Introduction

Background
Rapid recent progress in development of wireless and broadcasting technologies, such as
smartphones, wireless LAN, wireless sensor networks, RFID, GPS, digital TV, etc., has seen their
widespread adoption in most people’s daily lives.

A variety of measuring instruments, such as frequency counters, field strength meters, power
meters, etc., is used to measure, analyze, and evaluate RF signals including electromagnetic
(radio) waves broadcast by antennas; the spectrum analyzer plays a central role in detailed
measurement, analysis, and evaluation of RF signals. The importance of the spectrum analyzer
for engineers dealing with RF signals seems unlikely to change in the future.

At the same time, progress in wireless communications technologies and the appearance of
new applications have increased the complexity and level of the functions and performance
required by spectrum analyzers. Among these requirements, the need to evaluate signals
using new wideband digital modulation methods as well as to capture transient signal
spectrums to troubleshoot problems with electromagnetic noise resulting from use of high
sensitivity parts mounted at high densities requires new generations of spectrum analyzers
using frequency sweep principles.

As a result, the spectrum analyzer continues to evolve to meet the needs of engineers. As an
example, the digitization of internal processing has led to huge jumps in measurement speed,
accuracy and stability.

As devices and computation speeds

have become increasingly faster and
more accurate, Anritsu has been
incorporating developments in digital
processing as signal analyzer functions
into its MS2830A and MS2690A series.

Such improvements not only support

evaluation and analysis of digital
modulation signals, but also support
capture of various signals with non-
regular spectrums.

Figure 1

4 Understanding Spectrum & Signal Analysis

Engineers and technicians involved in modern RF or microwave communications have many
measuring instruments at their disposal, each designed for specific measurement tasks.
Among those available are:

a) The Oscilloscope – primarily developed for measuring and analyzing signal amplitudes in
the time domain. (Voltage vs. time) Often 2, 4 or more channels of voltage vs. time can be
viewed on the same display to show the relationships between signals. Extensive methods
to trigger signals are often available to capture and display rare events.

b) 
The Spectrum Analyzer – designed to measure the frequency and amplitude of
electromagnetic signals in the frequency domain. (Amplitude vs. Frequency & Amplitude
vs. time) Most modern analyzers also have the capability to demodulate analog modulated
signals. Spectrum analyzers are the most versatile tools available to the RF engineer. This
guide will describe the critical performance characteristics of spectrum and signal analyzers,
the types of signals measured, and the measurements performed

c) The Signal Analyzer – invaluable for measuring the modulation characteristics of complex
signals. These units capture and process blocks of spectrum to reveal amplitude and phase
relationships between signals. Newer models provide demodulation of digitally modulated
signals used in most of today’s communications systems.

d) 
The Signal Generator – an essential item of equipment for any communications test
laboratory or workshop. The cost of a signal generator largely depends on the additional
functions and facilities available as well as the type and quality of the frequency reference
used.

e) 
The Field Strength Meter (F.S.M.) – display the power density of an electrical signal
incident on a calibrated antenna and thus give a direct reading of field strength in dBµV/m.

f) The Frequency Counter – a digitally based instrument that measures and displays the
frequency of incoming signals. Some models can also count ‘pulse’ and ‘burst’ signals.

w w w. a n r i t s u . c o m 5
Frequency Domain / Time Domain
As mentioned in the introduction, electromagnetic signals can be displayed either in the time
domain, by an oscilloscope, or in the frequency domain using a spectrum or signal analyzer.
Traditionally, the time domain is used to recover the relative timing and phase information
required to characterize electrical circuit behavior. Many circuit elements such as amplifiers,
modulators, filters, mixers and oscillators are better characterized by their frequency response
information. This frequency information is best obtained by analysis in the frequency domain.
Modern Oscilloscopes provide frequency domain display modes and modern Spectrum and
Signal analyzers provide time domain displays. One key difference between oscilloscopes and
spectrum/signal analyzers is the resolution of the vertical axis. Oscilloscopes provide high
resolution along the time axis but low (8 bit) amplitude resolution. Spectrum and signal
analyzers provide high (16 bit or more) amplitude resolution to see small signals in the presence
of large signals.

In order to visualize these ‘domains’ refer to Figure 2.

f
t

Figure 2

This represents an electromagnetic signal as a 3 dimensional model using:

(i) a time axis (t)

(ii) a frequency axis (f) and
(iii) an amplitude axis (a)

Observing from position X produces an amplitude time display where the resultant trace is the
sum of the amplitudes of each signal present. This time domain view facilitates analysis of
complex signals, but provides no information on the individual signal components (Figure 3).

6 Understanding Spectrum & Signal Analysis

 1.5

1.0

0.5

Amplitude 0.0 Time

–0.5

–1.0

–1.5

Figure 3

Viewing the model in Figure 2 from position Y, however, produces an amplitude vs. frequency
display showing each component of the signal in the complex waveform. Observation in this
frequency domain permits a quantitative measurement of the frequency response, spurious
components and distortion of circuit elements (Figure 4).

