IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 21, NO. 2, FEBRUARY 2003 259

Bluetooth and IEEE 802.11b Coexistence: AnalyticalPerformance Evaluation in Fading Channels

Andrea Conti, Member, IEEE, Davide Dardari, Member, IEEE, Gianni Pasolini, Student Member, IEEE, andOreste Andrisano, Member, IEEE

Abstract—In this paper, the issue of Bluetooth and IEEE802.11bcoexistence in a heterogeneous environment is addressed by meansof an integrated analytical approach. The methodology proposedcarefully takes both physical (i.e., thermal noise, propagation,interference, modulation formats, and coding techniques) andmedium access control (frequency hopping, packet structures,traffic loads) aspects into account. This model can be easilyimplemented when developing network simulators, thus avoidingthe need of extensive bit level Monte Carlo simulations at thephysical level.

The mean packet error probability is evaluated as a function ofthe relative distance between the two systems for different condi-tions (e.g., propagation, packet type, traffic loading, etc). In partic-ular, how the presence or absence of line-of-sight propagation sig-nificantly affects the coexistence distance is emphasized. Further-more, for a fixed quality-of-service level we derive the coexistencedomain of the two considered systems in terms of relative distance.

Index Terms—Bluetooth (BT), fading channel, IEEE 802.11b,performance evaluation, standards coexistence, wireless local areanetwork (WLAN), wireless personal area network (WPAN).

I. INTRODUCTION

T HE INCREASING penetration of heterogeneous commu-nication devices sharing the unlicensed world-wide avail-

able industrial, scientific, and medical (ISM) band, makes theissue of mutual interference a key aspect to be investigated tobe able to ascertain whether their coexistence in the same sce-nario is possible or otherwise.

In the following, we focus our attention on two of themost important families of devices designed to operate in theISM band: IEEE 802.11b and Bluetooth (BT). In previouslyappeared papers, the coexistence issue has been generally inves-tigated separately considering the impact of one system on theother (IEEE802.11b versus BT, or BT versus IEEE802.11b). Asfar as the impact of the interference generated by IEEE802.11bon BT reliability is concerned, in [1] some simulation resultsare presented at bit level considering additive white Gaussiannoise (AWGN) and fading channel. The BT packet errorprobability (PEP) due to IEEE802.11b interference has beenconsidered in [2], nevertheless, the BT coded packet structureand the propagation channel are not taken into account. In[3], a measurement campaign was performed in order to

Manuscript received May 5, 2002; revised October 2, 2002. This work wassupported in part by Consiglio Nazionale delle Ricerche (CNR) and Ministerodell’Istruzione, dell’Università de della Ricerca (MIUR).

The authors are with EIS, IEIIT-BO/CNR, CNIT, University of Bologna,Bologna, 40136 Italy (e-mail: aconti@deis.unibo.it; ddardari@deis.unibo.it;gpasolini@deis.unibo.it; oandrisano@deis.unibo.it).

Digital Object Identifier 10.1109/JSAC.2002.807345

evaluate the performance of an interfered BT link consideringthe largest and least protected BT data packet. As far as thereciprocal scenario is concerned (BT versus IEEE802.11b),different studies have been presented in the literature. In [4],the coexistence issue has been investigated in the case ofIEEE802.11b low-rate link (1 Mb/s) interfered by a set ofBT piconets. It mainly focuses on the possibility to introducetraffic shaping techniques to reduce the effect of interferenceboth for data and voice traffic types. Other papers, e.g., [5],take different traffic models into account but adopt simplifiedchannels models without considering fading impairments; theeffect of fading is taken into account in [6] in a simplifiedapproach through simulation. In [1] and [7]–[9], it is shown thatthe effect of adjacent channel interference on the bit-error rate(BER) strongly depends on the frequency offset between usefuland interfering carriers, hence, this aspect can not be neglectedwhen the coexistence problem is addressed. This effect istaken into account in a more accurate analysis presented in[10], [11], where some performance measures, such as thepacket error rate, the number of packet retransmissions and thetransmission latency, are introduced. Therein, the log-normalshadowing effect is considered through the classical Jake’smethod [12], generally adopted in cellular systems design,and network performance is investigated considering severalBT piconets located in the same region of an IEEE802.11bstation. An approximated threshold-based approach on the PEPis proposed to determine the probability of packet collision,hence, the region where the network performance is likely tobe impaired by BT. In [13] and [14], both links are investigatedby means of simulations; moreover, in [14] a measurementcampaign is also performed.

Generally speaking, to obtain the overall system perfor-mance, a bit level simulation combined with a protocol levelsimulation does not represent a practical solution due to theprohibitive computational time required. Hence, there is thenecessity to introduce new accurate and simple methods to theBER evaluation that avoid simulation.

The aim of this paper is to perform a complete analyticalcoexistence study by means of an integrated approach [15],which properly takes all transmission aspects (propagationeffects, interference, thermal noise, modulation format, codingtechniques) and medium access control aspects (frequencyhopping, packet structure, traffic load) into account. An accu-rate expression for the mean PEP is carried out in the presenceof line-of-sight (LOS) and non LOS propagation conditions.The model developed can be easily implemented in protocolsimulators (or upper-layers analytical models) without the needof extensive bit level Monte Carlo simulations.

0733-8716/03$17.00 © 2003 IEEE

260 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 21, NO. 2, FEBRUARY 2003

Starting from this approach, performance evaluation and co-existence issues are addressed for different system configura-tions and propagation conditions. A joint analysis of BT inter-ference on IEEE802.11b and IEEE802.11b interference on BTis carried out in order to evaluate the minimum coexistence dis-tance as a function of the traffic load offered to each link and thedesired quality-of-service (QoS). To the authors’ knowledge, noanalytical study considering all the above mentioned aspects ispresent in the literature.

