Communication circuits

part 1

EE312

by

A. DAHIMENE

Institut de Génie Electrique et Electronique

Université M’Hamed Bougara

Boumerdes

1
 Communication circuits ...............................................................................................
Introduction ....................................................................................................................................................... 3
Chapter 1 ..........................................................................................................
Review of Electronic Devices ...........................................................................................
1.1 Some Generalities about networks. ............................................................................................................... 4
1.2 The Diode Model: ........................................................................................................................................ 5
1.3 The Bipolar Junction Transistor:................................................................................................................... 7
Biasing a BJT ................................................................................................................................................... 12
1.4 The Field Effect Transistor. ........................................................................................................................ 22
Biasing the JFET .............................................................................................................................................. 25
Chapter 2 ...........................................................................................................
Passive frequency dependent networks .....................................................................................
and transformer like networks............................................................................................
2.1 Impedance and Admittance ........................................................................................................................ 30
Parallel series transformations ...................................................................................................................... 31
2.2 Two port networks ..................................................................................................................................... 32
2.3 Mutual Inductance and transformer ............................................................................................................ 34
2.4 Parallel RLC circuit .................................................................................................................................... 36
2.5 Series Resonant Circuit .............................................................................................................................. 41
2.6 Parallel Resonant Circuit with Series Loss .................................................................................................. 42
2.7 Transformer like networks.......................................................................................................................... 44
Response of a tank circuit to a periodic input ................................................................................................... 59
Chapter 3 ...........................................................................................................
Non Linear Controlled Sources...........................................................................................
3.1 Piecewise linear characteristic .................................................................................................................... 62
3.2 Square law characteristic ............................................................................................................................ 67
3.3 Exponential characteristic ........................................................................................................................... 69
3.4 Resistively biased BJT ............................................................................................................................... 76
3.5 Differential Characteristic .......................................................................................................................... 80
3.6 Effect of series resistance ........................................................................................................................... 84
3.7 Clamp biased FET ...................................................................................................................................... 90
3.8 Non linear loading of tank circuits .............................................................................................................. 92

2
 Communication circuits

Introduction

This course is about electronic circuits applied to communication. We will

deal essentially with C.W. communication. In C.W. communication, we need to
produce carriers, amplify band pass signals, and multiply either a band pass signal
with a carrier or a baseband signal with a carrier. We will also consider raising
the level of a signal in order to either demodulate it or to transmit it. We see that
most of the signal processing needs some non linear processors. So, the different
electronic devices that we will consider in this course are going to be studied as
large signal amplifiers and their behavior will be essentially non linear.
In order to obtain analytically tractable models for the different electronic
components, we will assume that they are memoriless. This means that the type of
analysis that we will perform in this course is valid at frequencies below the
different cut-off frequencies. So, the models will be static and not dynamic.
We assume that the student has a basic knowledge of network analysis and
basic electronic devices (diodes, transistors and operational amplifiers). However,
we are going to provide a small review of that basic knowledge in the first chapter
of the course.

3
 Chapter 1

Review of Electronic Devices

1.1 Some Generalities about networks.

The electronic devices that we are going to analyze are essentially non
linear active devices. So, we start by defining the notions of linearity and activity
in networks.
Linear Network: A network is linear if superposition applies.
Active Network:

ik
i1
v1 vk

ii iN
vi vN

Consider the above N port network. The average power dissipated by the
network is:
P  v1i1  v2i2   vN iN
If P  0 , the network dissipates power and should be considered as passive.
If P  0 , the network provides power to some circuits connected to its ports (it is
amplifying power), and at that time the network is active. If P  0 , then we say
that the network is lossless. Networks built with pure inductors, capacitors and
transformers are lossless. In the above representation of the network, we should
not include the power supplies as input ports because in that case, all networks
4
will be passive. This is due to the fact that amplifiers (active devices) transfer
power from the power supply to the output port. A device like a diode is a
nonlinear one port device, but it is passive. A voltage source will be an active one
port device according to the above definition.

1.2 The Diode Model:

The diode is built using a junction of P and N doped semi-conductors. If we

apply a voltage across the junction, a current is going to flow though the junction
and is composed of two different types of carriers: The majority carriers which
compose the diffusion current and the minority carriers that compose the
saturation current.

id

vd

qvd
The diode equation is thus: id  I s exp I s where the first term is the
kT
diffusion current which depends mostly on the applied voltage vd while the
saturation current I s is independent on the applied voltage (as long as vd is
smaller than the avalanche or Zener voltage). The different constants are: q: the
electron charge = 1.6 10-19 C, k: Boltzman's constant = 1.38 10-23 J/°K and T: the
kT
temperature in °K. At ambient temperature (T = 300°K), the constant has the
q
value of 26 mV.
At this point, we should indicate that the electrical field in a reverse biased
junction (inside the "depletion region") accelerates the minority carriers.
The following figure shows the forward characteristic of a diode for currents that
do not exceed 10 mA.

5
 10

7
Diode Current (mA)

0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Diode Voltage (V)

Fig. 1-1 I-V characteristic of a Diode

The above figure shows clearly that as long as the voltage is smaller than a
threshold V0, the current is essentially zero, and that above threshold the current
is not limited at all. So, a good approximation of the diode is a voltage controlled
switch called the "ideal diode" followed by a battery.

ideal
Closed when vd>0

6
So, finally, a quite simple model (and at the same time fairly accurate) is the
following one.

Ideal V0

The threshold value will depend on the technology and the type of semi-
conductor used. Typical values for V0 valid for currents between 0.1 to 10 mA
are: 0.2 V for Ge, 0.7 V for Si, around 1 V for leds (GaAs, etc) and about 0.4 V
for Schottky diodes. Depending on the problem at hand, we will use one of the
different models seen above.
The exponential model will be used to model the Bipolar Junction
transistor (BJT).

1.3 The Bipolar Junction Transistor:

Consider a piece of silicon with three areas that will form two PN
junctions.

Emitter Collector

P N P

Base

Let the base-emitter (BE) junction be forward biased and the collector-base
(CB) junction be reverse biased. At that time, a majority carrier injected at the
7
emitter (a hole going from emitter to base or an electron going from base to
emitter) is going to find itself inside the depletion region of the reverse biased
collector base junction. From our previous discussion, we know that a minority
carrier is accelerated by the electrical field inside a reverse biased junction. A
majority carrier for the BE junction is a minority one for the CB junction. So,
most of the carriers injected in the BE junction will pass to the CB junction. In
other words, practically all of the emitter current will pass to the collector.

E C
IEN

FIEN

VCB
VEB

B
Fig.1-2 PNP transistor in normal operation

We can write I C   F I EN  I C 0 , where  F is a number that is close to one,

IEN is the current in the forward biased BE junction and IC0 is the saturation
current of the reverse biased CB junction. The collector current will depend only
on the injected emitter current and not on the applied voltage between the
collector and the base.

8
 IC

3mA IE=3mA

2mA IE=2mA

IE=1mA
1mA

VCB
-0.7V

Fig.1-3 Output characteristic for the BJT

The above set of curves shows that the output I-V characteristic of a bipolar
junction transistor is just a set of translated diode characteristic.
A more complete model can be obtained if we consider also a reverse
transistor (CB forward biased and BE reverse biased). At that time, we can
superpose the two transistors and we obtain the "Ebers-Moll" model.

IE IC
E C
IEN ICR
RICR FIEN
VEB

VCB

B
Fig.1-4 The Ebers-Moll model for NPN BJT

9
The Ebers-Moll model can be summarized by the following set of equations:

I E  I EN   R I CR
I C   F I EN  I CR
I E  IC  I B
And
qV BE
I EN  I E 0 (e kT  1)
qV BC
I CR  I C 0 (e kT  1)
The two saturation currents and the two gains are related via:
F I E 0  R I C 0

The above model is most useful in the analysis of common base circuits.
However, most of the circuits will use the BJT in common emitter. If we solve the
above equations, we can obtain the following relationship:
qV CE
e kT 
F
IC R
 hFE qV CE
IB hFE
e kT 
hFC
Where:
F
hFE 
1  F
R
hFC 
1  R
We see the usual relationship between the collector current and the base current
for large VCE, I C   I B ,   hFE , but for small values of VCE, we can remark
that the output characteristic does not pass by the origin, but all the curves start
from a voltage VCEsat given by:
kT  F
V CEsat  ln
q R

10
 IC

VCEsat VCE

Fig.1-5 IC-VCE characteristic for several values of IB

The above relationship does not take into account the "Early effect" which
implies a finite output impedance (inversely dependent on the collector current).
If we consider the two diodes (BE and CB), we can define four modes of
operation of the BJT.
BE forward biased, CB reverse biased: Normal operation.
BE reverse biased, CB forward biased: reverse transistor operation.
BE reverse biased, CB reverse biased: transistor is in cut-off mode.
BE forward biased, CB forward biased: transistor is saturated.
In order to use properly a BJT, we have to bias it correctly, i.e. make sure
that under all conditions, the BE junction remains forward biased and the CB
junction remains reverse biased. Usually, we select a given "Quiescent" point
(Q point), which means a voltage VCEQ and a current ICQ in the above set of
curves (Fig. 5) and we use resistors and power supplies in order to achieve the
required Q point.

11
Biasing a BJT
We assume given a Q point VCEQ, ICQ. The simplest biasing network is the
following two supply network:

T1

RC
RB

VBB VCC

Fig.1-6 Dual supply Biasing

Since we assume that the transistor is correctly biased, the equation

governing the BE junction simplifies to :
qv BE
i E  I ES e kT
And the Q point is achieved via:
V BB  R B I BQ V BEQ
I CQ   I BQ
I CQ
I EQ 

kT I EQ
V BEQ  ln
q I ES
V CEQ V CC  RC I CQ
The above set of equations is a transcendental set and can be solved by successive
approximation. However, we make a very small error by assuming that
12
VBEQ = 650 mV for discrete silicon BJT. Typical values of IES are about 2 10-16 A
for integrated silicon transistor, 2 10-14 A for a discrete silicon transistor and about
2 10-7 A for a germanium transistor. Solving for VBEQ for emitter currents between
0.1 mA and 10 mA produce a BE voltage that remains around 650 mV for a
discrete Si transistor, 750 mV for an integrated one and 220 mV for a Ge
transistor.
Since the two batteries have the same polarity, we can simplify the circuit
and use a single power supply. At that time the circuit becomes:

RB RC

T1 VCC

Fig.1-7 Single battery biasing

We can apply the same set of equations as for the previous circuit by
replacing VBB by VCC.

Example 1:
Consider the circuit of Fig.7 with a transistor having IES=2 10-14 A, RB = 1 MΩ,
RC = 5 kΩ, β = 120, VCC = 10V. Compute ICQ and VCEQ.
With β = 120, we obtain α = 0.9917. We have to compute ICQ. We start the
iteration by assuming a value of VBEQ = 650 mV. This provides the following
values for the different currents:

13
 kT I EQ
IBQ= 9.35 μA, ICQ=1.1 mA and IEQ=1.1 mA, replacing in V BEQ  ln , we
q I ES
obtain VBEQ = 643.7 mV. Another iteration will produce the same values. So, we
can say that ICQ = 1.1 mA and this provides VCEQ = VCC-RCICQ = 4.39 V.

Example 2:
Consider the same circuit with a transistor having 50    300 . We want
to bias it around the Q point ICQ = 1 mA and VCEQ = 5 V. Because of the wide
spread of the different values of β, we are going to use the geometric mean of 300
and 50. So   50  300 = 122. This provides IBQ = 8.2 μA and RB = 1.14 MΩ.
RC = 5 kΩ.
If we use these resistors with a transistor having an actual β = 50, the
collector current will reduce to: ICQ = 0.4 mA. And this will imply a collector
voltage VCEQ = 8 V. So, if the output ac voltage exceeds 2 V peak, we will have
distortion by cut-off.
If the true β = 300, then ICQ = βIBQ = 2.4 mA and this collector current will
produce a drop of voltage of 12 V across RC, which is evidently impossible. This
simply means that the transistor will be saturated at its Q point and V CEQ = VCEsat
V V CEsat
while ICQ will be given by I CQ  CC .
RC

Example 2 shows clearly that the above method of biasing is too dependent
on the value of β, and for most transistors, the fabricant can only guarantee that β
is within a wide spread of values. So, instead of trying to impose the base current,
a better method will consist of imposing the emitter current, since the relation
between emitter and collector current is via α which is always very close to one.
The next circuit that we will study will impose the emitter current via a
negative power supply.

