Analysis of Sub-Carrier Multiplexed Radio Over Fiber Link for the Simultaneous Support of WLAN and WCDMA Systems

Transcription

1 Wireless Personal Communications (005) 33: 1 0 DOI: /s C Springer 005 Analysis of Sub-Carrier Multiplexed Radio Over Fiber Link for the Simultaneous Support of WLAN and WCDMA Systems ROLAND YUEN AND XAVIER N. FERNANDO Ryerson University, Toronto, Ontario, Canada ryuen@ee.ryerson.ca; xavier@ieee.org Abstract. The present third generation (3G) wireless technology can provide data oriented applications. However, the bit rate is limited to around Mbps with limited mobility. Today, more applications demand high data rate and reasonable mobility. Therefore, by integrating 3G cellular system and wireless local area network (WLAN), there is a potential to push the data rate higher. This integration means 3G cellular users can enjoy high data rate at a location that is within WLAN coverage area. Similarly, WLAN users also can have data services as long as they are under the coverage of the 3G cellular system. The 3G cellular system has a much larger coverage than the WLAN. In this paper, we present the first step toward an integration of the two systems. This paper presents a fiber-wireless architecture that simultaneously supports the wideband code division multiple access (WCDMA) system and the IEEE 80.11b WLAN. Our approach uses sub-carrier multiplexed (SCM) architecture to combine and transmit.4 GHz WLAN and 1.9 GHz WCDMA signals through an optical fiber from a central base station (CBS) to a radio access point (RAP, single antenna unit). After the fiber, the signals continue to propagate through the air interface to respective mobile stations. The WLAN access point is also located at the CBS. For the SCM architecture, we investigate three areas: i) the signal to noise ratio of the uplink and the downlink, ii) the cell coverage area for the WCDMA and WLAN systems, and iii) the throughput of the IEEE 80.11b WLAN. Our results show that with up to.5 km cell radius, better than 18 db SNR is possible with 5 km fiber link for WLAN system. Simultaneously, the WCDMA system has at least 18 db SNR for a cell coverage radius of 8 km. These numbers depend on the relative RF power of each system in the fiber. Keywords: sub-carrier multiplexed, radio over fiber link, wideband code division multiple access, wireless local network, IEEE Introduction The sub-carrier multiplexed (SCM) architecture integrates both the 3G wideband code division multiple access (WCDMA) system and the IEEE 80.11b wireless local area network (WLAN, Figure 1). It uses the SCM technique to carry RF signals of both systems through optical fibers between a central base station (CBS) and a radio access point (RAP). The optical fiber that supports this communication is called the radio over fiber (ROF) link. The RAP operates like an extended antenna from the base station. It transmits and receives the signals of mobile stations (MSs). For the WCDMA system, an MS can be a cellular phone. For the WLAN system, the MS can be a laptop with IEEE 80.11b interface. The SCM architecture is illustrated in Figure. The integration of the two systems is responding to the demands for high data rate applications and reasonable mobility. The employment of the ROF link in the SCM architecture allows reduction in cell size that increases the frequency reuse, thus improves the spectrum efficiency. The RAP with relatively simple functions not only is inexpensive, and its compact

2 R. Yuen and X.N. Fernando Figure 1. Microcellular architecture that employs radio over fiber link. Figure. The sub-carrier multiplexed architecture. size can save estate cost. A centralized network can be constructed using ROF links to connect multiple RAPs to a single CBS. This is illustrated in Figure 1. The advantages of this network are: i) flexible radio spectrum management, ii) flexible data flow control, iii) sharing the cost of the CBS deployment and operation and iv) an efficient handover across systems. There is significant work done in the wireless access using the ROF link by many authors. Tonguz et al. [1] has investigated the personal communications access networks using SCM ROF link. They have derived the carrier to noise ratio of the link that includes the optical noise and the nonlinear distortion in the link. Moreover, they also investigated the cell coverage of the link. Fernando and Anpalagan [] also studied the ROF link, and they have derived the cumulative SNR for the combination of the ROF link and the air interface. They also studied the relationship between the cumulative SNR and the SNR after the optical link. Walker et al. [3] presents the criteria of optimizing the carrier-to-noise ratio of the sub-carrier multiplexed optical network. The criterion is the optimal choice of an optical modulation index that is used to modulate a laser. This optical modulation index is dependent on the number of channels and the nonlinearity of the entire optical link. The relative intensity noise in the ROF link is further improved by Fernando in [4] and this improved expression gives a more accurate model for the ROF link. In the paper by Kim and Chung [5], they have investigated several passive optical networks that support narrow band CDMA signal in microcellular communication system. One of the passive optical networks they investigated employs the SCM technique, and they derived the carrier to noise and distortion ratio for the network. Fan et al., in [6] investigated employment of ROF link in microcellular personal communication system. They have included the fading and the co-channel interference of the air interface to improve their model. They also compared the performance between the uplink and the downlink.