The paper is organized as follows: in Section II a briefoverview of BT and IEEE 802.11b is reported and in Section IIIthe channel model assumed is introduced. In Section IV,the basis of the proposed analytical methodology are given,whereas in Sections V and VI the performance, in termsof mean PEP, for BT interfered by IEEE802.11b and forIEEE802.11b interfered by BT are evaluated, respectively. Thecoexistence issue between the two standards is addressed inSection VII by means of a joint performance analysis and,finally, our conclusions are given in Section VIII.

II. BT AND IEEE 802.11B OVERVIEW

A brief description of the two considered standards is re-ported in this Section to recall the main characteristics that willbe considered in the following.

• Bluetooth. Designed to operate in a noisy and potentiallyinterfered radio frequency environment, BT adopts a time-slotted frequency-hopping spread-spectrum scheme witha forward error correction (FEC) coding technique [16].In the ISM band, a set of 79 hop carriers separated by1 MHz has been defined. A binary Gaussian shaped fre-quency-shift-keying (GFSK) modulation scheme is ap-plied in order to reduce cost and device complexity anda symbol rate of 1 Mb/s can be achieved for an occupiedbandwidth of about 1 MHz.

• IEEE 802.11b. As far as the physical layer of IEEE 802.11is concerned, three different basic transmission techniquesare defined: diffuse InfraRed (IR), 2.4 GHz ISM bandfrequency-hopping spread-spectrum (FHSS) and 2.4 GHzISM band direct sequence spread spectrum (DSSS) at 1–2Mb/s [17]. A high rate mode extension IEEE 802.11b hasbeen recently introduced enabling 5.5 and 11 Mb/s by em-ploying the complementary code keying (CCK) scheme[18]. In all modes, the occupied bandwidth is about 22MHz.

Even though our approach is quite general in this paper we focusour attention on the DSSS (1 Mb/s) and CCK (11 Mb/s) withregard to the IEEE 802.11b.

III. CHANNEL MODEL

In the scenario considered (sketched in Fig. 1), anIEEE802.11b and a BT link are simultaneously active inan indoor environment. Let us denote with the generalpath-loss law as a function of the distancebetween the trans-mitter and the receiver. Moreover, we assume that frequencyselective effects do not significantly affect the receiver per-formance [19], nevertheless, they have to be considered wheninvestigating the fading correlation among packets transmitted

Fig. 1. Scenario considered: interference of IEEE802.11b device on BT deviceand vice versa.

in different frequency hops (e.g., Bluetooth). Hereafter, whenLOS condition is satisfied, the Ricean fading statistics withRice factor is adopted [1], [20], whereas the Rayleighdistribution is considered in the opposite case (non LOS).

Now, let us introduce the following parameters (see Fig. 1):

• : distance between the two BT stations;• : distance between the BT receiver and the IEEE

802.11b transmitter;• : distance between the two IEEE802.11b stations;• : distance between the IEEE 802.11b receiver and the

BT transmitter.The useful and interfering power instantaneously received bythe receiving device can be expressed for each link, respectively,by

(1)

(2)

where:

• and are Rice/Rayleigh distributed random variableswith , denoting the sta-tistical average operator with respect to the random vari-able ;

• and are coefficients dependent on transmitting andreceiving antenna gains;

• and are the useful and interfering transmitted powerlevels, respectively;

• stands for or .Without loss of generality, since BT and IEEE 802.11b transmit-ters and receivers operate in the same frequency band, we canalso assume they have the same antenna gains, hence, .By averaging and over the fading statistics we obtain

(3)

(4)

CONTI et al.: BLUETOOTH AND IEEE 802.11B COEXISTENCE: ANALYTICAL PERFORMANCE EVALUATION IN FADING CHANNELS 261

Owing to the particular environment considered (low mobilityindoor scenario) the fading level has to be considered as constantover the entire packet duration (slow fading) for both BT andIEEE802.11b links.

IV. BLOCK ERRORPROBABILITY EVALUATION

A. Evaluation in Slow Fading Channel

Before giving the details of our analytical approach, in thisSection we derive a general result, that will be used throughthe paper, concerning the mean (averaged over Rice/Rayleighfading) error probability evaluation for a block of receivedsymbols. For this purpose let us assume that the instantaneoussymbol error probability can be described as a simple parame-terized function of the instantaneous signal-to-noise ratio (SNR)

as

(5)

with , , , and , , beingthe energy per received symbol, the single sided thermal noisespectral density and the symbol time, respectively. As will beexplained in the following, with a proper choice of parameters

and , the performance curves related to both the consideredsystems are well approximated by (5) for the whole range ofinterest for . Assuming that errors due to AWGN are indepen-dent, the instantaneous block error probability , i.e., theprobability to have at least an error in a block ofsymbols be-comes

(6)

Now, assuming Rice distributed slow fading channel (where theterm slow means constant during a block),is characterized bythe following probability density function (pdf) [21]

for (7)

where is the mean SNR, is the modifiedBessel function of the first order and the Rice factor. For

it reduces to the Rayleigh distribution.The mean (averaged over fading realizations) is

(8)

For Rice fading, using (7), the mean becomes

(9)

In the particular case of Rayleigh fading , (9) can beexpressed in the following compact form:

(10)

where is the Hypergeometric function [22]. In the fol-lowing, the dependence on and could be hidden forconciseness.

