Documente Academic
Documente Profesional
Documente Cultură
MILLIMETER-WAVE FREQUENCIES
WU YONG
Doctor of Philosophy
2012
Acknowledgments
ful to my supervisor, Professor Lu Yilong, for his excellent guidance, support and
encouragement.
Agency for Science, Technology and Research, for encouraging me to pursue this
PhD research work. I would like to thank my colleagues for their support and
discussions.
versity, Institute for High Energy Physics of the Chinese Academy of Sciences and
Huang, Li Tipei and Lin Baojun who have shown me how to carry out scientific
research work.
Finally, I would like to thank my family for their love, understanding, and
i
Contents
Acknowledgments i
Summary viii
List of Tables xv
1 Introduction 1
1.2 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Organization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2 Research Background 15
ii
Contents
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
iii
Contents
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
iv
Contents
4.7 Discussions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
v
Contents
vi
Contents
Bibliography 167
vii
Summary
viii
measurement accuracy for high attenuation devices. The accuracy of this proposed
system is much better than earlier published results. The phase-shift measurement
uncertainty is 0.09◦ –0.32◦ for a 0–140 dB coaxial step attenuator at 18 GHz. The
phase-shift measurement uncertainty of our system for a 100 dB attenuator at 18
GHz is ∼106 times smaller than that of a vector network analyzer (VNA). We have
analysed the mismatch uncertainty in insertion phase-shift measurement, which has
not been studied in the literature. A coaxial beadless air line has been used as a
phase-shift standard to verify the proposed phase-shift measurement system. The
calculated phase-shifts with correction of the line insertion loss agree with the
measured values within the expanded uncertainties.
ix
of a VNA. A waveguide line standard has been measured as a phase-shift standard
to verify the proposed phase-shift measurement system. The calculated phase-
shifts agree with the measured values of the line standard within their expanded
uncertainties.
These systems have been used as the Singapore national standards for attenu-
ation and phase-shift measurement at microwave and millimeter-wave frequencies.
Variable attenuators, air lines and waveguide shims calibrated by these systems
can provide traceability and serve as reference standards for calibration of VNAs,
power meters, radiometers, spectrum analyzers, etc. The dual-channel design using
multipliers and subharmonic mixers can be used at 0.5–1.1 THz and the receiver
noise floor is expected to be much lower than that of a VNA since it is mainly
determined by the receiver thermal noise. The proposed measurement techniques
can also be used to achieve high DR in instrumentation, wireless communication,
radar, imaging, spectroscopy, NDT, material measurement, etc.
x
List of Figures
xii
List of Figures
2.11 A beadless air line in our laboratory at the National Metrology Centre. 30
3.3 The magnitude and phase of the signal detected by LIA with ∼370
dB attenuation inserted between test ports in the dual-channel re-
ceiver at 26 GHz, data collected in 120 seconds (one sample per
second). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2 The SSB phase noise L(f) of the microwave sources, based on the
instrument specifications at 250–500 MHz, 3–10 GHz, 10–20 GHz
and 20–30 GHz. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.3 The maximum phase fluctuation measured by the LIA and the pre-
dicted values using the phase noise data shown in Fig. 4.2 (frequency
range: 0.05–26 GHz). . . . . . . . . . . . . . . . . . . . . . . . . . 74
xiii
List of Figures
4.5 (a): The magnitude and phase of the signal detected by LIA with
110 dB attenuation inserted between test ports in the dual-channel
receiver at 26 GHz, data collected in 120 seconds (one sample per
second), (b): smoothed magnitude data obtained by (4.28) using the
raw magnitude data shown in (a). . . . . . . . . . . . . . . . . . . 80
6.4 The SSB phase noise of the microwave sources (based on the in-
strument specification at 2.5 GHz and 13.75–20.83 GHz) and the
calculated phase noise of the sub-millimeter wave source obtained
using (6.14). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
xiv
List of Figures
xv
List of Tables
xvi
List of Tables
xvii
List of Tables
6.10 Measurement of the VA2 at 480 GHz at receiver input levels that
are 42 dB lower than those given in Table 6.9 . . . . . . . . . . . . 139
7.2 Differential phase-shift results of a 0–60 dB WR-03 RVA at 325 GHz 154
xviii
List of Acronyms
AC Alternating current
AF Audio Frequency
BP Band-pass
DC Direct current
DR Dynamic range
GHz Gigahertz
IF Intermediate Frequency
kHz Kilohertz
LF Low frequency
xix
LIA Lock-in Amplifier
LO Local Oscillator
LP Low-pass
MHz Megahertz
QO Quasi-optical
RF Radio Frequency
RSS Root-sum-of-the-squares
xx
THz Terahertz
VA Variable attenuator
xxi
“When you can measure what you are speaking about, and express it
in numbers, you know something about it; but when you cannot express
it in numbers, your knowledge is of a meagre and unsatisfactory kind;
it may be the beginning of knowledge, but you have scarcely, in your
thoughts, advanced to the stage of science whatever the matter may
be.”
“With better measuring tools, one can look where no one has looked
before. More than once, seemingly minute differences between measure-
ment and theory have led to major advances in fundamental knowledge.
The birth of modern science itself is intimately linked to the art of ac-
curate measurements.”
Introduction
surement accuracy and precision so that they can satisfy the ever-increasing de-
determine the measurement errors. The accuracy of the reference standard must
1
1.1. Motivation and Background
Precise and accurate measurement has played a key role in the advancement of
science and technology [3, 4]. The art of measurements has enabled scientists to
observe the nature with ever increasing precision and accuracy, which has led to
4.08 GHz [5,6]. The noise temperature of the radiometer has been calibrated using
a cryogenic noise source and microwave attenuators [7]. After careful analysis
cosmology to investigate the origin of the universe [8]. Currently, the most accurate
Cesium clocks have been the established primary standard for time and fre-
quency since 1950s. The resonance frequency of cesium atom is 9,192,631,770 Hz,
which can be used to steer an external microwave oscillator to match this stable
dards and Technology (NIST) has achieved a fractional uncertainty below 4 part
in 1016 [11].
It was discovered in 2000 that an optical frequency comb can be used as a pre-
2
1.1. Motivation and Background
(THz) [4, 12, 13]. The frequency ratio between the second harmonic of an Nd:YAG
laser and the fundamental frequency has been measured with a relative uncer-
tainty of 7 parts in 1019 using such a frequency comb [4, 14]. Very narrow optical
optical resonance can be used to lock a laser, which can serve as a stable oscillator
(optical atomic clock). The NIST has built an optical clock with a fractional fre-
quency uncertainty below 8.6 parts in 1018 , based on quantum logic spectroscopy
of an Al+ ion (with a transition frequency of 1,121 THz) [15, 16]. Such a precision
clock would remain accurate to within one second in ∼3.69 billion years. Precision
clocks are very useful in scientific research, communications, navigation and space
0.84184 ±0.00067 fm [17], which is different by 5.0 standard deviations from the
value recommended by the Committee on Data for Science and Technology (CO-
DATA) for international use [18]. This has been obtained by accurate measurement
of the Lamb shift frequency (49,881.88 ± 0.76 GHz) in muonic hydrogen atoms us-
ing an optical frequency comb [17]. The CODATA value for rp is 0.8768 ± 0.0069
fm. The accuracy of the latest measurement is 10 times better than the CODATA
value. The newly measured proton radius is 4% smaller than the CODATA value.
This finding implies that either the Rydberg constant has to be revised, or the
pected result shows precision measurement on the scale of 10−20 m could lead to
in molecules and dynamic behavior of matter using laser pulses with duration of
3
1.1. Motivation and Background
a time resolution of ≤ 150 as (1 as = 10−18 s) has been made possible using soft
X-ray pulse with duration of 650 ± 150 as and a few-cycle visible light pulse [21].
as to monitor electrons with sub-atomic resolution in both space and time, which
will have great impact on physics, chemistry, biology and future technologies [22].
trum
(DC) and low frequency (LF) standards have been carried out for more than 100
systems have been developed by many national metrology institutes [25, 26]. Over
the last two decades, quantum Hall resistance standards and voltage standards
based on Josephson junction array have been developed to achieve higher precision
from his equations [33]. In 1887-1888, Heinrich Hertz demonstrated the existence
crowaves at 56.5 MHz and 500 MHz [34]. Radio frequency (RF) and microwave
gan in 1920s [40–42]. The microwave power [43–46], attenuation [47–54], phase-
shift [55–64], reflection coefficient [65–70], thermal noise [71–79], frequency and
phase noise [80–83] are the key microwave parameters for measurement via coaxial
4
1.1. Motivation and Background
cable or waveguide connection. The field strength and antenna gain are the key
parameters for microwave measurement in free space [84, 85]. Reference and mea-
surement standards for these parameters have been developed and established in
Optical measurement standards for radiometry and photometry have been well
established in many countries [92]. Most of the electromagnetic spectrum has been
used for various applications, such as radar, microwave ovens, radios, wireless com-
GHz to 300 GHz and sub-millimeter wave refers to the radiation from 300 GHz
to 3000 GHz [93, 94]. They are between the microwave and optics band in the
ment standards for power, attenuation, reflection coefficient, thermal noise, phase
noise and antenna at frequencies beyond 100 GHz are being developed by several
national metrology institutes [95]. Fig. 1.1 shows the measurement standards for
Figure 1.1: Measurement standards for various bands in the electromagnetic spec-
trum.
5
1.2. Objectives
1.2 Objectives
cations, radar, remote sensing, security imaging and non-destructive testing (NDT)
[38, 96–102]. Communication systems at 60 GHz and 300 GHz, automotive radars
at 77 GHz and imaging radars at 220 GHz have been developed [37,96,103–108]. A
security imaging system at 500 GHz has been developed to produce high-resolution
to detect microwave and millimeter-wave signals with high dynamic range (DR).
The microwave attenuation per unit length for some low-loss materials (loss
tangent < 0.0005) could be below 0.1 dB/cm at 10 GHz [110]. For thin dielectric
materials with a thickness of 0.1 mm [111], the microwave attenuation through the
material could be merely 0.001 dB. Hence very accurate attenuation measurement
sensitivity radiometers and receivers which are widely used in high-speed com-
munications, radar, imaging, non-destructive testing and material analysis [37, 38,
96, 98, 112]. High DR and calibrated measurement at microwave frequencies can
greatly improve the quality of images formed by microwave cameras for security
and biomedical imaging systems [38, 98]. A microwave imaging system requires a
6
1.2. Objectives
many national standard laboratories since 1950s. The measurement DR, accuracy
the ever-increasing demands from the industry and scientific research community.
dards have limited DR of 80 dB at 18–26.5 GHz [116]. This thesis will study how to
methods. The phase noise effect plays a key role in the noise floor of various mi-
crowave receivers. We will analyse this effect and investigate on new ways to reduce
the noise floor to a record low level. We will also investigate the leakage error in
measurement [117].
engineering. In a K-band satellite ranging system used for mapping the global
antenna surface map is measured accurately at 100 GHz [119]. Precision measure-
ment of insertion phase delay of the radome (a structural enclosure that protects
has achieved good accuracy (around 0.1◦ ) for a device with attenuation of 140
analyzer (VNA) could reach 19◦ at 18 GHz for a device with attenuation of 100
dB. The increase in phase measurement uncertainty is mainly due to the decreased
7
1.2. Objectives
signal-to-noise ratio (SNR) as the device attenuation increases. This thesis will
study how to achieve accurate phase-shift measurement at 0.05–18 GHz for devices
with high attenuation up to 140 dB. We will also investigate the verification of such
dB at 110 GHz, 0.23 dB at 220 GHz and 0.78 dB at 330 GHz for a 30 dB at-
needed for higher frequencies beyond 75 GHz to provide traceability for such mea-
surements.
quencies within 75–500 GHz, which have important applications in radar, com-
munications and security imaging. The frequency bands at 75–110 GHz (WR-10
band), 140–220 GHz (WR-05 band) and 330–500 GHz (WR-2.2 band) will be se-
lected to study the measurement accuracy and receiver noise effect at increasing
the three selected bands will show the mismatch error, leakage error and phase-
noise induced instability effect. Similar techniques can be used for attenuation
measurement at other frequency bands within 50–500 GHz. The uncertainty anal-
ysis at WR-10, WR-05 and WR-2.2 bands helps to predict the performance of
8
1.2. Objectives
VNA and an air line with 1.85 mm connector at 5–65 GHz [125]. An air line (with
1.85 mm connector) can be used as the phase-shift reference standard and provide
at 60 GHz [103, 104]. The increasing demand for fast wireless data transmission
higher accuracy is needed for frequencies around 300 GHz to provide traceability
at 220–325 GHz (WR-03 band). The WR-03 band has been selected to study
sign. Similar design can be used for phase-shift measurement at other bands within
system at WR-03 band can be used to predict the performance of phase-shift mea-
surement systems using dual-channel design at other bands within 50–500 GHz.
pected to have slightly better accuracy compared to the WR-03 band, due to
that at the WR-03 band, due to lower SNR and higher mismatch uncertainty.
9
1.3. Major Contributions
DR of our proposed design is much higher than those reported in the literature
kHz. We have proposed new ways to analyse the leakage error and noise floor
in attenuation measurement, which have not been well studied in the literature.
For frequencies beyond 2 GHz, the system employs a coherent frequency reference
channel to minimize the phase noise effect on lock-in amplifier sensitivity. The
proposed receiver achieves a noise floor around −172 dBm and a record high DR
of 186.7 dB at 26 GHz. Our receiver noise floor is 47 dB lower than that of a VNA
at 26 GHz. The proposed system can measure a step attenuator with attenuation
up to 170 dB at 26 GHz with good accuracy, which is not possible using other
measurement systems.
developed using dual-channel heterodyne receivers. Our design can suppress the
phase noise effect of the microwave sources and significantly improve the phase
measurement accuracy for high attenuation devices. The accuracy of this proposed
system is much better than earlier published results. The phase-shift measurement
uncertainty is 0.09◦ –0.32◦ for a 0–140 dB coaxial step attenuator at 18 GHz. The
GHz is ∼106 times smaller than that of a vector network analyzer (VNA) [133].
10
1.3. Major Contributions
which has not been studied in the literature. A coaxial beadless air line has been
system. The calculated phase-shifts with correction of the line insertion loss agree
tems at 75–110 GHz, 140–220 GHz and 330–500 GHz. A commercial VNA has a
typical DR of 105 dB at 200 GHz and 104 dB at 500 GHz [134]. The DR of our
proposed system is estimated to be 175 dB at 110 GHz, 163 dB at 220 GHz and
152 dB at 480 GHz, respectively, which are much higher than that of a VNA. The
∼10 times smaller than that of a VNA. The attenuation measurement is traceable
are estimated to be 0.084 dB at 110 GHz, 0.11 dB at 220 GHz and 0.24 dB at
480 GHz, respectively. The DR of the proposed system is 58 dB higher than that
of a VNA at around 220 GHz. A voltage ratio of 3.0 × 10−5 at 480 GHz can be
proposed dual-channel heterodyne receivers have achieved a record low noise floor
at around −170 dBm at 480 GHz. The measurement systems have been verified
been proposed. The proposed dual-channel receiver design can measure the phase
for a 0–60 dB RVA at 325 GHz. The phase-shift measurement uncertainty of the
proposed system for a 50 dB attenuator at 325 GHz is ∼44 times smaller than
11
1.3. Major Contributions
that of a VNA [122, 124]. A waveguide line standard has been measured as a
calculated phase-shifts agree with the measured values of the line standard within
since their noise floor is quite high due to the phase noise effect. This thesis has
shown that our proposed dual-channel heterodyne receiver design has significantly
reduced the measurement error in magnitude and phase due to phase noise at
signal using a lock-in amplifier with precise knowledge of the signal’s instantaneous
frequency.
sis to characterise its performance [129, 135–138]. We have evaluated the measure-
ment errors and uncertainties of the six new measurement standards proposed in
this thesis. It has been very difficult to estimate the leakage error in high attenua-
tion measurement since the leakage signal can not be separated from the noise [117].
The leakage error in phase-shift measurement has not been well studied in the liter-
ature. We have proposed new methods to evaluate the leakage errors in attenuation
and phase-shift measurement for devices with high attenuation. This is critical for
These systems have been used as the Singapore national standards for attenu-
Variable attenuators, air lines and waveguide shims calibrated by these systems
can provide traceability and serve as reference standards for calibration of VNAs,
power meters, radiometers, spectrum analyzers, etc. The dual-channel design using
12
1.4. Organization
multipliers and subharmonic mixers can be used at 0.5–1.1 THz and the receiver
noise floor is expected to be much lower than that of a VNA since it is mainly
1.4 Organization
4, the phase noise effect, noise floor and dynamic range analysis of the attenuation
measurement receiver from 50 MHz to 26.5 GHz are presented. A 120–170 dB step
ment uncertainty has been carefully analysed. The measurement system is verified
110 GHz, 140–220 GHz and 330–500 GHz are presented in Chapter 6. The mea-
13
1.4. Organization
Finally, conclusions and recommendation for future work are given in Chapter
8.
14
Chapter 2
Research Background
Compared to signals at lower frequency range, microwave signals have higher prop-
agation loss during transmission through cables or waveguides. The reflected power
will be significant if proper impedance match between the devices are not achieved.
measurement system [139, 140] has been used for such measurements. Various
and noise temperature measurement which has led to the discovery of the CMBR
at around 3.5 K [5, 6, 76]. Microwave attenuators are also used to check the noise
15
2.1. Microwave Attenuation Measurement
check and correct the linearity errors of various microwave power meters, network
analyser and spectrum analysers [91, 139, 140]. Thus attenuation standard is very
important to other microwave standards for power, reflection coefficient and noise.
Insertion Loss
ΓG . If a 2-port network is inserted between the generator and the load, the power
dissipated in the load is denoted as P2 . The insertion loss [140, 141] of the 2-port
network is defined by
P1
LI = 10 log10 (2.1)
P2
The attenuation of the 2-port network is defined as the insertion loss when ΓG = 0
a1 b2
2-port
Generator ΓG ΓL Load
network
b1 a2
Fig. 2.1 shows the 2-port network, the generator and the load. Let bG denotes
b2 s21
= (2.4)
bG (1 − ΓG s11 )(1 − ΓL s22 ) − ΓG ΓL s12 s21
16
2.1. Microwave Attenuation Measurement
|b2 |2
P2 = (1 − |ΓL |2 ) (2.5)
Z0
The power dissipated in the load when the generator is directly connected to
the load can be derived by letting s11 = s22 = 0 and s21 = s12 = 1:
|bG |2 (1 − |ΓL |2 )
P1 = (2.7)
Z0 |1 − ΓG ΓL |2
1
A = 20 log10 (2.9)
|s21 |
Substitution Loss
When a 2-port network with scattering parameters s11b , s12b , s22b and s21b is con-
nected between the generator and the load, the power dissipated in the load is
17
2.1. Microwave Attenuation Measurement
Suppose this 2-port network is replaced by another one with scattering parameters
s11e , s12e , s22e and s21e , then the power dissipated in the load becomes
P3
Ls = 10 log10 (2.13)
P4
and we have
The attenuation difference between the initial and final 2-port network is
1 1
Ainc = 20 log10 − 20 log10 (2.15)
|s21e | |s21b |
Ainc is also called the incremental attenuation. We can see that Ainc = Ls when
able attenuation standard [1, 41, 47, 48, 52, 150–153]. It propagates a single evanes-
attenuator can be predicted from its dimensions with very high accuracy. A WBCO
18
2.1. Microwave Attenuation Measurement
with its field strength having an exponential decay along the axis. The amount
dimensions. A moving probe with coupling of the field inside the guide will give
which requires less correction due to skin depth effect. This frequency is also
The NIST has built a WBCO attenuator with a waveguide diameter of 3.2 in
and it has achieved an attenuation of 10 dB per inch [153]. The accuracy is close
WBCO attenuators have been made by the National Physical Laboratory (NPL)
over a range of 120 dB. The guide has a diameter of about 2 in, with an attenuation
rate of 16 dB per inch [153]. Fig. 2.2 shows a photo of the WBCO attenuator at
19
2.1. Microwave Attenuation Measurement
A rotary vane attenuator (RVA) has been widely used as a direct reading attenuator
where θ is the angle of rotating vane relative to the TE11 mode’s polarization. C
misalignment and mechanical components errors can cause the actual attenuation
to deviate from that predicted by the above equation. The deviation is small at
low attenuation values, but can be quite large at high attenuation values. The
RVAs are available from 2.6 GHz to 500 GHz. These attenuators, after cross
reference attenuators. Fig. 2.3 shows a RVA used in our laboratory at the National
Metrology Centre.
