Keysight Technologies

Cognitive Radio Algorithm

Development and Testing

White Paper
Introduction

The term “cognitive radio” (CR) means different things to differ-

ent people. However, most seem to agree that cognitive radios
have, as a fundamental building block, the requirement to sense
their environment. Specifically, they require spectrum and location
sensing. Fast and accurate spectrum sensing is critical to dynamic
spectrum sharing, whether for television (TV) whitespace, next
generation (XG) radio, or land mobile radio (LMR) applications.
The spectrum sensing block is common to most, if not all, dynamic
spectrum access technologies, where the goal is to find and
occupy underutilized, or “whitespace” spectrum. This is distinctly
different from multi-mode, multi-band cell phones which search
for occupied spectrum in the form of base-station downlink signals
and then take their instructions from the cellular system. Cellular
systems are quite sophisticated and achieve high spectral
efficiency using highly coordinated and controlled spectrum
access. Cognitive radio systems achieve increased spectral
utilization through opportunistic access to what the radio believes
is underutilized spectrum. As with most new technologies, CR
technology brings with it new design and performance verification
challenges. This paper describes some of these challenges and
introduces a new tool, the cognitive radio algorithm R&D testbed.

by Greg Jue, Keysight Technologies

Bob Cutler, Keysight Technologies
For the purposes of this paper, we will relevant information. For example, underutilized if additional RF energy
take a somewhat inclusive approach cognitive radios might share can be injected without causing
to the definition of cognitive radio, information to enable cooperative unwanted interference. A simple
including even the simplest of forms spectrum sensing. They might also example would be the addition of
dynamic spectrum-sharing technolo- access external information from a second spread-spectrum signal
gies, such as detect-and-avoid (DAA) sources such as protected transmit- to spectrum already “occupied” by
and dynamic-frequency-selection ter databases. As we increase the an existing spread-spectrum signal.
(DFS). With this admittedly broad sophistication of cognitive radios, we Clearly, accurate and reliable
definition, a cognitive radio is defined can perhaps evolve from simple DFS spectrum sensing is not without
as a radio that can sense its environ- and DAA radio technology to truly- its challenges.
ment and then modify its behavior, cognitive radios that fabricate signals
based on a set of rules (policy), and which fully utilize non-contiguous The key functional blocks of a cogni-
without operator intervention. The segments of underutilized spectrum tive radio are shown in Figure 1. The
term whitespace is used generically in a predictive manner. radio policy can be described as a set
in this discussion so as to not limit its of rules which provide operational
scope to TV whitespace. It is worth noting that underutilized boundaries and limitations. The actual
spectrum is not the same thing as operating characteristics of the
The performance of a cognitive radio unused spectrum. It is generally radio/modem are determined by the
system depends on many factors. true though, that unused spectrum behavior and control block using
Some spectrum sensing algorithms is underutilized – provided it is truly information from the other functional
are fast and low-complexity. Others unused. In some cases, spectrum that blocks. As an example, the behavior
provide high sensitivity and reliability, appears to be unused may actually be and control block could select a
but use more computational resourc- used. For example, if it is monitored modulation and frequency – based on
es and may require longer detection by receivers (e.g., radio astronomy) or what’s allowed by the policy – for the
intervals. The cognition engine may if use of the spectrum would result in geographic area determined by the
be implemented with simple logic or interference due to poor selectivity in location sensing block, using spec-
it may incorporate neural networks. a victim receiver tuned to an adjacent trum resources deemed as available
CR performance generally increases frequency. It is also generally true by both the spectrum sensing and
when the radio has access to more that occupied spectrum may be database blocks.

Emitter
Radio/modem
database

Sense Behavior
spectrum and control

Sense location Policy

Figure 1. A basic cognitive radio

3
 Sense location Policy

One motivation for cognitive radios

is increased spectrum utilization in
situations where it is not practical to
have a centralized and all-knowing
spectrum resource manager. In such
situations, cognitive radios determine
what spectrum resources are avail-
able and how best to use them.
This process, sometimes referred
to as “filling the whitespace,” is
depicted in Figure 2. Here, one of the

Time
whitespace radios appears to simply
change frequencies. The other radio
seems to anticipate the behavior of
the protected signals and completely
modify its signal’s spectral shape to
avoid interference, while filling the
new spectrum holes.

