Communication circuits

Chapter 4

EE312

by

A. DAHIMENE

Institut de Génie Electrique et Electronique

Université M'hamed Bougara

Boumerdes
 Chapter 4 3
Sinusoidal Oscillators ....................................................................................................................3
4.1 Basic definitions ..................................................................................................................3
4.2 Linear feedback analysis .....................................................................................................4
4.3 General conditions for oscillator design ..............................................................................8
4.4 LC oscillators ....................................................................................................................11
4.5 Stability factors .................................................................................................................15
4.5.1 Direct stability factor ..................................................................................................15
4.5.2 Indirect stability factor ...............................................................................................16
4.6 Crystal oscillators ..............................................................................................................20
 Chapter 4
Sinusoidal Oscillators

Sinewave generators are commonly used in communication circuits. We

find them in both transmitters and receivers. There exist many types of oscillators:
multivibrators, relaxation oscillators, etc. Most of them generate waveforms that
are not sinusoidal. In this chapter, we are going to analyze only the ones that can
generate sinewaves.

4.1 Basic definitions

A sinusoidal oscillator is a one port device that produces a sinewave. The
output signal is then:
x(t ) = A cos (ω0t + θ )
It is a simple matter to show that the above signal satisfies the differential
equation x′′(t ) + ω02 x(t ) = 0 . This equation represents a harmonic oscillator such as
the small amplitude pendulum in vacuum, a mass and spring without friction, etc.
In our case, we are interested in electrical (RC and LC) oscillators and also in
electromechanical ones (crystal).
The oscillators that we will study are usually composed of a passive
network that is used to set the frequency of oscillation and an amplifier that is
used to compensate for the losses in the passive network.

X K H ( s)

Fig. 4- 1 Typical sinusoidal oscillator

The amplifier is assumed to be memoriless. Any memory element in the

real amplifier will be integrated inside the passive network. The above network
can be analyzed using linear techniques. However, such analysis would
correspond only at the conditions that set the start of oscillations. When the
amplitude of the sinewave is small, we have seen that all of the active devices can
be considered as linear (incrementally linear). However, as the amplitude
increases, this assumption is no longer true. We then have to resort to nonlinear
analysis.
A general nonlinear analysis of oscillators would lead us to the study of
nonlinear differential equations. This is an extremely complex method of analysis.
In general, we use much simpler techniques.
A general study of a sinusoidal oscillator starts in general by a small signal
analysis (linear). This step allows us to compute the frequency of oscillation and
also the conditions on the gain of the amplifier in order for the oscillations to start.
The second step takes into consideration the nonlinearities and studies
conditions for setting the amplitude of the sinewave at a predetermined level. We
can also try to determine the distortion of the waveform at this point of analysis.

4.2 Linear feedback analysis

The block diagram shown in Fig. 4- 1 suggests the following closed loop:

vi + ε
K H ( s) vo
+

Fig. 4- 2 Closed loop

The block diagram shown in Fig. 4- 2 is equivalent to the one of Fig. 4- 1 if

the input signal vi is zero. Using results from linear feedback theory, we can
define an open loop gain:
AL ( s ) = KH ( s ) (1)
and a closed loop gain:
AL ( s ) KH ( s )
G (s) = = (2)
1 − AL ( s ) 1 − KH ( s )
In our analysis, we are going to consider only passive networks that can be
built using lumped elements. This means that the methods of analysis cannot be
used for studying oscillators implemented using transmission lines or delays, etc.
In the case of lumped elements, the transfer function of the passive network is a
ratio of polynomials.
N ( s)
H ( s) = (3)
D(s)
 where N(s) and D(s) are polynomials. It is clear that the roots of N(s) are
the zeros of the open loop gain AL(s) and the roots of D(s) are the poles of the
open loop gain AL(s). Replacing (3) in (2) provides:
KN ( s )
G ( s) =
D( s ) − KN ( s )
The above equation can be written as:
[ D( s) − KN (s)] v0 ( s) = KN ( s)vi ( s) (4)
Assimilating the operator s to the differentiation operator, we can see from
(4) that we can have an output vo different from zero when vi is zero if the
following constant coefficient differential equation is satisfied:
[ D( s) − KN ( s)] v0 (s) = 0 (5)
If the roots of D( s ) − KN ( s ) are distinct, the general solution of the
differential equation (5) is:
n
v0 (t ) = ∑ Ck exp ( ck t )u (t ) (6)
k =1

