Harmonic Oscillator Project

(Adapted from old notes)

The focus for this project is an evaluation of the generation of sinusoidal oscillations
using (nearly) linear feedback circuits. The underlying principle involved is a bootstrap
process that may be described crudely as follows. The output of a unity-gain amplifier is
by definition a copy of the input signal waveform that produces the output. If this output
is fed back as a replacement for the input the amplifier presumably drives it self! The
input necessary to start the process initially, i.e. to provide the initial amplifier output, is
provided by inevitable thermal noise. Regenerative feedback maintains the oscillation.
These are however matters to consider more fully.
Discussion
There are three essential constituents of a practical harmonic oscillator:
a) An amplifier to sustain the oscillation, transferring energy (usually from a DC source) into the
oscillation to replace inevitable circuit losses and oscillation energy supplied to a load;
b) Circuit components that establish the frequency of the oscillation;
c) A mechanism to define the amplitude of oscillation.
If the oscillator circuit used were truly linear there would be no amplitude limiting mechanism other
than that associated with the average energy initially stored in the system. In a linear system signal
amplitude can be scaled (i.e. multiplied by a constant) without affecting frequency, and the amplitude
will be whatever the available energy will sustain. In a practical oscillator however there is necessarily a
continual infusion of energy into the oscillation by the amplifier, and signal amplitude grows until a
circuit non-linearity of one sort of another constrains further growth. Sometimes a controlled nonlinearity is introduced deliberately for this purpose, rather than depending on the generally unpredictable
effect of an inherent non-linearity such as amplifier saturation to limit oscillation amplitude
Negative-Resistance Oscillator
An example of a negative resistance circuit is shown in the figure to
the right. Assume an idealized opamp for simplicity, and because of
this neglect the voltage difference across the amplifier input terminals,
so that the voltage drops across R1 and R2 then are equal. With
negligible amplifier input-terminal currents the circuit input voltampere relation is as shown in the figure. The description of the input
resistance of this circuit as a negative resistance reflects comparison
of the volt-ampere relation with Ohms Law.. However note that the idealized opamp assumption (and,
although for different reasons, for that matter Ohms Law) has a limited range of validity in practice, in
particular the range of applicability of the negative resistance
relationship is constrained by amplifier saturation limits.
The resistance RS in the series RLC circuit to the right
represents circuit losses, primarily the inductor wire
resistance. This is compensated by adding an equal negative
resistance, i.e. a mechanism for adding energy to the circuit to
replace circuit losses. A PSpice netlist for a representative
design accompanies the figure.

Harmonic Oscillator Project

M H Miller

*Negative Resistance Oscillator

X741A
3
1 6 7 2 UA741
V+
6
0
DC 12
V7
0
DC -12
R1
1
2
1K
R2
3
2
1K
.PARAM RVAL = 1K
R3
3
0
{RVAL}
LS
4
1
10M
CS
5
4
1U IC=1
RS
5
0
1K
.LIB
EVAL.LIB
.STEP PARAM RVAL LIST 1K, 970, 700
.TRAN .03m 3m 0 10u UIC
.PROBE
.END

The series-tuned LC circuit of this negative resistance

oscillator involves two poles; the poles are both
negative real for over-damped operation, and complex
conjugate with negative real parts for under-damped
operation. For the special case of zero net resistance
the poles become an imaginary conjugate pair. Zero net
resistance is obtained in effect by adding power to the
circuit as rapidly as it is dissipated. Computed output
data for oscillation, and for the cases of over- and
under-damping is plotted in the figure following. Since
PSpice computations do not include thermal noise
ordinarily the circuit is energized by assuming an initial
charge/voltage on the capacitor.
The computed frequency of oscillation may be
compared to the expected (calculated) value obtained
from 2LC = 1. Note the initial amplifier saturation
that limits the increase of energy stored in the circuit.

