See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/41586642

Spectrum and Network Analyzers

Article · January 1998

DOI: 10.1142/9789812818003_0029 · Source: OAI

CITATION READS

1 1,504

2 authors:

John Byrd Fritz Caspers

Argonne National Laboratory CERN (ESI)
246 PUBLICATIONS 2,275 CITATIONS 349 PUBLICATIONS 2,766 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Fritz Caspers on 27 December 2013.

The user has requested enhancement of the downloaded file.

 EUROPEAN ORGANISATION FOR NUCLEAR RESEARCH

CERN - PS DIVISION

CERN/PS 99-003 (RF)

SPECTRUM AND NETWORK ANALYZERS

J.M. Byrd*) and F. Caspers

Abstract

Spectrum and network analyzers (SPAs and NWAs) are found in every microwave
laboratory. This report describes the basic principles of operation of a SPA and NWA and also
describes several simple example measurements using each of these instruments.

*) LBNL, Berkeley, CA 94720

The Joint US-CERN-Japan-Russia School on Particle Accelerators

Course on “Beam Measurement”, 11 – 20 May, 1998, Montreux and CERN

Geneva, Switzerland
February 1999
 SPECTRUM AND NETWORK ANALYZERS

J. M. BYRD
Lawrence Berkeley National Laboratory, Berkeley, CA 94720
JMByrd@lbl.gov

F. CASPERS
CERN, Geneva, Switzerland
Fritz.Caspers@cern.ch

Abstract
Spectrum and network analyzers (SPAs and NWAs) are found in
every microwave laboratory. This report describes the basic principles of
operation of a SPA and NWA and also describes several simple example
measurements using each of these instruments.

1 Introduction
Spectrum and network analyzers are commonly found in microwave laborato-
ries and are the primary instruments for characterizing radiofrequency (RF)
and microwave systems for use in accelerators. The NWAs and SPAs in use
today are the result of over half a century of development. The capabilities of
these instruments are remarkable. The best commercially available SPAs to-
day have a dynamic range of greater than 100 dB (corresponding to 10 orders
of magnitude in power) and over six orders of magnitude in frequency. NWAs
can span over five to six orders of magnitude in frequency with up to 0.1 Hz res-
olution. Furthermore, automation and built-in software analysis have greatly
increased their ease of operation. However, because of the specialized nature
of these instruments, many accelerator scientists and engineers are not aware
of the capabilities of these instruments. This paper presents an introduction to
the basic principles of these devices, as well a specialized jargon, particularly
for accelerator–related measurements in a microwave laboratory.

2 Spectrum analyzers
2.1 Basic principle of superheterodyne detection
The basic function of a spectrum analyzer is to determine the amplitude of
the frequency components of a time domain signal. For a simple sinusoidal
signal, this could be easily done by measuring the frequency and amplitude
of the signal on an oscilloscope. However, for a signal containing multiple
frequencies, this becomes difficult, especially when the frequency components
have greatly differing amplitudes. A narrowband tuneable filter could be tuned
through the frequency range of interest and the power measured at each filter
frequency. However, with the technical difficulty of making a tuneable filter, it
is much simpler to have a filter with a fixed frequency and to translate the signal
frequency to the filter frequency. The frequency translation is accomplished
by a process known as mixing, or multiplying the signal with a local oscillator
(LO) signal in a device known as a mixer. A mixer is a passive element made up
primarily of diodes configured in a way to produce the product of two signals
at the RF and LO ports. Multiplication in an ideal mixer generates signals
at the sum and difference of the frequencies of the input signal and LO. The
LO frequency is chosen such that the difference component of the signal is at
an intermediate frequency (IF) which is then filtered by a bandpass filter with
adjustable bandwidth. By sweeping the frequency of the LO, the difference
signal from the mixer, including the information in the input signal, is swept
through the IF filter. This is known as the heterodyne principle. The SPAs
discussed here use a superheterodyne receiver.
A schematic diagram of a basic spectrum analyzer1,2 is shown in Fig. 1. The
input signal is passed through a lowpass (LP) filter which limits the frequency
range of the SPA. The input signal is then mixed with a variable frequency LO
to the IF filter. The envelope of the IF signal is detected and used to control
the vertical deflection of the cathode ray tube (CRT). The LO frequency is
swept through the frequency range of interest and is also used to control the
horizontal deflection of the CRT. In this way, the screen shows the signal power
as a function of frequency. The horizontal axis is also marked in units of time
as a reminder that the heterodyne process only measures a given frequency
component as it sweeps through that frequency. By the way, a generic radio
receiver uses the same heterodyne principle, except without the flexibility of a
SPA.
In single range SPAs, the IF is chosen to be above the highest frequency
of the SPA. For example, in Hewlett–Packard SPAs with frequency sensitivity
up to 2.9 GHz, the IF is chosen to be 3.6 GHz. In this case, the LO frequency
would need to range from 3.6 to 3.6+2.9 GHz to achieve the specified fre-
quency range. The bandwidth of the IF filter would determine the frequency
resolution of the measurement. Because narrow IF bandwidths are difficult to
achieve in the gigahertz range, most spectrum analyzers use several IF stages
to mix down to an IF where the narrowest filtering takes place. Shown in
Fig. multiheterodyne is a schematic showing several IF stages along with a
swept frequency LO at the first IF stage and fixed LOs at subsequent stages.
One useful feature of some SPAs is that the signal at one of the IF stages
(usually the 21.4 MHz) can be made available on the front or rear panel for
 input LP filter mixer IF BP filter
envelope detector
+