Figure 2.3: A 0–60 dB RVA in our laboratory at the National Metrology Centre.
20
2.1. Microwave Attenuation Measurement
switching repeatability and life span (greater than 5 million switching cycles per
section). High attenuation accuracy and low reflection coefficient are achieved using
miniature thin-film attenuation cards. Such coaxial step attenuators can serve as
uation measurement system. Fig. 2.4 shows a coaxial reference step attenuator at
0.01–26.5 GHz in our laboratory at the National Metrology Centre. This reference
'
)
Figure 2.4: A reference step attenuator used in our laboratory at the National
Metrology Centre.
21
2.1. Microwave Attenuation Measurement
An inductive voltage divider (IVD) is an autotransformer that has its winding di-
vided into a number of equal-turn sections (usually 10) so that when an alternating
voltage V is applied to the whole winding, the voltage across each section is nom-
inally V /10. The progressive voltages from one end to the section junctions are
thus V /10, 2V /10, 3V /10,..., and 9V /10. Fig. 2.5 shows a 4-decade IVD, where
OUTPUT
Figure 2.5: A 4-decade inductive voltage divider, where the input voltage is V and
This voltage division can be realized with errors considerably less than 1 part
per million of V , and such units find wide use as standards of LF voltage ratio in
the discipline of electrical measurements [25, 140, 159]. The division of voltage will
be in error if there are differences of resistance and leakage inductance from section
to section, and these errors will be significant if the differences are significant in
relation to the input impedance of the winding. IVDs have been carefully designed
to have almost equal resistance and leakage inductance of the sections and a very
high value of input impedance. Thus, voltage division at low audio frequencies can
be accurate to a few parts in 100 million. IVDs operate most accurately in the
22
2.1. Microwave Attenuation Measurement
1 MHz. Fig. 2.6 shows a programmable seven-decade IVD used in our laboratory
Metrology Centre.
to 1970s by the national metrology institutes in the US, UK and Germany are
using bolometer detector and stabilized signal source are required to measure the
microwave attenuation using microwave power ratio method [51, 54, 127, 129, 138,
140, 160]. Fig. 2.7 shows an attenuation measurement system using power ratio
method.
DUT Power
Detector
Source
23
2.1. Microwave Attenuation Measurement
comparing the attenuation of the device under test (DUT) with that of a gauge-
Intermediate frequency (IF) substitution technique has been widely used where a
WBCO standard attenuator is used to compare with the DUT at IF and provide
the reference attenuation values [51, 53, 54, 142]. A mixer is used to convert high
standard.
limited in DR due to the detection technique and is usually costly to maintain due
IF (30 MHz)
DUT WBCO Detector
Source
LO
tion method.
24
2.1. Microwave Attenuation Measurement
Audio frequency (AF) substitution method was developed later to make attenua-
tion measurement with higher accuracy [126, 140, 141, 161]. In this technique, the
AF (1 - 50 kHz)
DUT IVD Receiver
Source
Reference
(1 - 50 kHz)
LO
method.
is usually used as a traceable low frequency voltage ratio reference standard for the
The vector network analyzer (VNA) has been developed since 1960s to measure the
microwave transmission and reflection coefficient of a two-port device [89, 90, 167,
168]. A 2-port VNA uses four heterodyne receivers to measure the magnitude and
25
2.1. Microwave Attenuation Measurement
phase of the forward incident reference, reverse incident reference, reflected and
to collect the reflected microwave signal from the DUT. The four receivers convert
Source
Splitter Splitter
Splitter
LO
26
2.2. Microwave Phase-shift Measurement
Table 2.1 compares the merits and pitfalls of various methods for microwave at-
is below 100 dB for frequencies up to 18 GHz [126, 127, 129, 132, 138, 140]. It is
Power ratio [127, 129] Traceable to a power standard; DR limited to 90 dB at 0.01–18 GHz
DR limited to 100 dB
AF substitution [126, 132] Traceable to an IVD, more DR limited to 100 dB at 0.1–18 GHz;
phase-shift standards [57,61,62] have also been built to provide phase-shift reference
27
2.2. Microwave Phase-shift Measurement
θ = arg(s21 ) (2.18)
Differential Phase-shift
When a 2-port network with scattering parameters s11b , s12b , s22b and s21b is con-
nected between the generator and the load (as shown in Fig. 2.1), the output wave
b2b is given by
b2b s21b
= (2.19)
bG (1 − ΓG s11b )(1 − ΓL s22b ) − ΓG ΓL s12b s21b
When this 2-port network is replaced by another one with scattering parameters
s11e , s12e , s22e and s21e , the output wave b2e is given by
b2e s21e
= (2.20)
bG (1 − ΓG s11e )(1 − ΓL s22e ) − ΓG ΓL s12e s21e
when ΓG = ΓL = 0.
28
2.2. Microwave Phase-shift Measurement
Insertion Phase-shift
When a 2-port network with scattering parameters s11 , s12 , s22 and s21 is inserted
between the generator and the load (as shown in Fig. 2.1), its insertion phase-
shift can be measured. The insertion phase-shift is a special case of the differential
phase-shift, where the initial 2-port network is lossless, non-reflecting and has zero
characteristic phase-shift (these conditions are s11b = s22b = 0 and s21b = s12b = 1).
Notice that
θI = arg(s21 ) (2.25)
when ΓG = ΓL = 0.
64, 172, 173]. The phase difference between the two ports of the transmission line
θ = βl (2.26)
where β is the propagation constant of the transmission line, and l is the length of
Air-filled coaxial lines have been used as phase-shift reference standards in many
metrology laboratories. The phase-shift of an air line can be predicted with correc-
tion of its insertion loss factor [63, 64, 68, 69, 173–176]. Fig. 2.11 shows a beadless
29
2.2. Microwave Phase-shift Measurement
Figure 2.11: A beadless air line in our laboratory at the National Metrology Centre.
at 9 GHz have been described in [55, 60, 61], where the phase-shift of a DUT is
null system with a phase-shift standard, which provides phase-shift reference values
171, 178]. Expanded measurement uncertainties of 1–3 degrees have been achieved
attenuations up to 140 dB), with an accuracy of 0.1 degrees at 60 MHz [121, 179].
[89, 167] with an uncertainty of 0.2–7 degrees for a device with attenuation up to
30
2.2. Microwave Phase-shift Measurement
by a VNA could reach 19 degrees at 18 GHz if the device attenuation is 100 dB.
increased phase noise level at higher frequency and the decreased signal-to-noise
Table 2.2 compares the merits and pitfalls of various methods for microwave
at 0.1–18 GHz for a device with 140 dB attenuation, a new measurement system
Phase-shift standard based [55, 60, 177] Traceable to phase-shift standard Need several phase-shift standards
IF phase meter based [61, 171, 178] Have broad frequency coverage IF phase meter error is high (0.5◦ –3◦ )
Dual-channel RF phase-shift Uncertainties are 0.1◦ for DUT For low frequency only (below 100 MHz)
Vector Network Analyzer [133, 167] Wide frequency range, lower cost Uncertainties reach 19◦ for devices
31
2.3. Millimeter-wave Technology and Metrology
cations systems with high data rate have been developed at 60 GHz, 120 GHz and
300 GHz [103, 104, 106, 107]. Automotive high-resolution radars have been devel-
oped at 77 GHz, 94 GHz and 140 GHz [37,96,105]. Imaging radars at 94 GHz, 220
GHz and 350 GHz have been reported [108,189,194]. Sub-millimeter wave imaging
monolithic integrated circuits and modules at 300–500 GHz have been developed
Electronics sources can generate continuous wave (CW) output power of 3.5 mW
at 300 GHz and 100 µW at 1.2 THz [199, 200]. The first THz quantum cascade
laser (QCL) developed in 2001 can generate CW power of 2 mW at 4.4 THz [201].
QCLs operating at frequencies in the range from 0.84 THz to 5 THz have been
cies is much higher than that at microwave frequencies. On-chip metal-pipe rect-
nologies [203, 204]. These technologies reduce the transmission loss between the
32
2.3. Millimeter-wave Technology and Metrology
Millimeter-wave transmission loss through free space is high due to water vapor
currently. Thus highly sensitive receivers are required to measure low level signals
[205, 206]. The power measurement range of a calorimeter is limited by the sensor
drift and noise level. For millimeter-wave power below 1 microwatt, more accurate
measurement can be made using a heterodyne receiver with lower noise floor. A
provide linearity reference value for low-level power measurement made by a het-
wave noise standard is required to characterize the noise level in imaging and spec-
band can be used in a radiometer to provide linearity traceability for various noise
levels.
33
2.3. Millimeter-wave Technology and Metrology
IF attenuator was used as the reference attenuation standard. The systematic error
and random error are ±0.3 dB and ±0.1 dB, respectively, for measurement of 50
tion attenuation measurement system at 33–110 GHz was proposed in [210], which
used a video detector to reduce the system cost. A calibrated RVA was used as
system is about 97 dB with an averaging of 1.7 minutes to reduce the noise floor.
receiver is about −112 dBm at 140 GHz, and the attenuation measurement DR can
reach 142 dB if a backward wave oscillator (BWO) with output power of 1 watt
is used as the transmitter. However, a BWO is quite bulky and the phase noise
of a BWO is quite high which leads to high receiver noise floor. An attenuation
and 29 dB at 420 GHz. These two systems did not use any attenuation standard
extend the frequency coverage of a microwave VNA to cover 50–1100 GHz. These
ous waveguide bands from WR–15 to WR–1.0. And subharmonic mixers are used to
34
2.3. Millimeter-wave Technology and Metrology
provide a reference IF signal for VNA calibration and synchronous detection. The
down-convert the main and reference IF signals to a lower frequency for coherent
expanded uncertainty of 0.33 dB at 110 GHz, 0.23 dB at 220 GHz and 0.78 dB at
330 GHz, respectively [122–124]. The VNA extender from Virginia Diodes (VDI)
(ABmm) for attenuation measurement at 8–1000 GHz [134]. The ABmm network
analyzer has achieved a DR of 105 dB at ∼200 GHz and 104 dB at ∼500 GHz [134].
analyzer for attenuation measurement. In the ABmm network analyzer, the phase
which drives the harmonics generator and harmonics mixer, respectively [218].
However, the phase-lock loop used in the VNA or ABmm network analyzer has
some residual random phase error, which leads to random deviations between the
instantaneous frequency of the IF signal and the LO in the vector receiver. Such
random deviations increase the noise floor in the synchronous detection, which
has been developed at the National Metrology Institute of Japan [117]. The ex-
35
2.3. Millimeter-wave Technology and Metrology
to provide traceability for measurement made at 60 GHz, which is very useful for
wireless communications.
Single-channel heterodyne receiver [36] High DR (142 dB) Need bulky BWO with
at 500 GHz
Dual-channel VNA [122–124] Wide frequency range, easy to use DR limited to 100 dB
at 50–75 GHz
Table 2.3 compares the merit and pitfalls of various methods for millimeter-wave
36
2.3. Millimeter-wave Technology and Metrology
A 70 GHz phase-shift measuring system has been developed for plasma diagno-
sis [219]. This system used a single-channel homodyne receiver to derive the
coaxial air line can be used as millimetre-wave phase-shift reference standard for
the DUT in linear scale and U (|s21 |) is the expanded measurement uncertainty of
Table 2.4 compares the merit and pitfalls of various methods for millimeter-
37
2.3. Millimeter-wave Technology and Metrology
Single-channel homodyne receiver [219] Lower cost Limited DR, not suitable
Single-channel network analyzer [134] Wide frequency range, Uncertainties too high for DUT
Dual-channel VNA [122–124] Wide frequency range, Uncertainties too high for DUT
higher cost
system is required to provide traceable phase-shift reference values with high accu-
dard to provide traceability for VNA and phase measurement made by millimeter-
38
Chapter 3
Precision Measurement of
Microwave Attenuation
3.1 Introduction
for microwave power meters, spectrum analyzers and network analyzers [115, 221].
been developed by many national standards laboratories since 1960s to make very
dedicated attenuation measurement systems have been developed in the past two
decades to cover higher frequencies [116, 121, 127, 129–132, 145, 147, 160, 162, 163,
165, 166].
39
3.1. Introduction
Commercial VNA has been used in some national standards laboratories as at-
GHz using a VNA is around 0.3 dB. At 26 GHz, the direct receiver access input
quency bandwidth and its DR is around 125 dB with 0.45 dB receiver compression
error [223, 224]. The leakage error in VNA also limits its attenuation measurement
at 30 MHz with an uncertainty of 0.079 dB at the 150 dB range [150, 177]. At the
NPL, a measurement DR of 150 dB from 0.5 MHz to 100 MHz was achieved using
attenuation. These two systems only cover lower frequency range which is below
100 MHz. A voltage-ratio based attenuation measurement standard has also been
developed at NPL with a DR of 100 dB at 0.1–18 GHz [126]. This system did not
This system used an AF reference channel to provide a frequency reference for the
LIA. The main channel AF signal and the reference AF signal go into the A and B
40
3.1. Introduction
input of the LIA respectively (differential mode) so that a nulling can be detected.
its amplitude and phase can be adjusted before it enters the LIA. When the DUT
attenuation changes, an IVD in the main channel and the phase-shifter module in
the reference channel are adjusted respectively so that the nulling at the LIA can
a saturation amplifier and a bandpass filter [116, 130]. The saturation amplifier is
used to minimize the amplitude change in the output signal of the AF phase shifter
and the bandpass filter produces the fundamental mode of the saturated signal.
However, the phase noise in the AF reference signal is changed by the amplifier
and the bandpass filter, which causes more deviations between the instantaneous
frequency of the main channel AF signal and the reference AF signal at the LIA
input. This would lead to more fluctuations in the LIA reading and increased noise
In this chapter, novel measurement techniques are presented for accurate mea-
proposed for attenuation measurement at 2–26.5 GHz. Our proposed design did
not use the AF reference signal for nulling detection. We ensure that the phase
noise of AF reference signal is the same as that of the main channel AF signal, so
that there is negligible deviations between the instantaneous frequency of the main
channel AF signal and the reference AF signal at the LIA input. Thus the receiver
50 MHz to 2 GHz due to its simpler design and good accuracy. We have compared
ment errors due to receiver fluctuation and leakage, which has not been reported
41
3.2. Single-channel AF Substitution Attenuation Measurement
Measurement
0.05–26.5 GHz is shown in Fig. 3.1. In this system, the microwave signal is con-
verted to a 5.02 kHz AF signal. One mixer is used to cover 50 MHz to 2 GHz
and another mixer is used to cover 2–26.5 GHz. The additional 20 Hz shift away
from 5 kHz is to minimize the interference from the power line (50 Hz) harmonics.
The AF signal goes through a LC low pass filter, pre-amplifier, IVD and enters
the input of a lock-in amplifier (LIA) [226, 227]. A function generator is used to
provide a 5.02 kHz sinusoidal signal as the external frequency reference signal for
the LIA. This reference is used by the LIA to recover the AF signal. A 10 MHz
frequency reference signal synchronizes the frequency of the RF source and the
function generator.
Two RF isolators are placed before and after the matching tuner to reduce the
reflection from the RF source and the mixer RF input respectively. The isolators
are available in the following bands: 50–70 MHz, 80–140 MHz, 150–225 MHz, 225–
400 MHz, 400–500 MHz, 500–1000 MHz, 1–2 GHz, 2–4 GHz, 4–8 GHz, 8–12 GHz,
12–18 GHz and 18–26.5 GHz. To reduce the mismatch uncertainty [228], manual
tuners (for 800 MHz – 26.5 GHz) or matching pads (for 50 – 800 MHz) are used to
provide good test port impedance match. The tuners and the isolators connected
to the two sides of DUT have also been placed in two separate shield boxes to
42
3.2. Single-channel AF Substitution Attenuation Measurement
Gauge Block
Attenuator
RF Source (inserted when using
indirect attenuation
(50 MHz – 26.5 GHz) measurement) AF (5.02 kHz)
Port 1 Port 2
RF
Impedance Impedance LP Filter
Tuner or Tuner or
Pad Pad
Isolator LO
Level Set Attenuator Isolator
(inserted when using Pre-Amp
direct attenuation
measurement) DUT
BP Filter
IVD
LO Source
(50 MHz – 26.5 GHz)
10 MHz Lock-in
Reference Amp
Ref
Function Reference (5.02 kHz)
Generator
ation measurement system (50 MHz – 26.5 GHz). The frequency reference signal
Direct Measurement
The DUT is usually a step attenuator. When a DUT is set to its datum position
(0 dB range), the IVD ratio is set to D1 and the LIA detects an AF signal with
magnitude of V1 . After the DUT is set to a particular attenuation range, the IVD
ratio is adjusted via a computer program so that the LIA detects a signal with
D2
A = 20 log10 ( ) (dB). (3.1)
D1
The IVD is used as a traceable voltage ratio reference standard with an accuracy
43
3.2. Single-channel AF Substitution Attenuation Measurement
Indirect Measurement
For measurement of a DUT with attenuation higher than 30 dB, the LSA (30 dB) is
For measurement of DUT attenuation from 30 to 60 dB, the GBA is first set to
30 dB when DUT setting is 0 dB, and then changed to 0 dB when DUT is set to
a value higher than 30 dB. The difference between the incremental attenuation of
where AGB,30dB is the calibrated incremental attenuation of the GBA from its
where δA,2 is the measured difference between the incremental attenuation of the
DUT and the 60 dB GBA. AGB,60dB are the calibrated incremental attenuation of
44
3.3. Dual-channel AF Substitution Attenuation Measurement
The LIA relies on the frequency information provided by the external reference
signal to recover a weak signal accurately. Due to the phase noise of RF and LO
sources, the AF signal generated by the mixer has a random frequency fluctuation
which increases as the frequency of the microwave signal becomes higher than 2
GHz. Since the reference signal produced by the AF function generator does not
have the same frequency fluctuation, the voltage reading at LIA becomes less stable
as the frequency of the microwave signal increases. This causes larger uncertainty
GHz.
Measurement
the receiver sensitivity. The signal from the RF source is split via a directional
coupler. The main line signal goes into the main mixer input (signal channel) and
the coupled signal goes into the reference mixer input (reference channel), which is
used to provide a reference signal at 5.02 kHz for the LIA. The bandwidth of the
band-pass filter in the reference channel is set to 20 Hz. The LO signal is also split
via a directional coupler to provide LO drive signals for both mixers. The dual-
channel system can be used for attenuation measurement from 50 MHz to 26.5 GHz.
Two mixers are used to cover 50 MHz to 2 GHz and two mixers are used to cover 2–
45
3.3. Dual-channel AF Substitution Attenuation Measurement
26.5 GHz. The directional couplers used in the system are in the following bands:
50–500 MHz, 0.5–1 GHz, 1–18 GHz and 18–26.5 GHz. With the introduction
of the reference channel in the dual-channel system, the voltage reading at LIA
becomes much more stable for microwave signal at frequencies beyond 2 GHz.