Quadrature Oscillator
The analysis in the preceding illustration is done in the frequency domain, i.e. in terms of transfer
function poles and zeros. Another rather straightforward method of obtaining a sinusoidal oscillation is
by assembling a circuit whose operation in the time domain emulates a second-order differential
equation with constant coefficients; the natural solution of such an equation is sinusoidal. Actually it is
more practical to imagine the secondorder differential equation integrated
twice to obtain a technically more
convenient second-order integral
equation; assembling electronic
integrators generally is easier than
assembling differentiators. The
circuit diagram to the right illustrates
one such an oscillator configuration.

Harmonic Oscillator Project

M H Miller

The first stage to the left is an inverting Miller integrator. The other stage is a non-inverting integrator.
The opamp (assumed operating normally) maintains the non-inverting input voltage at vo/2 (where vo is
the second stage output voltage) A node equation at the capacitor node determines the current through
the capacitor is as shown. The integrating property of the second stage follows directly from this.
The output of the second stage is fed back to provide the input of the first, leading to a circuit with the
integral equation operating description desired. Note that because the output of one stage is the integral
of the output of the other stage both sine and cosine signals are available; hence the name 'quadrature'
oscillator. Design and analysis of an illustrative quadrature oscillator is left as an exercise.
Feedback Oscillators
A particularly productive method of studying harmonic oscillators is to view them as feedback
amplifiers for which the signal fed back from the output provides the entire input signal necessary to
produce the output. (This raises a question of how an oscillation that drives itself starts in the first place.
Although the question is fundamental the answer is rather straightforward; inevitable thermal motion of
atomic charge, i.e. electrical noise, provides an inherent startup signal. The more pertinent question to
consider is how the circuit must nurture this signal appropriately to sustain oscillation. The block
diagram below illustrates a basic feedback system. Si is an input signal applied to summing node
connected to an amplifier of gain GS; the amplifier output is So. An output sample fraction fSo is fed
back and subtracted from the input to produce the net amplifier input signal Si - fSo. The relationship
between Si and So is as shown.
For previous amplification
applications, we assumed that the
circuit designer made fGS >>1, i.e
designed for negative (degenerative)
feedback. For this case So/Si >1/f.
Suppose instead that a design makes
S
1+fG -> 0, implying the singular gain So/Si -> . Whereas for degenerative feedback the signal
sample is subtracted from the input the physical basis for this regenerative (positive) feedback is the
addition of the output signal sample to the input, i.e. to reinforce the input signal. An increase in input
signal strength further increases the output signal, and so causes still a larger reinforcing feedback
signal. The condition 1+fGS -> 0 corresponds to a theoretical feedback signal strength sufficient by
itself to produce the output level necessary to support the feedback signal; the operation becomes selfsustaining. However continuing growth is not a sustainable process; the amplifier ultimately is driven
into a saturated state for which the gain becomes essentially zero, and further signal growth stops.
The Barkhausen Condition 1+fGS -> 0 is interpreted as the condition for the onset of oscillation. This is
not really a 'bootstrap' affair; as a practical matter the oscillation starts because of signals generated by
random electron movement (i.e. currents) associated with thermal excitation. The circuit feedback,
appropriately designed, continuously reinforces a particular frequency component to form a selfsustaining output. The oscillation energy is obtained by conversion from an energy source that is an
integral part of the amplifier.
As noted before circuit non-linearity, whether inherent or specifically introduced, eventually must limit
the signal amplitude. Nevertheless a linear analysis is meaningfully applicable until the signal
amplitude is large enough for non-linearity to be significant, and so as stated may be applied usefully to
determining the conditions for the onset of oscillation. And, provided the nonlinear limiting is not too
Harmonic Oscillator Project