14
 +VCC

RC

T1

Vin

RE CE

-VEE

Fig.1-8 biasing using a negative supply

For the above circuit, the ac source is directly coupled to the base. From the
dc point of view, the base is directly connected to ground. So, the capacitor C E
will charge to VDCQ = VEBQ (we assume that Vin = 0 V, if Vin  0, then due to the
non linearity of the base-emitter junction, the voltage across the capacitor will
depend on the ac voltage also). We can compute the quiescent emitter current:
V V DCQ
I EQ  EE
RE
I CQ   I EQ
V CEQ V CC  RC I CQ

It is evident that changing the transistor will not affect (significantly) the Q
point and if VEE>>VDCQ, then the emitter current can be set with a very high
precision. The capacitor CE is a by-pass capacitor. Its impedance should be
smaller than the impedance in parallel at the lowest frequency. The impedance in
parallel with CE is the parallel combination of RE and the dynamic resistance
dv kT
re  BE  which is usually much smaller.
di E i  I qI EQ
E EQ

If we cannot use a negative power supply, we can use the same circuit if we
raise the dc voltage level of base. We can use a voltage divider to do so.

15
 VCC

R1 RC

IP

T1
IBQ

R2 RE
CE

Fig.1-9 single supply resistive biasing

The Thevenin equivalent circuit of the base emitter circuit is

T1
RB

CE
VBB RE

Fig.1-10 Base emitter equivalent circuit

Where:
R2
V BB  V CC
R1  R 2
R1R 2
RB 
R1  R 2
Using the circuit of Fig.1-10, we obtain the following equations:

16
 V BB  R B I BQ V BEQ  R E I EQ
I BQ  (1   )I EQ
This provides:
V BB V BEQ
I EQ 
R E  (1   )R B
kT I EQ
V BEQ  ln
q I ES
And:
V CEQ V CC  RC I CQ  R E I EQ
I CQ   I EQ
We will come back to the previous circuit later in the course. We are going to see
methods for biasing the circuit of Fig.1-9. The first method is based of the
I
stability factor SV BE  CQ . Using the above equations, we obtain:
V BE
 1
SV BE  
R E  (1   )R B R E
We know that the base emitter voltage has a variation that is inversely
proportional to variation of the temperature, V BEQ  2.2mV / C  T , so, if
we are given a variation of ICQ for a given variation of temperature T , we can
R
compute the value of RE and then RB such that (1   )R B  E .
10

Example 3:

We want to bias a 2N3904 transistor having βmin = 50 with ICQ = 1 mA

and we can tolerate a variation of  10% of collector current when the ambient
temperature varies by  30°C, the power supply is 10 V.
T  30C , so V BE  2.2mV / C  30C  - 66mV , so
V BE 66mV
RE     660 
I CQ 0.1mA
1
And then, since (1   )  , we obtain
 1
(   1)R E 51 660
RB    3366 
10 10
We leave the remaining calculation as homework (Compute R1, R2, RC and CE).

17
You should note that we have used the value of βmin. It is because it corresponds
to the maximum value of IBQ.
Another method of design is to use the following rule of thumb.
The voltage VE across the capacitor CE should be set to a value of about
10% of VCC as long as it is larger than 1 V. A smaller value of VE will lead to
thermal instability. The voltage divider current IP should be set to a value that is
at least equal to ten times the maximum base quiescent current. We can repeat the
same design as in example 3.
V 1V I 1 mA
V E  CC  1V so R E   1 k  , I BQ max  CQ   20  A
10 1 mA  min 50
The current IP is given by: I P  I BQ max 10  0.2 mA so the sum of the two
V CC 10V
resistors R1 and R2 is given by: R 1  R 2    50 k  and the
IP 0.2 mA
voltage at the base of the transistor is:
R2
V B V CC V E V BEQ  1V  0.65V  1.65V and we obtain
R1  R 2
R2 = 8.25 kΩ and R1 = 41.75 kΩ. The bypass capacitor can be computed if we
know the lowest frequency to be amplified. Let us assume that it is 100 Hz.
The dynamic resistance of the base emitter junction is:
kT 26 mV
re    26 
qI EQ 1 mA
1 re
So if Z CE  , we obtain C E  612  F .
2 fC E 10

The next biasing system consists of replacing the emitter resistor RE by a current
source. The type of current source we will demonstrate is the current mirror
which commonly used in integrated circuits. This circuit is shown in Fig.1-11.

18
 VCC

RC

T1

Vin

CE

IB2
T2
IRB

T3
RB IE3

-VEE

Fig.1-11 Biasing using a current mirror

Since we are considering integrated circuits, we can always take for granted that
the only way that two transistors will differ will be through their geometry. So, if
we use the same masks to diffuse the two transistors, they will be identical. So, in
the analysis of the above circuit, we assume that T2 and T3 are identical, which
qV BE k

means that I ES 2  I ES 3 . For each transistor, we can write I Ek  I ES k e kT , and

if we consider T2 and T3, we can remark that V BE 2 V BE 3 so I E 2  I E 3 . The
transistor T2 is connected as a diode with the collector and the base shorted. It is
quite common to use transistors as diodes in integrated circuits. They occupy the
same silicon area and it is easier to achieve identical transistors than identity
between a diode and a transistor. The currents obey: IRB = IB2 + IE3. Since the
transistors are identical, they have the same  . So, we can write: IRB = (2 – α)IE2.
We finally obtain the following equation for the current IE2:
V V BE 2
I E 2  EE
(2   )R B

19
And of course: IC2 = αIE2 and we can use the value of 750 mV for VBE2. We can
remark that if α = 1 then IC2=IRB, so the circuit mirrors the current IRB which is
produced by a non ideal current source to the current IC2 which is the collector
current of T2 and as long as VCE2 is larger than VCEsat, the transistor T2 will
behave as an ideal current source. So, as long as the lower pair of transistors
behaves as a current source, we can replace it in the schematic and we obtain the
following schematics.
+VCC

RC

T1

vi -
CE
IC2
+

Fig.1-12 Biasing using a current source

The model of Fig.1-12 is the simplest one to analyze because the dc current
is fixed by the current source and it does not depend on a voltage drop across a
resistor.
It is interesting to compute the dc voltage stored in the capacitor CE. We are
going to show that it depends on the applied voltage vi. The base emitter voltage
v BE is the sum of the ac voltage vi and the capacitor voltage V DC .
v BE  v i V DC
When vi = 0, the capacitor is charged to:
kT I C 2
V DCQ  ln
q I ES

20
When v i  0 , the capacitor charges to a different value V DC . This is due to the
non linearity of the transistor that does not amplify in the same way the two
alternances of the ac signal. Since the circuit is non linear, we cannot analyze it
for a general signal, so we assume that the input ac signal is sinusoidal:
v i V 1 cos 0t
qv BE qV DC qV 1
cos 0t
i E  I ES e kT  I ES e kT e kT
qV DC
 I ES e kT e x cos 0t
qV 1
Where x  .
kT
We can remark that the output current is not sinusoidal.
iC (t )   i E (t )
v C (t ) V CC  RC i C (t )
The current i E is periodic and it can be developed in Fourier series.

e x cos 0t
 I 0 (x )  2 I n (x )cos n0t
n 1
I n (x ) is the modified Bessel function of the first kind, of order n and argument x.
We obtain finally:
qV DC
  
I 0 (x ) 1   n
2I ( x )
i E  I ES e kT cos n0t 
 n 1 I 0 (x ) 
Thus, the average value of the emitter current  i E  is:
qV DC
 i E   I ES e kT
I 0 (x )
And it must be equal to I C 2 , since it is the only dc current present in the circuit
and no dc current can flow in a capacitor. The fundamental and the harmonics of
iE will flow through the by-pass capacitor, so we have the following expression
for the emitter current:
 
i E  I C 2 1   n
2I ( x )
cos n0t 
 I 0 (x ) 

And we can compute the capacitor voltage from the expression of <iE>:

21
 kT IC 2 kT I C 2 kT
V DC  ln  ln  ln I 0 (x )
q I ES I 0 (x ) q I ES q
kT
V DCQ  ln I 0 (x )
q
We remark that this voltage depends on x, i.e. on V1.

1.4 The Field Effect Transistor.

The field effect transistor operation is not based on the transformation of
majority carriers into minority ones but on the modification of the conduction of a
channel by an electrical field. The field can be developed inside a depletion
region of a reverse biased PN junction (junction FET or JFET) or using the
electrical field developed inside a capacitor (metal oxide semi-conductor or
insulated gate FET, MOSFET, IGFET). In MOSFET, the channel can be depleted
as in JFET or it can be enhanced. The following figures show a simplified
structure of the different FET transistors along with the associated symbol.

Gate Drain
Source Drain

Gate
P

N Source

Fig.1-13 N-channel JFET structure and symbol.

Oxide

Gate
Source Drain
Drain

N+ N+
N
Gate Substrate
P

Source

Substrate

Fig.1-14 Depletion mode N-channel MOSFET structure and symbol

22
 Drain
Oxide

Gate
Gate Substrate
Source Drain

Source
N+ N+

Substrate

Fig.1-15 Enhancement mode N-channel MOSFET and symbol

The arrow in the different symbols indicates the conducting direction of the
PN junction (gate-channel or substrate channel). If the channel is of P type, then
the arrow should be reversed.
The different FET transistors have a half square law transfer characteristic
(ID, VGS). We start with the description of the N channel JFET transistor. It is
evident that the gate-channel diode should never be forward biased. So, in normal
use, the gate voltage should always be lower than the source voltage. A good
approximation for the transfer characteristic is:

2
 v GS 
i D  I DSS 1   V P  v GS  0
 VP 
 0 v GS V P

The above relation is valid for a drain source voltage v DS  V P . This

transfer characteristic is plotted below (Fig.1-16). IDSS is the drain saturation
current and |VP| is the pinch off voltage.
IDSS
Drain Current ID

Vp Gate Source Voltage Vgs

Fig.1-16 N channel JFET transfer characteristic

23
 The output curves (ID, VDS) of the FET transistor are also quite different
from the ones of the BJT. The following curves are ideal in the sense that the
output impedance of the FET transistor is assumed to be infinite.

iD
vGS= 0
IDSS

vGS<0

vDS
-VP
Fig.1-17 Output curves iD vs vDS

We can remark two distinct regions in the above curves. In the area
corresponding to vDS > -VP, the transistor behaves as a current source (controlled
by vGS). It is called the saturation region and it corresponds to the complete pinch
off of the channel. It is the normal operating region. The other region is called the
ohmic region. For small values of vDS, the resistance of the drain to source
channel is variable and is controlled by the gate to source voltage vGS.

For vDS limited to a few hundreds of millivolts, we can use the following
expression:
V P2
R DS  V P  v GS  0
2I DSS (v GS V P )

and we can use the JFET as a variable resistor in circuits. The above
relation is valid even if vDS is negative as long as it is small enough.

24
 R

vGS
vin T1 vout

Fig.1-18 Voltage controlled attenuator

The above circuit is an example of a voltage controlled attenuator. If we

assume zero source impedance and infinite load impedance, its transfer is given
by:
v out R DS

v in R  R DS
as long as |vout| is limited to few hundreds of millivolts. The smallest value
of RDS is usually called Ron when we use the FET as a switch. If the control
voltage vGS is a square wave varying between 0V and Vmin< VP and if R >> Ron,
then the above circuit can be used as a chopper (for modulators).

Biasing the JFET

Biasing the JFET means fixing the values of IDQ and VDSQ. The following
circuit provides a very simple way of achieving the above Q point.

25
 VDD

ZL

CL

vin
RG
RS CS

Fig.1-19 Biasing using a source resistor.

The above biasing circuit is based on the fact that the gate to source diode
is reverse biased (under normal operation) and that there is practically no current
that flows through the resistor RG. So, the biasing relations are:
2
 V 
I DQ  I DSS 1  GSQ 
 VP 
V GSQ  R S I DQ
And V DSQ V CC V GSQ if the capacitor CS is large enough so that its voltage
remains constant when iD varies and if there is no dc drop across the load ZL. We
can guarantee the above result if the impedance of the capacitor is much smaller
than the impedance connected in parallel which is RS in parallel with the inverse
of the small signal transconductance at the Q point ( 1 g mQ ) at the lowest
operating frequency. A graphical representation of the above relations is shown
below:

26
 i
D

IDSS
-VGSQ
I =
DQ R
S

I
DQ

VP V v
GSQ GS

Fig.1-20 Biasing using a source resistor

The RG resistor should be such that the gate voltage is as close to zero volts
as possible. You can find the maximum reverse gate source current in
manufacturers' data sheets. The main problem with the above circuit is the fact
that the Q point is highly dependent on the transistor parameters IDSS and VP. A
smaller variation of IDQ is provided by the following circuit (based on a higher
voltage across the resistor RS.

