Phase Noise in Oscillators

Iulian Rosu, YO3DAC / VA3IUL http://www.qsl.net/va3iul

As well known from oscillator theory, two conditions are required to make a feedback
system oscillate: the open loop gain must be greater than unity; and total phase shift must
be 0° (or 360°) at the frequency of oscillation. The oscillations start when the loop gain
|Aβ| >1 and returns to unity |Aβ| =1 once oscillation commence.
An oscillator circuit can be a combination of an amplifier with gain A (jω) and a
frequency dependent feedback loop H (jω) = βA.
Oscillator has positive feedback loop at selected frequency.

• Frequency Stability is a measure of the degree to which an oscillator maintains the

same value of frequency over a given time.
• Phase Noise can be described as short-term random frequency fluctuations of a
signal; is measured in the frequency domain, and is expressed as a ratio of signal
power to noise power measured in a 1 Hz bandwidth at a given offset from the
desired signal.
• Phase Noise is a measurement of uncertainty in phase of a signal. It is measured as
the ratio of noise power in quadrature (90°out of phase) with the carrier signal to the
power of carrier signal. This is opposed to AM noise which is noise in phase with the
carrier signal.
• Two measurements of Phase Noise are common: the Spectral Density (SD) of
phase fluctuations, and the Single Side Band (SSB) Phase Noise.
Spectral Density is twice of SSB, since this is related to total phase change, which
includes both sidebands, when SSB Phase Noise corresponds to the relative level
on one sideband.

The Phase Noise of a signal can only be measured by a system that has equal or
better noise performance.

• Low oscillator Phase Noise is a necessity for many receiving and transmitting
systems. Adjacent Channel Rejection as well as transmitter signal purity are
dependent on the Phase Noise of the receiver local oscillator or transmit local
oscillator.
• The local oscillator Phase Noise will limit the ultimate Signal-to-Noise ratio (SNR)
which can be achieved when listening to a frequency modulated (FM) or phase-
modulated (PM) signal.
• In a heterodyne system, mixing a clean low-phase-noise RF signal, with a poor
phase noise (noisy) local oscillator, it will turn into a noisy IF.
 • The oscillator Phase Noise is transferred to the carrier to which the receiver is tuned
and is then demodulated. The Phase Noise results in a constant noise power output
from the demodulator.
• Reciprocal mixing is especially important in the presence of strong nearby
interferers. The skirt from the down-converted interferer raises the noise floor for the
down-converted signal well above Thermal Noise kTB.
• In a receiver, if a blocking interferer signal is much bigger than the desired signal,
than the reciprocal Phase Noise due to the blocker self noise would dominate the
noise at IF.

• The performance of some types of AM detectors or SSB detectors may be

degraded by the local oscillator Phase Noise. Reciprocal mixing may cause the
receiver noise floor to increase when strong signals are near the receiver’s tuned
frequency; this limits the ability to recover weak signals.
• Local oscillator Phase Noise will affect the Bit Error Rate (BER) performance of a
Phase-Shift Keyed (PSK) digital transmission system. A transmission error will occur
any time if the local oscillator phase, due to its noise, becomes sufficiently large that
the digital phase detection makes an incorrect decision as to the transmission
phase. For instance, a QPSK transmission system (used in Microwave Links,
CDMA, DVB, etc) will make a transmission error if the instantaneous oscillator
phase is offset by more than 45° since the phase detector will determine that baud
to be in the incorrect quadrant. Digital transmission systems with smaller phase
multiples are more sensitive to degradation due to local oscillator Phase Noise.
• Jitter is another factor that characterizes the oscillator signal and represents a
fluctuation in the timing of the signal and arises due to the Phase Noise.
Due to Jitter, the zero-crossing time of a periodic signal will vary slightly from the
ideal location since the signal is not strictly periodic due to noise.

All of these effects are due to local oscillator Phase Noise, and can only be reduced by
careful design decreasing the Phase Noise.

The Phase Noise of an oscillator is best described in the frequency domain where the
spectral density is characterized by measuring the noise sidebands on either side of the
output signal center frequency.

• Single Side Band (SSB) Phase Noise is specified in dBc/Hz at a given frequency
offset from the carrier.
SSB Phase Noise places limit on receiver Adjacent Channel Selectivity (ACS) and also
affects the receiver Signal to Noise Ratio.
A model for oscillator SSB Phase Noise was introduced by David B. Leeson in 1966.

where:

L PM = Single Side Band (SSB) Phase Noise density [dBc/Hz]

A = Oscillator output power [W]
F = device Noise Factor at operating power level A (linear)
k = Boltzmann’s constant, 1.38 x 10-23 [J/K]
T = Temperature [K]
Q L = Loaded-Q [dimensionless]
fo = Oscillator carrier frequency [Hz]
fm = Frequency offset from the carrier [Hz]

Leeson equation only applies between 1/f flicker noise transition frequency (f1) and a
frequency (f2) where white noise (flat) dominates.
Leeson equation provides several insights about oscillator SSB Phase noise:

• Doubling the Loaded-Q improves Phase Noise by 6dB.

