H

Fundamentals of
Quartz Oscillators
Application Note 200-2

Electronic Counters Series

1 1
Z 49° 35° Z

BT AT

Z1
57° 38°
X FT CT Y
52°
DT

Y X Y
Cut Cut X

1
 Table of Contents

Introduction .................................................................................. 3
Crystal Fundamentals ................................................................... 4
Piezoelectricity ....................................................................................... 4
Crystal Structure .................................................................................... 4
Crystal Cuts ............................................................................................. 5
Vibration Modes ..................................................................................... 6
Frequency Determination ..................................................................... 7
Crystal Mounting .................................................................................... 7
Electrical Equivalent ............................................................................. 8
Influences on Crystal Oscillator Frequency ............................... 10
Temperature Effects .............................................................................. 10
Time ......................................................................................................... 11
Long-Term .......................................................................................... 11
Short-Term ......................................................................................... 12
Drive Energy ........................................................................................... 13
Gravity .......................................................................................................... 13
Shock ............................................................................................................ 14
Vibration ....................................................................................................... 14
Electromagnetic .......................................................................................... 14
Retrace ......................................................................................................... 14
Typical Performance of
Oscillator Compensation Techniques ......................................... 15
Temperature ........................................................................................... 16
Time ......................................................................................................... 18
Long-Term .......................................................................................... 18
Short-Term ......................................................................................... 19
Line Voltage ............................................................................................ 19
Warm Up .................................................................................................. 20
Oscillator Influence on Measurement Accuracy ........................ 22
Frequency Counters ............................................................................... 22
Frequency Synthesis .............................................................................. 23
Appendix A .................................................................................... 24
References ..................................................................................... 25

2
Introduction

The accuracy of a generated frequency depends upon the reference

element selected. The purpose of this note is to provide a background
on the crystal reference element and its impact in an oscillator circuit.
This information is then used to explain the effects of the oscillator on
the accuracy of a frequency measurement and frequency generation.

While there are numerous crystalline substances which have the basic
requirements of a reference element, quartz, due to its many desirable
characteristics, has become the most widely accepted. A quartz refer-
ence element was first used in an oscillator circuit in 1920, some 40
years after the discovery of piezoelectricity. Since that time, improve-
ments have been and are still being made in converting the raw quartz
crystal into a usable reference element (Ref. 5). These changes in
processes have resulted in the discovery of behavioral changes of the
reference element. These behavioral aspects influence the oscillator
and as such warrant considerable discussion.

The basic application for the quartz resonator is to connect it in a

manner such that the mechanical vibrations stabilize the oscillator’s
frequency. This is possible since the crystal acts like a tuned circuit
when placed in an amplifier feedback arrangement. The electrical
equivalent of the quartz plate is shown in Figure 1. When connected
into an amplifier circuit, Figure 2, a small amount of energy is fed back
to the crystal which causes it to vibrate. These vibrations act to stabi-
lize the generated frequency at the resonate value. It is at this point that
the user of this generated frequency must decide upon the usefulness
of the source for the application. A detailed look at crystal
oscillator fundamentals can help make that decision.

C
0

Figure 1.
Equivalent circuit
of a crystal
includes the
capacitances
contributed by the
Schematic Symbol Equivalent wire leads and
the holder in C0.

Feedback

AMP

Figure 2.
Simplified
amplifier feedback
(oscillator) circuit
using a crystal
resonator.
3
 Crystal Fundamentals

Piezoelectricity
Piezoelectricity is the primary property of a crystal which makes it
usable as a resonator. Piezo is derived from the Greek word piezin and
means “to press.” Piezoelectricity as defined by Cady (Ref. 3) is
“electric polarization produced by mechanical strain in crystals
belonging to certain classes, the polarization being proportional to the
strain and changing sign with it.” This electric polarization can be
produced by strain such as bending, shear, torsion, tension, and
compression on a piece of quartz. The electric polarization provides a
source of electromotive force (voltage). Additionally, the inverse
effect can be created, i.e., a voltage applied across the crystal produces
mechanical movement (Figure 3).

Pressure

+
Quartz Quartz Voltage
–

Pressure
Applied pressure Applied voltage Figure 3.
generates voltage causes contraction Piezoelectricity
principles.

Crystal Structure
The quartz crystal is formed from silicon and oxygen (Si02). Its charac-
teristic form is a result of the unit cells by which the crystal grows.
These unit cells are identical and consist of atoms arranged in
a repetitive geometric pattern. Quartz crystals have a three dimen-
sional geometric body (Figure 4). Most the physical properties of a
crystal are anistropic (direction dependent), therefore, changes
during the growth of the crystal which affect anistropy result in crystal
imperfections. A change in the piezoelectric coefficient, for example,
will create a boundary across which the sign of the charge differs when
strain is applied. This twin boundary (referred to as twinning) prevents
the crystal piece from resonating, making it unsuitable for an oscillator
reference unit. Since a considerable amount of work is involved in
making a good crystal resonator, these defects should be detected
early. Crystal orientation and the presence of defects such as twinning
and fractures are detected through the use of polarized light, X-rays,
and chemical etching.

