) . . . . . . . . 78
II-7 Double-Tuned Circuits and Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . 80
Coupling between Identical Resonance Circuits . . . . . . . . . . . . . . . . 81
Example II-7-1 ( Foster-Seely FM detector ) . . . . . . . . . . . 85
Example II-7-2 ( quadrature FM detector ) . . . . . . . . . . . . . 88
Double-Tuned Amplifier Stages . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Asymptotic Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
II-8 Impedance Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
J.Vidkjær iii
Lumped Element Impedance Matching using Smith Charts . . . . . . . . 101
Example II-8-1 ( double and standard Smith charts ) . . . . . . 104
Example II-8-2 ( bandwidth estimations ) . . . . . . . . . . . . . . 108
APPENDIX II-A, Power Calculation and Power Matching . . . . . . . . . . . . . . 113
APPENDIX II-B Signal Flow Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Elementary Signal Flow Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Mason’s Direct Rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
References and Further Reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
iv J.Vidkjær
1
II RF-Circuits, Concepts and Methods
In RF-communication system, control of frequency bands and impedance matching

conditions between functional blocks or amplifier stages are problems that constantly face a
circuit designer. The tasks are so frequent that many analytical techniques and approximation
methods especially suited for high-frequency circuits have evolved and strongly influenced the
jargon of RF engineering. Below we shall introduce the most important basic concepts and
methods that are required to
- understand data sheets and literature,

- make simpler design decisions or calculations, and
- prepare and interpret simulation data.
The selection of topics and examples have furthermore been conducted to suit the needs in
the following chapters, which are still in preparation.
Scanning contents, the chapter starts summarizing basic properties of resonance

circuits. Although significant by themselves, the importance of acquiring familiarity with ideal
resonance circuits is the fact, that any narrowbanded resonance circuit may be approximated
by the ideal ones around resonance frequencies. This property reduces significantly the efforts
that are required to understand and explore operations of tuned bandpass circuits, which are
frequently used in RF-communication systems. Foundations of the simplifications are dealt
with in sections concerning narrowband approximations and series-to-parallel transformations.
The tuned amplifiers are introduced in ideal form concentrating on simple frequency charac-
teristics. Coupling techniques using transformers and coupled resonance circuits are still highly
useful methods in RF-designs, so they are considered in some details here. Finally, the very
general method of constructing lumped element matching networks using a Smith chart is
exemplified.
Power matching is fundamental for designing and understanding many RF circuits.

Although this concept is mandatory in basic circuit theory curriculums, it is repeated for
convenience in an appendix. Also the method of illustrating and solving network equations
by the signal flow graph method is summarized in an appendix.
J.Vidkjær
2
II-1 Parallel Resonance Circuits
(1)
Fig.1 Parallel resonance circuit
A basic parallel resonance circuit is shown in Fig.1. Besides component values the
combinations, which are summarized by Eq.(1), are frequently used. The resonance frequency
is the frequency where the capacitive and the inductive susceptances are equal in magnitude
as indicated in Fig.2a. When an external steady state sinusoidal voltage-source of frequency
ω0 is applied to the resonance circuit, the two opposite currents through the capacitor and the
inductor balance each other, and only the resistor current flows through the terminal. This
situation is sketched in Fig.2b, which also shows how the quality factor Q indicates the
magnitude ratio of the internal reactive currents over the resistive terminal current at reso-
nance.
Fig.2 Parallel resonance. (a) Susceptance composition as function of frequency. (b)

Current and voltage phasors at the resonance frequency ω0.
Another view upon resonance and the quality factor concerns the energy in the circuit
under steady state conditions. At instants where the two phasors iC and iL are perpendicular
to the real axis, no currents flow into the capacitor or inductor, but the capacitor hold
maximum energy
(2)
J.Vidkjær
II-1 Parallel Resonance Circuits 3
The first equation is the usual electrostatic energy expression. The second takes into account
that rms values - indicated by small letters - are conventionally used when dealing with steady
state linear circuits. A quarter of a period later the current phasors project in full onto the real
axis while the voltage is zero. The capacitor holds no energy, but the inductor energy peaks
with the same maximum that formerly was held in the capacitor,
(3)
Thus, a constant amount of energy laps between the capacitor and the inductor at resonance,
and the quality factor may be expressed
(4)
where the loss is calculated as the resistor power times the resonance period T0=2π/ω0. This
interpretation of resonance is often useful in the construction of lumped circuit equivalents for
the variety of electromagnetic and mechanical resonators that are used in RF-circuits.
Frequency Response
Expressed through circuit element values, the impedance function for the parallel
circuit in Fig.1 is
(5)
Introducing ω0 and Q from Eq.(1), the impedance expressed as a function of frequency s=jω
becomes
(6)
The frequency dependency of the impedance is kept in the quantity β(ω), which is zero at the
resonance frequency ω0. Here the denominator of Eq.(6) gets its smallest size and the
impedance has maximum Rp, the parallel resistance. The magnitude and phase of Zp(jω) are
(7)
The two functions are shown in Fig.3(a) while Fig.3(b) shows the corresponding admittance
characteristics,
J.Vidkjær
4 RF-Circuits, Concepts and Methods
Fig.3 Impedance (a) and admittance (b) magnitudes and phases of the parallel resonance
circuit in Fig.1. The curves are symmetric around ωo due to the logarithmic fre-
quency scales.
(8)
Upper and lower bounds of the 3dB bandwidth intervals W3dB, which are indicated
in Fig.3 , correspond to a denominator size equal to 2 in Eq.(7). The bounds are found
setting the imaginary part of the denominator equal in magnitude to the real part, i.e.
(9)
Both negative and positive frequencies are contained in the conditions. We call the largest
valued solutions, where the two terms in ωb have equal signs, the upper bounds ±ωbu. The
lower bounds ±ωbl are obtained with terms of opposite signs. Fig.4 summarizes how the
different solutions are formed. By definition, the 3dB bandwidth is taken to be the distance
between positive or zero-valued 3dB frequency bounds, and we get the result that was
incorporated in Fig.3,
(10)
It follows from the solutions in Eq.(9) that the resonance frequency ω0 is not centered between
the 3dB bounds but is the geometrical mean of the bounds,
J.Vidkjær
Fig.4 Upper and lower 3dB bound positions from Eq.(9) . Note, in linear frequency scale
the 3dB bands are not symmetric around the resonances at ±ω0 unless Q → ∞.
(11)
(12)
so in logarithmic frequency scale the upper and lower 3dB frequency bounds are symmetric
with respect to logω0. However, any other pair of frequencies, ωu,ωl that has the resonance
frequency as geometrical mean maps symmetrically around logω0. Since both frequencies
provide the same absolute |β|, i.e.
(13)
the impedance or admittance magnitude characteristics of the types in Fig.3 have even
symmetry with respect to the resonance frequency in a logarithmic frequency scale. Corre-
spondingly, the phase characteristics show odd symmetry because tan-1(-Qβ)=-tan-1(Qβ). At
the 3dB boundaries where |Qβ|=1, the phase angles of impedance Zp become ±¼π. Fig.5
shows plots of the impedance function with various Q-factors. The normalization in magnitude
corresponds to keeping the inductor and capacitor fixed while letting the parallel resistance
follows Q according to Eq.(1). The asymptotic behavior of the impedance approximating the
inductor reactance below and the capacitor reactance above resonance respectively are readily
observed. Calculating magnitudes, it may suffice to use the inductor or capacitor alone at
frequencies that differ more than a factor of three from resonance.
Summarizing the frequency characteristics, we have seen that the greater Q, the
smaller bandwidth and in turn, the steeper phase characteristics around the resonance frequen-
cy ω0. Moreover, the frequency characteristics were symmetric in logarithmic frequency scale.
Phase steepness is an important property when a resonance circuit is employed in an oscillator
and we shall return to this question later. Also, the symmetry property will be reconsidered.
Important classes of signal handling expect linear symmetry that may be approached with high
J.Vidkjær
Fig.5 Normalized magnitudes and phases in the impedance of parallel tuned circuits with
varying Q-factors.
Q circuits too.1 For obvious reasons such circuits are also called narrowbanded.
Poles and Zeros
Pole and zero positions are useful for investigating responses of frequency selective
networks that include parallel tuned circuits. Using parameters from Eq.(1), the impedance Zp
of Eq.(5) is rewritten,
(14)
Solving for the s-values, which set the numerator and the denominator equal to zero respec-
tively, gives the zero and the poles of the impedance function. Once poles and zeros are
known, the impedance may be cast in the form that suits the analysis of composite networks,
(15)
1) See section VI-1 on frequency stability in oscillators and section I-4 on transmission
of narrowband signals.
J.Vidkjær
While there is a tradition of using Q for characterizing frequency responses of

resonant RF circuits, the conceptually equivalent damping ratio ζ is often seen in pole-zero
and especially transient response calculations, where it leads to more compact expressions. We
proceed here with both forms and get,
(16)
Up to this point no attention was given to actual parameter values. The last result requires a
distinction between circuits having ζ≥1 and ζ<1 ( Q≤½ and Q>½ ). In first case the square
roots of Eq.(16) are real and the poles are on the real axis. In second case the poles move
from the real axis and make a complex conjugated pair. To emphasize this property we change
the last part of Eq.(16) to read
(17)
Fig.6 Position of poles s1, s2, and the zero s0 in the impedance
Zp(s) of the parallel resonance circuit with ζ<1 or Q>½.
The square roots in (a) are real valued if the poles are complex. The approximations in the
following lines apply to lightly damped circuits with higher and higher Q´s, where the
J.Vidkjær
expressions in Eq.(17)(b) are based on the estimate
(18)
Fig.6 sketches the geometry of complex pole and zero positions. Starting from damping ζ=1
( Q=½) the poles are by Eq.(17)(a) constrained to move along a circle of radius ω0 from the
real towards the imaginary axis with declining damping or growing Q. Once the poles and
zeros are known, the frequency characteristics of Zp may be calculated from geometrical
considerations as sketched in Fig.7 and Eqs.(19) to (21).
(19)
(20)
(21)
Fig.7 Calculation of Zp(ω) from poles

and zeros
Transient Response
Fig.8 Charging of capacitor C to voltage VC0 = QC0/C.
The projection of the poles on the imaginary axis determines the oscillatory modes
in the transient responses of the circuit. To see this, we consider the decay of circuit energy
when the circuit is left alone once the capacitor has been charged to a voltage of VC0 with the
inductor current initialized to zero. The initial charging of the capacitor is equivalent to
forcing a pulse current of strength
(22)
J.Vidkjær
through the circuit as indicated by Fig.8.2 The corresponding transient voltage decay Vdcy(t)
is the impulse response of the impedance function, which is given by its inverse Laplace
transform, i.e.
(23)
The last rewriting prepares for use of standard Laplace transform tables, from which we get
(24)
where the angular frequency is recognized as the size of the imaginary pole component if ζ<1.
Introducing the auxiliary phases and the identities
(25)
(26)
(27)
the time domain responses may be condensed to read
(28)
2) Note, the delta function δ(t) has dimension [sec-1] to comply with the requirement
J.Vidkjær
Fig.9 Transients from Eq.(28). Examples in (a) are heavily damped and (b) is lightly
damped. Time scales are in units of the resonance period T0=2π/ω0.
Fig.9 shows examples of the responses with different dampings. The case ζ=1, where the two
poles coincide on the real axis, sets the border between aperiodic and oscillatory solutions. In
the latter case, the leading exponential factor shapes the envelope to the solution
(29)
The corresponding time constant,
(30)
is sometimes called the logarithmic decrement of the circuit. Observe that the result is in
agreement with the previous equivalent baseband considerations in example I-4-2. In practical
terms we notice that the significant number of cycles through the decay approximates Q if Q
>≈ 5 so ζ≈< 0.1 and cosφ1≈1. It follows from the fact that at t=QT0, the exponential has fallen
from one to
(31)
so more than 99.8% of the initial energy is lost in the resistor at that instant.
J.Vidkjær
11
II-2 Series Resonance Circuit Summary
(32)
Fig.10 Series resonance circuit
A series resonance circuit is connected as shown by Fig.10 and Eq.(32) summarizes

its common parameter combinations. The circuit admittance and impedance are,
(33)
Compared to the parallel circuit impedance and admittance in Eq.(5) it is seen, that the
numerator and denominator are similar in structure with respect to s, but the coefficients are
different. Connected to a source, the two types of resonance circuits behave in duality. This
means that the reader could kindly be asked to repeat the preceding section exchanging terms,
voltage and current, parallel and series, impedance and admittance, inductance and capaci-
tance, resistance and conductance, and then we were done. To avoid confusion in future
references, however, the main concept of the series resonance is summarized below in its own
terms, but without detailed derivations.
Fig.11 Series resonance. (a) Reactance composition as function of frequency. (b) Voltage
and current phasors at the resonance frequency ω0.
At resonance in a series circuit the voltages across the inductive and the capacitive
reactances are equal in magnitudes but opposed in phases. With the same current flowing
through all components the requirement is that the two reactances balance each other at the
J.Vidkjær
resonance frequency as indicated by Fig.11(a). The phasor diagram in Fig.11(b) shows that
the quality factor now represents the ratio of the reactance voltage magnitudes over the voltage
vR across the series resistor Rs. At resonance this voltage equals the terminal voltage v. With
one terminal grounded, the potential at the interconnection between the inductor and the
capacitor becomes Q times as high as the potential at the driving terminal. This fact may
significantly influence the practical realization of high Q series circuits.
Like the parallel circuit in steady state resonance, a constant amount of energy is
exchanged between the capacitor and the inductor in the series circuit. In terms of energy
there are no differences between the two types of resonance circuits regarding the quality
factor, but the loss calculation must now be detailed as a series loss,
(34)
Using the resonance frequency and the quality factor from Eq.(32) the impedance of
the series circuit is expressed
(35)
The frequency is again accounted for through β(ω), so the impedance and admittance func-
tions become symmetric in logarithmic frequency scales as showed in Fig.12.
Fig.12 Impedance (a) and admittance (b) magnitudes and phases of the series resonance
circuit in Fig.10. The curves are symmetric around ωo due to the logarithmic fre-
quency scales.
Impedance and admittance magnitudes and phases in Fig.12(a) and (b) are given by
J.Vidkjær
II-2 Series Resonance Circuit Summary 13
Fig.13 Normalized magnitude and phases for the impedance of series resonance circuits
with varying Q-factors.
(36)
(37)
The equations are equivalent to Eqs.(8) and (7), so all results concerning bandwidth and
symmetry of bounds and characteristics may directly be overtaken from the forgoing section.
Fig.13 holds impedance characteristics with different Q-factors and shows the asymptotes set
by the capacitor and inductor below and above resonance respectively.
Poles and zeros for the series circuits are based on the expressions
(38)
Comparison to Eq.(14) reveals that the zeros of Zs(s) must follow the pattern of the poles in
the parallel circuit while the pole of Zs in origo corresponds to the zero of Zp there. In terms
of poles and zeros, the series circuit impedance is
(39)
where the geometrical properties of s0, s1, and s2 are the same as in Fig.6.
J.Vidkjær
14
II-3 Narrowband Approximations
Frequency dependency of the impedance in the parallel resonance circuits was

expressed through the nonlinear function β(ω). In consequence, we had to solve 2nd order
equations to find prescribed bandlimits. If the circuit is narrowbanded, a linear approximation
to the frequency relationship may suffice for design considerations around the resonance
frequency. A Taylor expansion provides
(40)
(41)
Following Eq.(9), the 3dB bounds are the frequencies where |Qβ(ω)|=1. With the approxima-
tion, bandlimits are placed symmetrically around ω0, and the 3dB bandwidth becomes
(42)
The same result was obtained in Eq.(10) from the nonlinear β(ω) expression, so the linear
approximation gives the right bandwidth. However, Fig.14 reveals that the approximation
places the 3dB interval below the correct one. But it is also seen in the figure that the greater
Q, the smaller spacing between the 1/Q and -1/Q lines, and the smaller is the error introduced
by approximating to symmetrical 3dB bounds through Eq.(41). To quantify this point
Fig.14 Comparison of true and approximate 3dB bandlimit and bandwidth calculation in
simple resonance circuits. The expression for the true middle frequency ωm is taken
from Fig.4.
J.Vidkjær
II-3 Narrowband Approximations 15
we calculate the relative difference between the approximated and the true value of the
bandlimits. They are equal to the relative difference between the corresponding middle
frequencies. It is ω0 in the linearization while the true middle frequency ωm depends on Q as
indicated by Fig.4. The relative error becomes
(43)
The figures show that the accuracy of the approximation is admirable for most applications
in circuits where Q is five or more. With Q exceeding one the accuracy may even suffice for
initial design estimations.
Inserting the linear expansion of β, the impedance of the parallel resonance imped-
ance circuit is approximated
(44)
In terms of frequency deviation from resonance, ω-ω0, the expression is as simple as a first
order lowpass characteristic, for instance the impedance of a resistor and a capacitor in parallel
as sketched in Fig.15. Thus, instead of the resonance curves in Fig.3 or Fig.5, we may take
Fig.15 Impedance of parallel RC circuit.

the impedance from a normalized first order lowpass characteristics like the one given in
Fig.16. Note, however, that compared to the lowpass case of Fig.15, the frequency deviation
in the last denominator of Eq.(44) is normalized with respect to half the 3dB bandwidth,
because there are 3dB limits on either side of ω0. The lower frequency bound in lowpass is
fixed to zero and not found from an expression. A standardized characteristic as the one in
Fig.16 has several interpretations, including
(45)
J.Vidkjær
Fig.16 Normalized first order lowpass characteristics. The sign of Ω must precede the
phase in bandpass interpretations, either complete or narrowband approximated.
It must still be kept in mind that the narrowband approximation is useful around ω0, but gives
wrong results far from this region. For instance, the true and the approximated bandpass
expressions in Eq.(45) are seen to give different results using ω = 0 or ω → ∞.
Fig.17 Contributions from poles s1, s2 and zero s0 to Zp in narrowband approximation

where (b) is an expanded view of the encircled region in (a).
Another approach to narrowband approximations takes an outset in the pole and zero
constellation. With high Q-values, the poles are close to the imaginary frequency axis. If we
limit our scope of investigation to the region around the upper pole, for instance the encircled
region in Fig.17, the distances to the other pole s2 and the zero s0 are long. As indicated by
the figure, their contributions to the impedance are taken constant, 2jω0 and jω0 respectively,
so the impedance becomes,
(46)
J.Vidkjær
II-3 Narrowband Approximations 17
Using the pole position estimates from Eq.(17)(c) and inserting the Q -factor from Eq.(1), the
approximation along the imaginary axis s=jω it calculated to yield
(47)
As seen, this is the same impedance approximation that was obtained in Eq.(44) on basis of
the Taylor expansion of β(ω). So it is concordant to approximate the impedance function by
the last parts Eq.(46) in s-plane calculations or by Eq.(44) along the jω axis.
Simplifications by the narrowband approximations for resonance circuits are beneficial

in many computations. Before leaving the subject we shall, however, expand the scope beyond
resonance circuits. The frequency dependency of a network is determined through the
reactances of capacitors and inductors. Suppose we know a network and its corresponding
frequency response H(s). This response may be transformed to another frequency range by
mapping frequencies while preserving individual reactances everywhere in the circuit, a
technique that is fundamental to filter design [1],[2],[3]. One such mapping - the only
one we consider - is the lowpass to bandpass transformation given by
(48)
Here subscripts "lp" and "bp" are introduced to distinguish between lowpass and bandpass
frequency planes. Along the imaginary axes where slp=jωlp and sbp=jωbp, the transformation
agrees with the lowpass and bandpass relations in (45). Applying the transformation directly
to circuit components maps inductors and capacitors in lowpass networks to series and parallel
resonance circuits in bandpass, where we get
(49)
(50)
Fig.18 Lowpass-to-bandpass trans-

formation of inductors and
capacitors.
J.Vidkjær
If the lowpass response of a network is known in form of a transfer or immitance3 function

