Spectral Occupancy and Interference Studies in support of Cognitive Radio Technology Deployment

Transcription

1 Spetral Oupany and Interferene Studies in support of Cognitive Radio Tehnology Deployment Dennis A. Roberson and Cynthia S. Hood Computer Siene Department Illinois Institute of Tehnology Chiago, Illinois Joseph L. LoCiero and John T. MaDonald Eletrial and Computer Engineering Department Illinois Institute of Tehnology Chiago, Illinois Abstrat This paper desribes the high value of ognitive radio tehnology and haraterizes the opportunity spae in four distint lasses. A Chiago-based spetrum oupany study illustrates the opportunity showing that 8.6% of the spetral apaity is unused. A set of spetral signatures is presented for ommon devies in the unliensed frequeny band with the view that this tehnique an be widely deployed aross the spetrum. The limitations of urrent network simulation tools in an interferene environment are identified. Finally, the paper briefly disusses several of the non-tehnology related issues that impat the deployment of ognitive radio tehniques. Keywords ognitive radio; interferene; spetrum oupany; wireless ommuniations; wireless networks I. INTRODUCTION We live in a world of spetrum sarity as exhibited by the billions of dollars being paid to obtain the right to exlusively utilize relatively modest portions of the eletromagneti spetrum. At the same time we are seeing a virtual explosion in the requirement for spetrum based on the rapidly growing number of new appliations, the rate of deployment of the more suessful appliations (ellular, Wi-Fi wireless fidelity, Bluetooth, et.), the inreasing utilization of these tehnologies (hours per day), and the performane expetations of the devies supporting these approahes (Mb/se). Colletively these trends are exponentially inreasing the demands for, and value of, our finite spetral resoures. This has reated an alarming state of spetrum sarity. At the same time, the limited snap-shot spetrum oupany studies that have been undertaken [1 7] suggest that there is still an abundane of real spetrum available and that the urrent issue is more related to our fixed, time independent approah to spetrum alloation, than it is to any real lak of spetrum from a timespae ontinuum perspetive. To address this issue, over the last deade or so, the Federal Communiations Commission (FCC) has established three fundamental diretions to satisfy the insatiable demand for additional spetrum, namely, unliensed spetrum, underlays (e.g., ultra-wideband (UWB)), and overlays (i.e., ognitive radio). Of these three issues, the unliensed approah is widely deployed and is generally viewed as a spetaular suess. The only storm louds on the horizon are the ever inreasing quality of servie hallenges assoiated with the over-deployment of tehnologies into this spae, i.e., the proverbial Tragedy of the Commons. This situation is partiularly pronouned in dense usage areas like our major tehnology based universities (e.g., Illinois Institute of Tehnology (IIT)). The underlay approah is still at an early stage and has unfortunately reeived mixed reviews based on among other things, the diffiulties experiened in the unsuessful standards efforts assoiated with the tehnology. The overlay approah requires the most researh and development effort to suessfully provide a deployable system. The approah also offers the greatest promise for the future sine it has an opportunity spae that overs all, or at least most, of the urrently alloated spetrum. The fundamental breakthrough that this tehnology offers is the exploitation of the time dimension to enable heterogeneous usage of previously alloated homogeneous signal spae. To optimally take advantage of the overlay based opportunity, there are several key tehnologies that need to be utilized. This inludes the use of dynami frequeny agile radios whih enable the sensing of available spetrum whih an be used for signal transmission. It also inludes the related ognitive tehnology apability to be rapidly pre-empted when a primary signal is deteted. Finally, when these soures are identified (espeially inumbent transmitters) agile tehnology enables a rapid move to unoupied bands to ontinue data transmission. The use of multi-band, diretional antenna tehnology and espeially array-based beam forming antennas is also of signifiant benefit in this arena sine the spatial region impated by the transmission is dramatially redued. The enhanement in the seletivity of reeiver tehnology used by both ognitive radios and traditional radios is also of extreme importane sine it will enable improved tolerane to both out of band and alternative modulation shemes. Within wireless networks, the most ritial element is the improved understanding of the nature of interferene at a fundamental level. As inreasing levels of understanding are attained, there is a related need to establish various strategies to radially enhane the ability to mitigate this interferene. This area has beome the entral researh fous at IIT s Wireless Interferene Laboratory (WIL). This paper desribes some of the insights that have been derived through the various researh thrust being undertaken by this Laboratory. 1