• Doubling the operation frequency results 6dB Phase Noise degradation.

Unloaded-Q means the resonant circuit is not loaded by any external terminating
impedance. In this case the Q is determined only by resonator losses.
Loaded-Q represents the width of the resonance curve, or phase slope, including the
effects of external components. In this case the Q is determined mostly by the external
components.

• It is a common design mistake to achieve high Loaded-Q values by using a very

loosely coupled resonator. The under-coupling results in increased overall resonator
loss requiring an extra amount of gain to compensate it, which in turn, results in
thermal noise increase.
• Resonator loss is a function of its unloaded and loaded Q-factors and is given by:

For example, in a simple feedback oscillator, the minimum Phase Noise is achieved when
the resonator Loaded-Q is set to one half of its Unloaded value (QL = 0.5*QU) that
corresponds to a 6dB resonator loss.
Other oscillator schemes may require different optimum coupling values due to different
design goals and trade-offs.

• In the figure above Phase Noise in dBc/Hz is plotted as a function of frequency

offset (fm), with the frequency axis on a log scale. Note that the actual curve is
approximated by a number of regions, each having a slope of 1/fx, where x = 0
corresponds to the "white" phase noise region (slope = 0 dB/decade), and x = 1
corresponds to the "flicker 1/f" phase noise region (slope = 20 dB/decade). There
are also regions where x = 2, 3, 4, and these regions occur progressively closer to
the carrier frequency.
 • Leeson equation assumed that the 1/f3 and 1/f2 corner occurred precisely at the 1/f
corner of the device. In measurements, this is not always the case.
• The Phase Noise of an oscillator depends by the noise of the open-loop amplifier
and by the half-bandwidth of the resonator. If the amplifier has no 1/f noise region,
the oscillator will have 1/f2 noise below the half-bandwidth. Unfortunately, all the
active devices have some sort 1/f region.
• If the 1/f “flicker” corner frequency is low, the oscillator will have 1/f2 noise slope until
that corner frequency is reached. This is the case with many LC oscillators.
• The 1/f region might be due to either active device or resonator. In many cases the
noise of the resonator dominates, especially in the case of crystals or SAW devices.
In this situation, the crystal should be presented with impedance that doesn’t
degrade the Q, or else the Phase Noise will also be degraded.

Oscillator harmonics can be filtered out by a simple Low Pass Filter, when the
spurious close to the carrier can only be minimized by careful oscillator design.

Rules for designing a low Phase Noise oscillator:

• Maximize the resonator Loaded-Q. To do this (but trading with gain), in the series
resonant circuits use a large Inductor, and in parallel circuits use a large Capacitor.
Coupling the resonator tightly to the oscillating device, and minimize the coupling of
the load to the circuit.
• A 10dB increase in Loaded-Q results in a 20dB improvement in Phase Noise.
• Build the resonator using high-Q components, having constant and quiet noise.
• Low losses are required in all of the constituent parts of the circuit including PCB.
To be carefully considered the series resistance of the reactive components.
Coupled losses in the rest of the circuit should be at most equal to the resonator
losses. To get best Phase Noise, the resonator losses should be x3 the circuit
losses.
• Use an active device with low noise figure at low frequencies.
• Use an active device with low 1/f flicker noise, with good bias circuit. The DC current
set to get the best 1/f flicker noise should be the oscillator device current.
• There is effectively a trade-off between Gain and Phase Noise performance in
microwave transistors, both for the additive or multiplicative noises.
• Maximize the output Signal Power vs Noise Power of the oscillator.
However, the output power increase should be implemented very carefully, since
severe Phase Noise degradation can occur because of the active device noise
elevation at compression.
• Extract the output signal through the resonator to the load, thereby using the
resonator transmission response selectivity to filter the carrier noise spectrum.
• Optimize (and do trade-offs) in noise reduction where is needed, especially consider
close-in noise vs large offset noise requirements.
• Power Supply (VCC) and tuning voltage (Vtune) returns must be connected to the
printed circuit board ground plane. VCO ground plane must be the same as that of
 the printed circuit board and therefore all VCO ground pins must be soldered direct
to the printed circuit board ground plane.
• Adequate RF grounding is required. Several chip decoupling capacitors must be
provided between the VCC supply and ground.
• Good, low noise power supplies must be used to prevent AM noise. Ideally, DC
batteries for both supply (VCC) and tuning (Vtune) voltages will provide the best overall
performance.
• The biasing circuit of the active device should be properly regulated and filtered to
avoid any unwanted signal modulation ore noise injection. Variations on the supply
voltages or currents may also cause undesirable output power fluctuations and
frequency drift.
• The active device should work in Class-A, to minimize the limitations in the stage
that drives the resonator.
• Carefully control the limiting amplitude mechanism, so as not introduce AM noise.
A signal limiter can be placed either before or after the active device, keeping its
output well below the compression level.
• AM-PM conversion is minimized by choosing a 90° crossing angle between the
device line and the load line.
• Phase perturbation can be minimized by using high impedance devices such as
FETs, where the Signal-to-Noise ratio of the signal voltage relative to the equivalent
noise voltage can be made very high.
• Output must be correctly terminated with good load impedance. It is also a good
practice to use a resistive pad between the VCO and the external load.
• Connections to the tuning port must be as short as possible and must be well
screened, shielded, and decoupled to prevent the VCO from being modulated by
external noise sources. A low noise power supply must be used for tuning voltage.
• Minimize Frequency Pushing by the Gate or Base voltage of the transistor.
Frequency Pushing is a shift in the oscillation frequency usually caused by a change
in the transistor bias voltage.
• Avoid saturation of the active devices at all cost, and try to have either limiting or
automatic gain control (AGC) without degradation of the Q of the resonator.
Saturation of the active device can also lower the loaded-Q since the device losses
will then add to those of the resonator.
• Use active components with low 1/f-noise. Flicker noise in active devices is also
known as 1/f noise because of the 1/f slope characteristics of the noise spectrum
(the amplitude varies inversely with frequency). Mainly traps associated with
contamination and crystal defects in the emitter-base depletion layer cause this
noise (in BJTs case). These traps capture and release carriers in a random fashion.
The time constant associated with the process produce a noise signal at low
frequencies.
• Transistors made in different processes have different 1/f noise corners.
JFETs are the best (~1kHz), followed by BJTs (~5kHz), then CMOS (~1MHz), and
GaAs are the worst (~10MHz).
• Consider using noise reduction via feedback, or feed-forward noise reduction
techniques.
Rules for designing a low Phase Noise Voltage Controlled Oscillator (VCO):