M H Miller

severe, a general continuity in nature suggests the results of the linear analysis can be 'close' in some
useful sense to the actual circuit performance.
Barkhausen Conditions
The unity loop-gain condition for the onset of oscillations (i.e. output = input) actually involves two
distinct requirements: (1) the magnitude of the net loop gain must be 1, and (2) the phase of the loop
gain must be 0 (or a multiple of 360). These two independent requirements together form the
Barkhausen conditions for the onset of oscillations.
In general for a particular linear circuit to support an oscillation the roots of the circuit determinant must
by definition have conjugate complex poles on the imaginary axis. Hence to make an oscillator we must
start with a circuit whose determinant involves at least two poles, and specify circuit parameters so that
these poles are placed properly on the imaginary axis. Unfortunately a circuit with just two poles is not
sufficient. The root locus for a two-pole system including loss simply does not cross the imaginary axis
whatever circuit element values used. At least one more singularity, either a pole or a zero, must be
present as a minimal requirement. Several oscillators meeting the minimal condition are studied here.
It can be noted that linear system may be scaled in frequency without changing the relative amplitudes
of circuit voltages or currents. Frequency is involved only as a factor in a product with either a circuit
inductance or capacitance, and only the product affects the voltage and current amplitudes. Hence the
condition of unity loop gain magnitude can be maintained while frequency is scaled arbitrarily; simply
scale inductance and capacitance by the inverse of the factor the frequency is scaled. It follows then that
it is the phase of the loop gain, and only the phase, which can determine the frequency of oscillation; the
oscillation frequency must be such that there is no net phase shift around the loop. Unity (or greater)
loop gain magnitude is necessary to initiate the oscillation and to replace energy losses, but this
requirement is quite separate from the determination of the oscillation frequency.

Harmonic Oscillator Project

M H Miller

(A) Tuned Circuit Oscillator

The parallel tuned circuit illustrated to the right is often
used in one form or another to satisfy the phase shift
condition for a sinusoidal oscillator; G represents
inevitable inductor (primarily) and capacitor losses
more than an intended circuit element. An expression
for the admittance Y looking into the circuit is shown to the right of the tuned circuit.
A simplified parallel tuned circuit oscillator is drawn to the left.
The amplifier on the left provides an adjustable output voltage,
of which a fraction 1/(1+RY) is returned to drive the amplifier; Y
is the admittance of the parallel combination of Go, L, and C.
Frequency selectivity is provided by the frequency-dependent
voltage divider.
A straightforward analysis provides

Note that Y contributes two poles and a zero to the expression, a minimal singularity requirement for
oscillation to be achievable. The Barkhausen oscillation initiation condition is that Vb = Va. Note also
that the expression is complex, i.e. s = j . The real and imaginary parts of the two sides of the
expression each separately must be equal, and so provide two requirements for oscillation. The
amplifier gain requirement is determined from the real part, and the frequency of oscillation from the
imaginary part.
Project: Design a tuned circuit oscillator for a nominal oscillation frequency of 10 kHz and nominal
peak amplitude of 6 volts or more. Use a inductor represented by a 10K resistance in parallel with 10
mH. Show explicitly how the individual Barkhausen conditions are met for your design and verify
performance expectations using PSpice. Plot the amplifier voltage output, and compare it to the voltage
across the tuned circuit. How does the improvement in waveform come about?
Note that the computer does not ordinarily provide thermal noise to initiate oscillation. Instead specify
an initial voltage across the capacitor to provide the start-up energy.

Harmonic Oscillator Project

M H Miller

(B) Phase-Shift Oscillator

Another circuit capable of sustaining oscillation is the RC phase-shift circuit that involves three poles.
A straightforward analysis can be made as a ladder development, i.e. note that the current into the
inverting amplifier through R is Va/R, from this calculate
the voltage across the capacitor in series with this R .
Obtain the equation

where as before s is used to represent j. As usual the

Barkhausen conditions are obtained by equating real and
imaginary parts of the two sides of the equation.
Project: Design a phase-shift oscillator for a nominal oscillation frequency of 10 kHz and nominal peak
amplitude of 6 volts or more. Use R = 1K. Show explicitly how the individual Barkhausen conditions
are met for your design and verify circuit performance using PSpice. Comment on the transient growth
of the oscillation until steady state is reached.
Note that the computer does not ordinarily provide thermal noise to initiate oscillation. Instead specify
an initial voltage across one of the capacitors to provide the start-up energy.