Figure 4

w w w. a n r i t s u . c o m 7
Spectrum Analyzers
A Spectrum Analyzer is a swept tuned analyzer. It is tuned by electronically sweeping its input
over the desired frequency range thus, the frequency components of a signal are sampled
sequentially in time (Figure 5). Using a swept tuned system enables periodic and random
signals to be displayed but does not allow for transient responses.

RF Input Log Envelope Video

Attenuator Mixer IF Gain IF Filter Amp Detector Filter

Input
ADC
Signal

Pre-selector, or
Low-pass Filter Step
Synthesizer Vert

Reference
Oscillator
Horiz
CPU

Display

Figure 5

8 Understanding Spectrum & Signal Analysis

Signal Analyzers
Signal analyzers sample a range of frequencies simultaneously, thus preserving the time
dependency and phase between signals. The signal analyzer makes use of high-level digital
processing technologies, such as high-speed A/D conversion, FFT processing, digital filters,
etc., to capture a spectrum that changes over time, such as for a burst signal, as well as a
spectrum with transient variations. Signal Analyzers and Spectrum analyzers have very similar
RF block diagrams, differing in frequency range (bandwidth) of the IF processing. The high
bandwidth processing offers many advantages, but at increased cost.

System
To make best use of the signal analyzer functions and performance, it is important to understand
how the signal analyzer captures the spectrum.

Figure 6 shows the signal processing flow and operation principles of a signal analyzer.

Signal analyzers capture the spectrum results using three main processes: frequency conversion,
digitization and storage, and conversion to the spectrum.

Figure 6: Signal analyzer Principle and Operation

First, the measurement signal input in a spectrum analyzer is converted to an intermediate

frequency (IF) by the frequency conversion section composed of a local oscillator, mixer and
bandpass filter.

The intermediate frequency Fif is found from the following equation where Fin is the
measured signal frequency and Flo is the local oscillator oscillation frequency.

Fif = Fin – Flo (Eq 3-1)

In a signal analyzer, the value of Flo is fixed during measurement. In addition, the IF has a wide
bandwidth corresponding to the measurement bandwidth (frequency span). As a result, the
measured signal spectrum is frequency-converted to IF without changing the shape at Fin.

w w w. a n r i t s u . c o m 9
Next, the IF-converted measured signal is converted to digital data using an A/D converter.

In other words, the signal analyzer captures the true form of the measured signal in a fixed time
period just by frequency conversion. The digitized time series waveform data is immediately
captured to the internal memory and this data can be saved to another hard disk.

Any part or range of the measured signal captured as time-series waveform data can be read
immediately and analyzed using digital processing. Deploying digital processing using Fast
Fourier Transform (FFT) in the frequency domain captures the spectrum in the read time range.

The data stored in memory or hard disk is the basis of the signal analyzer display. In other
words, the signal capture processing that stores the signal as digital data, and the analysis
processing that reads the data can be executed independently time wise. Since capture and
analysis are batch processed, either offline analysis can be performed in the free time after
capture, or the same signal can be analyzed repeatedly using different methods and settings.
This is a key feature of signal analyzers that is not supported by sweep-type spectrum analyzers.

For example, the measurement results can be confirmed using saved data by changing to a
different RBW at measurement.

Figure 7: Sequential display of wireless signal spectrum display captured in memory

10 Understanding Spectrum & Signal Analysis

Operation of each part

Input and frequency conversion

In an actual analyzer, an attenuator (signal attenuator) and preamplifier in the input section
assure the correct measured signal level.

At frequency conversion, although it is possible to obtain the IF signal using a single conversion,
in measuring instruments, the final IF signal is obtained using multiple frequency conversions
rather than a single conversion. In addition, if the frequency of the measured signal is higher
than the measured signal frequency, sometimes a frequency converter is used at the input
stage. The signal path until frequency conversion is the same in a signal analyzer and spectrum
analyzer and the items to consider are the same.

Noise and non-linearity generated in the analog section from input to IF conversion is added
to the measured value and is part of an analyzer’s basic performance. Consequently, high-level
analog circuit technology is essential when handling very small high-frequency signals.

As a result, in the MS2830A for example, a very low displayed average noise level of –150.5
dBm/Hz is assured to permit wide dynamic range measurements.

Signal Analyzer MS2830A specifications:

Displayed Average Noise Level (DANL): –150.5 dBm/Hz (Signal analyzer mode, 30 MHz ≤
Frequency < 1 GHz, Pre-amp: Off)

Local oscillator
The local signal oscillator handles frequency modulation of the measured signal. In a spectrum
analyzer, the local oscillator oscillation frequency is swept (changed continuously) at
measurement, but in a signal analyzer it oscillates at a fixed frequency during measurement.

In the same way that noise and non-linearity generated in the analog section from input to IF
conversion form the basis of an analyzer’s basic performance, the signal purity of the local
oscillator is directly related to basic performance. For example, if there is even a slight variation
in the oscillation frequency, the measured signal will appear to fluctuate at that moment too.

Similarly, short-term variations will cause Single Side Band (SSB) phase noise and widespread
errors in the spectrum tails.