3 Analysis of Sub-Carrier Multiplexed Radio 3 The performance of WLAN is also dependent on the throughput of IEEE 80.11b. The throughput is a medium access control (MAC) layer issue. The throughput might be lower in the proposed architecture; because of the extra distance (in the fiber) the radio signal has to travel. Many authors have worked to develop an accurate model for the throughput of the IEEE 80.11b standard. Bianchi in [7] developed the model using a Markov chain, and assumed unlimited packet retransmissions. Wu et al. in [8] improved Bianchi s work to include packet retransmission limit. Chatzimisos et al. in [9] further improved the model by including the bit error rate of the channel. In this paper, we analyse the quality of the sub-carrier multiplexed (SCM) architecture that provides the third generation WCDMA and the WLAN IEEE 80.11b services as well as the throughput of the WLAN IEEE 80.11b in this architecture. In the analysis, we derive the SNR expressions for both the uplink and the downlink at the optical receiver, the RAP, or the MS (see Figure ). This analysis allows the investigation into the quality of the two systems in the SCM architecture. It also provides system design parameters: the cell coverage radius and the length of the ROF link for both systems. The throughput analysis focuses on modifying the parameters in Chatzimisos s throughput model [9] to suit the SCM architecture. In section two of this paper, we derive the SNR expressions for both the uplink and the downlink of the SCM architecture. The SNR expressions include two parameters that are important for the system design. These parameters are the cell coverage radius and the length of the ROF link. In this section, we also investigate the throughput of the WLAN IEEE 80.11b in the SCM architecture. Section three of this provides the numerical results of the uplink SCM architecture. The results include the SNR of the stand-alone ROF link. The three results that generated for both systems are: i) the coverage area with respect to the cumulative SNR of the SCM architecture, ii) and the effect of one system on the other. The last result is the throughput efficiency of the WLAN IEEE 80.11b in both the SCM architecture and the normal architecture. We discuss the results in section four. The discussion includes the behavior of the ROF link, the reason for only considering the uplink, the effect of power distribution that affects the systems, and the throughput efficiency of the SCM architecture.. Sub-Carrier Multiplexed Architecture The SCM architecture analyzed in this paper employs a single mode ROF link. The WCDMA and the WLAN IEEE 80.11b signals are transmitted through the ROF link simultaneously and provide services to many mobile stations. Figure 1 illustrates the basic architecture of microcellular system that employs the ROF link. Here, a CBS with centralized processing is linked to many remote base stations, also known as the RAP, via fiber. The RAP serves as an extended antenna and provides wireless access to mobile stations. Each RAP can provide (low bit rate) service for large coverage area through the WCDMA interface and high-speed services for smaller coverage area through the WLAN interface. The RAP consists of simple devices and its main function is to convert received electrical signals to optical signals and vice versa. To study the performance of the SCM architecture, we derive the SNR at various locations (see Figure ) in this section. This section also investigates the cell coverage aea and the length of the ROF link of both systems, and the throughput efficiency of the WLAN IEEE 80.11b system. The characteristics of the WLAN IEEE 80.11b and the WCDMA signals are listed in Table 1.

4 4 R. Yuen and X.N. Fernando Table 1. The characteristics of the WLAN IEEE 80.11b signal and the WCDMA signal IEEE 80.11b WCDMA Transmission technology DSSS, FHSS DSSS Data rate 1,, 5.5, 11 Mbps Mbps Modulation scheme DBPSK, DQPSK, GFSK, 4GFSK QPSK Bandwidth MHz 5 MHz Frequency range GHz GHz, GHz.1. U PLINK The uplink starts with transmitted signals from mobile stations (MSs) propagate through the air interface, and then the RAP receives the signals. In the WCDMA system, there are n users denoted by i = 1,,...,n simultaneously sharing the same bandwidth. In the WLAN IEEE 80.11b system, there are many MSs trying to access the system. However, the IEEE 80.11b is designed to allow only one user to access at a time. Hence, the whole bandwidth is consumed by one user at a time in IEEE 80.11b. The SNR at the RAP of a ith WCDMA user is given as: SNR1 up, wcdma = P t i /L wl (r i ) (1) n wl B wcdma and the SNR of a WLAN IEEE 80.11b user is given as: SNR1 up, wlan = P t,wlan/l wl (r wlan ) () n wl B wlan where, P ti and L wl (r i ) are the transmitted RF power and the RF power loss in the air interface of the ith WCDMA user respectively, B wcdma is the WCDMA system bandwidth. Similarly, P t, wlan, L wl (r wlan ), and B wlan denote the WLAN user transmitted RF power, RF power loss in the air interface, and the system bandwidth respectively. The term n wl in the two SNR expressions (1) and () is the noise and the interference power per unit bandwidth in the air interface, and n wl is different for the WCDMA and the WLAN systems. The interference power of the WCDMA system increases with the number of active users, while the interference power of the WLAN system does not increase because only one user transmits at a time. The asynchronous transmission of the MSs causes interference among WCDMA users in the uplink. This nature of transmission leads to the near far effect. To avoid the near-far effect power P control is assumed, so the power at the RAP, ti L wl (r i is the same for all n users. ) The RF power loss in the air interface is expressed as a function of the distance r i or r wlan between the MS and the RAP; it is modeled by a large-scale propagation model. We modified the expression (13) for the received signal power in [1] to express it in terms of the RF power loss and the 90% confidence coverage radius that accounts for the statistical power fluctuation in the air interface. The modified expression is given as: ( 4π R90 ) γ (3) L wl (R 90 ) = 1 S λ10 ( 0.13σ/γ) where, S is the parameter that reflects the shadowing effect, γ is the path loss exponent, λ is the wavelength of the transmitted signal, and R 90 is the 90% confidence coverage radius.