Gauge Block
Attenuator
(inserted when using
indirect attenuation Main
RF Source
measurement)
(50 MHz ~ 26.5 GHz ) Isolator Mixer AF (5.02 kHz)
Isolator
Impedance Impedance RF
Tuner or Tuner or LP Filter
Pad Pad
LO
Level Set Attenuator Port 1 Port 2
(inserted when using
Pre-Amp
Attenuator
direct attenuation
(60 dB) measurement) DUT
Ref BP Filter
Mixer RF
LO
IVD
Isolators in
multiple bands LO Source
(50 MHz ~ 26.5 GHz)
Lock-in
10 MHz
Amp
Reference
AF (5.02 kHz)
LP Filter Pre-Amp BP Filter Reference (5.02 kHz)
system (50 MHz – 26.5 GHz). The frequency reference signal of the lock-in amplifier
There is an internal leakage path going through the reference mixer (from RF
to LO port), isolators, coupler and enters the main mixer RF port (from LO to
RF port). For large attenuation measurement, the leakage path attenuation must
dB. Ferrite RF isolators (reverse isolation nearly 145 dB) are inserted between
reference mixer LO port and the directional coupler to provide sufficient leakage
attenuation. These isolators have been placed inside several shield boxes to reduce
46
3.4. Measurement Uncertainty Analysis
The AF output of the reference mixer shares the same frequency fluctuation of
the signal channel AF signal since it is due to the phase noise of the RF source and
LO source. This is very important for weak signal recovery and coherent detection.
With the introduction of the reference channel, the voltage reading of the LIA
becomes much more stable for microwave signal at frequencies higher than 2 GHz.
A seven decade IVD is used as the voltage ratio reference standard. The maximum
where γ1 and γ2 are the IVD ratio maximum absolute error when IVD is set to D1
[−eIV D , eIV D ]. The standard uncertainty (coverage factor k=1 ) [135] due to the
The IVD is used between the pre-amplifier and the LIA. Thus the pre-amplifier
acts as the source of the IVD. The LIA acts as the load of the IVD. The loading
effect has been analyzed. The error limit due to the IVD loading effect is estimated
47
3.4. Measurement Uncertainty Analysis
estimation.
to have a U-shaped distribution and its standard uncertainty can be estimated from,
h
uM = 6.143 × |ΓG |2 (|s11b |2 + |s11e |2 ) + |ΓL |2 (|s22b |2 + |s22e |2 )
i1
2 2 4 4 2
+ |ΓG | |ΓL | (|s21b | + |s21e | ) (3.7)
where ΓG and ΓL are the test port reflection coefficients, s11b and s11e are the
reflection coefficients of the DUT’s port 1 before and after the DUT attenuation
switching [228]. Similarly, s22b and s22e are the reflection coefficients of the DUT’s
port 2, s21b and s21e are the transmission coefficients before and after the DUT at-
From 800 MHz to 26.5 GHz, impedance matching tuners can be used to reduce
the test port reflection coefficient. A VNA is used to monitor the tuning so as to
Below 800 MHz, two 10 dB attenuators are used as the matching pads at test
ment of high attenuation from 60 dB to 110 dB, two 6 dB attenuators are used as
matching pads so that the signal-to-noise ratio is sufficient for accurate measure-
ment.
48
3.4. Measurement Uncertainty Analysis
signal level at the mixer RF input is kept at least 30 dB below the LO level to
dB, a 30 dB attenuator is placed before the DUT to ensure the mixer is operating
step attenuator at decreasing initial mixer input levels ( −25 dBm, −35 dBm,...,
−85 dBm ) at 5 GHz. The mean value of four measurements taken at initial mixer
input levels from −25 dBm to −55 dBm is used as the reference attenuation value
of the 0–10 dB attenuator (Aref = 9.9814 dB). For initial mixer input levels from
−25 dBm to −65 dBm, the deviations of the measurements from the reference
attenuation value Aref are taken as the nonlinearity error (shown in Table 3.1).
The repeatability for the measurement at each mixer input level is obtained by
√
dividing its standard deviation by n, where n is the number of measurements for
each mixer input level (n = 4). The maximum nonlinearity error for the initial
mixer input levels from −25 dBm to −65 dBm is about eln = 0.001 dB. The
increased deviations for initial mixer input levels at −75 dBm and −85 dBm are
tribution in [−eln , eln ], and the standard uncertainty due to nonlinearity error is
estimated to be
eln
uln = √ = 0.0006 dB. (3.8)
3
49
3.4. Measurement Uncertainty Analysis
The detected signal by the LIA has small fluctuation in amplitude during mea-
surement. Such fluctuation will lead to some uncertainty in the attenuation mea-
surement since the IVD ratio setting is determined by the LIA voltage reading.
The dual-channel system uses a coherent frequency reference for the LIA and
has very small receiver fluctuation for low attenuation measurement up to 26.5
GHz. Compared to the single-channel system, the error due to receiver fluctuation
For high attenuation measurement, receiver fluctuation is mainly due to the noise
in the mixer, pre-amplifier and the LIA. For 90 – 110 dB attenuation measurements,
25 readings of LIA measurement taken within 25 seconds (one sample per second)
Let D1 and D2 denotes the IVD ratio before and after the DUT attenuation is
changed, and δ1 and δ2 represents the maximum absolute error in the IVD ratio
D1 and D2 due to the receiver fluctuation and noise, respectively. The maximum
50
3.4. Measurement Uncertainty Analysis
The error due to fluctuation and noise is assumed to have a rectangular distribution
in [−en , en ]. The standard uncertainty due to receiver fluctuation and noise can
be estimated by
en
uN = √ (3.10)
3
above 80 dB [140]. The resultant signal amplitude at the LIA due to the paths
where xA and xL are the magnitudes of the signals at the LIA via the DUT and the
leakage paths, respectively, with a phase difference φ between them. The value of
φ is unknown and depends upon the path differences. The maximum measurement
where Adut and AL are the attenuations through the DUT and leakage paths re-
To keep the eL to be smaller than 0.001 dB, the leakage path signal xL has to be
51
3.4. Measurement Uncertainty Analysis
External Leakage
There is an external leakage effect caused by radiated signals from the isolators,
slide screw tuners, coaxial cables, and coaxial connectors. These radiated signals
might enter the attenuation measurement receiver and introduce an error in atten-
foil tapes and semi-rigid cables with high shielding effectiveness are used to reduce
the radiation leakage effect. The slide screw tuners and the RF isolators connected
to the two ports of DUT have been placed in two separate shield boxes to reduce
their RF radiation. The main measurement receiver and reference receiver are also
Internal Leakage
reference mixer, isolator, directional coupler and enters the signal mixer RF port.
RF isolators are used in the path between LO port of two mixers to reduce such
to reduce the internal leakage. Directional couplers with high directivity are used
to provide ∼40 dB isolation between the two channels. The isolation between the
mixer RF port and the LO port is about 30 dB. The total isolation on the internal
leakage path is estimated to be ∼334 dB if four isolators (each with ∼36 dB reverse
between the test ports of the dual-channel system at 26 GHz. The receiver noise
52
3.4. Measurement Uncertainty Analysis
where nf (t) and θf (t) are the magnitude and phase of the noise floor in LIA, nt (t)
and θt (t) are the magnitude and phase of the random noise, and xL (t) and θL (t)
are the magnitude and phase of the residual leakage signal, respectively.
Fig. 3.3 shows the magnitude and phase of the measured signal at receiver. If
there is any significant leakage signal in the noise floor detected by LIA, it would
be expected to see a relatively stable phase in the data collected in 120 seconds.
However, the phase of the measured signal changes randomly between −180◦ and
180◦ , which suggests that the measured signal is actually the random noise. The
magnitude of the leakage signal must be much smaller than that of the random
noise. The measured signal is mainly due to the thermal noise in the receiver.
53
3.4. Measurement Uncertainty Analysis
Magnitude (microvolt)
1.5
0.5
0
0 20 40 60 80 100 120
time (sec)
200
150
100
Phase (deg)
50
−50
−100
−150
−200
0 20 40 60 80 100 120
time (sec)
Figure 3.3: The magnitude and phase of the signal detected by LIA with ∼370
and
54
3.4. Measurement Uncertainty Analysis
where AGB is the calibrated incremental attenuation of the GBA and δA is the
attenuation difference between the DUT and the GBA. The GBA has been cali-
brated previously. Thus the uncertainty for the indirect attenuation measurement
where n
P
xk
k=1
x̄ = (3.17)
n
and xk are the measurement data for each attenuation step and n is the number
nonlinearity, error due to receiver fluctuation and noise, IVD error, GBA uncer-
Us = k us (3.19)
55
3.5. Comparison of Single-channel and Dual-channel System
where k is the coverage factor and k=2. Us defines an interval estimated to have a
System
and can cover a broad band since it does not require multiple isolators to reduce
rate measurement at frequencies beyond 2 GHz. Fig. 3.4 shows the comparison of
to 26.5 GHz.
dB from 50 MHz to 26.5 GHz. The difference between the two systems becomes
56
3.5. Comparison of Single-channel and Dual-channel System
0.07
0.05
0.04
0.03
0.02
0.01
0
5 10 15 20 25
Frequency (GHz)
Figure 3.4: Comparison of the standard uncertainty due to LIA reading fluctuation
GHz due to its simpler design. From 2 to 26.5 GHz, a dual-channel system has to
For measurement of high attenuation above 80 dB, the error due to noise become
more significant compared with other measurement errors. To compare the perfor-
57
3.5. Comparison of Single-channel and Dual-channel System
replace the GBA so that the signal level at the receiver input can be reduced. The
intention is to obtain the same receiver input level when an 80–110 dB attenuator
(DUT) is inserted into the system. The DUT’s incremental attenuation from 0 dB
70 dB in 10 dB steps.
With various LSA settings, the receiver noise effect on high attenuation mea-
surement (80–110 dB) using the two systems can be compared. The signal mixer’s
RF input level is 76 dB below the LO drive signal level when input attenuator is set
to 40 dB and decreases to 106 dB below the LO drive signal level when input at-
tenuator is 70 dB, which simulates the mixer input level when a 110 dB attenuator
Table 3.2 compares the linearity of the single-channel and dual-channel systems
channel system. The deviation shown in Table 3.2 indicates the difference between
Table 3.2: Comparison of linearity of the single-channel and the dual-channel sys-
As the RF signal level at mixer input decreases to 106 dB below LO drive level
(LSA set to 70 dB), the measurement deviation increases to 0.036 dB using the
58
3.6. A Broadband Attenuation Measurement System at 0.05–26.5 GHz
the receiver noise effect at low signal level. The data shows that the dual-channel
When the frequency of microwave signal increases beyond 5 GHz, the single-
channel system will have even larger error for high attenuation measurement, as
suggested by the comparison shown in Fig. 3.4. This experiment shows that the
dual-channel system performs better in recovering the signal at very low level and
There will be large deviation due to microwave leakage if the internal isolation
and the shielding in the dual-channel system are not sufficient. single-channel
system does not have internal leakage problem and the large deviation of 0.036
dB (when the LSA is 70 dB) is due to the receiver noise. The smaller deviation
achieved by the dual-channel system suggests that the RF leakage error has been
reduced to a level below the receiver noise floor. Thus the internal isolation and
Matching pads are used to reduce the mismatch uncertainty in the frequency range
from 50 MHz to 800 MHz. Mismatch uncertainties are estimated using a step at-
tenuator with PC 3.5 mm connector as the DUT. The uncertainty due to receiver
fluctuation and noise is one of the dominant uncertainty components for mea-
59
3.7. Measurement of a 0–110 dB Step Attenuator
and 0.016 dB for a 110 dB attenuator. Table 3.3 gives the uncertainty budget for
Combined
0.0036 0.0058 0.0077 0.0075 0.0083 0.0073 0.0078
uncertainty
Expanded
0.0072 0.012 0.015 0.015 0.017 0.015 0.016
uncertainty (k=2)
dB synthesized step attenuator. The two step attenuators are called DU T1 and
measurement receiver.
Table 3.4 shows the settings of two step attenuators and the mean of 10 mea-
surements for each attenuation step. The repeatability and estimated expanded
60
3.8. Summary
measurement uncertainties (coverage factor k=2) are also given in Table 3.4. The
repeatability for each attenuation value is obtained using (3.16). The measure-
uator (formed by connecting two 0-90 dB step attenuators which are referred to as
DU T1 and DU T2 ) at 26 GHz.
3.8 Summary
have been proposed in this chapter. Our proposed design has a DR higher than
more suitable for attenuation measurement from 50 MHz to 2 GHz due to its simpler
design and good accuracy. A dual-channel system has to be used for attenuation
61
3.8. Summary
measurement from 2 to 26.5 GHz since the receiver reading is more stable in this
frequency range.
5.02 kHz. Attenuation from 0 to 110 dB at 0.05–26.5 GHz can be measured with
high accuracy. The measurement dynamic range of the proposed system at 18–26.5
earlier in [116].
Measurement uncertainty analysis of the broadband system has been given. The
0.016 dB for a 0–110 dB variable attenuator at 26 GHz. The systems have been
and 5 GHz [230,231]. Nine metrology institutes of APMP participated in this com-
parison. Our measurement values in this comparison agree with the comparison
62
Chapter 4
Microwave Attenuation
Measurement Receiver
4.1 Introduction
The phase noise has been an important factor which makes it difficult to precisely
In this chapter, the noise floor of the proposed attenuation measurement receiver
is analysed. The phase noise effect in a phase sensitive detector has been modeled,
which has not been well studied in previous literature. Attenuation measurement
63
4.2. Measurement Receiver Noise
error caused by the phase noise effect has been studied analytically. Our theoretical
analysis agrees with the measurement results. It is shown that the proposed dual-
channel system can effectively reduce the phase noise effect at frequencies beyond
that a record high DR of 186.7 dB has been achieved at 26 GHz when a microwave
amplifier is inserted before the input port. Our receiver noise floor is 47 dB lower
system and material analysis using non-destructive method [38, 39, 98]. The atten-
that, with a microwave amplifier inserted before the receiver input, attenuators
with attenuation up to 170 dB can be measured with good accuracy using our
tainty is estimated to be 0.046 dB for a 120 dB attenuator and 0.16 dB for a 150
dB attenuator, respectively.
ment. We have proposed a new method to evaluate the leakage errors in attenu-
ation measurement for devices with high attenuation. This is critical for reliable
64
4.2. Measurement Receiver Noise
its noise floor. The noise in our receiver comes from the thermal noise in pre-
amplifier, filters, mixers, IVD, LIA, and cables [71, 72, 75, 233, 234]. At higher
frequency offset from the microwave signal carrier frequency, the noise spectrum is
Nt = kT B, (4.1)
absolute temperature in Kelvin and B is the signal bandwidth in Hz [71, 72]. The
296 K .
The receiver noise also comes from the amplitude modulation (AM) and phase
modulation (PM) noise in the microwave oscillators, which are shifted to the AF
range after mixing. In the frequency range closer to the carrier frequency, the flicker
noise (also called 1/f noise or phase noise) is more significant [80–83, 235–238].
The Leeson-Cutler phase noise model for tuned tank oscillators [81, 82, 236]
where fm is the frequency offset from the carrier, fc is the flicker frequency corner,
F is the noise figure, Ps is the average power dissipated in the resistive part of the
tank, f0 is the oscillation frequency, and QL is the effective quality factor of the
tank with all the loadings in place. The phase noise L(f ) is expressed in dBc/Hz.
The phase noise of a microwave source increases with its carrier frequency. It has
65
4.3. Phase Noise Effect
been shown that phase noise in microwave devices is more meaningful in charac-
terizing the system noise level than the noise figure which is commonly used [232].
4.1. The source, LO and the function generator are synchronized through a 10
MHz reference signal. A band-pass filter with 1 Hz bandwidth inserted before the
LIA will not be effective to reduce the receiver fluctuation which is mainly caused
by the phase noise of microwave sources, especially when the frequency of the
signal is beyond 2 GHz. The phase noise of a microwave signal generator increases
significantly as the carrier frequency increases from 2 GHz to 26.5 GHz and this
Source
AF (5.02 kHz)
DUT LP
Filter
Amp IVD LIA
LO Function Generator
(5.02 kHz)
10 MHz
reference
receiver.
When a DUT is set to 0 dB range (datum position), the IVD ratio is set to a
reference value D1 and the LIA detects an AF signal with a magnitude of A1 . After
the DUT is set to a particular attenuation range, the IVD ratio setting is adjusted
66
4.3. Phase Noise Effect
an IVD ratio setting D2 . The microwave attenuation of the DUT is then calculated
as
D2 A1
A12 = 20 log10 (dB) (4.3)
D1 A2
A LIA requires a good frequency reference signal to recover a sinusoidal signal with
the same frequency when the signal-to-noise ratio is very low [226, 227]. The LIA
where θ = θs − θr .
the input signal amplitude and θ is the phase difference between the input signal
Similarly, a second phase sensitive detector with the reference signal shifted by
90◦ will produce Y1 = A sin(θ). Thus the input signal amplitude can be derived as
q
M = X12 + Y12 = A (4.5)
The LIA can measure a weak signal’s amplitude with better accuracy, rejecting
sufficient to filter out the AC noise. The phase difference between the input and
reference signal must be constant, otherwise X and Y will be AC signals and cut
67
4.3. Phase Noise Effect
In reality, the signal from the microwave source has random phase fluctuations,
which will be translated to AF range via mixing. The phase of the reference signal
that of the microwave sources, and can be assumed to be constant. Thus the phase
difference between the input and the reference signal will be time-varying. This will
GHz, the measurement uncertainty depends mainly on the LIA reading fluctuation,
the frequency domain [81–83,236,237]. The signal of the source can be modeled by
where Â(t) is the signal amplitude with random fluctuation, which is called AM
noise, and φ(t) represents the random phase fluctuation, also called phase noise.
The AM noise is usually much smaller than the phase noise. The phase fluctuation
will generate noise at a frequency offset from the carrier f0 . It has been shown
where the first term 2πf0 t represents an ideal phase and the second term is a
modulation (FM) signal and m << 1. The phase noise is the standard deviation of
the phase fluctuation and can be described by its root-mean-square (RMS) value
68
4.3. Phase Noise Effect
where cos[m sin(2πfm t)] ≈ 1 and sin[m sin(2πfm t)] ≈ m sin(2πfm t) since m << 1.
The model shows an ideal carrier with two sidebands, each having a peak amplitude
of m/2. The sidebands show the phase noise spectra in frequency domain, which
m2
φ2j = E[(m sin(2πfm t))2 ] = (4.9)
2
(4.8) shows that the spectral energy of each phase noise sideband in frequency
domain is m2 /4 and the total spectral energy of phase noise sidebands is m2 /2.
Thus for real signals with good spectral purity, the RMS phase jitter in radians2 is
equal to the phase noise spectral power. The RMS phase angle fluctuation can be
RMS phase fluctuation is due to the phase noise at frequency offset from f1 to f2
The phase noise of the microwave signal can also be viewed as fluctuating in-
69
4.3. Phase Noise Effect
instantaneous frequency is also fluctuating quickly. Thus the detected signal in the
LIA becomes
where θ(t) denotes the random phase shift caused by phase noise of microwave
sources, ∗ denotes convolution operation and h(t) is the impulse response of the
low-pass filter used in the LIA. The RMS magnitude of the input signal measured
p
M (t) = X 2 (t) + Y 2 (t) 6= A (4.12)
The phase shift angle θ(t) and M (t) can be assumed to have a rectangular distri-
bution. The microwave attenuation is derived from the voltage ratio, and it now
becomes
D2 M1
Â12 = 20 log10 (dB) (4.13)
D1 M2
where the measured signal amplitudes M1 and M2 are different from the true values
A1 and A2 respectively due to the phase noise effect. θ(t) is due to the phase noise of
the microwave source and LO source. The maximum value of θ(t) can be estimated
from the sum of the phase noise of the microwave signal source and the LO,
sZ
f2 Z f2
θmax = 2 LSG (f )df + LLO (f )df , (4.14)
f1 f1
where LSG and LLO are the SSB phase noises of the microwave signal generator
and LO respectively, f1 and f2 are the frequency offset range used to estimate the
phase noise effect on the possible phase fluctuations of the AF signal at the input
of LIA. The frequency offset can be from 1 Hz to 100 kHz, which is sufficient to
provide a good estimate since phase noise beyond an offset of 100 kHz is usually
quite small and negligible. θ(t) can assume any value within [−θmax , θmax ]. θmax
70
4.3. Phase Noise Effect
is usually smaller than 1 radian. Since the phase shift θ(t) mainly depends on the
close-in phase noise, the maximum phase shift value is not affected by the low-
pass filtering inside the LIA. The demodulated signal X(t) and Y (t) in the LIA
are actually random quantity. It can be assumed that X 2 (t) and Y 2 (t) have their
2
Xmax = A2
2
Xmin = A2 cos2 θmax (4.15)
and
2
Ymax = A2 sin2 θmax
2
Ymin = 0 (4.16)
Thus the detected magnitude by the LIA will have the following maximum and
minimum values,
p
Mmax = M0 1 + sin2 θmax ,
The phase difference between the reference and input signal is given by
Y
δφ = φs − φr = arctan( )
X
= ψ(t) + E[φs ] − φr (4.18)
where E[φs ] is the mean value of the phase of the input signal, φr is the phase of
reference signal, ψ(t) is a zero-mean random variable which is purely due to the
phase noise of the microwave source and LO source. The phase of the AF reference
signal (φr ) is very stable and can be assumed to be constant. The phase of reference
Y
ψ(t) = arctan( ) (4.19)
X
71
4.3. Phase Noise Effect
Ymax
ψm = arctan( )
Xmin
= arctan(tan θmax ) = θmax (4.20)
Thus, the maximum phase fluctuation can be estimated from the phase noise data.