B. Evaluation in the Presence of Co-Channel Interference

Now, we extend the result in (9) to the case with an interferingsignal present during the transmission of the block. Both theuseful and interfering links are affected by fading, the levels ofwhich are assumed to be independent in the two links.

Let us consider the exact instantaneousblock error probability expression in the presence of interfer-ence which, in general, depends on the SNRand the instan-taneous interference power level. According to the definition,the block error probability averaged over the fading statistic ofthe interfering signal and conditioned to the instantaneous valueof is

(11)

In [23], it has been demonstrated that when the interferer is anangle modulated signal and it is affected by Rayleigh fading, itsdisturbance contribution at the receiver has a Gaussian distribu-tion. Consequently, when evaluating (11), we can consider theinterfering signal as an additional contribution to the thermalnoise by simply replacing in (6) with

(12)

hence, writing

(13)

The parameter is a coefficient depending on the interferingsignal pulse shaping and the receiving filter, while 1 is acoefficient depending on the frequency offset between theuseful and interfering carriers. The parameter is normal-ized so that for MHz , where is tuned as aconsequence. These parameters will be specified for each linkin SectionsV.

Following the same procedure reported in the previous sub-section, it is possible to average over the useful received powerfluctuations obtaining, therefore, that the mean block error prob-ability in the presence of interference is given by

(14)

1With � , we will indicate the value of� for�f = xMHz.

262 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 21, NO. 2, FEBRUARY 2003

having defined

(15)

As can be noted, (14) has the same structure as (9), similarto what obtained in frequency-hopped cellular system contexts[24]. Hence, we can conclude that, even if the exact instanta-neous block error probability expression in the presence of in-terference, , is generally not known (its analytical deriva-tion is unaffordable, exception made for few particular cases[25], [26]), it is however possible to derive the mean block errorprobability by using the result of (9) and simply replacingwith

according to (15). Note that for a Ricean interferer withincreasing , we tend toward the condition of static channel,for which the Gaussian approximation leads to an upper bound[25], [26].

V. PERFORMANCEEVALUATION OF BT INTERFERED BY

IEEE802.11B

Hereafter, we derive the mean PEP of the BT link taking boththermal noise and interference generated by a IEEE802.11btransmission into account. For this purpose the different channelcoding techniques adopted by BT are carefully considered.We verified that, for noncoherent demodulation scheme, theexpression (5) with parameters and wellfits the instantaneous behavior in the region of interest, i.e.,

. Let us observe however, that the analytical approachis general in , . For different receiver structures the couples

for which (5) holds are derived in [27].Now, let us recall that BT data and voice packets are made

up of three different fields (access code, header, and payload)each one protected against errors by a different forward errorcorrecting (FEC) block code [16]. As an example, for data trans-mission the following FEC techniques are adopted.

• The access code is made up of three subfields: preamble(4 bits), trailer (4 bits), and Sync Word (64 bits); the lastone is the only subfield sensitive to errors and is protectedby a (64,30) BCH code that can correct at the mosterrors [28].

• The header is protected by a simple three-times repetitionFEC code which can correct, at the most error percodeword; it can be observed, furthermore, that the 1/3repetition code increases the product of the GFSKmodulation, hence, the demodulator performance is fur-ther improved.

• The payload, depending on the considered packet type,can be protected (DM1, DM3, DM5 packet types) ornot (DH1, DH3, DH5 packet types) by an FEC code. Inthe former case a (15,10) shortened Hamming code isadopted, which can correct at the most error percodeword.

To evaluate the instantaneous packet error probability,, as a function of the instantaneous SNR,, the error

correction capabilities of the FEC codes adopted in the differentpacket fields have to be considered, providing

(16)

where:

• is the access code error probability;• is the header error probability;• is the payload error probability.

Hereafter, the hard decision is considered due to the low com-plexity requirements of BT devices. The analytical derivation of

, , and requires the knowledge (for each FECcode adopted) of the codeword error probability , whichis a function of , the number of correctableerrors and of the codeword length , and can be generallyexpressed as

(17)

Having denoted with the error probability of a sequenceof codewords, we can write

(18)

For data packets (DM, DH), for example, it follows that

• ;• ;• with

for DM packets;• with

for DH packets.In Fig. 3, the exact instantaneous PEP following (16) is com-pared with for a DM packet in an AWGN channel for

and . As can be immediately observed thecurve coincides with , hence, an approximate

expression for the is given by

(19)

Let us observe that (19) is even more accurate for DH packetswhose payload is not protected at all. Now, we evaluate the meanPEP, averaged over fading, as

(20)

By using the well-known approximation [29], certainly valid for, on the codeword error probability

(21)

by substituting (5) in (21) results

(22)

and by recalling (8) and (9), we obtain

(23)

where and .

CONTI et al.: BLUETOOTH AND IEEE 802.11B COEXISTENCE: ANALYTICAL PERFORMANCE EVALUATION IN FADING CHANNELS 263

Now, to also include the interference effect, we simply followthe approach discussed in Section IV by replacing in (23) thevariable with according to (14), leading to

(24)

To validate the approximations, we verify that (24), through(15), provides a very good agreement with the simulated asymp-totical average packet error rate (PER) [interference limited op-erating conditions] for different values of . The parameter

is tuned as a consequence in order to match (24) with simu-lations in the case of interference limited operating conditions.Values of for different are then analytically evaluatedwith respect to the reference curve obtained with MHz.For this purpose, the interfering power at the output of the re-ceiving filter has been evaluated starting from the knowledge ofthe power spectral density. The analysis of the power spectraldensity of the IEEE 802.11b signal is carried out in theAppendix. The coefficient can now be obtained by calcu-lating the interfering power at the output of the BT receivingfilter, normalized to the interfering power in the reference caseof MHz

(25)

where , isthe actual BT carrier frequency and MHz is the3-dB bandwidth of the receiving filter (assumed to be 10 polesButterworth). For greater than 10 MHz the effect of theinterference can be neglected .