5 Analysis of Sub-Carrier Multiplexed Radio 5 The term 10 ( 0.13σ γ ) comes from the relationship between the 90% confidence coverage radius R 90 and the average coverage radius r [10], where σ and γ are the standard deviation of the channel loss and the path loss exponent respectively. At the RAP point, the WCDMA signals are amplified with a power gain of G up, wcdma to compensate for the loss in the air interface. Similarly, the WLAN signals are amplified with a power gain of G up, wlan. Then, the combined signal of both systems modulates the laser. The RF power of the n WCDMA users that modulates the laser is given as: P RF, wcdna = G up, wcdma n i=1 P ti L wl(ri ) = m wcdma P o (4) Similarly, the RF power of the WLAN user is given as: P RF, wlan = G up, wlan P t,wlan L wl (r wlan ) = m wlan P o (5) where, m wcdma and m wlan are the system optical modulation indices for the WCDMA and the WLAN systems respectively, and P o is the average optical power output from the laser. In expressions (4) and (5), there are a factor of 1. This factor comes from the power of the sinusoidal carrier, cos ω c t. The total RF power including the biased power that modulates the laser is given as: P RF, laser = P o + m wcdma P o + m wlan P o (6) The mean optical power P o is the same for the two systems because the same bias current is used for the laser. However, the system optical modulation indices for both systems are different. The square of each system optical modulation index is proportional to the RF power of the system that modulates the laser. Therefore, the two optical modulation indices (m wcdma and m wlan ) can be used to compare the RF power of both systems in the ROF link. The relationship between the system optical modulation index and the RF amplifier power gain for the WCDMA system is given as: G up, wcdma = m wcdma P o n i=1 L wl (r i ) P ti (7) Where, n indicates this power gain is applied to the whole WCDMA system with n users. Similarly, the relationship between the system optical modulation index and the RF amplifier power gain for the WLAN system is given as: G up, wlan = m wlan P o Lwl(r wlan ) P t,wlan (8)

6 6 R. Yuen and X.N. Fernando The RF amplifier power gain G wcdma and G wlan should bring the n WCDMA users and the WLAN user signals received at the RAP up to the optimal laser input level. The performance of the ROF link is determined by the optimal optical modulation index m opt that is optimized considering the optical noise and the nonlinear distortion in the ROF link. The optimal optical modulation index affects the level of total RF power that modulates the laser. The total RF power that modulates the laser not including the biased power is equal to the sum of the WCDMA and the WLAN RF powers: m opt P o = m wcdma P o + m wlan P o The relationship of the optimal optical modulation index to the two system optical modulation indices is given as: m opt = m wcdma + m wlan (10) When the sum of the m wcdma and m wlan is larger than m opt that means the total power input to the laser exceed the optimal level, and the nonlinear distortion in the ROF link will become dominant and limit the overall performance. On the other hand, when the summation is smaller than m opt, then the optical noise in the ROF link is the limiting factor. Finally, the performance of the entire uplink depends on the cumulative SNR at the output of an optical receiver. For a ith WCDMA user, the cumulative SNR is given as: (9) SNR up, wcdma = m wcdma P o /(Ln op) ( ( Gup, wcdma ) ) (11) nwl L op + nop Bwcdma + n NLD This cumulative SNR is divided by n, number of WCDMA users, because the power of an individual user is the RF power of the whole WCDMA system m wcdma P o divided by the number of users. The cumulative SNR for the WLAN user is given as: SNR up, wlan = m wlan P o /(L op) ( ( Gup, wlan ) ) (1) nwl L op + nop Bwlan + n NLD where, n op is the optical noise per unit bandwidth, n NLD is the nonlinear distortion, and L op is the RF power loss in the ROF link and it had been derived in []. The RF power loss in the ROF link is given as: L op,db = 0 log(g m R) + (n c l c + αd) L op = 10 (L op,db/10) (13) This expression assumes perfect impedance matching in electronics. It also accounts for four optical parameters: the modulation gain G m of the laser, the responsivity R of a photodetector, n c connectors each has loss l c, and the fiber attenuation α and the fiber length d in km. The noise in the ROF link are shot noise, thermal noise, and relative intensity noise [4]; the nonlinear distortion includes clipping distortion and third order intermodulation distortion. The shot noise and thermal noise are independent of the optical modulation index m opt. The

7 Analysis of Sub-Carrier Multiplexed Radio 7 relative intensity noise originates from the laser is dependent on the square of the photocurrent R P o and the optical modulation index m opt [4]. The clipping distortion and the third order intermodulation disortion also depend othe optical modulation index m opt [1]. The square of the optical modulation index m opt is proportional to the total RF power that modulates the laser. The cell coverage radius for the WCDMA and the WLAN systems in the uplink are determined from the cumulative SNR expressions (11) and (1) respectively. The cell coverage radius is the distance r i or r wlan in the cumulative SNR expressions that yields the required minimum SNR for the corresponding systems... D OWNLINK To complete the SNR evaluation of the SCM architecture, we now consider the downlink. The downlink is also illustrated in Figure. In the downlink, the signals for the WCDMA and the WLAN systems are first transmitted from the CBS to the RAP, then to the MSs. In order to simplify the expressions, we assume that the transmitted RF power of n WCDMA users are the same. The SNR at the optical receiver for a ith WCDMA user is given as: SNR1 down, wcdma = m wcdma P o /(nl op) n op B wcdma + n NLD (14) and for a WLAN user is given as: SNR1 down, wlan = m wlan P o /(L op) n op B wlan + n NLD (15) where, n is the number of WCDMA users, m wcdma is the system optical modulation index of the WCDMA system and m wlan is the system optical modulation index of the WLAN system. The term L op, n op and n NLD are the same parameters used in the cumulative SNR expressions (11) and (1) for the uplink. For the optimal performance, the sum of m wcdma and m wlan should not exceed the square of the optimal optical modulation index m opt. The performance of the entire downlink depends on the cumulative SNR at the MS. The SNR of a ith WCDMA user is given as: ( m wcdma Po SNR down, wcdma = /(nl op) )( G down, wcdma ) L wl (r i ) ( ( Gdown, wcdma ) ) (16) nrof L wl (r i ) + nwl Bwcdma and the SNR of a WLAN user is given as: ( m wlan Po SNR down, wlan = /(nl op) )( G down, wlan ) L wl (r wlan ) ( ( Gdown, wlan ) ) (17) nrof L wl (r wlan ) + nwl Bwlan where, the total noise and distortion power in the ROF link n ROF per unit bandwidth is n op + n NLD B wcdma for the WCDMA and n op + n NLD B wlan for the WLAN, G down, wcdma and G down, wlan are the RF amplifier power gain at the RAP for the WCDMA and the WLAN systems respectively, L wl (r i ) and L wl (r wlan ) are the RF power loss in the air interface of the ith WCDMA