Consequently, the local oscillator section of a signal analyzer incorporates a high-purity

oscillator.

w w w. a n r i t s u . c o m 11
Figure 8 shows an example of signal analyzer SSB phase noise characteristics (MS2830A
Option-066).

Figure 8: MS2830A Option - 066 SSB phase noise characteristics

IF circuit
The IF-converted measured signal is band-limited by the bandpass filter while being
simultaneously amplified to the level required for analysis. The signal analyzer IF requires a
frequency bandwidth that is equal to or greater than the analysis frequency width (span).

Consequently, the signal analyzer IF circuit is designed to achieve flat characteristics over a
wide band of about 30 MHz normally, and sometimes exceeding 100 MHz, depending on the
situation.

A/D converter
The frequency-converted IF (analog) measured signal is converted to a digital (time-series
data) signal by the A/D converter. At A/D conversion, the signal analyzer IF requires high-
speed conversion from the wideband signal.

In addition, supporting wide-dynamic-range analysis requires high-resolution A/D conversion.

Consequently, the signal analyzer has a built-in high-speed, high-resolution A/D converter.

Digital processing
The digitized measured signal is saved to memory. In addition, it can be either saved to internal
hard disk or transferred to an external storage device for later loading and replay.

At this time, the read data is converted to the necessary narrow range by FFT and repeated
data loading and FFT creates a spectrum series changing with time which is displayed on-
screen to give the impression of a spectrum changing dynamically in real time (Figure 9).

12 Understanding Spectrum & Signal Analysis

 Figure 9: Continuous analysis at short cycle

Moreover, as shown in Figure 10, after saving long-term data, loading the data by slightly
shifting the analysis time range and performing FFT makes it possible to sequentially inspect
spectrum transitions with time at any point and any speed.

Additionally, besides performing FFT on read data to obtain the spectrum, it is also possible
to obtain other information such as changes in power with time by performing processing on
read data using other parameters.

Figure10: Analysis of specific signal range recorded in memory

w w w. a n r i t s u . c o m 13
Sampling rate and sample count
The A/D-converted digital IF signal is resolved into the I-Q components (In-phase/Quadrature-
phase) after waveform correction, etc., and is stored in memory following various procedures
such as decimation, floating point data processing, etc.

Rather than leaving the data as a time-series waveform (scalar quantity), resolving and saving
as I-Q data maintains the signal phase data as a vector quantity, which can assure the flexibility
for post-processing, such as digital modulation analysis, etc.

The cumulative time range (capture time) can be saved to memory as data counts (samples)
allocated by capture period (sampling rate).

In an actual signal analyzer, the standard memory capacity is 1 Gbyte, supporting a maximum
data storage capacity of about 100 Msamples per measurement.

The sampling rate is limited by the frequency span at FFT processing, but in a signal analyzer
the sampling rate is selected automatically by setting the analysis frequency span.

Consequently, there is no necessity to be concerned about sampling rate at usage. The max.
capture time and max. sample counts are decided at the same time as the sampling rate, but
usually rather than capturing the max. sample count, the optimum sample count is set for the
measurement conditions.

Table 1 lists the sampling rates for the set frequency span and the relationship between the
max. capture time and max. sample count.

Max. sample
Frequency span Sampling rate Max. capture time
count
1 kHz 2 kHz 2000 s 4M
10 kHz 20 kHz 2000 s 40 M
100 kHz 200 kHz 500 s 100 M
1 MHz 2 MHz 50 s 100 M
10 MHz 20 MHz 5s 100 M
31.25 MHz 50 MHz 2s 100 M

Table 1: Frequency span, Sampling rate, Max. capture time and Max. sample count (Extract
for MS2830A)

14 Understanding Spectrum & Signal Analysis

FFT (Fast Fourier Transform)
Any time range of the I-Q data saved in memory can be read. The read data is FFT processed
to generate the spectrum for the relevant time range. Similarly, processing with other
parameters can also generate other signal displays, such as Power vs. Time, etc.

The spectrum obtained by FFT processing is a sequence of amplitude values for discrete
frequency points. In addition, the frequency span obtained by FFT depends on the sampling
frequency while the frequency gap is inversely proportional to the number of data points
(sample count). Consequently, the data used for FFT is digitized at the sampling rate that
satisfies the frequency span for spectrum display while also assuring a sample count supporting
the frequency resolution. Put another way, the data for FFT must be digitized as the “sampling
frequency” and “sample count” that achieves the expected “frequency span (frequency
range)” and “resolution” at spectrum display. Incidentally, in a spectrum analyzer, the frequency
resolution is evaluated as the Resolution Bandwidth (RBW); the time required for measurement,
or the sweep time, is determined by the RBW and frequency span.

A signal analyzer emulates the usability of a spectrum analyzer and uses the same concept of
RBW as a spectrum analyzer to improve the consistency of data measured using a spectrum
analyzer. To start with, the sampling rate and data length required for FFT processing
supporting the frequency span and RBW settings are set automatically.

Additionally, the RBW corresponding to the frequency span can be set automatically. At this
time, in the same way as a spectrum analyzer, the measurement time increases because the
number of calculations increases as the RBW becomes smaller (frequency resolution increases).