where ck are the roots of (5). They are the closed loop gain poles and Ck are
coefficients that depend on the initial conditions.
Since the passive network is built using real components (R, L, C and M),
its transfer function is a ratio of polynomial with real coefficients. If we want to
have as a solution a sinewave, we must have two closed loop poles on the jω axis
at ± jω0 and we have to make sure that the other closed loop poles will not appear
in the signal vo(t). This can theoretically be achieved by setting the initial
coefficients such that Ck = 0 except for the selected pair of poles. However, this
solution is not realistic. A much more sensible solution is to make sure that all the
unwanted poles are “stable”, i.e. with a negative real part. So, if we wait a while,
all the unwanted signals will decay and only the required signal will remain.
From the above discussion, we observe that the position of the closed loop poles
is fundamental. These closed loop poles are the solutions of the equation:
D( s ) − KN ( s ) = 0 (7)
This equation depends on the gain K. So, the position of the closed loop
poles will also depend on the gain K. their locus in the s-plane is called the “root
locus”. There are some general rules for drawing a root locus. You will learn
them in control courses. However, we can state some of them.
When K = 0, equation (7) becomes D(s) = 0. This means that the closed
loop poles are on the open loop poles for small gains. When K becomes very
large, equation (7) is practically N(s) = 0. So, the closed loop poles will be on the
open loop zeroes. For some value of the gain, the root locus will cross the jω axis
at the frequency ω0.
There are some minimal requirements on the transfer function H(s) so that
there can be oscillations and therefore closed loop poles on the jω axis. It is
shown that the transfer function must have at least two poles and one zero.
Consider the following example:
s 2 + cs + d
H (s) = 2
s + as + b
Equation (7) becomes:
(1 − K ) s 2 + (a − Kc) s + b − Kd = 0
which corresponds to the differential equation:
(1 − K ) v0′′ + ( a − Kc ) v0′ + ( b − Kd ) v0 = 0
In order to have oscillations, we must have:
b − Kd
a − Kc = 0 and > 0 . If these two conditions are satisfied, the
1− K
equation becomes:
b − Kd
s 2 + ω02 = 0 where ω0 =
1− K
If we have one pair of closed loop poles on the jω axis, this implies that
1 − AL ( s ) = 0 for s = jω0. In other words:
AL ( jω0 ) = 1 (8)
The above relation can be expressed as:
AL ( jω0 ) = 1 and arg [ AL ( jω0 )] = 0 mod 2π (9)
Equations (8) and (9) are necessary conditions. They are known as the
“Barkhausen Conditions”. Equation (9) simply means that, if we look at the
diagram of Fig. 4- 1 and if we consider signals on the right and on the left of the
point X, they are identical. There exists a sinewave which travels around the loop
with no attenuation and no phase shift. Of course, in order to have a good
sinewave oscillator, this condition should exist only for one single frequency.
Sometimes, it is easier to use the real and imaginary part of the open loop transfer
function.
Example:
Consider the open loop gain:
K
AL ( s ) =
(1 + RCs )
3
 K K (1 − 3ω 2 ( RC ) 2 ) KRCω ( 3 − ( RCω ) 2 )
AL ( jω ) = = −j
(1 + jRCω ) (1 + ( RCω ) ) (1 + ( RCω ) )
2 2 3 2 3

The application of the Barkhausen conditions provides:

3
Im [ AL (ω0 ) ] = 0 gives ω0 = and Re [ AL ( jω0 ) ] = 1 gives K = −8
RC
This circuit can be realized by cascading three first order lowpass circuits
separated by buffers.