8 8 R. Yuen and X.N. Fernando user and the WLAN user respectively. The RF power loss in the air interface is expressed in expression (3). In the air interface, the interference power of the WCDMA system in the downlink does not increase with the number of users because the synchronized transmission keeps the signals orthogonal. On the other hand, the WLAN is assumed to have only one user consuming the channel at a time. Therefore, multiple access interference does not contribute to the interference power. However, the co-channel interference does exist for both systems. The downlink cell coverage radius for the WCDMA and the WLAN systems are determined from the cumulative SNR expressions (16) and (17). The cell coverage radius is the distance r i or r wlan in the cumulative SNR expressions that yields the required minimum SNR for the corresponding system..3. T HROUGHPUT AND PACKET DELAY FOR IEEE 80.11B The IEEE 80.11b standard employs a distributed coordination function (DCF) for the medium access control. The DCF is based on carrier sense multiple access with collision avoidance (CSMA/CA) that provides asynchronous access to the medium with exponential backoff. Here, we briefly describe the DCF basic access method and if readers who want more details should refer to chapter 9 of [11]. When a packet is ready to be transmitted, the station listens to the channel for idle until the duration equals a distributed interframe space (DIFS). If the channel is sensed busy before the end of the DIFS duration, it will not transmit until the channel is idle for another DIFS duration. After a DIFS, the station backs off for few slot times before transmission. A slot time is defined as the time for a station to detect a transmission from any other station. The number of backoff slot times is uniformly chosen from (0, CW-1) where CW is the current contention window. In the first attempt, the current contention window size is at its minimum and it doubles for every unsuccessful transmission until the specified maximum is reached. During the backoff period if the channel is busy, the backoff counter will stop decrementing until the channel becomes idle for a DIFS. When the backoff counter reaches zero, the station will transmit the packet. When the receiving station receives a packet, it will transmit a positive acknowledgement (ACK) after a short interframe space (SIFS). The received station can transmit the ACK without collision because the immediate waiting time, SIFS, allows the station to access the medium faster. A SIFS time together with the channel propagation time is shorter than a DIFS time. That means the station can grab the medium faster than any other station. A successful transmission is when the ACK received within the ACK timeout period. An unsuccessful transmission is when the transmitted station does not receive the ACK or it received a packet. Then, the transmitted station needs to retransmit. The two interframe spaces, DIFS and SIFS, mentioned above are the time intervals between each frame. They are independent of the bit rate of the station. The duration of interframe spaces are determined by the characteristics of the physical layer. A different modulation implies a different physical layer. For a different physical layer, there is a different set of value for the interframe spaces. For IEEE specified in [11], the durations of the interframe spaces are specified at Table 57a for frequency-hopping and Table 59 for direct sequence. The tables does not list the durations for DIFS because DIFS can be determined from a timing relation defined in [11]. In section of [11], the timing relations for a slot time and a DIFS time

9 Analysis of Sub-Carrier Multiplexed Radio 9 is given as follows: aslottime = accatime + arxtxturnaroundtime + aairpropagationtime + amacprocessingdelay (18) T DIFS = T SIFS + aslottime (19) where the values for accatime, arxtxturnaroundtime, aairpropagationtime, amacprocessingdelay, T SIFS, and aslottime are listed in the two tables mentioned above. In the SCM architecture, the physical layer has an additional ROF link. This additional ROF link distance increases the propagation time of the signals, so the timing expression for a slot time should be modified to adjust for the change. To modify the expression, a propagation time through the ROF link is introduced. The propagation time in terms of the fiber core index of refraction n and the distance of the ROF link d is as follows: arofpropagationtime = n c d (0) where, c is the speed of light. The new expression for a slot time is given as follows: and aslottime = accatime + arxtxturnaroundtime + δ + amacprocessingdelay (1) δ = aairpropagationtime + arofpropagationtime () where, δ is the total propagation time including the air and ROF link propagation times. Expression (19) for the DIFS time depends on the slot time. For the SCM architecture, the DIFS time is also increased as the slot time is increased. Since the slot time and the DIFS time are dependent on the distance of the ROF link, for different ROF link distance there would be different values for the DIFS time and the slot time. Therefore, it is possible to assign a set of DIFS time and slot time values for a range of ROF link distances. It will be shown later that when the slot time is larger than the minimum required, the throughput only slightly reduces. The saturated throughput model has been derived in [7], then Chatzimisios et al. [9] improved it to a more accurate model that includes the bit error rate of the channel and the limited packet retransmission. The saturated throughput refers to the system condition where n contenting stations always have a packet ready to transmit. The saturated throughput efficiency, S, isgiven as: S = P tr P s l E[slot] = P tr P s l (1 P tr )σ + P tr P s T s + P tr P c T c + P tr P er T er (3) Where, P tr is the probability that at least one transmission occurs in a randomly chosen slot time, P s is the conditional probability that this transmission is successful, and l is the packet size. The E[slot] isthe average length of a slot time that includes the average duration of an empty slot (1 P tr )σ where σ is a duration of an empty slot, and the average time that a station senses for a successful transmission, a collision, or a transmission error. The probability P c and P er are the probability of collision in transmission and error in received packet respectively. The time T s, T c and T er are the average time that a station sense busy due to a successful