Frequency
Figure 2. Filling the whitespace

Orthogonal frequency-division multiplexing (OFDM) technologies are particu-

larly well suited to cognitive radio applications because of the ease with which
signals can be adapted to channel conditions and capacity requirements.
This is evident in radio standards such as Mobile WiMAX™, and Long Term
Evolution (LTE), both of which utilize a form of OFDM called orthogonal
frequency-division multiple access (OFDMA). Mobile WiMAX can support 5,
7, 8.75, and 10-MHz channel bandwidths with fast Fourier transform (FFT)
sizes of 512 or 1024. Similarly, LTE supports 1.4, 3, 5, 10, 15, and 20-MHz band-
widths. OFDM signals also have the potential to be modified on a subcarrier-
by-subcarrier basis. By selectively disabling subcarriers, wideband signals can
avoid interfering with narrow-band signals, even if they are located in the same
main channel. While not specifically designed for cognitive radio or whitespace
applications, we will use these standard signals to demonstrate key concepts.

4
Challenges and Barriers to Cognitive Radio Development

Cognitive radio development require- – Detect-and-avoid ‘what-if’ sce- Cognitive radio requirements
ments represent a new paradigm for narios are difficult to evaluate in for maximum rerformance and
R&D engineers by challenging estab- a lab environment, and are also minimum interference
lished design and test methodologies. difficult to test methodically in field
In particular, CR development poses environments.
significant challenges and barriers to 1. Quickly and accurately detect
– Metrology-grade test equipment
developing and testing CR whitespace a wide variety of protected
is needed to test radio algorithms,
algorithms, including: signals under dynamic channel
but off-the-shelf test equipment
typically does not support signal conditions. Signals can vary in
– Real-world spectral environments a number of ways, including:
types that represent the realistic
need to be used for algorithm bandwidth, duration, modulation,
CR environments that the radio
development and testing. However, and cyclo-stationarity properties.
would be exposed to.
the use of live signals creates
test conditions that cannot be – Algorithms need to be tested with 2. Identify underutilized spectrum in
replicated from one day to the next. real-world impairments such as the presense of strong adjacent
Short field trials may not experi- intermodulation distortion and channel signals while avoiding
ence the full range of conditions, spurious signals (both internally false-positive detections with
while long field trials are expensive and externally generated). varying levels of noise.
and may be impractical.
– Laboratory and R&D environments These challenges and barriers 3. Make timely and intelligent
provide the design tools and highlight the need to innovate new decisions within the constraints
equipment needed to develop and algorithm development and test meth- of the policy that maximize radio
test algorithms, but emulating real- odologies that combine real-world system performance, while mini-
world CR spectral environments CR spectral environments with the mizing interference and the use
to adequately design and test CR controlled, repeatable and toolset-rich of radio and spectral resources.
algorithms can be very difficult. R&D/lab environment to accelerate
algorithm development and testing 4. Do all of the above within the
before deploying the radio hardware practical constraints of hardware
to the field. performance limitations such as
RF performance (e.g., spurious
and distortion), battery consump-
tion and computational resources.

