Analog Converter (DAC) and then possibly upconverted from. IF to RF. The receiver .... Fig. 6. Power Spectrum Density (PSD) of a Typical NTSC Signal in US ...
Software Defined Radio Implementation of SMSE Based Overlay Cognitive Radio Ruolin Zhou1 , Xue Li1 , Vasu Chakravarthy2 , Reginald Cooper3 and Zhiqiang Wu1 Wright State University1 , Air Force Research Laboratory2 , Carnegie Mellon University3 Abstract—In this paper, we demonstrate an adaptive interference avoidance overlay cognitive radio implementation via software defined radio. Using spectrally modulated spectrally encoded (SMSE) framework, we implement a multi-carrier based overlay cognitive radio waveform via USRP (Universal Software Radio Peripheral) software defined radio boards and GNU radio software platform. This cognitive radio is capable of detecting primary users in real time and adaptively adjusting its transmission parameters to avoid interference to primary users. More importantly, this cognitive radio can take advantage of multiple spectrum holes by employing non-contiguous multicarrier transmission technologies. Additionally, we demonstrate that when the primary user transmission changes, the cognitive radio dynamically adjusts its transmission accordingly. We also demonstrate seamless real time video transmission between two cognitive radio nodes, while avoiding interference from primary users and interference to primary users operating in the same spectrum.
I. I NTRODUCTION Cognitive Radio is an intelligent radio which is capable of setting and configuring its own parameters including ’waveform, protocol, transmitting and receiving carrier frequency and networking’ autonomously [1]. Specifically, a cognitive radio should be able to detect spectrum holes (unused frequency bands) and transmit its own transmission over those spectrum holes without interfering the primary users. The innovation which makes engineers see Cognitive Radios a possible technology is the Software Defined Radio (SDR) where the ’software meets the antenna’ [2]. Software-Defined Radio (SDR) is one of the latest and most evolving technologies in the communications industries in civilian, military as well as commercial sectors [3]. For example, Joseph Mitola defines software-defined radio as a software radio capable of performing channel modulation and demodulation in software. That is, ’waveforms are generated as sampled digital signals, converted from digital to analog via a wideband Digital to Analog Converter (DAC) and then possibly upconverted from IF to RF. The receiver, similarly, employs a wideband Analog to Digital Converter (ADC) that captures all of the channels of the software radio node. The receiver then extracts, downconverts and demodulates the waveform using software on a general purpose processor’ [1]. An entirely hardware based radio gives no flexibility because of the fixed characteristics of the modules performing the radio functions. However, in a radio system built with SDR, the flexibility is very high. In our previous work, we have proposed and demonstrated a cognitive centric overlay/underlay waveform design through spectrally modulated spectrally encoded (SMSE) framework
to improve the BER performance and network throughput of cognitive radio [4][5][6]. In this paper, we use a software defined radio to implement the cognitive centric overlay waveform in [4][5][6] and demonstrate an adaptive interference avoidance cognitive radio. Specifically, we employ SMSE framework to generate multi-carrier transmission waveforms over non-contiguous frequency bands for the cognitive radio implementation. Combined with a spectrum sensing engine, the cognitive radio detects the availability of each and every subcarrier in the operational bandwidth. By turning off those subcarriers occupied by the primary users, the cognitive radio implements a non-contiguous SMSE transmission. There are a few unique features of our cognitive radio implementation: (1) we have demonstrated real time seamless video transmission without interference to primary users and from primary users; (2) our cognitive radio is capable of taking advantage of multiple spectrum holes and operating over multiple non-contiguous spectrum bands; (3) the cognitive radio dynamically adjusts which subcarriers to turn off according to the primary users’ transmission; (4) the cognitive radio can also easily adjust other parameters such as the total number of subcarriers, center frequency, bandwidth of each subcarrier, etc, making it a very flexible