10 10 R. Yuen and X.N. Fernando transmission, a collision, and an error in received packet respectfully. In [9], the authors did not define those times. However, the authors include [7] as their reference which defines T s and T c. The time T s and T c are given as follows: T s = T DIFS + H + P + δ + T SIFS + T ACK (4) T c = T DIFS + H + P + δ (5) where, T DIFS, T SIFS and δ are the time for a DIFS, a SIFS, and a propagation delay respectively, H is the duration of the physical layer header and the MAC layer header, P is the duration of the packet payload, and T ACK is the duration of a positive acknowledgment. The time T er can be derived according to the scenario where an error in received packet. When an error occurs in a receive packet, the sending station simply will not receive an acknowledgement and will retransmit after an acknowledgement timeout. Therefore, the time T er can be given as: T er = T DIFS + H + P + T ACKTimeout (6) where T ACKTimeout is the acknowledgement timeout. Chatzimisios et al. [9] presents the average packet delay E[D] as E[D] = E[X]E[slot] where, E[X]isthe average number of slot times required for a successful transmission. The WLAN signal in the SCM architecture travels the additional distance through the ROF link. This leads to 5 µs and 50 µs increase in the propagation delay for a 5 km and a 10 km ROF links respectively. This increase does not affect the accuracy of the saturated throughput and the average packet delay expressions. However, the duration for a DIFS and a slot time should be determined from the modified slot time expression (1) that is suitable for the SCM architecture. With the increase in the DIFS time and the slot time, we can expect a roll back in the throughput and an increase in the average packet delay. 3. Numerical Results In our calculation, we only consider uplink of the SCM architecture and the reason for this will be explained in the discussion section. The values of all parameters that are used in the calulation are listed in Table. The RF power loss in the air interface is calculated from expression (3). The RF power loss in the ROF link is calculated from expression (13). The RF optical noise power n op is calculated from the expressions found in [4], and the expressions are given as: n shot = qrp o n th = 4FK bt o R L n RIN = RIN R P o ( 1 + m rms / ) (9) where, n shot is the shot noise, n th the thermal noise and n RIN the relative intensity noise. In expression (7), q is the electron charge, R is the responsivity of the photodiode, and P o is the average optical power at the photodiode. In expression (8), F is the noise figure of the optical receiver, K b is Boltzmann s constant and R L is the load resistance of the optical receiver. In (7) (8)

11 Table. Parameters used in the numerical results Analysis of Sub-Carrier Multiplexed Radio 11 P ti, P t,wlan MS transmission power 10 dbm n wl Noise and interference power 10 5 W/Hz γ Path loss exponent σ Standard deviation of channel gain 4dB S Shadowing effect parameter 0.01 λ Average radio signal wavelength 0.15 m B wcdma Bandwidth of WCDMA 5 MHz B wlan Bandwidth of WLAN MHz P o Laser mean optical power 1 mw m opt Optimal optical modulation index 0.107, 0.14, and optimized for 1 km, 5 km, and 10 km G m Laser modulation gain 0.1 A/W R Photo diode responsivity 0.75 W/A n c Number of optical connectors l c Optical connector loss 1 db α Fiber attenuation 0.5 db/km RIN Relative intensity noise parameter 155 db/hz T o Optical receiver temperature 75 K R L Receiver load resistance 50 F Receiver amplifier noise factor 1 a 3 Third-order nonlinearity parameter 0. expression (9), RIN is the relative intensity noise parameter. The root means square (RMS) optical modulation index m rms in expression (9) is the RMS of all the individual system optical modulation indices. Under optimal condition m rms in expressions (30) (3) is the same as m opt in (10). The clipping distortion n cl and the third order intermodulation distortion n 3OI in the ROF link together give the total nonlinear distortion n NLD. The two distortions in [1] are given as: n cl = 1 ( 10 π R Po m5 rms exp 1 ) m rms (30) n 3OI = R Po a 3 m6 rms (31) Figure 3 shows the RMS optical modulation index m rms versus the SNR of the ROF link alone. This figure is generated using the following expression: SNR = m rms P o /(L op) n op (B wcdma + B wlan ) + n NLD (3) where, m rms is the RMS optical modulation index that varies from 0.01 to 1, and L op is the RF power loss in the ROF link that depends on the length of the ROF link. The three SNR curves are generated for 1, 5 and 10 km ROF links.

12 1 R. Yuen and X.N. Fernando Figure 3. Optical modulation index versus the SNR of 1, 5, and 10 km ROF links. Figure 4. Cumulative SNR of the uplink and downlink versus the received power at the RAP (uplink) or the MS (downlink). In the uplink, the ROF link is optimized for a received power of dbm. In the downlink, the MS is expected to receive dbm of power. Figure 4 illustrates the difference in quality requirement for the uplink and the downlink of the SCM architecture. This figure shows the cumulative SNR of the uplink and the downlink versus the received power at the RAP and the MS respectively. Expressions (11) and (16) are used to generate the uplink and the downlink cumulative SNRs respectively. The calculation assumed the ROF link is 5 km and the cell coverage radius is 1 km. In the uplink, the ROF link is optimized for a received power of dbm. In the downlink, the MS is expected to receive dbm of power. Figure 5 shows various uplink SNR of a ith WCDMA user versus the cell coverage radius. There are ten users assumed in the WCDMA system, the same is also assumed for Figures The four SNR curves are the SNR at the RAP, and the cumulative SNR for 1 km, 5 km and 10 km ROF links. Figure 6 shows the same set of SNR curves versus the cell coverage radius, but the curves are the SNR of the WLAN user. Both systems modulate the laser with the same