The major axis of quartz growth is called the optic axis. This axis is not
anistropic to light, therefore light passes readily. For the purpose of
cutting pieces of quartz to act as resonators, the optic axis is labeled
the Z axis in an orthogonal X, Y, Z coordinate system. A quartz crystal
having 6 sides has three separate X axes and three Y axes definable at
120° increments about the Z axis. The Y axes are perpendicular to the
prism faces while the X axes bisect the angles adjacent to the prism
faces (Figure 4).
4
 Z

X
Y

Y
X

Figure 4. Doubly
terminated quartz
crystal showing
axis orientation.

Crystal Cuts
A small piece of quartz material is obtained by cutting the crystal at
specific angles to the various axes. The choice of axis and angles
determine the physical and electrical parameters for the resonator. For
example, an X plate crystal, one which is cut with its major face normal
to the X axis (Figure 5), has a relatively large voltage generated when
compressed and decreases in frequency with temperature increases.
A similar voltage can be generated by a pure shear stress on a Y cut
plate. The Y cut, however, exhibits a positive temperature coefficient.
Numerous other cuts can be made simply by changing the angle and the
axis of reference.

Z Optic Axis

X Electrical Axis
Rotated
Plate
X Plate
Figure 5. A
Y Plate description of the
Y Mechanical Axis axis of quartz and
X, Y, and rotated
plates.

Figure 6 shows combinations of X, Y, and Z rotational cuts which are

arbitrarily labeled such as AT, BT, etc. These more common cuts
provide for various trade-offs between electrical-mechanical capability
and temperature coefficient.

5
 1
Z1 49° 35° Z

BT AT

Z1
57° 38°
X FT CT Y
52°
DT

Y X Y
Cut Cut X

Figure 6. Typical
crystal cuts from a
doubly terminated
Z quartz crystal.

Vibration Modes
When a piece of crystal is subjected to a voltage, a stress is produced.
If the voltage is caused to alternate at the proper rate, the crystal will
begin vibrating and produce a steady signal. The mode of vibration
depends upon the way the crystal was cut, i.e., an X cut exhibits an
extensional vibration mode whereas the AT, which is cut at 35 degrees
off the Y axis, vibrates in the thickness shear mode. An illustration of
the various vibration modes is given in Figure 7. The vibration set-up
in the quartz crystal may produce both harmonic and nonharmonic
signals and overtones. The harmonic overtones are desirable since
they allow the production of higher frequency crystal resonators using
essentially the same cut. Nonharmonic overtones, on the other hand,
are undesirable as they may lead to the generation of unwanted signals
at frequencies spaced close to the one desired. When unwanted signals
occur, they could also change with environmental influences. The
various vibrations may then cancel, causing the crystal to stop resonat-
ing. This phenomenon is termed an activity dip since the crystal
activity stops and starts due to a changing environment. Crystals
having unwanted signals could also shift from one resonate point to
another producing a frequency jump which would be an undesirable
effect.
6
 Fundamental

Third Overtone
Extensional Figure 7.
Fifth Overtone Vibration modes
of various crystal
Thickness Face Shear cuts and the
Shear thickness shear
overtone.

Frequency Determination
The frequency, or rate of vibration, is determined by the cut, size, and
shape of the resonator, e.g., a 10 MHz 5th overtone unit is only 1.2
centimeters in diameter and about 1.06 mm thick. The primary fre-
quency determining factor for the AT and BT cut is thickness since they
vibrate in the thickness shear mode. The precision with which the
thickness is controlled determines the variation from crystal to crystal
from a nominal center frequency. Final adjustment of the center
frequency is sometimes accomplished by plating small amounts of gold
on to the quartz. A monolayer (one atom thick) of gold can change the
frequency by 2 parts in 107. Circular crystals of the thickness shear
vibrating mode, when designed with the proper radius of curvature at
the center, will produce “clean” frequencies with no spurious tones. It
is for this reason that high performance crystal oscillators will typically
utilize highly polished and properly shaped quartz resonators. In fact,
these crystals are honed to a surface finish which is 10 times finer than
used for prescription eye glasses.

Crystal Mounting
The supporting structure and the means used to obtain the electrical
contacts is dictated by the vibrating mode. Care must be exercised in
mounting to avoid placing a strain on the crystal. The support must not
become a part of the resonator since it would absorb energy and could
cause an activity dip. A thickness shear mode crystal is supported by
the edges at an approximate null or zero node to avoid interfering with
the vibration. A typical 1 MHz crystal, due to its size, might be held by
tension wires at several points about the surface edge. A 10 MHz unit
being smaller could be supported at two points on a ceramic header.
Other support types are shown in Figure 8.

Cemented Lead Type Pressure Mounting Solder Lead Type Strap Type

Figure 8. Methods of mounting crystals depending upon cut and application.

7
 Once the crystal is mounted, a suitable encasement is selected. The
encasement reduces the effects of contamination, humidity, and
atmospheric changes. Glass has been used for many years since it is
easy to work when evacuation and inert gas backfill are required. A
newer technique is a cold weld copper lid over a ceramic header. This
method provides a cleaner environment and allows for uniform heat
distribution.

Electrical Equivalent
The electrical equivalent (Figure 9) provides the link between the
physical property of the crystal and the area of application, the
oscillator.

C0

Figure 9. The
electrical
C1 L1 R1 equivalent circuit
for a crystal
resonator.

The physical constants of the crystal determine the equivalent values

of R1, C1, L1, and C0. R1 is a result of bulk losses, C1, the motional
capacitance, L1 is determined by the mass, and C0 is made up of the
electrodes, the holder, and the leads. When operated far off resonance,
the structure is simply a capacitor C0 but, at the precise resonant
frequency the circuit becomes a capacitor and resistor in parallel. The
reactance of the crystal approaches zero at the point of series reso-
nance and reaches a maximum at the antiresonant frequency fA
(Figure 10).