VDD

ZL
R1

CG RG

C1 R2 RS CS

Fig.1-21 Biasing a JFET using a source resistor and a gate voltage.

27
 R2
The voltage divider R1, R2 raises the gate voltage to VGG =VDD . At
R1+R2
that time, the biasing load line becomes VGG -VGSQ = R SIDQ which is represented in
the figure below. We can remark the small variation of IDQ when we change the
transistor.

i
D

I DSS1

-V + VGG
GSQ
I DQ =
RS

I
DSS2

IDQ

V V V v
P1 P2 GG GS

Fig.1-22 biasing a JFET using a source resistor and a gate voltage

The different MOS transistors obey a transfer characteristic which also

square law. However, the gate source voltage is not restricted to negative
voltages. The transfer characteristic of a MOSFET is thus:

i D   (v GS V th )2 v GS V th
iD  0 v GS V th
If the transistor is of depletion mode, the threshold voltage V th is negative
and for enhancement mode MOSFET, it is positive.
The biasing circuits for depletion mode MOSFET can be the same as the
ones described above for JFET. The value of the gate resistor can be much higher
since the input of the transistor is a capacitor and not a reverse biased diode. The
depletion mode MOSFET can even be biased with VGSQ = 0V because vGS can be
positive.
The biasing circuits of enhancement mode MOSFET will be the same as
the one shown in Fig.1-21. The biasing load line is the same as the one shown
in Fig.1-22, with the transfer characteristic translated to the right.

28
 i
D

-V + VGG
GSQ
I DQ =
RS

IDQ

V v
VGSQ GG GS

Fig.1-23 biasing an enhancement mode MOSFET

If we want to avoid any variation of the biasing current IDQ, we can use a
current source to provide the current.

29
 Chapter 2

Passive frequency dependent networks

and transformer like networks

Practically all CW communication circuits are band pass circuits. A typical

circuit consists of a cascade of active devices and passive networks. The passive
networks are used to couple circuits together, to select a particular band of
frequency around some carrier and to match impedances. The basic building
elements of these networks are: resistors, capacitors, inductors and transformers
(mutual inductance).

2.1 Impedance and Admittance

One port networks are characterized by a relationship between the voltage

across the device and the current through the device. If the device is composed of
resistors, capacitors or inductors, the relationship is linear and time invariant.
The resistor is characterized by a memoryless relationship:
v(t )  Ri(t ) between time domain variables, V  RI between phasors and
the same relation between Laplace transformed variables V (s)  RI (s) .
The capacitor and the inductor have a differential relationship:
di (t ) 1 t
v(t )  L and v(t )   i( )d  in the time domain, V  jL I and
dt C 
1 1
V I between phasors and V (s)  LsI (s) and V ( s)  I ( s) between
jC Cs
Laplace transformed variables. Each one of the devices is characterized by an
attribute: Resistance for the resistor, inductance for the inductor and capacitance
for the capacitor. When we consider relations between phasors, the current and
the voltage across the one port device is a complex number. It is a function of
either the complex number s    j or just the purely imaginary number j .

30
 The impedance is the complex number:
Z  j   R  j   jX  j  , the real part R  j  is called the resistance of
the circuit and the imaginary part X  j  is the reactance of the circuit. The
current and the voltage for a general one port network are related by:
V  j   Z  j  I  j 
The inverse relationship I  j   Y  j V  j  is provided by the
admittance:
Y  j   G  j   jB  j  . The real part G  j  is called the conductance
of the circuit and the imaginary part is called the susceptance of the circuit.
If the network is passive, the resistance (the conductance) of the network is
always positive. The one port network will be inductive if X  j   0
( B  j   0 ). The network will be capacitive if X  j   0 ( B  j   0 ).

Parallel series transformations

One port networks can be seen either as a series representation of a
resistance and a reactance or as a parallel combination of a conductance and a
susceptance.

Rs

Gp jBp Rp jXp
jXs

Fig.2- 1 Series parallel transformations

As shown in the above figure, we have three possible representations of the

same impedance. We can thus write the following relations:
Z  Rs  jX s
1 1
Y  G p  jB p  
R p jX p
1
Since Y  , we have the following equations:
Z
R X
G p  2 s 2 and B p  2 s 2 . These relations are often stated as a
Rs  X s Rs  X s
function of the quality factor of the impedance. The quality factor Q is defined as:
31
 Xs B R
Q  p  p
Rs Gp Xp
The "Que" of the circuit indicates how reactive the circuit is. A pure
reactance will have Q = .
1 Q 2
Gp  ; Bp  .
Rs 1  Q2  X s 1  Q2 
The above expressions are frequently given in terms of Rp and Xp:
X s 1  Q2 
Rp  Rs 1  Q  and X p 
2
. If Q 10, these relations simplify
Q2
to: Rp Rs Q2 and X p X s (with an error of less than 1%).
We can use the above relations to solve single frequency matching
problems as shown in the following example.
Example:
We want to match a 50  load to a source having a resistance of 5 k at
the frequency of 1 MHz. We can use the following network:
Rs 5k C

L R
50
V1

If we use the above approximate formula, we can transform the series

combination of C and R to parallel Cp and Rp.
Rp  Q2 R giving Q2 = 100 or Q = 10. This value of Q justifies the use of the
Bp
approximate relation. From Q  , we obtain: Bp  C  QGp (Cp = C). This
Gp
provides:
C = 318 pF, we use the standard value C = 330 pF. L is used to cancel the
1
reactance of C: L  giving L = 79 H. We can use this value if we build the
C
inductor or use the standard value L = 82 H.

2.2 Two port networks

In this chapter, we are going to study coupling networks. These networks
are used to filter signals and match impedances.

32
 Linear time invariant two port networks are characterized by matrices
relating the different variables. The variables represent small ac voltages and
currents.

i1 i2

v1 Two port v2
network

Fig.2- 2 Two port network

Any pair of variables from  i1 , i2 , v1 , v2  can be associated with the other pair
of variables. Two of the commonly used parameters are: the “z” (open circuit) or
impedance parameters and the “y” (short circuit) or admittance parameters.
The impedance parameters are described by the following set of equations:
v1  z11i1  z12i2
v2  z21i1  z22i2
or in matrix form:
 v1   z11 z12  i1 
v    z z22  
 2   21  i2 
v1
z11  is the input impedance for an open output port.
i1 i 0
2

v1
z12  is the reverse transimpedance for an open input port.
i2 i 0
1

v2
z21  is the forward transimpedance for an open output port.
i1 i2 0

v2
z22  is the output impedance for an open input port.
i2 i1 0

The two port network is said to be reciprocal when the open circuit voltage
measured at one port due to a current excitation at the other port is unchanged
when the measurement and excitation ports are interchanged.
In this case, z12 = z21. The proof is left as an exercise. A network containing
only R, L, C and M elements is always reciprocal.
The two port is said to be unilateral if the reverse transimpedance is zero:
z12 = 0. In this case, there is no feedback from the output port to the input one.

33
 The disadvantage of the impedance parameters is the fact that they are
difficult to measure. One port must be open during the measurement. The
parasitic capacitances will affect the measurement. This problem does not exist
with the admittance parameters since they are short circuit parameters.
The admittance parameters are described by the following set of equations:
i1  y11v1  y12v2
i2  z21v1  z22v2
or in matrix form:
 i1   y11 y12  v1 
i    y y22  
 2   21  v2 
i1
y11  is the input admittance for an shorted output port.
v1 v 0
2

i1
y12  is the reverse transadmittance for an shorted input port.
v2 v1 0

i2
y21  is the forward transadmittance for an shorted output port.
v1 v 0
2

i2
y22  is the output admittance for an shorted input port.
v2 v1 0

It is evident that the (Z) matrix is the inverse of the (Y) matrix and vice
versa. Since the elements of the (Y) matrix are easier to measure at medium and
high frequency, the "y" parameters are commonly used to design HF, VHF and
even UHF amplifiers. At higher frequencies, it is better to use distributed
parameters such as the scattering parameters.

2.3 Mutual Inductance and transformer

A commonly used two port network consists of two inductors coupled
magnetically: the transformer.
M

L1 L2

Fig.2- 3 Mutual Inductance

The winding corresponding to L1 is called the primary, the other is the
secondary. The dots indicate the way the transformer is wound. The coupling is
due to the mutual inductance M. With the dots positioned as in Fig.2- 3, the

34
mutual inductance is positive. If the two coils are wound in opposing direction,
the value of M will be negative and the dots will be drawn on opposing ends of
the windings.
The basic equations are:
di di
v1 (t )  L1 1  M 2
dt dt
di di
v2 (t )  M 1  L2 2
dt dt
If we consider the transfer of energy in the transformer, we can show
M
that M 2  L1L2 . We can define a coupling coefficient k  . It is evident
L1L2
that k  1 . The sign of k is indicated by the dots positions. When |k| = 1, the
magnetic coupling is total. All the magnetic flux generated by one winding flows
inside the other one. An extreme case is the ideal transformer.
I1 1 : n I2

V1 V2

ideal
Fig.2- 4 Ideal Transformer
The direction of currents in Fig.2- 4 is non conventional for two port
networks. However, it simplifies the analysis of the circuit. n is called the turn
ratio. The basic equations for the ideal transformer are: V2  nV1 and the
conservation of power: V1I1  V2 I 2 . Consequently, I1  nI 2 . If we connect a load Z
V V V I Z
at the secondary, it will appear at the primary as 1  1 2 2  2 . This relation
I1 V2 I 2 I1 n
can be used for matching impedances. We can also remark that the ideal
transformer does not transform the type of impedance. If Z is resistive, its
transform remains resistive. The same thing results for inductive and capacitive
loads. In many cases, the direct use of the circuit of Fig.2- 3 is not very commode.
The following one is more useful.

La
n : 1
Lb

ideal
Fig.2- 5 Equivalent circuit

35
 L1
In the above circuit, La = (1  k²)L1, Lb = k²L1 and n  k . We can
L2
remark that n has the same sign as M. Furthermore, the circuit shows that a
physical transformer is equivalent to an ideal transformer if we have total
coupling (k² = 1) and infinite (very large) inductances for the primary (and the
L
secondary) winding. When we have perfect coupling, n   1 . We know that
L2
the value of an inductance is proportional to the square of the number of turns of
its windings. The constant of proportionality depends on its physical size. Since
the primary and the secondary are both wound on the same core, we can write
L1   n12 and L2   n22 , where n1 and n2 are respectively the number of turns of the
n
primary and the secondary. This means that n   1 and this justifies the name
n2
"turn ratio".

2.4 Parallel RLC circuit

In this section, we are going to analyze a one port network commonly used
as a frequency selective filter. This circuit is composed of a resistance R, an
inductance L and a capacitance C in parallel.

L C R
ii(t) v0(t)

Fig.2- 6 Parallel RLC circuit

The input of the above circuit is a current source. It may represent the
collector current of a bipolar transistor for example. The output is the voltage
across the parallel RLC circuit1. This means that the transfer function is the
impedance of the circuit.
1 1 s
Z (s)  
1 1 C 2 s 1
 Cs  s  
Ls R RC LC
1 1 
Let 02  and 2   0 . The transfer function becomes:
LC RC QT

1
Commonly called "Tank" circuit.
36
 1 s
Z ( s) 
C s  2 s  02
2

This transfer function possesses one zero at the origin and another one at
infinity. It has also two poles: p1,2     2  02 . The poles can be real or
complex. Since we use the tank circuit as a narrow bandpass filter, we will
1
consider only the complex case, i.e. 0   or QT  . In this case, we can express
2
the poles as: p1,2    j     j 02   2 . Replacing  by its value, we obtain:
0 1 
p1,2    j0 1  2
. If QT > 10, the poles become p1.2   0  j0 with
2QT 4QT 2QT
a very high precision. In this event, these poles will be also very close to the j
axis.
We can remark that the poles satisfy 02   2   2 . Given that   0, the
locus of the poles in the complex plane is a quarter of a circle on the top left
quarter of the plane for one pole and the symmetrical one with respect to the real
axis for its conjugate.