In a Phase-Locked Loop (PLL) a Voltage Controlled Oscillator (VCO) will always have
some spurious signals present on its output.
The amplitude and frequency of these spurious modulations may vary as the local
oscillator is tuned.

• Poor layout of the phase-locked loop oscillator circuitry (VCO) may increase the
amplitude and number of the output spurious signals.
• Oscillator Phase Noise has two components: Phase Noise resulting from direct
upconversion of white noise and flicker noise (1/f noise), and Phase Noise resulting
from the changing phase of the noise sources modulating the oscillation frequency.
• In VCO design another source of Phase Noise increase are the non-linear
capacitors (varactors) used in the LC resonator and its control lines.
• In a VCO, have to maintain the Q of the resonator by avoiding forward bias on the
varactor tuning diodes, limiting the signal swing across the tuning diodes to prevent
heating and thermal effects. This can be achieved by placing the varactor circuit in
the gate or base if possible.
• Two back-to-back varactors should be used to avoid self-rectification of the RF
signal across the varactors, which always results in phase noise degradation.
• The noise from the varactor diode resistance can also become the dominant noise
source. For good Phase Noise, the carrier signal effectively appearing across the
varactor noise resistance should be maximized to maintain good Signal-to-Noise
ratio at this point. By transforming the noise load resistance seen by the oscillating
device to a lower value in the matching circuit, the Power-to-Noise ratio across the
varactor can be maximized, although at the expense of tuning bandwidth since the
matching circuit will restrict the obtainable capacitance variation.
• There is a compromise in order to avoid breakdown, saturation, or overheating
effects in the varactor. These will all reduce the Loaded-Q.
• When frequency of the carrier increases, it is more difficult to achieve good Phase
Noise
• It’s easy to achieve good Phase Noise when the frequency range covered by VCO
is narrow; the tuning bandwidth must be small. Generated energy should be coupled
from the resonator rather than from another portion of the active device so that the
resonator limits the bandwidth.
• Increasing tuning sensitivity (measured in MHz / V) degrades Phase Noise.
• For a given frequency it’s easy to achieve good Phase Noise in VCO’s using a wide
tuning voltage range.
• Temperature affects the Phase Noise. In a range of –55’C to +85’C the variation is
+/- 3dB of the Phase Noise.
• Using of back-to-back varactor diodes in the tuning circuits has been found to
eliminate effects of tuning circuit diode noise on oscillator signal spectral
performance.
Characteristics of the ideal resonator for low Phase Noise oscillator:

• High Group Delay (high resonator Loaded-Q).

• High operating frequency.
• Low Loss.
• Moderate Drive Capability.
• Low frequency sensitivity to environmental stress (vibration, temperature, etc.).
• Good short-term and long-term frequency stability.
• Accurate frequency set-on capability.
• External frequency tuning capability.
• No undesired resonant modes or higher loss in undesired resonant modes or
undesired resonant mode frequencies far from desired operating frequency.
• High manufacturing yield of acceptable devices.
• In-circuit resonator effective Q can be determined by intentionally altering the circuit
phase shift by a known amount and measuring the resultant oscillator signal
frequency shift.