2 The next setion of the paper details the various lasses of ognitive radio opportunities. It also addresses the hallenges assoiated with establishing a theoretial apaity limit for a presribed spatial region, as well as the diffiulties in approahing these limits. Setion III fouses on providing spetral signatures of ommon signals found in unliensed bands and power in adjaent hannels. This provides a basis for similar efforts aross the alloated ommuniations spetrum. Setion IV disusses the value and the limitations of urrent network simulation apabilities in addressing the impat of interferene on network performane. Setion V introdues several of the non-tehnologial deployment limitations for ognitive radio tehnology. The paper is summarized in the final setion. II. OPPORTUNITIES, CHALLENGES, AND COMMUNICATION LIMITS The overlay / ognitive radio opportunity falls in at least four fundamentally different lasses. First there is the opportunity presented by spetrum that is assigned but rarely if ever utilized within a speifi geographi region. This is the proverbial low hanging fruit, where only minimal ognitive apabilities are required for exploitation. The seond opportunity is that presented by fixed signals, suh as television transmission where the spetrum appears to be fully utilized, but beause of the nature of the signal, has preditable, timebased opportunities for the insertion of data arrying transmissions. This opportunity is also very aessible, and given the large amount of prime spetrum alloated to television transmission, this opportunity should be exploited with high priority. The third opportunity is that presented by spetral regions that are infrequently utilized, where the ognitive radio must be partiularly apable in deteting inumbent transmissions when they our, and partiular strong in rapidly easing transmissions, or moving the transmissions to an unoupied hannel. Military and ivilian government bands are obvious examples of this kind of opportunity. The final lass is the regions where the spetrum is relatively well-utilized, but still has some apaity available, espeially in ertain period of the day, perhaps during lunh hour, or at night. In this ase, the ognitive radio apabilities will be highly stressed to insure that the radios do not interfere with inumbent spetrum users, and still deliver value as a ommuniations medium. Given the signifiant variation in the harateristis of these four ases, different measurement tehniques are required to disover the level of opportunity represented by eah. To support our investigations of available spetrum for the first ategory of relatively stati ognitive radio tehnology opportunity, we have undertaken passive monitoring of the radio spetrum in Chiago s business distrit (The Loop) from IIT s Chiago based ampus. In this snap-shot survey, we measured radio frequeny (RF) energy over a broad range of frequenies (30 MHz to 3 GHz) over a 48-hour time period. The study was undertaken on a Wednesday through Friday in November whih was deemed to be a relatively typial period of time during the fall of 005. These results are summarized in Table 1 and seletively illustrated in Figs. 1 and. Start Freq (MHz) Stop Freq Bandwidth (MHz) (MHz) Spetrum Band Alloation NYC Day 1 Spetrum Fration Used NYC Day Spetrum Fration Used NYC Avg Spetrum Fration Used NYC Oupied Spetrum (MHz) NYC Average Perent Oupied Chiago Day Chiago Day 1 Spetrum Spetrum Fration Fration Used Used Chiago Avg Spetrum Fration Used Chiago Oupied Spetrum (MHz) Chiago Average Perent Oupied PLM, Amateur, others % % TV -6, RC % % Air traffi Control, Aero Nav % % Fixed Mobile, amateur, others % % TV % % Maritime Mobile, Amateur, others % % Fixed Mobile, Aero, others % % Amateur, Radio Geoloation, Fixed, Mobile, Radioloation % % TV % % TV % % TV % % TV % % Cell phone and SMR % % Unliensed % % Paging, SMS, Fixed, BX Aux, and FMS % % IFF, TACAN, GPS, others % % Amateur % % Aero Radar, military % % Spae/Satellite, Fixed Mobile, Telemetry % % Mobile Satellite, GPS, Meteorologiial % % Fixed, Fixed Mobile % % PCS, Asyn, Iso % % TV Aux % % Common Carriers, Private Companies, MDS % % Spae Operation, Fixed % % Amateur, WCS, DARS % % Telemetry % % U-PCS, ISM (Unliensed) % % ITFS, MMDS % % Surveillane Radar % % Total Total Available Spetrum Average Spetrum Use (%) 13.1% 17.4% Table 1. Spetrum Utilization Survey onduted November 005, from IIT s Researh Tower in Chiago, IL In determining spetrum utilization, several riteria were used: the power of the signal present; the duration in time or the duty yle of the signal; and the ritial nature of the signals ontained in the bands. Some of the bands were highly utilized, others were used sporadially, and still others weren t used at all. The TV bands were a good example of high utilization with a 50% plus spetrum oupany. This results from the fat that TV hannels are well oupied in Chiago, and the signals are always broadasting. At the same time, nearly half the available hannels are not oupied to insure separation from transmitting hannels and to avoid overlap with neighboring metropolitan TV markets (see Fig. 1). While it an be argued that the information ontent in TV bands is low (some would say near nil), and they do generates signifiant revenue for the station owners.