R1

R
R2 R
R
C
C
C

The gain K is given by the ration of the two resistances R1 and R2.
R1
K =−
R2
In order to neglect the load on the last RC stage, we assume that the
resistance R2 is very large. If it is impossible to achieve this, we must add a high
input impedance buffer before.
The previous example shows that the value of the gain K is critical. If its
absolute value is smaller than 8, the closed loop poles will be in the left side of
the s-plane and the oscillation will not start. If the gain is too large, they will be in
the right half, the sinewave envelop will be a growing exponential and the
amplitude will increase until it will be limited by the amplifiers nonlinearities. So,
we need an infinite precision in order to set the conditions of oscillation.
Even if we are able to set exactly the gain, the linear theory cannot predict
the amplitude of the sinewave. Theoretically, this amplitude is set by the initial
conditions. These conditions cannot be predicted for general circuits. So, this
linear analysis is used only to determine the frequency of oscillations along with
conditions for starting these oscillations.
4.3 General conditions for oscillator design
There are certain conditions that must be satisfied in order to design a
practical oscillator. First, the oscillator must always start when we turn the power
on. Next, the amplitude of the waveform should be under our control and not set
by random initial conditions. We must also have good control on the distortion
and finally, the frequency of oscillation should not depend on parasitic elements
and on environmental conditions.
In general, the analysis of an oscillator should be performed in two steps.

• We do a linear small signal analysis and we must set the selected closed
loop poles on the right half of the complex plane (not to far from the j ω
axis). This will ensure that the oscillator will start what ever initial
conditions. Even if the memory elements (capacitances, inductances, etc.)
have zero initial conditions, the existence of thermal noise will ensure that
the oscillation will start. The fact that the closed loop poles are on the right
side implies that the waveform will have an exponentially growing envelop.

Fig. 4- 3 LC oscillator starting

Fig. 4- 3 shows the initial time of an LC oscillator (the one used in lab #4).
It takes practically 100 µs for the amplitude to stabilize at its final value.
• There should exist a mechanism of amplitude control that will fix the
amplitude at some pre-selected value. This mechanism can be an automatic
gain control or we can use the nonlinearities of the amplifying device to
push the closed loop poles back on the jω axis.
 An instructive example is the Wien bridge oscillator of lab #3.
4

R R3
1

VCC U1 VEE 5

3 10V -10V
VCC VEE

OPAMP_5T_VIRTUAL

C R

R1

Fig. 4- 4 Wien Bridge Oscillator

It is shown in lab #3 that the open loop transfer function of the Wien bridge
oscillator is:
− A s 2 − sδω0 + ω02
AL ( s ) =
3 + δ s 2 + 3sω0 + ω02
1
Where ω0 = and R3 = (2 + δ ) R1 . The previous method of analysis
RC
(Barkhausen conditions) gives that the circuit will oscillate at ω0 if the gain of the
9
amplifier is Amin = + 3 . If the gain is higher than Amin , the closed loop poles will
δ
be in the right half of the s-plane.
If we use an op-amp, its differential gain is very high. So, A >> Amin . In fact,
from what we have learned earlier, when the gain is very large, the closed loop
poles will be practically on top of the open loop zeroes. These zeroes are:
δ4 −δ 2
z1,2 = ω0 ± j ω0
2 2
So, if we want the oscillator to start, we must put the closed loop poles on
the right half of the s-plane. This implies that the open loop zeroes must also have
a real part that is positive. We must have δ > 0 or R3 > 2R1. However, the value
of δ must not be too large. If δ > 2, the open loop zeroes will be real and the
oscillations will not start. In the lab experiment, you will design an oscillator with
δ = 1. In this case, we will have exponentially growing oscillations at the
 3
frequency of ω0 . This waveform will be limited by the saturation of the op-
2
amp and the waveform displayed by the oscilloscope will be:

Fig. 4- 5 Output waveform for δ = 1

A better signal will be obtain in the second part of lab #3 by the use of a
voltage controlled resistance in place of R1. When the system is operating
correctly, the value of δ will be very close to zero and the sinewave will be very
pure.
10