H(s)=H(slp), the corresponding bandpass expressions, which accounts for the components
mappings, is the one obtained replacing slp by sbp. If the lowpass response is characterized by
poles, zeros, or other characteristic complex frequencies, we must solve for sbp in terms of slp
to get the corresponding bandpass properties, i.e.
(51)
A major reason to consider the lowpass to bandpass transformation is the fact that
many common, renowned frequency characteristics4 are described as lowpass prototypes. To
use them in bandpass circuits, we must use the lowpass to bandpass transformation. If the
transformation is based on pole-zero patterns, Eq.(51) must be employed. A center frequency
in bandpass, which is large compared to the bandwidth in translation, imply the simplified
solution
(52)
If the approximation applies, we have narrowband conditions. Here, a lowpass pole-zero

pattern around origo described by positions slp in the lowpass s-plane transforms to bandpass
by copying two linearly scaled versions of the patterns, one centering at jω0 and one at -jω0.
This is clearly much simpler than solving for the correct poles and zeros through Eq.(51).
Example II-5-2 in the next section discusses the question further. Like other narrowband
considerations the simplifications remain valid with poles and zeros close to the lowpass origin
or the bandpass centers. Far apart the true solution must still be employed, in particular with
respect to the zeros at infinity that are inherent properties of the lowpass characteristic. Zeros
at infinity means that the power of slp in the transfer function denominator is higher than the
power in the numerator. By Eq.(51) the lowpass plane point of infinity maps to two zeros in
bandpass, one at infinity and one in origo. The latter are not encompassed by the narrowband
approximation. It must be separately accounted for if the approximation method is used to
realize bandpass circuits.
3) Immitance is a collective name for impedance and admittance.
4) Names like Butterworth, Chebyshev, Legendre, and Cauer or features like elliptic or
equal-ripple are examples.
J.Vidkjær
19
II-4 Series-to-Parallel Conversions
Many practical situations include resonant circuits that are not ideal parallel or series
circuits. Clearly, it is always possible to elaborate impedance or transfer function expressions
in full details for the particular circuits in question. At single frequencies or under narrow-
band conditions, however, a technique known as series-to-parallel conversion may greatly
simplify the efforts to get useful results while keeping major insight on the circuit perfor-
mance. Due to the frequent - but often tacit - use of the method in technical literature and data
sheets, its foundation is presented here in some details.
Fig.19 and Fig.20 show a series connection and a parallel connection of a resistance
and a reactance. The resultant impedances and admittances are given in Eqs.(53),(54) and
(55),(56) respectively.
(53)
(54)
Fig.19
(55)
(56)
Fig.20
Forcing agreement between the two admittances gives the series-to-parallel conversions, i.e.
resistance and reactance conditions by which the parallel connection in Fig.20 may replace the
series connection in Fig.19.
(57)
Similarly, equating impedances gives the parallel-to-series conversions,
J.Vidkjær
(58)
If reactance is the dominating contribution to the impedance in consideration, that is either Rs

|Xs| in series or Rp |Xp| in parallel connections, the two types of conversion simplify to
the approximations
(59)
The last expressions are the ones most commonly associated with the concept of parallel-to-
series conversion. In this form they are particularly easy to memorize due to the symmetry of
converting back and forth between series and parallel representations. The following example
illustrates the technique of using the series-to-parallel method directly in design.
Example II-4-1 ( impedance matching )
Fig.21
A power transistor of known input impedance should match a 50Ω generator at 470 MHz
using the circuit in Fig.21, where biasing components are left out. Inductor L is fixed and C1,
C2 are trimmer capacitors. Find the trimmer settings and estimate the half power bandwidth
of transfer to the transistor.
Fig.22
To solve the problem we consider the equivalent circuit in Fig.22. A basic requirement for
matching is Rp = Rg, so the mapping of Rin to parallel form through the combined series
reactance Xs must equal Rg. When the required Xs is found, capacitor C1 is adjusted to tune
J.Vidkjær
II-4 Series-to-Parallel Conversions 21
out the corresponding parallel reactance Xp by setting X1=-Xp. Using terms from the figure
and the simplified conversions from Eq.(59), where Xp=Xs , we get
(60)
(61)
(62)
(63)
The value of Xs in Eq.(60) may seem marginal with respect to the prerequisites of the
simplified method. Without any assumptions about sizes of impedances, Eqs.(57),(58) provide
correspondingly
(64)
(65)
(66)
As seen only small changes follow from the more elaborate but correct conversions. To
estimate the bandwidth for power transfer, we consider the network as a parallel resonance
circuit and divide it into a capacitive and an inductive side as sketched in Fig.23. At the center
frequency the whole available power from the generator is transferred to the transistor since
there are no resistive losses in between. Therefore, the network determined above implies
conjugated matching across the cut. The parallel resistance Rpp, obtained by converting Rin
to parallel form through inductances only, must be equal to the parallel resistance, which
J.Vidkjær
Fig.23
originates from Rg and converts through the capacitive side. In a parallel resonance circuit the
Q-factor is the ratio of the total parallel resistance, here ½Rpp, over either the inductive or the
capacitive reactances. Continuing the simplified approach from Eqs.(60) to (63), which here
implies Xpp = XL +Xin, we get
(67)
Due to losslessness, the absolute level of the voltage transfer function at the center frequency
is most easily calculated letting Rin consume the available power, i.e.
(68)
This quantity wasn’t actually asked for, but now it possible to compare our results with
simulations as it is done in Fig.24. The fully drawn curve is calculated using capacitances
from the simplified method from Eqs.(62),(63) while the dotted curve is based on Eqs.(65),
(66). The observable consequence of the simplified method is a slight displacement of the
Fig.24
J.Vidkjær
center frequency where matching is obtained. However, the differences in capacitance values
between the two situations are easily compensated with trimmer capacitors.
One final point should be observed concerning the bandwidth estimation. The transistor
impedance to be matched was taken from data-sheets, and the positive reactance part was
treated by Eq.(66) like an inductor for bandwidth estimations. Here it is a tacit assumption that
the input reactance does not change faster than an inductor across our frequency band, i.e.
without sensible resonances from the transistor or its mounts. If this occurs, we have no means
for estimating bandwidth, but the center frequency matching procedure remains valid, if the
transistor data are reliable.
Example II-4-1 end
Conversions in Narrowband Applications
Fig.25 Conversion of a small inductor series resistance RLs to a parallel resistance RLp .
Circuit (b) is a narrowband approximation to (a) around resonance frequency ω0.
Reactances of inductors and capacitors depend on frequency so - strictly speaking -

a series-parallel conversion applies to a single frequency. However, we could estimate the
bandwidth of the resultant circuit in the example above, a consequence of the fact that in
narrowband circuits, component values obtained by series-parallel conversions at the center
frequency are useable throughout the passband. To see this from an analytical point of view
we consider the example in Fig.25. Inductor L is employed in a parallel resonance circuit, but
it has a small series resistance RLs, so the resonance circuit is no longer ideal. To deal with
this circuit easily, RLs is converted to the parallel resistance RLp and parallel resonance
methods are used on the circuit in Fig.25(b). It is supposed that the series resistance is much
smaller than the reactance Xs = Lω0, so the simplified relations from Eq.(68) are used
including the fact, that the inductor reactance remains unaffected of the conversion. Therefore,
resonance frequency ω0 is figured out the usual way and the corresponding Q-factor and band-
width follows from the sequence,
(69)
J.Vidkjær
Notice that the contributions from the two resistors to the resultant Q-factor are distinguishable
through a "parallelling" relationship
(70)
(71)
It is not self-evident that the circuit in Fig.25(b) with frequency independent RLp may
substitute the original circuit across a passband. To convince, the two impedance functions
must be compared. The ideal parallel circuit is given by Eq.(14),
(72)
Casting the impedance of the original circuit in Fig.25(a) into a comparable form we get
(73)
Identification of Q-factor through the first order denominator term gives
(74)
which agrees with the subdivision of Qtot in Eq.(71), provided that the resonance frequencies
ωm0 and ω0 are equal. But this is not so unless the resistors meet the condition
(75)
Implicitly the inequalities express a narrowband assumption. The result of a parallel connec-
tion cannot exceed any of its components. If Qm=Qtot 1 we must therefore require
J.Vidkjær
(76)
Under narrowband conditions the two circuits in Fig.25 will have the same poles because the
impedance functions get the same denominators. Regarding numerators Eq.(74) shows that
Zm(s) has a zero on the negative real axis compared to the zero in origo for Zp(s). Therefore,
the impedance Zm reduces correctly to the parallel combination RLs Rxl at dc while the
inductor short-circuits in the narrowband approximation.
(77)
Observe, however, that the difference in zeros between Zm and Zp is an asymptotic deviation
outside the scope of the narrowband assumption. If the response is dominated by poles, which
is the case in the passband, the parallel circuit obtained by series-to-parallel conversion at ω0
is usable to the same degree of confidence that accompanied other types of narrowband
approximations.
Parallelling of Q-factor contributions has wider implication than just being a vehicle
in the previous discussion. Imperfections in reactive components that cause loss of power may
be specified in data sheets or measurements by associating a quality factor directly to the
component. In agreement with Eq.(71), the Q-factor is either the ratio of the reactance over
a series loss resistance or the ratio of a parallel loss resistance over the reactance. With higher
Q´s, say three or more, it makes no difference for narrowband computations whether the
physical loss originates in series, parallel, or combined connection, we convert to the form
most suited for the problem at hand through the approximate conversion from Eq.(59),
(78)
The combination of Q-factors is sometimes expressed in a terminology of loaded, unloaded,

and external Q´s, which follows the paralleling expression, cf.[4] sec.7.1,
(79)
Subscript "unloaded" refers to unavoidable loss components in the reactances that make the
resonator and "external" to all other resistors. The resultant "loaded" Q determines the
bandwidth around resonance and is always smaller than any component Q-factor. In this
terminology the example from Fig.25 and Eq.(71) should read
(80)
J.Vidkjær
Conversions in Broadband Modeling
In narrowband applications of reactive components it suffices to specify its reactance

and Q-factor as a function of frequency. Then we may use Eq.(78) and the series-parallel
conversion around the center frequency to make design calculations. On the contrary charac-
terization of reactive components for broadband application, including modelling by frequency
independent circuit elements, requires detailed equivalent circuits. To find the models we may
still benefit from the series-parallel relationships. The following example illustrates this aspect.
Example II-4-2 ( spiral inductor )
Fig.26 Spiral inductor forward y-parameters. Reverse y12 and y22 are similar. Data are
converted from s-parameter measurements. g21´s are uncertain at high frequencies.
Data from an inductor in a GaAs microwave integrated circuit are shown in Fig.26.
The layout is a planar spiral of the type in Fig.27 and the measurements are made with the
inductor connecting an input and an output port. Due to small dimensions, thin metal and
isolation layers, the inductor is far from being ideal. We wish to find an equivalent circuit
based on the measurements. Like other integrated passive component a Π-structured model
is expected, so it is natural to translate the experiments to y-parameters. The definitions in
forward measurements take form of input and transfer admittances with shorted output as
summarized by Fig.28.
Fig.27 Spiral inductor layout example. Fig.28 Definition of y-parameters y11, y21.
J.Vidkjær
Fig.29 Spiral inductor equivalent circuit where (a) suffices below the resonances that are
included by (b). The shorts at the outputs are required to interpret y-parameters.
Fairly below 10 GHz the data show nearly ideal asymptotic behavior, where imaginary and
real parts are inversely proportional to frequency or squared frequency respectively. The latter
indicates that the dominant inductor loss is a series loss. To see this we convert the simple
equivalent circuit in Fig.29(a) to y-parameter parallel form through
(81)
The real conductance component shows the observed second order relationship. Extrapolating
the declining asymptotes to the frame of the figures at 31.62 GHz gives
(82)
The series resistance is high and troubles many designs using integrated spiral inductors. At
5GHz, for instance, the experimental data ( dots in Fig.26 ) are close to the maximum Q-
factor, which develops
(83)
Above 10 GHz the data indicate resonances in both y-parameters. The susceptances of the
inductance are here canceled by stray capacitors, at f01=11.52 GHz in y11 and at f02=18.16
GHz in y21. To include resonances the model is enhanced to Fig.29(b), which now gives
(84)
J.Vidkjær
(85)
(86)
A final look at the measurements shows a distinct asymptote in the real part of y11 above
resonance f01. The asymptote increases in proportion to the squared frequency, which might
be the effect of a small resistance in series with a capacitor. It is therefore tempting to make
Fig.30 Complete y11 model and the elements that are sensed above resonance.
one more step in the modeling. Considering y21, the real parts of the experimental data are
too noisy and uncertain for further identifications, and we concentrate on y11 as shown in
Fig.30. Now the asymptote for g11 at the frame of the data implies
(87)
Including the last extension, the equivalent circuit for the spiral inductor takes the shape of
Fig.31. The y-parameters, which this model accounts for, are shown in Fig.32. The only
observable discrepancies to the measurements are in g21 above resonance, where the experi-
Fig.31 Complete equivalent for the spiral inductor measurements.
J.Vidkjær
mental data are badly conditioned. In view of literature on GaAs IC design, for instance
[5], the model we have constructed from nothing but basic knowledge on series-parallel
transformations and resonance circuits is rather complete.
Fig.32 Simulated verification of the complete spiral inductor equivalent circuit from Fig.31.
The corresponding experimental data were shown in Fig.26.
Example II-4-2 end
J.Vidkjær
30
II-5 Tuned Amplifiers
Fig.33 Functional and simplified equivalent circuit of a single-tuned amplifier.
Resonance circuits are used to shape the frequency response of frequency selective
amplifiers. Fig.33. shows an example where a bipolar junction transistor is loaded by a parallel
resonance circuit. A simple transistor equivalent circuit is employed to keep the amplification
function uncomplicated. Using Eq.(45) the voltage gain v2/v1 is expressed
(88)
The transistor output capacitance and conductance add to the external tuning circuit compo-
nents to give center frequency amplification A0 and bandwidth,
(89)
Note, the input resistance and capacitance in the transistor model have no direct effects on
these expressions, but play the role of loads if more stages are cascaded. The frequency
response, which shapes like the parallel circuit impedance function Ztot(jω), may be calculated
either fully correct or narrowband approximated by the frequency expressions
(90)
Recall from the discussion on the narrowband approximation in section II-3 that the two forms
differ with respect to bandlimits but not bandwidths.
To compare frequency selective amplifiers, two figures of merits are the gain-band-
width product and the gain-bandwidth factor. The product is defined as the center frequency
voltage gain times the 3dB bandwidth,
(91)
J.Vidkjær
II-5 Tuned Amplifiers 31
5
With more stages the GW-product commonly refers to an average gain per stage, i.e.
(92)
where N is the number of tuned stages. The single stage amplifier above has
(93)
This product is used as a reference for the gain-bandwidth factor, GBF, which is defined by
(94)
It is supposed, that the transconductance and the total loading capacitance are the same in both
numerator and denominator. Transconductance gm is a property of the transistor. In the lower
limit Ctot holds the transistor output capacitance and other unavoidable contributions from
mounts and loads. Therefore, GW is often used as a goal for optimizing or comparing
transistor and IC performances. The normalized GBF is suited for comparing amplifier
structures without detailed regards to the devices that give the gain.
There exists a series of fundamental theorems concerning GBF’s of interstage

connections between amplifier stages, first introduced by Bode [6]. Their derivations and
detailed interpretations are beyond the present scope and needs. Briefly, when one amplifier
stage is connected to the next, it is stated that GBF cannot exceed two with a passive one-port
paralleled across the connection. A passive two-port interstage coupling has a theoretical maxi-
mum of four. Networks representing the maxima are complicated and of little practical use
with the high gain RF transistors of today. However, it is still informative to use GBF in
comparisons between different amplifier structures even when they fall considerably below
the theoretical limits.
An amplifier stage holding only one resonance circuit is called a single-tuned

amplifier. With more resonance circuits separated by transistors there are many possibilities
on how to organize their center frequencies and Q-factors. The simplest case is that all tuning
circuits have identical resonance properties. This well-defined situation is called synchronous
tuning, as opposed to a more diversified group of so-called stagger tuned amplifiers.
5) There are many competing definitions and notations around. Bandwidth in units of
radians per second is denoted GW here, while GBW is used if bandwidths are in
units of Hz. The term MGBW - "M" for mean - may be seen instead of GWav.
J.Vidkjær
Synchronously Tuned Amplifiers
Cascading N equally tuned amplifier stages of the type from Fig.33 gives a voltage
amplification of form,
(95)
The center frequency amplification AN0 at Ω = 0 is the product of the center frequency
amplifications in each stage. In normalized frequency a single stage amplifier has 3dB
bandlimits corresponding to Ω=±1. With N equally tuned stages we must solve for the Ω’s
that reduces the absolute amplification by a factor of 2, that is
(96)
Without normalization with respect to frequency, the result gives the bandwidth of N stages
in terms of one stage through
(97)
where the gain-bandwidth factor GBFN in this particular application also is called the
bandwidth reduction or bandwidth shrinkage factor. Examples of the reduction process are
Fig.34 Normalized amplitude and phase characteristics in synchronous tuning. Gain-band-

width factors GBF2, GBF3 give the bandwidth reductions for 2 and 3 stages.
J.Vidkjær
seen in the characteristics of Fig.34. The amplitude of N equally tuned stages rolls off
approaching an asymptote of -N×20dB/decade. The stronger the bending of the corresponding
curve, the smaller becomes the bandwidth. Since 21/N → 1 with growing N , the reduction
factor may be approximated from the following estimations,
(98)
True and estimated bandwidth reductions are compared below in Table I. Whenever more
stages are cascaded, the approximation gives reasonable results.
Table I True and approximated Gain-Bandwidth Factors ( Bandwidth Reduction Factors