In Fig. 4, the asymptotic (large SNR) PEP conditioned toas a function of is plotted and compared with simula-tion results (symbols) in the case of 1 Mb/s IEEE802.11b inter-ferer, DM1 BT packet type, and Rayleigh fading in both linksfor MHz. In this case, has been found tobe 0.11. It is worth noting that the values of calculated bymeans of (25), and reported in Table I, provide a remarkableagreement between simulation and analytical results, with theexception of MHz. For the DSSS scheme, the curverelated to the case MHz has been obtained assuming

dB which is different from the value dB,evaluated through (25), reported in Table I. This phenomenoncan be explained considering that the impact of an interferingsignal on an ongoing transmission generally depends not onlyon the interfering power level but also on the shape of the inter-fering power spectrum [30], [31]. As can be observed in Fig. 2,the IEEE 802.11b power spectrum shape denotes a high fre-quency selective behavior around the carrier frequency, whichdeeply affects the BT performance in the presence of co-channelinterference.

It is important to observe that simulation results also confirmthe validity of the adopted approach. Hence, by means of (14),(15), (24), and (25) we obtained a simple and efficient analyticalexpression for that depends on parametersand , which can be derived for any kind of receiving filter.

Fig. 2. IEEE802.11b @ 1 Mb/s: power spectrum and transmission mask.

Fig. 3. Comparison among instantaneous PEP obtained by considering errorsoccurring over all packet fields (Exact) or only on the payload (Approx) for BTpackets withN = 5; 50.

By assuming that the probability of time collision occurrenceis , being an indicator of the IEEE802.11b activity factor,that all BT bits are involved in the collision and averaging

over , we can write

(26)

where is the probability to have a BT transmission ina particular frequency hop. As far as is concerned, weassume that is uniformly distributed among all the allowed79 hops, hence, we can write , whilethe activity factor is a function of the traffic offered to theinterfering IEEE802.11b transmitter.

Being averaged over and over the collision events,, which is a function of and , is an overall

figure of merit which takes the effects of FEC codes, the

264 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 21, NO. 2, FEBRUARY 2003

Fig. 4. BT interfered by IEEE802.11b. AsymptoticalPEP as a function ofC =I conditioned to�f = 0; 1; 4; 8 MHz: comparison among analyticaland simulation results.

TABLE I� [dB] FOR BT INTERFERED BYIEEE802.11B

frequency hopping technique and traffic into account, hence, itis a meaningful indicator of the system performance.

In Fig. 5, we reported the as function of fordifferent values of and , in the case of DM1 packet type. Theimpact of the Rice factor on the performance can be clearly ap-preciated showing, therefore, that the LOS condition, normallyverified in a BT link, considerably improves the achieved per-formance. It is also possible to jointly observe the effect of boththermal noise and interference.

According to [5] and [32], in the rest of the paper, we adoptthe following path loss model:

(27)

being the distance between the transmitter and the receiver inmeters. This model assumes LOS propagation for the first 8 m.

From (3), (4), (26), and (27) it is possible to plot as afunction of the distances ratio (hereafter, ), as

Fig. 5. BT interfered by IEEE802.11b. Effect of the Rice factorK, and themean SNR< E > =N , on the mean packet error probability,PEP , as afunction of the ratioC =I . DM1 packet.

Fig. 6. BT interfered by IEEE802.11b. AsymptoticalPEP as a functionof d =d . a) d = 2 m, P = 0 dBm, b) d = 2 m, P = 4 dBm,c) d = 2 m, P = 20 dBm, d)d = 8 m, P = 0 dBm, e)d = 8 m,P = 4 dBm, f) d = 8 m,P = 20 dBm. DM1 packet type. IEEE802.11bactivity factorL = 1. Rice factorK = 5.

shown in the case of interference-limited operating conditions,in Fig. 6 where the activity factor the useful distanceand the BT transmitting power levels have been consideredas parameters. Here, the DM1 data packet has been adopted ina LOS propagation condition . By fixing a target meanPEP, it is possible to obtain the minimum distance ratio, ,for which the QoS requirement is satisfied. It is interesting tonote that, at the same , the case with m out-performs the case m. This is due to the greater prop-agation coefficient experienced by the interferer for increasingdistances . The strong dependence on the transmitted powerlevel suggests that a properly designed adaptive utilization ofthe transmitted power could reduce the interference effect im-proving the achieved performance [34].

CONTI et al.: BLUETOOTH AND IEEE 802.11B COEXISTENCE: ANALYTICAL PERFORMANCE EVALUATION IN FADING CHANNELS 265

VI. PERFORMANCEEVALUATION OF IEEE802.11B INTERFERED

BY BT

The above introduced methodology can also be adopted toinvestigate the reciprocal scenario in order to assess the per-formance of IEEE802.11b interfered by a BT transmission: infact, this analysis is indeed carried out by means of the sameapproach adopted in the previous section.

For this link proper values for parametersand in (5) havebeen chosen for DSSS and for CCK standards. In the formercase, they are and (exact expression from [29]),whereas for the CCK case they are and (fittingfrom results reported in [9]). As for the BT link, we assume thatthe PEP is mainly determined by errors occurring in the payload(i.e., MPDU); this is reasonable considering that the payloadis uncoded and generally considerably longer than other fields(only 192 bits in total). We assume a payload of 1024 Bytesas a reference. Hereafter, the following assumptions have beenmade.