2.2kΩ
R R3
R8
1 7

C
D2
DIODE_VIRTUAL
2
VCC U1 VEE 5 D1
3 10V -10V 5V
VCC VEE

OPAMP_5T_VIRTUAL

R6 8

51kΩ
R5 Q1
C R 510 Ω
9
6
R'1 JFET_N_VIRTUAL C3
R4 680kΩ 10nF
11kΩ R2
0

Fig. 4- 6 AGC using FET

 Fig. 4- 7 Output waveform

4.4 LC oscillators
LC oscillators use a parallel RLC circuit along with a transformer as
passive network. The amplifier is one of the nonlinear controlled sources studied
in chapter 3. So, the basic circuit is shown below (Fig. 4- 8). Gin represents the
load presented by the input of the active device at the oscillating frequency.

feedback

1 : n

f(v) L C GL
v Gin

ideal
Fig. 4- 8 Typical LC oscillator

If we analyze the response of the passive network, we remark that the

parallel RLC circuit has a zero phase response at the resonant frequency and at
any other frequency it is either capacitive or inductive. So, from the phase
condition of the Barkhausen relations, the frequency of oscillation will be the
resonant frequency:
1
ω0 =
LC
 If the total Q of the tank circuit is larger or equal to ten, the voltage at the
input of the active device will be sinusoidal, even if the output current is not. This
means that the only component of the current that will produce a voltage across
the tank circuit is the fundamental. For all voltage controlled current sources, this
current is given by:
I1 cos ω0t = GmV1 cos ω0t
Gm is the large signal transconductance and V1cosω0t is the input voltage
across the input of the active device. This means that the voltage across the tank
circuit is:
I1 GmV1
v L (t ) = cos ω0t = cos ω0t
GL + n Gin
2
GL + n 2Gin
At the output of the ideal transformer, we obtain the input voltage V1cosω0t,
and this voltage is equal to nvL(t). So, we obtain the equation:
nGm GL
= 1 or G = + nGin
GL + n 2Gin
m
n
If we divide the above relation by the small signal transconductance, we
obtain:
Gm GL nGin
= + (10)
g m ng m gm
Equation (10) must be satisfied for some value of the input. We have seen
G
that the ratio m depends on the amplitude V1 and that it decreases as the
gm
amplitude increases. Furthermore, for very small amplitude, this ratio is equal to
one. This means that the small signal transconductance must
G
satisfy g in ≥ L + nGin for the oscillator to start.
n
Gm
gm
1
GL nGin
+
ng m gm

input

Fig. 4- 9 Typical large signal transconductance

 Fig. 4- 9 shows a typical ratio of large to small signal transconductance. It
also gives the value of the amplitude of the oscillation. From this curve, we can
see how the oscillation starts and how it stabilizes. At start up, the amplitude is
very small, the gain is larger than required. The closed loop poles will be situated
in the right half of the s-plane and we will have exponentially increasing
oscillations. When the amplitude reaches the required one, equation (10) will be
satisfied and this means that the poles will be on the jω axis. Any further increase
of the amplitude will decrease the gain and the poles will move to the left half of
the s-plane and at that time the amplitude will decrease. So, we see that the
amplitude will stabilize at the amplitude for which equation (10) is satisfied. An
example of such oscillator is the Colpitts oscillator of lab #4.
Let us consider the following circuit: The transistors have β = 100.

Fig. 4- 10 Differential amplifier Colpitts oscillator

It consists of an oscillator built around a differential amplifier. The

coupling between the output (collector of transistor Q2) and the input is done
using a split capacitor transformer-like network. As seen in chapter 2, if we
assume that nQT 'QE > 100 and QE > 10 , we can transform the above circuit into
the one shown below.
 1 : n

Gm ( x)V1 cos ω0t L C GL vt Gin V1 cos ω0t

Fig. 4- 11 Equivalent circuit

C1C 2
The capacitance C is: C = = 2.25 nF so the resonant frequency is:
C1 + C 2
1
f0 = ≈ 1 MHz . The input admittance of the transistor is:
2π LC
G ( x)
Gin = m as seen in chapter 3. We need the value of x in order to
β
determine the value of the transconductance. However, we know that it is
bounded by the value of the small signal one. In order to determine the small
signal transconductance, we need to compute the biasing current.
V − 0.75
I k = EE ≈ 2 mA
(2 − α ) R3