and robust cognitive radio node. The rest of the paper is organized as following: Section II briefly reviews the overlay/underlay cognitive radio and SMSE framework, Section III describes the hardware and software we used, namely the USRP boards and GNU radio platform. Section IV describes spectrum sensing and the non-contiguous overlay/underlay cognitive radio we implemented. Conclusion is given in Section V. II. SMSE F RAMEWORK AND OVERLAY /U NDERLAY C OGNITIVE R ADIO Previous work provides a general analytic framework for SMSE signals that accommodates multi-carrier, CR-based waveforms. Specifically, an arbitrary CR waveform can be expressed in terms of its amplitude (A), phase (Θ) and frequency (F) characteristics. These three factors aid in SMSE waveform design through six design variables, namely Data modulation (d), Code (c), Window (w), Orthogonality (o) and two frequency allocation variables. Considering Nf total frequency components, the coding c = [c1 , c2 , . . . , cNf ], ci ∈ C, data modulation, d = [d1 , d2 , . . . , dNf ], di ∈ C, and windowing, w = [w1 , w2 , . . . , wNf ], wi ∈ C vectors account for component-by-component amplitude and/or phase variations. A phase only variable ø = [o1 , o2 , . . . , oNf ], oi ∈ C is used for
orthogonality between symbol streams and facilitate multiple access. The analytic SMSE framework development begins by considering data, code and window variables. The mth frequency component of the k th symbol is given by Sk [m] = cm dm,k wm ej(θdm,k +θcm +θwm ) ,
(1)
where m = 0, 1, ..., NF − 1 is the frequency index and cm , dm are magnitude and phase design variables. The expression in (1) is next modified to incorporate frequency and orthogonality variables. Frequency component selection is a function of two factors, including an available variable a = [a1 , a2 , . . . , aNf ], ai ∈ {0, 1} and a use variable u = [u1 , u2 , . . . , uNf ], ui ∈ {0, 1}. Given an Nf -point fast Fourier transform (FFT) process, Nf frequency components or spectral bands are available for waveform design. It is important to note that the frequency assignment variable takes on binary values 0 or 1 indicating the spectrum availability for secondary users. As a direct result, this pool of frequencies is reduced by component selection to create a number of CR available frequencies and usable frequencies. The mth component of the k th CR symbol corresponds to Sk [m] = am um cm dm,k wm ej(θdm,k +θcm +θwm +θom,k ) , (2) where the product ai ui ∈ {0, 1}. The discrete time domain SMSE waveform is obtained by taking the Inverse Discrete Fourier Transform (IDFT) of (2) according to
sk [n] =
f −1 NX
1 am um cm dm,k wm Re Nf m=0 o
ej(2πfm tn +θdm,k +θcm +θwm +θom,k )
(3) ,
where tk ≤ tn ≤ tk + T , fm = fc + m∆f , T is the symbol duration and ∆f = 1/T is the frequency resolution. The SMSE framework provides a unified expression for generating and implementing a host of multi-carrier type waveforms (e.g., OFDM, MC-CDMA, CI/OFDM, TDCS, etc) and satisfies current CR goals of exploiting unused spectral bands. However, it does not exploit underused spectrum. This section re-visits the original SMSE framework development and the frequency assignment variables to exploit both unused and underused spectrum to generate both overlay-CR and underlay-CR type waveforms. Figure 1 illustrates a conceptual view of the unused and underused spectrum utilization using an arbitrary interference threshold (IT). IT is assumed to be a limit set forth by the primary users based on the measured power spectrum density in a given bandwidth. Two cases of under utilized spectrum are demonstrated: 1) when the spectral assignment is based on a binary decision the bands adjacent to the primary users are unavailable to overlay-CR users and 2) primary users bands below the IT are unavailable to CR users as well. A soft decision CR (SDCR) will be able to exploit these underused
Fig. 1. Identification of primary users, unused and underused spectral region