Anti-resonance
Area of Usual
“Parallel
Resonance”
+
Reactance

fs
0
Series
Resonance
–
Frequency Figure 10. The
reactance of the
fA crystal varies with
1 the frequency of
2π fc0 operation near
resonance.

8
An area typically chosen for operation of the oscillator is either near
series resonance or at the more inductive area of parallel resonance.
The series resonant circuit (Figure 11) utilizes the characteristics of
the crystal where the reactance is just slightly inductive (above fs
Figure 10). Series capacitance is then added to obtain a tuned circuit.
The series capacitor is typically adjustable so that the phase of the
feedback can be changed slightly thus fine tuning the oscillator fre-
quency. The parallel resonant mode adds capacity in parallel or across
the crystal (Figure 12). This circuit typically operates highest on the
reactance curve, hence the crystal reactance is more inductive.

Figure 11. The Series Resonant Figure 12. The Parallel Resonant
Oscillator Circuit

9
 Influences on Crystal Oscillator Frequency

The desired result of the crystal and its associated oscillator circuit is a
precise frequency. The crystal frequency, however, is determined by
the thickness, density, elasticity, molecular changes, and area of
resonance over which the quartz plate is operating. Since these factors
are influenced by temperature changes, time, drive energy, and other
environmental conditions, it is only reasonable to expect that these
factors will influence the crystal oscillator frequency.

Temperature Effects
A major influence on the crystal frequency is that of operating over
variations in temperature. An oscillator exactly on frequency at
25° Celsius with a frequency variation of 5 parts per million (ppm)
per degree Celsius change could experience a frequency offset of
25 ppm with only a 5°C temperature rise. The amount of frequency
variation is due to the crystal temperature coefficient, and therefore,
depends upon the crystal cut. Graphs of temperature vs frequency for
various cuts are given in Figure 13.

0
–75
∆f PPM

–150
BT Crystal Cut
–225
–300
20 30 40 50 60 70 80 90 100 110 120 130140
Temperature in C°

+80 +100

+40
0
∆f PPM

∆f PPM

0
–100
–40 DT Crystal Cut
CT Crystal Cut
–80
–40 –20 0 +20 +40 +60 +80 +100 –55 0 +25 +85
Temperature in C° Temperature in C°

Figure 13. Frequency vs. temperature plots for DT, BT, and CT cuts.

The curves for some crystal cuts remain relatively flat over a limited
temperature range. Others have both positive and negative frequency
excursions creating an “S” shaped curve. Changing the angle of cut by
only a small amount can limit the excursions and make the slope less
steep (Figure 14). Obviously, it is not possible to completely avoid
these frequency variations if the crystal is to be used over a wide
temperature range. Therefore, other techniques must be used to reduce
this effect. This subject will be considered in a later section on com-
pensating techniques.

10
 +50
+40
A
+30
+20
+10
∆f PPM

B
0
–10
–20 Figure 14.
C A, B, and C are
–30 temperature vs.
–40 frequency plots of
–50 AT cuts which
–50 –40 –30–20 –10 0 +10 +20 +30 +40 +50 +60 +70 +80+90 have been varied
Temperature in C° by a few seconds
of angle rotation.

Time
Frequency variations relative to time are indicative of oscillator
stability. Stability is usually expressed as the fractional frequency
change over a period of time, i.e., long-term or short-term.

Long-Term
A gradual change in frequency over days or months is known as aging.
This occurs for various reasons, e.g., the physical properties of the
crystal mounting may change. The crystal coefficient of elasticity
changes when subjected to stress, or when trapped gasses escape, or
when contaminants attach to or leave the crystal surface. Aging occurs
at a relatively constant rate per decade for each crystal (Figure 15).
Therefore, to maintain an accurate frequency, periodic oscillator
adjustments must be made to remove the effects. Generally, the
frequency of an oscillator can be varied a few cycles by a slight change
in the phase of the feedback signal. This change is usually accom-
plished by an adjustable capacitor. A 10 MHz oscillator with an adjust-
ment range of 20 Hz can be corrected for 75,000 + hours (9 or 10 years)
of aging at a 5 × 10-10 per day rate.

0.30

0.25
PPM Change

0.20

0.15

0.10 Figure 15. Time

domain stability
of the fractional
frequency change
over time (days)
5 10 15 20 25 starting from a
Days point of
calibration.

11
 Short-Term
Short-term rms frequency variations or time domain stability is a
measure of the frequency or phase noise. This is specified as the
standard deviation of the fractional frequency fluctuations for a
specific averaging time. These small frequency changes are essentially
superimposed upon the aging curve (Figure 16).

Short Term
Stability

0.30

0.25
PPM Change

0.20

0.15

0.10 Figure 16. Short-

term time domain
stability or the
fractional frequency
change over time
5 10 15 20 25 (seconds) and its
Days relationship to
aging.

Time domain stability is typically specified for a one-second average.

