In the world of crystal
oscillators there are a myriad of design variations, with
an even wider spectrum of applications. Experience teaches
circuit designers that it’s always wise to consider
every factor influencing design complexity and reliability,
whether you’re designing an oscillator for a one-shot
application or a production run of thousands.
In deciding which type of oscillator
is appropriate, four primary factors must be considered: the
type of application (fixed or modulated); the frequency of
operation; the load that the oscillator will output into;
and the parts available.
The basic requirements for any functional
oscillator circuit are: 1) a phase shift through the oscillator
loop (consisting of an amplifier, reactive components, and
the crystal) of either 0 or 360° ; 2) an open loop gain
of greater than 1.0 at the operating frequency; and 3) a "negative
resistance" generated by the circuit greater or equal
in magnitude to the equivalent series resistance of the crystal.
Below are the basic design criteria for
the most popular types: both IC and transistor based Pierce,
Colpitts, and Common Base crystal oscillator circuits.
In many instances Pierce-type IC oscillators
may be appropriate. Stand-alone or incorporated into a complex
device, this configuration uses a simple inverter to provide
180° of phase shift, with the additional 180° supplied
by two "pi" capacitors. Loop gain is optimized by
specifying a lower output capacitance value than that of the
input. The crystal parallel-resonates
with the series combination of capacitors as the load, and
overtone operation can be accomplished by incorporating a
mode trap at the front end (see Fig. 1). The trap should be
tuned midway between the desired overtone frequency and its
predecessor. A resistor connecting the input and output of
the circuit adds linearity to the gate and adjusts amplification.
It can also be used for third overtone selection, negating
the need for the mode trap.
The simplest single pair inverter (equiv.:
Harris 74HCU04) is most desirable, as it more easily achieves
linearity, reduces the possibility of latch-up, and also provides
the fastest speed - limited only by the propagation delay
characteristics of the fabrication technology. All standard
IC inverters have a significant disadvantage: their amplifying
characteristics aren’t as easily fashioned, as input
impedances are fixed and output impedances are at best "negotiable."
Bandwidth can be adjusted to a limited extent by changing
the feedback resistor (usually from 2.2K to 100K). For optimum
performance and repeatability, these inverters shouldn’t
be utilized beyond their 3db point.
Supply-line bypassing techniques are
important when utilizing IC oscillators. Since most have a
relatively narrow linear region, their amplifying characteristics
can easily be affected by supply noise during system power
up. Neglecting to adequately bypass voltage transients will
result in: a lack of oscillator start, non-crystal controlled
oscillation (self-oscillation), or crystal-controlled oscillation
modulated by self-oscillation frequencies.
If the design is intended to drive circuits
or devices other than members of the standard digital logic
families, or if the application requires a more sinusoidal
output, the designer must consider alternative discrete-based
oscillator configurations. Completely adaptable and better
suited to load matching, a transistor-based oscillator is
more complex to design but significantly increases flexibility.
The simplest form of a Pierce-type transistor
oscillator (see Fig. 2) mirrors the IC design, with the amplifier
providing 180° of phase shift, and the opposing 180°
provided by the pi capacitors and the crystal. With amplifier
design following conventional rules for input and output impedance,
and since the transistor-based Pierce type is primarily applied
in a common-emitter configuration, it is more suitable for
driving medium to high impedance loads. Primarily utilized
for frequencies above 10M H z, as crystal drive is normally
quite high, sizable voltage gain provides a sinusoid output
of respectable amplitude. These characteristics can be advantageous
for high frequency applications, with an abundance of power
available for driving high impedance overtone modes. Excess
drive will cause unwelcome performance in the crystal however,
potentially exciting spurious & coupled mode responses.
The negative factor in this configuration
is the transistor’s base-collector capacitance ¾
which can be rather high if care isn’t exercised in
transistor biasing ¾ lumped in parallel with the crystal.
Loop gain is reduced with energy diverted through this capacitance,
and amplifier bandwidth is altered, complicating proper mode
selection. As neither side of the crystal is at
ground potential, phase jitter is characteristically higher
than in configurations such as the Colpitts oscillator.
The Colpitts configuration (see Fig.
3) places the crystal from the base of the transistor to ground.
As it employs an emitter-follower, the bulk of the phase shift
is provided by the capacitors, with the balance provided by
the feedback path through the crystal. The crystal is parallel-resonant
with the series combination of the Colpitts capacitors. Loop
gain is optimized with a base-emitter capacitor lower in value
than the emitter-ground capacitor. In applications above 70
M H z however, it is often beneficial to reverse this ratio,
as additional signal feedback to the base may be required.