RBW
1 kHz 3 kHz 10 kHz 30 kHz 100 kHz
Span
31.25 MHz 262144 65536 32768 8192 2048
10 MHz 131072 32768 8192 4096 2048
1 MHz 8192 4096 2048 2048 2048

MS2830A Spectrum trace marker result: Integration

Table 2: Frequency span, RBW and Data points (Extract)

FFT processes the time range to be analyzed as one period and performs the same continuous
repetitions on the signals before and after (the time range).

As a result, a difference in the values near both sides of the analysis time range causes an error
in which the spectrum appears to widen, etc. To prevent this, the signal analyzer performs
window function processing using a Gaussian window, but this requires there to be data before
and after the analysis time range (Figure 11).

w w w. a n r i t s u . c o m 15
 Figure 11: Required data range at analysis

As a result, the data length used for the analysis calculations (I-Q data length) becomes longer
than the target analysis time (width). This also happens with non-FFT analysis (trace) modes,
For example, Power vs. Time plots require data before and after the analysis time length to
perform processing for the detection and moving average, and for the spectrogram window
function and detection.

The length of the required before and after data differs with the trace type and the marker
measurement result read accuracy setting (Marker result). Usually, the minimum data length
required to capture the results as quickly as possible is set automatically (Capture time: Auto).

Obviously, analysis can be performed using the maximum memory by setting any time for the
captured data length (Capture time: Manual).

In either case, the analysis range and the data range used for calculation are expressed as
color bands on the time axis (Figure 12).

Figure 12: Data range display

16 Understanding Spectrum & Signal Analysis

Traces
The signal analyzer digitizes the signal and then uses FFT to capture the instantaneous
spectrum, but a spectrum is not the only method for displaying a changing signal. A signal
analyzer also has several trace display methods using captured data to measure and evaluate
3D signals from various viewpoints.

Spectrum trace
The Spectrum trace displays the basic spectrum
with frequency on the horizontal axis and power
on the vertical axis.

Figure 13: Example of spectrum

trace (W-CDMA signal
Power vs. Time trace
The Power vs. Time trace is a method for
displaying and observing changes in signal
power with time. It is good for understanding
signals that switch On/Off with time and
amplitude variations.

The horizontal axis is time and the vertical axis is

power.
Figure 14: Example of Power vs.
Time trace (ARIB T61 Emergency
radio burst signal

Frequency vs. Time trace

The Frequency vs. Time trace can be used to
grasp signal frequency drift and sudden
frequency fluctuations. The horizontal axis is time
and the vertical axis is frequency.

Figure 15: Example of Frequency

vs. Time trace (Frequency
deviation of FM waveform)

w w w. a n r i t s u . c o m 17
Phase vs. Time trace
The Phase vs. Time trace is good for analysis of
signal phase time stability and phase modulation,
etc. The horizontal axis is time and the vertical
axis is phase.

Main Traces
Figure 16: Example of Phase vs.
Time trace (Phase characteristics of
1.9 GHz band unmodulated signal)

CCDF trace
The CCDF trace is useful for comparing the
digital modulation signal peak and average as
well as for confirming the amplitude distribution
of non-periodic noise in a signal. The horizontal
axis is amplitude and the vertical axis indicates
the probability that the signal is at each
amplitude. The probability is displayed as CCDF
or PDF.
Figure 17: Example of CCDF trace
CCDF (Complementary Cumulative Distribution (CCDF of terrestrial digital TV
Function): signal)
Cumulative distribution of instantaneous power
deviation vs. average power

APD (Amplitude Probability Distribution):

Probability distribution of instantaneous power

vs. average power

18 Understanding Spectrum & Signal Analysis

Spectrogram trace
The Spectrogram trace is the equivalent of looking down from the top of the power axis at the
power, frequency and time data represented in 3D (Figure 19) to get a good overall grasp of
the signal spectrum and power changes.

The horizontal axis is time and the vertical axis is frequency.

Figure 18: Example of Spectrogram trace

(Frequency hopping of Bluetooth signal)

The power level is color-coded.

Figure 19: Signal observation direction

w w w. a n r i t s u . c o m 19
Sub-traces (Split screen display)
This function displays a sub-trace under the
main trace as a supplement to the normal main
trace.

The sub-trace can be selected to display either

the Power vs. Time or Spectrogram trace with
the trace displaying data for any time period.
When the sub-trace is displayed, the main trace
results are conveniently displayed immediately
above for the corresponding point in time. Figure 20: Example of split screen
Additionally, setting the analysis range at the display (Main trace: Spectrum, Sub-trace:
sub-trace and slowly shifting the analysis start Power vs. Timel)
point slightly using the main dial displays a time-
lapse animation of the spectrum on the main
trace.

Basic Operation
Both spectrum analyzers and signal analyzers are based on a super heterodyne receiver
principle (Figure 21). The input signal, fIN, is converted to an intermediate frequency, fIF, via a
mixer and a tunable local oscillator fLO. When the frequency difference between the input
signal and the local oscillator is equal to the intermediate frequency then there is a response
on the display.