Vert
rf IF
LO - -
f f
Horz

voltage ramp
to VCO frequency (time)
voltage controlled LO
to horz defl of CRT
t

optional tracking generator

Figure 1: Simplified schematic of a superheterodyne spectrum analyzer.

Figure 2: Schematic of a multiple IF stage SPA.

use in some other application. Note that the phase information in the original
signal is still available in the IF signal and is only lost after passing through
the envelope detector.
An economical technique for SPAs to extend their frequency range is called
harmonic mixing. In this technique, harmonics of the LO are generated and
also mixed with the input signal. The principle is similar to that of the single-
range SPA discussed above except that several additional mixing products are
generated and care must be taken to not confuse them. One step in this di-
rection is that a much lower IF of 321.4 MHz is typically used. Some of the
disadvantages of harmonic mixing include decrease in the sensitivity due to an
increase in conversion loss in the mixer and an increase in phase noise of the
LO at higher harmonics. The decreased sensitivity of a wideband SPA with
 0 RBW=3 kHz, ST=0.1 sec
a)
-20 RBW=10 kHz, ST=0.03 sec
Amplitude (dBm)

RBW=1 kHz, ST=0.3 sec

-40
RBW=0.3 kHz, ST=3.33 sec
-60

-80

-100

0
b) ST=305 msec
ST=10 msec
Amplitude (dBm)

-20
ST=20 msec ST=5 msec
-40

-60

-80

-100
299.96 299.98 300.00 300.02 300.04
Frequency (MHz)

Figure 3: a) SPA measurement with variable resolution band width (RBW) and b) fixed
RBW and variable sweep time.