) for N Synchronously tuned amplifier stages.
N GBFN ∼ GBFN N GBFN ∼ GBFN
2 0.6436 0.5893 5 0.3856 0.3727
3 0.5098 0.4811 6 0.3499 0.3407
4 0.4350 0.4167 7 0.3226 0.3150
Example II-5-1 ( synchronous tuning )
Fig.35 Amplifier principle, (a), and transistor equivalent circuit, (b).
The amplifier in Fig.35(a) should operate with Rg=RL=75Ω having a total bandwidth in
synchronous tuning of BWamp=55 MHz around f0=460 MHz. Transistor data are
(99)
Find the external components C0, L0, C1, L1, and the center frequency gain v2/Eg.
J.Vidkjær
Fig.36
Including the transistor model, a complete functional equivalent circuit for the amplifier
becomes the one shown in Fig.36 where the generator is changed to the Norton equivalent.
The two tuned circuits have equal bandwidths BW0 and quality factors Q0. Compensating for
bandwidth reduction, we get
(100)
At the input side, the parallel resistance R00 determines the impedance level and in turns give
the external components,
(101)
The corresponding calculations at the output side of the transistor read,
(102)
The inductances we have found here approaches the lower borderline for practical lumped
inductors. A guideline is that unwounded wires like component leads have inductances about
0.5nH/mm to 1nH/mm, a result that is verified in chap.4 of ref.[7]. The smallest inductor
available as a commercial component for surface mounting on PC-boards is presently 2nH.
J.Vidkjær
The impedances of the tuned circuits are equal to the parallel resistances at the center frequen-
cy, so the voltage gain becomes
(103)
The frequency characteristic of the amplifier is shown later in Fig.43.
Example II-5-1 end
The phase characteristic in synchronous tuning with N stages may be written
(104)
where the last equation introduces the Taylor expansion of tan-1(Ω). Instead of using Ω, which
is normalized with respect to the 3dB bandwidth of a single stage, we normalize frequency
with respect to the approximate 3dB bandwidth from Eq.(98) for the N stages in consideration.
The new frequency variable becomes
(105)
Inserting ΩN into the phase characteristic above gives
(106)
As N raises, the nonlinear third, fifth, and higher order terms loose significance compared to
the linear term. With more stages, the phase characteristic in the 3dB bandwidth |ΩN|<1
becomes more linear or - equivalently - the group delay stays more constant. The correspond-
ing amplitude characteristic approaches a Gaussian curve in the passband. To see this we
rewrite the amplitude expression using the fact that, cf.[8] p.228,
(107)
As bandwidth in measures of Ω reduces, the limit value is suited for large N, where we get
J.Vidkjær
(108)
The exponential gives a Gaussian characteristic due to the squaring of the frequency variable,
and the last version shows readily that the scaling from Eq.(105) gives 3dB limits with
ΩN=±1.
Butterworth Stagger Tuned Amplifiers
Each stage in a chain of amplifier stages provides a pole zero pattern corresponding
to its load impedance. Synchronous tuning let the poles and zeros from all stages coincide.
Staggering the pole-zero patterns gives provisions for a variety of amplification characteristics.
We shall demonstrate the concept of stagger tuning by picking a simple but common case, the
family of Butterworth characteristics of various order. They are defined by requiring maximal
flatness in magnitude, which means that derivatives with respect to frequency up to the order
2N-1 are zero in the center at Ω=0 6. The normalized N-th order Butterworth function is
(109)
Examples of the magnitude characteristics are given in the upper part of Fig.37. The 3dB
bandlimits corresponding to Ω=±1 are independent of order, but as N increases the roll-off
from the passband becomes more and more abrupt.
To find the pole patterns that give the normalized Butterworth characteristics, we start
considering the squared magnitude,
(110)
Tracing back from jΩ along the imaginary axis to the corresponding normalized complex
frequency variable S implies the substitutions
(111)
(112)
6) These and related concepts are part of the approximation problem in filter design.
Consult ref´s [1], [2], or [3] for further details including the multitude of cases that
are left out in this text.
J.Vidkjær
Fig.37 Normalized Butterworth characteristics of orders up to N=5. The phases are deter-
mined from pole positions like Fig.39.
Solving for the S values that set the denominator equal to zero provides poles for the squared
magnitude
(113)
The poles are confined to the unit circle where they are equally spaced with an angle of π/N.
When index k runs from 0 to 2N-1, all poles positions are covered once in the sequence
sketched by Fig.38.
Fig.38 Pole patterns in N-th order Butterworth squared magnitude |BN(S)|2.
J.Vidkjær
The Butterworth amplitude characteristic applies to lowpass as well as bandpass

amplifiers. Both types may be realized using tuned LC circuits as interstage networks. The
lowpass case ease the discussion because we transform from the normalized frequency Ω to
the lowpass ω simply by multiplying with the bandwidth W3dB. Thus, to represent a causal
lowpass system, the basic requirements of having even real and odd imaginary parts along the
imaginary frequency axis are met by imposing the condition
(114)
so the squared amplitude function may be rewritten
(115)
By the last factorization, poles of |BN|2 having positive real parts can be ascribed to BN(-S).
This assures that the transfer function BN(S), which is targeted in circuit design, fulfills a
necessary requirement of stability, as it only collects the N poles in the left half-plane. They
were given by the first N indices in Eq.(113), in summary
(116)
Pole positions in the second, third, and fourth order normalized Butterworth functions - also
called prototype functions - are detailed in Fig.39.
Fig.39 Examples of pole positions on the unit circle in normalized prototype Butterworth
functions BN(S).
Poles sp,k in a bandpass amplifier of center frequency ω0 and bandwidth W3dB are
related to the prototypes in Eq.(116) through the lowpass to bandpass transformation intro-
duced in section II-3. Without any assumptions they provide
(117)
J.Vidkjær
Under narrowband conditions, where W3dB/ω0 1, the square root is dominated by its second
term and the poles are approximated
(118)
Thus, in narrowband we scale the prototype to half the desired bandwidth and copy the pole
pattern from its lowpass center in origo to the bandpass centers at s=±jω0. This process is
demonstrated for a third order function in Fig.40.
Fig.40 Mapping of Butterworth prototype poles (a) to poles in a narrowbanded bandpass

amplifier (b). In (c) the poles are paired for stagger tuning that also requires three
zeros in origo.
To realize the third order Butterworth characteristics exemplified by Fig.40(b), we

need three basic parallel tuned stages. The idea of stagger tuning is to adjust the load of each
stage to a required pole pair as indicated by Fig.40(c) and Fig.41. Fortunately, the parallel
resonance circuits insert the necessary zeros at origo. As discussed in section II-3 they are not
automatically encompassed by narrowband techniques. The sequence of pole pairs is arbitrary,
but if we choose to let stage k realize the pole pair define by sp,k in Eq.(118), its resonance
frequency ω0k and quality factor Qk are given by,
Fig.41 Three stage tuned amplifier. By stagger tuning the pole pair of each resonance
circuit corresponds to at pole pair in the transfer function like the example in
Fig.40(c).
J.Vidkjær
(119)
Pole positions determine the shape of the frequency response but not the absolute amplifi-
cation level. The quantities ω0k,Qk constrain the tuning components by
(120)
It is a design decision to set the absolute impedance level. If the total amplifier is narrow-
banded in the sense of W3dB/ω0 1, each stage must be narrowbanded. By the approximation
from Eq.(46) the gain of stage k, which realizes pole pair sp,k becomes
(121)
The last rewriting is based on Eq.(118) and refers back to the prototype function poles on the
unit circle. At the center frequency ω=ω0, stage k has the voltage gain
(122)
In a complete chain of stages for a Butterworth characteristic, there is a real, negative center
frequency factor if the order is odd. The Sp,k’s in the remaining stages appear in complex
conjugated pairs, so the center frequency gain of a N-th order stagger tuned Butterworth
amplifier is given by
(123)
To elaborate further, the number of stages and an impedance strategy must be known.
Consider as an example a third order amplifier where all transistor transconductances and load
resistances are equal, gm0=gm1=gm2=gm and R0=R1=R2=Rp respectively. We observe from the
third order prototype in Fig.39 that the k=1 stage tunes to the center frequency. Using
Eq.(119) we get
(124)
The real values of poles with k=0 and k=2 are half the size of the k=1 pole, which gives
(125)
J.Vidkjær
Knowing all capacitances, Eq.(123) provides the center frequency gain in this particular setup
(126)
To estimate gain-bandwidth factors if loading capacitances differ among the stages, the one
with the smallest capacitance, i.e. the greatest single-stage GW, is commonly chosen as the
reference. In the present amplifier this is clearly the C1 stage, so we get
(127)
Example II-5-2 ( Butterworth amplifier )
By this example we change the components of the synchronously tuned amplifier from
Example II-5-1. Keeping bandwidth, center frequency, and impedance level specifications, the
frequency characteristic should now be of 2nd order Butterworth type. The task is to get the
new component values and calculate the center frequency gain.
First we find the required pole-positions. Assuming narrowband condition we consider the
pattern in the upper half-plane, Fig.42, where simple geometrical reflections give the coordi-
nates of the poles.
Fig.42
Associating the upper pole with the input circuit, the Q-factors of both tuned circuit are
(128)
Using resistance figures from Example II-5-1, the tuning components now become
J.Vidkjær
(129)
(130)
Recalling the discussion in the previous example, we are very close to practical bound on
small inductances. Note, however, that Example II-6-4 shows one way to transform so compo-
nent values stay practical. To find the center frequency gain Eq.(123) is used directly,
(131)
Compared to synchronous tuning, Eq.(103), maximal flatness halves the voltage gain.
The synchronously tuned amplifier from Example II-5-1 is shown with the present stagger
tuned Butterworth amplifier in Fig.43. The curves are simulated results from the equivalent
Fig.43 Simulated voltage gain magnitude, phase, and group delay for the synchronously
and stagger tuned amplifiers in Examples II-5-1 and 2. Narrowband approximated
circuit data are used.
J.Vidkjær
circuit in Fig.36 without further assumptions. It is therefore worth noticing how close we come
to the specifications regarding gain and bandlimits, although all underlying component
calculations were based on narrowband assumptions. The most visible consequence of the
simplifications is the lack of complete flatness in stagger tuning. Had the poles been correctly
found by solving Eq.(117) instead of the simpler scaling and copying approach in Eq.(118),
the result improves as will be seen below. From a design point of view it is dubious to go
further analytically. Practical component values need fine-tuning, either physically or by circuit
optimization, to compensate other design simplifications, for instance the employment of
uncomplicated transistor models or the ignorance of parasitic elements in the lay-out. The
simple narrowband methods give a good starting point for this process. However, in the
present context we shall enlighten the assumptions and approximations behind narrowband
methods whenever possible and therefore compare the previous result with their true counter-
parts. Applying prototype poles to Eq.(117) gives
(132)
Realize that by this calculation we give up the narrowband assumptions in the complete
Butterworth amplifier characteristics. To identify Q00,ω00 or Q11,ω11 from the pole positions,
narrowband assumptions about the individual stages are maintained. With Q-factors over 10,
Eq.(43) shows, that this is still a very satisfactory assumption. Having established the Q-factor
and resonance frequencies for the two stages, computations similar to Eqs.(129), (130) lead
to new components values,
Fig.44 Simulated voltage gain magnitude, phase, and group delay for stagger tuned
amplifier in Examples II-5-2. True pole positions are used to calculate circuit data.
J.Vidkjær
(133)
Fig.44 shows the simulated frequency response obtained by these data, and clearly the flatness
in magnitude has improved. The reason why group delay no longer gets symmetric appearance
is the fact, that the Butterworth characteristic is symmetric in the normalized frequency Ω, not
in ω with respect to which, the phase was differentiated.
Example II-5-2 end
J.Vidkjær
45
II-6 Transformers and Transformerlike Couplings
Needs for transforming signal and impedance levels in RF circuits are both frequent
and diversified, so a multitude of approaches and techniques are available for solving that sort
of problems. Among them are the conventional magnetically coupled transformer, which is
a highly useful component at frequencies up to approximately 3 GHz. We shall consider
transformers and circuits that behave similarly in some depth in this section. The scopes are
to gain basic understanding of advantages and limitations of transformer couplings, and to
provide a background for setting up simulator models. Some simulation programs are sparsely
equipped with the coupled inductor and transformer models that are needed in RF-design, so
they must be build from basic circuit functions and components.
Review of Mutual Inductances
Two or more inductors have mutual inductance if they interact through their magnetic
fields. To fix ideas we start considering two inductors where - as shown in Fig.45 - currents
are separately applied. Superposition of the two conditions gives the following expressions for
terminal voltages and the total magnetic fluxes Φ1 and Φ2 through the inductors,
(134)
Fig.45 Inductors with mutual inductance. Surfaces for the flux integrals are bounded by the
coils and lines through the terminals. Dots indicate orientation. Current entering a
dotted terminal support flux in the other coil.
The flux integrals condense in the inductances L1, L2, and the mutual inductances M12, M21.
They are factors of proportionality between inductor currents and the different flux
contributions. The factors are the elements of an inductance matrix, which finally gives the
two-port impedance matrix for two coupled inductors,
(135)
J.Vidkjær
Including more than two inductors, the equations are generalized to
(136)
Here v, Φ, and i are vectors holding the port voltages, fluxes, and currents respectively, while
L and Z are the inductance and impedance matrices. The inductance matrix is assumed to be
symmetric and positive definite or semidefinite. The first property expresses reciprocity, which
apply if the inductors have isotropic surroundings. It evolves from the renowned Lorentz
reciprocity theorem, [4] sec.2.12,4.5 , and causes the total magnetic energy to depend only
upon the instant current vector, no matter how it behaved in the past. The second property is
a passivity requirement ensuring that starting from zero initial conditions, the total stored
magnetic energy will always be positive or zero, [9] chap.15. In mathematical terms, the
requirements are that all subdeterminants of L, which can be taken symmetrically around the
diagonal, must be positive or zero, cf.[10] sec.7.2.
With two coupled inductors, the reciprocity and passivity conditions become
(137)
(138)
The limit cases of L1=0 or L2=0 implying M=0 are of no practical interest. Mutual inductanc-
es are often expressed by the coupling coefficient k,
(139)
In circuit schematics, mutual inductances may be represented as shown in Fig.46(a,b). If we

start fixing the orientations of the two ports, the sign of M and k is implicitly determined.
Applying a positive current to port one, M and k are positive if both port voltages are in
phase, negative if they are 180° out of phase. To emphasize phases rather than the more
arbitrary port orientation, the dot-convention, which defined flux directions in Fig.45, will also
tag a set of terminals that gives in-phase port voltages. Using port orientations opposite the
dot indication implies negative mutual inductances or couplings as shown in the figure.
Fig.46 Inductors with (a) positive and (b) negative mutual inductances following the dot-
convention from Fig.45. The signal flow-graph represents the z-parameter matrix.
J.Vidkjær
II-6 Transformers and Transformerlike Couplings 47
The borderline to passivity, |k|=1, is called tight or close coupling and is one goal
aimed upon when mutual inductances are employed in transformers. With k=±1 the two
voltage equations that stem from Eq.(135) become
(140)
However, the inductance matrix is singular in this case, so the two equations are linearly
related, and the lower equation follows from the upper one
(141)
The square root of the inductance ratio is called the winding or turns ratio N, because the
inductances of typical transformer coils are proportional to the squared number of windings,
cf.[7] sec.4.8. Equation (141) expresses the transformer voltage relationship
(142)
To share and confine the magnetic field, the inductances are often wound on a core of high
permeability material, for instance a toroide like Fig.47. The phrases of tight or close cou-
plings refer to the fact, that the two inductors must encompass the same magnetic flux when
|k|=1. Suppose the two coils are similar so they have the same ratio AL between inductance
and the squared winding count,
With k=±1, the flux per winding in L2 that originates from current i1 equals in size the flux
per winding in L1,
(144)
Fig.47 Toroidal transformer where both inductors share the same magnetic field.
J.Vidkjær
This result means that the two coils encompass the same magnetic flux when driven from the
primary port (no.1). Due to reciprocity this will also be the case if the current is applied to
the secondary port. With k less than one, part of the flux from L1 leaks outside L2 and vice
versa so the pictures tend towards Fig.45. We shall see later that flux leakage deteriorates the
transformer frequency response.
Two tightly coupled inductors are sometimes called a perfect transformer, cf.[10]
pp.44. It is a step towards the more familiar ideal transformer, and should not be confused
with it. From the lower equation in (135) we have
(145)
Letting the two inductances simultaneously rise towards infinity - keeping the winding ratio
constant and the voltage time integral limited - provides
(146)
(147)
which is the ideal transformer current condition. It is more restrictive than Eq.(142), as it
requires infinitely large inductors, while the voltage condition only assumes tight coupling.
In combination the voltage and current conditions define the ideal transformer. It has only
algebraic constraints between the terminal currents and voltages, and it is solely specified by
the winding ratio N. The ideal transformer is symbolized as shown in Fig.48(b), where the two
lines between the inductors indicate a hypothetical high permeability core that would be
necessary to approximate an ideal transformer in practice. Be aware that there is no conven-
tion on the side taken as unit reference for N although we often use port one, the primary side,
as reference.
Fig.48 Progressing form tightly coupled inductors in (a) to the ideal transformer in (b) if
the inductors approach infinity with a constant winding ratio N.
J.Vidkjær
Equivalent Circuits for Two Coupled Inductors
Instead of a circuit diagram that indicates mutual inductance, it is often more

informative to represent a pair of coupled inductors by an equivalent circuit, where the
inductors are uncoupled. The impedance matrix from Eq.(135) directly suggests the T-form
in Fig.49(a). However, this circuit has no DC separation between the two inductors. If it is
important to portray this property in a circuit diagram, the enhancement by an ideal 1:1 trans-
former as shown in Fig.49(b) is straightforward.
Fig.49 T-equivalent modeling two coupled inductors by three uncoupled inductors. Dia-
gram (a) suffices for dynamic responses. The ideal transformer in (b) emphasizes
DC separation between the inductors.
Including an ideal transformer, the circuit in Fig.49(b) is one particular choice among
all possible T-equivalent circuits for two coupled inductors. To see this we consider the T-
circuit and the corresponding z-parameter signal flow-graph in Fig.50.
Fig.50 T-equivalent circuit and a signal flow-graph for calculating z-parameters.