3 Fig to 16 MHz, 4 hr. period starting on Nov. 16, 005 The XM and Sirius satellite radio bands are fully utilized for similar reasons. There are however adjaent satellite radio bands whih are at present unused. Other spetral areas that are highly utilized inlude the various ell phone bands, whih are harateristially rowded in suh a densely populated area. Bands whih appear to be under-utilized are the amateur bands, various military and ivilian government bands, and to a lesser degree, the unliensed bands. Few people have amateur lienses or the equipment to utilize these under-served bands. Military bands are little used in a region like Chiago with relatively few military installations. In normal times (i.e., no large sale emergeny or major high profile event), ivilian government bands suh as polie, fire, and emergeny servies bands also display a relatively low level of utilization. Unliensed bands, on the other hand are being inreasingly utilized with produts and tehnology that gives rise to new data ommuniation servies, like wireless networking (Wi-Fi, Bluetooth and WiMAX), and wireless sensing (Zigbee) readily available to the publi. Still the urrent utilization level for the unliensed bands would allow for a reasonable level of ognitive radio ommuniations, espeially during the late night time hours (e.g., see Fig. ). Some bands appear under utilized but may not be available. Satellite ommuniations and positioning systems appear to be under-utilized due to their low power / low noise requirements. Fig to 500 MHz, 4 hr. period - Nov. 16/ With some knowledge of the loation of transmitters and their intended overage area, one an either improve the performane of reeivers to their intended target, or avoid transmitters that may interfere with the target transmitter. A small sale example, where knowledge of the loation of transmitters and their intended overage areas is needed is a Wi-Fi wireless network operating in the Industrial, Sientifi and Medial (ISM) band. Beause these systems are onstrained in power and are short in wavelength, they are onfined to small geographi areas, usually less than 100 feet. Knowledge of the existing overage is key to adding new aess points and expanding the apaity of new networks in an area already served by one or more wireless networks. To this end, we have onduted site surveys on the ampus of IIT to map out overage of the established wireless network, and determine where there is room for additional apaity. As an indiation of how rowded the ISM band has beome in ertain geographi areas, onsider the network map illustrated in Fig. 3. The map shows Galvin Library on the ampus of the IIT. A site survey was onduted with a global positioning system and a network monitoring devie. The dots indiate the estimated loations of aess points based on the loation of the greatest reeived signal strength. On the afternoon of August 31, 006, 63 unique aess points were identified in an area roughly the size of a football field. Of the identified beaons, ten were ative probes, six were ad-ho networks, and the rest were open aess points. Of the aess points, only 1 used wired equivalent privay (WEP) signal enryption and 30 were broadasting on hannel 6, the most ommonly used hannel. This shows that the field is rowded in partiular loations, like university libraries, and 3

4 yet even here there is room for improvement. In this ase, under-used hannels are identified and heavily utilized networks an be flagged and avoided. By loating the soure of traffi ongestion, appropriate mitigation shemes an be deployed. This opens an opportunity for ognitive radio to exploit otherwise poorly designed and implemented wireless networks. Fig. 3. Wi-Fi aess point site survey of IIT s Galvin Library Ultimately, it is important to understand the maximum apaity, i.e., the theoretial limit on data transmission apaity for a give spae-time-frequeny band. One this is understood, the researh, and later development and deployment hallenge for real wireless systems is to ome as lose to the theoretial limit as possible. If this is done onsistently aross geographial spaes (e.g., large urban environments), times (e.g., the normal business day), and frequenies (e.g., those spetral regions most eonomially addressable by today s available eletroni tehnologies), the available wireless data arrying apaity an be radially enhaned to meet the ommuniations needs and desires of people around the globe. This, of ourse, is a very hallenging undertaking given the diversity of physial environments to be dealt with, eah with their unique transmission harateristis and hallenges. Even more diffiult is sorting out what onstitutes an optimal seletion of spatial regions to be studied, and where and how to draw the appropriate boundaries. Even worse, these environments are very dynami based on the season of the year (hanges in foliage, power usage, lighting, et.), hanges in the density and diretion of human and automotive transportation, and the onstrution and demolition of buildings and other physial struture. Even more diffiult, the real environment generally does not obey analytial assessments sine there are