harmonic mixing is evident when viewing the full span with no input signal.
The noise floor appears to increase in steps at each harmonic band. It is use-
ful to point out some of the common misconceptions and limitations of SPAs.
Although the SPA purports to measure the frequency content of a signal, the
measurement is performed in the time domain (as are all measurements) and
thus is only an approximation of the actual frequency content. Furthermore,
because the SPA only measures the power at a given frequency, the phase
of a given frequency component relative to others is lost and thus the mea-
sured frequency spectrum is not sufficient for reconstructing a time domain
representation of the signal. Superheterodyne SPA measurements are also not
well suited for either single-shot signals or signals with low duty cycles since
the average signal power is very low. Discrete Fourier Transform methods on
digitally recorded data are more appropriate in this case.
2.2 Resolution, bandwidth, and sweep time
One of the primary specifications of a SPA is its ability to resolve signals of
different frequencies, known as the resolution bandwidth (RBW). This is de-
termined by the total bandwidth of the IF filters, usually determined by the
final stage filter. Consider a single frequency input to the SPA. As the mixing
product of the input signal and the LO is swept across the IF bandpass filter,
the shape of the filter mapped out. The measured signal is a convolution of
the width of the filter and the natural width of the signal. Clearly, the RBW
must be small compared with the natural width of the signal to observe it.
The general rule is that two equal amplitude signals can be resolved at a fre-
quency separation equal to the resolution bandwidth. One of the assumptions
in measuring the power at a given frequency as it is swept past the IF filter is
that the signal has reached steady-state. For example, consider the response
of the IF filter when it is subjected to a sinusoidal signal which is suddenly
turned on. The output signal will grow to a steady-state in a time approx-
imately equal to the inverse bandwidth of the IF filter. Narrow bandwidths
reach steady-state in a longer time than wider bandwidths. For the SPA to
accurately measure the signal amplitude, the signal must be swept past the IF
filter slowly compared to the response time of the filter. Shown in Fig. 3a is an
example of a 300 MHz sinusoidal input signal (the -10 dBm calibration signal
provided on many SPAs) measured with variable resolution bandwidths. The
sweep time is adjusted automatically to compensate for the changing RBW.
Fig. 3b shows the same signal with a fixed RBW but variable sweep times.
The optimum sweep time (ST) for a given RBW and frequency span can
be found by equating the time spent within the RBW for a given frequency
span with the risetime of the IF filter. The expression is
Span
ST = k (1)
RBW 2
where k is a constant relating the filter risetime to the resolution bandwidth
and depends in detail on the filter shape. For Gaussian filters it is about 2.5.
Usually, the RBW and sweep time are automatically adjusted according to the
frequency span such that a reasonable resolution is achieved. However, higher
resolution can usually be achieved but at the expense of longer sweep times.
Several SPAs offer the option of digitizing the IF signal to achieve even lower
RBWs. The video bandwidth (VBW) is another adjustable parameter on most
spectrum analyzers. It is a bandpass filter on the video signal observed on the
screen. Usually, the RBW and VBW are equal.
It is also interesting to note the variation of the noise floor of the signal
as the RBW is varied as is seen in Fig. 3a. For a constant background noise
signal density as a function of frequency, more noise power will pass through
the IF bandpass filter as its bandwidth is increased. Therefore, to observe low
level signals barely above the instrument noise, it is necessary to use narrow
RBWs, implying longer measurement times. Another source of noise inherent
in the SPA is amplitude and phase noise of the LO in the SPA receiver. In the
previous section, it was assumed that the LO was a pure sinusoid multiplied
with the input signal. Random phase or amplitude modulation of the LO can
lead to an apparent noise about a signal which may hide another signal.

2.3 Example measurement: Harmonic distortion of a nonlinear element

Many of the RF and microwave circuits used in accelerators are designed to
operate in their linear range. Therefore it is important to characterize the
distortion of a signal due to the nonlinear response of a circuit element. For a
weakly nonlinear element, the output voltage for a given input voltage can be
expanded in a power series
2 3
Vout = k0 + k1 Vin + k2 Vin + k3 Vin + ... (2)

where k0 is the DC response, k1 the linear gain, and kn the nth order non-
linearity. The expansion is usually truncated at third order. For a sinusoidal
input signal, the output signal contains products of the input with itself gen-
erating signals at harmonics of the input frequency. One common technique
for characterizing the nonlinearity is to use a two-tone input signal given by1

Vin = A1 sin ω1 t + A2 sin ω2 t (3)

The output voltage can be written as

Vout = c0 + c1 sin ω1 t + c2 sin ω2 t + c3 sin 2ω 1 t + c4 sin 2ω 2 t

+c5 sin 3ω 1 t + c6 sin 3ω 2 t
+c7 sin(ω1 + ω2 )t + c8 sin(ω1 − ω2 )t
+c9 sin(2ω 1 + ω2 )t + c10 sin(2ω 1 − ω2 )t
+c11 sin(2ω 2 + ω1 )t + c12 sin(2ω 2 − ω1 )t (4)

where the coefficients cn are determined by A1 and A2 . In addition to signals at