Forcing the graph to provide the original z-parameters from Fig.46(c) yields
(148)
Here, three equations constrain four unknowns, Lx, Ly, Lz, and the effective winding ratio n,
so one condition must be fixed initially. Above in Fig.49(b) n was set to one. To account for
J.Vidkjær
nearly perfect transformers it is more natural to get an n value close to the physical winding
ratio, which is achieved choosing Lx=0 or Ly=0. To solve for the first requirement we use
(149)
and the resultant circuit becomes the one in Fig.51(a). The circuit in Fig.51(b) has Ly=0,
which gives
(150)
In both cases n comes close to the physical winding ratio, which follows the square root of
the inductance ratio, when the coupling tightens towards k=±1. In the limit, furthermore, the
series inductances from Lx and Ly approach zero leaving alone the primary inductance L1
across the input port. Therefore, the T-type of equivalent circuit is especially suited for tight
coupling calculations in transformer designs.
Fig.51 Equivalent circuits suitable for tightly coupled inductances. For |k| → 1, the effec-
tive winding ratio n approaches the physical winding ratio and, simultaneously, the
series spreading inductances disappear.
Fig.52 Equivalent circuits suitable for tightly coupled inductances. The circuits are similar
to the cases in Fig.51, but now the ideal transformer is moved to the primary side.
J.Vidkjær
Had we from the outset placed the ideal transformer on the primary side, the resultant
equivalent circuits would be the ones in Fig.52. The parameters follow from mirroring
previous results by exchanging terms and inverting winding ratios.
An alternative to the T-equivalent circuit is the Π diagram in Fig.53. This circuit

structure is most conveniently described by y-parameters, so the component values are
calculated from term by term comparisons, inverting the original z-parameters to an admit-
tance matrix, i.e.
(151)
(152)
The last equation shows that the Π-diagram precludes tight coupling, because the denominator
determinant becomes zero if k=±1. For |k|<1, component values are calculated,
(153)
(154)
(155)
Fig.53 Π-equivalent for two coupled inductors where the DC separation is emphasized in
(b). Eqs.(153) to (155) give components in terms of original parameters.
J.Vidkjær
The results are reasonable as in the limit |k|→0, the two parallel inductors go towards the
inductances prior to coupling, La→L1, Lb→L2, while the connecting inductor Lc and its
pertinent impedance rise towards infinity and uncouples the two sides. Therefore, the Π-type
equivalent circuits are often preferable when dealing with loosely coupled inductors.
Fig.54 Π-equivalent circuit and a signal flow-graph for calculating z-parameters.
Like the T-circuit, the diagram in Fig.53 is one particular choice where the winding
ratio n is set to one and the remaining three unknowns La, Lb, and Lc have been solved for.
To get expressions for other possibilities, the easiest way is now to repeat the process from
the T-circuit, i.e. make a z-parameter signal flow-graph and identify terms. The graph in
Fig.54(b) holds two loops that give a denominator D different from one. We get
(156)
A particular solution, which is useful for considering the impedance conversions through
loosely coupled inductors, emerges if the two parallel inductors are set equal to each other.
Here a solution process leading to the result in Fig.55 is
(157)
J.Vidkjær
Fig.55 Π-equivalent circuits with equal parallel inductances. The circuit is suitable for
loosely coupled inductances.
Observe, that the effective winding ratio in the ideal transformer is independent of the
coupling coefficient k and equals the physical winding ratio N of the two coils. Even with
loose coupling it is easy to incorporate the effect of secondary side loading in computations.
The coupling coefficient in terms of the Π-equivalent circuit follows directly from Eq.(156),
(158)
The last equivalent circuit for two inductors with mutual inductance we consider is
shown in Fig.56(a). Here the coupling is accounted for by two voltage controlled voltage
sources. The method is verified by identifying paths in the signal flow-graph compared to the
z-parameters in Fig.46(b). This diagram may be useful in simulation programs that supports
controlled sources without including dedicated models for coupled inductors.
Fig.56 Equivalent circuit for inductors where the coupling is simulated by voltage con-
trolled voltage sources.
RF-Transformers
The most prominent use of two coupled inductors is the transformer, where the aim
usually is to get a coupling coefficient as close as possible to unity. If, first, the reactances of
the inductors are high compared to any connected impedance, the transformer approximates
the ideal one. Besides the current and voltage constraints from Eqs.(142),(147), impedances
are transformed as indicated by Fig.57,
J.Vidkjær
Fig.57 Ideal transformer with terminal voltage and current conditions and the correspond-
ing input and output impedances.
(159)
An ideal transformer was included in nearly all the equivalent circuits above, so these relations
are constantly used.
Ideal transformers are independent of frequency, but there are both upper and lower
frequency bounds in practical transformers. Direct causes are the finite inductances and less
than unit coupling, which were accounted for in all the equivalent circuits of Fig.51 or Fig.52.
If a transformer is supposed to match load resistances RL to generator resistance Rg, the
winding ratio must be chosen to transform RL to RL´=Rg, when it is seen from the generator
side of the transformer. Using the equivalent circuit from Fig.51(b), the frequency response
is described by the diagram in Fig.58(a). The low frequency 3dB limit fl is determined by the
shunting effect of inductor Lp across RL´. The high frequency bound, fu, is due to the voltage
division between the leakage inductance Lx and the ohmic generator and load impedances. If
the coupling coefficient is still close to one in magnitude, the two frequencies may be
estimated directly from the equivalent circuit if Lx is of no significance at frequency fl and
Lp of no significance at frequency fu, i.e.
Fig.58 Effects of parallel and leakage inductances in non-ideal transformer. The response
is the insertion loss, i.e. the deviation of v2´from the nominal output Eg/2. Note the
scale direction.
J.Vidkjær
(160)
(161)
Below fl and above fu the transfer characteristic rolls off towards 20 dB/decade asymptotes
as seen in the figure. To get a good high-frequency response, the coupling should be as tight
as possible, and to get a good low-frequency response, the inductance should be as large as
possible. It is difficult to satisfy the requirements simultaneously taking into account other
practical obstacles like ohmic losses or winding capacitances, which we have disregarded here.
The following example makes a brief account on the problems and demonstrates practical
transformer data.
Example II-6-1 ( practical RF transformers )
Fig.59 Inclusion of other transformer imperfections than leakage. Cw accounts for inter-
winding capacitances, Rws for winding series resistances, and Rcp for core losses.
Besides finite inductances and non-zero leakage contributions, a series of other

secondary effects limits transformer performances. Fig.59 shows a common equivalent circuit
holding first order approximations to an imperfect transformer. At RF-frequencies the
additional circuit components are briefly explained as follows.
J.Vidkjær
Fig.60 Extracts from Varil FP-518/FP-530 data sheets
Series resistance Rws comes from the inductor coils. At RF frequencies the skin effect
dominates the resistance, so currents are concentrated to the conductor region from the surface
down to one skin depth, [7] sec.3.16. Wound in coils the effect is worsened by the so-called
proximity effect, where the total flux causes unsymmetrical current distributions around the
conductor. The skin depth decreases with the square root of frequency, so resistance grows
with the square root of frequency. Measures against skin depth series resistances are to use
thick, sometimes silver plated wires; at lower frequencies also use of parallel isolated wires,
known as litze wires, which have a high surface to volume ratio.
The parallel resistance represent losses from the core in case a high permeability
material is used to improve coupling and raise inductances. At RF-frequencies it is eddy
currents that dominate losses. They are counteracted using fragmented materials, for instance
iron powder in an isolating binder.
A first order approximation to interwinding capacitances in the transformer places a

single, lumped capacitor across the inner inductor. The capacitance may limit the high
frequency 3dB cut-off and cause non-ideal roll-off characteristics. Interwinding capacitances
J.Vidkjær
are reduced with more spacing between the turns,7 which on the other hand could cause more
leakage, unless the flux is trapped by a high permeability core. Toroidal cores are especially
efficient in that respect.
The list of secondary effects above is not exhaustive, ref’s.[11] chap.2 and
[12] chap.2 give more comprehensive ones, but it suffices to show that patience and
experience are required to make good RF-transformers. Alternatively, a wide selection of RF-
transformers is commercially available. One example is given by the data sheet extracts in
Fig.60. There are two versions of this 6.3×6.3×3.2mm transformer, the most broadbanded has
winding ratio 1:2 (50Ω to 200Ω), the other one 1:3 (50Ω to 450Ω). Besides curves of the
insertion loss like Fig.58, impedance matching is specified by the VSWR curve ( voltage
standing wave ratio, cf.[7] sec.5.8 ). Be aware of the unusual frequency scaling in the upper
curves, which is introduced to emphasize roll-off details.
Example II-6-1 end
Tuned Transformers
Fig.61 Transformation of a secondary side load to primary side of a tightly coupled

transformer. Its equivalent circuit comes from Fig.51 using |k|=1.
In broadband applications the inductances of the transformer are supposed to be large

compared to the impedances that are connected. Besides ideal performance the transformer
should be transparent. In bandpass applications the inductances of the transformer may be a
part of the tuning circuits. Using the tight coupling idealization |k|=1 in the equivalent circuits
of Fig.51 or Fig.52, the impedance seen from at either side of the transformer now includes
the corresponding inductance in parallel to the load provided through the ideal transformer.
Fig.61 shows as an example that a secondary side load seen from the primary side is shunted
by the reactance of the primary side inductance L1. Here, the secondary side inductor mani-
fests itself solely through the winding ratio N of the ideal transformer. Seen from secondary
side the parallelling reactance corresponds to L2 as displayed by Fig.62, which shows how a
generator in Norton equivalent form transforms to the secondary side.
7) So-called transmission line transformers exploit interwinding capacitances. The

windings are made conductors in a transmission line that is wound around the core.
The method facilitates broad-band transformers of simple ratios, see [12] chap.5.
J.Vidkjær
Fig.62 Transformation of a generator from primary side to secondary side of a tightly

coupled transformer equivalenced by Fig.52 with |k|=1.
Commonly the easiest way to figure out tuning conditions with transformer couplings
is to refer all components to one side of the transformer before making design calculations.
The following amplifier example illustrates computational technique when dealing with tuned
transformers.
Example II-6-2 ( transformer-coupled tuned amplifier )
Fig.63
The task is to find components values and the voltage gain v2/Eg in the amplifier above,
which is required to operate at the following conditions:
1. Center frequency, fo = 120 MHz.

2. Bandwidth, BW3dB = 10 MHz, with frequency characteristic corresponding to
two synchronously tuned resonance circuits.
3. Generator and load impedances Rg=RL=50Ω, with center-frequency impedance
matching at the input side.
4. Transistor data: Rπ=305Ω, Cπ=19pF, gm=200mS, and Co=1.3pF. The internal
transistor feed-back is ignorable, if the following conditions are met8.
5. The total collector load at the center frequency should not exceed 88Ω to ensure
robustness of the frequency response with respect to parameter tolerances.
6. If Rg, RL, or both are absent stability is secured using Rp=220Ω or less.
8) Here we simply adopt the requirements. The theory behind is developed in the
"Stabilizing Active Two-Ports" paragraph in section III-1.
J.Vidkjær
Fig.64
Inserting a transistor model and converting the generator to Norton form provides the equiva-
lent circuit in Fig.64. Going to Fig.65, all components at the input side of the amplifiers are
collected at the primary side of the input transformer following the pattern from Fig.61.
Correspondingly, all components on the output side are referred to the secondary side of the
output transformer. The scaling of the transistor transconductance follows Fig.62 and the
scaling in control voltages from the input transformer.
Fig.65
The winding ratios of the two transformers are found directly from the basic requirements of
input matching and loading for stability,
(162)
(163)
Due to matching, the total resistance at the input port is ½Rg. At the output port we have
(164)
The bandwidth requirement pertains to the complete amplifier. To get the Q-factor for each
resonance circuit, the two stage bandwidth reduction, GBF2 in Table I, must be invoked
(165)
J.Vidkjær
Knowing the Q-factor, the remaining input circuit components in Fig.65 follow from basic
resonance circuit relationships from Eq.(1) applied to the primary side,
(166)
The primary side inductance is determined form the resonance condition and the secondary
side inductance follows from the winding ratio, i.e.
(167)
At the transistor output side we get correspondingly
(168)
Since we here refer all impedances to the secondary side, the tuning conditions give the
secondary inductor and the primary inductor is determined by the winding ratio,
(169)
Input matching implies v1=½Eg. By the transformed transconductance shown in Fig.65 it leads
to the voltage gain
(170)
The results are checked by frequency response simulations below in Fig.74.
Example II-6-2 end
J.Vidkjær
Autotransformers
Fig.66 Autotransformer in separate inductor form (a), as a reconnection of the conventional

transformer (b), and as a tapped inductor (c).
Fig.67 Signal flow-graphs for autotransformers, (d) corresponds to circuits (a) and (b) in
Fig.66. Flow-graph (e) to the tapered inductor in Fig.66(c).
Series connecting the two inductors in a transformer as shown by Fig.66(a) makes

a so-called autotransformer. To ease comparisons to previous results, the two coupled
inductors are connected to the new terminal conditions in Fig.66(b). The corresponding
embedding of the original z-parameters is indicated by the signal flow-graph in Fig.67(d).
Concentrating the graph to terminal nodes gives z-parameters of the autotransformer in terms
of the two series-connected, coupled inductors. The new parameters correspond to an inductor
with a tap as shown by Fig.66(c). This is a common way of making autotransformers.
Identifying similar paths through the flow-graphs provides
(171)
(172)
Observe that k=1 implies kA=1. Following the equivalent circuit of Fig.51(b), the effective
winding ration nA becomes
(173)
J.Vidkjær
The last tight coupling limit expression supports the fact, that the winding numbers are propor-
tional to the square roots of the inductances, and that the primary side in the present orienta-
tion has two parts separated by the secondary side tap.
Example II-6-3 ( autotransformers in a tuned amplifier )
Fig.68
The transformers in the amplifier from the previous Example II-6-2 could be autotransformers
as sketched in Fig.68, where the inductors to be tapped correspond to the largest inductors
from the conventional transformers. The placement of the tap with respect to the ground
terminal corresponds to the smallest inductor value in the original transformers. Tight
couplings with few windings due to the relatively small inductances, suggests employment of
iron-powder toroidal cores, where reasonably high inductor Q-factors are obtainable. An
example is the T25-0 core from Amidon, which has an outer diameter of 5mm, and which
give inductor Q-factors above 100 in the 100 MHz frequency range. Compared to the Q´s of
size 7.73 in the tunings, the inductor losses are insignificant. The core is specified to give an
inductance of 4.5μH with 100 windings, the so-called AL,100 value. Assuming proportionality
between inductance and the squared winding count, the number of windings for an arbitrary
inductance becomes
(174)
With this core, the inductances that were calculated in Example II-6-2 correspond to the
following number of turns,
(175)
Fig.69
J.Vidkjær
To make an autotransformer, we start as sketched in Fig.69 winding a number of turns - say

n1 - corresponding to the smallest inductor, then make a kink for the tap, and continue up to
the winding count n2 of the greatest inductance.
Note that the input tuning capacitor in the example of Fig.68 has moved to the transistor side
of the transformer compared to Fig.63. At this side it gets the value
(176)
Example II-6-3 end
Transformerlike Couplings
Fig.70 Transformation through uncoupled reactances. R |X2| and Rs |X1+X2| are assumed
in the parallel to series to parallel sequence. Note, the sequence may be reversed.
Commercial RF-transformers are available with simple winding ratios like 1: 2, 1:2,
1: 3 etc. To make transformers with non-standardized data in quantities beyond laboratory
scale may be tedious and costly. An alternative solutions to a transformation problem in
bandpass applications could be employment of uncoupled reactances. If the conditions for
simple parallel-to-series conversions following Eq.(59) are met, the conversions are applied
in a parallel-to-series-to-parallel sequence as shown in Fig.70. The resultant transformation
becomes,
(177)
If both reactances are of same signs, fulfillment of the assumption in the first part of (177)
automatically implies that the requirement in the lower part is met. This is the case with either
two capacitors or two uncoupled inductors.
J.Vidkjær
Fig.71 Uptransformation of resistor through uncoupled capacitances or inductances. The

expressions assume R 1/ωC2 and R ωL2 respectively.
The capacitive utilization of (177) is called a capacitive transformer. It is shown in

Fig.71(a). The equivalent winding ratio is expressed through sizes of the reactances,
(178)
With two uncoupled inductors, the sequence in (177) gives the result in Fig.71(b). The
equivalent winding ratio is expressed,
(179)
Note the difference to the autotransformer in Eq.(173), where it was the square roots of the
inductances that gave the winding ratio.
Example II-6-4 ( uncoupled reactance transformers in tuned amplifiers )
By this example we consider again the transformer-coupled tuned amplifier that was started
using conventional transformers in example II-6-2, changed to autotransformers in example
II-6-3. Below it is modified to make the transformations by either uncoupled inductors or
capacitors. The original requirements from page 58 shall still be met, so the quality factors
and by that the total reactances on either side of the transistor are the same as before. The
transformations by uncoupled inductances shown in Fig.72 require
(180)
J.Vidkjær
Fig.72
Similarly, the output side gives
(181)
The capacitances in the tuning are unaltered from Example II-6-2, i.e. C1=26.5 pF and
C2=75.9pF. To check assumptions, we get at the center frequency 120 MHz,
(182)
Both comparisons are above the factor of three that may be taken as a practical lower bound
where the simplifications behind Eq.(177) apply. We may go that low because it is actually
the square of the ratio that determines the approximations, cf. Eqs.(57) to (59).
In the capacitive transformer coupling of Fig.73, the input winding ratio defined by Eq.(178),
and the requirement that the total capacitance seen from the secondary side must equal C1,
provide
(183)
Fig.73
J.Vidkjær
At the output side we get correspondingly,
(184)
The tuning inductors are here equal to the transformer inductors towards the transistor, i.e.
(185)
Before using the results, we should again check the assumptions. Now we get
(186)
so the method is still usable although the input condition is not as good as it was in Eq.(182).
The reason for the difference is that the whole inductive reactance is engaged in the inductive
transformations, but only a part of the capacitive reactance is available for transformation
because a fraction belongs to the transistor.
The agreements between the transformation methods are illuminated by Fig.74. It compares
the simulated responses of the transformer-coupled amplifier from II-6-2, heavy lines, with
the responses of the two configurations above. As seen, the simplifying assumptions manifest
themselves mostly by a slight displacement in phases with the inductive coupling lagging
behind and the capacitive leading the transformer-coupled amplifier.
Fig.74
Example II-6-4 end
J.Vidkjær
If the requirements for simple series-to-parallel conversions do not apply, it is

possible to calculate components in the transformations without any assumptions from more
elaborates relations in Eqs.(57) and (58). Under such circumstances, where also the reactive
components are influenced, there is no benefit of referring to transformer concepts as the
problem behaves more likes that of establishing impedance transformation through a lossless
network. Smith chart constructions, which are considered in section II-8, are here alternatives
to calculations by the two formulas mentioned above.
Three-Winding Transformers
Transformers with three windings are often encountered in RF-circuits to change

signals back and forth between unbalanced and balanced forms, or to isolate parts of the
circuits from each other. Setting up general purpose equivalent representations based on
uncoupled inductors and ideal transformers in the three-winding case is both tedious and
confusing, so we shall limit ourself to consider idealized situations that closely follow the
intentions behind the use of the transformers. Before doing this, however, an equivalent
representation suitable for computer simulations of less ideal cases is briefly discussed.
(187)
The inductance matrix of three coupled inductors is given by Eq.(187). The corre-
sponding three port circuit may be equivalenced by the network in Fig.75, which is the three
port counterpart to Fig.56. The last version of the inductance matrix includes the assumption
of reciprocity using only three different coupling coefficients because Mij=Mji. To ensure
passivity the matrix must be positive semidefinite, and the diagonal elements satisfy
(188)
Zero-valued inductors are uninteresting and are disregarded below. Taking two-dimensional
subdeterminants symmetrically around the diagonal shows that no coupling coefficient may
exceed one in size, i.e.
(189)
The final three-dimensional determinant requirement relates the three coupling coefficients
(190)
J.Vidkjær
Fig.75 Equivalent circuit for three inductors where the couplings are simulated by voltage
controlled voltage sources. The signal flow-graph shows z-parameters.
To illuminate the last requirement, suppose that inductor number one and two are tightly
coupled, k12=±1. Then the size of the coupling from inductor number three must be the same
to both number one and two. The condition develops
(191)
Taking negative value of a squared number cannot exceed zero, so only the equality condition
is in effect making the two remaining couplings equal in size. Signs are adjusted so there are
either none or two negative coupling coefficients. The result is the limit case of the require-
ment from Eq.(190), which in a less stringent form is phrased that if two inductors are
strongly coupled, a third one cannot be loosely coupled to one of them and strongly to the
other.
Sign conventions with three windings are more confusing than the two winding case,
because it is now possible both to exchange terminal orientations, equivalently winding senses,
and to alter the magnetic flux paths from a serial to a parallel one. In transformer idealizations
without leakage the two last alternatives are sketched in Fig.76, where the flux orientations
correspond to currents entering the positive terminals. With the toroidal core, where the flux
path follows the inductors in series, it is possible to organize port orientations to let all fluxes
support each other, so here dot tagging remains meaningful. Without leakage all coupling
coefficients are simultaneously of unit size with either none or two being negative as we saw
in Eq.(191). Following the dots all couplings are positive, so turning the direction of one coil
lets its flux opposes the fluxes from the other two, thus changing signs of two coupling
coefficients.
J.Vidkjær
Fig.76 Three-winding flux-paths. Couplings are positive between coils with supporting
fluxes, negative on opposing fluxes. In series paths (a) there may be none or two,
in parallel (b) one or three negative couplings.
In parallel flux paths, like Fig.76(b), there will always be at least one path opposing
the others, so either one or three coupling coefficients are negative. Here, a single dot
indication is meaningless because it is no longer possible to establish an unambiguous refer-
ence direction for fluxes, so it seems better to resort to signed mutual inductance or coupling
coefficient indications in that case. Moreover, the flux from either port divides between the
two others so none of the coupling coefficients may approach one in size if all inductors
should be coupled together. The requirement was foreseeable from the passivity condition in
Eq.(190). If all the k’s are negative, all k-terms in the last factor are negative and so the sizes
of the couplings are severely restricted. Suppose the fluxes from either coil in Fig.76(b)
distributes evenly between the two others. Then we have k12=½, k13=½, and k23=-½, which
is the borderline for the passivity criterion.
RF-applications of three winding transformers are dominated by the series flux