The small signal transconductance is:

qI 2 mA
gm = α k = = 1.9 × 10−2 Ω −1 so the small signal input
4kT 4 × 26 mV
conductance is: gin = 1.9×10−4 Ω−1.
C1
The turn ratio is: n = = 0.0625 . We can now bound the value of
C1 + C 2
QT' and QE.
ωC
QT ' = 20 . The conductance G is the sum of the input conductance and the
nG
conductance of the resistor R1.
1
G≤ + gin = 2.3 × 10−4 Ω −1 giving QT' ≥ 16361 and QE ≥ 1090. The
22 kΩ
assumptions are largely verified. Furthermore, the very small value of G allow us
to neglect the second term in equation (10). So, we can write:
Gm GL
= = 0.17
g m ng m
 So, using Fig.3-19, we obtain x = 13.5. With this value, the output
I
fundamental current is given by 1 = 0.6366 from Fig.3-18. So, the output
Ik
voltage is:
vt (t ) = VCC − RL I1 cos ω0t = 10 V − (4.7 k Ω × 0.6366 × 2 mA) cos ω0t = 10 V − (5.98 V ) cos ω0t
The same analysis can be repeated for any LC oscillator built around a
nonlinear controlled current source.

4.5 Stability factors

When we design and implement an oscillator, we must make sure that its
parameters are not going to vary with different conditions such as aging,
temperature, etc. The most important parameter that must be fixed is the
frequency of oscillation. In general, frequency is allocated by government
agencies to users and in order to avoid interferences, the frequency must not
change with time and should not be affected by external conditions. We
distinguish two different stability factors.

4.5.1 Direct stability factor

The direct stability factor indicates the sensitivity to a variation of the
frequency setting elements of the circuit. We have seen that the frequency of
many RC oscillators is inversely proportional to the product RC.
k
ω0 =
RC
In this case, we can write:
∆ω  ∆R ∆C 
= − + 
ω0  R C 
The above expression indicates that a relative variation of the frequency of
the oscillator depends on the relative variation of the frequency setting elements.
If these elements vary when external conditions such as temperature vary, we can
appreciate the induced frequency variation and correct it. One commonly used
technique is to select components with inverse temperature coefficients in order
to have compensation. This technique is mostly used in LC oscillators. In this
case, we have:
1 ∆ω 1  ∆L ∆C 
ω0 = giving: =−  + 
LC ω0 2 L C 
 We can also put the oscillator inside a temperature stabilized oven in order
to eliminate the temperature dependence of the circuit.

4.5.2 Indirect stability factor

When we design an oscillator, we assume that the elements are ideal.
However, there are many parasitic elements that intervene in a circuit. The wires
of the different components are inductive. There exists a distributed capacitance
between the turns of an inductor and there exist many nonlinear memory elements
that are present in active devices. We can model these elements as "spurious"
poles and zeroes that are introduced in the open loop transfer function. In general,
these spurious poles and zeroes have a very large frequency (imaginary part) and
a very negative real part. We can express the real open loop transfer function as:
ALreal ( s ) = AL ( s ) ALspur ( s ) (11)
where AL(s) is the theoretical open loop transfer function, ALspur(s) is the
part of the transfer function due to the spurious poles and zeroes and ALreal(s) is
the actual open loop function. From the Barkhausen conditions, we know that the
frequency of oscillation is given by the frequency for which the phase of the open
loop transfer function is zero. From equation (11), we can write:
φreal (ω ) = φ (ω ) + φspur (ω ) (12)
where φ (ω ) = arg [ AL ( s )]s = jω .
Consider the following phase responses:

φ 2(ω )

φ 1(ω )

