Cognitive Radio Test Bed Experimentation using USRP and Matlab®/Simulink®

Nuzli Mohamad Anas1, Hafizal Mohamad2, Mohammad Tahir2 1Wireless Software Development, 2Wireless Communications Cluster

MIMOS Berhad, Technology Park Malaysia, Kuala Lumpur, Malaysia

[nuzli.anas, hafizal.mohamad, mohammad.tahir]@mimos.my Abstract—Sensing of spectrum holes constitute the fundamental step to utilize spectral resources efficiently in Cognitive Radio (CR). This new technology trends intends to alleviate the scarce and underutilized spectrum issues occurred in traditional fixed spectrum allotment. CR defined as dynamic spectrum usage of unlicensed (secondary) user in opportunistic manner without causing harmful interference to licensed (primary) user. Much recent works presented as either theoretical or simulation approach. However, this paper presents a design prototype of indirect and non-parametric of spectrum sensing method implemented using low-cost Universal Software-defined radio Platform (USRP™). A spectrum-sensing algorithm designed based on energy detection built on top of Matlab®/Simulink® interfaced with a USRP™ main board and SBX transceiver daughterboard; both are Ettus Research product.

Keywords Cognitive Radio, Spectrum Sensing, Energy Detection, USRP™

I. INTRODUCTION Current fixed spectrum allotment began spectrum

dearth additionally atrophies by underutilized usage. In 1998 [1], cognitive radio (CR) concept introduced by Joseph Mitola has proliferated a new technology paradigm to the spectral congestion solution. Intuitively, CR builds upon software-defined radio (SDR) technology that is able to aware of its operating environment and automatically adapts itself to the desired communications. Operating frequency, power output, antenna orientation, modulation, and transmitter bandwidth are just a few of the operating parameters that can be adjust automatically in a CR system. Therefore, the next generation wireless network and devices philosophy is aimed such that self-configuring, self-organizing, self-optimizing, and self-protecting of CR system [2].

CR, as an agile radio system, utilizes available spectrum dynamically and opportunistically without causing harmful interference. CR users equipped with accurate sensing able to fill in spectrum holes and adapting to its radio transmitting parameters. However, one of the most challenges for CR users is to detect the presence of licensed (primary) users at particular instant and specific location. Furthermore, it must continuously sense the spectrum being used to detect reappearance of licensed users, and withdraw the resources at instant.

This paper describes the initial work of cognitive radio studies. It presents a detailed survey of spectrum sensing based on energy detection and also presents an experimental prototype system for cognitive radio built on top of Matlab/Simulink to take advantages of extensive signal processing and communication toolboxes. This prototype is rapidly designed using a low-cost platform called universal software defined-radio platform (USRP™) interfaced with RF front-end daughterboard which provides flexibility and reconfigurable of CR test bed.

The rest of the paper organized as follows: Section II presents the decision hypothesis theory and reviews indirect spectrum sensing algorithm based on energy detection, used in the USRP™ and Simulink experimentation. This followed by the design of CR test bed system in Section III. Then, the implementation and design analysis discussed greatly in section IV. A great emphasize has been given on the spectrum estimation. Section V, concludes current work and suggests future direction towards efficient spectrum sensing techniques.

II. SPECTRUM SENSING Spectrum sensing involves the detection of primary

user signals, which constitutes the fundamental step in spectrum opportunity usage. The secondary user devices, typically run the spectrum sensing algorithm to sense and adapt into the environment, by reconfiguring its radio transmission parameters to make use of the available spectrum efficiently. Spectrum sensing is broadly categorized into two types [3], direct and indirect spectrum sensing, which are known as primary user’s receiver, and transmitter detection respectively. Spectrum holes detected based on primary user receiver’s local oscillator leakage power falls under direct spectrum sensing category. Other proactive detection approach based on close loop control suggested in [4][5] has widely been deployed using power control, adaptive modulation/coding or automatic request retransmission.

Detection of primary user transmitter performed based on the received signal at CR users. Matched filter detection, energy based detection, covariance based detection, waveform-based detection, Random Hough transform based detection are among the indirect spectrum sensing approach which have been given extensive survey in [2][6]. This paper reviews a non-parametric method of energy-based detection as spectrum sensing algorithm experimented on CR test bed.

978-1-4673-3033-6/10/$26.00 ©2012 IEEE

2012 International Symposium on Computer Applications and Industrial Electronics (ISCAIE 2012), December 3-4, 2012, KotaKinabalu Malaysia

229

A. Decision Theory Spectrum sensing essentially performs a binary

hypothesis test on primary users’ existence in a particular bandwidth. Assuming the received signal, y(k) scenario at the CR users

( ) ( )kwkyH =:0 (1)

( ) ( ) ( )kwkskyH +=:1 (2)

where s(k) and w(k) denotes the transmitted primary users’ signal and additive white Gaussian noise (AWGN) respectively. H0 and H1 are consider as the absence and presence of the primary users’ signal and index k represent the received signal samples. Intuitively, the received signal is essentially the AWGN under the idle scenario, while under the busy scenario; the received signal would consist of the primary users’ signal deterioted by AWGN. The received signal will have more energy when the spectrum occupied than when it is idle, thus forming the underlying concept of the energy detector discussed in detail in the next section.