13 Analysis of Sub-Carrier Multiplexed Radio 13 Figure 5. Cell coverage radius versus various uplink SNR of the ith WCDMA user. The WCDMA system modulates the laser with 50% of the optimal RF power (0.5(m opt P o /)). So, the system optical modulation index of WCDMA is m wcdma = 0.5 m opt. Figure 6. Cell coverage radius versus various uplink SNR of the WLAN user. The WLAN system modulates the laser with 50% of the optimal RF power (0.5(m opt P o /)). So, the system optical modulation index of WLAN is m wlan = 0.5 m opt. RF power. The RF power is 50% of the optimal RF power (0.5(m opt P o /)), so the system optical modulation indices are m wcdma = 0.5 m opt and m wlan = 0.5 m opt. The SNR at the RAP for the ith WCDMA user and the WLAN user are generated from expressions (1) and () respectively. The cumulative SNR for the ith WCDMA user and the WLAN user are generated from expressions (11) and (1) respectively. Figures 5 and 6 are generated with the same scale for the ease of comparison across systems. Figure 7 illustrates the SNR curves from a different perspective. It shows the length of ROF link versus the uplink cumulative SNR. This figure has six SNR curves where three of them are the cumulative SNR of a ith WCDMA user and the other three are the cumulative SNR of the WLAN user. Both systems modulate the laser with same RF power. The RF power is 50% of the optimal RF power (0.5(m opt P o /)). The set of three SNR are generated for a cell coverage radius of 100 m, 500 m, and 1 km using the expressions (11) and (1).

14 14 R. Yuen and X.N. Fernando Figure 7. Length of ROF link versus three different uplink cumulative SNR of the cell coverage radius: 100 m, 500 m, and 1 km. Both systems modulate the laser with the same RF power. The RF power is 50% of the optimal RF power (0.5(m opt P o /)), so the system optical modulation indexes are m wcdma = 0.5 m opt and m wlan = 0.5 m opt. Figure 8. Cell coverage radius versus cumulative SNR. The SNR are generated for ROF link distance of 1, 5, and 10 km. The WCDMA and the WLAN systems modulate the laser with 90% and 10% of the optimal RF power respectively. The system optical modulation index for the WCDMA is m wcdma = 0.9 m opt and the WLAN is m wlan = 0.1 m opt. Figure 8 illustrates the changes to the cumulative SNR when the WCDMA and the WLAN systems modulate the laser with 90% and 10% of the optimal RF power respectively. In this figure, there are six cumulative SNR curves versus the cell coverage radius. Three of the cumulative SNR curves are the cumulative SNR of a ith WCDMA, and other three are the cumulative SNR of a WLAN user. The set of three cumulative SNR curves are generated for 1, 5, and 10 km ROF links. The system optical modulation index for the WCDMA is m wcdma = 0.9m opt and the WLAN is m wlan = 0.1 m opt. The cumulative SNR curves are generated using expressions (11) and (1). Figures 9 and 10 illustrate how the RF power that modulates the laser of one system can affect the other system. The figures show the cumulative SNR to the RMS optical modulation

15 Analysis of Sub-Carrier Multiplexed Radio 15 Figure 9. Cumulative SNR to the RMS optical modulation index m rms. The WLAN system modulates the laser with 50% of the optimal power, while the WCDMA system varies its RF power from 10 00% of the optimal power. m rms is the RMS of the individual system optical modulation index. It is given as 0.5 m opt + P wcdma m opt where P wcdma ɛ [0.1,.0] and m opt is Figure 10. Cumulative SNR to the RMS optical modulation index m rms. The WCDMA system modulates the laser with 50% of the optimal power, while the WLAN system varies its RF power from 10 00% of the optimal power. m rms is the RMS of the individual system optical modulation index. It is given as 0.5 m opt + P wlan m opt where P wlan ɛ [0.1,.0] and m opt is index m rms. The cumulative SNR curves are calculated for a 100 m cell coverage radius and a 5 km ROF link. In Figure 9, the WLAN system modulates the laser with 50% of the optimal RF power, while the WLAN system varies its RF power from 10% to 00% 0.5m opt + P wcdma m opt of the optimal RF power. The RMS optical modulation index m rms is where P wcdma ɛ [0.1,.0]. Figure 10 shows the opposite by keeping the modulated RF power of the WCDMA system constant at 50% of the optimal RF power and the WLAN system varies the RF power from 10% to 00% of the optimal RF power. The RMS optical modulation 0.5m opt + P wlan m opt where P wlan ɛ [0.1,.0]. The cumulative index m rms in Figure 10 is SNR curves are generated using expressions (11) and (1). Figures 9 and 10 are generated with the same scale for the ease of comparison.