frequency bands to improve spectral efficiency and increase channel capacity. To support the envisioned SDCR system, the original SMSE framework is extended to account for both unused and underused frequency bands. The proposed SD-SMSE framework is first illustrated using Fig. 2 and Fig. 3, then the design variables are re-defined to extend the SMSE expression to account for both unused and underused spectrum. Figure 2a and Fig. 2b show how the current CR framework identifies the used and unused spectrum based on binary decisions. Figure 2c shows the weighted spectrum estimation resulted from spectrum sensing block in Fig. 3. In Fig.3, IT represents the interference threshold, PU and SU are the accumulated signal strengths from all the primary users and all the secondary users respectively, and N represents the total observed noise. The weighted spectrum estimate (WSE) (a) is further processed taking into account inputs from the IT estimator, primary users, other secondary users requirements and channel conditions. Specifically, the weighted spectrum estimate provides a metric of the allowable transmission power density at each and every frequency component in the entire bandwidth. Hence, the WSE divides the entire bandwidth into unused (u) and underused (b) frequency components and both the unused and underused spectrum can be exploited. Notice in Fig. 2 that different underused frequency components have different allowable CR transmission power densities. It is envisioned that a CRbased SDR will have the option to choose an overlay-CR, underlay-CR or hybrid overlay/underlay waveform to improve performance based on the scenario, situation and need. The first step in SD-SMSE framework development is to reexamine the design variables in the original SMSE framework. For the SD-SMSE development, frequency related factors are termed primary variables while amplitude and phase related factors are termed secondary variables. Since the objective here is to optimize the spectrum usage, only frequency components related design variables are considered. From this point forward the SD-SMSE framework development is based on the
½ bm =
0 am
am = 1 am 6= 1
(8)
for m = 0, 1, · · ·, Nf − 1. Note that when am = 1 the value of bm = 0. This is because when am = 1 the spectral component is being unused and accounted for in the assignment of um . It is obvious that if one frequency component is underused it cannot also be counted as unused and vice versa, i.e., um = 0 if bm > 0 and bm = 0 if um = 1. The remaining waveform design variables, i.e., code (c), data (d), window (w) and orthogonality (o), remain unchanged from the original SMSE framework. Applying all these design variables, the mth component of the k th data symbol of the SD-SMSE can be expressed as Fig. 2. Spectrum parsing using weighted spectrum estimation in realization of SD-SMSE waveform.
Sk [m] = am cm dm,k wm ej(θdm,k +θcm +θwm θom,k ) (9) ( j(θdm,k +θcm +θwm θom,k ) um cm dm,k wm e am = 1 = j(θdm,k +θcm +θwm θom,k ) bm cm dm,k wm e am 6= 1 The expression in (9) can be decomposed into unused and underused SMSE waveform representing the new SDCR architecture shown in Fig. 3. Applying the IDFT to (9) results in the discrete time domain waveform given by: f −1 NX 1 sk [n] = Re am cm dm,k wm Nf m=0 o ej(2πfm tn +θdm,k +θcm +θwm +θom,k )
Fig. 3.
Block diagram representation of SD-SMSE framework [?].
scenario depicted in Fig. 2. As shown in Fig. 2c, the weighted spectrum estimate represents all frequency components which can be utilized for secondary user applications. It is represented by variable a with the range changed from binary values (hard decision) to real values (soft decision), i.e., a = [a0 , a1 , . . . , aNf −1 ], 0 ≤ am ≤ 1 .
(4)
From the weighted spectrum estimate a, the unused spectrum vector u can be derived as u = [u0 , u1 , . . . , uNf −1 ] ,
(5)
1 if am = 1 m = 0, 1, · · ·Nf − 1 0 else
(6)
where, ½ um =
The original SMSE hard decision CR design transmits over the unused spectrum specified by u. Now introducing a new design variable b to account for the underused spectrum, b = [b0 , b1 , . . . , bNf −1 ] , where,
(10)
(7)
f −1 NX 1 um cm dm,k wm sk [n] = Re Nf m=0 o ej(2πfm tn +θdm,k +θcm +θwm +θom,k ) f −1 NX 1 + bm cm dm,k wm Re Nf m=0 o ej(2πfm tn +θdm,k +θcm +θwm +θom,k )
(11)
where the first summation in (11) represents the unused frequency components and the second summation accounts for underused frequency components. The SMSE expression in (10) was demonstrated by applying it to a number of OFDM based multi-carrier signals. The process of generating these waveforms can be viewed as a two step approach 1) generating the frequency related primary variables and 2) applying the secondary variables such as the code code, data modulation, windowing and orthogonality to the frequency vector. Since the SD-SMSE only focused on manipulating the primary variables, all the OFDM based multicarrier modulations expression such as NC-OFDM, NC-MCCDMA, NC-CI/MC-CDMA and NC-TDCS are applicable to both overlay-CR and underlay-CR scenarios.