Regardless of the precise signal model or detector used, sensing errors are inevitable due to additive noise, limited observations, and the inherent randomness of the observed samples. The key metrics that defined the performance analysis of such detectors are characterized by its probability of detection, Pd and probability of false alarm, Pf. Assuming the detection metrics and it’s respective threshold denoted as E and , thus the probability of detection considered when the signal truly detected, written as

( )1Pr HEPd λ>= (3)

while a false alarm occur when the spectrum detected incorrectly as occupied such that

( )0Pr HEPf λ>= (4)

Consequently, miss detections occur when the spectrum detected as unoccupied however, it actually is not, and its probability of missed detection denoted as

dm PP −= 1 (5)

The choice of the decision threshold reflects the detector system performance. A high probability of false alarms may lead to a potentially wasted opportunity for the CR users to transmit. On the other hand, a missed detection may result in intolerable interference to the licensed users. A typical complementary receiver operating characteristic (ROC) curve illustrates the performance of a binary hypothesis detection system as its discrimination threshold is varied. It is a plotting of the probability of detection versus probability of false alarm, at various signal-to-noise ratio settings, as shown in Figure 1.

Figure 1. Complementary ROC at different SNR

B. Energy Detection Detection of the energy spectra regarded as the most

common way of spectrum sensing because of its low computational and complexities [7]. In this method, CR users do not need any kind of knowledge of the primary users’ signal. The signal detection is performed by comparing the received energy spectra with a predetermined threshold value. Parametric and non-parametric methods are two way of representing the energy spectra, either the received samples belong to any distribution or not. The decision metric for the energy detector with N size of observation vector can be written as

( )=

=N

n

nyE0

2 (6)

Assuming the AWGN can be modeled as zero-mean random variable with variance 2, i.e. w(n) ~ N(0, 2), the decision metric for parametric method follows chi-square distribution with 2N degrees of freedom can be modeled as [8]

( )=iu

uiE

γχχ

2,

22

22 ,

1

0

HH

(7)

where 22uχ denotes a central chi-square distribution with 2u

degrees of freedom and ( )iu γχ 222 denotes a non-central

chi-square distribution with u degrees of freedom and a non-centrality parameter 2 i, respectively. The instantaneous SNR of the received signal at the ith CR is

i, and u is the time–bandwidth product. Meanwhile in periodogram based non-parametric

method, the received signal pass through ADC is transformed into frequency domain using Fast Fourier Transform (FFT). The output of FFT is squared and then averaged it over the samples to get test statistic. The underlying assumption is that with the presence of a signal in the channel, there would be significantly more energy than if there was no signal present. Figure 2 depict the process flow of energy detection in digital domain.

2012 International Symposium on Computer Applications and Industrial Electronics (ISCAIE 2012), December 3-4, 2012, KotaKinabalu Malaysia

230

Figure 2. Energy Detection with Periodogram Method

Although energy detection is less computational and low complexity, it comes with several shortfalls. Energy detection is not able to differentiate between interference from licensed users and noise prominently. Furthermore, the probability of detection is very poor at low signal-to-noise ratio and does not work well for spread spectrum systems [6].

III. PROTOTYPE DESIGN: CR TEST BED The CR test bed consists of two USRP™ N210 of Ettus

Research™, shown in Figure 3, each connected to host PC running baseband processing algorithm built on top of Matlab®/Simulink®. USRP™ Hardware Driver (UHD) provides host driver and API to connect USRP with system object supported by Simulink® in Communication with USRP™ Toolbox. Design and verification of CR system is built using extensive Communication System Toolbox and Signal Processing Toolbox incorporated with Simulink™. RF front-end SBX daughterboard is interfaced with USRP™ supporting frequency range of 400 MHz to 4.4 GHz turn the test bed into a complete RF transceiver system. An omni-directional vertical antenna provides 3 dBi gain is mounted for two-way high-bandwidth communication.

The USRP™ N210 architecture includes a Xilinx® Spartan® 3A-DSP 1800 Field Programmable Grid Array (FPGA) optimized for signal processing application at high sampling rates [9]. Two onboard digital down-converters (DDC) mixed, filtered and decimated the incoming signals from Analog-to-Digital converter (ADC) at 100 MS/s. On the other hand, the baseband signal is interpolated at 400 MS/s on dual Digital-to-Analog converter (DAC) before translating to the required carrier frequency. A Gigabit Ethernet connectivity provide data streaming to host processors enables simultaneously sending up to 50 MHz RF signal in and out of the USRP device. Figure 3 depicts the USRP N210 system architecture while figure 4 shows the test-bed setup.