method in Fig.76(a), so the reason for considering the parallel case is mainly to show how
easily passivity criteria are violated in a simulation, for instance by setting all couplings close
to one in size but let one of them be negative. Transformers that are not passive may supply
power and cause unrealistic results or introduce instability in computations.
The impedance matrix for three coupled inductors, which is based on the inductance
matrix in Eq.(187), leads to the requirement,
(192)
A winding scheme like Fig.76(a) is assumed, so all coupling coefficients are taken positive
and port orientations follow the dot marking. Tightly coupled, Eq.(192) implies
(193)
J.Vidkjær
Fig.77 Ideal three-winding transformer and a flow-graph of its terminal constraints.
Taking the first inductor as reference, the three-winding transformer voltage relations become
(194)
N12 and N13 hold winding ratios of inductor no.2 over no.1 and inductor no.3 over no.1
respectively. The ideal voltage conditions require tight coupling only. If the inductances are
raised towards infinity keeping unaltered winding ratios, Eq.(193) gives the current relation
for an ideal three-winding transformer,
(195)
Fig.77 shows the transformer symbol and a flow-graph representation of the ideal, algebraic
current and voltage constraints.
The three equations in Eq.(192) may be written
(196)
These equations have form of the ideal transformer current relation using the current sets
(197)
Each hatted component is adjusted for the current through an uncoupled inductor. To account
for finite inductances in the three-winding case, we may use the same approach that formerly
applied to two windings, namely to include one of the inductors across the corresponding port
of an otherwise ideal transformer. The three possibilities are shown in Fig.78.
J.Vidkjær
Fig.78 The three possibilities of taking into account finite inductances in a three-winding
transformer with tight couplings.
The input impedance of a loaded three-winding transformer is calculated as shown

by Fig.79. There are no loops in the flow-graph, which directly gives
(198)
If the reactances of the transformer inductances compare to the load impedances, the effect
of finite transformer inductances may be accounted for by connecting inductor L1 across the
input-port as indicated in part (b) of the figure.
Fig.79 Determination of the input impedance of a loaded, three-winding transformer. The

flow-graph assumes ideality. L1 may be added to account for finite inductances.
To get the generator equivalent circuit at a transformer port, when the generator is
connected to another port and the third port is loaded, we consider Fig.80. The loop provides
(199)
By that, the EMF and generator impedance at port number two becomes
(200)
J.Vidkjær
Fig.80 Signal flow-graph for determining a generator equivalent circuit at port 2.
(201)
This result is interpreted as shown by Fig.81, where it is seen that the third port loads the
input generator before it is transformed to port 2. Again, one inductor may be included to
account for finite transformer inductances.
Fig.81 Interpretation of Eqs.(200),(201) where (a) holds the generator components that are
transformed to (b). L2 may account for finite transformer inductances.
Transformer Hybrids
A coupling method based on the three winding transformer that has great practical
importance is the so-called hybrid in Fig.82. It is used to combine or isolate signals from
different sources. To achieve such goals, winding ratios and load resistances must be realized
accurately. Failures to meet requirements set the performance limits, so we shall start analyz-
ing the circuit without any assumptions, and then simplify to see how the desired properties
evolve. As indicated in the diagram, there is an impedance and an EMF connected to each
port. We consider the corresponding port voltages - the vpj’s - as the responses when the
EMF’s are successively applied. With the transformer conditions from Eqs.(194),(195), the
connections in Fig.82 imply
(202)
J.Vidkjær
Fig.82 Three-winding transformer coupled as a hybrid. On proper conditions there are no

transmission from E1 to vp4 , from E2 to vp3 , and vice versa.
The port currents and voltages are constrained by the branch relations
(203)
The two last sets of relationships are drawn together in the signal flow-graph of Fig.83. There
are six loops and two sets of non-touching loop products in the graph yielding the common
transfer function denominator
(204)
The two sets of ports between which it is possible to get isolation are 1,4 and 2,3 respectively.
For the first set the flow-graph provides,
Fig.83 Signal flow-graph for the hybrid in Fig.82. The insets show six loops, where the
upper ones contribute loop-loop and loop-transmission terms in gain computations.
J.Vidkjær
(205)
A simple criterion where these transmissions become zero is symmetry,
(206)
Transmissions between the second pair of potentially isolated ports give
(207)
The joint conditions for isolation between the two port pairs are called double balance, where
(208)
Under double balance conditions the denominator reduces to
(209)
Conditions at port 1 and 4 are now expressed by
(210)
(211)
Fig.84 Conjugated matching.
J.Vidkjær
Fig.84 and Eq.(211) summarize conjugated matching condition at the j’th port including the
port voltage, vpj, the power delivered to the port Ppj, and the available power at the port, Pav,j.
Applied to the last expression in Eq.(209), conjugated matching or shortly just matching is
obtained if
(212)
In case impedances are ohmic, the voltage ratios are simple and equal to one half,
(213)
Obtaining matching conditions at the two other ports is a little lengthy, because there are four
transmission paths in the flow-graph, two including non-touching loop products. We find
(214)
Imposing the conditions from Fig.84 on the second but last expression above shows that
conjugated matching at port 2 and 3 resorts to the requirement from port 1 and 4 in Eq.(212).
Again, ohmic conditions simplify to halving the port voltage compared to the generator EMF.
The transfer functions that now remain are all the desired signal-branchings. From
port 1 we get
(215)
Note the sign-shift in the transmission to port 3, which was required to bring the two transmis-
sions on final common forms. To admire the matching result we show that the power Pp2
J.Vidkjær
delivered to port 2 is half the available power from the source (the other half goes to port 3).
Incorporating the balancing and matching conditions from Eqs.(212),(213) we get
(216)
Transmission from port 4 becomes
(217)
Observe here, that none of the paths required sign-shift to reach the common forms. The
power branching when transmitting from this side of the transformer is expressed
(218)
To economize writings we take the remaining transmission in output port order, i.e. to the
same sides of the transformer. Then we get
(219)
where the lower ratio has changed sign like its opposite direction counterpart in Eq.(215). Due
to the similarities in the expressions, power transmission from port 2 and 3 to port 1 are the
same, for instance
(220)
J.Vidkjær
Finally, the transmissions to port 4 are
(221)
where the power balance with a source on port 2 or equivalently on port 3 is,
(222)
A summary of the possible transmissions between port voltages is given in Fig.85.

It assumes ohmic matching conditions, where the voltage across any driving port is half the
EMF. Due to the sign-shifts between port 1 and port 3, the coupling is referred to as a 180°
hybrid and it has the property of being broadbanded. With tuned transformers or in transmis-
sion line techniques it is also possible to make 90° hybrids, cf.[4] sec.6.5.
Fig.85 Port-voltage relations summary in ohmic matched trans-

former hybrid, classified functionally as a 180° hybrid.
One reason for working out details of the three-winding transformer hybrid is its
widespread use in mixer circuits. Mixers are often the most critical component in communica-
tion systems, and derivations above are helpful tools for exploring their performance limits.
As a brief introduction, the following example shows a diode-ring mixer that includes two
three-winding transformers, one operating as a hybrid.
J.Vidkjær
Example II-6-5 ( diode-ring mixer preliminaries )
Fig.86 Diode-ring mixer. A large local oscillator signal switches the two diode branches
instantly and reverts sign of transmission from input to load on positive (b) and
negative (c) VLO’s.
The basic function of the diode-ring mixed in Fig.86 is most easily explained if the
mixer is employed to generated a double sideband AM signal with suppressed carrier, the so-
called DSB-SC modulation. For simplicity it is supposed that both transformers have winding
ratios equal to one and we let a low-frequency sinusoidal represent the baseband signal, VBB.
It is applied to the isolated winding port of the input three-winding transformer, T1. The
carrier signal is inserted by the local oscillator signal VLO at the isolated port in transformer
T2. Proper operation of the mixer requires that this signal be large compared to both VBB -
Fig.87 is not drawn to scales - and the voltage across two conducting diodes. The role of the
local oscillator is merely to switch between the two branches in the diode-ring as showed by
inserts (b) and (c) in Fig.86. The switching changes direction of transmission from the
baseband input to the modulated RF output VRF across the load. This process corresponds to
multiplying VBB by function fsw that switches evenly between 1 and -1. It has no DC term
and a fundamental frequency equal to the carrier frequency of VLO. Proper low-pass or
bandpass filtering of the VRF signal in the figure gives therefore a DSB-SC waveform.
No sophisticated balancing and transmission conditions are required to explain this

basic function. In practical operation, however, a series of secondary phenomena related to
unavoidable nonlinearities produces spurious components. They are unwanted signals at
frequencies that might disturb further processing or other communications. One measure to
prevent their build-up is to keep signal sources isolated from each other. In the present case,
it is required that the two secondaries in T1 have exactly equal winding ratios and that the
J.Vidkjær
diode impedances in the two switching directions are the same, to prevent a baseband signal
component in the output. Both in Fig.86(b) and in (c) the output load resistor and one of T1´s
secondary sides make the isolated port of T2 when seen from the local oscillator. Therefore,
the accuracy of the double balance determines how good the carrier is suppressed in the output
signal. Furthermore, the balance prevents baseband signals directly to enter the oscillator
circuitry. It is for the quantitative analysis of this type of problem, we need details of the
three-winding transformer and its use as a hybrid.
Fig.87 Diode-ring mixer waveform example. Baseband signal VBB is chopped at the local
oscillator frequency. Proper filtering of the resultant VRF signal gives a DSB-SC
output.
Example II-6-5 end
J.Vidkjær
80
II-7 Double-Tuned Circuits and Amplifiers
Two resonance circuits coupled so loosely that they nearly maintain their individual
resonance properties constitute a double-tuned circuit. It can be build with both parallel and
series resonance circuits, but only coupled parallel circuits are considered. This is the type that
is most commonly encountered in RF circuit designs, but the concepts and results below
translate to double-tuned series circuits like the translation between single tuned parallel and
series circuits we have seen before.
Parallel resonance circuits are in principle coupled by a connecting element as

illuminated by Fig.88(a) and (b), which show the important cases of purely capacitive and
inductive couplings. To make the coupling loose while keeping transmission between the
ports, the admittances of Cc or Lc must be comparable in magnitude to the admittances of the
parallel resonance circuits. We shall be more precise about this matter below. Comparing the
inductive case with Fig.53 in section II-6 shows that inductors La, Lb, and Lc could be the
equivalent circuit of two coils with mutual inductances.
Fig.88 Double-tuned circuits based on two capacitively (a) and inductively (b) coupled
parallel resonance circuits Ca, Ra, La and Cb, Rb, Lb.
Fig.89 Circuit structure and signal flow-graph for calculating transfer and input impedances
of the double-tuned circuit.
The structure of the circuit to be considered is shown in Fig.89. The transfer and
input impedances are derived directly from the diagram or guided by the flow-graph to yield
(223)
J.Vidkjær
II-7 Double-Tuned Circuits and Amplifiers 81
(224)
To cast the expressions in symmetric forms we define a sum and a difference admittance,
(225)
Now the last parts of Eqs.(223) and (224) can be rewritten,
(226)
(227)
Coupling between Identical Resonance Circuits
The advantage of Eqs.(226) and (227) are, that they expose symmetry in the resultant
impedances from the beginning even if the two coupled circuits are different. But we start
considering the complete symmetrical situation, where the two resonance circuits are identical
and we have
(228)
The sum admittance depends on the type of coupling and takes one of the forms,
(229)
Both expressions represent parallel resonance circuits, which include the coupling element,
either Cc or Lc. The resonance circuit described by YΣ appears across either port in the two-
ports of Fig.88 when the opposite port is short-circuited. To express both coupling methods
equivalently, we denote
(230)
J.Vidkjær
By that the resonance frequency and quality factor for YΣ are expressed by the usual relations
(231)
To obtain a unified description of both coupling methods, the coupling coefficient k is

introduced by the definitions
(232)
Observe that the definition with respect to inductive coupling agrees with the k definition for
two inductors having mutual inductances, cf. Eq.(158) and the equivalent circuit in Fig.55. The
Q-value of the resonance circuit in YΣ is supposed to be high, so the narrowband approxi-
mation from Eq.(46) applies. We may therefore approximate
(233)
in vicinity of the upper half-plane pole s1, particularly on the imaginary axis around ω0. In
the present symmetrical case Eqs.(226) and (227) reduce to
(234)
(235)
Before inserting terms into the equation for z21, narrowband considerations are applied once
more. The resonance circuit admittance YΣ(s) exhibits its strong frequency dependency around
s1 because it contains two counteracting reactive components, inductor La and capacitor Ca.
Contrarily, Yc(s) is designated by a single component and shows no significant variations
around s1. In a narrowband approximation it is therefore considered constant,
(236)
J.Vidkjær
These expressions are used jointly with Eq.(233) in Eqs.(234),(235) to give the narrowband
approximation to the transfer and input impedances of the symmetrical double-tuned circuit,
(237)
(238)
The sign in the numerator of z21 is determined by the method of coupling, positive if it is
capacitive and negative if it is inductive. Note, however, that k may be either positive or
negative if two inductively coupled coils are employed. Coupling two identical resonance
Fig.90 Double-tuning. Encircling around s1 in (a) is expanded to show narrowband poles

and zeros in z21 (b) or z11 (c). See page 99 about the number of zeros in origo.
circuits together has the effect of splitting pole positions symmetrically apart from s1 in direc-
tions parallel to the imaginary axis as indicated by Fig.90. The coupling does not affect the
zero of z11 that remains at s1. Observe that to stay consistent with the narrowband assump-
tions, the poles are still confined to the region around s1. Besides the basic high Q require-
ment, k must be small. In summary upper half-plane poles are
(239)
Inserting the poles and s=jω into Eq.(237), using Eq.(231), gives the frequency response by
the narrowband approximated transfer impedance
J.Vidkjær
Fig.91 Double-tuned circuit normalized magnitudes and phases (inductive coupling) with
various kQ settings. Part (a) shows transfer impedances z21 and (b) input imped-
ances z11.
(240)
It is the same frequency normalization that formerly was introduced for a single tuned circuit
by Eq.(45), which is applied here. Magnitude and phase curves of z21 using different kQ
products are shown in Fig.91(a). When kQ is below one, the amplitude characteristic shapes
likes single tuned responses. If the kQ products are greater than one, the two poles are so
widely separated that they are sensed individually to produce two maxima in the amplitude
curves. In between is the so-called critical coupling, where kQ equals one. It has the widest
J.Vidkjær
bandwidth that can be achieved without ripples in the amplitude characteristics. The coupling
admittance, which by Eqs.(231) and (236) is expressed
(241)
becomes equal in size to the admittances of either parallel circuit at resonance, if the coupling
is critical. The sign still follows the coupling method being positive with capacitive coupling.
A narrowband approximation for the input impedance z11 is obtained by a procedure

similar to the one above based on Eqs.(231),(233), and (238). It gives
(242)
Curves of the impedance are shown in Fig.91(b). Due to an intervening zero compared to the
transfer impedance, cf.Fig.90(c), the impedance magnitude gets a minimum at the center
frequency with critical coupling, kQ=1. Flatness in z11 requires that the two poles moves close
to compensate the effect of the zero. As seen, this happens if kQ=0.5.
Example II-7-1 ( Foster-Seely FM detector )
In frequency modulation, FM, signals of constant amplitude are transmitted with an instanta-
neous frequency that differs Δω from a fixed carrier frequency ωc in proportion to an informa-
tion bearing low-frequency signal. Recovering information in demodulation requires a circuit
whose output is proportional to the frequency deviation between the instantaneous frequency
Fig.92 Foster-Seely FM detector. Besides the overcoupled signal v2, primary voltage v1
adds as a common signal to the secondary side by coupling capacitor Cm.
J.Vidkjær
and the carrier frequency. A classic construction having this property is the Foster-Seely
detector in Fig.92. Here, current source ig introduces the FM signal of instantaneous angular
frequency ωinst=ωc+Δω. The detector outputs a LF voltage proportional to Δω.
We shall consider how a double-tuned circuit is used to achieve the goal above with good
linearity. Source ig drives the primary side of two inductively coupled parallel resonance
circuits, each holding elements Ra, C, L. The resonance frequencies are equal to the carrier
frequency ωc. Transimpedance z21 causes the secondary voltage v2, which is shown by two
halfs on either side of a center tap in the secondary inductor. Left alone this point would have
zero potential as it is connected to ground through the choke Lm, which is a high impedance
inductor at the signal frequency. However, the primary voltage v1 is added through the
coupling capacitor Cm as a common voltage on the secondary side. Due to the symmetry of
grounding, the voltages presented at the envelope detector inputs, vA and vB respectively, get
geometrical interpretations as shown by Fig.93. The detectors are supposed to provide no
loading compared to Ra.
Fig.93 Phasors for secondary side voltages in the Foster-Seely detector. The instantaneous
frequency, ωinst=ωc+Δω, equal to, above, and below carrier frequency ωc.
To realize the three phase conditions in the figure, we observe that the FM signal frequencies
are related to our previous frequency variables by
(243)
The phase difference θ between v2 and v1 follows from Eqs.(240),(242), and becomes
(244)
where the last expression is based on a first order expansion of tan-1. We assume k>0. An
angle difference of -Ωs offsetted by -½π agrees with the situations in Fig.93, since there is
proportionality between Δω and Ωs from Eq.(243). At Ωs=0, where v1 and v2 are perpendicu-
lar to each other, the voltages are said to be in quadrature .
The two voltages vA and vB are AC voltages of instantaneous frequency ωinst. Lengths of their
phasors, |vA| and |vB|, are envelopes to the AC signals, and these voltages are the assumed
J.Vidkjær
outputs from the ideal envelope detectors. The difference in envelope values makes up the
detector output, which is expected to be proportional to Δω or, equivalently, Ωs. To investigate
the question, we calculate again from Eqs.(240),(242)
(245)
(246)
(247)
Curves illustrating the last equation are plotted in Fig.94 for various values of the kQ product.
As seen, kQ-products about three give the largest, linear output interval.
Fig.94 Foster-Seely FM detector characteristics for different kQ’s. The curve in heavy line,
kQ=3, has rather good linear output span.
Before leaving the Foster-Seely detector, we shall make a few comments on its implementa-
tion and use. First, practical envelope detectors are commonly based on diode rectification,
so there is voltage loss in the detection. The detector circuit contributes nonlinearily to both
the primary and secondary side load resistors, so a thorough analysis of this case becomes
rather involved. Second, the detector is sensitive to input level, in addition to the frequency
deviations. This is expressed here by the ig factor in Eq.(247). Therefore, a separate so-called
limiter circuit, which ensures constancy of the ig amplitude, must precede a Foster-Seely
detector.
Example II-7-1 end
J.Vidkjær
Example II-7-2 ( quadrature FM detector )
Fig.95 Quadrature detector principles. Quadrature of v2 is made by voltage division