Shorter or longer averaging times may be required in the accuracy
computation for some applications. The manufacturers of high perfor-
mance oscillators will usually include measurement from 10–4 to 102
second averages in power of 10 increments (Figure 17). Time domain
stability is related to the frequency domain measurement termed phase

Time Domain Stability

10–9
X
Short Term Stability σy (2, τ, B)

–10 X
10

10
–11 X
X
–12 X
10
X
10–13

–14 Figure 17. Time

10 domain stability
–3 –2 –1 0 1 2 3
10 10 10 10 10 10 10 (short-term) for
Averaging Time τ Seconds specific averaging
times.

12
spectral density. This measurement is sometimes provided in lieu of
the time domain stability. Phase spectral density is related to the signal
to single sideband phase noise ratio normalized to a 1 Hz bandwidth at
various offsets from the carrier frequency (Figure 18). These two
techniques and the correlation between them are discussed in a
number of application notes and technical papers (Ref. 2, 11) Appendix
A6, 7, 9, 10, and 14.

X Estimate of Mean Value

–80
SSB/Carrier Phase Noise (dB/Hz)

–90 X σ of Measurements
–100
X
–110
X
–120
X
–130
–140
Figure 18. Phase
1E-01 1E 00 1E 01 1E 02 1E 03 1E 04 spectral density at
Offset from Carrier (Hz) specific offsets
from the carrier.

Drive Energy
A quartz crystal is analogous to a mechanical block. The clock relies
upon the main spring energy to keep the pendulum going. The crystal
requires energy to sustain the mechanical vibration which in turn
maintains the piezoelectric action. The resonate frequency of the
oscillator will change with a variation in the drive energy. The fre-
quency of an AT cut crystal will change by 1 × 10–9 with a variation in
drive of one microwatt. The drive level requirements will, of course,
vary depending upon the crystal impedance. Excessive drive level may
cause the mechanical vibrations to exceed the quartz elastic limits
resulting in a fracture. The usual operating point is for minimum
amplitude drive level since this is where maximum oscillator stability
is achieved.

Gravity
The earth’s gravitational force causes a stress related frequency effect
which varies with the physical orientation of the crystal oscillator, i.e.,
a calibrated oscillator in one orientation will change frequency when
rotated to another orientation. Typical offsets for 180° of rotation may
be in the order of 2 × 10–9 (Figure 19) which is 1 × 10–9 per G. This
effect is also experienced when crystals are subjected to acceleration.

13
 Shock
Striking a crystal oscillator places a sudden stress on the crystal by
temporarily deforming the mounting structure. Shock can result in a
change of frequency by 1 × 10–9 per G. While not normally subjected to
such extremes, a crystal oscillator should be able to withstand a 30G
1/2 sine wave shock for 11 milliseconds without permanent damage
(Figure 19).

Temperature Turn
Vibration Off On 180° Rotation
Change
Shock
Aging
∆f
f

Retrace
Figure 19. Graphic
representation of
TIme environmental
conditions.

Vibration
Shaking the crystal oscillator also causes stress in the crystal. These
variations are typically of a longer duration but less severe than shock.
Long-term, the effects of vibration tend to the average frequency,
however, for short-term applications, the frequency can be expected to
change by 1 × 10–9 per G (Figure 19).

Electromagnetic
Electromagnetic interference comes from sources located physically
close to the oscillator. These effects are a result of signals being
coupled into the electronics of the circuit rather than on the crystal.
Careful design, like the use of torroids in the oscillator and placing the
oscillator away from fans and transformers, etc., will reduce the effects
to insignificant values. Care should also be exercised in placement of
measurement instruments using oscillators since large signals, i.e.,
from a radio transmitter, could couple into the oscillator.

Retrace
Turning on an oscillator begins the generation of a usable frequency,
which as explained earlier, changes with time (aging). When an oscilla-
tor is turned off and then back on, it will not necessarily start at the
same frequency at which it had been operating. Eventually the oscilla-
tor will begin to age at its previous rate but will most likely be offset
slightly from its original frequency. A typical retrace offset may be in
the order of 1 × 10–8 (Figure 19). Another offset error which is similar
to retrace is that of temperature hysteresis, i.e., an oscillator whose
temperature is raised by several tens of degrees and then lowered may
not return to the same start frequency.

14
Typical Performance of
Oscillator Compensation Techniques

There are three crystal oscillator configurations which are in common

use for instrument frequency sources. These configurations attempt
to compensate for variations due to crystal characteristics primarily
that of the temperature coefficient. The three types are (1) room
temperature crystal oscillator (RTXO), (2) temperature compensated
crystal oscillator (TCXO), and (3) oven controlled temperature crystal
oscillator.

The RTXO typically uses a hermetically sealed crystal and individual

components to build the oscillator circuit. The TXCO encases the
crystal, the temperature compensating components, and the oscillator
circuit in a container. The oven controlled approach adds a heater and
the heater control to the oscillator circuit and puts the temperature
influenced elements in a thermally insulated container.

Selecting the right oscillator for the job requires a comparison of the
specifications between the various oscillator types and determining
how this affects the measured results. The user can do little in the way
of changing the crystal structure, oscillator design, and oven circuitry
but by comparing specific performance an intelligent choice can be
made. The major specifications to be considered for each type oscilla-
tor are temperature, time (aging and short-term stability), line voltage,
and warm up.

Before embarking on a comparison program, it is wise to define the

units to be used. Since the main objective is an accurate frequency,
variations from this can be defined as a fractional frequency error, ∆f/f,
where ∆f is the frequency error and f is the nominal frequency. Frac-
tional error is a dimensionless number and can, therefore, be used to
describe variations due to time, temperature, voltage, etc. The total
variation is the sum of the individual variations.