Figure 3. USRP N210 System Architecture [9]

Figure 4. Cognitive Radio Test bed Setup

The test bed operates at frequency of 2.4GHz in which the transmitter, as in Figure 5, includes the random bit generator, the Quadrature Phase Shift Keying (QPSK) modulation, and the raised-cosine pulse shaping filter. A frame payloads contains 200 random bits is later scrambled to mitigate burst error by guarantee a balance distribution of zeros and ones. Consequently, the scrambled bits are modulated with gray-mapped QPSK symbols. The modulated symbols are upsampled by four by the filter with a roll-off factor 0.5 makes the output rate set to be 200k samples/second. The default IP address of USRP™ transmitter block is set to 192.168.10.2, while the host's Ethernet interface is set to 192.168.10.1 and subnet mask is 255.255.255.0. Simulink™ are set with ‘Rapid Accelerator’ mode to facilitate higher compiler optimization level and performance. The modelling of receiveng system follows the energy detection method described in earlier section.

Figure 5. QPSK Transmitter with USRP™ N210

IV. CR EXPERIMENTAL RESULTS The test-bed allows to experiment with spectrum

sensing capabilities while examine the linked between USRP™ N210 with Simulink™ USRP system objects. However, this model assumes perfect timing and frequency offset to further faciltate energy detection method deployment. Figure 6 shows the noise floor, in which the USRP™ receiver does not sensed any signal energy. While in the next figure, a signal gain of atmost 50dB is detected when USRP™ is being transmitted.

Figure 6. Transmit Signal Power Spectrum

2012 International Symposium on Computer Applications and Industrial Electronics (ISCAIE 2012), December 3-4, 2012, KotaKinabalu Malaysia

231

Figure 7. Sensed signal using Periodogram Methods

V. CONCLUSION AND FUTURE WORK Spectrum sensing is a key requirement for dynamic

spectrum allocation in CR networks. This paper demonstrated a studies on energy detection based implemented on low-cost software-defined radio platform, products from Ettus Research LLC. The design philosophy of this preliminary work is re-configurability and flexible CR system to explore indirect spectrum sensing based on non-parametric energy detection techniques. CR test bed has been demonstrated using USRP™ N210 device interfaced with MATLAB™/Simulink® and able to sense occupied spectrum. Further exploration and extension works are expected for others cognitive engine includes spectrum adaptation and management between licensed and unlicensed users.

REFERENCES [1] J. Mitola and G. Q. Maguire, “Cognitive radio: Making software

radios more personal,” IEEE Personal Communications Magazine, vol.6, no. 4, pp. 13–18, Aug. 1999.

[2] D.B. Rawat and Gongjun Yan, “Spectrum Sensing Methods and Dynamic Spectrum Sharing in Cognitive Radio Networks: A Survey,” International Journal of Research and Reviews in Wireless Sensor Networks, vol.1, no.1, Mar. 2011.

[3] Y.C. Liang, K.C Chen, G.Y. Li, P. Mahonen, “Cognitive Radio Networking and Communications: An Overview,” IEEE Transactions on Vehicular Technology, vol.60, no.7, pp. 3386- 3407, Sept. 2011.

[4] G. Zhao, G.Y. Li, C. Yang, J. Ma, “Proactive Detection of Spectrum Holes in Cognitive Radio,” in proc. IEEE International Conference on Communications, Dresden, Germany, pp. 1-5, June 2009

[5] F. E. Lapiccirella, S. Huang, X. Liu, and Z. Ding, “Feedback-based access and power control for distributed multiuser cognitive networks”, in proc. IEEE ITA, San Diego, CA, Feb. 2009, pp. 85–89.

[6] T. Yucek and H. Arslan, “A survey of spectrum sensing algorithms for cognitive radio applications,” IEEE Communications Surveys & Tutorials, vol.11, no.1. pp. 116-130, 2009

[7] D. Cabric, S.M. Mishra and R.W. Brodersen, “Implementation issues in spectrum sensing for cognitive radios,” in Proc. Asilomar Conf. Signals,Syst., Comput., Pacific Grove, CA, Nov. 2004, vol. 1, pp. 772–776.

[8] F.F. Digham, M.S. Alouini and M.K. Simon, “On the Energy Detection of Unknown Signals Over Fading Channels,” IEEE Transactions on Communications, vol.55, Issue: 1, pp. 21-24, 2007

[9] https://www.ettus.com/

2012 International Symposium on Computer Applications and Industrial Electronics (ISCAIE 2012), December 3-4, 2012, KotaKinabalu Malaysia

232