• The collision event involves all the bits belonging to theBT packet, i.e., we consider the worst case with regard tothe timing synchronism between packets.

• Fading realizations in different interferers BT slots are in-dependent (due to frequency hopping).

• Symbols not collided are correctly detected (both systemsworking inside their respective operative ranges).

Based on the above mentioned hypothesis and consideringthat the fading process on the useful link is constant over the en-tire useful packet duration, the mean probability (averaged overthe fading) to have an erroneous packet reception, conditionedto time collisions, can be written as

(28)

where is the number of IEEE802.11b symbols collidingwith one BT packet ( for the 11 Mb/s system and

for the 1 Mb/s one). This expression can be upperbounded by moving the averaging operator inside the productgiving

(29)

Let us denote with the maximum number of potential BTpackets colliding with an IEEE802.11b packet. The accuracy ofthe bound (29) can be verified in Fig. 7 where the upper bound(29) is compared with (28) in cases of DSSS and CCK sys-tems experiencing both Rayleigh and Rice fading and

(100% traffic load). Note that, even though the DSSS

Fig. 7. IEEE802.11b interfered by BT DM1 packets. Comparison between(28) and the upper bound (29) in cases of DSSS and CCK systems for Rayleighand Rice(K = 5) fading. 100% traffic load.

scheme is more robust than CCK against co-channel interfer-ence, the fact that its packet duration is 11 times greater (hence,more subject to packet collisions), leads to similar performancefor both schemes.

Let us now define as the probability to have BT trans-missions during the same packet transmission. By averagingover all the possible number of BT time collisions, the meanPEP becomes

(30)

To investigate the impact on the performance of different trafficmodels characterizing the interfering (BT) device, let us con-sider the following cases.

• BT voice traffic @64 Kb/s. Using the HV3 packet format,transmissions occur on 33% of the BT time slots in anygiven direction [5], hence, , forCCK and , for DSSS.

• BT full loaded e-mail download/upload traffic[5]. UsingDH1 packets for both CCK and DSSS, we have ,

.• BT binomial traffic. DM1 packet used. Let be the prob-

ability to have a BT packet transmission during the generictime slot, we have

(31)

Other traffic models [4], [5], [34] can be easily taken into ac-count by changing the form of (31). Moreover, starting from thepresented analytical methodology, different traffic managementstrategies could be tested to further reduce the interference im-pact [33], [34].

The values for and in the CCK case have beenevaluated according to the semi-empirical model presented in[8], which considers the interference effect, as a function of,to be shaped as a function combined with a low-pass

266 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 21, NO. 2, FEBRUARY 2003

TABLE II� [dB] FOR IEEE802.11B INTERFERED BYBT

Fig. 8. IEEE802.11b interfered by BT. AsymptoticalPEP as a function ofC =I dB for IEEE802.11b @ 1 or 11 Mb/s: a) CCK,K = 0, voice traffic; b)CCK,K = 5, voice traffic; c) DSSS,K = 5, voice traffic; d) DSSS,K = 5,binomial trafficL = 1=6; e) DSSS,K = 5, e-mail traffic; f) DSSS,K = 0,voice traffic; g) DSSS,K = 0, binomial trafficL = 1=6; and h) DSSS,K = 0, e-mail traffic.

Butterworth filter mask. The DSSS case has been obtained in asimilar way by taking the effect of the Barker code into account.These values are reported in Table II and in this case .

The analytical results related to IEEE802.11b interfered byBT are reported in Figs. 8 and 9. Fig. 8 shows the asymptotical

as a function of dB for IEEE802.11b @ 1 or11 Mb/s: a) CCK, , voice traffic; b) CCK, , voicetraffic; c) DSSS, , voice traffic; d) DSSS, , bino-mial traffic (@64 Kb/s); e) DSSS, , e-mailtraffic; f) DSSS, , voice traffic; g) DSSS, , bino-mial traffic ; and h) DSSS, , e-mail traffic. Itcan be observed, in particular, that at the same BT mean packettransmission rate (e.g., voice and binomial @64 Kb/s) the per-formance does not strongly depend on traffic model. As ex-pected, the worst performance occurs for email download trafficwhich is characterized by 100% activity factor.

Fig. 9. IEEE802.11b interfered by BT. AsymptoticalPEP as a functionof d =d for IEEE802.11b @ 11 Mb/s,K = 5, and differentd and BTtransmitted power: a)d = 8m, P = 0 dBm; b) d = 8m, P = 4 dBm;c) d = 8 m, P = 20 dBm; d) d = 16 m, P = 0 dBm; e)d = 16 m,P = 4 dBm; f) d = 16 m, P = 20 dBm. 100% traffic load. DM1 packettype.

In Fig. 9, we reported the asymptotical as a func-tion of the coexistence distances ratio (hereafter

) for CCK, and 100% BT activity factor (e.g.,e-mail download). The IEEE802.11b transmitted power level isset to 15 dBm, and different cases are considered: a) m,

; b) m, ; c) m,; d) m, ; e) m,

; f) m, . denotes theBT transmitted power. Differently to what happens when BT isinterfered by IEEE802.11b, better performance is obtained forthe lowest value of due to the fact that, for small values of

, the interferer (BT) operates in LOS condition. More-over, in this case we can verify that IEEE802.11b is able towork with lower distance ratios than BT. As an example, fixing

, dBm and m, a minimum dis-tance of about 2 meters is tolerable (this result is confirmedby experimental tests [34]). The deep impact of the BT trans-mitting power level on the minimum tolerable distance ratio canalso be noted.