Passive components in the oscillator circuit also exhibit short-term instability.

• Passive components (resistors, capacitors, inductors, reverse-biased, varactor

diodes) exhibit varying levels of flicker-of-impedance instability whose effects can be
comparable to or higher than to that of the sustaining stage amplifier 1/f AM and PM
noise in the oscillator circuit.
• The oscillator frequency control element (i.e., resonator) can exhibit dominant levels
of flicker-of-resonant frequency instability, especially acoustic resonators.

Rules to select a transistor and its bias for designing a low Phase Noise oscillator:

• The best oscillator transistor is a device with the lowest possible noise figure and
lowest fT. A commonly used criteria is: fT ≤ 2 * fosc.
• Meantime, doing a trade-off, have to use a high frequency transistor having small
junction capacitance and operate at moderately high bias voltage to reduce phase
modulation due to junction capacitance noise modulation.
• Low 1/f noise of the transistor in the oscillator is very important, because the 1/f
noise appears as sideband noise around the carrier frequency of the oscillator
output signal.
• The 1/f noise is directly related to the current density in the transistor.
Transistors with high Icmax used at low currents have best 1/f performance.
For low Phase Noise operation use a medium power transistor. If you need your
output power to be achieved at 6-9 mA, select a transistor with Icmax of 60-90 mA.
However, the ft of a transistor drops as current is decreased. Additionally, the
parasitic capacitances of a high current transistor are higher due to the larger
transistor structure required.
 • In BJTs as VCE increases, the flicker corner increases as the white noise increases,
but the magnitude of the 1/f noise is constant. As base current increases, the flicker
corner frequency increases with the magnitude of the 1/f noise and the increased
shot noise current.
• The effect of flicker noise can be reduced through RF feedback. An unbypassed
emitter resistor of 10-30 Ω in a BJT circuit can improve the flicker noise by as much
as 40 dB. The proper bias point of the active device is important.
• In a well-designed near-class-A oscillator, the frequency is determined primarily by
the resonator. As the loaded-Q is increased, the active device parasitic reactances
become less significant in determining the oscillation frequency.
Thus, changes in these parameters from device to device, with temperature and
with supply voltage, have less effect. A simple test of how well the active device
reactances are isolated from the resonator is to observe the operating frequency as
the supply voltage is varied.
• Precautions should be taken to prevent modulation of the input and output dynamic
capacitances of the transistor; which will cause amplitude-to-phase conversion and
therefore introduce noise.
If phase shift in the transistor changes, the oscillation frequency will change until the
loop phase shift returns to zero. Thus, phase modulation in the amplifier causes
frequency modulation of the oscillator.
• Device with low noise figure combined with a small correlation coefficient.
• Device with relative high output power.
• Device with low output conductance.
• Device with reasonably high input impedance.
• Meeting an impedance condition at the input of the active device, which can be
achieved by optimization of the feedback factor and which leads to optimum
impedance noise matching.
• Device with low multiplicative noise (1/f AM and especially 1/f PM).
• Device having drive capability consistent with resonator drive level and loss.
• Low noise in ALC/AGC circuits and/or in-compression amplifier operation.
• Low gain and phase sensitivity to DC supply and circuit temperature variations.
• Device with low Group Delay (wide bandwidth).
• Device with high load circuit isolation.
• Device with minimal number of adjustable and bias components.
• Ease of alignment and test.
• Device with good DC efficiency.

Phase Noise in PLL (Phase Lock Loop) oscillators

A Phase Lock Loop (PLL) is a type of a controlled oscillator where the frequency stability
is of critical importance. PLL designers are interested in both, long-term frequency
stability (over a long period of time, hours, days, or months) and short-term stability (over
a period of seconds or less).
Frequency stability it is usually specified as the ratio, Δf/f for a given period of time,
expressed as a percentage or in dB. These frequency variations can be random or
periodic.
• A spectrum analyzer can be used to examine the short-term stability of a signal, as
in the figure below where is shown a typical spectrum, with random and discrete
frequency components causing a broad skirt and spurious peaks.

• The discrete spurious components could be caused by known clock frequencies in

the signal source, power line interference, and mixer products.
• The broadening caused by random noise fluctuation is due to phase noise.
It can be the result of thermal noise, shot noise, and/or flicker noise (1/f noise) in
active and passive devices.
• The phase noise spectrum of an oscillator shows the noise power in a 1 Hz
bandwidth as a function of frequency.
Phase noise is defined as the ratio of the noise in a 1 Hz bandwidth at a specified
frequency offset, fm, to the oscillator signal amplitude at frequency fo.