between Ck and the impedance of a single-tuned (a), or double-tuned circuit (b).
Alternatives to the Foster-Seely detector are the quadrature detectors in Fig.95. Here the FM
modulated input signal v1 and the quadrature component v2 are multiplied and filtered to
recover the baseband signal, which is proportional to Δω. This last part of the process gives
(248)
where b is the multiplicator scaling factor. This result is based on the trigonometrical identity
(249)
and the assumption, that the LP filter totally removes the high-frequency component at twice
the carrier frequency. To see the significance of quadrature, observe that it is exact if θ=±½π,
and around this angle we approximate
(250)
To detect FM, the divider circuit must provide a deviation angle φ from quadrature, which to
a first order approximation must be proportional to the deviation Δω from the carrier frequen-
cy ωc. More generally, the complete processing of the divider in combination with the
nonlinear multiplication should possess linearity between frequency deviation and detector
output voltage. The two detectors in Fig.95 have different spans of linearity where the more
complicated solution with double-tuning is the better one. As a reference, however, the simpler
single-tuned circuit in Fig.95(a) is considered first.
J.Vidkjær
Using narrowband techniques including the impedance Zp of the single tuned circuit,
cf.Eq.(45), the quadrature voltage may be written
(251)
The resonance circuit is tuned to the carrier frequency of the FM signal, so ω0=ωc. To assure
quadrature at Ωs=0, the last term in the denominator of Fs must be insignificant, i.e.
(252)
In its first form, the inequality shows that the impedance of the divider capacitor Ck must be
large compared to the peak-impedance of the resonance circuit. The second form quantifies
the requirement in terms of the tuned circuit Q-factor. Including the assumptions we express
(253)
Factors at the right-hand side of Vout from Eq.(248) are now related through,
(254)
Expressed in either normalized or absolute frequency variables, the detector output becomes
(255)
It is clearly the frequency dependent term in the denominator that limits linearity in the output
voltage versus frequency deviation detector characteristics. To improve the linear range , the
input impedance of a double-tuned circuit may be employed instead of Zp. Then we get the
division ratio, cf. Eq.(242),
(256)
J.Vidkjær
The requirement of Eq.(252) must still apply, so the contribution from the quadrature compo-
nent is calculated in analogy with the previous development, i.e.
(257)
(258)
where the deviation from linearity is kept in the function U,
(259)
The detector output now becomes
(260)
and corresponding detector characteristics are shown in Fig.96. To make easy comparisons the
outputs are scaled so all slopes in the center at Ωs=0 are equal. The scaling is based on the
U-function with arguments set to zero,
(261)
Fig.96 Normalized quadrature detector characteristics. The curves show that kQ=.45 is
near-optimal with respect to linearity in double-tuning. The kQ=0 curve corresponds
to single-tuning.
J.Vidkjær
Note, that kQ=0 gives the function that controls departures from linearity in the single tuned
case from Eq.(255), so the figure includes both type of tunings. In double tuning, the curves
confirm that a value kQ=0.45, which is often aimed upon in design, comes close to an
optimum regarding linearity. The important consideration is here the span of normalized
frequency where linearity is maintained, so double-tuning is approximately twice as good as
single-tuning in that respect.
Quadrature detectors, compared to the Foster-Seely detector from the foregoing example, have
the advantage that multiplication is better suited for integration than envelope detection. An
example of a quadrature detector IC and its use is given by the data-sheet extracts9 in Fig.97,
Fig.98, and Fig.100. The block diagram shows that the chip contains an "input amplifier", two
coupling capacitors to the external tuned circuit, a demodulator that holds the multiplication
function, and finally a video amplifier that includes lowpass filtering to a bandwidth of 10
MHz. The input amplifier terms are set in quotes above because it is not a conventional linear
amplifier but a limiting one, which keeps its output amplitude constant across an input range
from 10 to 300 mV rms. It is clear from the quadratic V1 amplitude dependency in Eq.(260),
Fig.97
9) GEC Plessey Semiconductors, Consumer IC Handbook, Sept.1991, pp.2-9,12
J.Vidkjær
Fig.98
that the input to the detector must stay very constant, claimed as excellent threshold among
the highlighted features in the data. The RC network around the input amplifier shows, that
the inputs are biased internally.
Like other integrated circuits, the RF signal path is differential, i.e. the signal voltage is the
difference between two identical processing chains. Difficulties of establishing good RF signal
ground inside the IC are avoided this way, which benefits from the fact, that many basic IC
constructions are differential anyway. A multiplicator that gives differential output as the
product of two differential inputs - denoted demodulation in the block diagram - is a standard
construction called a Gilbert cell [13],[14], or a double-balanced modulator (mixer)
[15]. Thinking differential instead of single-sided is therefore easy in the present case,
where the interface to the external tuning circuits is sketched in Fig.99. To get a v2 voltage
here that equals v2 in Fig.95(b), for same v1´s, the voltage across each of two differential
mode dividing capacitors, the Cd’s, must be half the difference voltage v1-v2. If all other
Fig.99 Details in differential mode connection of the double-tuned circuit to the block
diagram in Fig.98.
J.Vidkjær
impedance levels are unaltered from the single-sided case, the differential mode capacitors
should be twice the Ck capacitance to halve the corresponding impedances. We could repeat
this arguing with respect to coupling capacitance Cc, so the balanced coupling in the figure
requires two capacitors, each twice the size of the corresponding single-sided coupling
capacitor. Strictly speaking it is unnecessary to balance the coupling capacitors in this
application although it is shown in the data-sheets, where also the low number of external
components required for this IC is praised. So here is a suggestion for further savings.
Fig.100
J.Vidkjær
The use of the IC is illustrated by single and double tuned examples in Fig.100, which also
gives examples of the detector characteristics for various Q values and, with double-tuning,
the corresponding coupling capacitors. The strategy has been to keep the tuning components
L and C fixed and let the resistor control the Q by the usual relationship, cf. Eq.(252). In
agreement with Eqs.(255),(260) the resultant outputs get slopes proportional to Q2 - R is
proportional to Q - as seen in both types of detector characteristics. Comparing single and
double tuned curves, it is also observed that the latter get the best span of linearity with a
given center slope.
Example II-7-2 end
Double-Tuned Amplifier Stages
Fig.101 Bandpass amplifier with symmetrical, critically coupled, double-tuned output circuit.
The transfer function is shown in Fig.102(b).
The flat amplitude response in critical coupling is a common reason for using double-
tuned circuits in amplifiers. If the transistor is modeled by a transconductance, letting possible
output impedance components contribute to the first resonance circuit, the voltage gain of the
amplifier in Fig.101 can be expressed,
(262)
where the denominator D is taken from z21 in Eq.(240). Inserting kQ=1 for critical coupling,
the denominator and its squared absolute value become
(263)
To get upper and lower 3dB bandlimits we solve for the Ωs´s where the squared denominator
has doubled compared to its center frequency value at Ωs=0, i.e.
(264)
J.Vidkjær
Using Eq.(231), the corresponding unnormalized 3dB bandwidth is expressed
(265)
Considered separately, the two parallel resonance circuits have bandwidths corre-
sponding to Ωs=±1. Broadening in bandwidth by a factor of 2 in z21 is readily seen in the
comparison by Fig.102(b). The corresponding flatness of the amplitude response is maximal
as the frequency characteristic with critical coupling is the Butterworth characteristic, which
was discussed in section II-5, page 36. One way to realize this is by the pole positions in
Fig.102(a). They equal the second order Butterworth pole patterns in Fig.39. Alternatively, the
defining Butterworth magnitude equation, Eq.(109), follows directly from Eqs.(262),
(263), by introduction of a new frequency normalization Ωbtw, which has 3dB bandlimits at
Ωbtw=±1. We get
(266)
Taking the total capacitance CΣ in the first resonance circuit as reference, the gain-
bandwidth factor of the amplifier with symmetrical double-tuned output circuit becomes,
(267)
so this stage has a gain-bandwidth product below that of a single tuned stage.
To get flat response and simultaneously keep gain-bandwidth like a single tuned
stage, symmetry must be given up so only the output circuit is resistively loaded .
Fig.102 Critical coupling. Poles of z21,(a), and amplitude response,(b). Scaling is chosen to
compare the shape with a single tuned response of equal Q, i.e. with one pole in s1.
J.Vidkjær
Fig.103 Bandpass amplifier with one-sided load of the double-tuned output circuit .
This is exemplified by the amplifier in Fig.103, where the secondary resonance circuit is
resistively loaded. Both resonance circuits have identical reactive elements, and are therefore
tuned to the same frequency. The quality factors are different being infinitely large at the
transistor side and finite at the output side. To get the transfer impedance we use the technique
developed by Eqs.(223) through (226). The sum admittance becomes
(268)
The reactive elements are defined to encompass both capacitive and inductive couplings by
the definitions in Eq.(230). It is repeated here for convenience,
(269)
Following Eq.(268), the sum admittance YΣ has the resonance frequency and quality factor
(270)
where Qb is the quality factor of the output side resonance circuit alone. The narrowband
approximation for YΣ is
(271)
With an unsymmetrical circuit we get a non-zero difference admittance
(272)
The coupling admittance is narrowband approximated as before in Eq.(236), where positive

sign refers to capacitive coupling, negative to inductive coupling,
(273)
J.Vidkjær
All elements in the transfer impedance are now collected to yield
(274)
The last expression is a narrowband approximation around s1 in the upper half-plane. Casted
into standard form the expression should read
(275)
so the poles s11 and s12 are easily identified. Using Eq.(271), they become
(276)
Suppose the coupling k is a parameter that can be independently adjusted. Starting from k=0,
where the two resonators are isolated from each other, the two upper half-plane poles have
imaginary part jω0 and real parts 0 and -ω0/2Qb respectively as sketched in Fig.104(a). With
Fig.104 Poles of z21 in the unsymmetrical double-tuned circuit from Fig.103. (a) Poles at
k=0 and directions of movements for increasing k. (b) Poles at kQb=1/ 2.
growing k the poles attract horizontally until they met at s1 when the square root becomes
zero for kQb=½. Further rise in k turns the square root imaginary and the poles split apart in
parallel to the imaginary axis. A pole pattern like Fig.104(b) gives maximally flat transfer
characteristics. It is reached when the square root is j and the kQb-product becomes
J.Vidkjær
(277)
Inserting the radius r from the pole plot in Fig.104(b) into the denominator of Eq.(275) gives
the center frequency gain and subsequently the gain-bandwidth product,
(278)
(279)
Hence, a double-tuned amplifier with one undamped resonator rise the gain-bandwidth product
to the level of the single tuned amplifier and keeps the maximally flat transfer characteristic.
The contrast to the symmetrical case may be understood taking the output side resistors as a
load receiving power from the transistor. The intervening parts of the unsymmetrical double-
tuned circuit are lossless, while the input side resistor causes an additional loss in the symmet-
rical circuit.
Asymptotic Behavior
With exception of signs, neither qualitative nor quantitative distinction between

capacitive and inductive couplings was required in the discussions above. They were all based
on narrowband assumptions around the passband of the double-tuned circuits. Outside the
passband, where the narrowband approximations are no longer adequate, dissimilarities show
up as differences in the response asymptotes with frequency approaching either zero or
infinity. We shall make a brief account on these questions.
Fig.105 Dominant contributions to asymptotic responses when the frequency is approaching

zero, (a),(b) or infinity, (c), (d).
J.Vidkjær
When the frequency is approaching zero, Fig.105(a) and (b) indicate the circuit
elements of the double-tuned circuit that are significant. The limit values are found using
Eqs.(223),(224) to yield
(280)
(281)
Both input impedances z11 go towards zero like ω as does z21 with inductive coupling. In
capacitive coupling z21 follows ω3. The rate of decline, taken as the power in frequency that
the response follows towards zero frequency, is equal to the number of zeros in origo.
In the high frequency limit the double-tuned circuit responds like the highlighted parts
of Fig.105(c) and (d), where we get
(282)
(283)
Both input impedances z11 are approaching zero inversely proportional to frequency as is the
transimpedance z21 if capacitive coupling is employed. The inductive coupling shows a z21
decline proportional to the third order frequency term in the denominator. It is common
terminology to say that the impedance or transimpedance has a number of zeros at infinity that
corresponds to the resultant power, by which the denominator goes towards infinity with
frequency. Hence, capacitively coupled circuits have one and inductively three zeros at
infinity.
J.Vidkjær
Fig.106 Low and high frequency asymptotes of |z21| in double-tuned, critically coupled
circuits using capacitive (a) or inductive (b) couplings.
The asymptotic behaviors of the input impedances follow the pattern we have already
seen with single-tuned circuits in Fig.5. The results should not be surprising as they illuminate
the fundamental lumped network property, that the driving point impedance of a passive
network cannot approach zero faster than a single shunting inductor in the zero frequency limit
and faster than a single shunting capacitor in the infinite frequency limit10. Trans-
impedances, which have no similar restrictions, are summarized by Fig.106. The central parts
in the figures are the regions encompassed by the narrowband approximation, where no
distinctions between the two types of couplings were necessary. It is clear from the figure, that
signal suppression requirements in the low or high-frequency range of a given application
favors either a capacitive or an inductive solution. In practice, care must be exercised when
realizing the selected coupling type in limits where the connecting impedance Zc is supposed
to be large. At high-frequencies with inductive coupling, stray capacitances between the
primary and secondary circuits will destroy the steep roll-off asymptote. A small mutual
inductance between the coils would similarly change the low-frequency roll-off if capacitive
coupling is employed. Placing coils perpendicular to each other is a first measure against
inductive coupling.
10 ) See most books on circuit analysis and synthesis, for instance chap.5 in ref.[1].
J.Vidkjær
101
II-8 Impedance Matching
Transformer and transformer-like couplings of the types considered before are not
always the best suited solutions to impedance matching problems. Once build, traditional
transformers have few provisions for adjustments. If the impedance to be matched has wide
tolerances, the matching circuit should include trimming and tuning components, often two
in number to compensate spreadings in both the real and the imaginary impedance compo-
nents. The technique to be considered in this section has the advantage of being a direct,
graphical approach to matching problems. It is not restricted to certain classes of circuit
topologies, even not to certain types of circuit components, but we concentrate on lumped
element networks here leaving mixed and pure transmission line methods to discussions in
application and microwave literature, for instance ref’s [16], [12] or [4]. The new ap-
proach has the drawback of being a single frequency method, which only admits a crude
bandwidth estimation. However, if bandwidth control is in question we may get good starting
points for adjusting to specifications by subsequent design optimization processes.
Lumped Element Impedance Matching using Smith Charts
Smith charts are commonly associated with problems of calculating impedances and
designing matchings in transmission line networks, cf.[7] sec.5.9 or [4] chap.5. Staying solely
with lumped elements - as we do in this section - the Smith chart is still a powerful tool in
construction of lossless matching circuits. A given problem has often many solutions. With
a Smith chart it is possible to overview and chose among different alternatives. Before
exemplifying the method a few basic properties on reflection coefficients and the Smith chart
are reviewed.
The reflection coefficient Γ of an impedance Z with respect to a reference impedance

Z0 is defined
(284)
Although written like a complex impedance, the reference Z0 is purely resistive in all practical
applications we consider, commonly Z0 is 50Ω or 75Ω, but lower values like 20Ω and 10Ω
are sometimes used in transistor data. Once Z0 is known, it is convenient to normalize all
impedances in a given problem with respect to Z0. As indicated by the last equation above,
we let small letters refer to normalized impedances or admittances, while capitals stands for
the absolute counterparts. Expressed by admittances Y = Z-1, Y0 = Z0-1, the reflection coeffi-
cient is
(285)
J.Vidkjær
Coming from normalized admittances, the mapping to reflections is seen to be mathematically

equivalent to the mapping from normalized impedances, but turned 180° due to the leading
minus sign. A few basic reflection coefficients, i.e. the shorted, open, and matched load are
summarized in Fig.107(a), which also shows the following important properties,
(286)
i.e. a passive impedance maps inside or on the unit circle in the reflection plane. The circle
itself corresponds to the lossless case of a pure reactance. Impedances were used in the
demonstrations of Eq.(286) but admittances could be used as well. Fig.107(b) repeats the basic
properties taking outset in admittances.
The Smith chart is an overlay to the reflection plane that holds the mapping of a
Cartesian coordinate net from either normalized impedance or normalized admittance planes.
Both forms are shown in Fig.108. They are contrasted by the 180° rotation implied by the sign
difference of the transforming equations. Without details, the mapping given by the normalized
parts of Eq.(284) or (285) are known as the bilinear or the linear fractional transformation,
[17] chap.16. It is conformal, which implies angle conservation. A major consequence is
that circles are mapped onto circles where straight lines are included as a limit case of infinite
radius. For the present purpose the most important point is that constant r lines in the z-plane,
i.e. lines parallel to the imaginary axis map into a system of constant r circles in the Smith
chart as indicated by the upper part in Fig.108. When a reactance is series-connected to an
impedance the impedance of the connection is to be found along the constant r circle through
the original impedance. Likewise, constant g lines in the y-plane map to constant g circles in
the Smith chart. Parallel connecting a susceptance let the resultant admittance move along the
circle through the original admittance Y.
Fig.107 Mapping of short, open and matched component to reflection coefficients. Ohmic
components are transferred to the real Γ axis and reactive components to the unit
circle in the Γ-plane.
J.Vidkjær
II-8 Impedance Matching 103
Fig.108 Mappings of impedance or admittance coordinate nets to Smith charts in the

reflection plane. Series connected reactance or paralleled susceptances map onto
constant r and g circles respectively.
Knowing the reflection coefficient of a one-port, practical translations to both

normalized impedances and normalized admittances are easily made by reading the coordinates
in the appropriate Smith chart. Covering a Γ-plane with two Smith charts, one in impedance
and one in admittance orientation, provides a tool for making graphical inversions from
impedance to admittance representations and vice-versa. This is the property that makes the
Smith chart highly useful in lumped element matching problems. The technique to find a path
that connects the impedance to be matched to the complex conjugated of the matching goal
where we go along the constant r circles in impedance representation when a reactance is
series-connected and along the constant g circles in admittance representation when a suscep-
tance is parallel-connected. The following example illustrates the method step-by-step.
J.Vidkjær
Example II-8-1 ( double and standard Smith charts )
The task in consideration is the design of a lossless network, which matches a power transistor
of 6.92Ω-j16.8Ω input impedance at 150 Mhz to a 50Ω source. Using Z0=50 Ω, the starting
point, (1), and the target point, (4), are fixed in the Smith chart. Our job is to find a route
Fig.109 Impedance matching steps using a double Smith chart. The dots correspond to the
coordinates in the 1,2,3,4 sequence that is detailed by Eq.(287).
between the two points following constant r and g circles only. The choice with two interven-
ing points determines the particular network topology shown in the figure. Corresponding
components are calculated using z and y values from chart,
(287)
J.Vidkjær
Fig.110 Example of a double Smith chart ( Analog Instruments Company, N.Y.)
The double Smith chart in Fig.109 is simplified for clarity. A more dense chart is required to
reed the normalized impedances and admittances with sufficient accuracy. Fig.110 is an
example of a printed double Smith chart. It is sometimes called an immitance chart. Unfortu-
nately, the example is also one of the few double charts that are commercially available.
Doing without a double chart, several mirrorings through the center are required to substitute
the 180° rotations between impedance and admittance representations. The method is demon-
strated by Fig.111, where the previous design is repeated. Clearly, the process is more tedious
Fig.111 Impedance matching using a single Smith chart. The steps are similar to the steps
in Fig.109, but inversions between impedances and admittances require mirroring
through the center.
J.Vidkjær
than before, and it is also more difficult to overview. One guide, however, is to observe that
matching to Z0 always have a final step along either the constant r or the constant g circle
through the center. In the step prior to the last one, we may aim upon the inverse of the final
circle, so it is helpful to draw it into the Smith chart from the beginning as seen in the figure.
Example II-8-1 end
Fig.112 Lossless matching. Zin absorbs the available power Pav at the design frequency f0.
The reflection coefficient indicates the impedance difference to Z0, and it is a direct
measure for checking the performance of a design. Complete match, Γ=0, means that all the
generator available power, Pav, is transferred to the load. Since the matching network is
lossless, the only place to deposit power is the resistive part of the input impedance to be
matched. The bandwidth of power transfer in the resultant matching network is , therefore, the
3dB limits around the design frequency f0 of the ratio
(288)
Eg is the EMF in the generator of internal impedance Z0.
Fig.113 Conjugated matching across a cut in the lossless matching network.