III. GNU S OFTWARE D EFINED R ADIO AND USRP We have chosen GNU SDR as the platform for this CR implementation. GNU Radio, founded by Eric Blossom, is an open source Python-based architecture for building SDR project. The typical TX/RX path for GNU SDR is shown in Fig.4. The USRP, the brainchild of Matt Ettus, is the hardware solution for GNU Radio. USRP connects the computer and the real Radio Frequency (RF) world perfectly via flexible USB interface. A block diagram of USRP is shown in Fig. 5.
Fig. 4.
Fig. 5.
Typical TX/RX Path for GNU SDR
two transmit daughterboards and two receive daughterboards. RF front-end is implemented on the daughterboard. There are a variety of daughterboards that are available to work at different frequency band.[9] In this paper, USRP daughterboards transceiver FLEX400 and receive-only TVRX are used. IV. SMSE BASED C OGNITIVE R ADIO I MPLEMENTATION A. Spectrum Sensing and Adapting The designed CR observes the band from 400MHz to 500MHz, which is the working range of USRP daugterboard FLEX400. From the observation, it determines which TV channels are occupied, which are unoccupied, and what other transmissions are working on this band. The unoccupied bands can be used by the CR. Because this CR applies ’Listenbefore-talk’ (LBT) concept, estimation of the power spectral density and some TV signal characteristics are useful for power/energy based spectrum sensing. Before sensing the spectrum, it is necessary to know the analog TV and digital TV signal characteristics. The analog/digital TV bandwidth in the United States is 6 MHz. The analog TV follows NTSC standard. The amplitude-modulated video signal uses the carrier at 1.25 MHz above the lower edge of the channel. The quadrature-amplitude-modulated color signal uses the carrier approximately at 3.58 MHz above the video carrier. The frequency-modulated audio signal uses the carrier at 4.5 MHz above the video carrier. Fig. 6 shows the PSD of a typical NTSC signal. [10] The scale of the y-axis is not normalized. The video, color, and audio carriers are detected obviously. The digital TV follows the ATSC standard. 8level vestigial sideband modulation scheme is applied, and the modulated signal uniformly occupies almost the entire 6 MHz TV channel. A pilot tone is located at around 310 kHz above the lower edge of the channel. Fig. 7 shows the PSD of a typical ATSC signal. [10] The scale of y-axis is not normalized. The pilot carrier near the lower edge of the TV channel shows clearly.
Universal Software Radio Perepheral Diagram
The USRP consists of a motherboard that provides up to four 12-bit ADCs at 64M samples/sec, four 14-bit DACs at 128M samples/sec, a million gate-field programmable gate array (FPGA) and a programmable USB 2.0 controller. ADCs and DACs are the bridge between the continuous analog signals and the discrete digital samples. All the ADCs and DACs are connected to the FPGA. The functionality of FPGA is to perform the high-speed general purpose operations, such as digital-up conversion and digital-down conversion, decimation, and interpolation, and to reduce the data rate feeding to USB2.0. FPGA connects to a USB2.0 interface chip. Then, all of the waveform-specific processing are performed on the host CPU. Each fully populated USRP motherboard can support
Fig. 6.
Power Spectrum Density (PSD) of a Typical NTSC Signal in US
Fig. 9.
Fig. 7.
Power Spectrum Density (PSD) of a Typical ATSC Signal in US
Fig. 8.
Spectrum Sensing via USRP
obtain the time domain SMSE symbols. Cyclic prefix (CP) is added to the time domain signal to avoid residual InterSymbol-Interference (ISI) from the previous SMSE symbols. Digital-to-analog converter (DAC) converts baseband digital signal to analog signal which fed to the RF front-end. The RF front-end upconverts the signal to the RF frequency for transmitting through the antenna.
The USRP Pathway of Receiving Signal
A power spectrum density (PSD) estimation is performed on the USRP received signal samples. The USRP pathway of receiving signal is shown in Fig. 8. First, the source sample stream is converted into vector with length 128 by streamto-vector block, then the vector is decimated to a rate for real-time processing. A Blackman-Harris window is applied to every vector with length 128 to minimizes the side-lobes of the spectrum.. A complex Fast Fourier Transform (FFT) is taken, then averaging the magnitudes of each bin is performed in the average bin block. Last block analysis PSD estimation. The spectrum for band 400 MHz to 500 MHz is sensed by USRP, shown in Fig. 9. The occupied bands, CH16, CH18 and an emission at 448 MHz, are sensed, as well as the unoccupied bands. Then, our cognitive radio adapts itself to the environment using unoccupied bands to communicate without any interference with the current users within the band. B. SMSE Based Adaptive Transmission A SMSE based multi-carrier transceiver block diagram is shown in Fig. 10. Channel coding is used to reduce the degradation of system performance due to the frequency selective fading. Interleaving randomizes the occurrence of bit errors and introduces system immunity to burst errors. Then, coded and interleaved data is mapped to the constellation points to obtain data symbols. Through S/P block, the serial data symbols are converted to parallel data symbols which are fed to the Inverse Discrete Fourier Transform (IDFT) block to
Fig. 10.