In a PLL the design of the loop filter can affect the Phase Noise of the system:

• Within the loop bandwidth, the Phase Noise of the oscillator will tend to cancel itself,
leaving a Phase Noise essentially equal to the frequency multiplied Phase Noise of
the crystal reference.
• Multiplied Phase Noise of the crystal reference at particular frequency offset is equal
with reference Phase Noise at the same frequency offset plus 20*LOG(N VCO_divider)
plus 1dB (multiplication efficiency factor).
• Outside the loop bandwidth, the Phase Noise of the oscillator is not canceled, and
will continue to decrease, until reaching its half bandwidth, ωo/2Q or 1/f corner
frequency.
Since the Q of the crystal reference is very large, its half bandwidth is very small,
and its frequency multiplied Phase Noise will remain relatively flat down to very
small frequency offsets. Further, at some moderate frequency offset, this multiplied
phase noise power spectral-density will be crossed by the decreasing oscillator
phase noise power spectral-density.
• The bandwidth of the loop should be chosen equal to the frequency offset of this
crossover.

• Typically, crystal-based oscillators have good phase noise performance close to the
carrier frequency and the VCO of this PLL offers a low noise floor (>1 MHz from the
carrier). The PLL loop bandwidth should be set in such a way that the output phase
noise can take advantage of the low phase noise of reference clock at close to
carrier frequency and the low phase noise floor of the VCO. The in-band phase
noise of the PLL contributes up to the loop bandwidth frequency.
• To minimize the output noise due to the VCO, the loop bandwidth must be as large
as possible.
• To achieve minimum phase noise within the loop bandwidth, the in-band noise
contributed by the other loop components should be kept to a minimum.
• The loop bandwidth must be less than the input reference frequency to keep the
loop stable and suppress the spurs at the output due to the reference leakage
signal.
• The PLL loop bandwidth is not a barrier frequency with a discontinuity on either side
of the barrier; it can be approximated as such with the proviso that small errors
around the offset frequency equal to the loop bandwidth are accepted.
• The role of the loop filter, which is a low-pass filter inserted between the phase
comparator and the VCO control voltage circuit, eliminates the high frequency
component of the phase correction pulse generated by the phase comparator so
that the only the DC component is provided to the VCO.
• As a rule of thumb, the cut off frequency of the low-pass filter is chosen as equal or
less than comparison frequency divided by ten; Fcutooff < (Fcomparision / 10)
• Usually the low-pass filter is an RC network. The analysis of the Phase Noise
performance shows that the Phase Noise depends on the resistor value, part of the
low-pass filter. The higher the resistor, the higher is its contribution to the Phase
Noise.

• The noise spectrum of a frequency source is made up of (at least) five components:
 1. Random frequency walk
2. Flicker frequency noise
3. Random phase walk (white FM)
4. Flicker phase noise
5. White phase noise

These components each dominate within a particular region of the phase noise spectrum,
and the conversions are a piecewise linear approximation to the spectrum.
Each component is converted separately. Therefore, it may well be quicker and more
accurate to make the measurement in question than to convert from one to the other.

In the figure below is shown a spectrum analyzer presentation of the output of a typical
phase lock frequency synthesizer.

PLL reference spurs shown on a spectrum analyzer

The spurs, which could be down about 40dB to 80dB, are at the reference frequency and
its harmonics, and are the most difficult to suppress unwanted output from a phase lock
synthesizer. They are produced by undesired frequency modulation of the VCO by the
reference frequency output of the phase detector, which should ideally be rejected by the
loop filter.

• When considering oscillator phase noise, we assume a signal such as that from a
local oscillator, and consider that the noise is made up of two components in
quadrature, one of which is an amplitude noise component and the second a
phase noise component.
• However, because it is regenerative, the action of an oscillator is such that its output
amplitude continues to build up until it cannot increase any more.
This means that an oscillator’s output normally suppresses amplitude noise,
although it can appear as AM-to-PM noise but at a greatly reduced level.
• The higher the Q of the resonator, the lower is the effect of AM-to-PM noise
conversion, as well as the phase noise itself. The overall result is that there is very
little AM noise to start with, and what little there is, is converted to phase noise.
The best approach to phase noise reduction is to increase the Q of the resonator in
the signal source.
 • There are three different means of generating high-frequency signals, and each of
them produce different single-sideband (SSB) spectrum.

1. Direct multiplication of a quartz crystal-stabilized oscillator, generating a

low- frequency signal, to a high frequency.
2. Use of a phase lock loop synthesizer to multiply a low crystal frequency
reference signal to a high frequency.
3. Use of a high-frequency surface acoustic wave resonator in an oscillator
to generate a high-frequency signal directly.

Comparison of SSB phase noise of three techniques for high-frequency local oscillator signal generation

- The test results carried out to compare the single-sideband phase noise at offsets of
more than approximately 8 kHz, shows that the direct, surface acoustic wave
oscillator approach has the lowest phase noise.
- Below 1 kHz offset, however, the phase lock technique is lowest in phase noise.