the input frequencies, the output contains signals at frequencies, |nω1 ± mω2 |,
where n and m are integers. This is commonly referred to as intermodulation
distortion. The amplitude of the intermodulation products are determined
by the amplitudes of the input signals, A1 and A2 , raised to the powers n
and m. The sum |n| + |m| is called the order of distortion. The order of
intermodulation distortion which is observed is always smaller for lower input
signals. Expressed in logarithmic terms, the second and third order terms
increase 2 and 3 dB per dB of increase in input signal.
In practice, the intermodulation distortion is usually specified to third
order. To characterize it, the two input tones are chosen close together in
frequency in the center of the frequency response of the device. Shown in
Fig. 4 is a spectrum of a low–level signal amplifier with a response from 1–30
MHz with an input of two signals at 0 and −30 dBm at frequencies of 10
and 11 MHz summed together. The signal harmonics are clearly seen. The
intermodulation products appear as sums and differences of harmonics of the
two input signals. Each signal is marked with the index n and m. Shown in
Fig. 5 is a plot of the amplitudes of the first, second and third order products
as a function of increase in input amplitude of each tone. Eventually the first
order signal stops increasing as the amplifier reaches its linear limit known as
gain compression. The point where extrapolation of the first order response to
a level where it intercepts with the second and third order response is known
and the second and third order intercept (SOI or TOI). The linearity of RF
and microwave components is often specified with the SOI and/or the TOI.
The SPA is well-suited for this type of measurement because it is designed
to receive signals at any frequency as opposed to the NWA discussed below
whose receiver is meant to respond at the excitation frequency.

3 Network analyzers

Network analyzers (NWAs) are used in such a variety of RF and microwave

measurements that they are roughly equivalent in utility to an oscilloscope
in an electronics laboratory. Their basic purpose is to measure the response
of a system to an applied sinusoidal input over a range of frequencies. The
capabilities of modern NWAs vary widely but essentially all consist of a signal
source and two input channels as shown in Fig. 6. In this simple schematic,
the NWA is configured to measure the transmission of a test signal through
the device under test (DUT) in channel A relative to the applied signal in
channel R. NWAs with the ability to measure the relative phase of the signals
are called vector NWAs (VNWAs). Scalar NWAs can measure only the relative
amplitude and are typically less expensive. SPAs equipped with a signal source
synchronized to the LO, called tracking generators, can be used to make scalar
transmission measurements as shown in Fig. 1. This section discusses the
principle of operation of vector NWAs and how they can be used to make
useful measurements.
Figure 6 shows a NWA configured to measure either the transmitted or
Figure 4: Spectrum analyzer measurement of intermodulation distortion in an amplifier. The
amplifier input is a sum of 10 and 11 MHz signals at levels of 0 and -30 dBm, respectively.
Intermodulation products are notated with indices m and n.

40
third-order intercept
20

0
Amplitude (dBm)

second-order intercept
-20

-40

-60
m=1
-80 m=1, n=1 harmonic
m=2, n=1 harmonic
-100
-20 0 20 40

Input signal amplitude (dB)

Figure 5: Plot of first, second and third order intermodulation products measured on an
amplifier illustrating the intercept points.
 incident transmitted
DUT

source reflected

R A B
Receiver Display

Figure 6: Schematic diagram of a NWA measuring the ratio of reflected and transmitted
signal with respect to the incident signal.

reflected signal from the DUT. A swept frequency signal source (R) is used to
generate a wave along a transmission line (coaxial cable or waveguide) which
is transmitted to the DUT. Some fraction of the incident wave is reflected,
transmitted, and attenuated or amplified by the DUT. From the ratio of the of
the reflected wave to the incident (A/R), quantites such as the standing wave
ratio (SWR), reflection coefficient, impedance and admittance, and return loss
of the DUT can be determined. From the ratio of the transmitted wave to
the incident (B/R), the gain or insertion loss, the transmission coefficient, the
insertion phase and group delay of the DUT can be found. Note that both the
reflected and transmitted ratios are complex quantities implying that both the
relative amplitude and phase are measured. Thus, it is possible to characterize
a 2-port device by measuring the transmitted and reflected components of a
signal incident to a device as a function of frequency.

3.1 NWA test sets and receivers

The directional sense of the forward and reflected signals is measured using
a test set. A test set consists of a combination of directional couplers and
switches configured as shown schematically in Fig. 7. A simple directional
coupler works by combining signals coupled from a transmission line such that
signals traveling one direction constructively interfere and signals in the oppo-
site direction destructively interfere. An example of such a test set is shown
in Fig. 7. Depending on the NWA model, the test set is either separate or
 input signal input signal

switch

R R
dir. coupler

dir. coupler
A B A B

DUT DUT

Transmission/reflection test set S-parameter test set

Figure 7: Schematic diagram of a NWA test set. Left) Setup for a reflection measurement
from port 1 or a transmission from prot 1 to 2. Right) Setup for a full two–port measurement.