many seondary effets that are always interating with a given environment. These inlude the fat that transmissions in one spatial region do interat and interfere with the transmissions in adjaent regions, espeially near the boundary dividing the regions. There are also many unintentional transmitters (e.g., PCs, power tools, automobiles, trains, eletri lights, et.) reating dynamially hanging interferene in any given environment. Even the natural environment reates a dynami noise floor that must be dealt with in sorting out the ultimate apaity limits of a given environment. III. INTERFERENCE TEMPERATURE / SIGNATURES / MODELS To understand a speifi environment suffiiently to suessfully deploy ognitive radio, it is important to identify the transmitters that are ative in the environment. With this identifiation in hand, the ognitive radio an adjust its spetral usage strategy to optimally operate in the environment. To support this approah, a yli researh strategy has been pursued omposed of analytial modeling, laboratory-based experimentation, and physial and network level simulation. The following desribes some of the early suesses in this arena fousing on ommon devies operating in the ISM frequeny band. A wireless loal area network (WLAN) is inherently interferene limited, partiularly in the unliensed frequeny bands where users are vying for a lean piee of the temporal, spetral, spatial domain, in a rather unonstrained manner. The need for effetive ognitive radio methodologies is unmistakably lear, partiularly when one begins to delineate the multipliity of interfering emitters and the array of modulations shemes found in both the unliensed frequeny regions and the frequeny bands that are dediated to liensed ommuniations. The throughput and apaity of a WLAN an radially hange when an aess point (AP) is near an interfering transmitter, and when non-data arrying interferes are present in the frequeny band of the RF information bearing signal. This interferene is partiularly prevalent in the unliensed ISM bands. The ISM-900 band is MHz; the ISM-.4 band is GHz; and the ISM-5.8 band is GHz. To somewhat alleviate gross interferene, the FCC plaes a transmitter signal power limit in the ISM bands of 1 watt, with a maximum effetive isotropi-radiated power (EIRP) of 4 watts in the ISM-900 and ISM-.4 bands, but a maximum EIRP of 00 watts in the ISM-5.8 band. The FCC also defines Unliensed National Information Infrastruture Bands (UNII) that an ontribute additional eletromagneti radiators. The UNII Indoor band is GHz; the UNII Low Power band is GHz; and the UNII/ISM band is GHz. The Indoor and Low Power UNII bands have muh lower maximum transmitted signal power speifiations: 50 mw and 00 mw, with 00 mw and 50 mw maximum EIRP, respetively. Commonly referred to as Wi-Fi, data arrying signals in the ISM-.4 band use three different modulation formats, well defined by the IEEE 80.11b and 80.11g standards. These modulation formats are also found in the IEEE 80.11a 4

5 standard, where the radiated signals are in the 5 GHz range. The three formats are: diret sequene spread spetrum (DSSS) using an 11-bit Barker ode as bipolar amplitude modulation (AM); DSSS using 8-bit omplementary ode keying as phase modulation; and orthogonal frequeny division multiplexing (OFDM) with a yli prefix ode (CPC). The interferene aused by ompeting WiFi signals an be lassified as homogeneous modulation disturbane or heterogeneous modulation disturbane, if the modulation format is the same or different than the vitim signal. There are many other types of data arrying emitters with modulation formats quite different from the ommon Wi-Fi types. The original standard speified a frequeny hopped spread spetrum (FHSS) signal using Gaussian frequeny shift keying (GFSK) as its RF waveform with binary or quaternary data. Common ommerial devies use the Bluetooth speifiation, whih alls for a FHSS signal where GFSK is employed with a small modulation index and a peak data rate of 1 Mbps. Bluetooth applies time division duplexing (TDD), where separate 65 µs time-slots are used to transmit and reeive pakets of data. There are typially 79 arrier frequenies in the ISM-.4 band, spaed 1 MHz apart, and the hopping rate is 1.6 khops/seond. Intentional interferers that orrupt the ISM bands are UWB signals, where a pulse position modulation (PPM) temporal format would inlude a monopulse waveform, with an effetive width of ns, allowing the UWB spetrum to fall in any of the unliensed bands. Cordless phones are yet another ommon soure of intentional ommuniators that generate interferene in the unliensed bands. There are DSSS and FHSS phones that transmit binary data representing digitally enoded voie signals, in additional to wideband frequeny modulated (FM) phones that transmit analog speeh signals. Devies suh as last mile wireless internet servies, baby monitors, last mile wireless Internet servies, and even video game ontrollers produe intentional data arrying signals in the ISM bands. These different modulation formats all must be onsidered when employing ognitive radio methodologies to improve throughput and apaity in WLANs. The array of