Phase Noise in Crystal Oscillators

One of the most important characteristics of crystal oscillators, besides they can
provide good frequency stability, is that they can exhibit very low Phase Noise.
In many oscillators, any spectral energy at the resonant frequency will be amplified
by the oscillator, resulting in a collection of tones at different phases.
In a crystal oscillator, the crystal mostly vibrates in one axis, therefore only one
phase is dominant.
• At lower offset frequencies approaching the carrier, the Phase Noise is determined
by the quality Q of the crystal resonator. For example, a 100MHz crystal has a
considerably lower Q than a 10MHz crystal, so the noise is higher at the low offsets.

The amplitude of 1/f flicker noise in crystal resonators is a very important parameter
of oscillators used in various applications.
 • To get accurate models of the 1/f frequency noise in the resonator itself, this should
be independent of the noise generated by the afferent electronic circuit.
• Was discovered that the amplitude of the 1/f frequency noise in a crystal depends
not only on the Q of the resonator but also on the volume between the electrodes.
Since the amplitude of 1/f noise depends on active crystal volume, to get low close-
in Phase Noise we have to use the lowest overtone and lowest resonator frequency.
• Extra noise source is associated with electrode-crystal interface.
A resonator having smaller electrodes would have lower 1/f flicker noise than other
with the same resonant frequency and Q, but with larger diameter electrodes.
• The decrease in electrode area would increase the impedance and degrade the
wideband noise, but for most resonators the wideband noise is dominated by the
electronics of the oscillator.
The increase in series resistance decreasing the electrode diameter by a factor of 4
would be probably the limit from the standpoint of wideband noise.
This change might lead in a change of the oscillator loop gain.
• Thus, for application specifically requires minimum close-in Phase Noise, lower
frequency crystals may be used, when for low noise floor applications (wideband
noise), the highest frequency crystal which satisfies long term stability requirements
should generally be used.
• Also was discovered that 1/f frequency noise in a crystal is virtually independent of
the loaded-Q of the resonator, when we know that in a practical oscillator circuit
there is a dependence of the Phase Noise on loaded-Q, because the sustaining
electronics contribute to the overall noise level.
• The crystal resonator plate can be cut from the source crystal in many different
ways. The orientation of the crystal cut influences the crystal's frequency stability,
Phase Noise characteristics, aging characteristics, thermal characteristics, and
other parameters. A special cut (SC - Stress Compensated), is a double-rotated cut
developed for oven stabilized oscillators with low Phase Noise, and good aging
characteristics. This special cut SC is less sensitive to mechanical stresses, and
has faster warm-up speed, higher Q, better close-in Phase Noise, less sensitivity to
spatial orientation, and less sensitivity to vibrations.

Various topologies of crystal oscillators exhibit different performances mainly due to

the limiting functions of the circuit, and of the loaded-Q of the crystal resonator.
In many instance decisions of selection of particular type of crystal oscillator
configuration is made on the basis of short-term frequency stability.

Pierce Miller Butler Bridge-Tee Driscoll

Crystal Oscillator Topologies

 In most anti-resonant circuit configurations (such as Pierce and Miller configurations),
the out-of-band impedances may become reactive due to the sharp reactance vs
frequency characteristic exhibited by the crystal unit.
The series-mode circuits (as Butler and Bridge-Tee configurations) are more effective
in reducing the wideband noise floor (up to 10dB compared to anti-resonant circuit).
The main disadvantage of series-mode circuits is the large degradation in crystal unit
loaded-Q (due to limiting of the transistor). For example, the effective value for crystal unit
loaded-Q is about 120.000 for the Pierce circuit, and 24.000 for Bridge-Tee circuit.
• When limiting occurs, the transistor is turned OFF for a time portion of the signal
waveform, time when the impedance seen by the crystal resonator at the transistor
emitter contains a large value component at the signal frequency. This component
of transistor impedance (which becomes increasingly large as the excess gain in the
sustaining stage is increased) it will degrade the oscillator loaded-Q.
• In addition, the degradation in crystal loaded-Q can produce degradation in
oscillator long-term frequency stability. This includes changes caused by
environment (temperature, humidity) long-term power supply variation, and short-
term effects (vibration and power supply ripple).
• Thus, better output noise spectrum could be obtained using a crystal oscillator in
series-mode configuration, employing class-A non-limiting action in the sustaining
stage transistor.
• The limiting function in a crystal oscillator may be controlled by:
1. Auxiliary low-noise AGC circuits (a portion of the amplified RF signal is rectified
and used to control the RF gain of the sustaining stage).
2. Back-to-back Schottky diodes incorporated in the oscillator circuit, so that the
diode RF impedance presented to the sustaining stage (and hence the RF gain)
decreases with increasing the RF level.
3. Incorporation of a second self-limiting transistor stage in the oscillator sustaining
circuit, in a manner such that its effect on crystal unit loading is insignificant.