integral to the NWA. The signal source itself can also be either separate or an
integral part of the NWA, depending on the level of precision required. For
example, the HP8510 series NWA uses a separate signal source which allows
the customer to choose the precision and stability of the source needed.
Shown in Fig. 8 is a schematic diagram of the receiver. The harmonic
mixing is similar to that described in the superheterodyne receiver for a SPA
except that it is applied to two channels at the same time. A further require-
ment is that the relative phase of the two signals be maintained. This is the
function of the phase locked loop on the LO shown on the reference channel.
The automatic gain control loop shown on the IF is to maintain optimal signal
levels for subsequent components. Because the receiver for a NWA is expected
to respond at only the frequency of the applied signal source, the receiver de-
sign is considerably less sophisticated than that of a SPA and typically has less
sensitivity than a SPA.

3.2 2-port networks

Network analyzer measurements are based on the concepts of 2-port networks4,5 .

Consider the 2-port shown in Fig. 9 with voltage V and current I at ports 1
and 2. Voltages at the ports can be written as a linear combination of the
Figure 8: Schematic diagram of a NWA measuring the ratio of reflected and transmitted
signal with respect to the incident signal.

currents I1 and I2 given by

V1 = z11 I1 + z12 I2
V2 = z21 I1 + z22 I2 (5)

where the Z matrix values are called impedance parameters. For example,
Z12 is the transfer impedance for I1 equal to zero. One can similarly relate
the currents I1 and I2 to the voltages via a matrix called the admittance
parameters.
In network analyzer measurements, a 2–port device is characterized in
terms of the transmission of forward and reflected waves rather than the volt-
ages and currents at the two ports as shown in the lower part of Fig. 9. The
reflected waves, VR , from ports 1 and 2 can be written as a linear combination
of the incident waves, VI , given by

VR1 = S11 VI1 + S12 VI2

VR2 = S21 VI1 + S22 VI2 (6)

The values of the matrix relating the reflected and forward waves are called
scattering or S–parameters. For example, with port 2 terminated by its char-
Figure 9: Upper) A 2-port network expressed in terms of the voltages and currents at the
ports. These can be related by impedance or admittance parameters. Lower) The same
network can be expressed in terms of forward and reflected voltage waves.

acteristic impedance, the reflection coefficient is given by

VR1 VR2
S11 = , S21 = (7)
VI1 VI1

To generalize the results, the incident and reflected voltages are usually
normalized by the root of the characteristic impedance of the system, Z0 . The
normalized reflected voltages are given by

b1 = S11 a1 + S12 a2
b2 = S21 a1 + S22 a2 (8)

where
VI1,2 VR1,2
a1,2 = √ , b1,2 = √ (9)
Z0 Z0
Note that the S–parameters are implicitly a function of frequency. From
the point of view of a NWA, a two–port device resembles the schematic shown
in Fig. 10. The lower half of the figure is called a signal flow diagram.
 S21 a1
incident
a1
b2

reflected, S22
DUT
reflected, S11

b1 a2
S12 incident

S21
a1 b2

S11 S22

b1 a2

S12

Figure 10: Upper) Schematic relationship of the S-parameters to the forward and reflected
signals. Lower) The same relationship expressed as a signal flow diagram. This is often
printed on the front panel of S-parameter test sets. Other parameterizations of the network
such as impedance and admittance can be related to the S-parameters.
 Ex

a1 S 21a E tt b1

E df E sf S 11a S 22a El

S 12a a2
b2 E rt

port 1 port 2

Figure 11: Signal flow diagram for a forward S-parameter measurement with error terms.
Esf =forward directivity, Esf =source match, Erf =reflection tracking, Ett =transmission
tracking, El =load match, Ex =isolation. Reverse flow has a similar diagram.

3.3 NWA calibration

One of the necessities in making accurate NWA measurements is calibration.
Because of imperfections in the test set, i.e. the directional couplers, attenua-
tors, large measurement errors can arise. To minimize the effect of these im-
perfections, it is possible to measure the response of standards with well known
responses and thus normalize subsequent measurements from the effects of the
test set. Thus the calibration removes a large part of the systematic errors of
the measurement but does not affect the random errors.
Consider the signal flow diagram of a reflection measurement with errors
as shown in Fig. 11. The error terms Erf , Edf , and Esf represent the imper-
fections in forward transmission, forward directivity, and the forward source,
respectively. The measured reflection coefficient is related to the actual by