intentional data transmitting devies, with a plethora of modulation formats; ombine in an additive manner to inrease interferene power for any desired ommuniations signal in the unliensed bands. In addition, however, there are a number of unintentional emitters that further add to the noise. The most ommon unintentional emitter is the residential mirowave oven whose nominal frequeny of osillation is.45 GHz, almost in the middle of the ISM-.4 band. The magnetron is turned on and off with the 60 Hz line voltage. The mirowave oven signal is a form of swept FM that has distint on and off pulses ausing energy to spread throughout the ISM band, affeting many 80.11b and g hannels. Unwanted radiation that an affet Wi-Fi signals in a WLAN ome from: ar lamps, industrial heaters and dryers, and even household applianes, PCs and motors with high-speed eletronis used for the operation and ontrol of these devies. There are many metris that an be applied in a ognitive radio methodology. Although it is unlikely that one metri will be all that it is needed, interferene power, P I, and interferene temperature, T I, are related metris that serve to provide a baseline parameter for WLANs. The term interferene temperature [10], in kelvins is defined as T I = (P I + N ) / Bk, (1) where P I is the interferene power, in watts, in the bandwidth B, in Hz, N is the ambient noise power, in watts, in the bandwidth B, and k is Boltzmann s onstant, J/K. The use of interferene temperature or interferene power neessitates a speifiation of the bandwidth of interest, B. When dealing with signals in the ISM-.4 band, the speifiation of bandwidth is oupled with the hannel spaing and the frequeny mask defined by the standard. In the ISM-.4 band, there are a total of 14 hannels defined, although only the first 11 are used in North Ameria. The arrier frequenies start at 41 MHz (hannel 1) and are inremented by 5 MHz. Commonly, hannel 1, hannel 6 (537 MHz) and hannel 11 (46 MHz) are used in WLANs to provide a separation of 5 MHz. The frequeny mask defined for the Barker spread signals arrying data at 1 Mbps or Mbps, and the CCK spread signals with data at 5.5 Mbps or 11 Mbps, speifies that the signal is attenuated by at least 30 db from its peak power level at ±11 MHz from the arrier frequeny and attenuated by at least 50 db from its peak power level at ± MHz from the arrier frequeny. Following this spetral mask speifiation and the power spetral density (PSD) of the Barker and CCK spread signals, the bandwidth, B, is taken to be MHz. Our researh efforts in the area of wireless interferene are based on an iterative approah to model the soures of interferene. This approah ombines theoretial analysis, omputer simulation, and experimental measurements. The model is ontinually refined by repeating these steps in the proess. Outomes of this proess are analyti and graphial spetral signatures, in partiular the PSD of a set of signals, and a plot of interferene power as a funtion of ISM-.4 hannel number. It is antiipated that his approah an be applied to both the unliensed band and to liensed band ommuniations. The modulation formats used for the Barker and CCK spread signals an be unified into one form for the data rates of 1,, 5.5 and 11 Mbps, that is, y(t) = x(t) os(π f 0 t + θ (t)), () where f 0 is the arrier frequeny; x(t) has unit power; and the spread data is ontained in x(t) and θ(t) as speified by the data and spreading formats. Using this unified form, we have insured that y(t) has 1 watt of power. The analyti expression for the PSD of y(t) is given by S ( f ) ( 1/ ) [ S ( f f ) + S ( f f )] =. (3) x 0 x + Theoretial studies were undertaken that produed exat analyti expressions of the PSD of the Barker-spread signals and the CCK spread signals. Computer simulations of these signals and experimental emulation of these spread spetrum signals, along with laboratory measurement of their spetral signatures in a WLAN environment follow the initial analytial work. 0 5

6 A 1 Mbps signal with 11-bit Barker ode spreading employs Differential Binary Phase Shift Keying (DBPSK), while the Mbps Barker spread signal uses Differential Quaternary Phase Shift Keying (DQPSK). In both ases, an exat theoretial analysis, with no approximations, yielded the 11-bit Barker spread baseband signal PSD as [11] x ( f ) = Tsin ( ft )[ 1 ( /11) os( 4 i f T )] S π, (4) where the hip rate R = 1 / T is 11 Mps. When this signal was simulated and the PSD generated, there was a near perfet math between the theoretial result and the omputer simulated result, as shown in Fig. 3a and Fig. 3b. The 11-bit Barker spread baseband signal was emulated in the WIL using Teleommuniations Instrutional Modular Systems (TIMS) equipment by Emona Tehnologies with 1 kbps random data. The resulting PSD, show in Fig. 4, supports the theoretial and experimental results. The spetral signature of the Barker spread signal, with its harateristis nothes every 5.5 MHz, allows for identifiation of this type of signal, whih proves useful for interferene mitigation using adaptive arrays. Adjaent hannel interferene power and interferene temperature from a 1 Mbps Barker spread signal in