15
 Fractional frequency error will generally be given as ≤ or ± to signify
a band of errors, i.e., ±1 × 10–7. Figure 20 shows various other ways
in which errors can be specified and a conversion chart for quick
reference.

Parts per 0.001 0.01 0.1 1 10 100 1,000 10,000

Million 2 3 4 56789

(PPM)
0.0000001 0.000001 0.00001 0.0001 0.001 0.01 0.1 1
%

0.000000001 0.00000001 0.0000001 0.000001 0.00001 0.0001 0.001 0.01

Decimal

10–9 10
–8
10
–7
10
–6
10
–5
10
–4
10
–3
10
–2
Powers
of 10
1 1 1 1 1
Etc. 1,000,000 100,000 10,000 1,000 100
Fraction

Figure 20. These factors are specified in various ways: parts per million (PPM),
percent, as a deviation from a nominal frequency expressed as some power of 10,
or as a fraction.

Temperature
RTXO: Room temperature crystal oscillators use crystals which have
been manufactured for minimum frequency change over a change in
temperature. This is accomplished primarily through the choice of the
crystal cut and finishing process.

The frequency variation for an AT cut may be less than 2.5 ppm,
2.5 × 10–6, over a 0 to 50 degree range (Figure 21). Since the user is not
informed as to how the crystal actually acts on a per degree basis, it
must be assumed that the entire variation could occur over a small
portion of the temperature range. Thus, the 2.5 × 10–6 specification
could be reached by the changes in ambient temperature inside an
instrument.

+2.5
PPM Change

Figure 21.
–2.5 Frequency
temperature plots
0 50 for several room
Temperature in °C temperature
oscillators.

16
TCXO: The temperature compensated oscillator uses components
external to the crystal to offset the temperature effects. This could be
in the form of a relatively simple circuit, such as capacitors, ther-
mistors, etc., having opposite temperature coefficients, or a series of
compensating elements and an amplifier used to control a voltage
tuned capacitor which is part of the oscillator circuit. The individual
compensating elements may only provide for limited correction over
portions of the temperature range. This can result in several small
frequency excursions over the entire operating range as evidenced in
an actual temperature vs frequency plot (Figure 22). The TXCO
typically has a temperature characteristic which is 5 times better than
an RTXO or less than 5 × 10–7 for a 0–50° change.
Midnight
Noon

Noon
8 am

4 pm

8 pm

4 am

8 am

2 pm
4 pm

8 pm
10 am
start
temp 44°C to 25°C
cycle

1 × 10–7
4 hours

0°C +2°C +9°C +16°C +22°C +29°C +38°C +44°C

Figure 22. Actual plot of a TCXO over 0–44° change and the excursion
encountered for a rapid change from 44° to 25°.

Since the TCXO may have nonlinear performance, the unit may require
an offset in frequency at some temperature in order to maintain the
specification over the entire range. Figure 23 illustrates how the
frequency must be offset by –5 Hz at 25 degrees Celsius to meet the
manufacturer’s specification for a particular TCXO. The offset number
for each TCXO is typically printed on the side of the case. The user
who always operates at a specific temperature should adjust the
oscillator frequency to remove this bias effect since it could result in
different values from two similar products.

10 MHz + 5 Hz

Offset
Removed
Frequency

10 MHz Figure 23. The

adjustment of
Offset this TCXO for
compliance with
temperature over
10 MHz – 5 Hz its full range
0 25 50 results in a
Temperature in °C frequency offset
at 25°.

17
 Oven Controlled Crystal Oscillators: The third technique of
temperature compensation is to place the crystal and temperature
sensitive elements within a temperature controlled environment. A
heating element is used to maintain the temperature of operation at
the crystal turn-over point. This is the point of minimum change as
indicated on the frequency vs temperature curve and is determined by
the crystal cut (Figure 24). The best oscillator stability is achieved
when the operating point is 15 to 20 degrees above the highest tem-
perature to which the unit will be exposed.

∆f
Figure 24.
f
Selection of
oven turn over
Turnover Point point based
upon the crystal
–25 0 25 80 characteristic
curve.

Two oven controller techniques are in common use for the purpose of
maintaining a constant temperature on the crystal assembly. They are
the switching controller and the proportional controller. The switch-
ing controller turns the power off when the maximum temperature is
reached and on at the minimum level much like a home thermostat.
The proportional oven oven controller varies the current to the heater
or the duty cycle of the heater voltage inversely based upon the offset
of the oven temperature from the desired level. The switching oven
may actually have a worse temperature specification than the TCXO
due to heating and cooling the crystal. Once the proportional oven has
stabilized to its operating temperature, the frequency of the oscillator
will remain very stable, typically <7 × 109 over a 0 to 50 degree Celsius
temperature variation.

Time

Long-Term
The frequency of an oscillator changes due to aging. Aging is the
second largest source of measurement error, the first being tempera-
ture. Calibration removes the aging offset and returns the oscillator to
its nominal operating frequency. Therefore to ensure an accurate
frequency the oscillator must periodically be calibrated. The rate at
which the crystal ages and the time since calibration must be consid-
ered in determining the accuracy of the measurement. Specifications
for aging may be given in terms of days or months, when given in days
a monthly rate is found by multiplying the daily rate times 30 days.
The monthly rate, however, cannot be divided by 30 to obtain a daily
rate since this number would be masked by the effects of temperature

18
and other environmental conditions. Aging for a RTXO will typically
be less than 3 × 10–7 for a 30-day period.