S11A Erf
S11M = Edf + (10)
1 − Esf S11A

where S11A and S11M are the actual and measured reflection coefficients.
Clearly the error terms can have a large influence on the measurement.
By measuring the reflection coefficient of three known loads, the values of
Erf , Edf , and Esf can be solved at each frequency over which the measurement
is made and their effect can be removed from subsequent measurements of a
DUT. For example, shown in Fig. 12 is a measurement of S11 for a standard
50 Ω termination with and without a calibration. For an ideal termination, no
reflection should be present. The effect of the calibration in this case improves
the measurement by 20 dB.
 -10
50 S11 w/o calibration
-20 50 S11 w/calibration
Amplitude (dB)

-30

-40

-50

-60

-70
2 4 6 8 10
Frequency (GHz)

Figure 12: S11 measurement of a 50 Ω termination with and without calibration. The
calibration provides and 20 dB improvement over this frequency band.

For a full 2-port measurement, the error model must be expanded to in-
clude errors at the receiving port, requiring calibration of each port using 3
known loads in reflection as well as a standard in transmission.
For measurements of devices with standard connectors, calibration stan-
dards such as a termination, open and short circuit are available. It is necessary
to either load the response of the standards into the NWA or internally spec-
ify which calibration standards are being used. In current NWA models, it is
important to remember that the calibration is valid only over the frequency
range over which the calibration was performed. Although the calibration pro-
cess is tedious, it is fairly well-automated and only requires the connecting of
standards as indicated by the analyzer. Future NWA models may include in-
ternal standards which are automatically switched to perform fully automated
calibration.

3.4 Example NWA measurements

RF cavities are one of the accelerator devices most commonly characterized
using NWAs. Among the desired properties to measure are the frequencies and
quality factors (Q) of the resonant modes, particularly the fundamental mode,
and the match of the coupler to the fundamental mode. The most common
method for measuring the mode frequency and Q is to measure transmission
through the cavity (S21) using probes lightly coupled to the cavity. Shown
in Fig. 13a is an example of such a measurement on the fundamental mode
of a pillbox cavity. The Q can usually be determined automatically by the
 a)
3
S21 Amplitude (dB)

-30

Phase (rad)
2

-40 1

0
-50
4
b)

0
2
S11 Amplitude (dB)

Phase (rad)
0
-5

-2

-10
-4
0.7830 0.7835 0.7840 0.7845
Frequency (GHz)

Figure 13: NWA measurements on the fundamental mode of a pillbox cavity. a) Transmission
(S21 ) measurement to the determine the resonant frequency and quality factor. b) Reflection
(S11) measurement to measure the coupling to the cavity.

NWA by finding the peak width at the 3 dB points (i.e. half power points).
The match of the coupler to fundamental mode can be measured by measuring
the reflection of the cavity on resonance. The coupler is the means by which
external power is injected into the cavity and usually takes the form of a
waveguide coupled to the cavity with either a loop or an aperture. If the
cavity is uncoupled to the external source, all of the input power is reflected.
If the cavity presents a perfect match to the external source, no power is
reflected. Shown in Fig. 13b is an example of a reflection (S11) measurement
of the fundamental mode via an input loop coupler. Off resonance, the power
is reflected with either a ±180 phase shift, indicating an inductive or capacitive
load. It is particularly important to calibrate the reflection measurement to
get an accurate measurement of the cavity coupling.
There are numerous RF and microwave devices in any accelerator sys-
tem and they are usually characterized using a NWA. One of the simplest
components is a coaxial cable. Because of the skin effect on the conductors,
characteristics of the coaxial line such as signal loss, and group velocity vary as
a function of frequency, leading to a dispersive distorion of a signal. Although
this may seem like a trivial concern, accelerator signals sometimes must be
cleanly transmitted over distances exceeding kilometers, requiring innovative
designs.
Shown in Fig. 14 is a transmission measurement of a 30 meter length of
RG-233/U coaxial cable (i.e. standard BNC cable). The increase in signal loss
at higher frequencies is evident as well as a nonlinear slope of the phase as a
function of frequency. Most of the effect of long time delay in the cable has been
artificially removed in the NWA using a feature called a port extension. For
example, a delay, ∆T , in the time domain creates a phase shift in the frequency
domain of ω∆T . This corresponds to linear variation of the phase as a function
of frequency. Because the phase shift due to the time delay is constant, it can
be removed numerically. This is the function of a port extension, which acts as
if one the ports were physically extending closer to the other port, effectively
removing the time delay. For example, the phase response shown in Fig. 14
has an added port extension of 150 nsec, almost exactly equal to the electrical
delay of the cable at low frequency. Without the port extension, the phase
would change by hundreds of radians over a 1 GHz bandwidth. Note that
most NWAs have a feature to measure the delay directly from the slope of the
phase response. Shown in Fig. 15 is another example of a common microwave
device: a bandpass filter. Among the important characteristics to measure
are the bandwith, frequency rolloff above and below the bandpass, and the
insertion loss.