16 16 R. Yuen and X.N. Fernando Figure 11. Saturated throughput efficiency of the SCM architecture and the normal IEEE 80.11b architecture. The SCM architecture has a 10 km ROF link. Figure 11 shows the saturated throughput efficiency of the WLAN IEEE 80.11b versus the number of contenting stations. The throughput effciency is the throughput bit rate normalized by the bit rate that is provided by the physical layer. The two throughput curves in Figure 11 are the WLAN IEEE 80.11b throughput in the SCM architecture and in the normal WLAN architecture that only has an air interface. According to Table 59 of [11] for direct sequence physical layer characteristics, the normal WLAN architecture has a slot time of 0 µs and a SIFS time of 10 µs. Using timing relation in (19), the DIFS time is 50 µs. For a SCM architecture with a 10 km ROF link, the SIFS time stays the same, while the slot time is calculated using the modified expression (1). The corresponding slot time and DIFS time are 70 µs and 150 µs respectively. The throughput curves are generated first by solving two nonlinear expressions (1) and () in [7] numerically, then uses expressions (3) (6) to calculate the throughput efficiency. 4. Discussion In this section, we discuss the role of the RMS optical modulation index, the quality of the uplink and the downlink, the effect of one system on the other, and the saturated throughput of the IEEE 80.11b in the SCM architecture. The quality of the ROF link is closely related to the RMS optical modulation index. This optical modulation index is the RMS of the sum of individual system optical modulation indices. The RMS optical modulation index limits the amount of RF power that modulates the laser. When only considering the optical noise, a large SNR can be achieved simply by increasing the RMS optical modulation index. However, as this optical modulation index passes its optimal level, the nonlinear distortion becomes dominant and degrades the SNR. Figure 3 illustrates such behavior. This figure also shows there is an optimal optical modulation index m opt for each SNR curves that gives the highest SNR. These optimal optical modulation indices are 10.7%, 1.4% and 15.0% for 1 km, 5 km and 10 km ROF links respectively. The optimal optical modulation index increases with the length of the ROF link, which means more power can be modulated the laser for a longer link distance. However, the SNR decreases with the length of the ROF link, even with the optimal optical modulation index. This is because of fiber attenuation.

17 Analysis of Sub-Carrier Multiplexed Radio 17 The reason for only considering the uplink of the architecture is due to the quality of the uplink is more demanding than the downlink. Generally, the quality of the SCM architecture is limited by the lower SNR of the two interfaces; the ROF link and the air interface. However, depending on which interface goes first, the quality can be different. In the downlink, the WCDMA and the WLAN systems modulate their signals to the optimal level of the ROF link, and then signals propagate through the air interface to MSs. The quality of the downlink is determined from the cumulative SNR at the MS. In the uplink, signals first propagate through the air interface, then through the ROF link. The quality of the uplink is determined from the cumulative SNR at the optical receiver. The air interface has large power fluctuation because of the shadowing effect and the fading effect. We can observe from Figure 4 that the cumulative SNR of the uplink exhibits the same shape as the SNR of Figure 3. This SNR curve shows that if the received power is larger or smaller than the optimized received power of dbm, it results in a lower SNR. The SNR curves of the downlink is very much proportional to the received power at the MS. Comparing the uplink and the downlink SNR curves, there is only a small received power range from dbm to dbm where the uplink SNR is better than the downlink SNR, and everywhere else the downlink SNR is better than the uplink. The range of received power in Figure 4 is well within the typical power fluctuation of the air interface. Therefore, the uplink, where the air interface follows by the ROF link is more demanding than the downlink. We can conclude that the quality of the uplink limits the quality of the whole SCM architecture, thus most of the calculations consider only the uplink. The performance of the SCM architecture for the WCDMA and the WLAN systems can be observed from the SNR curves of the uplink in Figures 5 8. In Figure 5, the three cumulative SNR curves (SNR up, wcdma ) follow the shape of the SNR at the RAP (SNR1 up, wcdma ) after 4 km of the cell coverage radius. That means when the cell coverage radius is larger than 4 km, the SNR of the air interface is the limiting factor. When the cell coverage radius smaller than 4 km, the SNR of the ROF link is the limiting factor. The similar trend is even more clearly shown in Figure 6 where the cumulative SNR (SNR up, wlan )isclosely resemble the SNR at the RAP (SNR1 up, wlan ) for the cell coverage radius greater than 3 km. Beyond the 3 km cell coverage radius, the length of the ROF link does not have much effect on the SNR. That means the SNR of the air interface is the limiting factor. When comparing the two systems, the SNR of a ith WCDMA user is better than the SNR of a WLAN user for the cell coverage radius larger than 4 km. This is because the ith WCDMA user has a smaller bandwidth, so the noise and the interference power is also smaller and results in a better SNR than the WLAN user. In other words, the WCDMA system can provide a larger coverage radius in this case. Figure 7 gives insights to the WCDMA and the WLAN system performances versus the length of the ROF link as well as the performance versus the power distribution of systems. From this figure, we observe a general decline of the cumulative SNR curves as the length of the ROF link increases. Although the power that modulates the laser for both systems are equal, Figure 7 shows the SNR of the WCDMA user is 3 to 5 db lower than the SNR of the WLAN user. This is because there are ten users in the WCDMA system who share the available power that modulates the laser, while the WLAN system has only one user at a time in the system. Figure 8 shows a fairly even power distribution among all the users in both systems. The WCDMA system with ten users modulates the laser with 90% of the total optimal RF power and the WLAN system modulates 10% of the total optimal RF power. For the cell coverage radius within 1 km, the SNR of the WCDMA user is about db better than the SNR of the WLAN user. As the cell coverage radius increases, the margin between the SNR of the WCDMA user and the SNR of the WLAN user also increases. For a large cell coverage