We have estimated the maximum phase shift caused by the phase noise of mi-
crowave source using (4.14) , based on the phase noise specification at 250–500
MHz, 3–10 GHz, 10–20 GHz and 20–30 GHz. The SSB phase noise of the mi-
crowave signal source and the LO is assumed to be the same as the instrument
specification, as given in Fig. 4.2. To verify our modeling and prediction of the
maximum phase and magnitude fluctuation at the LIA output, we also measured
shown in Fig. 4.1. Measurements have been made for microwave frequencies set at
50 MHz, 100 MHz, 500 MHz, 1 GHz, 2 GHz, 4 GHz, 8 GHz, 12 GHz, 18 GHz and
26 GHz.
Fig. 4.3 shows the maximum phase fluctuation measured by the LIA and the
predicted values by (4.14) using phase noise data shown in Fig. 4.2. The phase noise
data is an average for each band, thus the predicted maximum phase fluctuations
are also average values for each band. The maximum phase fluctuation measured
of the detected signal by LIA are 1.5059 mV and 1.4801 mV respectively. The
72
4.3. Phase Noise Effect
θmax =7.55◦ using (4.14) . This value agrees with the measured value for θmax
which is about 8◦ .
−20
250−500 MHz
3− 10 GHz
10 − 20 GHz
−40
20 − 30 GHz
SSB Phase noise (dBc/Hz)
−60
−80
−100
−120
−140
0 1 2 3 4 5
10 10 10 10 10 10
Frequency offset (Hz)
Figure 4.2: The SSB phase noise L(f) of the microwave sources, based on the
instrument specifications at 250–500 MHz, 3–10 GHz, 10–20 GHz and 20–30 GHz.
73
4.3. Phase Noise Effect
10
9 measured value
predicted value
8
7
Maximum phase fluctuation (deg)
−1
2 3 4
10 10 10
Microwave carrier frequency (MHz)
Figure 4.3: The maximum phase fluctuation measured by the LIA and the predicted
values using the phase noise data shown in Fig. 4.2 (frequency range: 0.05–26 GHz).
74
4.3. Phase Noise Effect
It has been shown that the detected signal by the LIA has a fluctuation in mag-
nitude from Mmin to Mmax due to the phase fluctuation of the AF signal at the
LIA input, which is not phase-locked to the LIA external reference signal in our
single-channel system.
x 0 T 1 D1 G s = M0
x 0 T 2 D2 G s = M0 , (4.21)
where x0 is the input signal level at the test port 1, T1 and T2 are the DUT initial
and final attenuation (in linear scale) respectively, D1 and D2 are the corresponding
IVD ratio used to get the same signal magnitude of M0 at the LIA, and Gs is the
receiver gain (in linear scale). Then the derived incremental attenuation of DUT
is
D2
A = 20 log10
D1
T1
= 20 log10 (dB). (4.22)
T2
Assume the signal magnitude detected by the LIA is Mmax when IVD ratio is set
to D̄1 and Mmin when IVD ratio is set to D̄2 , then we have the following,
D̄2
Ā = 20 log10 ( )
D̄1
!
T1 cos θmax
= 20 log10 ( ) + 20 log10 p . (4.25)
T2 1 + sin2 θmax
75
4.3. Phase Noise Effect
to be uniformly distributed in [−epn , epn ], where epn is the maximum error given
by
epn = A − Ā
p !
1 + sin2 θmax
= 20 log10 . (4.26)
cos θmax
effect is obtained as
epn
upn = √ . (4.27)
3
are measured to be 0.03◦ to 8◦ for microwave signal from 50 MHz to 26.5 GHz.
The upn estimated from the measured maximum phase fluctuation in LIA using our
below 30 dB at frequencies beyond 1 GHz, the phase noise effect is more significant
In our dual-channel receiver, the reference signal of the LIA is from the reference
receiver which shares the same signal source and LO, thus the phase fluctuation
of the AF signal at the LIA input is very small. The maximum phase fluctuation
θmax is measured to be below 0.05◦ , and the phase noise effect on the attenuation
found that the phase noise effect on final signal detection in LIA is similar from
uncertainty in attenuation due to phase noise effect is below 0.00015 dB using our
dual-channel receiver.
76
4.3. Phase Noise Effect
−4
x 10
7
−1
200 400 600 800 1000 1200 1400 1600 1800 2000
Carrier frequency (MHz)
(a)
0.09
Attenuation measurement uncertainty (dB)
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
5 10 15 20 25
Carrier frequency (GHz)
(b)
Figure 4.4: The standard attenuation measurement uncertainty due to phase noise
effect using single-channel system for the frequency range: (a) 0.05–2 GHz, (b)
2–26 GHz.
77
4.4. Noise Floor in a Dual-channel System
the phase noise effect is negligible (upn ≤ 0.0005 dB). For frequencies from 2 to
26.5 GHz, a dual-channel receiver need to be used to ensure phase noise effect is
minimized.
For measurement of attenuation higher than 90 dB, the magnitude of the signals
This smoothing helps to reduce the effect of random noise. The LIA time constant
has been set to 2 seconds. The measurement process for each attenuation value
takes less than 40 seconds to complete. Thus the signal magnitude fluctuation
at LIA output within 120 seconds would be able to indicate the noise effect on
measurement result. We have processed the data measured by the LIA in 120
Let yk (k = 1, 2, ..., 120) denotes the magnitude data detected by LIA in 120
seconds with one sample per second, the smoothed samples ȳi (i = 1, 2, ..., 96) are
obtained by
i+24
1 X
ȳi = yk , i = 1, 2, ..., 96 (4.28)
25 k=i
Fig. 4.5(a) shows the magnitude and phase of the signal detected by LIA with
110 dB attenuation inserted between the two test ports at 26 GHz (the LSA is not
used and the GBA is set to 0 dB). Fig. 4.5(b) shows the smoothed magnitude data
ȳi which has been obtained using (4.28) from raw data measured in 120 seconds.
The pre-amplifier gain is set to 54 dB and IVD ratio is set to 0.96 when the DUT
attenuation is set to 110 dB. It can be seen that the signal phase is quite stable.
78
4.4. Noise Floor in a Dual-channel System
The mean value of the smoothed signal’s magnitude is s = 704.4 µV and the
maximum fluctuation around its mean value is about n1 = 0.4 µV. Therefore, the
s
SN R = 20 log10 ≈ 65 (dB). (4.29)
n1
The equivalent noise floor at receiver input at port 2 (in Fig. 3.2 ) is defined as
and SN R110dB refer to the SNR estimated at LIA output when a 110 dB attenuator
The equivalent noise floor Ne includes the effect of noises from the mixer, the
filters, the pre-amplifier and the LIA. The SNR at the LIA output is estimated
The Ne indicates the minimum discernible signal level of the measurement receiver.
79
4.4. Noise Floor in a Dual-channel System
706
Magnitude (microvolt)
705.5
705
704.5
704
703.5
703
0 20 40 60 80 100 120
time (sec)
63.75
Phase (deg)
63.7
63.65
63.6
63.55
0 20 40 60 80 100 120
time (sec)
(a)
704.9
704.8
704.7
704.6
Magnitude (microvolt)
704.5
704.4
704.3
704.2
704.1
704
0 10 20 30 40 50 60 70 80 90 100
Sample
(b)
Figure 4.5: (a): The magnitude and phase of the signal detected by LIA with 110
data collected in 120 seconds (one sample per second), (b): smoothed magnitude
data obtained by (4.28) using the raw magnitude data shown in (a).
80
4.4. Noise Floor in a Dual-channel System
maximum signal power level at the receiver input and the receiver noise floor. In
a VNA, the receiver input power level is usually specified to be below a certain
to ensure the nonlinearity error is around 0.001 dB [140, 164, 229]. For DUT atten-
uation below 30 dB, we have placed a 30 dB LSA before the DUT to ensure the
For DUT attenuation higher than 30 dB, a GBA is inserted into the circuit to
serve as a reference and the attenuation of the DUT is compared with that of the
GBA. Thus the signal level at mixer input in this indirect measurement process is
also below the 0.001 dB compression point. Therefore, the combination of direct
system to be extended beyond the mixer compression point. The upper limit of
DRs = P1 − Ne , (4.32)
The DRs gives the maximum range of attenuation that can be measured by the
receiver. The P1 of the system is around 14 dBm at 26 GHz. Thus the measurement
81
4.5. Attenuation Measurement Uncertainty due to Noise
to Noise
presence of noise in the detected signal by the LIA, assuming the worst case when
the noise magnitude after smoothing has the maximum value, the model for the
x0 T2 D̄2 Gs = M0 + Nf , (4.34)
where Nf is the noise floor level in the measurement made by the LIA. When DUT
attenuation is set to 0 dB and GBA is set to 60 dB or 110 dB, the noise effect on
x0 T1 D̄1 Gs = M0 , (4.35)
D̄2
Ā = 20 log10 ( )
D̄1
T1 M0 + N f
= 20 log10 ( ) + 20 log10 . (4.36)
T2 M0
The error in attenuation measurement caused by the noise can be assumed to have
en = Ā − A
Nf
= 20 log10 1 + . (4.37)
M0
82
4.6. Measurement of Attenuation up to 170 dB
factor k = 1) is obtained as
en
un = √ . (4.38)
3
and un can be estimated from the SNR at LIA output by
20 1
un = √ log10 1 + SN R/20 (dB), (4.39)
3 10
where
and Adut is the attenuation of DUT in dB. The SNR for measurement of 110 dB
The receiver noise will cause a dominating error in the measurement of large
noise (coverage factor k=1) at 26 GHz is 0.0028 dB and 0.74 dB for attenuation
before the main mixer input to improve the measurement accuracy. The microwave
amplifier has a noise figure of about 13 dB at 26 GHz. The receiver noise floor is
estimated as
83
4.6. Measurement of Attenuation up to 170 dB
where P1′ is the power at port 1, Adut,170dB is the attenuation of a 170 dB attenuator
and SN R170dB refer to the SNR estimated at LIA output when a 170 dB attenuator
is inserted at test port. Tuner is not used for measurement of 110-170 dB to reduce
leakage error. P1′ is measured to be 14.7 dBm at 26 GHz. The SNR for measurement
The receiver noise floor without using the microwave amplifier has been estimated
to be −161 dBm. We can see that the microwave amplifier inserted before port
2 helps to reduce the receiver noise floor by about 11 dB. The LIA bandwidth
is 0.06 Hz. Therefore the theoretical thermal noise limit in the receiver at room
temperature (296 K) is about −186 dBm. The Ne′ approaches this limit. The DR
of the system using the microwave amplifier is about 186.7 dB at 26 GHz. This
receiver has achieved very sensitive detection using LIA with a coherent frequency
reference.
A 110 dB GBA is used for measurement of DUT attenuation from 120 to 170 dB,
where δA,3 is the measured difference between the incremental attenuation of the
DUT and the 110 dB GBA. AGB,110dB are the calibrated incremental attenuation
of the GBA from its datum position to the 110 dB setting. If a DUT has an
attenuation setting of 110 dB, it can be used as a GBA and inserted between the
84
4.6. Measurement of Attenuation up to 170 dB
0-170 dB synthesized step attenuator. The two step attenuators are called DU T1
performed by indirect measurement using the 110 dB setting as the GBA (DU T1
Table 4.1 shows the settings of two step attenuators and the mean of 10 mea-
surements for each attenuation step. The repeatability and estimated expanded
measurement uncertainties (coverage factor k=2) are also given in Table 4.1. The
The standard uncertainty due to receiver noise is 0.27 dB and 0.75 dB for mea-
surement of 160 dB and 170 dB attenuation, respectively. Table 4.1 shows that
the deviations of the mean measured attenuation from their nominal values are less
than 1 dB.
120-170 dB range (formed by connecting two 0-90 dB step attenuators which are
85
4.6. Measurement of Attenuation up to 170 dB
We have conducted another experiment to check the noise and leakage effect on
170 dB were measured with reference to the 110 dB setting of the synthesized step
attenuator (DU T1 set to 90 dB and DU T2 set to 20 dB) and the results are listed in
at a much higher signal level. The expanded uncertainties of A1 and A2 are listed
The differences between these measurements (A1 − A2 , listed in Table 4.3) will
indicate the noise and leakage effect on the high attenuation measurement. Table
4.3 shows that the difference (A1 − A2 ) for each incremental attenuation value
and A2 , which means that they agree with each other within their corresponding
measurement uncertainties. Thus we can conclude that the error due to leakage for
These measurement results show that our estimation of the receiver DR is valid.
Our measurement receiver has achieved very low noise floor and wide dynamic
86
4.7. Discussions
with reference to the 110 dB setting at 26 GHz (DU T1 set to 90 dB, DU T2 varying
from 20 dB to 80 dB).
Mean
DUT1 DUT2
Measured Expanded
Nominal Nominal Incremental Uncertainty
Attenuation Attenuation Attenuation, of A1 (dB)
Step (dB) Step (dB) A1 (dB)
90 20-30 10.130 0.043 0 20-30 10.096 0.00
90 20-40 20.221 0.051 0 20-40 20.178 0.01
90 20-50 30.188 0.076 0 20-50 30.155 0.01
90 20-60 40.363 0.16 0 20-60 40.263 0.01
90 20-70 50.179 0.62 0 20-70 50.256 0.01
90 20-80 60.364 1.6 0 20-80 60.185 0.01
ments A1 .
Mean
DUT1 DUT2
Measured Expanded
Nominal Nominal A1 − A2
Incremental Uncertainty
Attenuation Attenuation (dB)
Attenuation, of A2 (dB)
Step (dB) Step (dB)
A2 (dB)
20-30 10.130 0.043 0 20-30 10.096 0.0072 0.034
20-40 20.221 0.051 0 20-40 20.178 0.012 0.043
20-50 30.188 0.076 0 20-50 30.155 0.012 0.033
20-60 40.363 0.16 0 20-60 40.263 0.012 0.10
20-70 50.179 0.62 0 20-70 50.256 0.015 -0.077
20-80 60.364 1.6 0 20-80 60.185 0.015 0.18
4.7 Discussions
A VNA is a powerful microwave instrument, which has been used in many national
provide a reference IF signal for VNA calibration and synchronous detection. The
reference IF signal is used to phase-lock the LO drives in the vector receivers which
The phase-lock loop used in a VNA has some residual random phase error, which
87
4.7. Discussions
and the LO in the vector receiver. Such random deviations behave like a phase noise
effect which increases the noise floor in the synchronous detection. A VNA has a
receiver noise floor of about −125 dBm and a typical attenuation measurement DR
of 125 dB at 26 GHz (with 10 Hz IF bandwidth) [223, 224]. The noise floor of the
our receiver noise floor to −172 dBm at 26 GHz using a LIA-based dual channel
design. A LIA can extract a weak AC signal with known carrier frequency from ex-
tremely noisy environment [227, 239, 240]. The background noise can be effectively
In our proposed design, the main channel AF signal is detected by a LIA. The
LIA employs the coherent frequency reference signal at 5.02 kHz provided by the
reference channel receiver to detect the main channel AF signal with low SNR. The
deviations between the instantaneous frequency of the main channel and reference
channel AF signal are negligible since they are generated from the same microwave
source and LO. This means that the phase noise effect in the receiver is essentially
removed, which ensures that the synchronous detection by LIA has very good
88
4.8. Summary
4.8 Summary
We have studied the thermal noise and phase noise effect in our single-channel and
phase noise effect play a key role in limiting the attenuation measurement DR of
can significantly suppress the phase noise effect and improve the measurement DR.
and analysed. The system has been designed carefully to reduce the leakage to
a level which is negligible compared to the receiver noise. The DR of the dual-
receiver noise floor is estimated to be −172 dBm and the measurement DR is about
89
Chapter 5
Measurement
5.1 Introduction
have been developed [55, 60–62]. Measurement uncertainties of 0.1◦ to 0.3◦ have
been achieved for devices with low attenuation below 10 GHz [61]. The phase-
shift for a device with attenuations up to 140 dB has been measured with an
this dual-channel system only works at 0.5–100 MHz. High accuracy phase-shift
measurement system for high attenuation devices with uncertainty below 0.4◦ at
0.1–18 GHz has not been reported in the literature. Such a measurement standard
panded uncertainty of 0.2◦ –7◦ at 18 GHz for a device with attenuation less than 80
90
5.1. Introduction
dB [133]. The phase-shift measurement uncertainty using a VNA could reach 19◦
at 18 GHz if the device attenuation is 100 dB. The significant increase in trans-
mission phase measurement uncertainty is mainly due to the decreased SNR as the
which has measurement uncertainties of 0.09◦ –0.32◦ at 18 GHz for a coaxial device
of our system for a 100 dB attenuator at 18 GHz is ∼106 times smaller than that
of a VNA [133]. In this chapter, the system design and measurement uncertainty
The leakage error in phase-shift measurement has not been well studied in the
literature. We have proposed a new method to evaluate the leakage error in phase-
shift measurement for devices with high attenuation. This is critical for reliable
still not available. In this chapter, we have investigated the mismatch uncertainty
A beadless coaxial air line has been used as a phase-shift reference standard
to verify the system performance. The theoretical insertion phase-shift of the air
line is calculated with consideration of the line’s insertion loss and compared to
the measured values given by the measurement system. The calculated phase-
shift value agrees well with the measured values within the estimated measurement
uncertainties.
91
5.2. A Precision Phase-Shift Measurement System at 0.05–18 GHz
There are two types of phase-shift: insertion phase-shift defined for a two-port
network with fixed attenuation and phase-shift, and differential phase-shift defined
MHz to 18 GHz is shown in Fig. 5.1. The LIA uses the frequency reference signal
provided by the reference channel to achieve sensitive detection with phase resolu-
tion of 0.01 degrees. The dual-channel receiver is coherent and can make accurate
RF Source
(0.05 − 18 GHz)
Gauge Block
Attenuator DUT
(inserted for indirect
phase-shift Main
Insertion Point
measurement)
Mixer
AF (5.02 kHz)
RF Amp
RF
Tuner Tuner
or Pad or Pad
LO
Level Set Attenuator Isolator Isolator
(inserted for direct
Attenuator phase-shift
measurement) Port 1 Port 2 Low-pass Filter
Attenuator
Ref
RF
Mixer Pre-Amp
LO
Band-pass Filter
Multiple Isolators in 11 bands
10 MHz IVD
Reference LO Source
(0.05 − 18 GHz)
AF (5.02 kHz)
Low-pass Pre-Amp Band-pass Lock-in Amp
Filter Filter Frequency Reference
Before the DUT is inserted, the phase of AF signal detected by the LIA is
92
5.2. A Precision Phase-Shift Measurement System at 0.05–18 GHz
θ1 . After the insertion of a DUT, the IVD ratio are adjusted till the LIA detects
a signal with magnitude similar to that before DUT insertion (voltage difference
< 1 µV ), and the phase of the AF signal detected by the LIA is θ2 . Thus the
obtained from
where θi′ and θf′ are the phases measured by the LIA when the DUT is at its initial
When the signal frequency is above 800 MHz, a tuner is placed before and after
the insertion point to reduce the reflection from the source and load side. Below
800 MHz, a matching pad is used instead of a tuner. The tuner or the pad helps
There is an internal leakage path going through the reference mixer, isolators,
directional coupler and enters the signal mixer RF port (from LO to RF port).