To control bandwidth while constructing matching networks in the Smith chart we
get only crude guidelines. The result has to be checked afterwards by a circuit analysis or
simulation. Transfer of the available power at the design frequency requires conjugated
matching at the generator terminal and the load terminal. In the design process we work
through a series of cuts in the lossless matching network where conjugated matching is
assumed to maintain maximum power transfer11. Each cut can be considered as shown by
11 ) A formal proof is given in section III-1, where the so-called mismatch factor
is introduced and shown to be invariant across any lossless two-port.
J.Vidkjær
Fig.113. Including all impedances or admittances across a cut, the two imaginary parts cancel
and the two real parts add. Had the cut separated pure inductance from pure capacitance, the
ratio of absolute reactance value over twice the resistance or the absolute susceptance value
over twice the conductance would be the Q-factor of an ideal resonance circuit, either series
or parallel. The ratio of the matching network design frequency f0 over a Q-factor, which is
calculated from less ideal imaginary part, gives the so called inherent bandwidth at the
junction in consideration. In the k’th step that is
(289)
(290)
The resultant bandwidth in a design, which has bandpass characteristics, will be less than any
inherent bandwidths. The inherent Q´s are not independent from step to step, so this way of
estimating bandwidths is much more crude than any method we have used before. Neverthe-
less, inherent bandwidth may guide design decisions on where to adjust a matching network
to met bandwidth specifications. Constant Q´s patterns in a Smith chart follow the impedance
or admittance angles according to
(291)
A mapping of a polar z or y grid to the reflection plane give a system of constant angle circles
as shown in Fig.114. The grid in the reflection plane is sometimes called a polar Smith chart.
Translating to Q-values the chart shows curves of constant inherent bandwidths.
Fig.114 Mapping of a polar grid in the impedance plane to a polar Smith chart by the same
bilinear transformation, that gave the usual Smith cart.
J.Vidkjær
Example II-8-2 ( bandwidth estimations )
To exemplify bandwidth estimations, data from the previous example in the steps that have
imaginary parts give
(292)
The smallest inherent bandwidth in step 2 clearly overestimates the corresponding value of
55MHz in Fig.115(b), which is based on Eq.(288).
Fig.115 Reflection and transmission through the matching network, which was constructed
by Fig.109 and Eq.(287).
Fig.116 Impedance matching steps corresponding to Fig.109, but enforcing a higher Q and
a correspondingly lower bandwidth.
J.Vidkjær
To get a more narrowbanded design, the inherent bandwidth in step 2 should be decreased.
Enlarging inductor L1 moves the joining point closer to the border of the Smith chart and
enlarges the inherent Q. The design in Fig.116 is controlled by
(293)
Fig.117 Matching network from Fig.116 and Eq.(293) ( dotted curves from Fig.115 ).
A simulated response of the design is given in Fig.117. As seen, the bandwidth has reduced
to 31 MHz, which is close to the calculated inherent bandwidth in step 2 above. To get the
frequency response curves it was assumed that the input impedance originates from a parallel
connection of a resistor and a capacitor. It is the most probable configuration in a transistor.
In the present case, cf. step 1 in Eq.(293),
(294)
J.Vidkjær
To make bandwidth wider than the one in the first design, the inductance L1 should be
reduced. The smallest inductance, and the smallest inherent Q in step 2, corresponds to the
situation in Fig.118. Here the matching network simplifies to only two components.
Fig.118 Impedance matching to the same requirements as the circuit in Fig.109, now using
a simpler, low-Q circuit.
(295)
The resultant matching network is no longer a bandpass circuit, so bandwidth estimation

becomes somewhat misleading, although the simulated response in Fig.119 indicates a widen-
ing. Matching to transistors, as this and the previous examples were supposed to do, com-
monly require two tunable components to adjust for parameter spreadings in both the real and
the imaginary parts of device impedances. Since trimmer capacitors are far the easiest
J.Vidkjær
Fig.119 Matching network from Fig.118 and Eq.(295) ( dotted curves from Fig.115 ).
component to adjust, and since a transistor matching network must often include an inductor,
most matching circuits for this purpose commonly hold three or more components, so our last
example is a limit case. A little afterthought reveals that using three components, it is possible
to match from everywhere to everywhere in the Smith chart. Different networks are used to
avoid impractical component values with distinct topologies and matching goals.
Example II-8-2 end
If the possibility of easy tuning is unimportant in an application, two element

matchings may suffice. However, different configurations have different ranges of coverage.
They are summarized by Fig.120. where the hatched areas are the impedances that can be
matched to the center impedance in the four possible LC combinations. The regions are
covered by the element closest to the unmatched impedance, when the component nearest to
the generator in the final step follows the particular constant g or r half-circle, along which
each component type and configuration can reach the center.
Fig.120 Matching to impedance Z0 in the Smith chart center by two element LC networks.
Hatched areas show the admittance or impedance ranges that can be matched.
J.Vidkjær
J.Vidkjær
113
APPENDIX II-A, Power Calculation and Power Matching
This appendix summarizes basic power relations and impedance matching properties.
Consider first an load impedance ZL, which is driven by a single tone of frequency fo Hz. At
a given time t, the instantaneous current and voltage are expressed
(296)
IL and VL are amplitude or peak values of current and voltage respectively while ϕI and ϕV
are the corresponding phases. It is assumed that current and voltage in ZL are sensed for
power consumption as indicated by Fig.121. The instantaneous power absorbed by the load
at any time t, PL(t), is given by the product
(297)
Under stationary conditions, the average power given to the load is the mean value of the
instantaneous power over a period, i.e.
(298)
In this calculation, the last cosine integral vanishes - it would even do so had we taken only
half a period as the averaging interval. At frequency f0 average power consumption is
positive and the impedance termed passive if the cosine is positive, which requires that the
absolute phase difference between current and voltage is less than 90°. Expanding to complex
notation the average power - in brief often just the power - may be expressed
Fig.121 Current and voltage orientations for power calculations in a load impedance.
J.Vidkjær
(299)
In the rewritings above we have first introduced the scalar rms (root mean square) current and
voltage. Taking current as the main example they are given by
(300)
Evaluation of the integral follows the pattern from Eq.(298). Complex current and voltage
(301)
were subsequently substituted. Here and in the main text lower case i’s and v’s always
represent complex currents and voltages in rms scale while capitals are more freely used.
Introducing the complex impedance ZL with real and imaginary components RL and XL, i.e.
(302)
the average power is given by
(303)
Using the impedance in polar form
(304)
the power may also be expressed,
J.Vidkjær
II-A - Appendix - Power Calculation and Power Matching 115
(305)
Driving a load impedance ZL from a generator with rms-scaled EMF (electromotive

force) Eg and impedance Zg=Rg+jXg, as sketched in Fig.122(a), the load current and voltage
become
(306)
The corresponding power in the load, using Eq.(299), becomes
(307)
For a given generator Zg, Eg, where the impedance is supposed to be passive, the concept of
power matching corresponds to the situation where a passive load impedance is adjusted to
absorb maximum power from the generator. Since reactances may be both positive and
negative, one step towards optimum power is to let the last term in the denominator vanish
using
(308)
In this case power given to the load impedance is
(309)
Passive load resistances are positive so the value that gives optimum power must be found by
setting the differential coefficient of the load dependent factor above equal to zero. We get
Fig.122 Thévenin (a) and Norton (b) form of driving a load impedance ZL or admittance YL
from a generator
J.Vidkjær
(310)
Both resistances must be positive due to the passivity assumptions. The condition corresponds
to an optimum as it is easily seen that the second order derivative is negative if the two
resistances are equal. Combining the two requirements from Eqs.(308) and (310), optimum
power is absorbed in the load, if the load impedance is set equal to the complex conjugated
of the generator impedance. Under this condition, the power given to the load is the maximum
that the generator can provide, and it is termed the available power. It is expressed through
Eq.(307), which gives
(311)
Had we used the Norton form in Fig.122(b) instead of the Thévenin form from Fig.122(a) to
describe the driving of the load from a generator, the load voltage and current are expressed
(312)
Compared to Eq.(306) these expressions have the same structure if we replace generator EMF
by the short-circuit current Ig, generator impedance with generator admittance, and load
impedance with load admittance. In a strict Thévenin to Norton transformation we should use
(313)
Due to the duality between the two equivalent representations, we get similar results with
respect to the optimal power transfer from the generator to the load. The results are in
summary
(314)
For obvious reasons, power matching is sometime referred to as conjugated matching or

impedance matching.
J.Vidkjær
117
APPENDIX II-B Signal Flow Graphs
Signal flow graphs are tools for visualizing the relationships between the variables
in linear systems like electronic circuits under small-signal excursions12. Furthermore, flow
graphs make provisions for writing down transfer functions in symbolic form between pair of
variables from inspection of the graph. The latter is an alternative to solving the pertinent
system of equations by eliminating all superfluous variables. With small or sparse circuits the
signal flow graph approach is often the easiest way of the two.
Elementary Signal Flow Graphs
Basic elements in flow graphs are nodes and directed branches that connect the nodes.
The nodes represent variables, typically currents and voltages in electronic circuits. There is
a number associated with each branch. It is the factor that is multiplied on the variable at the
node, where the branch has its outset, to give its contribution to the variable at the node in
the receiving end of the branch. The branch factor is sometimes called the gain of the branch
regardless of its actual units. Nodes that have outgoing branches only are called source nodes,
and they represent the independent variables in a problem. Nodes with at least one incoming
branch hold a dependent variable. As examples the simple proportionality relationship in
Ohm’s law from Eq.(315) may be visualized by either of the two flow graphs in Fig.123(b),
the upper one using current as independent source variable and voltage as dependent variables,
the lower one using voltage as source variable and current as dependent variable.
(315)
Fig.123 Simple impedance,(a), with two flow

graph representations,(b).
The amplifier circuit in Fig.124(a), where gain and input current are given by
Eq.(316), may be represented by the flow graph in Fig.124(b) that has two final, dependent
variables, input current iin and output voltage v2. To express the gain A=v2/v1 we follow a so-
called path in the flow graph from source node v1 through the gm branch to the dependent
current variable ic and further on through the Ztot branch to the dependent output variable v2.
12 ) Signal flow graphs should not be confused with the network graphs that are used in
circuit theory to represent Kirchhoff’s equations.
J.Vidkjær
Comparing with Eq.(316), the resultant gain illustrates the rule, that the gain of two cascaded
flow graph branches is the product of the two branch gains.
(316)
Fig.124 Amplifier equivalent circuit,(a), and a corre-

sponding flow graph,(b).
Adding contributions from more variables to a dependent variable appears in a signal

flow graph by a node that has more than one incoming branch. This is the case for the
dependent terminal currents i1 and i2 in the signal flow graph representation of y-parameters
for a two-port where the port voltages v1 and v2 are the independent variables. The relations
are expressed by Eq.(317) and shown in Fig.125. The corresponding expressions and graphs
for a two-port that is characterized by z-parameters, where voltages v1, v2 are the dependent
variables and currents i1, i2 are the independent sources, are given in Eq.(318) and Fig.126.
(317)
Fig.125 Two-port y-parameters in block form,(a), or

the corresponding signal flow graph,(b).
(318)
Fig.126 Two-port z-parameters in block form,(a), or

the corresponding signal flow graph,(b).
For larger circuits with more branches or more complicated problems than the
demonstrations above, the corresponding signal flow graphs become more involved as do the
underlying system of network equation, we have to solve. One way of solving large system
of equations is by eliminating superfluous variables successively. In a signal flow graph
J.Vidkjær
II-B - Appendix - Signal Flow Graphs 119
(319)
(320)
(321)
(322)
Fig.127 Elementary reduction operations with signal

flow graphs.
representation, this process may be visualized by repeated use of a few basic rules for signal
flow graph reductions. Fig.127 and Eqs.(319) through (322) show how to eliminate a variable,
x3, and how to simplify two parallel branches to one branch. In the process of simplifying and
redrawing a signal flow graph, we may get self loops. This means that a dependent variable
is included at both sides of an equation. To simplify here, we must divide all incoming
branches, whether there is one or more, by one minus the loop gain as seen from
Eqs.(323),(324) and the corresponding signal flow graphs in Fig.128.
(323)
(324)
Fig.128 Elimination of self-loops in signal flow

graphs.
J.Vidkjær
Fig.129 Equivalent circuit for an amplifier, (a), and signal flow graphs, (b) to (e), which
illustrates the successive elimination approach to get the gain A=v2/Eg.
A brief illustration of the signal flow graph simplification approach is given in

Fig.129, which shows how to incorporate the generator and input impedance interaction in the
amplifier from Fig.124 in order to calculate the voltage gain v2/Eg. The extension to the signal
flow graph in (b) above, compared to the previous one, expresses that the transistor input
voltage v1 is the EMF of the generator minus the voltage across the generator impedance Zg.
By successive eliminations we first remove the two current variables iin and ic to get the graph
with a self loop in Fig.129(c). The loop is removed following the rule from Fig.128 and
Eq.(323), and in the final step, the transistor input voltage v1 is eliminated. The voltage gain
becomes the branch gain of the final branch in Fig.129(e), which directly connects the
independent variable Eg to the dependent variable v2, i.e.
(325)
The elimination process here is included to emphasize that the signal flow graph is
just another way of representing linear equations like network equations, and that it is possible
to operate on the graphs like we otherwise should handle the corresponding system of
equations. It is clear, however, that to conduct the successive reduction method on larger
problems would require a series of redrawings. It is not evident, that this would be beneficial
in many practical cases. However, the potential of the signal flow graph approach lies not in
tedious simplification techniques. It stems from the fact, that any desired transfer function -
like the amplifier gain above - can be formulated directly from an inspection of the first signal
flow graph that we may draw for a given problem.
J.Vidkjær
Mason’s Direct Rule
Masons direct rule provides us with a recipe for writing down transfer functions
between any particular set of an independent and a dependent variable directly from inspection
of the signal flow graph. Before phrasing the method we should be a little more strict about
terms and definitions. We use
Path: Sequence of branches between two nodes, a starting node and an end node.
All branches must be directed the same way and any node included in the
path must be encountered only once (there are no loops).
Loop: Like a path but with coinciding start and end nodes.
Nth order Loop: A set of N nontouching loops (i.e. loops without common nodes).
Path Gain: The product of all branch gains in a path.
Loop Gain: The product of all branch gains in a loop (also higher ordered loops).
The transmission Tij from an independent variable xsj represented by a source node to a
dependent variable xi represented any non-source node may be written in the form
(326)
where k sums over all distinct paths from node xsj to node xi, and where
Δ: Graph determinant, calculated by
(327)
( Note the alternating signs )
Pk: Path gain of the kth path from node xsj to node xi.
Δk : Sum of terms from Δ including the leading "1" followed by contributions

that come from loops, which don’t touch the kth path.
The proof for the approach is somewhat complicated and beyond the scope here, so we
familiarize ourself on how to use the rule in practice below. Mason, who formulated the rule,
gave an intuitive derivation in ref´s [18],[19]. However, it couldn’t satisfy mathemati-
cians and more formal proofs may be found elsewhere, for instance in [20]. For obvious
reasons, Mason’s direct rule is also known as "the non-touching loop rule".
J.Vidkjær
Fig.130 Indication of the path P1 and the loop L1 in the signal flow graph from Fig.129.
Our first demonstration is to show how Mason’s direct rule works with the example
from Fig.129. The signal flow graph for the problem is repeated in Fig.130(a). It has one path,
P1, and one loop L1, which touches the path in the v1 node, so Δ1 to be used as a factor on
P1 becomes one. We get
(328)
As seen, the resultant voltage gain equals the gain in Eq(325), which formerly was found by
successive reductions.
Fig.131 Calculation of input admittance of a loaded two-port, (a), from the corresponding
signal flow graph (b). The paths and the loop are indicated in (c).
The problem in Fig.131 is to find the input admittance of a loaded two-port, which
is characterized by its y-parameter matrix. As indicated, there are two paths and one loop in
this problem. The first path, P1, does not touch the loop, so the Δ1 factor includes the
contribution to the determinant from the loop ( in this simple, one-loop example the factor
even equals the determinant ). From the signal flow graph we identify the terms
(329)
J.Vidkjær
Fig.132 Calculation of voltage gain through a loaded twoport,(a), from the signal flow graph
in (b), where (c) indicates paths and loops. Note the 2nd order loop L1L2.
By the final example we shall find the voltage gain v2/Eg through a loaded two-port -
specified by y-parameters - which is driven from a generator with given impedance Zg. The
signal flow graph for this problem is shown in Fig.132(b). It is seen that there are three loops
in the graph, L1, L2, and L3. Furthermore, there is a second order loop made of L1 and L2,
since these two loops are non-touching. There is only one path P1 from the independent
voltage Eg to v2 and it touches all loops, so the Δ1 factor simplifies to one. The gain becomes
(330)
The last result demonstrates how the signal flow graph approach using Mason´s direct
rule - even in relatively small problems - greatly simplifies the efforts that are required to set
up transfer functions in electrical networks. To get the result in Eq.(330) by traditional
network calculations, for instance from node equations, would require much more work.
J.Vidkjær
J.Vidkjær
II Problems 125
Problems
P.II-1
Draw phasor diagrams like Fig.2b showing iC, iR, iL, i, and v at the frequencies
ω=0.1 ω0, ω=0.5 ω0, ω=2 ω0, and ω=10 ω0.
P.II-2
Fig.133
A simple receiver for the medium frequency band is shown in Fig.133(a). The
oscillator frequency, fo, is controlled by a parallel resonance circuit as is the RF
filtering after the antenna, which is tuned to the carrier frequency fRF between 550
kHz and 1.550 MHz. The mixer downconverts to the intermediate frequency fIF=fo-
fRF ≈ 450kHz, so the frequency difference between the two resonance circuits should
stay constant in tuning. Tuning elements are two identical capacitors, the Ct´s, which
are adjustable from 6pF to 100pF. Resonance frequencies of the tuned circuits are
nonlinear function of the capacitances, so exact IF frequency can be obtained at two
input frequencies only. They are chosen to fRFl=617 kHz and fRFu=1.483 MHz.
Find the components Lrf, Crf, Lo, Co, and draw a curve showing the deviation from
the nominal IF frequency of 450 kHz, the so-called tracking error. What is the
maximum error?
The IF error may be substantially improved using a so-called padding capacitor, Cp
in the oscillator circuit as shown in Fig.133(b). Recalculate all components using Cp
= 118 pF, and draw a new curve of frequency deviations. What is the new maximum
tracking error?
P.II-3
Fig.134
J.Vidkjær
Derive an expression for the transient decay voltage Vdcy(t), t≥0, when the switch in
Fig.134 breaks the dc-current IL0 at t=0. What is the maximum value of Vdcy if
L=100nH, C=47pF, Rp=3.3kΩ, and IL0=15 mA ?
P.II-4
Fig.135
The Π-circuit C1, L, C2 is inserted to match a transistor of input impedance 150Ω
paralleled by 12.3 pF at 100 MHz. C2 is chosen to 67.7 pF. Use series-to-parallel
techniques to find L and C1. Calculate the voltage ratio vin / Eg at the center frequen-
cy and estimate the 3dB bandwidth of power transfer to the transistor.
P.II-5
Fig.136
Show that the amplifier above in Fig.136(a) with the rudimentary transistor model
Fig.136(b) can realize a third order lowpass Butterworth frequency characteristic.
Find the components C0, C1, and L1 that give the amplifier 230MHz bandwidth when
gm=50mS and Rg=RL=300Ω.
Why is it not possible to use a parallel tuned circuit at the output to realize the
complex conjugated pole pair as it is done in bandpass stagger tuned amplifiers?
P.II-6
Fig.137
J.Vidkjær
II Problems 127
Fig.138
A bipolar transistor is described by the equivalent circuit in Fig.137. It is used in a

tuned amplifier, which has functional diagram as shown by Fig.138. The two induc-
tors L1 and L2 are tapped and used as autotransformers where tight coupling is
assumed. Design the amplifier, i.e. find L1, N1, C1, C2, L2, Rp, and N2 to meet the
specifications:
- center frequency = 110 MHz,
- 3dB bandwidth of v2/Eg = 8 MHz in synchronous tuning,
- impedance matching at the input side
- impedance matching at the output side with a total collector load of 80Ω
at the center frequency
Calculate - in dB - the center frequency voltage gain v2 / Eg and the operating power
gain Pout / Pin.
P.II-7
Fig.139
A bipolar transistor, for which we use the equivalent circuit in Fig.139, is employed
in a tuned amplifier that has the functional diagram shown in Fig.140.
Fig.140
The requirements to the amplifier are,