Extra care in manufacturing the crystal resonator can reduce the

effects of aging. Since the effects of temperature are reduced by the
addition of selected compensating elements in the TCXO, it is
advantageous to reduce the aging component. Aging of a TCXO can be
slightly less than that of a room temperature crystal. A typical specifi-
cation for aging is less than 1 × 10–7 per month.

Aging for oven controlled oscillators is usually specified in terms of

days and must be measured after warm-up has been completed. A high
stability proportional oven will have a typical aging rate of less than
5 × 10–10 per day. Translating this in terms of months (5 × 10–10 × 30
days) = 1.5 × 10–8, nearly an order of magnitude better than the TCXO.

Short-Term
Short-term variations in the oscillator output may be further degraded
by where the oscillator circuit is located. It is, therefore, important
that the design engineer in placing the oscillator within an instrument,
be aware of the sources of interference which cause short-term varia-
tions. For some oscillator applications time domain stability is of little
importance since other characteristics mask the effects. Other applica-
tions require extremely stable oscillators for short periods of time so
time domain stability becomes extremely important.

RTXO, short-term, time domain stability is typically 2 × 10–9 for a

1-second average. TCXO units are typically twice as good as the room
temperature unit or 1 × 10–9 for a 1-second average. Oven units may be
2 orders of magnitude better than the TCXO with high performance
units being 1 × 10–11 for a 1-second average.

Line Voltage

The line voltage specification includes effects from a number of

sources. Some of these are power consumption in the power supply
resulting in temperature changes, voltage changes in the oscillator
circuit, amplitude changes in the oscillator drive level which can cause
instability, and even phase changes in the feedback which pulls the
frequency slightly. Variations between the oscillator types is primarily
due to a better quality of regulation, the use of electronic frequency
control circuits and the selection of better components. Line voltage
specifications for an instrument refer to changes in the AC line supply
power. RTXO variations will typically be less than 1 × 10–7 for a 10%
line voltage change. TCXO variations are 5 × 10–11 for a 10% line
voltage change, a 2 times improvement over the room temperature
unit. Oven units are 1 × 10–10 for a 10% line voltage change which is
1,000 times better than the RTXO and 500 times better than the TCXO.

19
 Warm Up

Warm up is a special case of temperature variation which is brought

about by the temperature rise from oscillator turn-on until stable
operating point is reached. It may not be apparent that an RTXO and
TCXO would have a warm up specification, and in fact, it is typically
not specified. However, any instrument when placed into operation
will generate a certain amount of heat. This heat elevates the tempera-
ture surrounding the crystal, and therefore, causes a frequency change.

Examples of warm up are provided in Figures 25A-E for the various

oscillator types. Note that the actual time required to reach a stable
frequency is significantly longer for the RTXO and TCXO during which
the frequency may continue to change by parts in 107 and more.

A B C
5 5 5

10 000 010.7 Hz
4 4 –7
4
1 × 10
3 +1.19 Hz 3 3 –7
1 × 10
10 000 000.2 Hz

2 2 1 hour 2
1 1 1
0 0 10 MHz 0 –.6 Hz
–7
1 × 10
9 999 979.6 Hz

9 9 9
8 8 8 1 hour
–2.7 Hz
7 1 hour 7 7
6 TCXO 6 TCXO 6 TCXO
NO 218 NO 219 NO 20
5 5 5

D E F
5 5 5
4 4 4
–7
3 3 3 1 × 10
–7
2
+2.0 Hz
2 1 × 10 2
6 Min
1 1 1 1 hour
0 10 Hz –7
0 10 Hz 0 10 Hz
1 × 10
9 998 244.0 Hz

9 9 6 Min 9 Oven Oscillator

9 998 282.8 Hz
10 000 008.8 Hz

8 8 8 998244.0 Hz Start
1 hour 8751.1 1 min
7 1 hour 7 7 9219.9 2
9555.8 3
6 TCXO 6 Oven Oscillator 6 782.4 4
NO 218 922.8 5
5 5 5
993.4
000.6 5-1/2
000.5 6
000.0 7
999.9 20 min

Figure 25. A is a RTXO warm up curve. B, C and D are typical TCXO warm up characteristics. The long time (3 hours) is
due to changes in temperature inside the instrument. E and F are oven oscillators. Stability is achieved in ≈20 minutes.

20
Warm up specifications for oven controlled oscillators are given in
terms of the final frequency value after warm up is completed. A
typical time for warm up might be 20 minutes at which time the
frequency will be within 5 parts in 10$9 of its final value. Any calibra-
tion adjustments on the oscillator frequency should be performed after
the crystal has reached its stable range.

Figure 26 shows the effects of warm up and the offset contributed by

retrace. From this it should be apparent that to maintain the accuracy
and stability of an oscillator, the unit should be kept at its operating
temperature. Many instrument manufacturers now provide the capa-
bility of keeping power to the oven oscillator.

∆f
f

Retrace
Warm up Figure 26. Graphic
Time presentation of
OFF ON warm up and
retrace.

A comparison chart of the various crystal oscillators and their typical

specifications for aging, temperature, line voltage, etc., is provided in
Figure 27.