5
A Paradigm Shift in Algorithm Development and Testing

A new tool, the cognitive radio algo- The remote field portion of the CR cognitive radio and spectrum-sensing
rithm R&D testbed, combines remote R&D testbed consists of one or more algorithms. The receiver’s tuning range
field sensing with a toolset-rich Keysight Technologies, Inc. N6841A is 20 MHz to 6 GHz. The RF sensors
controlled lab environment (Figure 3). RF sensors that are installed at vari- are small and weatherproof, making
Doing so allows CR algorithms to be ous geographic locations. The sensor them easy to install--an important
developed and tested with real-world locations are selected to gather data consideration when the installation is
CR environment signals. Using this from different environments, or, to temporary, or when greater flexibility
testbed, engineers can: test cooperative sensing algorithms in site selection is desired. More
within an environment. information on the N6841A sensor
– Capture live signals at one or more can be found at:
locations for live testing, or record The N6841A RF sensors are effec- http://www.Keysight.com/find/RFsensor
for later playback. tively wideband RF receivers/digitizers
that supply the spectrum-sensing algo- The lab environment portion of the
– Synthesize signals using simulated
rithms with baseband I/Q time-series CR R&D testbed utilizes Keysight’s
sources and channel models (to
data. The data is accurately time- SystemVue software to capture the
model transmitters and impair-
stamped with time-of-day information CR environment data from the remote
ments).
obtained either from the integrated N6841A sensors. SystemVue’s math
– Combine captured and synthesized GPS receiver, or the built-in IEEE 1588 language and math language debug-
signals. network-based precision time protocol ger are used to write and test CR
(PTP). The receiver’s information band- whitespace algorithms with the remote
– Test spectrum-sensing algorithms
width (and bandwidth-proportional field-captured signals (Figure 4). More
(single or cooperative).
sample rate) can be reduced from a information on SystemVue can be
– Test cognition engine performance. maximum of 20 MHz, to something found at:
– Test radio link performance. better suited to the needs of the http://www.keysight.com/find/SystemVue

Download field and simulated

Cognitive radio R&D testbed signals to signal sources
Controlled lab environment

Remote field environment

Field
sensor #1

Prototype hardware
– Sense/capture field signals and dev. boards
LAN
network
– Develop and evaluate algorithms
Field with actual field signals
sensor #3
Field
– Evaluate simulated radio link
sensor #2 performance (e.g. WiMAX, LTE)
in detected whitespaces

– Evaluate “what if” scenarios

(e.g., DAA interferers, jammers)

– Download field signals and

simulated signals to test
equipment for prototype
HW testing

– Capture DUT signals and bring into

simulation to process with simulated
algorithms and designs

Capture signals with signal analyzers, logic

analyzers, oscilloscopes and use in simulation

Figure 3. Cognitive radio R&D algorithm development testbed

6
 CR algorithm development & testing environment
SystemVue math algorithm development
Download and capture signals
to/from test equipment

N6841A

Setup WiMAX or LTE link

Detect whitespace
and determine
WiMAX link parameters

Simulate link performance

Figure 4. Using SystemVue’s math algorithm environment to develop and test CR whitespace algorithms

prime example of which is determin- the captured field-sensor data to

The OFDMA simulation modeling ing the impact of swept interferers be downloaded to signal generator
capability in SystemVue can then on OFDMA link coded-BER/block- arbitrary waveform generators to
be used to configure and evaluate error-rate (BLER) performance for re-create the field-captured signal
OFDMA Mobile WiMAX and LTE link detect-and-avoid scenarios. in the lab environment. Once the
performance. SystemVue’s math lan- waveform is downloaded to the signal
guage scales the OFDMA bandwidths It can be useful to download and generator arb, it can be stored in
and establishes the transmission store remote field-sensor data to non-volatile memory. To facilitate DUT
center frequencies within the metrology-grade test equipment for testing, DUT outputs can be captured
detected CR whitespace to evaluate hardware device-under-test (DUT) and then read back into simulation
the coded, OFDMA bit-error-rate testing in the lab environment. to post-process signals. An example
(BER) performance of the radio link. Simulation provides the link between of this is testing the coded BER/
Simulation also provides the flexibility the remotely located field sensors BLER of a CR RF receiver. Here,
to evaluate “what-if” scenarios, a and signal generators. It allows simulation provides the baseband
post-processing required to support
the coded-BER/BLER measurement.