1
SF =
C

 0

Fig.2- 7 Pole and Zero Plot

The pole and zero plot is a plot of the location of the poles (indicated by )
and the zeros (indicated by ) along with the value of the scale factor (SF) for a
transfer function H(s):
H ( s)  SF
 s  z1  s  z2   s  zm 
 s  p1  s  p2   s  pn 

37
 SF

lp1

p1
p1

lz1
z1
z1

Fig.2- 8 Typical Pole and Zero Plot

The pole and zero plot allows a graphical evaluation of the transfer function
in the frequency domain. The transfer function in the frequency domain is the
evaluation of H(s) for s on the j axis:
( j  z1 )( j  z2 ) ( j  zm )
H ( j )  H ( s) s  j  SF (1)
( j  p1 )( j  p2 ) ( j  pn )
The modulus and phase versus frequency are given by:
j  z1 j  z2 j  zm
H ( j )  SF
j  p1 j  p2 j  pn
l l l
H ( j )  SF z1 z 2 zm (2)
l p1l p 2 l pn
m n
and Arg  H ( j )  Arg  SF    Arg  j  zi    Arg  j  pk 
i 1 k 1
m n
Arg  H ( j )  Arg  SF   zi   pk (3)
i 1 k 1

where lzi is the length of the vector joining the zero zi to the point of
coordinate  on the imaginary axis while zi is the angle that this vector makes
with the real axis. lpk is the length of the vector joining the zero pk to the point of
coordinate  on the imaginary axis while pk is the angle that this vector makes
with the real axis as shown on Fig.2- 8.

In our case, the existence of a zero at the origin and another one at infinity
imply that the value of |Z(j)| at  = 0 and at  =  is zero. This means that the

38
system is bandpass and |Z(j)| is maximum at some frequency. In fact, the pole
and zero plot of Fig.2- 7 shows that this frequency is 0. We have also
|Z(j0)| = R (at this frequency, the susceptance of L and C cancel each other). In
our case, the simple shape of Z(s) allows a straightforward algebraic evaluation of
Z(j).

R
Z ( j ) 
 2  02
1  jQT
0
R

R
2
|Z(j

  
9π/20

2π/5

7π/20

3π/10

π/4

π/5
Arg[Z(j
3π/20

π/10

π/20

 
-π/20

-π/10

-3π/20

-π/5

-π/4

-3π/10

-7π/20

-2π/5

-9π/20


Fig.2- 9 Modulus and Phase Response of the Tank circuit

We remark that the impedance is resistive at  = 0 (its value is R); it is

inductive below 0 and capacitive above. From Fig.2- 9, we observe clearly the
bandpass nature of the transfer function. The above expression can be used to
compute the "3 dB" bandwidth of the circuit. The bandwidth B is defined as the
difference between the two "3 dB" cutoff frequencies: B = 2  1. These two
frequencies are the ones for which |Z(ji)| = |Z(j0)|/ 2 , i = 1,2. Using the
previous equation and solving for 1 and 2, we obtain:
39
 0
0  12 and 2  1  B 
QT
The resonant frequency is the geometric mean of the two cutoff frequencies
(it is not in the middle). This result shows clearly that the transfer function is not
symmetrical with respect to the resonant frequency. However, when QT is high,
the two frequencies will become close to each other and the geometric mean will
be close to the arithmetic one. This narrow band approximation can be obtained
using the technique of graphical evaluation of transfer functions seen previously.
Using the pole and zero plot of Fig.2- 7 and equations (2) and(3), we
obtain:
1 lz 0
Z ( j )  and Arg  Z ( j )  z 0   p1   p 2
C l p1l p 2
When we evaluate the above equations for  around 0, we obtain:
lz0 =   0, lp2  20 because  is very small. So:
1 0 1
Z ( j )  
C 20l p1 2Cl p1
We have also: z0 = /2, p2  /2, giving: Arg  Z ( j )   p1 . Grouping
the two results, we finally get:
1
Z ( j )  j
2Cl p1e p1
j p1
The complex number l p1e is the vector connecting the pole p1 to the point
j in the complex plane. So,
j
l p1e p1  j  (  j  )    j (   )    j (  0 ) .
This implies that we can approximate Z(j) very closely by:
1 1 1 R
Z ( j )    (4)
2C   j   0  2C 1  j   0 1  j 2Q   0
 T
0
The above result is valid for positive frequencies around 0. For negative
frequencies, we can use the fact that Z(j) = Z*(j). Fig.2- 10 (for QT = 10,
0 = 5 rd/s and R = 10 ) shows that the two responses are very close. In fact, the
approximation is very accurate for all frequencies between the cutoff frequencies
1 and 2. A closer look at the approximation shows that the curve is
symmetrical with respect to the line  = 0. In fact, the resonant frequency is now
the middle of the interval [ 1, 2]. Equation (4) is commonly used when we want
to obtain the equivalent lowpass filter for the analysis of bandpass signals.
40
 5
|Z(j)|
4.5
Exact
Approximate
4

3.5

2.5

1.5

0.5


Fig.2- 10 Exact and Approximate Amplitude response

2.5 Series Resonant Circuit

When the signal source has low impedance, it is usually modeled as an
ideal voltage source. At that time, we cannot use a parallel tank circuit since any
circuit in parallel with an ideal voltage source will have no effect. A series
resonant circuit corresponds to the circuit shown below.

L
C

v
R

Fig.2- 11 Series Resonant Circuit

We don’t have to repeat the analysis for the above circuit. By using the
principle of duality, we can immediately derive the admittance function of the
circuit. We simply replace L by C, C by L, R by G, Z by Y and i by v.
The parallel tank circuit had an impedance function given by:
1 s
Z ( s) 
C s2  s  1
RC LC
The application of the duality principle gives:
1 s 1 s
Y (s)  
L s2  s  1 L s2  R s  1
LG LC L LC

41
 Having the same transfer function, the analysis of this circuit will be
identical to the one derived for the parallel RLC circuit and is left as an exercise.

2.6 Parallel Resonant Circuit with Series Loss

If one of the reactive elements in a parallel connection of an inductor and a

capacitor possesses series loss, the zeroes of the impedance function are going to
move. If the loss is in series with the capacitor, the zero at infinity moves to a
finite value.

L
r

Fig.2- 12 Parallel Resonant Circuit with Series Loss

The impedance transfer function in this case is given by the parallel

combination of the inductance L and the series arrangement of the capacitance
and resistance. So:

 1   1 
Ls  r   rs  s  
Z ( s)  
Cs   rC 

1 r 1
Ls  r  s2  s 
Cs L LC

r 1 1
Let   , 02  and QC  . The impedance transfer function
2L LC rC0
1
has two real zeros: at the origin and at  . It has two poles: at
rC
  j     j 02   2 .

42
 l p1
SF = r

l z1 lz 0

lp2

1
rC


Fig.2- 13 Pole and Zero Plot

According to the above pole and zero plot, if  is small, the two poles will
be complex and their imaginary part  will be practically 0. A simple
0
manipulation shows that   . The fact that this transfer function has two
2QC
zeros and two poles implies that the value of the impedance at infinity is different
from zero. In fact Z ()  r . From the pole and zero plot, we also have:
rl l
Z ( j0 )  z 0 z1
l p1l p 2

1 r
along with: lz 0  0 , lz1  22
 02 , l p1  and l p 2  20 . This gives:
rC 2L
Z ( j0 )  L0 1  QC2 . We also have L0  rQC . So, Z ( j0 )  rQC 1  QC2 . If
QC  10 , Z ( j0 )  rQC2 . It is also apparent from the same pole and zero plot that
Arg  Z ( j0 )  0 if QC  10 . Then, we can say that if QC  10 , the impedance of
the circuit at 0 is resistive and is equal to:
R  rQC2 (5)
For all frequencies around 0 , the circuit is equivalent to a parallel RLC
circuit. Equation (5) can be used for impedance transformation (matching). For
the high Q case and if r is small, we can say that most of the time |s| can be
1
neglected compared with . The transfer function becomes:
rC
43
  1 
rs  
Z ( s)  2   2 
rC 1 s
s  2 s  0 C s 2  2 s  02
1 1
Using (5), we can see that QC   RC0 and that   . This is
rC0 2RC
the same as the one of the parallel RLC circuit.
If now the loss is in series with the inductor, the same analysis can be
repeated, but now it is the zero at the origin that moves to a real negative value.

Fig.2- 14 Inductor with series loss

1  r
 r  Ls  1  s  L 
Z ( s)  Cs 
1 r 1
 r  Ls C s 2  s 
Cs L LC
L
In this case also, if QL  0  10 , then we can replace the series circuit
r
composed of r and L by a parallel one composed of L in parallel with R  QL2r .
Both circuits can be used for impedance matching. The value of Q is used to
transform a small resistive load to a large apparent load. So, in this part, we find
the same results as the ones derived in the series parallel transformations. To see
an example of application, refer to the example in page 31.

2.7 Transformer like networks

The matching networks described above allow impedance transformation at
a single frequency. However, their frequency response around that frequency is
not very accurate. They are bandpass, but since they are built around only two
components (a capacitance and an inductance), we have only two degrees of
freedom. We can set the resonant frequency and the impedance transformation.
The impedance transformation is due to a multiplication by Q². This fixes the
value of the bandwidth. In general, we need circuits that have three degrees of
44
freedom if we want to set three quantities: The impedance transformation, the
resonant frequency and the 3 dB bandwidth. One solution is to use a tank circuit
in cascade with an ideal transformer.

1 : n

L C G

ideal
Fig.2- 15 Ideal transformer matching

In the above circuit the conductance is reflected to the input of the

transformer to a value equal to n2G . The resonant frequency is given by
1 C
02  and the quality factor is Q  20 . So, we have three elements that can
LC nG
be adjusted in order to set the above cited three parameters. The ideal transformer
can be approximated by a transformer wound around a toroidal core such as the
ones illustrated below.

Fig.2- 16 Toroidal core transformers

The problem with these transformers is that they are rather bulky and if we
have to use them, they are quite far from ideal. We are going to analyze circuits
that have the same behavior but that do not use ideal transformer.
The first system is a split capacitor network.

45
 1

C1

L
G
C2

Fig.2- 17 Split capacitor matching network

Under circumstances that will be stated later, this circuit is equivalent to the
one shown in Fig.2- 15. To show this equivalence, we have to show that the two
circuits have the same input impedance (loaded) and the same transfer function.
The input impedance of the circuit is given by:
 G 
s s  
1 1  C1  C2 
Z11 ( s )  
1 1 C s3  G s 2  1  G

Ls 1 1 C2 LC LC1C2

C1s C2 s  G
CC
where C is the series equivalent capacitance of C1 and C2: C  1 2 .
C1  C2
G
We see that Z11(s) has two finite real zeroes (at the origin and at  )
C1  C2
and three poles. One pole is always real; the other two can be either real or
complex conjugate. In order for the system to be narrow band bandpass, the poles
must be complex at   j  . The denominator can then be written as:
 s    s    j  s    j 
The pole and zero plot is shown in Fig.2- 18. If we want to have
equivalence with a parallel RLC circuit, we must have a cancellation between the
G
real pole at  and the real zero at  . To find conditions for this to occur,
C1  C2
let us identify the development of the denominator in terms of ,  and  and the
one in terms of G, L, C1 and C2.

46
 1
 SF 
C
 o
G 
C1  C2
 o o o




Fig.2- 18 Pole and Zero Plot

Developing the denominator, we obtain:
s3    2  s 2   2   2  2  s    2   2 
Identifying the coefficients in the two polynomials produce the following
three equations:
G
  2 
C2
1
 2   2  2   02
LC
  2   2  
G
LC1C2
We know that in the impedance function of the parallel RLC circuit, the
root locus of the complex poles is a circle of radius 0, i.e.  2   2  02 . So, we
can rewrite the second equation as follows:
 2 
 2   2  02 1  2  in order to be able to appreciate the precision of
 0 
02
the approximation (in percent). Let us call   . The three equations are
2
then:

47
  1 
 1
n 
2
nG
2   
C  1 1 
  
 1
 2   2  02 1  
 
G 1

C1  C2 1  1

C1
where n  .
C1  C2
G
Now, if  > 100, we have   with an error that is less than 1% and
C1  C2
of course  2   2  02 . The real part of the poles satisfies:
n 2G  1 
2  1  
C  n 
Even if  is larger than 100, we cannot eliminate the product n because n
is less than 1. So, if  is larger than 100, we have the pole and zero cancellation
and the input impedance will be the one of a parallel RLC circuit.
1 s
Z11 ( s) 
C s  2 s  02
2

So, if we can control the value of , we can find conditions for the
equivalence. The problem is that  does not correspond to physical quantities. In
order to have criteria that depend directly on the circuit elements, let us develop
the expression of .
 1  1
C 1    C1  C2  1  
 2
   
  0  02
2  1  G
n 2G  1  
 n 
We can introduce two “Que’s”:
C   C  C2 
QT '  20 and QE  0 1 .
nG G
QT’ corresponds to the Q of a tank circuit composed of the capacitance C in
1
parallel with the inductance L and in parallel with RT '  2 . QE on the other
nG
hand corresponds to the Q of the parallel combination of G with C1 and C2.