- center frequency, f0 = 50 MHz,
- 3dB bandwidth, BW3dB = 4 MHz,
J.Vidkjær
- 3rd order Butterworth frequency characteristics for v2 / Eg, and

- center frequency impedance matching at the input.
The Butterworth characteristic is achieved by tuning the input circuit to the center
frequency and let the double-tuned output circuit be coupled overcritically. It is
assumed that the two output resonance circuits operates identically, i.e. with same
resonance frequencies and q-factors. To ensure stability - and usefulness of the
transistor equivalent circuit - R2 is set to 100Ω.
Find the components C11, C12, L1, L2, Ck, C21, and C22.
Calculate the operating power gain, Pout / Pin at the center frequency .
P.II-8
Fig.141
The amplifier stage in Fig.141 is loaded with a critically coupled double-tuned circuit,
where the two sides are equal using C = 60 pF. The transistor is modeled by a simple
transconductance gm=18 mS. Find R, L, and k so the gain v2/v1 gets 3dB bandwidth
equal to 300 kHz around a center frequency of 10.7 MHz.
Two of the stages above are cascaded. What is the resultant 3dB bandwidth? Sketch
the low and high frequency asymptotes of the voltage gain.
P.II-9
Fig.142
Explain the function of the FM detector in Fig.142.

Calculate ( Matlab, Matematica, etc. ) detector characteristics and find the kQ that
gives the best linearity.
J.Vidkjær
II Problems 129
Compare performance with the quadrature demodulator ( could show why the
principle in Fig.142 was never successful )
P.II-10
Fig.143
Find component values Ca, Cb, Cc, La, and Lb so the amplifier in Fig.143(a) gets a
maximally flat gain characteristics, v2/v1, of bandwidth 8 MHz and center frequency
100 MHz. Fig.143(b) shows the transistor equivalent circuit. It is supposed that the
two sides of the output double-tuned circuit have equal Q-factors and center frequen-
cies but differs in component values due to the impedance transformation between
the load RL=50Ω and the transistor output resistance, Ro=215Ω. Calculate the center
frequency voltage gain using gm=70mS.
P.II-11
A transistor has the input impedance 87Ω-j82Ω at 160 MHz. Use a Smith chart to
design network that matches the transistor to a 50Ω source with an inherent band-
width not exceeding 80 MHz. The network should include two capacitors to make the
network suitable for adjustments.
P.II-12
Fig.144
A matching network for a RF power MOS transistor is shown in Fig.143, Fig.144.

The input impedance is the nominal value from the data sheets (Motorola, MRF136).
The two capacitors are identical trimmers that are adjustable from 15pF to 115 pF.
J.Vidkjær
What is the value of the inductor L if it is supposed that matching is achieved when
C1 is set to 45 pF ?
Show in a Smith chart the range of input impedances that can be matched using the
full adjustment range of the two trimmers.
P.II-13
Repeat problem P.II-4 using a Smith chart (exclude gain computations)
J.Vidkjær
131
References and Further Reading
[1] L.Weinberg, Circuit Analysis and Synthesis, McGraw-Hill 1962
[2] G.C.Themes, S.K.Mitra ed´s, Modern Filter Theory and Design, Wiley 1973
[3] R.Shaumann, M.S.Ghausi, K.R.Laker, Design of Analog Filters, Passive, Active RC

and Switched Capacitor, Prentice Hall 1990.
[4] R.E.Collin, Foundations for Microwave Engineering, 2nd ed. McGraw-Hill 1992.
[5] R.Goyal ed., High-Frequency Analog Integrated Circuit Design, Wiley N.Y., 1995.
[6] H.W.Bode, Network Analysis and Feedback Amplifier Design, D.Van Nostrand,
1945.
[7] S.Ramo, J.R.Whinnery, T.Van Duzer, Fields and Waves in Communication Elec-
tronics , 3rd.ed., Wiley 1994.
[8] R.A.Adams, A Complete Course in Calculus, 3rd.ed., Addison-Wesley, 1995.
[9] R.A.Rohrer, Circuit Theory, An Introduction to the State Variable Approach, Mc-
Graw-Hill 1970.
[10] N.Balabanian, T.A.Bickart, Electrical Network Theory, Wiley, 1969.
[11] F.E.Terman, Electronic and Radio Engineering, 4th.ed, McGraw-Hill 1955
[12] P.L.D.Abrie, The Design of Impedance-Matching Networks for Radio-Frequency and

Microwave Amplifiers, Artech House, 1985.
[13] B.Gilbert,"A Precise Four-Quadrant Multiplier with Subnanosecond Response",

IEEE.J. Solid-State Circuits, Vol. Sc-3,pp.365-373, Dec.68, reprinted in [15].
[14] P.R.Grey, R.G.Meyer, Analysis and Design of Analog Integrated Circuits, 3rd
ed.,Wiley 1993.
[15] A.B.Grebene, ed., Analog Integrated Circuits, Part V: Multipliers and Modulators (
5 papers incl. [13] ) , IEEE Press, 1978.
[16] B.Becciolini,"Impedance Matching Networks Applied to RF Power Transistors",

AN721, Motorola RF-Application Reports, 1995 ( includes many more application
notes on matching methods and problems ).
[17] E.Kreyzig, Advanced Engineering Matemathics, 7th ed., Wiley 1993.
[18] S.J.Mason,"Feedback Theory: Some Properties of Signal Flow Graphs", Proc.IRE,

vol.41, pp.1144-1156, 1953
[19] S.J.Mason,"Feedback Theory: Further Properties of Signal Flow Graphs", Proc.IRE,

vol.44, pp.920-926, 1956.
J.Vidkjær
[20] K.Thulasiraman,"Signal Flow Graphs", Chapter 8 in W.K.Chen (editor) "The Circuits

and Filters Handbook", CRC and IEEE Press, 1995.
J.Vidkjær
133
Index
β(ω) . . . . . . . . . . . . . . . . . . . 3, 12, 14 identical resonance circuits . . . . 81
A_L value (inductors) . . . . . . . . 47, 62 inductive coupling . . . . . . . . . . 82
Autotransformer . . . . . . . . . . . . . . . . 61 LF and HF asymptotes . . . . . . . 98
Bandwidth unsymmetrical load . . . . . . . . . 96
double tuning, kQ=1 . . . . . . . . 94 Envelope Detector
parallel circuit . . . . . . . . . . . . . 4 ideal . . . . . . . . . . . . . . . . . . . . 87
series circuit . . . . . . . . . . . . . . 13 External Q-factor . . . . . . . . . . . . . . . 25
Bandwidth Reduction Factor . . . . . . . 32 FM Detector
Bandwidth Shrinkage Foster-Seely . . . . . . . . . . . . . . 85
!see bandwidth reduction . . . . . 32 quadrature . . . . . . . . . . . . . . . . 88
Broadband Modeling . . . . . . . . . . . . 26 Foster-Seely FM detector . . . . . . . . . 85
Butterworth Characteristic . . . . . . . . . 36 Frequency Modulation, FM
double-tuned circuit . . . . . . . . . 95 Foster-Seely detector . . . . . . . . 85
Capacitive Coupling quadrature detector . . . . . . . . . . 88
double-tuned circuits . . . . . . . . 82 Gain-Bandwidth factor, GBF . . . . . . . 31
Capacitive Transformer . . . . . . . . . . . 64 double-tuning . . . . . . . . . . 95, 98
Close Coupling . . . . . . . . . . . . . . . . 47 maximum . . . . . . . . . . . . . . . . 31
Conjugated Matching . . . . . . . . . . . 116 Gain-Bandwidth Product, GW . . . . . . 31
Coupled Inductors . . . . . . . . . . . . . . 49 Gaussian frequency characteristic . . . . 35
PI-equivalent circuits . . . . . . . . 51 Gilbert Cell
T-equivalent circuits . . . . . . . . . 49 quadrature FM detector . . . . . . 92
Coupled Resonance Circuits Hybrid
! see Double-Tuned Circuit . . . . 80 transformer . . . . . . . . . . . . . . . 72
Coupling Immitance . . . . . . . . . . . . . . . . . . . . 18
tight or close inductive . . . . . . . 47 Immitance Chart . . . . . . . . . . . . . . 105
Coupling Coefficient . . . . . . . . . 46, 82 Impedance Matching . . . . . . . 101, 116
autotransformer . . . . . . . . . . . . 61 lumped elements . . . . . . . . . . 101
capacitive . . . . . . . . . . . . . . . . 82 two element LC . . . . . . . . . . . 111
inductive . . . . . . . . . . . . . . 46, 82 Inductive Coupling . . . . . . . . . . . . . . 45
PI equivalent circuit . . . . . . . . . 53 double-tuned circuit . . . . . . . . . 82
Critical Coupling . . . . . . . . . . . . . . . 84 Inherent Bandwidth . . . . . . . . . . . . 107
Damping Ratio Loaded Q-factor . . . . . . . . . . . . . . . 25
parallel circuit . . . . . . . . . . . . . 2 Logarithmic decrement . . . . . . . . . . . 10
series circuit . . . . . . . . . . . . . . 11 Lowpass to Bandpass Transfor-
Diode-Ring Mixer . . . . . . . . . . . . . . 78 mation . . . . . . . . . . . . . . 17
Double-Balanced Modulator Mason’s Direct Rule . . . . . . . . . . . 121
quadrature FM detector . . . . . . 92 Mixer
Double-Sideband Suppressed Carrier, diode-ring . . . . . . . . . . . . . . . . 78
DSC-SC Mutual Inductances . . . . . . . . . . . . . 45
modulator . . . . . . . . . . . . . . . . 78 Narrowband Approximation . . . . . . . 14
Double-Tuned Circuits . . . . . . . . . . . 80 _beta(omega) . . . . . . . . . . . . . 14
capacitive coupling . . . . . . . . . 82 accuracy . . . . . . . . . . . . . . . . . 15
critical coupling . . . . . . . . . . . . 84 double-tuning . . . . . . . . . . 82, 96
J.Vidkjær
pole-zero conditions . . . . . . . . . 16 polar . . . . . . . . . . . . . . . . . . 107

resonance circuit . . . . . . . . . . . 15 Spiral Inductor . . . . . . . . . . . . . . . . 26
Padding Capacitor . . . . . . . . . . . . . 125 Stagger Tuned Amplifier
Parallel Resonance . . . . . . . . . . . . . 2 Butterworth . . . . . . . . . . . . . . . 36
bandwidth . . . . . . . . . . . . . . . . 4 Tapped Inductor
damping ratio . . . . . . . . . . . . . 2 !see autotransformer . . . . . . . . . 61
inductor series loss . . . . . . . . . 23 Three-Winding Transformer . . . . . . . 67
poles and zeros . . . . . . . . . . . . 6 passivity . . . . . . . . . . . . . . . . . 67
quality factor . . . . . . . . . . . . . . 2 Tight Coupling . . . . . . . . . . . . . . . . 47
resonance frequency . . . . . . . . . 2 Transformer . . . . . . . . . . . . . . . . . . 45
transient response . . . . . . . . . . 8 autotransformer . . . . . . . . . . . . 61
Parallel-to-Series Conversion capacitive . . . . . . . . . . . . . . . . 64
!see series-to-parallel conver- hybrid . . . . . . . . . . . . . . . . . . 72
sion . . . . . . . . . . . . . . . . 19 ideal . . . . . . . . . . . . . . . . . . . . 48
Passivity perfect . . . . . . . . . . . . . . . . . . 48
coupled inductors (2) . . . . . . . . 46 RF . . . . . . . . . . . . . . . . . . . . . 53
coupled inductors (3) . . . . . . . . 67 three-winding . . . . . . . . . . . . . 67
Poles and Zeros tuned . . . . . . . . . . . . . . . . . . . 57
Butterworth amplifier . . . . . . . . 39 uncoupled inductors . . . . . . . . . 64
double-tuned circuits . . . . . 83, 97 Transient Response
parallel circuit . . . . . . . . . . . . . 6 parallel circuit . . . . . . . . . . . . . 8
series circuit . . . . . . . . . . . . . . 13 Tuned Amplifier . . . . . . . . . . . . . . . 30
Power autotransformer coupling . . . . . 62
calculation . . . . . . . . . . . . . . 113 Butterworth characteristic . . . . . 36
matching . . . . . . . . . . . . . . . . 113 double-tuning . . . . . . . . . . . . . 94
Q-factor Gaussian characteristic . . . . . . . 35
see quality factor . . . . . . . . . . . 2 stagger tuning . . . . . . . . . . . . . 36
Quadrature Detector, FM . . . . . . . . . 88 synchronous tuning . . . . . . . . . 31
Quality Factor transformer-coupled . . . . . . . . . 58
inherent . . . . . . . . . . . . . . . . 107 Tuned Transformer . . . . . . . . . . . . . 57
loaded, unloaded, external . . . . . 25 Unloaded Q-factor . . . . . . . . . . . . . . 25
parallel circuit . . . . . . . . . . . . . 2 Winding Ratio . . . . . . . . . . . . . . . . . 47
reactive components . . . . . . . . . 25 capacitive transformer . . . . . . . 63
series circuit . . . . . . . . . . . . . . 11 effective inductive . . . . . . . . . . 49
Reflection Coefficient . . . . . . . . . . . 101 ideal, inductive . . . . . . . . . . . . 47
RF-Transformer . . . . . . . . . . . . . . . . 53 physical . . . . . . . . . . . . . . . . . 48
frequency response . . . . . . . . . . 54 uncoupled inductors . . . . . . . . . 64
Series Resonance . . . . . . . . . . . . . . . 11
damping ratio . . . . . . . . . . . . . 11
poles and zeros . . . . . . . . . . . . 13
quality factor . . . . . . . . . . . . . . 11
resonance frequency . . . . . . . . . 11
Series-to-Parallel Conversion . . . . . . . 19
Signal Flow Graphs . . . . . . . . . . . . 117
Smith Chart . . . . . . . . . . . . . . . . . 102
J.Vidkjær

CH 2

Uploaded by

Copyright:

Available Formats

CH 2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

CH 2

Uploaded by

Copyright:

Available Formats

Class Notes, 31415 RF-Communication Circuits

II RF-Circuits, Concepts and Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

II RF-Circuits, Concepts and Methods

In RF-communication system, control of frequency bands and impedance matching

- understand data sheets and literature,

Scanning contents, the chapter starts summarizing basic properties of resonance

Power matching is fundamental for designing and understanding many RF circuits.

II-1 Parallel Resonance Circuits

Fig.1 Parallel resonance circuit

Fig.2 Parallel resonance. (a) Susceptance composition as function of frequency. (b)

Poles and Zeros

While there is a tradition of using Q for characterizing frequency responses of

expressions in Eq.(17)(b) are based on the estimate

Fig.7 Calculation of Zp(ω) from poles

Fig.8 Charging of capacitor C to voltage VC0 = QC0/C.

the time domain responses may be condensed to read

The corresponding time constant,

II-2 Series Resonance Circuit Summary

Fig.10 Series resonance circuit

A series resonance circuit is connected as shown by Fig.10 and Eq.(32) summarizes

II-3 Narrowband Approximations

Frequency dependency of the impedance in the parallel resonance circuits was

Fig.15 Impedance of parallel RC circuit.

Fig.17 Contributions from poles s1, s2 and zero s0 to Zp in narrowband approximation

Simplifications by the narrowband approximations for resonance circuits are beneficial

Fig.18 Lowpass-to-bandpass trans-

If the lowpass response of a network is known in form of a transfer or immitance3 function

If the approximation applies, we have narrowband conditions. Here, a lowpass pole-zero

3) Immitance is a collective name for impedance and admittance.

II-4 Series-to-Parallel Conversions

Similarly, equating impedances gives the parallel-to-series conversions,

If reactance is the dominating contribution to the impedance in consideration, that is either Rs

Example II-4-1 ( impedance matching )

Example II-4-1 end

Conversions in Narrowband Applications

Reactances of inductors and capacitors depend on frequency so - strictly speaking -

Identification of Q-factor through the first order denominator term gives

The combination of Q-factors is sometimes expressed in a terminology of loaded, unloaded,

Conversions in Broadband Modeling

In narrowband applications of reactive components it suffices to specify its reactance

Example II-4-2 ( spiral inductor )

Fig.31 Complete equivalent for the spiral inductor measurements.

Example II-4-2 end

II-5 Tuned Amplifiers

Fig.33 Functional and simplified equivalent circuit of a single-tuned amplifier.

There exists a series of fundamental theorems concerning GBF’s of interstage

An amplifier stage holding only one resonance circuit is called a single-tuned

Synchronously Tuned Amplifiers

Fig.34 Normalized amplitude and phase characteristics in synchronous tuning. Gain-band-

Table I True and approximated Gain-Bandwidth Factors ( Bandwidth Reduction Factors

Example II-5-1 ( synchronous tuning )

Fig.35 Amplifier principle, (a), and transistor equivalent circuit, (b).

The corresponding calculations at the output side of the transistor read,

The frequency characteristic of the amplifier is shown later in Fig.43.

Example II-5-1 end

The phase characteristic in synchronous tuning with N stages may be written

Inserting ΩN into the phase characteristic above gives

Butterworth Stagger Tuned Amplifiers

Fig.38 Pole patterns in N-th order Butterworth squared magnitude |BN(S)|2.