7
Case Study: A Mobile WiMAX CR Whitespace Algorithm

To better understand how the CR R&D calculated. Next, the coded BER of 1. Capture and read
algorithm testbed can be applied, the configured Mobile WiMAX radio N6841A to sense
consider the development and test link is evaluated to determine radio CR environment
of a Mobile WiMAX CR whitespace link performance.
algorithm using the flow shown in
Figure 5. The captured CR environment is also 2. Detect whitespace
downloaded to a signal generator to space & check
The flow involves capturing the CR re-create and analyze the CR environ- against policy
spectral environment from the remote ment. Lastly, the CR environment
field location. The sensor is remotely spectrum is de-modulated both in
located in western Washington State, simulation and with test equipment 3. Determine WiMAX
while the simulation and test equip- using the Keysight 89601 vector link parameters and
ment setup is located in an office signal analysis (VSA) software to configure simulated
environment in eastern identify the CR environment. WiMAX link
Washington State.
More details on these specific steps
SystemVue’s math algorithm are as follows. 4. Perform simulated
processes the captured CR spectral WiMAX link analysis
environment to determine potential Step 1: Capture remote
whitespace. The potential whitespace
CR environment and
frequencies are evaluated against a
policy to determine valid whitespace read into SystemVue 5. Re-create field signal
frequencies and available bandwidth The first step in this case study is to on test equipment to
to support an OFDMA Mobile WiMAX capture the CR environment using identify signal type
link. The math language algorithm an N6841A sensor remotely located
then configures and scales the in western Washington State (Figure Figure 5. Cognitive radio
Mobile WiMAX center frequency 6). This is accomplished from the whitespace algorithm case
and channel bandwidth to either convenience of an office environment study flow
5 or 10 MHz, depending on the in eastern Washington State using a
available whitespace it sensed and custom N6841A link for SystemVue.

Instructions on installing and downloading the required dynamic link libraries

(DLLs) for the custom N6841A link can be found at:

http://edocs.soco.keysight.com/display/eesofkc/SystemVue+N6841A

N6841A sensor
(custom model)
N6841A

Capture and read N6841A

to sense CR environment

Figure 6. Capture remote CR environment from field and read into SystemVue

8
 N6841A
sensor

N6841A

Capture and read N6841A

to sense CR environment

Detect whitespace and

Sensing algorithm detects rising/falling edges check against policy
and checks policy to determine whitespace

Frequency rules policy

Valid whitespace
determined within policy

Figure 7. Sense whitespace, compare to policy and determine valid whitespace for OFDMA transmission

The custom N6841A link serves as The primary whitespace algorithm Real-world whitespace algorithms can
a simulation signal source to read calls a secondary algorithm to detect be quite involved and sophisticated,
the captured signal into SystemVue rising and falling edges on the and thus challenging to develop and
where it is then stored using a sensed, CR environment spectrum. debug. The algorithms used in this
spectrum analyzer sink. This enables This secondary algorithm effectively case study are fundamental; however,
the stored spectrum data to be post- integrates the spectral power within they demonstrate how the CR R&D
processed and evaluated using the a user-defined resolution bandwidth, algorithm testbed can be used to
CR whitespace algorithm modeled in sweeps the frequency span and develop and test more elaborate
SystemVue as per Step 2. compares the integrated power to and robust whitespace algorithms.
user-defined power thresholds. The SystemVue’s math language debug-
Step 2: Whitespace rising and falling edges, graphically ging capability was quite useful in
illustrated in Figure 7 as the vertical debugging algorithm issues during
algorithms sense the the development of this whitespace
bars on the lower spectrum, are com-
spectrum to determine pared against a frequency rules policy algorithm code. The screen capture
valid whitespace to determine the validity of detected of the debugger in Figure 8 shows
whitespace. Valid transmission where a breakpoint is set in the
Once the CR environment has
frequencies are displayed as the blue algorithm code.
been captured and stored in the
SystemVue dataset, several math spectral mask on the upper spectrum
in SystemVue. Once valid whitespace Once the breakpoint is set, the
algorithms are utilized to determine
is determined, it is displayed as the algorithm code can be single-stepped.
the available whitespace and
green spectral mask on the Calculated values from the code are
determine if it is valid within the
lower spectrum. displayed on the left-side of the math
constraints imposed by the radio’s
language interface. Breakpoints can
policy (Figure 7).
also be set in functions or other algo-
rithms called from the main algorithm
(e.g., the algorithm which detects
rising and falling spectrum edges).
9
 Figure 8. Debugger and single stepping
through whitespace algorithm code