18 18 R. Yuen and X.N. Fernando radius, the SNR of the WLAN user always has a lower SNR because the larger bandwidth of the WLAN user significantly reduces its SNR in the air interface. If a system needs a larger coverage area, then a larger RF power needs to be transmitted. In Figure 9, the SNR for the WCDMA user peaks at around 13% of the RMS optimal modulation index which is closed to the optimal optical modulation index of 1.4% for a 5 km ROF link. The cumulative SNR curve of the WLAN user is always decreasing because the power that modulates the laser keeps constant while the noise and distortion power increases due to the increase of the total power that modulates the laser. The rate of the WLAN SNR decline is larger when the RMS optimal modulation index passes the optimal optical modulation index of 1.4%. The increase in the decline rate is due to the nonlinear distortion of the ROF link. In Figure 10 where the WCDMA system RF power that modulates the laser is constant, there is a much sharper decline in the SNR of the WCDMA user. The sharp decline in the SNR is because the power of individual WCDMA user is small compares to the noise and the distortion power brought by the power of the WLAN system. When the signal power of a user is small, the power of the other system can have large effect on the SNR of that user just as the WCDMA user in Figure 10. Moreover, the best SNR of the WLAN user does not happen at the optimal optical modulation index, and the SNR curve is relatively flat compared to the SNR curve of the WCDMA user in Figure 9. This is mostly due to the larger power of the WLAN user. The IEEE 80.11b router is located at the CBS, and the additional distance to the RAP increases the collision detection time, and Figure 11 shows the reduction in the throughput of the WLAN IEEE 80.11b system is about 0.01 of saturated throughput efficiency. This is very low even for significantly long (10 km) fiber. Therefore, reduction in throughput is not a concern with this scheme. 5. Conclusion In this paper we have analyzed a SCM architecture that supports both WLAN IEEE 80.11b and WCDMA services. This architecture employs ROF links with centralized processing in the CBS. SCM refers to the transmission of RF signals from both systems through the fiber in frequency multiplexed manner. The main advantage of this architecture is having relatively simple single antenna RAP that translates to lower deployment cost. Our numerical analysis indicates that this system has the potential to support both WLAN and WCDMA signals from reasonably long fibers into reasonable size cells. For example considering 5 km ROF link, it is possible to have better than 18 db SNR within.5 km cell radius for WLAN system and, 8 km cell radius for WCDMA system simultaneously. The cell sizes can be varied by changing the relative RF power through the fiber. In the SCM scenario, WLAN system throughput should also be analyzed because of the additional delay involved in the ROF link. However, the reduction in the throughput is not seemed significant from our analysis. The reduction is about 0.01 in saturated throughput efficiency for 10 km ROF link. This is because the effect of propagation delay on the distributed interframes space (DIFS) is very small. References 1. O.K. Tonguez and H. Jung, Personal Communications Access Networks Using Subcarrier Multiplexed Optical Links, Journal of Lightwave Technology, Vol. 14, No. 6, pp , 1996.

19 Analysis of Sub-Carrier Multiplexed Radio 19. X.N. Fernando and A. Anpalagan, On the Design of Optical Fiber Based Wireless Access Systesm..., in Proc. Int. Conf. on Communication, Paris, France, 004, pp S.D. Walker, M. Li, A.C. Boucouvalas, D.G. Cunningham, and A.N. Coles, Design Techniques for Subcarrier Multiplexed Broadcast Optical Networks, Selected Areas in Communications, IEEE Journal, Vol. 8, No. 7, pp , September X.N. Fernando, An Improved Expression for Dynamic Relative Intensity Noise in Radio over Fiber Applications, Under review to be published in IEEE Transactions on Communications, H. Kim and Y.C. Chung, Passive Optical Network for CDMA-Based Microcellular Communication Systems, Journal of Lighwave Technology, Vol. 19, No. 3, pp , J.C. Fan, C.L. Lu and L.G. Kazovsky, Dynamic Range Requirements for Microcellular Personal Communication Systems Using Analog Fiber-Optic Links, Microwave Theory and Techniques, IEEE Transactions on, Vol. 45, No. 8, pp , G. Bianchil, IEEE Saturation Throughput Analysis, IEEE Communications Letters, Vol., No. 1, pp , December H. Wu, Y. Peng, K. Long, S. Cheng and J. Ma, Performance of Reliable Transport Protocol Over IEEE Wireless LAN Analysis and Enhancement, in Proc. Twenty-First Annual Joint Conference of the IEEE Computer and Communications Societies, INFOCOM 00, IEEE, 00, Vol., pp P. Chatzimisios, A.C. Boucouvalas and V. Vitsas, Influence of Channel BER on IEEE DCF, Electronics Letters, Vol. 39, No. 3, pp , W.C.Y. Lee, Mobile communications Engineering, New York, McGraw Hill, ISOL/IEC :1999(E), IEEE standard for Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, Roland M.C. Yuen received a Bachelor of Electrical Engineering degree in 003 from Ryerson University, Toronto, Canada. He is pursuing a Master of Applied Science degree at Ryerson University. He has a conference paper. His research interests are in the area of optical and wireless communications. Currently, he works on unique fiber based architecture to extend the capability of cellular networks and support wireless LANs simultaneously. Xavier N. Fernando ( fernando) obtained B.Sc. Eng. (First Class Honors) degree from Sri Lanka, where he was first out of 50 students. He got Master s degree

20 0 R. Yuen and X.N. Fernando from the Asian Institute of Technology (Bangkok) Ph.D. from the University of Calgary, Canada in affiliation with TRLabs. He has worked for AT&T for three years as an R&D Engineer. Currently he is an Assistant Professor at Ryerson University, Toronto, Canada. Dr. Fernando one US patent and about 38 peer reviewed publications in journals and conference proceedings. His research focuses on signal processing for cost-effective broadband multimedia delivery via optical wireless networks. Dr. Fernando s work won the best research paper award in the Canadian Conference of Electrical and Computer Engineering for the year 001. His student projects won both the first and second prize at Opto Canada the SPIE regional conference in Ottawa in 00. He is a senior member of IEEE, member of SPIE, Chair of the IEEE Communications Society Toronto Chapter and licensed Professional Engineer in Ontario, Canada. He has many research grants including Canadian Foundation of Innovations (CFI), Ontario Innovations Trust (OIT) and Natural Sciences and Engineering Research Council (NSERC) of Canada.