Multiple isolators and a directional coupler are used in the leakage path between
before the reference mixer RF input to reduce the internal leakage. There is also
an external leakage due to radiation from tuners and isolators. Thus the tuner and
the isolator attached to it are put inside two shield boxes to reduce the radiation
leakage.
The system can accurately measure the phase-shift of a DUT with attenuation
93
5.3. Measurement Uncertainty Analysis
inserted before the input of the main mixer. The RF amplifier is used to improve
below 40 dB, a 30 dB LSA is inserted before the DUT. For measurement of a DUT
with attenuation higher than 40 dB, the LSA (30 dB) is replaced by a GBA which
used as the GBA for indirect measurement. The GBA is first set to 40 dB when
DUT setting is 0 dB, and then changed to 0 dB when DUT is set to a value higher
than 40 dB. The difference between the phase-shift of the DUT and the GBA is
where θGBA is the GBA’s phase-shift. Thus the phase-shift of the DUT is derived
as
For a DUT with attenuation from 90 to 140 dB, a 0–90 dB step attenuator is used
The IVD is used to adjust the gain of the measurement receiver for various atten-
uation range of the DUT, so that the signal level at the LIA input before DUT
insertion is similar to that after DUT insertion. The maximum phase error caused
94
5.3. Measurement Uncertainty Analysis
by the change of IVD ratio is about 0.02◦ . The standard uncertainty due to the
√
IVD phase error is estimated to be uI = 0.02/ 3 = 0.012◦ assuming a rectangular
distribution.
The detected signal by the LIA has a RMS magnitude around 1 mV and it has a
maximum fluctuation of 0.01◦ in its phase reading when the DUT attenuation is
less than 110 dB at 2 GHz. The maximum phase fluctuation for DUT with low
The fluctuation increases when the noise effect is more significant at higher
attenuation levels. 600 phase readings are taken for a DUT with attenuation above
100 dB (at a rate of 10 samples per second) to get an averaged estimate of the phase-
shift value. The averaging reduces the random noise error in phase measurement.
As DUT attenuation increases to 140 dB, the maximum deviation of the phase
reading θx (after averaging) from its mean value (θ0 ) was measured to be ef = 0.15◦
It is assumed that the probability for the value of θx to lie within [θ0 −ef , θ0 +ef ]
is equal to one and the probability that θx lies outside this interval is essentially
zero. Since there is no specific knowledge about the possible values of θx within the
interval, we can only assume that it is equally probable for θx to lie anywhere within
the phase reading θx . In [130], it is also assumed that the receiver fluctuation error
95
5.3. Measurement Uncertainty Analysis
as
√
uf = ef / 3. (5.5)
The detected phase of the main channel has a small drift error due to the phase
noise of the two microwave sources. The dual-channel heterodyne receiver uses
a phase reference signal generated from the same RF and LO sources which are
also used by the main channel, thus the phase noise effect on the detected phase
difference between two channels is greatly reduced. The maximum phase drift at 2
rectangular distribution.
coherent single channel system, the maximum phase drift at 2 GHz will increase to
For DUT with attenuation in the range from 40 dB to 140 dB, its phase-shift is de-
the GBAs have been calibrated previously with certain measurement uncertainties,
which should be included in the uncertainty for the DUT phase-shift measurement.
96
5.3. Measurement Uncertainty Analysis
Let
and
Thus
value when all of the vectors have the same directions, i.e.,
lim |Ep | = |ΓG s11b | + |ΓG s11e | + |ΓL s22b | + |ΓL s22e | +
|ΓG ΓL |(|s11b s22b | + |s11e s22e | + |s12b s21b | + |s12e s21e |) (5.10)
maximum and minimum of ǫd will occur when the |Ep | is at its maximum and
97
5.3. Measurement Uncertainty Analysis
Ed Ep
lim İd
– lim İd
1
ment.
written as
|ΓG ΓL |(|s11b s22b | + |s11e s22e | + |s12b s21b | + |s12e s21e |) (5.11)
where ΓG and ΓL are the test port reflection coefficients, s11b , s11e , s22b , s22e , s12b ,
s12e , s21b and s21e are the reflection and transmission coefficients of DUT before
mismatch is
lim ǫd
um = √ (5.12)
3
98
5.3. Measurement Uncertainty Analysis
The approximation in the above equation is made since the reflection coefficients
of the test port and DUT all have small magnitude values, which are much less
than 1. Let
and
Thus
For arbitrary phases of the coefficients, the magnitude of Ep′ is at its maximum
value when all of the vectors have the same directions, i.e.,
lim |Ep′ | = |ΓG s11 | + |ΓL s22 | + |ΓG ΓL |(|s11 s22 | + |s12 s21 | + 1) (5.17)
Similarly, the maximum ǫI will occur when |Ep′ | is at its maximum and the vector
Ep′ is perpendicular to Ed′ , as shown in Fig. 5.3. Thus the error limit of the phase
sin(lim ǫI ) = |ΓG s11 | + |ΓL s22 | + |ΓG ΓL |(|s11 s22 | + |s12 s21 | + 1) (5.18)
uniformly distributed in [− lim ǫI , lim ǫI ] and the standard uncertainty due to mis-
99
5.3. Measurement Uncertainty Analysis
Eǯd Eǯp
lim İI
– lim İI
1
nector is obtained using (3.16). The uncertainties due to IVD phase error, receiver
fluctuation, phase drift, mismatch error, GBA uncertainty and measurement re-
dard uncertainty:
sX
us = ( u2i ) (5.20)
i
Us = kus (5.21)
where k is the coverage factor and k = 2. Table 5.1 gives the phase-shift measure-
100
5.4. Phase-shift Measurement of a 0–140 dB Step Attenuator
degrees).
Combined
0.044 0.071 0.091 0.087 0.093 0.16
uncertainty
Expanded
0.088 0.14 0.18 0.17 0.19 0.32
uncertainty (k=2)
5.4 Phase-shift
Measurement
of a 0–140
dB Step
Attenuator
5.4.1 Measurement Results
Two 0-90 dB coaxial step
attenuators
with
PC 3.5
mm connectors
are connected
in series to form a 0–140 dB synthesized step attenuator. The two step attenuators
measurement results and their expanded uncertainties (k=2) are listed in Table
101
5.4. Phase-shift Measurement of a 0–140 dB Step Attenuator
port 1 and DU T2 . The results, the uncertainties and deviation from the calibration
value are shown in Table 5.3. When DU T2 is set to 50 dB, the total attenuation
between the test ports is 140 dB. The calibration values for the differential phase-
deviations are smaller than the sum of the corresponding uncertainty for each
attenuation step listed in Table 5.3 and Table 5.4. Thus the leakage effect in
102
5.5. Microwave Phase-Shift Reference Standard
50 dB at 18 GHz
DUT2 Nominal Calibrated Expanded
Attenuation differential Measurement
phase-shift Uncertainty
(dB) (Degree) (Degree)
0-10 -21.05 0.09
0-20 -8.28 0.09
0-30 -29.21 0.09
0-40 -21.82 0.09
0-50 -9.48 0.14
Beadless coaxial air line has been used as precision impedance reference standard
[65, 173]. Such air lines are nearly reflectionless and have very low insertion loss.
They can be used as phase-shift reference standard to verify our dual-channel phase-
A coaxial line can be characterised using a circuit model involving series resis-
unit length of the line. If the air line is assumed to be lossless, R = 0 and G = 0.
µ loge (b/a)
L0 = (5.22)
2π
103
5.5. Microwave Phase-Shift Reference Standard
2πǫ
C0 = (5.23)
loge (b/a)
where µ is the permeability of the line’s dielectric, ǫ is the permittivity of the line’s
dielectric, a is the radius of line’s center conductor and b is the radius of line’s
outer conductor. Fig. 5.4 shows the cross-section of an air line. For a 7 mm air
µ, ε
a
b
θ = 2nπ − βl (5.24)
where n is an integer so that θ is within (−π, π), l is the measured length of the
It has been shown in [63,64,69] that air line loss need to be considered to estimate
104
5.5. Microwave Phase-Shift Reference Standard
k 2 a2 F 0
R = 2ωL0 d0 1 − (5.27)
2
k 2 a2 F 0
L = L0 1 + 2d0 1 − (5.28)
2
G = ωC0 d0 k 2 a2 F0 (5.29)
C = C0 (1 + d0 k 2 a2 F0 ) (5.30)
where
2π
k= (5.31)
λ
and
(b2 /a2 ) − 1 (b/a) loge (b/a) a + b
F0 = − − (5.32)
2 loge (b/a) (b/a) + 1 2a
δs (1 + (b/a))
d0 = (5.33)
4b loge (b/a)
where ρ is the resistivity of the line’s conductors (Ωm). The attenuation constant
α can be calculated from the transmission coefficient magnitude |s21 | of the air line,
ln|s21 |
α=− (5.35)
l
ment system [241]. Based on the first-order approximation of the field equations
105
5.5. Microwave Phase-Shift Reference Standard
for a slightly lossy coaxial line (derived by Daywitt [174]), the resistivity ρ of the
air lines are calculated using (5.35) and (5.36) from their measured attenuation
values at 0.05–18 GHz, and it is shown that the frequency dependence in the cal-
culated resistivity values is negligible in this frequency range. Thus the resistivity
Once ρ is estimated using (5.36), the R, L, G and C of a lossy air line can
be calculated using (5.27)-(5.34). Then βL can be calculated from (5.26) and the
θL = 2nπ − βL l (5.37)
Daywitt [175] has derived the exact principal mode field for a lossy coaxial air
line. In [175], the propagation constant (γ = α + jβL ) of a lossy air line derived by
the exact field solution has been numerically compared to that derived by the first-
the error in the line loss per unit length is around 6.2×10−7 dB/cm and the error in
the phase-shift per unit length is below 3.0×10−9 degrees/cm at 0.05–18 GHz [175].
For an air line (with 7 mm connectors) having a length of ∼10 cm, its attenuation
is around 0.04 dB at 2 GHz and 0.15 dB at 18 GHz. The error in its calculated
loss due to the first-order approximations is only 6.2 × 10−6 dB, which is negligible.
The expanded uncertainty in the calculated phase-shift of such an air line is mainly
due to the line length measurement uncertainty, which is around 0.01◦ at 2 GHz
and 0.13◦ at 18 GHz, respectively. The error in its calculated phase-shift due to the
106
5.6. System Verification using a Beadless Air Line
first-order approximations is below 3.0 × 10−8 degrees, which is much smaller than
can conclude that the error in the calculated line phase-shift due to the first-order
A 50 ohm beadless air line fitted with precision 7 mm connectors is used as a phase-
shift reference standard. The length of the air line is measured to be 99.9017 mm
with expanded uncertainty of 0.0060 mm. This air line has been used to verify the
performance of the phase-shift measurement system. The air line attenuation has
The calculated phase-shift of the air line at 2 GHz and 18 GHz using (5.37) and the
associated uncertainty due to the line length measurement uncertainty are listed
in Table 5.5.
and cable movement errors are more significant and need to be included in the un-
certainty budget. The standard uncertainty due to connector and cable movement
errors is estimated to be 0.012◦ at 2 GHz and 0.087◦ at 18 GHz. Table 5.5 also
shows the measured phase-shift of the air line at 2 GHz and 18 GHz and their
the calculated phase-shift values of the airline are smaller than the corresponding
measurement uncertainties. The comparisons in Table 5.5 show that the measured
phase-shift values agree with the calculated values within the expanded measure-
ment uncertainties.
107
5.7. Summary
Table 5.5: Measured and calculated phase-shift of a beadless air line with 7 mm
5.7 Summary
18 GHz has been proposed. The measurement uncertainty of the proposed system
GHz for an attenuator with attenuation of 0–140 dB. The phase-shift measurement
uncertainties at 18 GHz are about 106 times smaller than that of VNA for a 100
dB attenuator.
A coaxial beadless air line has been used as a phase-shift standard to verify the
insertion loss agrees with the measured values within the measurement uncertain-
ties. Beadless air lines and variable attenuators can be measured by our proposed
phase-shift measurement system with high accuracy and provide traceability for
posed in this chapter can also be used for precision phase measurement in other
applications.
108
Chapter 6
Millimeter-wave Attenuation
Precision Measurement
6.1 Introduction
was proposed in [211]. A calibrated IF attenuator was used as the reference at-
tenuation standard. The systematic error and random error are ±0.3 dB and
at 75–110 GHz was proposed in [210], which used a video detector to reduce the
system cost. A calibrated RVA was used as the reference attenuation standard at
ing of 1.7 minutes to reduce the noise floor. An attenuation measurement system
420 GHz. This system did not use any attenuation standard to provide traceable
109
6.1. Introduction
and a reference channel receiver is used to provide a reference IF signal for VNA
lock the LO in two vector receivers, which down-convert the main and reference IF
signals to a lower frequency for coherent detection [168]. A VNA has been used to
The phase-lock loop used in a VNA has some residual random phase error, which
and the LO in the vector receivers. Such random deviations behave like a phase
noise effect which increases the noise floor in the synchronous detection. A VNA has
a receiver noise floor of about −125 dBm and a typical attenuation measurement
cations [103, 104]. New techniques need to be developed for attenuation measure-
ment at higher frequencies (beyond 75 GHz) with improved DR beyond 150 dB.
and radiometers.
110
6.1. Introduction
ment system at 75–110 GHz, 140–220 GHz and 330–500 GHz, respectively. These
security imaging [37, 96, 105–108, 189, 194]. Our proposed dual-channel AF substi-
IVD [25] which is a voltage-ratio reference standard at 5.02 kHz. In our proposed
design, the main channel AF signal is detected by a LIA. The LIA employs a co-
herent frequency reference signal at 5.02 kHz (provided by the reference channel
receiver) to detect the main channel AF signal with low SNR. The deviations be-
tween the instantaneous frequency of the main channel and reference channel AF
signal are negligible since they are generated from the same microwave source and
LO. The phase noise effect in the receiver is essentially removed, which ensures
that the synchronous detection by the LIA has much better sensitivity.
The system has been verified using several precision RVAs. The DR of these
novel measurement systems is much better than that of a VNA. The measurement
dB at 220 GHz and 0.078 dB at 480 GHz, respectively, which are much smaller
than that of a VNA. The DR of the proposed system is 58 dB higher than that of
It has been very difficult to estimate the leakage error in high attenuation mea-
surement since the leakage signal can not be separated from the noise [117]. We
have proposed a new method to evaluate the leakage errors in attenuation mea-
surement for devices with high attenuation. This is critical for reliable estimation
The frequency bands at 75–110 GHz (WR-10 band), 140–220 GHz (WR-05
band) and 330–500 GHz (WR-2.2 band) have been selected to study the measure-
111
6.2. Attenuation Measurement System at 75–110 GHz
ment accuracy and receiver noise effect at increasing frequencies within 75–500
GHz. Measurement uncertainty analysis performed at the selected bands will show
the mismatch error, leakage error and phase-noise induced instability effect. The
ment system at other frequency bands within 50–500 GHz. Although the phase
noise in the source increases significantly from 75 GHz to 500 GHz, the noise floor
in our proposed receivers at WR-10, WR-05 and WR-2.2 band remains at around
−169 dBm. This shows that our design can suppress the phase noise effect in the
a similar noise floor level. As the frequency increases from 75 GHz to 500 GHz,
the source output power level decreases which causes the system DR to decrease
GHz
(75–110 GHz) is shown in Fig. 6.1. A microwave signal (12.5−18.33 GHz) drives
a multiplier (×6) to generate a signal in 75–110 GHz. The signal level at test port
1 is about 5 dBm at 110 GHz. Two subharmonic mixers are used to down-convert
signal at 3.2 GHz, respectively. The frequency of LO source 1 is set to fLO which
satisfies (fRF − fLO ) × 6 = 3.2 GHz, where fRF is the frequency of the RF source.
112
6.2. Attenuation Measurement System at 75–110 GHz
Ref Subharmonic LO X3
X3 Splitter
Mixer
Attenuator LO LP Filter
RF
Pre-Amp
LO
Ref RF Mixer
LO Source 2
Isolator Isolator Isolator Isolator (3.20000502 GHz) IVD
10 MHz
Reference
AF (5.02 kHz)
Reference (5.02 kHz) Lock-in
LP Filter Pre-Amp
Amp
The main and reference IF signals go through BP filters which are used to
suppress the harmonic signals. The IF signals are further down-converted by two
RF mixers to AF signals at 5.02 kHz. The cut-off frequency of the LP filter is 100
kHz. The reference AF signal provides a frequency reference to the LIA to ensure
ratio reference standard with an accuracy of a few parts in 106 . The voltage ratio
of the millimeter-wave signal can be derived from the IVD voltage ratio at 5.02
kHz.
The phase noise of the millimeter-wave source will be translated to the IF signal,
which will cause significant fluctuation in LIA reading if the non-coherent frequency
reference signal is used for the LIA. To stabilize the LIA reading for accurate
an AF signal, which is used as the external frequency reference of the LIA. The LIA
reading in our measurement system is very stable and the standard uncertainty due
113
6.2. Attenuation Measurement System at 75–110 GHz
RF isolators are placed between the main and reference RF mixers to reduce the
A waveguide level-set attenuator (LSA) is used to control the signal level at the
receiver input and avoid compression error in the main subharmonic mixer and the
RF mixer. The isolator in the source side and receiver side are used to reduce the
reflection from the multiplier and the main subharmonic mixer, respectively. Two
matching pads (waveguide attenuators) are used to reduce the mismatch error in
For a DUT with attenuation higher than 50 dB, an indirect measurement method
described in 3.2 is used. A variable 0–50 dB GBA is inserted between the Isolator
A and the Matching Pad A (shown in Fig. 6.1), and the LSA is removed. The
IVD error
A seven decade IVD is used as the voltage ratio reference standard. The maximum
error in attenuation due to the IVD ratio error is obtained using (3.5). The error
114
6.2. Attenuation Measurement System at 75–110 GHz
eIV D
uIV D = √ . (6.1)
3
The uIV D is estimated to be 0.0004 dB, 0.0008 dB and 0.002 dB for measurement
For a DUT beyond 50 dB, indirect measurement is made by comparing the DUT
and 90–100 dB is estimated to be 0.0004 dB, 0.0008 dB and 0.002 dB, respectively.
Receiver nonlinearity
The nonlinearity of the subharmonic mixer, the RF mixer and the pre-amplifier
systems [140,229]. The signal level at the subharmonic mixer millimeter-wave input
error. For measurement of low attenuation, a waveguide LSA is used to ensure the
RVA can be used as a reference waveguide attenuation standard [155, 158]. A RVA
with a 10 dB attenuator connected at each port is used as the DUT to provide the
0–10 dB attenuation with very good repeatability. When the DUT is set to 0 dB,
the LSA is adjusted to reduce the initial signal level at subharmonic mixer input
from −35 dBm to −95 dBm in 10 dB step. The DUT is measured eight times at
each signal level and the measurements are listed in Table 6.1. The measurement
The measurement values obtained at initial mixer input level from −35 dBm
115
6.2. Attenuation Measurement System at 75–110 GHz
Table 6.1: Measurement of the nonlinearity and leakage error of the attenuation
to −75 dBm are averaged to obtain the reference value (9.9658 dB), and their
maximum deviation from the reference value is eLIN = 0.0013 dB. The nonlinearity
eLIN
uLIN = √ = 0.0008 dB. (6.2)
3
The increased deviations from the reference value at mixer input level of −85 dBm
and −95 dBm are due to the millimeter-wave leakage error and receiver noise.