Fig. 4- 12 Phase responses

 In Fig. 4- 12, we can observe two different theoretical phase responses that
pass by zero at the same abscissa. According to equation(12), the actual phase
response is going to shift up or down by a random small amount. It is evident that
the intersection with the frequency axis is going to shift by a much smaller
amount for the phase response φ2 (ω ) than for φ1 (ω ) . So, the spurious poles and
zeroes will have a much smaller effect on the frequency of oscillation for φ2 (ω ) .
From the figure, we can observe that the main difference between the two curves
is their slope around the frequency ω0. It is clear that the steeper the curve around
ω0, the less sensitive the circuit is to spurious poles and zeroes. So, we can define
an "indirect frequency stability factor" from the phase response of the theoretical
open loop gain. We define it as the ratio of the variation of phase over the relative
frequency change around the frequency of oscillation.
∆φ
SF = (13)
 ∆ω 
 
 ω0 
From which we deduce:
ω0 ∆φ
∆ω =
SF
The above relation indicates that the frequency shift is small if SF is large.
For example, a phase shift of 1° implies a frequency shift of 174.5 Hz around an
oscillation frequency of 1 MHz if the value of the indirect stability factor has a
value of 100.
In many cases, the variations are very small, so we can replace the
variations by differentials, and we can write:
dφ
S F = ω0 (14)
dω ω =ω0
Equation (14) allows the computation of SF from the expression of the open
loop transfer function. For lumped circuits, the open loop transfer function is a
ratio of real polynomials. This means that is has zeroes and poles that are either
real or they occur as pairs of conjugate complex numbers. The open loop transfer
function can be written as:
AL ( s ) = A0
( s − z1 )( s − z2 )L( s − zM )
( s − p1 )( s − p2 )L( s − pN )
For a real A0, the phase of this transfer function is the sum of the phases
due to the zeroes plus the sum of the phases due to the poles.
 M N
φ (ω ) = arg [ AL ( jω ) ] = ∑φzk + ∑φ pi
k =1 i =1

The phases introduced by the poles are negative because they occur at the
denominator.
From equation(14), the stability factor can then be written as:
M N
S F = ∑ S Fzk + ∑ S Fpk
k =1 k =1

dφzk dφ pk
where S Fzk = ω0 and S Fpk = ω0 .
dω ω =ω0 dω ω =ω0

We just have to know the contribution of a real zero (pole) and a pair of
complex conjugate zeroes (poles) in order to be able to compute the stability
factor for any transfer function. This computation is performed using a pole and
zero plot.
Indirect stability factor corresponding to a real zero (pole):
Consider a zero z1 = −ω1 ∈  .

ω
φ1

ω1

ω 
The angle φ1 is given by: φ1 (ω ) = tan −1   and this gives a stability factor
 ω1 
equal to:
ω0
dφ ω1
S Fz = ω0 1 =
dω ω =ω0
ω
2
1 +  0 
 ω1 
Let us now consider a pair of complex conjugate zeroes: z1,2 = −α ± j β .
The corresponding plot is:
 ω

φ1

β
−α

−β
φ2

The total phase contribution corresponding to the pair of zeroes is given by:
ω−β ω+β
φ (ω ) = φ1 + φ2 = tan −1 + tan −1
α α
This gives an indirect stability factor of:
dφ 2αω0 (ωres2
+ ω02 )
S F = ω0 = 2 where ωres
2
=α2 + β2.
dω ω =ω0 (ωres − ω0 ) + 4α ω0
2 2 2

ωres is the resonant frequency corresponding to the pair of complex zeroes.

It is quite usual for an oscillator to operate at the resonant frequency. At that time,
we have ω0 = ωres and the stability factor simplifies to:
ωres
SF = = 2QT
α
where QT is the "Que" of the pair of zeroes. If we consider poles instead,
we have:
S F = −2QT
Example: Consider the oscillator seen in section 4.2.
K −1
AL ( s ) = . It has 3 real poles at the position: . So the indirect
( + )
3
1 RCs RC
ω
−3  0 
stability factor is: S F =  ω1  along with ω = 3 and ω = 1 . After
ω
2 0 1

1+   RC RC
0

 ω1 
3 3
replacement, we obtain: S F = . This is a quite small value. With a stability
4
factor that small, it is very hard to adjust exactly the value of the frequency.
Furthermore, any variation in the parasitic elements will have a large influence on
the oscillation frequency.
The LC oscillators all have a transfer function having a zero at the origin
and a pair of complex conjugate poles. This gives a stability factor equal
to: S F = −2QT . With typical capacitors and inductors, we can achieve maximum
values of QT of 150. So, the largest value we can expect for the indirect stability
factor is 300. This is much better than the RC oscillator seen before. We can
improve the stability of the oscillator by having a very small direct stability factor.
This can be achieved by selecting capacitors with a negative temperature
coefficient equal in absolute value to the positive temperature coefficient of the
inductor at the ambient temperature.