Room Temperature TCXO High Stability Oven

<1.5 × 10–8/mo. usually

Aging Rate <3 × 10–7/mo. <1 × 10 –7/mo.
specified <5 × 10–10/day

Short-Term
<2 × 10–9 rms <1 × 10–9 rms <1 × 10–11 rms
(1 s average)

Temperature
<2.5 × 10–6 <5 × 10–7 <7 × 10–9
0°C — 50°C

Line Voltage
<1 × 10 –7 <5 × 10–8 <1 × 10–10
10% Change

Warm up — — 20 Minutes (5 × 10–9)

Figure 27. Comparison Chart

21
 Oscillator Influence on Measurement Accuracy

The quartz oscillator is a relatively inexpensive method of obtaining an

accurate time standard which when periodically calibrated, will
exhibit traceability to the basic unit of time, the second, as defined by
the XIII General Conference of Weights and Measures, in October
1967. When incorporated in a measurement instrument, any errors
which are a result of the crystal oscillator must be given consideration
before the accuracy of the instrument can be determined. This section
will explore the application of the quartz oscillator when used in a
frequency counter and a frequency synthesizer. The basic operation of
the instrument will not be explained in detail since this may be found
in other readily available sources. Appendix A.

Frequency Counters

The Universal Frequency Counter can make a number of measure-

ments such as frequency, period, period average, ratio, time interval,
and time interval-average. These measurements are really extensions
of a ratio measurement using the fundamentals of frequency and time.
When measuring frequency, the crystal oscillator is used to determine
a precise measurement interval during which the input frequency is
totalized.

The error associated with a frequency measurement consists mainly of

two parts: the ±1 count and the ± time base error. The second error of
the specification ± time base error, is the part contributed by the
crystal oscillator and its associated circuitry. This error is a direct
result of changes in frequency of the oscillator and is made up of
temperature effects, aging, short-term stability, voltage changes, and
warm up.

To compute the time base error for a particular measurement, the

individual fractional frequency errors must be determined. These
errors will typically be specified on the data sheet for the instrument
as follows:

Aging per month 3 × 10–7

Temperature 0 to 50° 2.5 × 10–6
Line Voltage ±10% Change 1 × 10–7

The total error for aging is the rate per month times the number of
months since calibration: 3 × 10–7 × 3 months = 9 × 10–7. Temperature
and line voltage must be considered over the entire variation range so
that the total error is:

Temperature 2.5 × 10–6

Aging 9 × 10–7
Line Voltage 1 × 10–1
∆f/f Total 3.5 × 10–6

22
The error in the measurement equals the nominal frequency f multi-
plied times the fractional frequency error $f/f.

i.e.: f = 10 MHz; ∆f/f = 3.5 × 10–6

10 × 106 × 3.5 × 10–6 = ±35 Hz

The actual frequency for this measurement is 10 MHz ±35 Hz or falls

somewhere between 9.999965 MHz and 10.000035 MHz. These calcula-
tions were made using worst case error. It is possible that some errors
will be positive and some negative producing offsetting effects.

This accuracy may or may not be adequate depending upon the appli-
cation. If not, a look at the computations reveals that a major contribu-
tor to the ∆f/f total, is from temperature. Since this calculation used an
RTXO, selection of a TCXO or an oven oscillator will reduce this
factor. Once this factor has been reduced below 1 × 10–7 . the other
elements of the accuracy statement must be examined if an improve-
ment is required.

Frequency Synthesis

The frequency synthesizer translates the stable frequency of a preci-

sion frequency standard to one of thousands or even billions of fre-
quencies over a spectrum range of DC to 18 GHz. Two methods
have been used in frequency synthesis - the direct and indirect method.
Direct synthesis is accomplished by multiplying, dividing, and mixing
while the indirect method derives its frequency from one or more
phase locked voltage tuned oscillators. The stability and accuracy in
either case is derived from the reference source. The fractional
frequency errors contributed by aging, temperature, and line voltage as
they affect the oscillator must be summed to determine the frequency
error.

Calculations for the synthesizer accuracy can be taken directly from

the data sheet. For ∆f/f:

Temperature 0 — 50 degrees = 5 × 10–9

Line Voltage ±10% change = 5 × 10–10
Aging 5 × 10–10 × 20 days = 1 × 10–8
∆f/f Total 1.55 × 10–8

Selecting a frequency of 1 MHz, the error is 1 × 106 × 1.55 × 10–8 or 1.55

× 10–2 = ±0.0155 Hz, at 10 GHz the error is 10 × 109 × 1.55 × 10–8 = 1.55
× 102 or 155 Hz. The error percentage is the same for both frequency
settings. Whether or not this amount of error is tolerable
depends upon the application for which the frequency is being used.

Improvements in the frequency accuracy can be obtained through

calibration of the oscillator or by using an external reference such as a
cesium standard.

23
 Appendix A

1. “Understanding Microwave Frequency Measurements”, AN 144,

Hewlett-Packard Co.

2. “Measuring the Stability of a Frequency Source”, AN 174-6,

Hewlett-Packard Co.

3. “Measuring Fractional Frequency Standard Deviation (Sigma)

Versus Averaging Time (Tau)”, AN 174-7, Hewlett-Packard Co.

4. “Timekeeping and Frequency Calibration”, AN 52-2,

Hewlett-Packard Co.

5. “Fundamentals of Time and Frequency Standards”, AN 52-1,

Hewlett-Packard Co.