VII. COEXISTENCEANALYSIS

By means of the methodology proposed in Sections V andVI, we can now analyze the coexistence issue, between the twostandards, evaluating their performance. In fact, the two systemscan simultaneously operate in the same environment only if thetarget QoS, in this case the target mean PEP, is guaranteed forboth of them. To this aim, by fixing the target QoS for eachsystem, we define thecoexistence domain, in terms of relativedistances, as the maximum between the two minimum ratios

.In this regard, in Fig. 10, we show for the BT link the min-

imum ratio as a function of the IEEE802.11b activityfactor, , which respects two target of in-terest, e.g., and , for the lowest BT transmitted power

CONTI et al.: BLUETOOTH AND IEEE 802.11B COEXISTENCE: ANALYTICAL PERFORMANCE EVALUATION IN FADING CHANNELS 267

Fig. 10. BT interfered by IEEE802.11b. Coexistence distance ratio for BTinterfered by IEEE802.11b as a function ofL for different BT packet typesand targetPEP . BT transmitted power 0 dBm. Rice factorK = 5. d = 2

m.

Fig. 11. IEEE802.11b interfered by BT. Coexistence distance ratio forIEEE802.11b interfered by BT as a function ofL for different targetPEPand fading conditions. BT transmitted power 0 dBm,d = 8 m, K = 0

andK = 5, PEP = 0:01, and 0.1. DM1 packet type.

level (0 dBm). Here, we considered LOS propagation conditions(Rice factor ), which is likely to occur when consideringa BT link and, for comparison purposes, the DM1 and the DH1BT packets (let us recall that the payload of DH1 packets is notprotected by any FEC code). As expected the adoption of theshortest and best protected DM1 packet considerably reducesthe minimum ratio needed to achieve the target per-formance. For the less stringent requirement ,the benefic effect of the frequency hopping leads to a minimumvalue of the activity factor . Below that limit, anyinterferer distance satisfies the PEP requirement.

Similarly, for the IEEE802.11b link, in Fig. 11 we show theminimum ratio as a function of the BT traffic load,

(binomial traffic), considering the same target QoS and dif-ferent fading conditions (Rice and Rayleigh). The useful dis-tance has been set as .

VIII. C ONCLUSION

In this paper, we derived an analytical methodology to eval-uate the performance of a BT link interfered by IEEE802.11b,andvice versa, in a Rice/Rayleigh fading channel with AWGN.The QoS of the two systems is evaluated in terms of the meanPEP. The methodology proposed carefully takes the presence ofthermal noise, propagation impairments, interference, modula-tion formats, and coding techniques into account together withfrequency hopping, packet structures and traffic load. Whennecessary the methodology assumptions are verified by sim-ulations. This model can be easily implemented when devel-oping network simulators, thus avoiding the need of extensivebit level Monte Carlo simulations at the physical level. In par-ticular, it has been emphasized how the presence or absence ofLOS propagation and different choices of system parameters,e.g., the transmitted power level, significantly affect the coex-istence distance. Moreover, the quite relevant coexistence issuebetween the two standards has been addressed deriving the co-existence domain, in terms of relative distances, for a fixed QoS.This methodology can be easily extended to include more so-phisticated traffic models and to test suitable techniques to coun-teract the reciprocal interference.

APPENDIX

Since IEEE802.11b @1 Mb/s adopts a direct sequence spreadspectrum technique, the baseband pulse amplitude modulation(PAM) signal, is multiplied by the spreading signal ,where

(32)

(33)

and is the transmitted bit, is the bit-time,is the rectangular pulse of duration, is the

spreading sequence,is the spreading factor, is thechip time and is the rectangular pulse of duration. In theconsidered case, the adopted spreading sequence is the Barkersequence with

[17]. The spread spectrum signalbecomes

(34)

where

(35)

From (34), we can note that is still a PAM signal withmodulation pulse given by expression (35). Assuming

268 IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 21, NO. 2, FEBRUARY 2003

independent and ergodic, we can state that the two-sidedpower spectral density of is given by

(36)

(37)

and being the Fourier transform of and ,respectively. For the DBPSK scheme adopted by IEEE 802.11b@1 Mb/s the two-sided power spectrum of the modulated signalis, exception made for a scale factor, the translation of of

( being the carrier frequency); it follows that in order toderive the power spectrum of the modulated signal we can referto .

The power spectral density , as expressed by (36), isreported in Fig. 2 where it is denoted by the letter “A.” In Fig. 2,the transmit spectrum mask defined by the specification is alsoreported. In order to fulfill the requested mask [17], the modu-lated signal is passed through a transmission filter with base-band equivalent amplitude characteristic . Hence, thepower spectrum of the transmitted signal is, apart from the al-ready discussed frequency translation, proportional to

(38)

which is plotted in Fig. 2 and denoted with the letter “B.”As far as the CCK scheme is concerned, it can be viewed as

a coded QPSK modulation based on the same pulse andthe same transmit spectrum mask as for the DSSS case. Theevaluation of the power spectral density follows the classicalresult for QPSK [29].

ACKNOWLEDGMENT

The authors would like to thank the anonymous reviewers fortheir suggestions.

REFERENCES

[1] A. Soltanian and R. E. Van Dyck, “Physical layer performance for coex-istence of Bluetooth and IEEE802.11b,” presented at the Virginia Tech.Symp. Wireless Personal Communications, Blacksburg, VA, June 2001.

[2] N. Golmie and F. Mouveaux, “Interference in the 2.4 GHz ISM band:impact on the Bluetooth access control performance,” inProc. IEEEInt. Conf. Communications, vol. 8. Helsinki, Finland, June 2001, pp.2540–2545.