Block Diagram of SMSE Transceiver
On the receiver side, the received signal passes through a bandpass noise rejection filter, and is downconverted to baseband by the RF front-end. The analog signal is digitized by analog-to-digital converter (ADC). After synchronization, CP is removed. Converted parallel symbols go through Discrete Fourier Transform (DFT) block to get the frequency domain symbols. Then, symbols are demodulated, deinterleaved, and decoded to obtain the transmitted information bits. To implement adaptive interference avoidance, we implemented non-contiguous SMSE waveform. Specifically, as shown in Fig. 11, the spectrum sensing engine detects which subcarriers are occupied by the primary users and other narrowband transmissions. Next, by switching off those subcarriers, we implement a non-contiguous SMSE waveform which only transmits over spectrum holes (white space). The system block diagram of the adaptive interference avoidance CR is shown in Fig. 12. The webcamera with microphone captures the real-time video and audio signals. Video and audio codec is implemented by VideoLAN, which is a complete software solution for media streaming developed under the GNU General Public License. The video and audio streaming is truncated to
subcarrier that narrow band emission occupies. Fig. 13 and Fig. 14 are the state diagrams of the transmitter and the receiver.
Fig. 11.
Contiguous SMSE vs. Non-contiguous SMSE
Fig. 12. NC-SMSE Based Adaptive Interference Avoidance CR System Block Diagram
the frames which are mapped to the constellation points to obtain modulated symbols. The known preambles are inserted for frame detection and synchronization. The serial data are convert to parallel which fed to the IFFT block to get the time domain SMSE symbols. CP is added on it to avoid ISI from the previous SMSE symbols. After scaling, the data are fed to the USRP via USB to convert to analog signal and transmit after RF front-end upconverting. The receiver USRP downconverts the received signal and digitizes it. After passing through a channel filter, SMSE symbol synchronizes using PN correlation [12]. Next, CP is then removed. The serial symbols are converted to parallel which pass through the FFT block to obtain the frequency domain symbols. SMSE preamble correlation detector also estimates the input signal frequency offset. Then build the frame from the demodulate symbols. VideoLAN decodes the video and audio signal and real-time play it. Transmitter USRP senses the spectrum, adapts itself to the environment, finds an unoccupied band to communicate. Next, we employ a control channel to inform the receiver important parameters necessary to decode the data. Such parameters include center frequency, total number of subcarriers, which subcarriers are tuned off, etc. Then the non-contiguous SMSE signal is transmitted. It is important to note that the spectrum sensing keeps sensing the environment while the CR is operating. Hence, if the spectrum sensing engines senses any primary user (or other narrowband transmission) becomes active in the band, the transmitter will dynamically adapt to the environment quickly. If the bandwidth of the primary user is larger than 2 MHz, the transmitter will hop to the next available unoccupied band; otherwise, it turns off the
Fig. 13.
Transmitter State Diagram
Receiver USRP listens to the control channel. Once it obtains the control signal, it will tune itself ready to receive. In this system, more than one receivers can be operating simultaneously, forming a cognitive radio network.
Fig. 14.
Receiver State Diagram
Fig. 15 shows the spectrum captured by the USRP board when the cognitive radio is operating over one contiguous spectrum hole. Evidently, if no narrow band transmission is existing in the transmission band, the designed CR will transmit a SMSE symbol. Fig. 16 shows the non-contiguous SMSE symbol spectrum. It is evident from this figure that the cognitive radio operates over two non-contiguous bands by transmitting a non-contiguous SMSE symbol by turning off a few subcarriers around the frequency of 430 MHz of the primary user. Fig. 17 shows when the primary user jumps to
a different frequency, the cognitive radio dynamically changes the waveform by turning off some other subcarriers to avoid interference dynamically. As a direct result, the real time video and audio transmission supported by the cognitive radio link experiences no interruption.