hannel 1 was omputed, without and then with the spetral mask defined in the standard, as this signal served as an interferer into signals operating in hannels 1 through 11, using a MHz main lobe bandwidth. Composite interferene urves, after the spetral mask, are displayed in Fig. 5. These interferene results serve as a limitation metri for the operating of a WLAN in the.4 GHz unliensed band. A study of a 5.5 Mbps signal using an 8-hip CCK spreading signal and DQPSK modulation was onduted. Here, two bits were alloated for the CCK spreading ode and two bits were used for the modulation, as per the standard definition. A detailed, exat theoretial analysis produed the PSD of this signal, and omputer simulations supported the spetral signature. Although the definition of the CCK spreading signal is rather involved, the exat PSD took a very simple form. The expression given in Eq. (3) is again valid with the baseband PSD given by [1] S x ( f ) T sin ( ft ) 5 i= 1 =, (5) where the value of T is the same as given in the Barker spread ase. Thus the envelope of the PSD shown is Fig. 3, without the nothes, is the PSD for the 5.5 Mbps signal. As with the 1 Mbps and Mbps investigations, the adjaent hannel interferene power and temperature was omputed, using a MHz main lobe bandwidth without and with the spetral mask, and the results ompared to the Barker spread signals. This is all shown in Fig. 5. To omplement the theoretial and omputer simulation results with experimental verifiation, the CCK spread DQPSK signal was emulated in the WIL using a Field Programmable Gate Array (FPGA) system, and the signal onstellation displayed as a demonstration, along with the aptured PSD in a WLAN environment. Fig. 3a. Theoretial PSD of Barker spread signals Fig. 3b. Simulated PSD of 11-bit Barker spread signals Fig. 4. PSD of emulated baseband Barker spread 1 kbps signal 6

7 Fig. 5. Interferene power into adjaent hannels in the.4 GHz band A similar study was onduted for an 11 Mbps CCK spread signal. Here, six bits were used to pik the 8- hip CCK spreading ode and again two bits used for DQPSK modulation. The theoretial analysis performed here was onsiderably more detailed. However, an exat expression was obtained; it again employs Eq. (3) along with the baseband PSD given as [13] S ( f ) = T sin ( ft ) 1+ [ os( π f T ) os( 6π f T )] x [ os10 ( π f T ) os14 ( π f T )]. Observe that the envelope of the baseband PSD is always the same for these signals, but now there is a ripple, without nothes in the spetrum. Simulation results, FPGA emulation, and laboratory measurements all supported the theory. Again adjaent hannel interferene power and temperature was omputed and ompared to the previous results, with a MHz main lobe bandwidth, without and with the spetral mask, as show in Fig. 5. Higher rate signals use OFDM. The 80.11b and g standards speify 6 and 9 Mbps using OFDM with BPSK and ode rates of 1/ and 3/4, respetively; they also detail 1 and 18 Mbps rates using OFDM with QPSK and ode rates of 1/ and 3/4, respetively. Even higher rates use OFDM with quadrature amplitude modulation (QAM), speifially 16- QAM to ahieve 4 and 36 Mbps, and 64-QAM to obtain 48 and 54 Mbps, all with onvolutional oding. The PSD of the OFDM signal was determined analytially and via omputer simulation, with very good agreement. This work was extended to OFDM with a CPC. Here the analysis was muh more involved, but an exat expression for the PSD was derived, with no approximations. The exat expression for a transmitted N omponent OFDM signal u(t) with a Fast Fourier Transform (FFT) time T F and CPC time T F / 4, is found as [14] S u ( f ) 1/ 4)[ S ( f f ) + S ( f )] ( g 0 g f0 (6) =, (7) where the omplex baseband signal g(t) with CPC has a PSD S g a a ( f ) = Ts sin ( TF f k) + Ts sin [( TF f k) / 4] σ a + 5 4σ 5 N / 1 N / 1 k = N / σ 0 N / 1 k = N / k Ts ( 1) os[ ( 5TF f k )( π / 4) ] sin( TF f k ) sin[ ( TF f k )/ 4]. k = N / (8) Here the symbol time T s = T F / N and σ a is the OFDM omponent variane used to set the OFDM signal power to 1 watt. The CPC makes the intended DC (and arrier) zero imperfet and generates a ripple in the normally flat PSD, as shown in Fig. 6 with N = 5 and T F = 3. µse. Computer simulations verified this spetral signature, and the interferene power and temperature was omputed without the with the spetral mask. A omparison between all signal formats, in terms of interferene power, is shown in Fig. 5. An analyti model was postulated for an interfering mirowave oven (MWO) signal in the.4 GHz unliensed band. This onsisted of a linearly swept Frequeny Modulated (FM) signal with an on/off envelope at the 60 Hz line voltage rate, with 50% duty yle. A theoretial analysis of the spetral signature and subsequent omputer simulation was performed. A trapezoidal envelope was also investigated. Fig. 6. PSD of baseband OFDM signals with CPC This modeling effort was expanded into a ombined AM- FM model with the addition of transient turn-on and turn-off wideband signal bursts. This new model has 15 parameters that an be applied to any residential mirowave oven and is desribed as sum of ON-yle wave-shapes, (t), that is [15]: n= v ( t) = ( t nt ), (9) where T = 1 / fa = 1/ 60. The wave-shape is modeled analytially as ( t) = A1 p( t + ta; b1 )os(πf1t ) + A p( t + ta; b )os(πf t) + s( t), + A1 p( t ta; b1 )os(πf1t ) + A p( t ta; b )os(πf t) (10) 7