3.5 Time domain analysis

A useful feature that has been added recently to several VNWA models is the
ability to transform the measured frequency response of a device into the time
domain displayed as either the impulse or step response. This feature simulates
the measurements usually made using time domain reflectometry (TDR). Note
that the synthetic pulse technique is only applicable to linear time invariant
networks. Otherwise a real pulse approach is needed.
Although the time domain display of the measurement data may be of only
passing interest, the most useful feature is the ability to filter or gate the data
in the time domain and subsequently transform back to the frequency domain.
For example, the effect of imperfect connectors on a measurement is not clearly
distinguished from the response of the device itself when viewing the measure-
ment in the frequency domain. However, by transforming the measurement to
the time domain where the reflection at the imperfect connector is evident, it is
 0
Amplitude (dB)

-5

-10

-15

0.0

-0.5

-1.0
Phase (rad)

-1.5

-2.0

-2.5

-3.0
0.2 0.4 0.6 0.8 1.0
Frequency (GHz)

Figure 14: Transmission measurement of a 30 m standard BNC cable.

0 amplitude 4
phase
2
Amplitude (dB)

-20
Phase (rad)

0
-40
-2
-60
-4

2.0 2.5 3.0 3.5 4.0

Frequency (GHz)

Figure 15: Transmission (S21 measurement of a 2.7–3.3 GHz bandpass filter.

possible to remove the effect of the connector by gating out the reflection and
transforming back to the frequency domain. This feature is particular useful
in bench impedance measurements where the mechanical arrangement of the
connection to the device often leads to step impedance mismatches which tend
to obscure the effect of the device.
Shown in Fig. 16 are the same NWA measurements shown in Figs. 13, 14, 15
tranformed to the time domain. Fig. 16a shows the simulated impulse response
of the fundamental mode of an RF cavity. The high frequency component is
at the cavity resonant frequency and the slow exponential decay of the am-
plitude is related to the relatively high cavity Q. Fig. 16b shows the impulse
response of the bandpass filter. The ringing signal is at 3 GHz, the filter center
frequency, and the inverse of the width in time of the main tone burst corre-
sponds to the frequency bandwidth of the filter. Fig. 16c shows the impulse
response of a long cable, where the long delay of the cable has been removed
using a NWA port extension. An ideal lossless cable would have an delta func-
tion impulse response. The real dispersion of a cable spreads the frequency
components of the signal in time resulting in a reponse as shown.

4 Acknowledgements
The authors would like to thank the Hewlett-Packard Company for providing
the equipment for use in the microwave measurements laboratory and Stuart
Turner and Mel Month for encouraging completion of this manuscript.

References
1. R. White, Spectrum and Network Measurements, PTR Prentice Hall,
New Jersey (1993).
2. B. Peterson, Spectrum Analysis Basics, Hewlett–Packard Application
Note 154 (1989).
3. S–Parameter Design, Hewlett–Packard Application Note 154 (1989).
4. G. H. Bryant, Principles of Microwave Measurements, IEE, London,
(1993).
5. F. Caspers, Proc. CERN Accelerator School, RF Engineering for Particle
Accelerators, Oxford, UK (1991).
 1.0 a)
Linear Amplitude (a.u.)

0.5

0.0

-0.5

-1.0

0 200 400 600 800 1000

b)
0.5
Linear Amplitude (a.u.)

0.0

-0.5

0 2 4 6 8 10

20 c)
Linear Amplitude (a.u.)

15

10

0
0.0 0.2 0.4 0.6 0.8 1.0
Time (nsec)

Figure 16: Examples of synthetic time domain on measured on an HP8510 NWA. a) Impulse
response of the fundamental mode. b) Impulse response of a bandpass filter. c) Impulse
response of a long cable with a port extension set to remove the long time delay.

View publication stats