Mismatch uncertainty
The standard mismatch uncertainty for a variable attenuator can be estimated from
for measurement of attenuation from 0.01 dB to 40 dB. For a DUT below 50 dB,
two 10 dB matching pads are used. Two 3 dB matching pads are used for DUT
beyond 50 dB. The upper limit of uM for the RVA at 110 GHz is 0.0076 dB for the
attenuation range below 50 dB and 0.010 dB for the attenuation range from 50 dB
116
6.2. Attenuation Measurement System at 75–110 GHz
The detected signal by the LIA has small fluctuation in amplitude during mea-
surement. Such fluctuation will lead to some uncertainty in the attenuation mea-
surement since the IVD ratio setting is determined by the LIA voltage reading.
The dual-channel system uses a coherent frequency reference for the LIA, which
significantly reduces the fluctuation due to phase noise effect [81,236,243,244]. For
noise effect in the mixers, pre-amplifier and the LIA. For attenuation measurement
in 80−100 dB, the LIA receiver bandwidth is set to 0.06 Hz to reduce the random
noise effect.
The maximum error in attenuation due to the fluctuation and noise is obtained
from (3.9). The error due to fluctuation and noise is assumed to have a rectangular
Leakage error
surement of high attenuation above 80 dB. The leakage error can be determined
by a measurement experiment. The linearity test result has indicated that there
GHz. Thus a 0–50 dB RVA (referred to as RVA1 ) is first measured by the system
to get the reference values for the 0–50 dB attenuation setting in 10 dB step. The
117
6.2. Attenuation Measurement System at 75–110 GHz
!"#$%#&'(#)*+,-.*/0
RVA1 is measured again with another RVA (set to 50 dB attenuation step, referred
)*+#,..
to as RVA2 ) cascade-connected with it. In this case, the 4,5*-
)1- 2,-+# .,3 6
total inserted nominal
7% 778%9:$ 7787%;%;: <%8%;$= <%8%7:>
;% 778%?@ <%8%%=% !"!!#$
attenuation value between the test ports becomes 7787%;%;:
60, 70, 80, 90 and 100 dB as the
>% 778%??7 7787%;%;: <%8%%;? !"!!%&
=% 778779 7787%;%;: %8%7:% !"!!'(
RVA1 changes to 10,:% 20, 30,778%?=
40 and 50 dB, respectively.
7787%;%;: The<%8%%@%
deviations between
!"!!)(
$% 778%@@ 7787%;%;: <%8%7=% !"!!'%
these two measurement
9% experiments
7787= are shown in7787%;%;:
Table 6.2. %8%>@% %8%;7?
@% 778%@ 7787%;%;: <%8%;;% <%8%7;9
?% 7%8?@ 7787%;%;: <%87;;% <%8%9%:
The deviation of eL = 0.067 dB when the total nominal incremental attenua-
tion is 50–100 dB is due to the leakage error, thus the standard uncertainty due
uncertainty due to leakage for measurement of 90 dB and 80 dB attenuator are =?8$9@ 7%8
=?8$9@ ;%8
estimated to be 0.011 dB and 0.0031 dB respectively at 110 GHz. When the total =?8$9@ >%8
=?8$9@ =%8
=?8$9@ :%8
nominal incremental attenuation is 60 and 70 dB, the deviations in Table 6.2 are
quite small and should be due to the receiver noise and mismatch uncertainties.
Therefore this experiment shows that the leakage error is negligible for measure-
Table 6.2: Measurement of the leakage error in the attenuation measurement sys-
118
6.2. Attenuation Measurement System at 75–110 GHz
comparing the attenuation of the device to that of a 50 dB GBA. The GBA has
been calibrated previously. Thus the GBA calibration standard uncertainty (uGBA )
DUT repeatability
A RVA has an inherent repeatability error due to the dial resettability which in-
creases as the attenuation becomes higher [155, 158]. Two cascade-connected pre-
cision RVAs (referred to as RVA1 and RVA2 ) with 0–50 dB attenuation in each
RVA are used to form a 0–100 dB synthesized variable attenuator (SVA), which
measurements are made at each attenuation step. The standard uncertainty due
The standard uncertainty components include the IVD error, nonlinearity, mis-
match uncertainty, receiver fluctuation and noise, leakage, GBA calibration uncer-
tainty and DUT repeatability, which are listed in Table 6.3. These uncertainty
119
6.2. Attenuation Measurement System at 75–110 GHz
U = k uc (6.5)
where k is the coverage factor and k = 2 [135]. The combined uncertainty (uc ) and
the expanded uncertainty (U ) for measurement of a 0–100 dB SVA at 110 GHz are
Table 6.3: Uncertainty estimation for measurement of a 0–100 dB SVA at 110 GHz
The SVA is measured at 10–100 dB in 10 dB step at 110 GHz to verify the perfor-
mance of the system. Eight measurements are made at each attenuation step and
the results are given in Table 6.4. The nominal attenuation values from 60 dB to
nominal attenuation value, hence the accuracy of the SVA is about 1% in 10–50
120
6.2. Attenuation Measurement System at 75–110 GHz
!"#$%#&'(#)*+,-.*/0
dB and 2% in 50–100 dB. Table 6.4 shows that, at each )*+#,..
attenuation step, the
)1- 2,-+# .,3 4,5*- 6
7% measured
deviation of the mean 778%9:$ value from the nominal
7787%;%;:
value<%8%;$= <%8%7:>
is less than 1% of
;% 778%?@ 7787%;%;: <%8%%=% !"!!#$
>%
the nominal incremental 778%??7
attenuation. A calibrated7787%;%;: serve as a!"!!%&
RVA can <%8%%;? reference
=% 778779 7787%;%;: %8%7:% !"!!'(
:% 778%?= 7787%;%;: <%8%%@% !"!!)(
waveguide attenuation standard in 75–110 GHz to provide traceability in linearity
$% 778%@@ 7787%;%;: <%8%7=% !"!!'%
9%
for other millimeter-wave 7787=
instruments and systems.7787%;%;: %8%>@% %8%;7?
@% 778%@ 7787%;%;: <%8%;;% <%8%7;9
?% 7%8?@ 7787%;%;: <%87;;% <%8%9%:
Table 6.4: Measurement results of a 0–100 dB SVA at 110 GHz
figure of 0.5 dB is used to improve the SNR for high attenuation measurement.
Let xk and yk (k = 1, 2, ..., 240) denotes the real and imaginary part of the com-
plex data detected by LIA in 4 minutes with one sample per second, the smoothed
Each sample in z̄i (i = 1, 2, ..., 121) is obtained by averaging the complex data
121
6.3. Attenuation Measurement System at 140–220 GHz
on the value of z̄i which has reduced the effect of random noise. The mean of
fluctuation of n1 = 0.97 µV ), when the total insertion loss between the test ports
is 104 dB at 110 GHz. Thus the SNR in LIA measurement after smoothing using
(6.6) is estimated to be
s
SN R = 20 log10 ≈ 69 (dB). (6.7)
n1
The signal level at port 1 is 5 dBm at 110 GHz and the noise floor at receiver input
(port 2) is estimated to be
GHz
Fig. 6.2. A microwave signal (11.67–18.33 GHz) drives a multiplier chain (×12)
to generate a signal in 140–220 GHz. The signal level at Port 1 is about −6 dBm.
a main and a reference IF signal at 2.5 GHz, respectively. The frequency of the
local oscillator (LO1 ) is set to fLO which satisfies (fRF − fLO ) × 12 = 2.5 GHz,
122
6.3. Attenuation Measurement System at 140–220 GHz
where fRF is the frequency of the RF source. The main IF signal goes through
noise figure of 8 dB at 2.5 GHz is used to amplify the IF signal. The reference AF
signal provides a frequency reference to the LIA to stabilize the LIA reading. The
about 0.004 dB. A waveguide LSA is used to control the signal level at the receiver
input and minimize the compression error in the main subharmonic mixer and the
RF mixer. The isolator and matching pad in the source side and receiver side are
used to reduce the reflection from the multiplier chain and the main subharmonic
mixer, respectively.
5)6RXUFH '87
*+] 3RUW 3RUW
,VRODWRU$ 0DLQ6XEKDUPRQLF
,VRODWRU%
0L[HU
; /6$
0DWFKLQJ 0DWFKLQJ
3DG$ 3DG% ,)*+]
7HVW3RUWV /2 %3)LOWHU
/2
,VRODWRU ,VRODWRU ,VRODWRU ,VRODWRU *+] ,9'
0+]
5HIHUHQFH
$)N+]
5HIHUHQFHN+] /RFNLQ
/3)LOWHU 3UH$PS
$PS
method described in 3.2, using a matching pad of 6 dB at each side of the DUT.
(described in 3.2) using a 0–60 dB variable attenuator as the GBA and the matching
123
6.3. Attenuation Measurement System at 140–220 GHz
the 10 dB attenuation with good repeatability. When the DUT is set to 0 dB, the
)*+#,.. LSA cascade-connected with another variable attenuator are adjusted to reduce the
4,5*- 6 1+. 3
7%;%;: <%8%;$= <%8%7:>
initial signal level at subharmonic mixer input from −53 dBm to −133 dBm in 10
7%;%;: <%8%%=% !"!!#$
7%;%;: <%8%%;? !"!!%&
dB step. The DUT is measured eight times at each signal level and the results are
7%;%;: %8%7:% !"!!'(
7%;%;: <%8%%@% !"!!)(
7%;%;: <%8%7=% listed
!"!!'% in Table 6.5. The measurement repeatability for each signal level is obtained
7%;%;: %8%>@% %8%;7?
7%;%;: <%8%;;% using (3.16). The measurements for initial mixer input level from −53 dBm to
<%8%7;9
7%;%;: <%87;;% <%8%9%:
−103 dBm are averaged to obtain the reference value Ar = 9.9908 dB, and their
maximum deviation from Ar is eLIN = 0.0026 dB. The standard uncertainty due
to nonlinearity is estimated to be
%87;
%8;= √
%8>$= eLIN / 3 = 0.0015 dB.
uLIN (6.10)
%8=@
%8$
The increased deviations from78;the reference value at mixer input level below −113
78=
78$ noise.
dBm are due to the leakage and
78@
;
Table 6.5: Measurement ;8;of the receiver nonlinearity error at 220 GHz
;8=
%!
%$
&
"
&
&
#$
+# :=8%
# # -/#;;%#&'( 124
)E11K1+. 16I'-.2#2*A,.#*3
1*&#FEG,. <7:$
6.3. Attenuation Measurement System at 140–220 GHz
Two cascade-connected precision RVAs (referred to as RVA1 and RVA2 ) with 0–60
GHz to verify the performance of the attenuation measurement system. Eight mea-
surements are made at each attenuation step and the results are given in Table 6.7.
The measurement uncertainty components include the errors due to IVD, nonlin-
uncertainty and measurement repeatability. Table 6.6 gives the uncertainty budget
Table 6.6: Uncertainty estimation for measurement of the 0–120 dB SVA at 220
GHz
&'()!*$
%
#" (( ((
5
( ( ( ( ( ( ( ( ( (
6
&
3 *( 3( * * 3
7 '
7
6
(( '* 3 ( 3 * *
/
( ('
' * * ** '
!"#$
The standard uncertainty due to receiver noise is about 0.038 dB and 0.12 dB
when the DUT attenuation is 110 dB and 120 dB, respectively. The standard
125
6.3. Attenuation Measurement System at 140–220 GHz
The leakage errors in measuring high attenuation values are estimated using the
method described in section 6.2.3. The 0–60 dB RVA1 is first measured by the
system to get the reference values for the 0–60 dB attenuation setting in 10 dB
step. The RVA1 is then measured again with another RVA (set to 60 dB attenuation
step) cascade-connected with it. In this case, the total inserted nominal attenuation
value between the test ports becomes 70, 80, 90, 100, 110 and 120 dB as the RVA1
changes to 10, 20, 30, 40, 50 and 60 dB, respectively. The deviations between the
measured and the reference values of RVA1 (0–60 dB) show that there is leakage
error in measurement of attenuation beyond 100 dB. The standard uncertainty due
!"#$%#&'(#)*+,-.*/0
to leakage at 220 GHz is about 0.044 dB, 0.13 dB and 0.37 dB for a DUT with
)*+#,..
)1- 2,-+# .,3 4,5*- 6
100 dB, 110 dB and 7%120 dB attenuation, respectively.
778%9:$
Table 6.7 shows that the
7787%;%;: <%8%;$= <%8%7:>
;% 778%?@ 7787%;%;: <%8%%=% !"!!#$
deviation of the mean>%measured
778%??7 value from the nominal value
7787%;%;: is within
<%8%%;? the 2% of
!"!!%&
=% 778779 7787%;%;: %8%7:% !"!!'(
:% each778%?=
the nominal values for attenuation step. 7787%;%;: <%8%%@% !"!!)(
$% 778%@@ 7787%;%;: <%8%7=% !"!!'%
9% 7787= 7787%;%;: %8%>@% %8%;7?
@% 778%@ 7787%;%;: <%8%;;% <%8%7;9
Table 6.7: Measurement results of a 0–120 dB synthesized rotary vane attenuator
?% 7%8?@ 7787%;%;: <%87;;% <%8%9%:
at 220 GHz
126
A:%%
B C#D##+E*1,#3)EE. -2F#&-*+# :=8%
6.3. Attenuation Measurement System at 140–220 GHz
The LIA time constant is set to 5 s and the equivalent noise bandwidth in the LIA
is 0.024 Hz. The complex data measured by the LIA in 4 minutes (one sample
per second) is averaged to achieve a SNR of about 58 dB when the total insertion
loss between the test ports is 105 dB at 220 GHz. Thus the measurement dynamic
The signal level at port 1 is −6 dBm at 220 GHz and the noise floor at receiver
127
6.4. Attenuation measurement System at 330–500 GHz
GHz
System
The signal level at Port 1 varies from −18 dBm to −20 dBm at 340–497 GHz,
decreasing to −23 dBm at the edge of the WR–2.2 band. A subharmonic mixer is
Another synthesizer is used as the local oscillator (LO1 ) and its frequency is set
to fLO so that (fRF − fLO ) × 24 = 2.5 GHz, where fRF is the frequency of the
filter with a pass-band from 5.01 kHz to 5.03 kHz is used to reduce the noise in
port 1 and port 2 to improve the port match. A function generator provides a 5.02
kHz frequency reference signal for the LIA. A 10 MHz reference signal is used to
synchronize the RF source, LO1 , LO2 and the function generator. Accurate mea-
In the frequency range closer to the carrier frequency, the flicker noise (also called
1/f noise or phase noise) is more significant [81–83, 235–237, 245]. The phase noise
128
6.4. Attenuation measurement System at 330–500 GHz
/2 ;
5)
$)N+]
0DLQ5)0L[HU
/2
/3)LOWHU
0+]
5HIHUHQFH 3UH$PS
/2
*+] %3)LOWHU
,9'
)XQFWLRQ*HQHUDWRU /RFNLQ
&RPSXWHU
N+] $PS
([W5HIHUHQFHN+]
of an oscillator increases with its carrier frequency. It has been shown that phase
level than the noise figure which is commonly used [232]. The SSB phase noise at
where L(f ) is the SSB phase noise at the input of the ×N frequency multiplier
the frequency of a microwave signal (13.75–20.83 GHz) by N = 24, thus the phase
where LM (f ) and LT (f ) are the SSB phase noises of the microwave synthesizer
and the sub-millimeter wave source, respectively. The phase noises of the LO2
129
6.4. Attenuation measurement System at 330–500 GHz
drive and the sub-millimeter wave sources are translated to the AF signal at 5.02
kHz after the mixer down-conversion. The phase noise contribution (in AF signal)
The random phase jitter of the AF signal has a root-mean-square (RMS) value (in
radians) given by s Z
f2
φj = 2 [2LT (f ) + LLO2 (f )]df . (6.15)
f1
The SSB phase noises of the LO2 drive (2.5 GHz), the microwave synthesizer (13.75–
20.83 GHz) and the sub-millimeter wave source (330–500 GHz) are shown in Fig.
6.4. The 95% confidence interval of the phase of the AF signal is estimated to be
[θ̄ − 174.6◦ , θ̄ + 174.6◦ ] where θ̄ is the mean of the phase (in degrees).
0
2.5 GHz
13.75 − 20.83 GHz
−20 330 − 500 GHz
SSB Phase noise (dBc/Hz)
−40
−60
−80
−100
−120
−140
0 1 2 3 4 5
10 10 10 10 10 10
Frequency offset (Hz)
Figure 6.4: The SSB phase noise of the microwave sources (based on the instrument
specification at 2.5 GHz and 13.75–20.83 GHz) and the calculated phase noise of
The phase of the AF signal has been measured by the LIA with a 5.02 kHz fre-
130
6.4. Attenuation measurement System at 330–500 GHz
synthesizers via a 10 MHz signal. Fig. 6.5 shows the measured magnitude and
phase of the AF signal down-converted from a signal at 480 GHz, with 10 dB at-
tenuator inserted between the test ports. The LIA time constant is set to 1 s. The
measured AF signal shows random phase fluctuation from −180◦ to 180◦ , which is
Such a large phase jitter causes significant fluctuation in the magnitude of the
AF signal measured by the LIA. Thus it is very difficult to make accurate attenu-
ation measurement using LIA with a 5.02 kHz frequency reference from a function
signal. The LIA requires a coherent frequency reference signal to accurately detect
131
6.4. Attenuation measurement System at 330–500 GHz
200
magnitude, microvolt
150
100
50
0
0 20 40 60 80 100 120
time, s
200
150
100
phase, degree
50
−50
−100
−150
−200
0 20 40 60 80 100 120
time, s
Figure 6.5: The measured magnitude and phase of the AF signal down-converted
from a signal at 480 GHz using the single-channel system, with 10 dB attenuator
132
6.4. Attenuation measurement System at 330–500 GHz
To reduce the significant receiver fluctuations caused by the phase noise effect in
Fig. 6.6. Two subharmonic mixers are used to down-convert the sub-millimeter
wave signals to a main and a reference IF signal at 2.5 GHz, respectively. The
frequency of LO1 is set to fLO so that (fRF − fLO ) × 24 = 2.5 GHz, where fRF is
the frequency of the RF source. The main and reference IF signals go through BP
Ref Subharmonic LO X 12
X 12 Splitter
Mixer
IF (2.5 GHz)
LO1 RF
BP Filter AF (5.02 kHz)
Main RF Mixer
LO
LP Filter
Attenuator
RF
LO Pre-Amp
Ref RF Mixer
BP Filter
Isolators
10 MHz
Reference LO2 IVD
(2.49999498 GHz)
AF (5.02 kHz)
Lock-in
LP Filter Pre-Amp BP Filter Amp
Ext Reference (5.02 kHz)
Computer
The reference channel receiver in Fig. 6.6 is used to provide a coherent frequency
reference to the LIA to stabilize its magnitude reading. A BP filter with a pass-
band from 5.01 kHz to 5.03 kHz is used to reduce the noise in the main channel and
133
6.4. Attenuation measurement System at 330–500 GHz
signal in the reference channel is very close to that of the main channel. RF isolators
with total reverse isolation of 180 dB are placed between the main and reference
RF mixers to reduce the crosstalk between the main and reference channel.
Fig. 6.7 shows the measured magnitude and phase of the AF signal down-
converted from a signal at 480 GHz, with a 10 dB attenuator inserted between the
test ports and the LIA reference is provided by the coherent reference channel. The
LIA time constant is set to 1 s. It is clear that the receiver fluctuations in magnitude
and phase have reduced significantly. The phase fluctuation around its mean value
LIA voltage reading fluctuation reduces to 0.004 dB using this dual-channel system.
134
6.4. Attenuation measurement System at 330–500 GHz
1200.5
magnitude, microvolt
1200
1199.5
0 20 40 60 80 100 120
time, s
52.2
52
phase, degree
51.8
51.6
51.4
51.2
0 20 40 60 80 100 120
time, s
Figure 6.7: The measured magnitude and phase of the AF signal down-converted
from a signal at 480 GHz using the dual-channel system, with 10 dB attenuator
135
6.4. Attenuation measurement System at 330–500 GHz
Dual-channel System
noise figure of 0.5 dB at 2.5 GHz is inserted before the main RF mixer. The LNA
Two waveguide matching pads are placed at test ports to improve the port match.
method, with a matching pad of 10 dB at each side of the DUT. For a DUT
with attenuation higher than 50 dB, it is measured using the indirect measurement
dB.