48
 The expression of  becomes:
2
 1
QT 'QE 1   2
  1  1
 giving    QT 'QE 1  
 1  n  
1  
 n 
1
So if  is large, then we obtain:   QT 'QE  .
n
The above expression shows clearly that, if QT 'QE  100 , then  will be
even larger. At that time, from the input impedance point of view, the circuit
becomes equivalent to a resistance RT in parallel with C in parallel with L where:
1  1 
GT   n2G 1   (6)
RT  nQ Q
T' E  1 
1
However, if nQT 'QE  100 , then GT   n 2G .
RT
Under the above condition, the circuit is then equivalent to the circuit
shown in Fig.2- 15. To have the complete equivalence, we must show that the
voltage transfer of the circuit is the one of the ideal transformer. In other words,
the voltage transfer of the circuit must be H (s)  n .
H(s) is the transfer of the following voltage divider:

C1

v1
C2 G

v2

C1 s
H ( s) 
C1  C2 s  G
C1  C2
Evaluating the above expression on the imaginary axis provides:

49
 
j
0
H ( j )  n
 1
 j
0 QE
So, if QE > 10, we can safely say that H ( j )  n at all frequencies. To
resume the different approximations, we can say that if QT 'QE  100 , the split
capacitor circuit is equivalent to a parallel tank circuit loaded by a resistance
given by equation (6). If we have nQT 'QE  100 , the circuit is equivalent to the
circuit shown in Fig.2- 15, but only from the input impedance point of view. If we
add the condition QE  10 , the equivalence becomes complete and we can replace
the circuit of Fig.2- 17 by the one of Fig.2- 15. This approximation will greatly
simplify the analysis of circuits.
In the next circuit, we use a voltage divider built by means of two
inductors.
2

L1

1
i C

L2
G

Fig.2- 19 Split Inductance Circuit

In the circuit shown above, we assume that there is no magnetic coupling

between the two inductances. We don’t have to redo the same analysis for this
1
circuit. The tank circuit resonant frequency is 0  . L is the inductance of
LC
the series connection of the two inductances L1 and L2, L  L1  L2 . The ideal
transformer turn ratio n is the voltage transfer of the circuit with no load (G = 0).
L2
So, n  . Applying the definitions of QT’ and QE given previously we get:
L1  L2
C
QT '  20 (Q of a tank circuit composed of the capacitance C in parallel
nG
1 L  L2
with the inductance L and in parallel with RT '  2 ) and QE  1 (Q of the
nG L1L20G
parallel combination of G with L1 and L2).

50
 If there exist magnetic coupling between the inductors, we obtain a
transformer.

M
4 1
 
i
L1 L2
C G

3 2

Fig.2- 20 Tuned Primary Transformer

Using the equivalence derived in (2.3 Mutual Inductance and transformer),

we can replace the transformer composed of L1, L2 and M by the equivalent
circuit shown in Fig.2- 5, we obtain the following circuit.
(1  k ) L1
2

4 1 a : 1 2

i
2
C k L1 G
3 5

ideal

G
It corresponds to the split inductor circuit loaded by a conductance ,
a2
where a is the turn ratio of the ideal transformer. The coupling coefficient k is
M L
given by k  and the turn ratio a is given by a  k 1 . The turn ratio of
L1L2 L2
the ideal transformer corresponding to the split inductor is
k 2 L1
n'   k 2 and the series combination of the two inductances is
(1  k ) L1  k L1
2 2

L = L1. The final equivalence is a cascade of two transformers corresponding to a

n' L M
single transformer with turn ratio n   k 2 . Replacing k  we obtain
a L1 L1L2
M C C 1
n ; QT '  0  0 and QE  .
L1 n '²G ' nG L1 (1  k 2 )n 2G0
If the transformer is tightly coupled ( k  1), QE will be very large and the
equivalence with the ideal transformer is always valid. The above analysis can
also be extended to the autotransformer.
The following table is taken from the textbook and summarizes the
different conditions for equivalence for all circuits.

51
 Except for the transformer based circuits, the previous networks are step
down. The equivalent turn ratio is always less than one. The next circuit, on the
other hand, is step up. In fact, it is a split capacitor circuit used in reverse.

C1

L G
C2

Fig.2- 21 Step Up Transformer

52
 The input impedance is:
1
Z in 
1
C2 s 
1 1

C1s G  1
Ls
After simplification, we obtain:
G 1
s2 
s
1 C1 LC1
Z in 
C2 s s 2  G s  1
C LC
CC 1 1 G G
with C  1 2 . Replacing 02  , 12  ,  and 1  ,
C1  C2 LC LC1 2C 2C1
we obtain:
1 s 2  21s  12
Zin 
C2 s s 2  2 s  02
There are two finite zeros and 3 poles:
z1,2  1  j 12  12 and p0  0 , p1,2    j 02   2
We assume that the zeros and the poles are complex. We obtain the
following pole & zero plot:
l
 1
SF =
C2

0

1

1

An exact analysis (like the previous case) cannot be performed. The next
analysis is thus valid for the high Q case.

53
 Since C  C1 , we have 0  1 and   1 . If we assume that we operate
only around 0 , then, if  is very small, we can write:
1 0  1 0  1  1 02  12 1
Zin and Arg  Zin   giving: Zin
C2 0  20  l 2C2 02 le j
lej is the vector represented in the above figure joining the point
  j0 to the point j in the s-plane. Using the definition of 0 and 1 , we
obtain:
02  12 C C1 C C
 1  . Let n  , then C  and C  and finally:
02 C1 C1  C2
2
n
1
1 n
02  12 n 1
 n . So: Zin  .The final result is:
02
2C2
1 j
   0 

n2 1
Zin 
G
1 j
  0 

and for frequencies  around 0 , the circuit is equivalent to the following
figure for the input impedance point of view.
n:1

L C G

ideal
Fig.2- 22 Step Up equivalent circuit

To obtain a complete equivalence, the voltage transfer of the circuit must

be the one of the ideal transformer. The voltage transfer is given by:

s2
H ( s) 
s 2  21  12
Under the same previous hypotheses, we obtain:
02 1
H  and Arg  H  0 for frequencies around 0 .
02  12 n
This approximate analysis is valid for bandlimited signals and
C
for Q  0  10 . The bandwidth of the signals must be much smaller than
G
2 0  1  in order to remain always in the vicinity of the pole.

54
 The final transformer like network is the “pi” circuit. This circuit is
commonly used in power amplifiers. We will not make a complete analysis of the
circuit. The advantage of the “pi” circuit is that it can be used either as a step up
or as a step down transformer.
L
2
1

C1 C2 G

Fig.2- 23 The Pi Network

In this case also, we are going to study the input impedance and the transfer
function. The input impedance is given by:
G 1
s2  s 
1 1 C2 LC2
Z11 ( s)  
C1s 
1

1 C1 s 3  G s 2  C1  C2 s  G
Ls C2 s  G C2 LC1C2 LC1C2
CC 1 1 C C
Introducing: C  1 2 ; 02  ; 22  ; Q2  0 2 and N  2 ,
C1  C2 LC LC2 G C1
the input impedance is expressed as:
0
s2  s  22
1 Q2
Z11 ( s ) 
C1 0N002
s  s  s 
3 2 2
0
Q2 Q2
This function has two finite zeroes. If Q2 is large enough, these zeroes are
complex:
0 02
z1,2   j   2
2 .
2Q2 4Q22
It has three poles: at  and at   j  .
Using the same technique as for the split capacitor, we obtain the following
three equations:

55
 0
2   
Q2
 2   2  2  02
N022
      Q
2 2

02
In this case also we introduce the variable   . The above three
2
equations become:

N 1
0 1
2  
 N  1 Q2 1  1

 1
 2   2  02 1  
 
N22 1

Q20 1  1


If   100 , we obtain the following equations:

 2   2 02
N22

Q20
0  N 1
2 1  
Q2  N  1   
Replacing the expressions of  and  in the definition of , we obtain:
02Q22 ( N  1)  1 
2

  ( N  1)  1  
22 N  
22 1
We also have 2  . So, when  is large, we obtain:
0 N  1
 N  1
2

 Q22  N  1
N
N 1 
So, if Q2  10 ,   100 . Furthermore, if Q2  10 ,  100 . At
N N 1
that time:

56
  2   2 02
N22

Q20
0
2
Q2  N  1
In order to progress in our analysis, we must resort to narrowband analysis.
It is not possible to obtain a more general result. The following figure shows the
pole and zero plot.

l
 1
SF =
C1

0

2

2

Using the above plot, we obtain: for  around 0

1 0  2 0  2  1 02  22
Z11 ( j )  
C1 202l C1 202l
The argument is practically . Given that
le  j     j0     j   0  , the expression of the impedance becomes:
j

1 02  22
Z11 ( j ) 
C1 202   j   0  
02 0
Replacing   2
and 2  , we obtain the following
N 1 Q2  N  1
2

expression:
NQ2 1
Z11 ( j ) 
C1 1  j   0

57
 0C2
Since Q2  , the final result is:
G
N2 1
Z ( j ) 
G 1  j   0

So, from the input impedance point of view, the impedance is the same as
N2
the one of a parallel tank circuit loaded by a resistance of value .
G
The transfer function is the transfer of the following voltage divider:
3 1

C2 G

The transfer function is:

1
C2 s  G 22
H (s)  
1 
Ls  s 2  0 s  22
C2 s  G Q2

Using a narrowband approximation and operating at a frequency around

0 , we obtain from the following pole and zero plot:

SF = 22

2

2

58
 22 22 1
H ( j )   2 
0  2 0  2  0  2 N2

 
and arg  H ( j )    
2 2
N 1
So for the stated conditions: Q2  10 and narrowband signals
N
operating around 0, the pi network is equivalent to:

1
N : 1 2

L C G

3 4

ideal

Response of a tank circuit to a periodic input

If we apply a periodic signal to a parallel RLC circuit, its response will be
also periodic, but with harmonics that will be much reduced. Consider the
following periodic current:

i(t )   I n cos n0t
n 0

applied to a parallel RLC network tuned at 0. The voltage across the tank
circuit is given by:

v(t )   Z ( jn0 ) I n cos  n0t  arg  Z ( jn0 ) 
n 1

If the Q of the tank circuit is larger than 10, we can use the pole zero plot
represented in Fig.2- 24 to evaluate the impedance of the parallel RLC circuit. We
obtain:
1 lz 0 n0 n
Z ( jn0 )   
C l p1l p 2 C (n  1)0 (n  1)0 C0 (n2  1)

and arg  Z ( jn0 )   for n > 1
2
along with Z ( j0 )  R . So:

nI n  
v(t )  RI1 cos 0t   cos  n0t  
n  2 C0 ( n  1)  2
2


nI n
 RI1 cos 0t   sinn0t
n  2 C0 ( n  1)
2

59
 1
SF =
C

l p1  (n  1)0

lz 0  n0

0

l p 2  (n  1)0

Fig.2- 24 pole zero plot for the harmonics

The above relation allows us to evaluate the distortion of the waveform

before and after filtering.
Before filtering, the distortion coefficient is:

I n2
Dbefore   2  100%
n  2 I1

And after filtering, we have:

2

 n  I n2
Dafter    2  100%
n  2  RC0 ( n  1)  I1
2

60
 Chapter 3

Non Linear Controlled Sources

Communication circuits are built using passive matching networks and

active devices. Most active devices act as voltage control current sources (BJT in
Common Emitter or in Common Base, FET in Common Source or in Common
Gate, etc). If the input voltage exceeds few millivolts, their transfer becomes
highly nonlinear. In this chapter, we assume that the nonlinearity is memoriless.
This simplifies the analysis of the different elements but it restricts the frequency
range (in general, we cannot exceed the VHF band).
When we analyze nonlinear circuits, we cannot make use of the
superposition theorem. This means that an analysis made for one type of
waveform cannot be generalized to a linear combination of these waveforms. In
our analysis of nonlinear controlled sources, we are going to restrict our analysis
to systems excited by sinewaves or square waves. The output signal will be
periodic with the same fundamental frequency as the input signal. Furthermore,
since the nonlinear element is assumed memoriless, then, if the input is a sum of
cosines (even function), the output will have the same parity and will also be a
sum of cosines. The extreme waveforms that we will consider are the periodic
train of impulses and the square wave.
The periodic train of impulses is useful in modeling very narrow current
pulses.