• In 1972 M.M. Driscoll developed a very low Phase Noise series-mode crystal
oscillator employing two transistors connected in cascode configuration.
- The quartz crystal resonator is used as an un-bypassed emitter load on Q1.
- Unlike the common Butler or Bridged-Tee circuits, Q1 is ON during the full cycle of
the signal waveform, since the limiting function is provided in Q2.
- Connecting the crystal between emitter of Q1 and ground increase the crystal
loaded-Q. Reducing the emitter impedance by increasing the bias current of Q1
avoid over-dissipating power into the crystal.

• In crystal oscillators the phase noise at frequency offsets near the resonant
frequency (close-in phase noise) is found to be inversely proportional to the fourth
power of the circuit Q, i.e. it is proportional to Q-4, and as such this performance
deteriorates rapidly with the choice of higher resonator frequency.
For example, an oscillator using a 5 MHz, 3rd overtone resonator, has phase noise of
-100 dBc/Hz at 1 Hz offset from the carrier, while at 100 MHz oscillator using the same
overtone resonator achieve only a phase noise of -60 dBc/Hz.
 • Noise contributions from the active and passive components of the crystal oscillator
(other than resonator) are proportional to dx/df within the crystal’s bandwidth and
thus contribute to increased phase noise at higher frequencies.
• The crystal oscillator phase noise floor (wideband phase noise) is governed by the
subsequent amplifier noise and by the passive components, so remains relatively
constant with frequency. Values of -160 to -170 dBc/Hz at the phase noise floor are
typical for many precision crystal oscillators regardless of the frequency.

Many applications require multiplication of the crystal reference frequency up to the

system local oscillator (LO) frequency. In such cases a design tradeoff lies between
obtaining either a good close-in phase noise or a good phase noise floor.
• When multiplying, the phase noise performance is degraded at the rate of
20*log(multiplication factor).
• For applications requiring a good phase noise floor, it is preferable to multiply from
the highest reference frequency practical.
• On the other hand, if good close-in phase noise is required, multiplying from a lower
frequency oscillator produces the best results.

Measuring the Oscillator Phase Noise

Generally, the Spectral Density, or Phase Noise, of an oscillator is measured in dBc (dB
below the carrier) in a bandwidth of 1Hz at an offset frequency fn.
The Phase Noise, therefore, is related to the output power.
The Noise Power and the curve can have different shapes based on the noise sources.

Phase_Noise[dBc/Hz] = 10*LOG (Pn / Ps)

Pn = Noise Power in 1Hz Bandwidth at particular frequency offset (fn) in Watts
Ps = Carrier signal power in Watts

A. The simplest and fastest method of determining the Phase Noise of an oscillator is
the direct measurement using a Spectrum Analyzer.

For this measurement, the tested oscillator must fulfill the following conditions:
• The oscillator drift must be small relative to the Spectrum Analyzer sweep time
since otherwise the oscillator frequency varies during the sweep, leading to
distorted results. The synthesizers commonly used in radio communications fulfill
this condition since they are locked to a stable reference.
 • The Phase Noise of the local oscillators of the Spectrum Analyzer must be low
enough to ensure that the characteristics of the tested oscillator and not those of the
Spectrum Analyzer are determined.

Factors that limit the analyzer’s ability to correctly measure the Phase Noise of a signal:
• IF (RBW) filter bandwidth, verses noise bandwidth.
• IF filter type and shape factor.
• Analyzer’s local oscillator stability - residual FM.
• Analyzer’s local oscillator stability - noise sidebands.
• Analyzer's detector response to noise - peak detector introduces errors.
• Analyzer's log amplifiers response to noise.
• Noise floor of the analyzer.

When measure the oscillator Phase Noise using a Spectrum Analyzer the following
equation can be used for direct reading in dBc/Hz, for particular Resolution Bandwidth
(RBW) set on the analyzer:

Phase_Noise[dBc/Hz] = Carrier_Power[dBm] – Noise_Power@Freq_offset[dBm] – 10*LOG (RBW[Hz])

Spectrum Analyzers generally only measure the scalar magnitude of noise sidebands of
the signal and are not able to differentiate between amplitude noise and phase noise.
In addition, the measurement process is complicated by having to make a noise
measurement at each frequency offset of interest, sometimes a very time consuming task.

B. Another Phase Noise measurement can be done using a Reference Oscillator and a
Phase Detector:

• The Phase Detector converts phase difference between its two inputs into a voltage
When the phase difference between the two inputs is 90° (quadrature) the Phase
Detector output will be 0 volts.
• Any phase fluctuations around the quadrature point will result in a voltage
fluctuation at the output of the Phase Detector.
• The Phase Detector output can then be digitized and processed to obtain the phase
noise information desired.
• Additionally, the Phase Detector technique also enables residual/additive noise
measurements for two-port devices.