[3] R. J. Punnoose, R. S. Tseng, and D. Stancil, “Experimental results forinterference between Bluetooth and IEEE802.11b DSSS systems,” inProc. IEEE Vehicular Technology Conf., vol. 1. Atlantic City, NJ, Oct.2001, pp. 67–71.

[4] C. F. Chiasserini and R. R. Rao, “Performance of IEEE802.11 WLANsin Bluetooth environment,” inProc. IEEE Wireless Communications andNetworking Conf., vol. 1, IL, Sept. 2000, pp. 94–99.

[5] Reliability of IEEE802.11 Hi Rate DSSS WLANs in a High DensityBluetooth Environment, J. Zyren. [Online]. Available: http://www.in-tersil.com.

[6] M. Fainberg and D. Goodman, “Analysis of the interference betweenIEEE802.11b and Bluetooth systems,” inProc. IEEE Vehicular Tech-nology Conf., vol. 2. Atlantic City, NJ, Oct. 2001, pp. 967–971.

[7] A. Soltanian and R. E. Van Dyck, “Performance of the Bluetooth systemin fading dispersive channels and interference,” inProc. IEEE GlobalTelecommunications Conf., vol. 6. San Antonio, TX, Nov. 2001, pp.3499–3503.

[8] I. Howitt, V. Mitter, and J. Guitierrez, “Empirical study for IEEE802.11and Bluetooth interoperability,” inProc. Vehicular Technology Conf.,vol. 2. Rhodes, Greece, May 2001, pp. 1109–1113.

[9] IEEE P802.15 wireless local area networks. IEEE documentation, IEEEP802.15-01 418r0.

[10] I. Howitt, “WLAN and WPAN coexistence in UL band,”IEEE Trans.Veh. Technol., vol. 50, pp. 1114–1124, July 2001.

[11] , “IEEE802.11 and Bluetooth coexistence analysis methodology,”in Proc. Vehicular Technology Conf., vol. 2. Rhodes, Greece, May 2001,pp. 1114–1118.

[12] W. C. Jakes,Microwave Mobile Communications. New York: Wiley,1974.

[13] N. Golmie, R. E. Van Dyck, and A. Soltanian, “Interference of Blue-tooth and IEEE802.11: simulation modeling and performance evalua-tion,” presented at the ACM Int. Workshop on Modeling, Analysis, andSimulation of Wireless and Mobile Systems, Rome, Italy, July 2001.

[14] J. Lansford, A. Stephens, and R. Nevo, “Wi-Fi (802.11b) and Bluetooth:enabling coexistence,”IEEE Network Mag., vol. 15, no. 5, pp. 20–27,Sept./Oct. 2001.

[15] O. Andrisano, D. Dardari, and G. Mazzini, “An integrated approachfor the design of wide-band wireless LANs,” inProc. IEEE Int. Conf.Telecommunications. Porto Carras, Greece, June 1998, pp. 121–126.

[16] Specification of the Bluetooth System. Core, Version 1.0 B. [Online].Available: http://www.bluetooth.com

[17] IEEE Standard for Wireless LAN Medium Access Control (MAC) andPhysical Layer (PHY) Specification. IEEE Standard 802.11.

[18] S. Halford, K. Halford, M. Webster, and C. Andren, “Complementarycode keying for rake-based indoor wireless communication,” inProc.IEEE Int. Symp. Circuits and Systems, vol. 4. Orlando, FL, May/June1999, pp. 427–430.

[19] G. F. Pedersen and O. Eggers, “Inital investigations of the Bluetoothlink,” in Proc. IEEE Vehicular Technology Conf., vol. 1. Tokyo, Japan,May 2000, pp. 64–69.

[20] S. C. Kim, H. L. Bertoni, and M. Stern, “Pulse propagation character-istics at 2.4 GHz inside buildings,”IEEE Trans. Veh. Technol., vol. 45,pp. 579–592, Aug. 1996.

[21] 10 158-0012M. K. Simon and M. S. Alouini,Digital Communicationover Fading Channels, 1st ed. New York: Wiley, 2000.

[22] R. L. Graham, D. E. Knuth, and O. Patashnik,Concrete Mathematics,2nd ed. Reading, MA: Addison-Wesley.

[23] J. Chow and N. C. Beaulieu, “Analisys of cochannel interference inangle modulation Systems,” inProc. IEEE Pacific Rim Conf Communi-cations, Computers and Signal Processing, Victoria, Canada, Aug. 1999,pp. 123–127.

[24] M. Chiani, A. Conti, and O. Andrisano, “Outage evaluation for slowfrequency-hopping mobile radio systems,”IEEE Trans. Commun., vol.47, pp. 1865–1874, Dec. 1999.

[25] M. Chiani, “Performance of BPSK and GMSK with multiple cochannelinterfereres,” inProc. 7th IEEE Int. Symp. Personal, Indoor and MobileRadio Communications, Oct. 1996, pp. 833–837.

[26] , “Analytical distribution of linearly modulated cochannel inter-ferers,”IEEE Trans. Commun., vol. 45, pp. 73–79, Jan. 1997.

[27] O. Andrisanoet al., “Bluetooth physical layer parameters optimization,”, submitted for publication.

[28] W. Stahl, S. Zurbes, K. Matheus, and J. Haartsen, “Radio network per-formance of Bluetooth,” inProc. IEEE Int. Conf. Communications, vol.3. New Orleans, LA, June 2000, pp. 1563–1567.