4.6 Crystal oscillators

In the previous section, we have seen that we can achieve indirect stability
factors of about 300 with LC oscillators. This corresponds to a frequency shift of
58 Hz for a phase shift of 1° when the oscillation frequency is set to 1 MHz.
However, this is the limit of what we can achieve with LC resonators. In order to
achieve better results, we have to resort to mechanical resonators. A very high Q
can be achieved with "crystal" resonators. It is possible to obtain Q's of about 104
at resonant frequencies around 1 MHz. This means that the indirect stability
factor will be around 2×104. This corresponds to a frequency shift of 0.87 Hz for
a phase shift of 1° (less than 1 Hz).So, if we put the oscillator circuit in a
temperature stabilized oven, we can guarantee the frequency stability to the Hertz
value when the oscillation is around 1 MHz. In fact, many test equipments have
such oven stabilized references.

Fig. 4- 13 Crystal enclosures and mounting

 Some materials such as quartz possess the piezoelectricity property. In
other words, a quartz crystal submitted to a mechanical strain along a given axis
will develop an electrical polarization along another axis. This process can be
reversed. If we apply a voltage across this axis, the crystal will have a mechanical
deformation along the first axis. This property is used to build transducers like the
ones used in SONAR and in echographic imaging. However, in our case, we are
interested in mechanical resonators. A crystal cut has some elasticity and a mass.
So, it can act as a mechanical resonator. The piezoelectricity allows us to transfer
this mechanical resonance to an electrical one.
We can show that a crystal has the following equivalent electrical circuit.
C r
L

Co

Fig. 4- 14 Crystal equivalent circuit

The inductance L is the electrical equivalent to the mechanical inertia. The

capacitance C is the electrical equivalent to the elasticity and the resistance r is
the electrical equivalent to the mechanical loss. These losses are greatly reduced
by enclosing the piece of crystal inside a vacuum container. The capacitor C0 on
the other hand is due to the metallization on both faces of the crystal used to make
the electrical connection. For example, typical values for an 8 MHz crystal are:
L = 14 mH, C = 27 fF = 0.027 pF, r = 8 Ω and C0 = 5.6 pF. Except for C0 and r,
the other values are impossible to realize using discrete components.
The impedance of the circuit shown in Fig. 4- 13 is:
r 1
s2 + s +
1 L LC
Z ( s) = (15)
C0 s s 2 + r s + C + C0
L CC0 L
We introduce the following frequencies:
1
ω0 = : It is the resonant frequencies of the series arm composed of L,
LC
C and r .We call it the resonant frequency.
C + C0 C  C 
ω1 = = ω0 1 + ≈ ω0 1 +  : It is the resonant frequency of a
LCC0 C0  2C0 

circuit composed of the parallel connection of L, C in series with C0 and a

transformed r. We call it the antiresonant frequency. From the above definitions
of frequencies, we can also define the "pulling range" as ∆ω = ω1 − ω0 . It is equal
to:
C
∆ω ≈ ω0
2C0
We can re-express equation (15) using the above notations:
1 s 2 + 2α s + ω02
Z ( s) =
C0 s s 2 + 2α s + ω12
It has two finite zeroes at z1,2 = −α ± j ω02 − α 2 and three poles at
p0 = 0 and at p1,2 = −α ± ω12 − α 2 . The value of α is:
r
α=
2L
ω0
Using the fact that α = for resonant circuits (see chapter 2), we obtain:
2Q
Lω0
Q=
r
For the 8 MHz crystal described before, the different values are:
Resonant frequency: f0 = 8.186046961 MHz
Antiresonant frequency: f1 = 8.205757451 MHz
Giving a pulling range of ∆f = 19.710490739 kHz
The Q corresponding to the complex zeroes or poles is: Q = 90010. With
such high value of Q, the zeroes and the poles are practically:
z1,2 = −α ± jω0 and p1,2 = −α ± jω1 . The corresponding pole and zero plot is:

We remark that the complex poles and zeroes are practically superposed
and they have a very small real part.
 The magnitude of the impedance is shown below.