6. “Characterization of Frequency Stability”, IEEE Transactions on

Instrumentation and Measurement, pp 105-120, May 1971.

7. “The Measurement of Frequency and Frequency Stability of

Precision Oscillators”, NBS Technical Note 669, May 1975.

8. “Wave Analyzer Dynamic Range and Bandwidth Requirements for

Signal Noise Analysis”, Technical Report RE 76-26, U.S. Army
Missile Command. Redstone Arsenal, Alabama 35809.

9. “Frequency Domain Stability Measurements: A Tutorial

Introduction”, NBS Technical Note 679, U.S. Department of
Commerce/National Bureau of Standards.

10. “Frequency Stability Specification and Measurement: High

Frequency and Microwave Signals”, NBS Technical Note 632, U.S.
Department of Commerce/National Bureau of Standards.

11. “Spectrum Analysis ... Noise Measurements”, Spectrum Analyzer

Series Application Note 150-4, Hewlett-Packard Co., April 1974.

12. “Short-Term Stability for a Doppler Radar: Requirements Measure-

ments and Techniques”, Proceedings of the IEEE Vol. 54, No. 2,
February 1966. D.B. Leeson, Senior Member, IEEE, and G.F.
Johnson, Senior Member, IEEE.

13. Luis Peregrino/David W. Ricci, “5390A Technical Paper”, presented

at the Thirtieth Annual Frequency Control Symposium, June 2-4,
1976.

14. Michael C. Fischer, “Frequency Stability Measurements

Procedures”, Technical Paper presented at the Eighth Annual
Precise Time and Time Interval Applications and Planning Meeting,
December 1976.

24
References

1. “Basic Electronic Instrument Handbook”, Clyde Coombs, Editor,

McGraw Hill, 1972, Chapter 9, Chapter 33, Chapter 35.

2. “Understanding and Measuring Phase Noise in the Frequency

Domain”, AN 207, Hewlett-Packard Co.

3. W.G. Cady, “Piezoelectricity, An introduction to the Theory and

Applications of Electromechanical Phenomena in Crystals”, (New
revised edition in two volumes), Dover Publications, Inc., New
York, 1964.

4. Paul Bryan, “Precision Quartz Crystals and How They Relate to

the Electronics Manufacturing Industry”, paper 29 presented at
Wescon Electronic Show and Convention, San Francisco,
September 19-21, 1977.

5. Walter L. Bond, John A. Kusters, “Making Doubly Rotated Quartz

Plates” presented at the 31st Frequency Control Symposium,
Atlantic City, June 1977.

6. John A. Kusters, Jerry G. Leach, “TTC’s - Further Developmental

Results”, presented at the 31st Frequency Control Symposium,
Atlantic City, June 1977.

7. “Frequency Standards and Clocks: A Tutorial Introduction”, NBS

Technical Note.

8. Bottom, Virgil E., “The Theory and Design of Quartz Crystal Units”,
McMurray Press, Abilene, Texas, April 1968.

9. Dwyer, D.F.J., “Characteristics of Quartz Crystal Oscillators”,

Electronic Engineering, pp 24-25, April 1977.

10. Ottowitz, Peter J., “A Guide to Crystal Selection”, Electronic

Design, pp 48-51, May 10, 1966.

11. “Characterization of Frequency Stability”, NBS Technical Note 394,

U.S. Department of Commerce/National Bureau of Standards.

12. “Quartz Crystal Oscillators”, Bench Briefs, Volume 13, No. 3, July-
August 1973, Hewlett-Packard Co.

13. “Calibration of an Oscillator”, Bench Briefs, Volume 13, No. 4,

September-October 1973, Hewlett-Packard Co.

14. “Fundamentals of Electronic Counters”, AN 200,

Hewlett-Packard Co.

25
26
27
 H

For more information about Hewlett- Japan:

Packard Test and Measurement products, Hewlett-Packard JApan Ltd.
applications, services, and for a current Measurement Assistance Center
sales office listing, visit our web site, 9-1, Takakura-Cho, Hachioji-Shi,
httpL//www.hp.com/go/tmdir. You can Tokyo 192, Japan
also contact one of the following centers (81) 426 56 7832
and ask for a test and measurement sales (81) 426 56 7840 Fax
representative.
Latin America:
United States: Hewlett-Packard
Hewlett-Packard Company Latin American Region Headquarters
Test and Measurement Call Center 5200 Blue Lagoon Drive
P.O. Box 4026 9th Floor
Englewood, CO 80155-4026 Miami, Florida 33126
1 800 452 4844 U.S.A.
(305) 267 4245/4220
Canada:
Hewlett-Packard Canada Ltd. Australia/New Zealand:
5150 Spectrum Way Hewlett-Packard Australia Ltd.
Mississauga, Ontario 31-41 Joseph Street
L4W 5G1 Blackburn, Victoria 3130
(905) 206 4725 Australia
1 800 629 485
Europe:
Hewlett-Packard Asia Pacific:
European Marketing Centre Hewlett-Packard Asia Pacific Ltd.
P.O. Box 999 17-21/F Shell Tower, Time Square,
1180 AZ Amstelveen 1 Matherson Street, Causeway Bay,
The Netherlands Hong Kong
(31 20) 547 9900 (852) 2599 7777
(852) 2506 9285

Data Subject to Change

Hewlett-Packard Company
© Copyright 1997
Printed in U.S.A. May 1997
28 5965-7662E