1st Mixer B.P.F 2nd Mixer 3rd Mixer

Input
Signal

Step Vert
Synthesizer

Reference
Oscillator Horiz
CPU

Display

Figure 21

20 Understanding Spectrum & Signal Analysis

 fIN = fLO ± fIF
This is the basic tuning equation that determines the frequency range of a spectrum/signal
analyzer. Using the super heterodyne technique enables high sensitivity through the use of
intermediate frequency (IF) amplifiers and extended frequency range by using the harmonics
of the local oscillator (LO). This technique is not, however, real time and sweep rates must be
consistent with the IF filter bandwidth charge time.

Characteristics
Spectrum and signal analyzers have the following characteristics:
• Wide frequency range.
• Amplitude and frequency calibration via internal calibration source and error
correction routines.
• Flat frequency response where amplitude is independent of frequency.
• Good frequency stability using synthesized local oscillators and reference source.
• Low internal distortion.
• Good frequency resolution.
• High amplitude sensitivity.
• Linear and logarithmic display modes for amplitude (voltage and dB scaling).
• Absolute and relative measurement capabilities.

Frequency Range
The lower frequency limit of a spectrum analyzer is determined by the sideband noise of the
local oscillator. The local oscillator feedthrough occurs even when there is no input signal
present.

The sensitivity at the lower frequency is also limited by the LO. sideband noise. Figure 22
shows typical data of average noise level vs. frequency for two IF bandwidths.

10 kHz RBW

1 kHz RBW

Figure 22

w w w. a n r i t s u . c o m 21
It should be noted however, that as the IF bandwidth is reduced so the time to sweep a given
frequency range increases since the charge time of the IF filter increases. This means that the
sweep time is increased to allow the IF filter to respond and therefore present an undistorted
signal to the detector. These variables are generally taken into account automatically and are
referred to as ‘coupling’. Beyond the detector can be more filtering known as Video Bandwidth
and this can also be coupled to IF bandwidth and sweep time. These functions are coupled
together since they are all inter dependent on each other, i.e. change one parameter setting
and it affects the others.

An additional facility available on most modern analyzers is a Zero Frequency Span mode. As
mentioned earlier, most analyzers are based on the super heterodyne receiver design, where
the local oscillator is swept continuously. If the local oscillator is manually tuned, the spectrum
analyzer becomes a fixed tuned receiver whose frequency is determined by that of the local
oscillator. In this mode the analyzer will display the time domain function since the frequency
component is fixed even though the scan generator is still sweeping the display i.e. the display
is now amplitude vs. time (Figure 23).

Figure 23

22 Understanding Spectrum & Signal Analysis

Frequency Resolution
The frequency resolution (typically called “resolution bandwidth”) of a spectrum/signal
analyzer is its ability to separate and measure two signals in close proximity. This frequency
resolution is determined by three primary factors:
• the IF filter bandwidth used
• the shape of the IF filter and
• the sideband noise of the IF filter
The IF bandwidth is normally specified by ∆f at 3 dB (Figure 24). From this it can be seen that
the narrower the filter bandwidth the greater the frequency resolution. However, as mentioned
earlier, as the IF band width is reduced so the charge time for the filter increases hence
increasing the sweep time. As an ex ample, narrow IF bandwidths are required to distinguish
the sidebands of amplitude and frequency modulated signals (Figure 25).

Δf @ –3 dB

Figure 24

Figure 25

w w w. a n r i t s u . c o m 23
When measuring close in spurious components, the shape of the IF filter becomes important.
The filter skirt inclination is determined by the ratio of the filter bandwidth at –60 dB to that at
–3 dB (Figure 26).
Δf @ –3 dB

Δf @ –60 dB

Figure 26

This skirt inclination is known as the ‘shape factor’ of the filter and provides a convenient guide
to the filter quality. The most common type of IF filter is known as the Gaussian filter, since its
shape can be de rived from the Gaussian function of distribution. Typical shape factor values
for Gaussian filters are 12:1/ 60 dB:3 dB, while some spectrum analyzers utilize digital filters
where the shape factor can be as low as 3:1. Digital filters appear to be better in terms of
frequency resolution, but they do have the drawback of sharply increasing the scan time
required to sweep a given frequency range. Figure 27 shows the effects of scanning too fast for
a given IF bandwidth filter. As the scan time decreases, the displayed amplitude decreases and
the apparent bandwidth increases. Consequently, frequency resolution and amplitude
uncertainty get worse, and some analyzers will warn you that you are now in an ‘UNCAL’ mode.

Figure 27

24 Understanding Spectrum & Signal Analysis

A spectrum analyzer’s ability to resolve two closely spaced signals of unequal amplitude is not
only dependent on the IF filter shape factor. Noise sidebands can reduce the resolution
capabilities since they will appear above the skirt of the filter and so reduce the out of band
rejection of the filter.