Several methods have been developed based upon the phase detector concept.
Among them, the Reference Source / PLL (Phase Locked Loop) is one of the most widely
used methods.
 Phase Noise measurement using a Reference Oscillator and a Phase Detector

• The reference oscillator is synchronized to the measured oscillator by means of a

PLL of a very small bandwidth.
• The PLL sets the phases of the two oscillators to a difference of 90°.
• The Phase Noise of the DUT is eliminated within the loop bandwidth.
• The sum noise power of the reference and the test oscillator obtained outside the
loop bandwidth is present at the output of the phase detector.
• This output signal is amplified by means of an LNA (Low Noise Amplifier) and
displayed on a Spectrum Analyzer starting at a frequency of 0Hz.

This method offers the advantage of a very wide dynamic range, provided that the
reference oscillator is of a very high spectral purity. Often, two identical oscillators are
used for measurements on crystal oscillators, and the assumption made that the two
oscillators have the same Phase Noise. In this case, 3 dB is subtracted from the result
because the noise powers add up.
Also, this method yields the widest measurement coverage (e.g. the frequency offset
range is 0.01 Hz to 100 MHz).
This method is insensitive to AM noise and capable of tacking drifting sources.

The disadvantages of this method are:

• The method requires two oscillators at the same frequency that have to be
synchronized to each other.
• An extra PLL and a Low Noise Amplifier are needed.
• Calibration is complex because the gain of all components is included in the result.
Calibration is made by mistuning the two oscillators relative to each other and
measuring the AC voltage obtained at the output of the LNA.
• Requiring a clean, electronically tunable reference source, and that measuring high
drift rate sources requires reference with a wide tuning range.

C. The Frequency Discriminator method is another variation of the Phase Detector

technique with the requirement of a reference source being eliminated.
 • The signal from the tested oscillator is split into two channels.
• The signal in one path is delayed relative to the signal in the other path.
• The delay line converts the frequency fluctuations to phase fluctuations.
• Adjusting the delay line or the phase shifter will determine the phase quadrature of
the two inputs to the mixer (phase detector). Then, the mixer (working as a phase
detector) converts phase fluctuations to voltage fluctuations, which can then be read
by the baseband Spectrum Analyzer as a frequency noise.
• The frequency noise is then converted as a phase noise reading.

The Frequency Discriminator method degrades the measurement sensitivity (at close-in
offset frequencies) but is useful when the tested oscillator is a noisier source that has
high-level, low-rate phase noise, or has high close-in spurious sidebands which can make
problems for the phase detector PLL technique.
• A longer delay line will improve the sensitivity but the insertion loss of the delay line
may exceed the source power available and cancel any further improvement.
• Longer delay lines limit the maximum offset frequency that can be measured.
• This method is best used for free-running sources such as LC oscillators or cavity
oscillators.
References:

1. Low Noise Oscillator Design and Performance – M.M.Driscoll

2. Two-Stage Self-Limiting Series Mode Type Quartz-Crystal Oscillator Exhibiting Improved Short-
Term Frequency Stability (June 1973) – M.M.Driscoll
3. A simple model for feed back oscillator noise- IEEE 54(2):329 (Feb 1966) D. B. Leeson
4. Oscillator Design and Computer Simulation - R.W. Rhea
5. RF/Microwave Circuit Design for Wireless Applications - U. Rohde and D. Newkirk
6. Microwave Circuit Design using Linear and Nonlinear Techniques - Vendelin, Pavio, U. Rohde
7. RF Design Guide – P. Vizmuller
8. Practical RF Circuit Design for Modern Wireless Systems - vol 2 - Gilmore, Besser
9. Phase Noise Reductions in Microwave Oscillators – A. Chenakin
10. California Eastern Laboratories - AN1026 - “1/f Noise Characteristics Influencing Phase Noise”
11. Infineon Technologies - AN023 - Designing Oscillators with low 1/f-noise
12. Analysis and Prediction of Phase Noise in Resonators and Oscillators – HP
13. A New Model of 1/f Noise in BAW Quartz Resonators – Walls, Handel, Besson, Gagnepain
14. Precision Quartz Oscillator Tradeoffs - M.Vaish, D.Trepanier
15. Radio Receiver Design - Dixon
16. RF Oscillators - Besser Associates, Inc.
17. Mini-Circuits – VCO Designers Handbook 2001
18. Applied Microwave and Wireless, 1990-2002
19. Analog Devices Application Notes, 2005
20. Phase Noise Measurements with Spectrum Analyzers of the FSE family - R&S
21. Phase Noise Measurement Methods and Techniques - Agilent
22. RF Design Magazine, 1993-2002
23. Microwave Journal, 1997-2012

http://www.qsl.net/va3iul