The phase of the circuit is:

The impedance is real at two frequencies. At the resonant frequency, the

impedance is real and is minimum at a value of 8 Ω. At the antiresonant
frequency, the impedance is also real and has a maximum value of 1.5 MΩ. At
frequencies below the resonant frequency and above the antiresonant frequency,
the crystal is capacitive. Between these two frequencies, the crystal is inductive.
The value of the impedance at the antiresonant frequency can be computed easily
from the pole and zero plot and is approximately given by:
1 C2
R≈ 1+ Q 2
2

C0ω1 C0
When we want to use a crystal in an oscillator, we can operate it at the
resonant frequency ω0. In this case, we say that we have a series mode oscillator.
The impedance of the crystal at this frequency is resistive and is minimal. In some
other cases, we can operate the oscillator at a frequency between ω0 and ω1. This
type of oscillator is called a parallel mode oscillator. In this case, the crystal is
inductive and we must use a capacitance CL in parallel with the crystal. In general,
crystals are designed to be used at given particular frequency. In this case, the
capacitance is specified. The frequency of oscillation is the parallel resonant
frequency given by the following circuit:

C r
L

Co

CL

We see that C0 is replaced by C0 in parallel with CL. This gives an

oscillation frequency of:
 C 
ωL ≈ ω0 1 + 
 C0 + CL 
A typical parallel mode oscillator is the Pierce oscillator. Its circuit is
shown below.

VDD

RL
RFC

Crystal Q1
CL

JFET_N_VIRTUAL
RG
C1 C2

Fig. 4- 15 Pierce Oscillator

Before starting the analysis of the circuit, we must remark that the FET is
clamp biased by the pair CL and RG. The RFC is an inductor that has a very high
value. It is used to bypass the dc current and acts as a very large impedance for
the ac current in the drain circuit.
Another point worth noting is that the circuit composed of the crystal and
the two capacitors C1 and C2 must produce a phase shift of 180° at the frequency
of oscillation to satisfy the phase requirement of the Barkhausen conditions. In
order to produce such phase shift, the crystal must be inductive. This means that
the frequency of oscillation is going to be inside the pulling range of the crystal,
i.e. between the resonant and the antiresonant frequencies. So, we can replace the
crystal by an inductance in the circuit of Fig. 4- 15.

VDD

RL
RFC

Q1
CL
L

JFET_N_VIRTUAL
RG
C1 C2

In this part, we are going to perform a small signal analysis of the above
circuit. The small signal equivalent circuit is shown below (Fig. 4- 16).

− g m 0 v (t ) RL C1 C2
RG v(t)

Fig. 4- 16 Small signal equivalent circuit

We have already studied the coupling circuit formed by the two

capacitances C1 and C2 and the inductance L at the end of chapter 2. It is the "pi"
circuit. Using the equivalence derived in chapter 2, we can replace the above
circuit by:
 N : 1

L C
g m 0v(t ) RL RG

ideal
C1C2 C
where C = and N = 2
C1 + C2 C1
g m 0v(t )
The voltage at the primary is: and at the secondary, we must
1 1
+
RL N 2 RG
have this voltage divided by N. So, the voltage at the secondary must be:
g m 0 v (t ) 1
× . This voltage must be larger than v(t) at the frequency of
1 1 N
+
RL N 2 RG
oscillation in order to have closed loop poles on the right half of the s-plane. This
implies that:
N 1 C 1 C1 1
g m0 ≥ + = 2 +
RL NRG C1 RL C2 RG