Sweep Speed
Spectrum analyzers, incorporating swept local oscillators have the issue of needing to manage
the sweep speed to prevent uncalibrated displays. Signal analyzers do not. Blocks of spectrum
are processed together in a signal analyzer. See Figure 6. The sample rate of the A to D
converter determines the span of spectrum that can be processed. The span is approximately
½ the A to D sample rate. The difference in sweep speed performance between the spectrum
analyzer mode and signal analyzer mode is especially visible when very narrow resolution
bandwidths are used. Most modern analyzers combine swept spectrum and signal analyzer
technology. The spectrum analyzer mode offers very wide span views and the signal analyzer
mode offers fast spectrum displays for narrow spans. The sweep speed for signal analyzer-
based spectrum displays depends on the FFT computation speed. Dedicated FFT processing
circuitry can speed up spectrum display rates to support searching for intermittent signals.

Sensitivity and Noise Figure

The sensitivity of a spectrum analyzer is defined as its ability to detect signals of low amplitude.
The maximum sensitivity of the analyzer is limited by the noise generated internally. This noise
consists of thermal (or Johnson) and non-thermal noise. Thermal noise power is expressed by
the following equation:
PN = kTB
where

PN = Noise power (in Watts)

k = Boltzman’s constant (1.38 x 1023 JK–1)

T = Absolute temperature (Kelvin)

B = System Bandwidth (Hz)

From this equation it can be seen that the noise level is directly proportional to the system
bandwidth. Therefore, by decreasing the bandwidth by an order of 10 dB the system noise
floor is also decreased by 10 dB (Figure 28).

w w w. a n r i t s u . c o m 25
 10 kHz RBW

1 kHz RBW

Figure 28

When comparing spectrum analyzer specifications it is important that sensitivity is compared

for equal bandwidths since noise varies with bandwidth.

An alternative measure of sensitivity is the noise factor FN:

FN = (S/N)IN / (S/N)OUT
where S = Signal and N = Noise
FNF == (S/N)
10 log
IN /(FN) dB
(S/N) OUT
Since the noise factor is a dimensionless figure of merit we can derive the noise figure as:

F = 10 log (FN) dB

Using the equation PN = kTB it is possible to calculate the theoretical value of absolute
sensitivity for a given bandwidth. For example, if a spectrum analyzer generates no noise
products at a temperature of 17 degrees Celsius, referred to a 1Hz bandwidth, then:

absolute sensitivity = 1.38 x 10–23 x 290

= 4x1021 W/Hz

= –174 dBm/Hz

To determine the noise figure of a typical spectrum analyzer where the average noise floor is
specified as 120 dBm referred to a 300 Hz bandwidth:

–120 dBm = –174 dBm/Hz + 10 log 300 + F (dB)

F (dB) = –120 + 174 –24.8

Noise Figure = 29.2 dB

26 Understanding Spectrum & Signal Analysis

Video Filtering or Averaging
Very low level signals can be difficult to distinguish from the average internal noise level of
many spectrum analyzers. Since analyzers display signal plus noise, some form of averaging or
filtering is required to assist the visual detection process. As mentioned earlier, a video filter is
a low pass, post detection filter that averages the internal noise of the analyzer.

Because spectrum analyzers measure signal plus noise, the minimum signal power that can be
displayed is the same as the average noise power of the analyzer. From this statement it would
appear that the signal would be lost in the analyzer noise but:

if signal power = average noise power

then by definition, the minimum signal power that can be displayed will be:

Where
S+N
S = signal power =2
N
N = average noise power

When the signal power is added to the average noise power, the resultant signal power
displayed will be 3 dB greater (Figure 29). This 3 dB difference is sufficient for low level signal
identification.

Figure 29

Signal Display Range

The signal display range of a spectrum/signal analyzer with no input attenuation is dependent
on two key parameters.

w w w. a n r i t s u . c o m 27
 • The minimum resolution bandwidth available and hence the average noise level of
the analyzer and
• The maximum level delivered to the first mixer that does not introduce distortion or
inflict permanent damage to the mixer performance.
Typical values for these two factors are shown in Figure 30.

As the input level to the first mixer increases so the detected output from the mixer will
increase. However, since the mixer is a semiconductor diode the conversion of input level to
output level is constant until saturation occurs. At this point the mixer begins to gain compress
the input signal, and conversion reverts from linear to near logarithmic. This gain compression
is not considered serious until it reaches 1 dB.

Input levels that result in less than 1 dB gain compression are called linear input levels (Figure
31). Above 1 dB gain compression, the conversion law no longer applies and the analyzer is
considered to be operating nonlinearly and the displayed signal amplitude is not an accurate
measure of the input signal.

Distortion products are produced in the analyzer whenever a signal is applied to the input.
These distortion products are usually produced by the inherent nonlinearity of the mixer. By
biasing the mixer at an optimum level internal distortion products can be kept to a minimum.
Typically, modern spectrum analyzer mixers are specified as having an 80 dB spurious free
measurement range for an input level of –30 dBm. Obviously the analyzer will be subjected to
input signals greater than –30 dBm and to prevent exceeding the 1 dB compression point, an
attenuator is positioned between the analyzer input and the first mixer. The attenuator
automatically adjusts the input signal to provide the –30 dBm optimum level.
+30 dBm Damage Level

Total –10 dBm < 1 dB Gain Compression

Measurement
Range

–30 dBm Max Input for Specified Distortion

Optimum Operating Range

(80 dB, Spurious-Free

–110 dBm Noise Level (10 kHz BW)