Single-step
through code

Add/remove
breakpoint

Code variable values are displayed as code is single-stepped

Steps 3 and 4: Whitespace algorithms sense the spectrum to determine valid whitespace
Next, the rising and falling edges of the sensed CR spectrum are compared to the frequency rules policy to determine the
valid whitespace within the lower and upper frequencies defined in the policy. In Figure 9, the policy frequency limits are
displayed by the blue rectangle in the spectrum, while the calculated valid whitepace is displayed as the green rectangle.
The CR algorithm then determines if the whitespace is sufficient to support a 5- or 10-MHz OFDMA Mobile WiMAX or LTE
link. It scales the OFDMA WiMAX or LTE channel bandwidth configuration accordingly, to fill the available whitespace and
centers the OFDMA spectrum within the valid whitespace. A user-defined preference is set in the algorithm to specify
whether a Mobile WiMAX link or LTE link will be used to occupy the spectrum.

Algorithm determines if whitespace is sufficient for 5 MHz or 10 MHz WiMax

signal, then configures and simulates to evaluate radio link performance Capture and read N6841A
to sense CR environment

Detect whitespace and

check against policy

Determine WiMAX link

parameters and configure
simulated WiMAX link

BER Perform simulated

Sum captured field signal measurement WiMAX link analysis
or simulated RF interferers

WiMAX
receiver

WiMAX
source

Figure 9. Determine valid whitespace by comparing to policy, configure and scale

the Mobile WiMAX channel bandwidth and center frequency, and evaluate coded
BER for the OFDMA link

10
The rectangular-shaped spectrum Step 5: Identify CR allow this analysis to be performed
(middle spectrum) on the spectrum environment signals over adjustable time intervals, such
plot in Figure 9 is the Mobile WiMAX as when analyzing a portion of a sig-
spectrum. It has been centered in When sensing spectrum in the real nal searching for training sequences.
the valid whitespace and scaled to world, it is very likely that one will Auto-correlation functions can be
occupy the available whitespace observe signals that raise questions used to search the signal for periodic-
bandwidth. The two red spectrums on about their origin and perhaps even ity. Many demodulators both analog
either side are the sensed CR envi- their legitimacy. In TV whitespace (AM/FM/PM/IQ) and digital (too
ronment that was detected with the applications, TV and wireless micro- numerous to list) are also available
N6841A sensor. The Mobile WiMAX phone signals must be protected, for more detailed analysis.
schematic on the left is configured by while most other signals--especially
the math algorithm and simulated to other whitespace signals--are not. For The 89601 VSA can be used with
evaluate the coded BER performance these reasons, there is often a need Keysight test equipment, as well as
of the Mobile WiMAX link. to identify over-the-air signals to ver- with Keysight’s SystemVue software.
ify correct behavior of the spectrum When used with SystemVue, it
To evaluate detect-and-avoid scenari- sensing and cognition algorithms. processes simulated signals instead
os, simulated interferers can be intro- of DUT hardware signals with test
duced into simulation and swept to One tool that has proven very useful equipment. Several possibilities exist
evaluate the impact of the interferer for analyzing and identifying unknown for identifying remotely sensed signals:
on OFDMA link performance. Figure signals is the 89601 VSA. It is unique identifying the signal in simulation by
10 shows a differential quadrature in its ability to isolate and analyze analyzing SystemVue’s N6841A sensor
phase-shift keying (DQPSK) narrow- signals in time, frequency and data with the 89601 VSA software
band interferer being swept across modulation domains. The tool can be or using SystemVue to download
the CR frequency environment, along used to observe the burst length of the N6841A sensor data to a signal
with its impact to the Mobile WiMAX a signal, its power statistics and its generator and then using the 89601
OFDMA BER performance. spectral characteristics. Gate markers VSA to analyze the signal. Both
possibilities are examined below.