The LIA time constant is set to 10 s to improve the SNR in measurement. The
equivalent receiver bandwidth in the LIA is 0.011 Hz. When the attenuation from
Port 1 to Port 2 is A21 = 117 dB, the receiver SNR is measured to be 35 dB at 480
GHz if the phase sensitive measurements made in 2 minutes are averaged using the
The signal level at port 1 is −18 dBm at 480 GHz and the noise floor at the receiver
136
6.4. Attenuation measurement System at 330–500 GHz
The measurement uncertainty components include the errors due to IVD, nonlin-
tainty and measurement repeatability. The standard uncertainty due to IVD ratio
at initial input levels (at port 2) deceasing from −36 dBm to −86 dBm in 10 dB
The matching pads at test ports help to reduce the mixer compression error for
The receiver fluctuation for high attenuation measurement is mainly due to the
thermal noise and phase noise in the system. The standard uncertainty due to
receiver noise is about 0.0085 dB and 0.088 dB when the DUT attenuation is 80
dB and 100 dB, respectively. The standard mismatch uncertainty is about 0.037
dB and 0.052 dB for DUT attenuation below and above 50 dB, respectively.
VA1 and VA2 ), each with 0–50 dB attenuation, are used to form a 0–100 dB
SVA are estimated to be 0.078–0.24 dB at 480 GHz. Table 6.8 gives the uncertainty
at 480 GHz to verify the performance of the system. The results are given in Table
6.9. The VA2 is first measured at 10–50 dB with VA1 set to 0 dB. The incremental
137
6.4. Attenuation measurement System at 330–500 GHz
Table 6.8: Uncertainty estimation for measurement of a 0–100 dB SVA at 480 GHz
!"#$%#&'(#)*+,-.*/0
!"#$%#&'(#)*+,-.*/0
attenuation of the VA1 at 0–50 dB step is measured to be AGBA = 50.340 dB, which
!"#$%%
&'( )$*"# %$+ ,$-!* .
is used as the GBA reference
12 for measuring the SVA113127275
1132456 at its 60–1008232769
dB range using the
823215:
72 1132;< 113127275 8232292 1%2%%34
indirect measurement
:2 method.
1132;;1With VA1 set to 50 dB, the measured
113127275 1%2%%56
823227; attenuation
92 113114 113127275 232152 %2%%7$
52
of VA2 at its 0–50 dB step1132;9
is added to the 113127275
AGBA to obtain 82322<2 1%2%%8$
the attenuation of SVA
62 1132<< 113127275 8232192 1%2%%75
42
at its 0–100 dB step. 11319 113127275 232:<2 23271;
<2 1132< 113127275 8232772 8232174
;2 123;< 113127275 8231772 8232425
Table 6.9: Measurement results of a 0–100 dB SVA at 480 GHz
%-*#%$8 *= $,#=>#./ 62,=#?@*"A$,#BC##52#,=#)*%D
138
6.4. Attenuation measurement System at 330–500 GHz
When the SVA is set to 100 dB and a matching pad of 6 dB is connected at each
test port, the power level at port 2 is around −135 dBm. There might be error
sub-millimeter wave signal or cross-talk signal from the reference channel. We have
!"#$%#&'(#)*+,-.*/0
conducted an experiment using the SVA to check whether the measurement error
!"#$%%
&'( )$*"# %$+
,$-!* .
due to leakage
12 is significant.
1132456 113127275 8232769 823215:
72 1132;< 113127275 8232292 1%2%%34
:2 1132;;1 113127275 823227; 1%2%%56
With VA92 1 set113114
to 50 dB, the VA2 is measured
113127275 again at its 10–50
232152 %2%%7$dB range in 10
52 1132;9 113127275 82322<2 1%2%%8$
62 results
dB steps with 1132<<given in Table 6.10.
113127275 also shows1%2%%75
Table 6.108232192 the deviations from
42 11319 113127275 232:<2 23271;
<2 1132< 113127275 8232772 8232174
measurements
;2 of123;<
VA2 made at a power level which
113127275 is 42 dB 8232425
8231772 higher (used as the
%-*#%$8 *=reference
$,#=>#./ 62,=#?@*"A$,#BC##52#,=#)*%D
values, given in row 1 to row 5 in Table 6.9). The deviations show that
measurements made at power level from −85 dBm to −135 dBm (Table 6.10) agree
with the reference values for VA2 (given in Table 6.9) within their measurement 232:2 23
231;< 13
uncertainties, indicating that leakage error in attenuation measurement is negligible 234:4 13
73191 73
at 480 GHz. 13551 73
823:17 :3
823119 :3
Table 6.10: Measurement of the VA2 at 480 GHz at receiver input levels that are 23996 93
13;9< 93
1379< 53
42 dB lower than those given in Table 6.9
6.5 Summary
GHz and 330–500 GHz have been developed for precision attenuation measurement.
able to a low frequency voltage ratio standard at 5.02 kHz. The measurement
0.11 dB at 220 GHz and 0.24 dB at 480 GHz, respectively. The noise floor of
the measurement receiver is around −168 dBm at 110 GHz and −169 dBm at 220
GHz, when a LNA is used to boost up the main channel IF signal. The DR of the
220 GHz. The proposed dual-channel heterodyne receivers in WR–2.2 band has
GHz.
The DR of these new measurement systems is much higher than that of a VNA
[122, 123, 134]. The measurement systems have been verified using precision RVAs.
These systems have been used as Singapore national standards for millimeter-wave
tors calibrated by these systems can serve as reference and transfer standards for
With antenna connected at the test ports, these systems can also be used for high-
can be used for the frequency range at 500–750 GHz or 750–1100 GHz and the
receiver noise floor is expected to be much lower than that of a VNA since it is
140
Chapter 7
Millimeter-wave Phase-shift
Precision Measurement
7.1 Introduction
for plasma diagnosis [219]. This system used a single-channel homodyne receiver
and an air line with 1.85 mm connector for frequencies up to 65 GHz [125]. An
air line can be used as a phase-shift reference standard and provide traceability
[103, 104].
141
7.1. Introduction
110–170 GHz using a VNA [122]. The expanded uncertainty of transmission mea-
The phase-lock loop used in a VNA has some residual random phase error, which
and the LO in the vector receivers. Such random deviations behave like a phase
noise effect which increases the instability in phase measurement for a DUT with
network analyzer [134] or a dual-channel VNA [122] is too high for a DUT with
systems.
The increasing demand for fast wireless data transmission applications requires
communication systems at 300 GHz [106]. In this chapter, we will investigate the
signal to detect the main channel AF signal with low SNR. The deviations be-
tween the instantaneous frequency of the main channel and reference channel AF
signal are negligible since they are generated from the same microwave source and
142
7.1. Introduction
LO. This ensures that the synchronous detection by the LIA can make accurate
shift of a 0–60 dB WR-03 RVA can be measured at 325 GHz with uncertainties
system for a 50 dB attenuator at 325 GHz is ∼44 times smaller than that of a
The leakage error in millimeter-wave phase-shift measurement has not been stud-
ied in the literature. We have proposed a new method to evaluate the leakage errors
to verify the system performance. The measured phase-shift of the waveguide shim
agrees with its calculated phase-shift value within their estimated uncertainties.
The frequency band at 220–325 GHz (WR-03 band) has been selected to study
design can be used for phase-shift precision measurement at other frequency bands
within 50–500 GHz. Chapter 6 has shown that the noise floor of dual-channel
systems using our proposed design is expected to be around −169 dBm at 50–
GHz is expected to have slightly better accuracy compared to WR-03 band, due
than that at WR-03 band, due to lower SNR and higher mismatch uncertainty.
143
7.2. A 220–325 GHz Phase-shift Measurement System
tem
level at port 1 is about −10 dBm at 325 GHz. Two subharmonic mixers are
used to down-convert the signal to a main and a reference IF signal at 2.5 GHz,
respectively. The frequency of the local oscillator (LO1 ) is set to fLO which satisfies
fRF × 18 − fLO × 16 = 2.5 GHz, where fRF is the frequency of the RF source. The
the LIA to stabilize the LIA reading. If the reference of the LIA is generated by
10 MHz reference signal, the maximum phase reading fluctuation at 325 GHz with
DUT. Several RF isolators with total isolation of about 180 dB are placed between
the main and reference RF mixers to reduce the RF leakage signal introduced by
the reference channel. A LSA is used to minimize the compression error in the main
subharmonic mixer and the RF mixer. The matching pads are used to reduce the
144
7.2. A 220–325 GHz Phase-shift Measurement System
RF Source DUT
Port 1 Port 2 Main Subharmonic
(12.22 - 18.06 GHz )
Mixer
X 18 Matching Matching
LSA IF (2.5 GHz)
Pad A Pad B
Test Ports LO
BP Filter
Ref Subharmonic LO X8
X8 Splitter
Mixer
IF (2.5 GHz)
LO1 RF
BP Filter AF (5.02 kHz)
Main RF Mixer
LO
Attenuator LP Filter
Attenuator
RF
LO Pre-Amp
Ref RF Mixer
Isolators IVD
10 MHz
Reference LO2
(2.49999498 GHz)
AF (5.02 kHz)
Reference (5.02 kHz) Lock-in
LP Filter Pre-Amp
Amp
by the LIA. The time constant of the LIA is set to 2 s. Before a DUT is inserted,
the phase detected by the LIA is θ1 . After the insertion of the DUT, the IVD ratio
is adjusted till the LIA detects a signal with magnitude similar to that before DUT
insertion (magnitude difference < 10 µV ), and the phase detected by the LIA is
θI = θ2 − θ1 . (7.1)
obtained from
where θi′ and θf′ are the measured phases when the DUT is at its initial and final
145
7.3. Measurement Uncertainty Analysis
GBA with variable attenuation from 0 to 20 dB. The GBA is set to 20 dB before
DUT insertion, and set to 0 dB after DUT insertion. The phase-shift difference
of the DUT and the GBA is measured to be δθ . The phase-shift of DUT is then
obtained from
through an IVD before its phase is measured by the LIA. The maximum phase error
eI
uI = √ , (7.4)
3
The detected signal by the LIA has a maximum fluctuation of ef = 0.1◦ in its phase
reading when the DUT attenuation is less than 20 dB at 325 GHz. The fluctuation
increases when the noise effect is more significant at higher attenuation levels. The
time constant of the lock-in amplifier is set to 2 seconds to reduce the noise effect.
146
7.3. Measurement Uncertainty Analysis
becomes ef = 0.5◦ at 325 GHz. The standard uncertainty due to the receiver noise
is estimated from
ef
uf = √ , (7.5)
3
assuming the error has a rectangular distribution.
The detected phase of the main channel has a small drift error due to the phase
noises of the two microwave synthesizers and the multiplier chains. The LIA uses
a phase reference signal produced by the reference channel receiver, which helps to
stabilize the phase reading in the LIA. However, the LO signal for the reference
subharmonic mixer still has some random phase jitter which is not synchronized
with the LO for the main subharmonic mixer. Thus some phase drift in the LIA
is expected.
The maximum phase drift in one minute has been measured to be ed = 0.3◦
when the DUT attenuation is below 10 dB, and ed increases to 0.8◦ as the DUT
attenuation increases to 60 dB. The phase drift error is assumed to have a rectan-
gular distribution. The standard uncertainty due to phase drift error is estimated
to be
ed
ud = √ . (7.6)
3
For DUT with attenuation in the range from 20 dB to 60 dB, its phase-shift is
the GBA has been calibrated previously with a measurement uncertainty, which
should be included as one of the uncertainty components for the DUT phase-shift
147
7.3. Measurement Uncertainty Analysis
UGBA
uGBA = , (7.7)
2
to mismatch is
lim ǫd
um = √ (7.8)
3
The standard uncertainty due to mismatch error in differential phase-shift mea-
surement is estimated to be 0.22◦ for a WR-03 RVA at 325 GHz. The maximum
The error limit of the mismatch error in insertion phase-shift measurement is de-
lim ǫI
um = √ (7.9)
3
148
7.3. Measurement Uncertainty Analysis
A millimeter-wave leakage signal refers to the signal which arrives at the receiver via
a path different from the main path which goes through the DUT. For a DUT with
high attenuation, the leakage signal will cause a significant phase-shift measurement
error when the main path signal level is quite low. When the DUT is set to a high
where x~M is the main path signal and x~L is the leakage signal. The angle between
the two vectors x~2 and x~M is referred to as θe . When the DUT is set to its datum
position (0 dB attenuation), the signal detected by the LIA is x~1 . If the magnitude
of x~1 is assumed to be much higher than that of the leakage signal, then the error
The phase difference between x~M and x~L is unknown. The maximum error in
|x~L |
θL = sin−1 (7.11)
|x~M |
x1
ADU T = 20 log10 (dB) (7.12)
xM
x1
AL = 20 log10 (dB) (7.13)
xL
149
7.3. Measurement Uncertainty Analysis
x2 xL
șL
– șL xM
thus we have
xM
AL − ADU T = 20 log10 (dB) (7.14)
xL
and
ADU T −AL
θL = sin−1 10 20 (7.15)
Fig. 7.3 shows the computed maximum error θL for various values of AL − ADU T
ranging from 20 dB to 80 dB. To keep the θL to be smaller than 0.1◦ , the leakage
path signal xL has to be at least 55 dB below the main path signal xM , i.e.,
AL − Adut ≥ 55 dB.
The maximum error in phase-shift measurement due to leakage for DUT with
100 dB, 110 dB, 120 dB, 130 dB and 140 dB. Fig. 7.4 shows that the maximum
error θL is 1.81◦ and 5.74◦ for DUT attenuation at 70 dB and 80 dB, respectively,
150
7.3. Measurement Uncertainty Analysis
0
20 30 40 50 60 70 80
A −A (dB)
L DUT
leakage for various attenuation differences between DUT path and leakage path.
mixer, RF isolators, the directional coupler and the main RF mixer RF port. Six
RF isolators are inserted in the path between LO port of two RF mixers to reduce
such leakage. Directional couplers with high directivity are used to provide 50 dB
isolation between the two channels. The isolation between the main RF mixer RF
port and the LO port is about 30 dB. The total isolation on the internal leakage
path is estimated to be 350 dB, which is sufficient to ensure the leakage effect is
nals from the waveguide flange connectors. These radiated millimeter-wave signals
might enter the measurement receiver and introduce an error in phase-shift mea-
components have been connected carefully through a flange mating process and
151
7.3. Measurement Uncertainty Analysis
6
A = 140 dB
−1
10 20 30 40 50 60 70 80
Attenuation of DUT (dB)
each flange connection is fixed by four screws to reduce radiation leakage to the
minimum level. It is assumed that the leakage path attenuation through waveg-
uide flange interface coupling and free space propagation is large enough so that
the error in phase-shift measurement due to leakage is negligible for DUT attenu-
ation below 60 dB. This assumption will be verified in section 7.5 by measurement
experiments.
The uncertainties due to IVD phase error, receiver noise, phase drift, mismatch er-
152
7.3. Measurement Uncertainty Analysis
Us = kus (7.17)
where k is the coverage factor. The expanded uncertainties (coverage factor k=2)
0.58◦ –1.32◦ at 325 GHz. Table 7.1 gives the phase-shift measurement uncertainty
153
7.4. Phase-shift Measurement of a 0–60 dB RVA
A WR-03 RVA used in our laboratory at the National Metrology Centre is shown in
Fig. 7.5. This precision device can provide variable phase-shift values at specified
dial setting with good repeatability. The differential phase-shift of a WR-03 RVA
Figure 7.5: A WR-03 0–60 dB RVA used in our laboratory at the National Metrol-
ogy Centre.
signal level at Port 2 during measurement is around −93 dBm when the RVA is set
to 60 dB and the GBA set to 0 dB. The measurement results and their expanded
uncertainties (coverage factor k=2) are listed in Table 7.2. The differential phase-
at 325 GHz.
Table 7.2: Differential phase-shift results of a 0–60 dB WR-03 RVA at 325 GHz
154
7.5. Investigation of Leakage Error
of the same RVA is measured again at 325 GHz with a 20 dB attenuator inserted
at Port 1. Then the minimum signal level at port 2 during measurement is around
−113 dBm when the RVA is set to 60 dB and the GBA set to 0 dB. A low-noise
RF amplifier with a noise figure of 6.2 dB is inserted before the main RF mixer
attenuation steps 0–30 dB, 0–40 dB, 0–50 dB and 0–60 dB are listed in Table 7.3.
The deviations from the measurements in Table 7.2 for the same attenuation
range are also listed in Table 7.3. Comparing to results in Table 7.2, the deviation
in differential phase-shift for attenuation steps 0–30 dB and 0–40 dB is 0.03◦ and
0.08◦ respectively, which is much smaller than the measurement uncertainty given in
Table 7.2. Thus we can conclude that the leakage error in phase-shift measurement
is negligible when the signal level at port 2 is above −93 dBm. This means that
which is larger than the uncertainties due to receiver noise and phase drift error.
leakage at 325 GHz, which becomes significant as the signal level at Port 2 decreases
to −113 dBm.
155
7.6. Millimeter-wave Phase-Shift Reference Standard
Table 7.3: Differential phase-shift measurement of the same WR-03 RVA at 325
GHz when the signal levels at port 2 are 20 dB lower than those in Table 7.2
dard
standard (line standard). Fig. 7.6 shows a WR-03 waveguide shim with a length
of 0.997 mm, which is used in our laboratory at the National Metrology Centre.
Figure 7.6: A WR-03 waveguide shim used in our laboratory at the National
Metrology Centre.
λ
λg = q , (7.18)
λ 2
1 − ( 2a )
c
λ= √ , (7.19)
f ǫr
156
7.7. System Verification using a Waveguide Shim
and a is the waveguide broad wall dimension. For WR-03 waveguide, a = 0.8636
mm.
from
l
θ = 360 × (n − ), (7.20)
λg
The phase-shift of the waveguide shim is computed from the longitudinal length,
measurement.
A WR–03 waveguide shim has been measured as a line standard to verify the
system. The longitudinal length of the waveguide shim has been measured to be
0.997 mm with an expanded uncertainty of 0.007 mm. Thus its insertion phase-shift
GHz, and 112.26◦ with an expanded uncertainty of 1.74◦ at 270 GHz, respectively.
uncertainty.
The measured insertion phase-shift of the waveguide shim at 325 GHz and 270
GHz is 28.65◦ and 110.98◦ , with an expanded uncertainty of 0.61◦ and 0.57◦ , respec-
tively. Six measurements have been made at each frequency. The 95% confidence
intervals of the measured insertion phase-shift and the calculated value overlap
with each other at 325 GHz and 270 GHz respectively. The measured and calcu-
157
7.8. Summary
lated phase-shift of the shim, along with their expanded uncertainties, are listed in
Table 7.4. The comparison in Table 7.4 shows that the measurements agree with
uide shim
Mean Expanded
Expanded Calculated Difference between
Measured Uncertainty for
Frequency Uncertainty for Insertion Measured and
Insertion Calculated
Measured Value Phase-shift Calculated Value
Phase-shift Value
(GHz) (Degree) (Degree) (Degree) (Degree) (Degree)
270 110.98 0.57 112.26 1.74 -1.28
325 28.65 0.61 30.89 2.31 -2.24
7.8 Summary
0.58◦ –1.32◦ for a RVA with attenuation up to 60 dB at 325 GHz. This system can
provide accurate phase-shift measurement for various devices at WR-03 band. The
phase noise effect has been suppressed significantly by the proposed dual-channel
tem design can also be used in other applications for accurate phase measurement
at millimeter-wave frequencies.