The Colpitts can be utilized in either
a common-emitter or common-collector configuration. This allows
compatibility whether you are driving high, medium, or low
impedance applications. In the common-collector configuration,
the resistor between the collector and source voltage is normally
eliminated for optimum power transfer to the load. The load’s
equivalent impedance should also be considered when arriving
at the physical value for the emitter-ground capacitor.
A well-designed Colpitts oscillator can
be utilized through the K H z range, up to approximately 130M
H z. Overtone operation is achieved by adding an inductive
component in parallel with the emitter-ground capacitor. Common-emitter
wiring will provide the best high frequency performance as
it minimizes the effects of and variations in load characteristics.
The transistor’s Miller capacitance also has a reduced
impact in the common-emitter configuration.
The common-base oscillator configuration
is the most complicated to design. The version shown in Fig.
4 is designed for frequencies above 130MHz, using crystals
operating on and above the fifth overtone. As the name infers,
it follows conventional common-base rules, providing mid-range
voltage gain and sinusoid output. The operating frequency
is set by the resonant frequency of the reactive components
in the collector arm of the circuit. Output signal amplitude
is adjusted by altering the ratio between capacitors on the
output. Input impedance is low, a disadvantage at high frequencies
(the in-circuit Q of the crystal is reduced). Medium output
impedance characteristics allow it to drive a wide range of
The crystal operates in a series-resonant
condition, requiring less gain margin from the amplifier.
Since the crystal comprises the primary feedback path for
oscillation, there is a resulting loss of loop gain from power
diversion through its holder capacitance. This configuration
is also more susceptible to stray, temperature, and component
tolerance effects, as the inductive component is used to set
the operating point instead of for band/mode trapping.
Design & Analysis:
Historically, designing oscillators was
accomplished through trial-and-error, with innumerable iterations
of breadboarding and testing. Now, with computerized circuit
simulation, designers have the ability to be exacting in the
design exercise - leaving breadboarding as merely a final
working model of their concept.
The key to reliable crystal oscillator
design is concise analysis and circuit modeling under both
open loop (without crystal) and closed loop (with crystal)
conditions. Correlated with carefully measured data, this
analysis aids in reducing breadboard iterations and increasing
first-time accuracy. Physically, the key performance element
is the interface between the crystal andcircuitry. A well
designed and constructed interface produces a reliable oscillator.
There are two useful techniques for design
and analysis of crystal oscillators. The first technique involves
analysis and measurement of the gain and phase response of
the circuit. A broad band analysis is performed with the crystal
removed from the circuit to identify potential conditions
for oscillation outside the intended frequency. A second broad
band sweep with the crystal included is used to ensure the
leads or holder capacitance aren’t inducing an undesirable
response like an
LC mode. A narrow band sweep is used to indicate proper conditions
at the intended oscillation point. A well designed oscillator
has a characteristically steep slope on the phase crossing
at 0 or 360° , and a sufficient gain margin (³ 2dB
– refer to Fig. 5). A steep phase slope and sufficient
gain create an oscillator more resilient to change in frequency
and output amplitude relative to temperature and voltage variations.
For open loop gain phase analysis, a common method is opening
the AC feedback loop, reflecting the impedance at the open
The second analysis technique involves
measuring negative resistance. In general, measurement becomes
easier when one side of the crystal is tied to circuit ground,
as with a basic Colpitts design. Offering practical advantages
over the gain/phase technique, negative resistance facilitates
a quicker one-port measurement for comparison with computer
broad and narrow band negative resistance measurements on
IC oscillators with a "floating" crystal (i.e: Pierce-type)
are more difficult.
The fundamental strength of the negative
resistance technique is its ability to analyze and measure
the "real" impedance of the circuit and the crystal.
The designer also needs a solid understanding of the crystal’s
equivalent circuit parameters and their effect on the circuit.
With these details in hand, it is a simple matter to measure
the "imaginary" portion of the circuit and the crystal.
By overlaying these four pieces of data the designer can predict
whether or not the circuit will oscillate, where the circuit
will oscillate, and what the design margins are.
The ability to closely correlate
analytical versus measured data is no easy task. A solid understanding
of active and passive devices and their equivalent circuits
is required. A large volume of work and an understanding of
key parameters is mandatory to develop computer models and
techniques that accurately predict results. Once the models
and measurement have been refined so theoretical data correlates
with measured data, the designer has the tools necessary to
decrease the burden and time required to design crystal oscillators.