–120 dBm Noise Level (1 kHz BW)
–130 dBm Noise Level (100 Hz BW)
–140 dBm Noise Level (10 Hz BW)

Figure 30

28 Understanding Spectrum & Signal Analysis

 1 dB

P in

P out
Figure 31

Dynamic Range
The dynamic range of a spectrum/signal analyzer is determined by four key factors.
• A verage noise level.
This is the noise generated within the spectrum analyzer RF section, and is
distributed equally across the entire frequency range.
• Residual spurious components.
The harmonics of various signals are mixed together in complex form and converted
to the IF signal components which are displayed as a response on the display.
Consequently, the displayed response is present regardless of whether or not a
signal is present at the input.
• Distortion due to higher order harmonics.
When the input signal level is high, spurious images of the input signal harmonics
are generated due to the nonlinearity of the mixer conversion.
• Distortion due to two signal 3rd order intermodulation products.
When two adjacent signals at high power are input to a spectrum/signal analyzer,
intermodulation occurs in the mixer paths. Spurious signals, separated by the
frequency difference of the input signals are generated above and below the input
signals.
The level range over which measurements can be performed without interference from any of
these factors is the dynamic range. This represents the analyzers performance and is not
connected with the dis play (or measurement) range. The four parameters that determine
dynamic range can normally be found in the analyzer specifications.

For simplicity, some analyzer specifications state the dynamic range as “Y dB for an input level
of X dBm”. The following example shows how these parameters are related to dynamic range:

w w w. a n r i t s u . c o m 29
Amplitude Dynamic Range: 70 dB for a mixer input signal level of –30 dBm (Atten. = 0 dB) In
order to achieve this value of dynamic range the following conditions are required:

a) the IF bandwidth must be narrow enough such that the average noise level is better than
–100 dBm.

b) the residual spurious components must be less than –100 dBm.

c) for an input level of 30 dBm the higher harmonic distortion must be better than –70 dB (i.e.
better than –100 dBm).

Analyzer manufacturers often relate the above specifications at a particular frequency or over
a range of frequencies.

Frequency Accuracy
The key parameter relating to frequency accuracy is linked to the type of reference source built
into the spectrum/signal analyzer.
• Synthesized
The analyzer local oscillator is phase locked to a very stable reference source, often
temperature controlled to prevent unwanted frequency drifting. In this case, a
precision crystal is often used and the overall frequency accuracy and stability, both
short term and long term depend on its quality. Portable analyzers, intended for
outdoor use, often have GPS receivers that can significantly improve the stability of
the internal local oscillator.
• Non Synthesized
The analyzer local oscillator operates as a stand-alone voltage controlled source

Uses of Spectrum Analyzer and Signal Analyzers

Signal analyzers incorporate a wide bandwidth digitizer in the IF to capture a time block of
spectrum for analysis. Frequency, time and phase relationships of signals can be analyzed
within the bandwidth and time limits of the captured spectrum. Digital modulation can be
characterized in many ways not possible with a swept tuned spectrum analyzer. Figure 32
compares the block diagrams for a spectrum analyzer and signal analyzer.

Figure 33 show example time blocks of spectrum with a variety of modulations. A signal
analyzer is often used to measure the characteristics of analog and digital modulation.

30 Understanding Spectrum & Signal Analysis

 RF Input Log Envelope Video
Attenuator Mixer IF Gain IF Filter Amp Detector Filter

Input
ADC
Signal

Pre-selector, or
Low-pass Filter Step
Spectrum
Synthesizer Vert
Analyzer
Reference
Oscillator
Horiz
CPU

RF Input
Attenuator Mixer

Input
ADC DDC
Signal
I/Q
Display
Pre-selector, or
Low-pass Filter
Signal Synthesizer
CPU / FFT
Analyzer
Reference
Oscillator

Display

Figure 32

1 0 1 0 Digital Data
Digital Baseband
Modulation Signal
Amplitude
t

Frequency
t

Phase
t

Both Amplitude
and Phase t

U(t) = U(t) • sin(2πfct + φ(t))

AM FM PM

Figure 33

Figure 34 shows a signal analyzer display of QPSK modulation in polar display format. The
polar display is called a constellation or vector diagram.

w w w. a n r i t s u . c o m 31
 Tb Q
1 0 0 1 1 0 1 0 0 1 0 0 1 0 1 0

Information 01 11
Bits: 0 t
–A

A
I-Data:
I
0 t

A
Q-Data:
00 10
0
–A
t

T = 2Tb
QPSK

Figure 34

Structural differences and similarities

The measured signal is converted to an intermediate frequency (IF) signal at the mixer and
then passed through a narrowband filter. The amplitude of the frequency in the measured
signal (Fin) is determined by the detector.

Additionally, since the input frequency Fin is composed of Flo + Fif (local oscillator frequency
+ intermediate frequency), sweeping (continuously changing) Flo makes it possible to obtain
the spectrum corresponding to the sweep range. In other words, the frequency width (span)
that can be analyzed by a spectrum analyzer is determined by the sweep range of the local
oscillator.