158
Chapter 8
8.1 Conclusion
DR, accuracy and frequency range of these continuous-wave (CW) systems need to
be increased to meet the ever-increasing demands from the industry and scientific
space is very high and the output power of millimeter-wave CW source is quite
and millimeter-wave signals with high DR. There is a growing demand for accurate
microwave and millimeter-wave frequencies, which are critical for component and
system development.
159
8.1. Conclusion
The DR of our proposed design is much higher than those reported in the litera-
kHz. We have proposed new ways to analyse the leakage error and noise floor
in attenuation measurement, which have not been well studied in the literature.
For frequencies beyond 2 GHz, the system employs a coherent frequency reference
channel to minimize the phase noise effect on lock-in amplifier sensitivity. The
proposed receiver achieves a noise floor around −172 dBm and a record high DR
of 186.7 dB at 26 GHz. Our receiver noise floor is 47 dB lower than that of a VNA
at 26 GHz. The proposed system can measure a step attenuator with attenuation
up to 170 dB at 26 GHz with good accuracy, which is not possible using other
measurement systems.
developed using dual-channel heterodyne receivers. Our design can suppress the
phase noise effect of the microwave sources and significantly improve the phase
measurement accuracy for high attenuation devices. The accuracy of this proposed
system is much better than earlier published results. The phase-shift measurement
uncertainty is 0.09◦ –0.32◦ for a 0–140 dB coaxial step attenuator at 18 GHz. The
GHz is ∼106 times smaller than that of a vector network analyzer (VNA) [133].
which has not been studied in the literature. A coaxial beadless air line has been
160
8.1. Conclusion
system. The calculated phase-shifts with correction of the line insertion loss agree
tems at 75–110 GHz, 140–220 GHz and 330–500 GHz. A commercial VNA has a
typical DR of 105 dB at 200 GHz and 104 dB at 500 GHz [134]. The DR of our
proposed system is estimated to be 175 dB at 110 GHz, 163 dB at 220 GHz and
152 dB at 480 GHz, respectively, which are much higher than that of a VNA. The
∼10 times smaller than that of a VNA. The attenuation measurement is traceable
are estimated to be 0.084 dB at 110 GHz, 0.11 dB at 220 GHz and 0.24 dB at
480 GHz, respectively. The DR of the proposed system is 58 dB higher than that
of a VNA at around 220 GHz. A voltage ratio of 3.0 × 10−5 at 480 GHz can be
proposed dual-channel heterodyne receivers have achieved a record low noise floor
at around −170 dBm at 480 GHz. The measurement systems have been verified
been proposed. The proposed dual-channel receiver design can measure the phase
for a 0–60 dB RVA at 325 GHz. The phase-shift measurement uncertainty of the
proposed system for a 50 dB attenuator at 325 GHz is ∼44 times smaller than
that of a VNA [122, 124]. A waveguide line standard has been measured as a
calculated phase-shifts agree with the measured values of the line standard within
161
8.1. Conclusion
since their noise floor is quite high due to the phase noise effect. This thesis has
shown that our proposed dual-channel heterodyne receiver design has significantly
reduced the measurement error in magnitude and phase due to phase noise at
signal using a lock-in amplifier with precise knowledge of the signal’s instantaneous
frequency.
sis to characterise its performance [129, 135–138]. We have evaluated the measure-
ment errors and uncertainties of the six new measurement standards proposed in
this thesis. It has been very difficult to estimate the leakage error in high attenua-
tion measurement since the leakage signal can not be separated from the noise [117].
The leakage error in phase-shift measurement has not been well studied in the liter-
ature. We have proposed new methods to evaluate the leakage errors in attenuation
and phase-shift measurement for devices with high attenuation. This is critical for
These systems have been used as the Singapore national standards for attenu-
Variable attenuators, air lines and waveguide shims calibrated by these systems
can provide traceability and serve as reference standards for calibration of VNAs,
power meters, radiometers, spectrum analyzers, etc. The dual-channel design using
multipliers and subharmonic mixers can be used at 0.5–1.1 THz and the receiver
noise floor is expected to be much lower than that of a VNA since it is mainly
162
8.2. Recommendations for Future Work
0.75 THz and 0.75–1.1 THz using multipliers, couplers and subharmonic mixers.
High precision in fabrication will be required due to the reduced size of the waveg-
163
8.2. Recommendations for Future Work
mixers and horn antenna operating at frequencies up to 2 THz are already avail-
Such a QO system can calibrate a set of silicon plates which can be used as
QO attenuation reference standards at THz band. Each silicon plate can provide
3.1 dB transmission attenuation at 0.3–3 THz, which has been measured using a
plates as THz phase-shift reference and transfer standards, which can be measured
These research works will help to develop new THz-wave metrology standards
which will fill up the ‘gap’ between the current millimeter-wave and optical stan-
164
Author’s Publications
165
dard in the frequency range from 10 MHz to 26.5 GHz, in: IEEE Instrumen-
tation and Measurement Technology Conference, Singapore, pp. 1620-1624,
2009.
166
Bibliography
[3] F. K. Richtmyer, “The romance of the next decimal place,” Science, vol. 75,
no. 1931, pp. 1–5, 1932.
[4] T. Hänsch, “Nobel Lecture: Passion for precision,” Rev. Mod. Phys., vol. 78,
no. 4, pp. 1297 – 309, 2006.
[7] R. Wilson, “The cosmic microwave background radiation,” Science, vol. 205,
no. 4409, pp. 866–874, 1979, (Nobel Lecture).
[8] J. C. Mather, “Nobel Lecture: From the Big Bang to the Nobel Prize and
beyond,” Rev. Mod. Phys., vol. 79, pp. 1331–1348, Nov 2007.
[10] R. Cowen, “State of the universe: Microwave glow powers cosmic insights,”
Science News, vol. 173, no. 11, p. 163, 15 Mar 2008.
167
Bibliography
[17] R. Pohl et al., “The size of the proton,” Nature, vol. 466, pp. 213–217, 8 Jul
2010.
[20] Y. Bhattacharjee, “Measuring the immeasurable,” Nature, vol. 412, no. 6846,
pp. 474–476, 2001.
[27] J. Flowers, “The route to atomic and quantum standards,” Science, vol. 306,
no. 5700, pp. 1324–1330, Nov. 2004.
168
Bibliography
[28] K. von Klitzing, “The quantized Hall effect,” Rev. Mod. Phys., vol. 58, pp.
519–531, Jul 1986, (Nobel Lecture).
[41] D. Harnett and N. Case, “The design and testing of multirange receivers,”
Proc. IRE, vol. 23, no. 6, pp. 578 – 592, june 1935.
169
Bibliography
[43] A. Fantom, Radio Frequency and Microwave Power Measurement. The In-
stitution of Engineering and Technology, 1990.
[44] M. P. Weidman, Direct Comparison Transfer of Microwave Power Sensor
Calibrations. U.S. Dept. of Commerce: NIST Tech. Note 1379, 1996.
[45] J. R. Juroshek, “A direct calibration method for measuring equivalent source
mismatch,” Microwave Journal, vol. 40, no. 10, pp. 106–118, 1997.
[46] ——, “NIST 0.05-50 GHz direct comparison power calibration system,” in
2000 Conference on Precision Electromagnetic Measurements, 2000, pp. 166
–167.
[47] G. Gainsborough, “A method of calibrating standard signal generators and
radio-frequency attenuators,” J. IEE - Part III: Radio and Communication
Engineering, vol. 94, no. 29, pp. 203 –210, may 1947.
[48] R. E. Grantham and J. J. Freeman, “A standard of attenuation for microwave
measurements,” Trans. American Institute of Electrical Engineers, vol. 67,
no. 1, pp. 329 –335, jan. 1948.
[49] B. O. Weinschel, “Development in attenuation measurements and standards,”
J. Res. Nat. Bur. Stand., vol. 64D, no. 6, pp. 599 – 600, 1960.
[50] A. V. James, “A high-accuracy microwave-attenuation standard for use in
primary calibration laboratories,” IRE Trans. Instrum., vol. I-11, no. 3, pp.
285 –290, dec. 1962.
[51] R. W. Beatty, Microwave Attenuation Standards and Measurements. U.S.
Dept. of Commerce: National Bureau of Standards Monograph 97, 1967.
[52] D. Russell and W. Larson, “RF attenuation,” Proc. IEEE, vol. 55, no. 6, pp.
942 – 959, Jun. 1967.
[53] B. O. Weinschel, “New attenuation standard uses laser interferometer.” Mi-
crowave Journal, vol. 27, no. 8, pp. 145–148, 1984.
[54] H. Bayer, F. L. Warner, and R. W. Yell, “Attenuation and ratio - national
standards,” Proc. IEEE, vol. 74, no. 1, pp. 46 – 59, Jan. 1986.
[55] M. Magid, “Precision microwave phase shift measurements,” IRE Trans. on
Instrum., vol. I-7, no. 3, pp. 321 –331, Dec. 1958.
[56] G. E. Schafer, “Mismatch errors in microwave phase shift measurements,”
IRE Trans. Microwave Theory Techniques, vol. 8, no. 6, pp. 617 –622, Nov.
1960.
[57] G. E. Schafer and R. W. Beatty, “Error analysis of a standard microwave
phase shifter,” J. Res. Nat. Bur. Stand., vol. 64 C, no. 4, pp. 261 – 265, Oct.
1960.
[58] S. Cohn and N. Weinhouse, “An automatic microwave phase-measurement
system,” Microwave Journal, vol. 7, no. 2, pp. 49 – 56, 1964.
170
Bibliography
[59] R. W. Beatty, “Some basic microwave phase shift equations,” J. Res. Nat.
Bur. Stand., vol. 68D, no. 4, pp. 349–353, 1964.
[61] ——, “UHF and microwave phase-shift measurements,” Proc. IEEE, vol. 55,
no. 6, pp. 960 – 969, Jun. 1967.
[63] S. Protheroe and N. Ridler, “Using air lines as references for VNA phase
measurements,” in ARMMS RF & Microwave Society Conf. Digest, Buck-
inghamshire, UK, Nov. 2005.
[67] ——, “Errors in coaxial air line standards due to skin effect,” Microwave
Journal, vol. 33, no. 11, pp. 131–43, 1990.
[70] N. M. Ridler and A. G. Morgan, “New primary reference standard for vec-
tor network analyser calibration at millimetre wavelengths in coaxial line,”
Measur. Sci. Technol., vol. 19, no. 6, pp. 065 103 – 065 112, 2008.
[73] H. Friis, “Noise figures of radio receivers,” Proc. IRE, vol. 32, pp. 419 – 422,
1944.
171
Bibliography
[75] B. M. Oliver, “Thermal and quantum noise,” Proc. IEEE, vol. 53, no. 5, pp.
436–454, 1965.
[80] W. Schottky, “Small-shot effect and flicker effect,” Phys. Rev., vol. 28, pp.
74 – 103, 1926.
[82] L. S. Cutler and C. L. Searle, “Some aspects of the theory and measurement
of frequency fluctuations in frequency standards,” Proc. IEEE, vol. 54, no. 2,
pp. 136 – 154, 1966.
[83] A. L. Lance, W. D. Seal, and F. Labaar, Phase noise and AM noise mea-
surements in the frequency domain, in Infrared and Millimeter Waves, K. J.
Button, Ed. New York: Academic Press, 1984, pp. 239 – 289.
172
Bibliography
[91] R. Collier and D. Skinner, Eds., Microwave Measurements, 3rd ed. The
Institution of Engineering and Technology, 2007.
[102] ——, “Multistatic synthetic aperture imaging of aircraft using reflected tele-
vision signals,” in Proc. of SPIE, vol. 4382, 2001, pp. 1–12.
[103] P. Smulders, “Exploiting the 60 GHz band for local wireless multimedia
access: prospects and future directions,” IEEE Communications Magazine,
vol. 40, no. 1, pp. 140 –147, jan 2002.
173
Bibliography
[104] R. Daniels and R. Heath, “60 GHz wireless communications: emerging re-
quirements and design recommendations,” IEEE Vehicular Technology Mag-
azine, vol. 2, no. 3, pp. 41 –50, sept. 2007.
[106] H.-J. Song and T. Nagatsuma, “Present and future of Terahertz communi-
cations,” IEEE Trans. Terahertz Science Technology, vol. 1, no. 1, pp. 256
–263, sept. 2011.
[115] Y.-S. Lee, “Testing dynamic accuracy of vector network analyzers using the
40 GHz step attenuator,” in 65th ARFTG Microwave Measurements Confer-
ence Digest, 2005, pp. 159–168.
174
Bibliography
[127] J.-G. Lee, J.-H. Kim, J.-I. Park, and U.-T. Kang, “A broadband attenuation
standard,” Measur. Sci. Technol., vol. 15, no. 1, pp. 55–57, 2004.
175
Bibliography
[134] P. Goy and M. Gross, “Millimeter and submillimeter wave vector measure-
ments,” in Proc. SPIE, vol. 3465, USA, 1998, pp. 288 – 303.
[141] R. W. Beatty, “Insertion loss concepts,” Proc. IEEE, vol. 52, no. 6, pp. 663
– 671, Jun. 1964.
176
Bibliography
[143] F. K. Weinert, “An automatic precision IF-substitution vector ratio meter for
the microwave frequency range,” IEEE Trans. Instrum. Meas., vol. IM-29,
no. 4, pp. 471 – 477, 1980.
[147] W. Ruhle and D. Stumpe, “Voltage ratio technique used with the national
RF standard attenuation measuring equipment at the PTB,” IEEE Trans.
Instrum. Meas., vol. 42, no. 6, pp. 1009 –1013, Dec. 1993.
[151] D. L. Hollway and F. P. Kelly, “A standard attenuator and the precise mea-
surement of attenuation,” IEEE Trans. Instrum. Meas., vol. 13, no. 1, pp. 33
–44, march 1964.
[154] P. Mariner, “An absolute microwave attenuator,” Proc. IEE - Part B: Elec-
tronic and Communication Engineering, vol. 109, no. 47, pp. 415 –419,
september 1962.
177
Bibliography
[160] J.-G. Lee, J.-H. Kim, J.-S. Kang, and T.-W. Kang, “Novel attenuation stan-
dards at microwave frequencies and evaluation of their uncertainty,” Measur.
Sci. Technol., vol. 18, no. 7, pp. 1929–1933, 2007.
[163] Q. L. Gao, W. Liang, H. Huang, and Y. Li, “A new synchronous method for
attenuation measurement system,” in Conference on Precision Electromag-
netic Measurements Digest, 2008, pp. 262–263.
[168] D. Sharrit, “Vector network analyzer with integral processor,” U.S. Patent
4 703 433, 1987.
178
Bibliography
[172] N. M. Ridler, “Connectors, air lines and RF impedance,” in The 14th IEE
Microwave Measurements Training Course, May 2005, pp. 9 – 9/16.
[173] K. H. Wong, “Using precision coaxial air dielectric transmission lines as cal-
ibration and verification standards,” Microwave Journal, vol. 31, no. 12, pp.
83–92, 1988.
[175] ——, “Exact principal mode field for a lossy coaxial line,” IEEE Trans.
Microw. Theory Tech., vol. 39, no. 8, pp. 1313 –1322, Aug. 1991.
[176] ——, “The propagation constant of a lossy coaxial line with a thick outer
conductor,” IEEE Trans. Microw. Theory Tech., vol. 43, no. 4, pp. 907 –911,
Apr. 1995.
[181] ——, “THz instruments for space,” IEEE Trans. Antennas Propag., vol. 55,
no. 11, pp. 2957 –2965, Nov. 2007.
[183] P. Han and X.-C. Zhang, “Free-space coherent broadband terahertz time-
domain spectroscopy,” Measur. Sci. Technol., vol. 12, no. 11, pp. 1747–1756,
2001.
[184] S. Mickan and X.-C. Zhang, “T-ray sensing and imaging,” Int. J. High Speed
Electronics and Systems, vol. 13, no. 2, pp. 601–676, 2003.
179
Bibliography
[187] P. Goldsmith, C.-T. Hsieh, G. Huguenin, and J. Kapitzky, “Focal plane imag-
ing systems for millimeter wavelengths,” IEEE Trans. Microw. Theory Tech.,
vol. 41, no. 10, pp. 1664–1675, 1993.
[188] C. Tang and K. Aydin, “Scattering from ice crystals at 94 and 220 GHz mil-
limeter wave frequencies,” IEEE Trans. Geoscience Remote Sensing, vol. 33,
no. 1, pp. 93 – 99, 1995.
[190] M. Kemp, “Millimetre wave and terahertz technology for the detection of
concealed threats: a review,” in Proc. SPIE, vol. 6402, USA, 2006, pp. 64 020
– 64 021.
[194] D. Sheen et al., “Active imaging at 350 GHz for security applications,” in
Proc. of SPIE, vol. 6948, 2008.
[196] P. H. Siegel and R. J. Dengler, “Terahertz heterodyne imaging Part II: Instru-
ments,” Int. J. Infrared and Millimeter Waves, vol. 27, no. 5, pp. 631–655,
2006.
[197] L.-P. Schmidt, S. Biber, G. Rehm, and K. Huber, “THz measurement tech-
nologies and applications,” in 14th Int. Conf. on Microwaves, Radar and
Wireless Comm., 2002, vol. 2, 2002, pp. 581 – 587.
180
Bibliography
[200] H. Eisele, “State of the art and future of electronic sources at terahertz
frequencies,” Electron. Lett., vol. 46, no. 26, pp. S8–S11, 2010.
[206] Erickson Instruments LLC, Erickson Instruments PM4 Power Meter Oper-
ating Manual, 2008.
[210] C. Canada, “A 33-110 GHz millimeter wave loss scanner,” IEEE Trans. In-
strum. Meas., vol. IM-20, no. 4, pp. 332 – 337, 1971.
181
Bibliography
[213] A. Fung, D. Dawson, and et al, “Two-port vector network analyzer mea-
surements in the 218-344 and 356-500 GHz frequency bands,” IEEE Trans.
Microw. Theory Tech., vol. 54, no. 12, pp. 4507–4511, 2006.
[218] P. Goy and M. Gross, “Millimeter and sub-millimeter network vector ana-
lyzer,” U.S. Patent 5 119 035, 1992.
[224] Anritsu, VNA dynamic range and linearity accuracy. Application Note No.
11410-00487, Rev. A, 2008.
182
Bibliography
[234] A. R. Kerr and J. Randa, “Thermal noise and noise measurements: A 2010
update,” IEEE Microw. Mag., vol. 11, no. 6, pp. 40 –52, oct. 2010.
[237] B.-G. Goldberg, “Phase noise theory and measurements: A short review,”
Microwave Journal, vol. 43, no. 1, pp. 112–120, 2000.
[238] D. Howe and J. Ostrick, “100-GHz cooled amplifier residual PM and AM noise
measurements, noise figure, and jitter calculations,” IEEE Trans. Microw.
Theory Tech., vol. 51, no. 11, pp. 2235 – 2242, nov. 2003.
183
Bibliography
[240] J. Bollinger, “Filtering noise with a quantum probe,” Nature, vol. 473, pp.
39–40, 2011.
[242] ——, “Investigation of RF leakage in single channel and dual channel mi-
crowave attenuation measurement systems,” J. Astronautic Metrology Mea-
surement, vol. 28, no. 5, pp. 1–5, 2008.
[243] ——, “Comparison of a single channel and a dual channel microwave at-
tenuation measurement system,” in Proc. of the 71st ARFTG Microwave
Measurement Symposium, Atlanta, USA, Jun. 2008, pp. 62 –67.
[246] F. Walls and A. Demarchi, “RF spectrum of a signal after frequency mul-
tiplication - measurement and comparison with a simple calculation,” IEEE
Trans. Instrum. Meas., vol. 24, no. 3, pp. 210–217, 1975.
[247] P. Goldsmith, “Quasi-optical techniques,” Proc. IEEE, vol. 80, no. 11, pp.
1729 – 47, 1992.
[249] B. Thomas et al., “Terahertz cooled sub-harmonic Schottky mixers for plan-
etary atmospheres,” in Proc. 5th ESA Workshop Millimetre Wave Technol.
Appl, Noordwijk, The Netherlands, 2009, pp. 101–108.
184