8 where the pulse waveform, p(t), is p(t;b) = sin (bt), t < 0.5T p. (11) The power in the transient pulses is ditated by the amplitudes, A 1 and A, and the enter of their spetra is determined by the arrier frequenies, f 1 and f. The time loations of the transient pulses are at ±t a and their duration is T p. The bandwidths of the two transients are determined by b 1 and b, respetively. The AM-FM signal, with sinusoidal modulation, an be written as: s( t) Aos(π f t) os(πf t + β sin(πf t)), t < 0. 5T. (1) = a a s The power in s(t) is ditated by the amplitude A and the sweep time is T s. The peak frequeny deviation is determined by the modulation index, β, while T s and β together determine the frequeny-swept band. The enter frequeny of the mirowave oven magnetron is given by f. Real-world experimental measurements lead to the AM-FM model that uses a sinusoidal AM envelope and a sinusoidal frequeny sweep, along with band limited, flat, wideband, turn-on and turn-off spetral shapes. Measurements in the WIL were in good agreement with omputer simulations in terms of spetral signatures. The simulated PSD is shown in Fig. 7 and the experimentally measured PSD of a residential mirowave oven is given in Fig. 8. A mitigation tehnique has been derived that allows up to 88% avoidane of a mirowave oven interfering signal by an signal operating on hannel 1. Separate paradigms for mirowave oven interferene avoidane by signals operating on the other hannels. Fig. 8. Experimental PSD for a MWO (enter.4 GHz, 1 MHz / division) Additional work is underway on the spetral signatures of ordless phones in the ISM bands, unliensed WiMAX signals, and Canopy TM (Motorola last mile tehnology) transmissions. Similarly, baby monitors and wireless video game ontrollers using unliensed bands are also being studied. The ritial next step is a paradigm that will apply interferene power and other network metris to a WLAN so as to minimize the effets of interferene and maximize the apaity and throughput (or more aurately goodput, i.e., useful data) of the network. IV. CRITICAL ROLE OF SIMULATION & IMPROVED MODELS Network simulation is inreasingly relied on as an evaluation tool, yet many of the assumptions and tehniques ommonly employed have been under adverse srutiny by the networking ommunity [8][9]. One of the issues with wireless network simulation tools is the limited interferene modeling apability. The ommonly used network simulation tools (Qualnet, ns- and Opnet) have similar apabilities for modeling interferene. These apabilities are primarily used to simulate o-hannel interferene within a Wi-Fi environment, and there has been very limited modeling of interferene from other soures. Clearly, more general interferene modeling will be ritial to suessful development and deployment of ognitive radio networks. Fig. 7. Simulated PSD of the MWO signal (nominally at 1 GHz) To begin to understand the impat of interferene on network performane, a mirowave oven model was developed. Speifially, a physial layer model was developed in Opnet based on the spetral and temporal harateristis of mirowave oven emissions. There are many physial layer variables to onsider to realistially model a mirowave oven signal and the resulting interferene with the transmission of Wi-Fi signals. More important though, is how the effets of interferene are propagated up the protool stak. Before finetuning the mirowave oven model, it is ritial to validate the propagation of errors due to interferene. To validate this propagation it is neessary to do multi-layer network 8

9 monitoring. Multi-layer network monitoring an be aomplished with a wireless ard that is apable of taking measurements from multiple layers and software drivers that an ollet the measurements. Experiments with mirowave ovens employing multi-layer monitoring are urrently underway. Given the goal of understanding the impat of interferene on network performane, it quikly beame evident that physial layer simulation models would be of limited utility. Although physial layer interferene models an be useful for lower layer modeling or modeling of speifi networking environments, general purpose network interferene testing annot be easily performed with these models. To address this issue, we have begun to investigate reating general error models at different layers in the protool stak. These general models will be based on experimental multi-layer measurements of interferene and will allow network simulation users to realistially model interferene and understand its impat on network performane. These models will enable users to test performane under interferene senarios without having to onfigure the speifi models of the physial layer and the propagation up to the layer of interest. V. WIRELESS CAPACITY EXPLOITATION (SPECTRUM / SPACE / TIME) As we ontinue to move forward in our understanding of the ognitive radio opportunity, it is important to note