In passing, the part of Figure 35 enclosed by the dotted line (post-IF signal processing and
local oscillator control) have been digitized, resulting in much faster measurement speed and
higher accuracy. Figure 35 shows the operation principles of the spectrum analyzer.

Figure 35

32 Understanding Spectrum & Signal Analysis

If we treat the digital part of the spectrum analyzer (part inside dashed line in Figure 35) as a
black box, comparison with the signal analyzer operation principles shows the circuit diagram
of both the spectrum analyzer and signal analyzer to be very similar. One reason is because the
signal analyzer offers added spectrum analyzer functions. On the other hand, compared to the
signal analyzer, which captures the spectrum using FFT, a big difference between the two is
that the spectrum analyzer captures the spectrum according to the operating principles in
Figure 35 using filtering and digital log conversion/detection.

Operation differences
Due to the different principles of spectrum capture, the signal analyzer is ideally suited for
capturing irregular signals where the spectrum changes in the short term. This is a key functional
difference compared to the spectrum analyzer, which is better for measuring stationary signals
with a stable spectrum over time. In addition, the spectrum analyzer and signal analyzer have
other differences as a consequence of the differences in the operation principles.

The first difference is the analysis frequency width (span). The spectrum analyzer analysis
frequency width (span) is dependent on the local oscillator sweep range; the local oscillator
supports sweeping of a wide frequency range exceeding 1 GHz. Conversely, the signal analyzer
analysis frequency width (frequency span) is determined by the IF bandwidth. The actual IF
bandwidth is several MHz to 100 MHz. As a result, the spectrum analyzer is used to measure
spectrums with a wide frequency of more than several 100 MHz.

The second is the time required for measurement. The problem with spectrum analyzer
measurement time is the time required for the local oscillator sweep. Consequently, sweep
speed (frequency conversion rate) is limited by the frequency resolution (RBW) and a drop in
the sweep speed is required to delay the response as resolution increases (RBW becomes
narrower).

In comparison, the signal analyzer measurement time depends only on the time to capture
and FFT-process the signal and is much shorter than the local oscillator frequency sweep. As a
result, when comparing both measurement times for a relatively narrow span, the signal
analyzer completes measurement and obtains the measurement in the shortest time.

The difference is striking when measurement is repeated, such as when averaging measured
values (Figures 36 and 37).

Incidentally, the effect of using averaging to decrease randomness in measured results

depends on the video filter VBW setting for spectrum analyzers and on the number of
measurements used at averaging processing for signal analyzers.

w w w. a n r i t s u . c o m 33
Spectrum Analyzer Signal Analyzer
Measurement time (Sweep time): 3 s Capture time: 50 µs
(VBW: 100 Hz) (Average trace points: 1024)

Figure 36: Example of narrowband high resolution measurement

(Center frequency: 1 GHz, Span: 10 kHz. RBW: 100 Hz)

Spectrum Analyzer Signal Analyzer

Measurement time: 1.2 s Capture time: 10 ms
(Sweep time: 12 ms; Averaging count: 100, (Analysis time length: 10 ms, Average
Frequency span: 2 MHz) trace points: 2049, Frequency span: 5
MHz)
Figure 37 : Example of Averaging measurement time
(Center frequency: 1 GHz, RBW: 10 kHz)

34 Understanding Spectrum & Signal Analysis

The third is that the spectrum analyzer captures measurement results while the measured
signal is connected whereas the signal analyzer uses batch processing so it can perform
measurement both in real time and when offline. As a consequence, various analyses can be
performed by changing the analysis method and parameters (RBW, time specifications, etc.).
In addition, since the signal analyzer can record and save the measured signal to a storage
device, such as a hard disk, it is possible to take data recorded on-site back to the laboratory
where it can be replayed and analyzed.

Applications
As stated in the introduction, spectrum analyzers are used to display the frequency and
amplitude of signals in the frequency domain. Efficient transmission of information is
accomplished by a technique known as modulation. This technique transforms the information
signal, usually of low frequency, to a higher carrier frequency by using a third, modulation
signal. But why modulate the original signal? The two primary reasons are:

1) modulation techniques allow the simultaneous transmission of two or more low frequency,
or base band signals onto a higher, carrier frequency and

2) high frequency antenna are small in physical size and more electrically efficient.

In this section we will consider three common modulation formats:

• Amplitude Modulation or AM.
• Frequency Modulation or FM.
• Pulse Modulation or PM.
Each modulation technique places emphasis on a particular area of the analyzer’s specification.

Amplitude Modulation
As the name suggests, amplitude modulation is where the carrier signal amplitude is varied by
an amount proportional to the amplitude of the signal wave and at the frequency of the
modulation signal. The amplitude variation about the carrier is termed the modulation factor
‘m’. This is usually expressed as a percentage called the percent modulation, %M.

The complex expression for an AM carrier shows that there are three signal elements.
• the unmodulated carrier.
• the upper sideband whose frequency is the sum of the carrier and the modulation
frequency.
• the lower sideband whose frequency is the difference between the carrier and the
modulation frequency.

w w w. a n r i t s u . c o m 35