Consider the current i(t )  q   t  nT0  . We have already encountered
n 0

such signal in a previous course. It is periodic (period T0) and can be developed in
Fourier series. The coefficients of the series are all equal and the Fourier series is:

q     

i(t )   e jn0t  1    e jn0t  e jn0t   I dc 1  2 cos n0t 
q
(7)
T
n  0 T0  n 1   n 1 
We remark that the fundamental and the harmonics have a peak amplitude
equal to twice the dc current.
The other extreme case is the current that switches between a peak value Ip
and zero with a period T0 (square wave).

61
  T T
I p  0 t  0
i (t )   4 4

 0 in the remaining part of the period
In this case, the Fourier series is:
I 2I 2I 2I
i(t )  p  p cos 0t  p cos30t  p cos50t  (8)
2  3 5

3.1 Piecewise linear characteristic

The first nonlinear source that we will consider is the piecewise linear one.
It provides an adequate model for many MOSFET power amplifiers. This model
is also encountered when we consider the effect of series resistance in many
nonlinear amplifiers. Let us consider the following voltage controlled current
source.

v1 i2 = f(v1)

Fig.3- 1 Voltage Controlled Current Source

The transfer characteristic of the above controlled source is:

G  v1  V0  v1  V0
i2  
 0 v1  V0
Graphically, this relation is represented by:

i2

V0 v1

Fig.3- 2 Piecewise Linear Characteristic

The voltage V0 is a threshold voltage. The slope G of the transfer function

is a transconductance. The input signal is a sum of a biasing voltage Vb and an ac

62
signal v(t): v1 (t )  Vb  v(t ) . If the whole signal has a value larger than the
threshold voltage, the transfer is incrementally linear. This means that the output
current is expressed as:
i2 (t )  Ib  i(t ) along with Ib  G(Vb  V0 ) and i(t )  Gv(t ) .
We remark that the ac signals are linearly related. However, we do not have
the same relation between the dc signals. When the whole ac signal is amplified,
we say that the amplifier is operating in “class A”.
Another simple analytic case occurs when the biasing voltage Vb is equal to
the threshold voltage V0. This corresponds to “class B” operation. In this case,
only the positive half of the ac signal is amplified and the operation is completely
nonlinear. This implies that we cannot use an arbitrary ac voltage as input. Let us
assume that this signal is sinusoidal: v(t )  V1 cos 0t . The output current is a half
rectified sinewave as shown below.

i2(t)

Ip = GV1

  t
Fig.3- 3 Class B Output
We can develop the above signal in Fourier series. The result is:
I I 2I 2I
i2 (t )  p  p cos 0t  p cos 20t  p cos 40t 
 2 3 15
and I p  GV1 .
GV1
We remark that the dc output current is . Its value depends on the

amplitude of the input signal. This means that if the device is biased using a
current source, we must adjust the value of the biasing current every time the
input voltage changes in order to maintain the biasing voltage at the value
Vb = V0.
Another point worth taking into account is that the peak value of the
fundamental current is proportional to the input voltage. If we load the output
circuit with a parallel RLC circuit tuned at the fundamental and with a Q high
enough so that the harmonics are practically eliminated, the output voltage will
63
 RI p RG RG
be: v0 (t )  cos 0t 
V1 cos 0t  v(t ) . This means that the output
2 2 2
voltage is proportional to the input one. The amplitude of the output is
proportional to the amplitude of the input. Thus, an amplifier biased in class B
can be used to amplify “linearly” an amplitude modulated signal. The device
amplifies half of the waveform and the tank circuit recovers the other half. If the
signal is modulated (AM, VSB, SSB, QAM), we must make sure that the
bandwidth of the parallel RLC circuit is wide enough to let the modulation pass
without affecting it adversely.
If now the biasing voltage Vb is different from the threshold voltage V0, but
the ac signal does not pass completely, we have to define a “conduction angle”.
In this case also, we study the case of a sinewave drive v(t )  V1 cos 0t .

i2

Vx Ip

Vb V0 v1 -    0t

V1

Fig.3- 4 Class C Biasing

In the above figure, we define Vx  Vb  V0 and I p  G V1  Vx  . Using the

variable   0t , we remark that the controlled source produces an output current
for      , i.e. for a “conduction angle” equal to 2. The angle  is given by:
V 
  cos 1  x 
 V1 
64
 Using the above definition of a conduction angle (2 ), we can give now a
precise definition of the different classes of amplification:
 Class A: Conduction angle = 2 = 360°. The whole sinewave passes
through the system.
 Class B: Conduction angle = 180°. Only the positive half of the
sinewave is amplified.
 Class AB: Conduction angle is such that 180  2  360 .
 Class C: Conduction angle is such that 0  2  180 . Only a small
tip of the sinewave is amplified.
If we consider the voltage Vx, the class C corresponds to Vx > 0, i.e. Vb < V0
while class AB corresponds to Vx < 0, i.e. Vb > V0. The output current is not
sinusoidal and can be developed in Fourier series.
The system is memoriless and since the input is an even function, the
output current is also even and its Fourier series is then a sum of cosine functions.

i2 (t )   I n cos n0t
n 0

The computation of the Fourier coefficients is simplified if we use the

variable   0t . We have:
1  
 2   
 0  0   0 2  0 
I0  i2   d and I n  i   cos n d

The expression of the output current in the interval   ,  is just a shifted

sinewave:
i2 (t )  G V1 cos 0t  Vx  for     0t   . So:
 
i2    G V1 cos  Vx 
 0 
Replacing in the expression of the Fourier coefficients, we obtain 2:
I sin    cos
I0  p
 1  cos 
I   cos sin 
I1  p
 1  cos 
2 I cos sin n  n sin  cos n
In  p n2
 n  n2  1 1  cos  
These currents are given below (Fig.3- 5) as a function of the conduction
angle (2). The next figure (Fig.3- 6) provides the same information, but with the
2
See Appendix to Chapter 4 of textbook for details.
65
 V 
normalized voltage offset  x  as argument. They are also plotted in your
 V1 
textbook in pp.94-95, Fig.4.2-3 and Fig.4.2-4 but with a logarithmic scale.

0.6

0.55

0.5
(I1/Ip)
0.45 (I0/Ip)
0.4

0.35

0.3

0.25

0.2

0.15 (I2/Ip)
0.1

Conduction Angle (2 )

0.05

-π/5 -π/10 π/10 π/5 3π/10 2π/5 π/2 3π/5 7π/10 4π/5 9π/10 π 11π/10 6π/5 13π/10 7π/5 3π/2 8π/5 17π/10 9π/5 19π/10 2π 21π/10 11π/5

Fig.3- 5 DC, Fundamental and Second Harmonic for Sine Tips

0.5

0.45
I1/Ip

I0/Ip 0.4

0.35

0.3

0.25

0.2

I2/Ip 0.15

0.1

0.05

-1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1

Normalized voltage offset (Vx/V1)

Fig.3- 6 DC, Fundamental and Second Harmonic vs Normalized Voltage Offset

66
 If we use Taylor approximations for the coefficients In, we can remark that
In
lim  2 for n  1 . This means that the current approaches an impulse train as
 0 I
0

the conduction angle becomes small. We will see that many amplifiers have this
behavior when the input amplitude becomes large.

3.2 Square law characteristic

Another characteristic commonly used in communication circuits is the

transfer characteristic of the FET devices. The transfer that we will study is the
one of the N channel JFET. The obtained results can easily be extended to
MOSFET transistors. By reversing polarities, the same results are valid for P
channels of both types.
The square law characteristic is given by:
2
 v 
i2  I DSS 1  1  ; V p  v1  0
 V 
 p 

0 ; v1  V p
The region corresponding to positive values of input voltages is a forbidden
region.

i2

IDSS

Vp v1
Fig.3- 7 Square law Characteristic

As usual, we assume that the input signal is a dc biasing voltage added to

an ac signal. v1 (t )  Vb  v(t ) . The first case we will analyze is the one of a very
small input ac voltage ( v(t )  1 ). We use Taylor’s approximation and we obtain:

67
 i2 (t )  Ib  i(t ) where i(t )  gmv(t ) . g m is the small signal transconductance
at the Q point Vb, Ib. It is defined by:
2 I DSS
gm 
di2
dv1 v V

V p2
Vp  v1 
1 b v V
1 b

The expression of the small signal transconductance is:

2 I DSS
gm  Vx
V p2
Vx  Vp  Vb
For JFET, the small signal transconductance varies between 0 (for Vb = Vp)
and a maximum value:
2 I DSS
gm0  obtained when Vb  0 .
Vp
The voltage gain is then maximum when the biasing voltage is zero (the
biasing current is equal to IDSS).
If the ac signal is large, we restrict the analysis to the sinewave case
v(t )  V1 cos 0t . There are two possible cases: The ac signal is completely inside
the square law region ( Vp  Vb  V1  Vb  V1  0 ) or only the tip of the sinewave is
amplified.
The first case is fairly simple. Starting from:
2
 v  I
i2  I DSS 1  1   DSS2 Vp  v1 
2

 V  Vp
 p 

and replacing v1  Vb  V1 cos 0t , we obtain the following Fourier series:

i2  I 0  I1 cos 0t  I 2 cos20t
I DSS  2 V12  I DSS I DSS V12
I 0  2  Vx   ; I1  2 2 VxV1 ; I 2  2 . Vx  Vp  Vb .
Vp  2  Vp Vp 2
The Fourier series contains only three terms: dc, fundamental and second
harmonic. If we load the output with a parallel RLC circuit tuned at the
fundamental, the coefficient I1 will be the only one that will produce an output.

L C R

68
 For such system, we can define a “large signal transconductance” Gm as
being the ratio of the peak amplitude of the fundamental of the output current
over the peak amplitude of the input voltage.
I
Gm  1
V1
And the output ac voltage is v0 (t )  RGmv(t )  Gm RV1 cos 0t (The dc and
the second harmonic are eliminated by the tank circuit). The voltage gain is:
A  Gm R . In our case, the large signal transconductance is given by:
2I
Gm   DSS Vx
Vp2
The expression of the large signal transconductance is identical to the small
signal one. However, the large signal transconductance applies only to a sinewave
drive along with a high Q parallel RLC as a load. The small signal
transconductance, on the other hand, applies to the case of an arbitrary small input
drive.
If now the signal has values outside the interval [Vp, 0], we will have more
harmonics in the output current. This is due to the fact that only the tip of the
sinewave is amplified. In this case also, we can define a conduction angle 2  with
V 
  cos 1  x  . The output current is also given as a function of this conduction
 V1 
angle.

i2 (t )   I n cos n0t
n 0

I 
with a peak output value I p   DSS2  V1  Vx  .
2
V 
 p 
The coefficients are plotted in your textbook pp. 102-103, Fig.4.4-4 and
Fig.4.4-5. The analytic formulation is provided in the appendix of chapter 4,
p.147.

3.3 Exponential characteristic

This transfer characteristic is a good model of a bipolar junction transistor
driven by an ideal voltage source. In order to apply it to actual circuits, we must
make sure that the transistor is operating in normal mode. One particularity of this
transfer characteristic is the absence of a threshold. The output current is related
to the input voltage by
69
  qv 
i2  I s exp  1 
 kT 
We have already seen that kT / q is approximately 26 mV at 300°K.
The input voltage is v1  Vb  v(t ) as in the previous sections. The small
signal transfer is obtained when v(t )  1 . At that time, we can use a first order
approximation for the exponential e z  1  z . We obtain:
 qV   qv(t )   qVb  qv(t ) 
i2  I s exp  b  exp    I s exp  1  
 kT   kT   kT  kT 
The output current is equal to a dc current added to an ac current:
 qV  qI
i2  I dc  i(t ) where I dc  I s exp  b  and i(t )  dc v(t ) .
 kT  kT
The ac signals are linearly related by the following small signal
transconductance:
qI
g m  dc
kT
For example, if we consider an NPN transistor biased by a current source
IEQ

VCC

IEQ

The voltage at the collector is given by:

qI EQ
vc (t )  Vcc  RiC (t )  Vcc   RI EQ  gm Rv(t ) where g m   when the
kT
input is less than few millivolts (as in Lab #1).