that there are at least four fundamental onstraints on the deployments of the overlay approah to spetrum utilization. First, there is the hallenge of the operation environment identified earlier. Seond, there are the onstraints of tehnology where there are for instane finite times between sensing the initiation of an inumbent transmitter s signal and the time when the ognitive radio s signal an be shut down or transferred to another spetral region. Third, there are eonomi realities to be dealt with. For example, if there were an effetively infinite number of sensors whih ould always provide an aurate eletromagneti desription, and a listing of the devies operating a speifi region, and if this information ould be broadast on a separate ontrol hannel (suh as advane traffi ontrol systems on motor ways), then the deployment of highly suessful ognitive radio systems would be relatively trivial. Unfortunately, as the number of sensors approahes infinity, so does the ost. Finally there is the soial onstraint assoiated with wide-spread antenna deployments. Based on worries of unsubstantiated health risks, and on onerns about the maintenane of the beauty of the natural environment, there are signifiant limitations on the plaement of antennas independent of the ost onsideration. These issues olletively onspire to inrease the diffiulty in deploying ognitive radio based ommuniations systems. VI. SUMMARY The suessful deployment of ognitive radio tehnology is one of, if not the most signifiant urrently identified opportunities to dramatially expand the data arrying apaity of our finite wireless spetrum. There is however signifiant fundamental, as well as appliations oriented, researh required in order to realize this opportunity. In this paper we have identified a few areas where progress is being made whih will help to suessfully implement ognitive radios in speifi areas of opportunity. It is also reognized that additional work will be required in a variety of domains to expand and aelerate the realization of the benefits antiipate from the deployment of ognitive radio tehnology. ACKNOWLEDGEMENTS The authors would like to aknowledge the onsiderable efforts and ontributions of Dr. Mark A. MHenry, and D. MCloskey of Shared Spetrum Co., Professor Donald R. Ui and graduate students A. Z. Al-Banna, T. R. Lee, T. M. Taher, and X. L. Zhou all at or formerly at IIT. REFERENCES [1] Spetrum Oupany Measurements, Loation 1 of 6: Riverbend Park, Great Falls, Virginia, M. A. MHenry, K. Steadman, Shared Spetrum Company Report, Aug [] Spetrum Oupany Measurements, Loation of 6: Tyson s Square Center, Vienna, Virginia, April 9, 004, M. A. MHenry, K. Steadman, Shared Spetrum Company Report, Aug [3] Spetrum Oupany Measurements, Loation 3 of 6: National Siene Foundation Building Roof, April 16, 004, Revision, M. A. MHenry, S. Chunduri, Shared Spetrum Co. Report, Aug [4] Spetrum Oupany Measurements, Loation 4 of 6: Republian National Convention, New York City, New York, August 30, September 3, 004, Revision, M. A. MHenry, D. MCloskey, G. Lane-Roberts, Shared Spetrum Company Report, Aug [5] Spetrum Oupany Measurements, Loation 5 of 6: National Radio Astronomy Observatory (NRAO), Green Bank, West Virginia, Otober 10-11, 004, Revision 3 M. A. MHenry K. Steadman, Shared Spetrum Company Report, Aug [6] Spetrum Oupany Measurements, Loation 6 of 6: Shared Spetrum Building Roof, Vienna, Virginia, Deember 15-16, 004, M. A. MHenry, D. MCloskey, J. Bates, Shared Spetrum Company Report, Aug [7] Chiago Spetrum Oupany Measurements & Analysis and a Long-term Studies Proposal, M. A. MHenry, P. A. Tenhula, D. MCloskey, D. A. Roberson, and C. S. Hood, in Pro. of TAPAS Conferene, Aug. 006 [8] Experimental Evaluation of Wireless Simulation Assumptions, D. Kotz, C. Newport, R.S. Gray, J. Liu, Y. Yuan, C. Elliot, in Pro. of MSWim 04, pp. 78-8, Ot [9] On Credibility of Simulation Studies of Teleommuniation Networks, K. Pawlikowski, H.-D. Jeong, and J.-S. Lee, IEEE Communiations Magazine, vol. 40, no. 1, pp , Jan

10 [10] Federal Communiations Commission, Spetrum Poliy Task Fore, P. Kolodzy, Chair, Rep. ET Doket no , Nov. 00 [11] Spetral Signatures and Interferene in Wi-Fi Signals with Barker Code Spreading, T. R. Lee, A. Z. Al- Banna, X. L. Zhou, and J. L. LoCiero, and D. R. Ui, in Pro. IEEE Dynami Spetrum. Aess Networks Conferene (DySPAN), Nov [1] Spetral Signatures and Interferene of Wi-Fi Signals in the.4 GHz, ISM Band, X. L. Zhou, Master s Thesis, Chiago, IL, ECE Department, Illinois Institute of Tehnology, De. 005 [13] 11 Mbps CCK - Modulated 80.11b Wi-Fi: Spetral Signature and Interferene, A. Z. Al-Banna, T. R. Lee, J. L. LoCiero, and D. R. Ui, in Pro. Sixth IEEE International Conf. on Eletro/Information Tehnology (EIT'06), May 006. [14] Effet of Cyli Prefix and Symbol Shaping on Inter- Carrier and Inter-Channel Interferene in OFDM Systems, A. Z. Al-Banna, J. L. LoCiero, and D. R. Ui, in Pro. World Wireless Congress (WWC), May 006. [15] Charateristis of an Unintentional Wi-Fi Interferene Devie The Residential Mirowave Oven, T. M. Taher, A. Z. Al-Banna, J. L. LoCiero, and D. R. Ui, in Pro. IEEE Military Communiations Conferene, Ot