Harmonic Oscillator Project

M H Miller

(C) Colpitts Oscillator

The Colpitts circuit (drawn to the right) is another wellknown three-pole oscillator configuration. Note that
because the inverting input of the opamp is a virtual ground
the resistor r effectively shunts C2. Assume an idealized
opamp and as a customary simplification C1 = C2; verify
that the transfer function is

(To derive the transfer expression note that the voltage across C2 is -va/G., and work backwards to
calculate vb.)
The last term in the parentheses on the right in the denominator (coefficient of sCR) ordinarily can be
neglected (with capacitance values in microfarads, resistances in kilohms, and inductance in
millehenries the order of magnitude of the term generally can easily be made much smaller than 2). It is
not difficult to assure this design simplification over a wide range of oscillation frequencies, and in any
event it is at least useful as a way of estimating the oscillation frequency. Within this approximation
verify (explicitly) that oscillation occurs for 2LC = 2, for an amplifier gain magnitude G 1+ (R/r).
Project: Design a Colpitts oscillator for a nominal oscillation frequency of 10 kHz and nominal peak
amplitude of 6 volts or more. Use L = 10 mH. Show explicitly how the individual Barkhausen
conditions are met for your design and verify circuit performance using PSpice. Observe the amplifier
voltage output, and compare it to the voltage across C2. How does the improvement in waveform come
about?
Note that the computer does not ordinarily provide thermal noise to initiate oscillation. Instead specify
an initial voltage across the capacitor to provide the start-up energy.

Harmonic Oscillator Project

M H Miller

(D) Hartley Oscillator

The oscillator circuit drawn to the right is the dual of the
Colpitts oscillator; the roles of the inductor and the
capacitors in the Colpitts circuit are interchanged to obtain
the Hartley circuit. Assume an idealized opamp and, as is
usual, L1 = L2); verify that the transfer function is

To derive the transfer expression note that the voltage

across L2 is va/G.
The last term in the parentheses on the right in the denominator ordinarily can be neglected (with
capacitance values in microfarads, resistances in kilohms, and inductance in millehenries the order of
magnitude of the term generally can easily be made much smaller than 2). It is not difficult to assure
this design simplification over a wide range of oscillation, and in any event it is at least useful as a way
of estimating the oscillation frequency. Within this approximation verify (explicitly) that oscillation
occurs for 2LC = 1/2, for an amplifier gain magnitude G>1+ (R/r). Note that the frequency of
oscillation is half that of the Colpitts circuit for the same L and C values.
Design a Hartley oscillator for a nominal oscillation frequency of 10 kHz and a nominal peak amplitude
of 6 volts or more. Use L1 = L2 = 10mH. Show explicitly how the individual Barkhausen conditions
are met for your design and verify circuit performance using PSpice. . Observe the amplifier voltage
output, and compare it to the voltage across the tuned circuit. How does the improvement in waveform
come about?
Note that the computer does not ordinarily provide thermal noise to initiate oscillation. Instead specify
an initial voltage across the capacitor to provide the start-up energy.

Harmonic Oscillator Project

M H Miller

(E) Transfer Function Oscillator

The active filter circuit drawn to the right provides another
illustration of a circuit that can be made to oscillate. (The
circuit is one used elsewhere to illustrate two-pole pulse
response in general.) A straightforward transfer function
analysis obtains the relationship

Equating real and imaginary parts of the two sides of the equation obtains the two Barkhausen
conditions G = 3 and RC = 1.
Design an oscillator using this circuit for a nominal oscillation frequency of 10 kHz and nominal peak
amplitude of 6 volts or more. Show explicitly how the individual Barkhausen conditions are met for
your design and verify circuit operation using PSpice.
Note that the computer does not ordinarily provide thermal noise to initiate oscillation. Instead specify
an initial voltage across the capacitor to provide the start-up energy.

Harmonic Oscillator Project

M H Miller