Fig. 15. SMSE
real time video and audio transmission over the cognitive link is supported seamlessly. Fig. 19 illustrates the spectrum of a non-contiguous SMSE symbol, which takes advantage of multiple spectrum holes, captured by Agilent spectrum analyzer, confirming the effectiveness of our non-contiguous frequency band transmission.
Adaptive Interference Avoidance CR Demonstration - Contiguous
Fig. 18.
CR Demonstration
Fig. 16. Adaptive Interference Avoidance CR Demonstration - Noncontiguous SMSE 1
Fig. 19. NC-SMSE Spectrum Observed by Spectrum Analyzer Taking Advantage of Multiple Spectrum Holes
Fig. 17. Adaptive Interference Avoidance CR Demonstration - Noncontiguous SMSE 2
Fig. 18 shows the setup of the interference avoidance cognitive radio. As can be seen on the two laptop screens,
V. C ONCLUSION AND F UTURE W ORK In this paper, we extended our previous work to implement and demonstrate an overlay cognitive radio via SMSE framework using USRP software defined radio boards and GNU radio software. The implemented interference avoidance cognitive radio quickly senses the spectrum, learns to adapt to the environment, adaptively transmits and receives real-time video without interference with the primary users on the band. When the primary user transmission changes, the cognitive radio dynamically changes its transmission to avoid any harmful interference to the primary users. The implementation and demonstration of a SMSE based underlay cognitive radio and hybrid overlay/underlay cognitive radio will be presented in future publications.
R EFERENCES [1] J. Mitola. ”Cognitive Radio - An Integrated Agent Architecture for Software Defined Radio”, Ph.D. Dissertation, Teleinformatics, Royal Institute of Technology - Sweden, 2000. [2] Blossom, Eric. ”Software Radio.” Comsec. Blossom Research, LLC. Feb. 2008 http://www.comsec.com/ [3] Boyd, Susan, and Weber Shandwick. SDR Forum. Software Defined Radio Forum. Feb. 2008 http://www.sdrforum.org/ [4] V. Chakravarthy, Z. Wu, M. Temple, and F. Garber, “Novel Overlay/Underlay Cognitive Radio Waveforms Using SD-SMSE Framework to Enhance Spectrum Efficiency - Part I: Theoretical Framework and Analysis in AWGN Channel,” IEEE Transactions on Communications, vol. 57, no. 12, December 2009. [5] V. Chakravarthy, Z. Wu and M. Temple, “Novel Overlay/Underlay Cognitive Radio Waveforms Using SD-SMSE Framework to Enhance Spectrum Efficiency - Part II: Analysis in Fading Channel,” to appear in IEEE Transactions on Communications [6] V. Chakravarthy, Z. Wu, A. Shaw, F. Garber, ”Cognitive Radio Centric Overlay/Underlay Waveform,” IEEE DySPAN 2008, 2008, Chicago [7] H. Arslan. Cognitive Radio - Software Defined Radio and Adaptive Wireless Systems, Springer-Verlag New York, LLC, 2007. [8] T.A. Weiss and F.K. Jondral. ”Spectrum pooling: an innovative strategy for the enhancement of spectrum efficiency”, IEEE Communications Magazine, Vol. 42, No. 3, March 2004, pp. S8 - S14. [9] R. Zhou, X. Li, R. Husnay, Z. Wu, and S. Hong. ”Low-Cost Wideband Electronic Emission Detection via Software Defined Radio”, IEEE NAECON, July 2008 [10] S. Shellhammer, A. Sadek, and W. Zhang. ”Technical Challenges for Cognitive Radio in the TV White Space Spectrum”, Qualcomm, 2009. http://ita.ucsd.edu/workshop/09/files/paper/paper 1500.pdf [11] A. K. Jain. Fundamentals of Digital Image Processing. Prentice Hall, Upper Saddle River, NJ, 1989. [12] T.M. Schmidl and D.C. Cox. ”Robust Frequency and Timing Synchronization for OFDM”, IEEE Trans. Communications, vol.45, No. 12, 1997