70
 If the amplitude of the input becomes larger, we cannot use the above
approximation and the system is highly nonlinear. We consider the sinewave case
only. So, let the input be:
v1  Vb  v(t ) where v(t )  V1 cos 0t .
 qV   qV 
The current is then: i2  I s exp  b  exp  1 cos 0t  .
 kT   kT 
qV
Let x  1 be the input amplitude normalized to 26 mV. The current is:
kT
 qV 
i2  I s exp  b  exp  x cos 0t  . This current has values that vary between a
 kT 
peak value Ip when cos0t = +1 and a minimum value Im when cos0t = 1. The
peak value is:
 qV 
I p  I s exp  b  e x
 kT 
If we normalize i2 to Ip, we obtain the following:
i e x cos 0t
Wx (t )  2 
Ip ex

1
Wx(t)
0.9
x = 0.1
0.8

0.7

x=1 0.6

0.5

0.4

x=5 0.3

0.2

0.1

0t
-11π/10 -π -9π/10 -4π/5 -7π/10 -3π/5 -π/2 -2π/5 -3π/10 -π/5 -π/10 π/10 π/5 3π/10 2π/5 π/2 3π/5 7π/10 4π/5 9π/10 π

Fig.3- 8 Normalized current

This normalized current is plotted above for different values of x. We see

that when x is small, the current is practically sinusoidal. This character
disappears for larger values. You can observe this behavior in lab#1. The
following figure is a display of the output of part 2 of lab#1 for an input of
26 mV, i.e. x = 1.

71
 Fig.3- 9 Lab#2 Scope display for x = 1

We can remark the typical distortion of the exponential characteristic. The

figure is reversed with respect to Fig.3- 8 because we are displaying the voltage at
the collector of the transistor: vC (t )  Vcc  Rci2 (t )  Vcc  RcWx (t ) I p .
Now, the function exp  x cos 0t  is periodic and it can be developed in
Fourier series. We have seen in chapter 1 that:

exp  x cos 0t   I 0 ( x)  2 I n ( x)cos n0t
n 1

where the functions I n ( x) are the modified Bessel functions of the first
kind. These functions are tabulated in your textbook. Using the above relation, we
can write:
 qV   
2 I ( x) 
i2 (t )  I s exp  b  I 0 ( x) 1   n cos n0t 
 kT   n1 I 0 ( x) 
 
2 I ( x) 
or i2 (t )  I dc 1   n cos n0t 
 n1 I 0 ( x) 
Idc is the dc current flowing in the output branch of the circuit. If the circuit
is biased using a current source, its value is fixed and does not depend on the
input voltage. However, if the circuit uses resistive biasing, its value will depend
on that voltage. We also remark that the fundamental and the harmonics are given
2I ( x)
by I dc n . This ratio of modified Bessel functions is provided in the next table
I 0 ( x)
(following page) taken from your textbook.

72
Fig.3- 10 Normalized current harmonics

73
 2 I n ( x)
One particularity of the modified Bessel function is that lim  2.
x  I ( x )
0

This means that for large x, the Fourier development of the current becomes the
one of a periodic impulse train. Another point worth noting is that it is very hard
to define a conduction angle. In fact, the current is never zero. In your textbook,
the conduction angle is defined for a current equal to 5% of the peak value.
If the amplifier is loaded with a high Q tank circuit tuned at the
fundamental, the output voltage will be sinusoidal.

L C R

As we did with the square law characteristic, we can define here also a
large signal transconductance:
2 I ( x)
I dc 1
I 0 ( x) qI dc 2 I1 ( x) 2 I ( x)
Gm ( x)    gm 1
V1 kT xI 0 ( x) xI 0 ( x)
Giving a normalized value of:
Gm ( x) 2 I1 ( x)

gm xI 0 ( x)
This function is tabulated below.

This function is plotted below.

74
 Fig.3- 11 Normalized large signal transconductance

We can use this large signal transconductance to express the output voltage
across the tank circuit:
vo (t )  Gm RV1 cos 0t
We can remark that the gain decreases with increasing values of the input.
This behavior can be used to stabilize the amplitude of many systems such as the
sinusoidal oscillators. This decrease of gain can be explained by the f act that, as
the amplitude increases, most of the power is distributed to the harmonics rather
than to the fundamental. So, even though the output amplitude is increasing as x
is increasing, its ratio with the input is decreasing.
Example:

VCC

C R L

Q1

v(t )  (52 mV )cos 0V1

t

Idc CE

Let us consider the above circuit. The tank circuit is tuned at 0 and the
resistance R has the value of 1 k. The power supply voltage VCC has the value
52
of 10 V. The current Idc has the value of 1 mA. We have: x   2 , giving a
26
75
fundamental current I1  1.3955  (1 mA) . The voltage at the collector of the
transistor is equal to
vC (t )  VCC  RI1 cos 0t  10V  (1 kΩ  1.3955 mA) cos 0t  10 V  (1.3955 V) cos 0t
We can find the same result using transconductance calculation. The small
1 mA
signal transconductance is: g m   0.0385 1 . For x  2 , we have
26 mV
Gm (2)
 0.698 giving Gm (2)  0.0268 1 . This gives:
gm
vC (t )  VCC  Gm ( x) RV1 cos 0t
vC (t )  10V  0.0268 1000  52 103 cos 0t
vC (t )  10 V  (1.3936 V)cos 0t
We obtain practically the same result using both methods. The differences
in the results are mainly due to rounding errors.
If we want to increase the output without increasing the distortion, we can
increase the biasing current (this will increase the value of the small signal
transconductance) or we can increase the value of the load resistance (in this case
we are limited by the value of the finite output resistance).

3.4 Resistively biased BJT

When the biasing current is fixed by a current source, we have seen that the
analysis of a BJT amplifier is straightforward. The dc current is given. However,
if we use resistors to produce the biasing, the voltage across these resistors is
stored in capacitors and this voltage will depend on the applied ac voltage.

VCC

R1 RC

Q1

CB 2N3904
V1

R2 RE CE

Fig.3- 12 Resistively biased BJT

76
 The above figure shows a resistively biased BJT. By using Thevenin
theorem, we can transform the base circuit as follows:

VCC

VCC

RC

Q1

CB 2
2N3904
V1

RB
RE CE
3

VBB

Fig.3- 13 Base equivalent circuit

R1R2 R2
where RB  and VBB  VCC .
R1  R2 R1  R2
The above circuit is shown with a resistive load; however, the load RC can
be replaced by a parallel tank circuit. In the analysis of the circuit, we assume that
the transistor is operating in the normal mode. So, the emitter current is related
with the base emitter voltage by the usual exponential characteristic:
 qv 
iE  I ES exp  BE 
 kT 
i
and iC   iE along with iB  (1   )iE  E .
 1
If the base emitter voltage is only a dc one (no signal), the dc current
flowing through the emitter is given by:
VBB  VdcQ
I EQ  (9)
RE  (1   ) RB
The voltage VdcQ  vBEQ is the dc voltage that appears across the series
connection of the two capacitors CB and CE (the ac voltage source at the input is
shorted to ground). Its value is also related to the current flowing in the transistor
by:

77
 kT I EQ
VdcQ  ln (10)
q I ES
For a discrete transistor biased with a current between 0.1 to 10 mA, the
value of VdcQ remains around 650 mV. If we want to compute its value, we can use
the iterative method shown in chapter 1.
Now, if we apply an ac voltage at the input ( v(t )  V1 cos 0t ), the value of
the dc voltage changes from VdcQ to Vdc and this will entail a change in the dc
current from IEQ to IE0. The new equations are now:
vBE (t )  Vdc  V1 cos 0t
giving
 qV   qV   qV 
iE (t )  I ES exp  dc  exp  1 cos 0t   I ES exp  dc  exp  x cos 0t 
 kT   kT   kT 
Using the Fourier series development seen in the previous section, we
obtain:
 qVdc   
2 I n ( x) 
iE (t )  I ES exp   I 0 ( x) 1   cos n0t 
 kT   n1 I 0 ( x) 
 
2 I ( x) 
or iE (t )  I E 0 1   n cos n0t 
 n1 I 0 ( x) 
We can now rewrite equations (9) and (10) for a sinewave input as:
VBB  Vdc
IE0  (11)
RE  (1   ) RB
and
 qV 
I E 0  I ES exp  dc  I 0 ( x) (12)
 kT 
Since Vdc is different from VdcQ, let us write:
Vdc  VdcQ  V
If we replace in (11), we obtain:
VBB  VdcQ V
IE0  
RE  (1   ) RB RE  (1   ) RB
V  V 
I E 0  I EQ   I EQ 1   (13)
RE  (1   ) RB  V 
where V   RE  (1   ) RB  I EQ .
Equation (12) on the other hand produces:
 qV 
I E 0  I EQ exp    I 0 ( x) (14)
 kT 
78
 Equating (13) and (14), we obtain:
 V   qV 
1    exp    I 0 ( x)
 V   kT 
The above equation does not have an analytic solution. However, it is
shown in the textbook, that, if V is larger than 520 mV, we can safely neglect the
V
term in the above equation. We finally get:
V
kT
V  ln I 0 ( x)
q
Now, we can express the relationship that exists between the two dc
currents:
 
 ln I ( x) 
I E 0  I EQ 1  0

  qV  
  kT  
So, if x is small and V large, we can safely assume that the two currents
are the same. However, if the design is not very good (small V), the two currents
will be quite different when the amplitude of the input signal increases.
If we want to neglect the effect of the input drive and have practically
IE0 = IEQ within 5%, we must have:
 qV 
20ln I 0 ( x)    
 kT 
This relation is shown in the next figure.

Fig.3- 14 Condition on V

79
 For example, if the input signal has a peak value of 260 mV, the voltage V
must be larger than 4 V.

3.5 Differential Characteristic

In our collection of nonlinearities, the differential characteristic is one of
the most important. This circuit is commonly used in integrated circuits. We will
see later that it is the basic building block of the double balanced mixer-
modulator. We also find it in the implementation of IF amplifiers used in FM
receivers. This is due to the fact that it is a very good amplitude limiter.

iC1 iC 2
Q1 Q2

V1 i1 i2
V2

Ik

Fig.3- 15 Differential Amplifier

In the circuit of Fig.3- 15, the current Ik is divided between the two
transistors. This means that any value of the input voltages, the current in each
transistor will never exceed the value of Ik. The node equation at the emitter of the
two transistors is:
i1  i2  I k
This implies that the variations of the two currents are opposite. When one
increases, the other one decreases by the same amount.
We assume that the two transistors are identical. The basic equations are:
 qv   qv 
i1  I S exp  BE1  and i2  I S exp  BE 2 
 kT   kT 
i  q   q 
So: 1  exp   vBE1  vBE 2    exp   v1  v2  
i2  kT   kT 
q
Let us define: z   v1  v2  . The emitter currents become:
kT

80
 Ik I
z
i1 
and i2  k z
1 e 1 e
The above expressions do not show clearly the fact that when one current
increases by a given amount, the other decreases by exactly the same value. The
I
average value for both currents is k . Let i be a variation around this value.
2
I I
i1  k  i and i2  k  i . The variation of current is:
2 2
I z
i  k tanh   and finally the two currents can be expressed as:
2 2
I   z  I   z 
i1  k 1  tanh    and i2  k 1  tanh    .
2  2  2  2 

0.9 i1
i2
Ik
Ik
0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1
z
-4.5 -4 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0.5 1 1.5 2 2.5 3 3.5 4 4.5

Fig.3- 16 Emitter currents

The above figure shows the variation of the two currents as a function of
the normalized input. It shows clearly the opposite variation of the two currents.
Another point worth noticing is the fact that when the input is sinusoidal with
large amplitude, the output becomes practically a square wave.
If the input signal has a small amplitude, we can use a first order Taylor
z z
series approximation of the hyperbolic tangent ( tanh  ) as follows:
2 2

81
 I 
q k 
 z  I k gin
gin   
Ik
1  2   2  2  v1  v2 
2
i1  where is the input
2 kT
Ik
conductance of a transistor biased by an emitter current equal to (seen from
2
 z  I k gin Ik
the emitter). The other current is i2  1  2   2  2  v1  v2  .
2
Since the output current is iCk = ik, we can define the small signal
transconductance as:
 gin
gm 
2
So, we can remark that the gain of the differential pair is half of the gain of
I
a single ended transistor biased by an emitter current equal to k .
2
If (v1  v2 )  V1 cos 0 t , we can develop the current i in Fourier series.
qV
z  1 cos 0t  x cos 0t giving:
kT
I x 
i  k tanh  cos 0t  (15)
2  2 

i
x=∞
Ik

2
x=1
↓
0t
-2π -9π/5 -8π/5 -7π/5 -6π/5 -π -4π/5 -3π/5 -2π/5 -π/5 π/5 2π/5 3π/5 4π/5 π 6π/5 7π/5 8π/5 9π/5 2π

x=5 I
 k
2

Fig.3- 17 Sketch of i for several values of x

The above sketch shows the shape of the current i for several values of x. It
goes from a sinusoidal shape for small input to a square wave when the input
becomes large. Equation (15) represents a periodic waveform. Its Fourier series is
given by:

i   I 2 n1 cos(2n  1)0t
n 1

 I k  a2 n1  x  cos(2n  1)0t
n 1

82
 The coefficients are tabulated in p.117 of the textbook. This table is
reproduced below.

Fig.3- 18 Normalized harmonics

The last row of the above table represents in fact the Fourier series
development of a square wave having zero dc and peak amplitude of ½.
If we use a tank circuit as a load for one of the transistors, we can use the
large signal transconductance. This transconductance is defined as:
 I   I k a1 ( x)  a ( x) 4a ( x )
Gm ( x)  1   2 gin 1  gm 1
V1  kT  x x
 q 
x
 
It is represented in the figure below.