OFDMA coded BER vs. narrowband swept frequency QPSK interferer

Narrowband
interferer

Sweep narrow-
band interferer
vs. frequency to
evaluate the
impact on
OFDMA BER

Figure 10. Impact of swept narrowband interferer on OFDMA BER

11
The first possibility is to identify the The second possibility is to download Summary
signal in simulation using the VSA the N6841A captured signal to an
simulation measurement capability Keysight ESG or MXG signal genera- This white paper details a novel
as shown in the upper-left corner of tor, as shown in the lower-right cor- new R&D testbed for developing and
Figure 11. The sensed spectrum from ner of Figure 11. This is accomplished testing cognitive radio whitespace
the remote N6841A sensor is read using the sink on the lower-right of algorithms. It combines simulation
into SystemVue and then demodu- the SystemVue schematic. The sink and test to enable sensor data to be
lated by the 89601 VSA software. downloads the simulated I and Q captured at remote field locations
While the 89601 VSA is measurement to the signal generator’s arbitrary and used in the convenience of an
software which runs on Keysight test waveform generator, turning the office or lab environment to develop
equipment (e.g., RF signal analyzers, simulated signal into a physical and test cognitive radio algorithms.
logic analyzer and oscilloscopes), real-world test signal for DUT testing. This offers the benefit of using real-
here it measures simulated signals In this example, the captured signal world environment signals (captured
from SystemVue, where the simulated from the N6841A is downloaded to remotely) to develop and test
signal is the captured sensed spec- an ESG arbitrary waveform generator algorithms in an R&D environment,
trum from the N6841A sensor. The and the RF output of the ESG is con- with a rich set of simulation tools
sensed spectrum shows two spectral nected to an MXA signal analyzer. and metrology-grade test equipment.
peaks--the blue spectrum traces on The same 89601 VSA software used Flexible simulation can be used
the lower left of the VSA display for the demodulation analysis in to evaluate “what-if” interference
in Figure 11. By setting the VSA’s SystemVue is also used in the MXA scenarios for detect-and-avoid situa-
center frequency to center on one of to demodulate the ESG RF output. tions. Seamless integration between
the spectrums, and trying different Once the CR environment signal has simulation and test equipment
demodulation configurations pre-set been downloaded to the ESG or MXG, enables remotely detected signals to
on the VSA software, this signal it can be stored in non-volatile be identified in simulation, or on the
type can be identified as a WCDMA memory for further testing in the testbench, by downloading and ana-
downlink signal. The code domain is lab environment. lyzing signals with test equipment.
shown on the upper left, constellation
on the upper right, and EVM on the
lower right of the VSA display.

Identify signal in System Vue with VSA software Capture and read N6841A
to sense CR environment

Detect whitespace space

and check against policy
N6841A sensor

Determine WiMAX link

parameters and configure
simulated WiMAX link

Perform simulated
WiMAX link analysts

Re-create field signal

on test equipment to
identify signal type

Figure 11. Identifying remotely-sensed signals using the CR R&D testbed

12
13 | Keysight | Cognitive Radio Algorithm Development and Testing - White Paper

For more information on Keysight

Technologies’ products, applications or
services, please contact your local Keysight
office. The complete list is available at:
www.keysight.com/find/contactus

Americas
Canada (877) 894 4414
“Mobile WiMAX” is a registered trademark of the WiMAX Forum. Brazil 55 11 3351 7010
Mexico 001 800 254 2440
www.keysight.com/find/eesof-eda United States (800) 829 4444

Asia Pacific
Australia 1 800 629 485
China 800 810 0189
Hong Kong 800 938 693
India 1 800 112 929
Japan 0120 (421) 345
Korea 080 769 0800
Malaysia 1 800 888 848
Singapore 1 800 375 8100
Taiwan 0800 047 866
Other AP Countries (65) 6375 8100

Europe & Middle East

Austria 0800 001122
Belgium 0800 58580
Finland 0800 523252
France 0805 980333
Germany 0800 6270999
Ireland 1800 832700
Israel 1 809 343051
Italy 800 599100
Luxembourg +32 800 58580
Netherlands 0800 0233200
Russia 8800 5009286
Spain 0800 000154
Sweden 0200 882255
Switzerland 0800 805353
Opt. 1 (DE)
Opt. 2 (FR)
Opt. 3 (IT)
United Kingdom 0800 0260637

For other unlisted countries:

www.keysight.com/find/contactus
(BP-07-01-14)

This information is subject to change without notice.

© Keysight Technologies, 2009-2014
Published in USA, July 31, 2014
5990-4389EN
www.keysight.com