[29] J. G. Proakis,Digital Communications, 4th ed. New York: McGraw-Hill.

[30] O. Andrisano and R. Siroli, “Performance of partial response CPM sys-tems with narrowband channels and limiter-discriminator detection,”Alta Frequenza, Focus on Mobile Radio Systems, no. 1, 1988.

[31] O. Andrisano and N. Ladisa, “On the spectral efficiency of CPM sys-tems over real channel in the presence of adjacent channel and cochannelinterference: a comparison between partial and full response systems,”IEEE Trans. Veh. Technol., vol. 39, pp. 89–100, May 1990.

[32] A. Kamerman,Coexistence Between Bluetooth and IEEE802.11 CCKSolutions to Avoid Mutual Interference: Lucent Technologies, Bell Lab-oratories, Jan. 1999.

[33] C. F. Chiasserini and R. R. Rao, “Coexistence mechanisms for interfer-ence mitigation between IEEE 802.11 WLANs and Bluetooth,” inProc.IEEE 21st Annu. Joint Conf. IEEE Computer and Communications So-cieties, vol. 2. New York, June 2002, pp. 590–598.

[34] [Online]. Available: http://www.ieee802.org/15/pub/WG2.html

CONTI et al.: BLUETOOTH AND IEEE 802.11B COEXISTENCE: ANALYTICAL PERFORMANCE EVALUATION IN FADING CHANNELS 269

Andrea Conti (S’99–M’01) was born in Bologna,Italy, on December 20, 1972. He received theDr.Ing. degree (with honors) in telecommunicationsengineering and the Ph.D. degree in electronicengineering and computer science, both from theUniversity of Bologna, Bologna, Italy, in 1997 and2001, respectively.

In 1999, he joined the Consorzio Nazionale In-teruniversitario per le Telecomunicazioni (CNIT) atthe Research Unit of Bologna, working in the designof a DSP-based CDMA satellite modem within the

Italian Space Agency (ASI)-CNIT project “Integration of Multimedia Serviceson Heterogeneous Satellite Networks.” In the Summer 2001, he joined theWireless Section of AT&T Labs-Research, Middletown, NJ, working on theperformance of digital telecommunications systems with diversity reception.His research interests include mobile radio resource management, frequencyhopping, coding in faded MIMO channels, nonlinear effects in CDMA, anddigital signal processing.

Davide Dardari (S’96–M’98) was born in Rimini,Italy, in 1968. He received the Laurea degree (summacum laude) in electronic engineering and the Ph.D.degree in electronic engineering and computer sci-ence, both from the University of Bologna, Bologna,Italy, in 1993 and 1998, respectively.

In 1998, he joined the Dipartimento di Elettronica,Informatica e Sistemistica, University of Bologna,to develop his research activity in the area ofdigital communications. Moreover, he worked onthe European project PROMETHEUS regarding

short-range communications systems for cooperative driving. From 1998to 2001, he was involved with the Italian Space Agency (ASI)-ConsorzioNazionale Interuniversitario per le Telecomunicazioni (CNIT) project on het-erogeneous satellite networks. Since 2000, he has been a Research Associateat the University of Bologna. His research interests are in OFDM systems,spread-spectrum communications, cellular mobile radio, vehicle-to-vehicleshort-range communication systems, satellite systems, and wideband wirelesslocal-area networks.

Gianni Pasolini (S’00) was born in Cesena, Italy,on June 22, 1970. He received the Laurea degree intelecommunications engineering from the Universityof Bologna, Bologna, Italy, in March 1999. In May1999, he joined the Centre for Studies in ComputerScience and Telecommunication Systems of the Na-tional Research Council (CSITE-CNR) and is cur-rently working toward the Ph.D. degree.

His research interest are in wireless local-area net-works, digital communications, and radio resourcemanagement.

Oreste Andrisano (M’83) was born in Bologna,Italy, on February 14, 1952. He received the Dr.Ing.degree in electronic engineeringcum laudefrom theUniversity of Bologna, Bologna, Italy, in 1975.

In 1975, he joined the University of Bologna,where he became a Professor of Electrical En-gineering in 1985. Since 1992, he has been theDirector of Centro di studio per l’Informatica e iSistemi di Telecomunicazioni (CSITE), Universityof Bologna and Consiglio Nazionale delle Ricerche,Roma. Since 2000, he has been the Director of the

Laboratorio Nazionale di Comunicazioni Multimediali, Napoli (ConsorzioNazionale Interuniversitario per le Telecomunicazioni, CNIT). His researchactivity has been concerned with different fields in the digital communicationarea, such as digital signal processing, data transmission for satellite andfixed radio links applications, and local wireless and mobile radio networks.He has also been active in the Intelligent Transportation Systems (ITS) area,with reference to vehicle-to-vehicle and vehicle-to-infrastructure communi-cation systems. In 1987 and 1988, he cooperated in the definition phase ofPROMETHEUS (EUREKA), as an European Coordinator for the transmissionsystems research area. Then, he was in the Steering Committee of Project DataAcquisition and Communication Techniques and their Assessment for RoadTransport (DACAR) in the framework of DRIVE I, ECC, 1988–1991. From1991 to 1997, he was responsible for the ITS communication activities in Italy(coordination of PROCOM and TELCO). Since 1998, he has been the NationalCoordinator of the project multimedia systems funded at national level byMIUR and CNR (Roma).

Dr. Andrisano is a member of the IEEE Communication and Vehicular Tech-nology Societies and of the IEEE Radiocommunication Committee. From 1996to 2001, he was an Editor of the IEEE TRANSACTIONS ONCOMMUNICATIONS

(Modulation for fading channels).