Fig.3- 19 Large signal transconductance vs x

Here again, we see the decrease of the gain as the input amplitude
increases.

83
3.6 Effect of series resistance
When a resistance R is connected in series with a nonlinear element, it has
the net effect of making the characteristics practically (piecewise) linear. In this
section, we are going to study the effect of the resistance on the exponential
characteristics. The same analysis can be repeated on other nonlinearities such as
the square law.
Consider the following circuit:

Q1

v R
1

Fig.3- 20 NPN transistor with series resistance in the emitter

From Fig.3- 20, we can write:

v1  vBE  RiE
and the transistor is described by the exponential characteristic:
 qv  kT  iE 
iE  I ES exp  BE  or vBE  ln   . This gives:
 kT  q  I ES 
kT  iE

v1    RiE
ln 
q  I ES

We can define a small signal dynamic resistance as:
dv kT
rin '  1   R  rin  R
diE i  I qI dc
E dc

The output current being the collector current, the small signal
transconductance of the compound device is:
   gin
gm '   
rin ' rin  R 1  gin R
gin  1 is the small signal dynamic conductance (seen from the emitter).
rin
We can remark that the small signal gain decreases with increasing R. To show
the effect of piecewise linearization, let us introduce the dc current flowing in the
emitter circuit.

84
 kT  iE I dc  kT  I dc  kT  i 
v1  ln    RiE  ln   ln    RiE
q  I ES I dc  q  I ES  q  I dc 
kT  iE  kT
Let V0  ln   and Vco  I dc (rin  R)  (1  gin R) , the above
q  I ES  q
expression can be evaluated as:
ln  E 
i
v1  V0  gin R  iE   I dc 
    (16)
Vco 1  gin R  I dc  1  gin R
Two extreme cases for equation (16) are of interest. The first one
corresponds to gin R  0 . This is the case of the exponential characteristics seen in
section 3.3. The other extreme case corresponds to gin R   . In this case, the
second term in the above expression will be zero and we obtain:
v1  V0 iE
 for v1  V0 and of course, no current can flow if v1  V0 . In this
Vco I dc
case, R >> rin and Vco = RIdc. Finally, the piecewise linear characteristic is the one
of an ideal diode in series with a battery of value V0 and a resistance R. The base
emitter is replaced by the following circuit:
R

Ideal

v
1

Vo

The current flowing in the above circuit is of course the emitter current.
Consider the following circuit:
VCC

ZL

Q1

CE
v(t)
Idc

85
 It can be analyzed using the following equivalent circuit:

iE
R ZL
CE

Ideal
v(t)

 iE
VCC
Idc

Vo

The capacitor is going to develop a biasing voltage Vb that will be added to

the input ac voltage v(t )  V1 cos 0t . The average value of the emitter current is
fixed by the dc current source. So, we will have the following situation:

Slope 1/R
iE

Idc

V0 v1
Vb

Fig.3- 21 Input transfer

The voltage applied to the resistance in series with the base emitter junction
is: v1  Vb  V1 cos 0t . From the above figure, it is clear that if the amplitude of
the ac input voltage satisfies: V1  RI dc  Vco , we are in the class A and the emitter
current satisfies:
V  V 
iE  I dc  1 cos 0t  I 0  I1 cos 0t  I dc 1  1 cos 0t 
R  Vco 

When V1  Vco , we leave the class A and the current becomes a periodic
train of sinewave tips as seen in section 3.1. The dc value of the current is fixed
by the biasing network. However, since the current is now asymmetrical, the
voltage across the capacitor is going to move in order to maintain the current at its
biasing value Idc.
iE  I 0  I1 cos 0t  I 2 cos20t 
From the equations developed in section 3.1, we have:
86
 V1  Vx sin    cos
I 0  I dc 
R 1  cos
V  V   cos sin 
I1  1 x
R 1  cos 
Vx
while   cos 1 and Vx  Vb  V0 .
V1
Using the fact that V1  Vx  V1 1  cos  , we can relate
V1
to  by:
Vco
V1 
 (17)
Vco sin    cos
and
I1   sin  cos
 (18)
I 0 sin    cos
The above two equations form a parametric evaluation of the relation
I V
between 1 and 1 for V1  Vco and using the result of class A also
I dc Vco
I V
( 1  1 ), we can plot the following curve:
I dc Vco
I1
I dc
2 g inR = 0

1.8 g inR = 

1.6

1.4

1.2

0.8

0.6

0.4

0.2

V1
Vco
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

Fig.3- 22 Normalized fundamental vs. normalized input voltage

In the above graph, two curves are drawn: the one corresponding to
gin R   and the one corresponding to gin R  0 . The second one is just the curve

87
 2 I1 ( x) kT
expressing vs. x. This is due to the fact that Vco  rin I dc  in this
I 0 ( x) q
V1 qV1
case, so   x . We can use the above set for any value of R by
Vco kT
interpolating between the two curves.
If the load of the transistor is a high Q tank circuit, here again, we can use
the concept of large signal transconductance.
 I1 
 I1 I1 Vco  I dc  I dc 
Gm    gm '
V1 I dc V1 Vco  V1 
 V 
 co 
I  I dc
g m '  dc  is the small signal transconductance evaluated at
Vco  rin  R  I dc
the Q point. We can use the previous set of curves (Fig.3- 22) to draw the curve
G V
giving m for different values of 1 . This can be done for the two cases:
gm ' Vco
gin R   and gin R  0 .

Fig.3- 23 Normalized large signal transconductance vs. normalized input voltage

We see here again the decrease of the gain as the distortion of the output
current increases. Consider the following circuit:

88
 VCC
10V

C 2k L

ac short

V1

130mV 2.6mA
10MHz
0Deg

The tank circuit composed of L and C is tuned a 10 MHz. We assume that

  1. We will compute the output voltage for the three cases: R = 10 , R = 0
and R = 100 .
First, we compute gin and rin .
qI 2.6mA
Idc = 2.6 mA gives gin  dc   0.11 , so rin  10 .
kT 26mV

Case R = 10 : gin R  1 . We must compute Vco.

Vco  I dc  rin  R    2.6mA   20   52mV .
V1 130 I
  2.5 . From the curves in Fig.3- 22, we read 1  1.5 . The
Vco 52 I dc
fundamental of the output current is then: I1  1.5   2.6mA  3.9mA . So, the
voltage at the collector of the transistor is:
vo (t )  10V   2k     3.9mA cos  2 107 t   10V   7.8V  cos  2 107 t 

V1 130
Case R = 0: gin R  0 . We have Vco  26mV . We obtain:   5 giving
Vco 26
I1
 1.8 from the curve gin R  0 . In this case, the output voltage is:
I dc
vo (t )  10V   9.36V  cos  2 107 t 

Case R = 100 . gin R  10 . This value is quite large.

89
 Vco   2.6mA  10  100   286mV . In this case, we have V1  Vco . So,
130mV
I1  g m 'V1   1.18mA . The output voltage is:
110
vo (t )  10   2.36V  cos  2 107 t 
If we double the input voltage (V1 = 260 mV), we will still have V1  Vco .
The amplification will still be linear.

3.7 Clamp biased FET

We have seen that N channel junction FET must be biased with a negative
VGSQ to keep the gate to channel diode reverse biased. One technique for
automatically achieving this is to use clamp biasing.
Consider the following circuit:

Q1
CG CG
G

V1 cos 0t VC
RG
 RG ideal

The circuit is equivalent to the one on the right. We assume that the time
2
constant of the RC network satisfies RGCG  . In this case, the capacitor will
0
charge to the peak value of the input sinewave (VC = V1) and it will not discharge.
So, the voltage vGS will satisfy:
vGS  V1 cos 0t  VC  V1  cos 0t  1
The above equation show clearly that the voltage vGS is going to be
Vp
clamped to zero and it will remain negative. If V1  , the FET will be operated
2
Vp
completely in the square region. If V1  , the drain current will consist of
2
squared sine tips of peak value I p  I DSS and it will contain more harmonics.
iD  I 0  I1 cos 0t  I 2 cos20t  I3 cos30t 
The following two curves provide the values of the first components of the
current along with the normalized transconductance.

90
 Fig.3- 24 Normalized Fourier Coefficients

Fig.3- 25 Large signal transconductance

2 I DSS
The large signal transconductance is normalized to g m 0  .
Vp
91
 For example, consider an FET with IDSS = 6 mA and Vp = 4 V. If we apply
V1
an input voltage v(t )  (2 V )cos 0t , we have  0.5 . From the curve in
Vp
I
Fig.3- 24, the normalized fundamental current is 1  0.5 , giving a
I DSS
fundamental current (3mA)cos 0t .

3.8 Non linear loading of tank circuits

It is quite common that the input of the different nonlinear controlled
sources is tuned using a parallel tank circuit. If the Q of the tank circuit is high
enough, the voltage across the circuit will be sinusoidal, even if the current
absorbed by the nonlinear load is not. Since all the harmonics produced by the
load will be absorbed (shorted to ground) by the tank circuit, the nonlinear load
will behave like an equivalent conductance GNL given by:
I
GNL  1 where I1 is the amplitude of the fundamental current absorbed by
V1
the load and V1 is the amplitude of the sinusoidal voltage across the load.

Common base equivalent input:

Let us consider the following circuit:

Q1

ac short

I cos 0t
R1
L C Idc

Fig.3- 26 Tank circuit at emitter

We assume that 02  1 and that the Q of the tank circuit is high enough
LC
in order to keep the voltage across it sinusoidal. We have seen that the emitter
current of the transistor is given by:
 
2 I ( x) 
iE (t )  I dc 1   n cos n0t 
 n1 I 0 ( x) 
qV
where x  1 and V1 cos 0t is the sinusoidal voltage appearing across the
kT
tank circuit.
92
 2 I1 ( x)
The fundamental current has an amplitude: I1  I dc and the base
I 0 ( x)
emitter junction will behave like a conductance:
I 2 I I ( x) qI dc 2 I1 ( x) Gm ( x)
GNL  1  dc 1  
V1 V1I 0 ( x) kT xI 0 ( x) 
We must verify that the Q is high enough:
C0 1
Q  10 where G1 
G1  GNL R1
Common emitter equivalent input:
If we consider now the base emitter junction, but seen from the base, the
current is smaller iB  (1   )iE . So, the conductance is:
G ( x) Gm ( x)
GNL  (1   ) m 
 
Clamp biased FET equivalent input:

V1 Q1

CG

I cos 0t CG L
C R1 V1 cos 0t RG

The current flowing in the gate channel diode consists of very short
impulses occurring at each peak of the sinewave V1 cos 0t . This means that we
can develop this gate current in the following Fourier series:
 

iG  I dc 1  2 cos n0t 
 n 1 
Since no dc current can flow through the capacitor, this dc current is going
to flow through the resistor RG. The dc voltage across RG is the peak value V1 of
the ac voltage stored in the capacitor CG. So:
V
I dc  1 and the amplitude of the fundamental current in this diode is then:
RG
V V
2 1 . There is also an ac current of amplitude 1 at 0 flowing in this
RG RG
resistance since there is an ac voltage V1 cos 0t across it. The total ac current at

93
0 flowing inside the circuit is then the sum of the two currents. Finally, the
equivalent conductance across the tank circuit is:
3V1
R 3
GNL  G 
V1 RG

94