Downlink Interference Analysis of CDMA-based LMDS Networks Applying Multiuser Detection

Csaba Novák, Tamás Pálfalvi, András Tikk and János Bitó

Budapest University of Technology and Economics – Department of Broadband Infocommunication Systems, Goldmann Gy. tér 3., Budapest, H-1111, Hungary

csaba.novak@mht.bme.hu, palfalvi@docs.mht.bme.hu, tikk@docs.mht.bme.hu, bito@mht.bme.hu

Abstract

The subject of this paper is the investigation of code division multiple access (CDMA) in broadband wireless point-to-multipoint networks, such as local multipoint distribution services (LMDS). The most critical point in LMDS networks is inter- and intracell interference. In our work we examine worst-case interference situations on the downlink of LMDS networks applying CDMA. Comparison between CDMA and time division multiple access (TDMA) approaches will be drawn in terms of downlink interference. For effective interference cancellation in CDMA, multiuser detection will be introduced. Downlink investigations of an LMDS sector by carrier to interference ratio (C/I) and bit error rate (BER) maps, BER vs. SNR simulations will be presented.

1. Introduction

Increasing demand on wideband applications set wireless fixed networks into focus as a cheap alternative to wired solutions such as digital subscriber line (DSL). Local Multipoint Distribution Service (LMDS) is a broadband wireless point-to-multipoint (PMP) communication system that provides two-way high speed multimedia transmission [1]. LMDS has to fulfil the requirements of copper- and fiber-based networks besides offering high flexibility in network configuration. However, several disturbing effects occur on the air interface of wireless PMP networks. One of the most critical factors is interference caused by other sectors of the network. Efficient cell planning tool is required to reduce intra system interferences. However, critical subscriber positions in the service area of the LMDS system could exist caused by inter- and intracell interference. To overcome this problem of unacceptable interference situations, sophisticated sectored antenna systems, adaptive modulation techniques, diversity solutions are addressed. Considering multiple access schemes for LMDS, mostly TDMA and FDMA approaches are favored. Although, code division multiple access (CDMA) provides the possibility of applying advanced interference suppression techniques, i.e. multiuser detection (MUD), CMDA-based LMDS networks are not foreseen at the present time.

In our paper we investigate the feasibility of applying MUD-aided CDMA in LMDS networks. This work exploits the experiences of our former investigations, which addressed the performance evaluation of uplink analysis of CDMA-based LMDS networks [2], and the effects of applying different antenna systems in wireless PMP networks [3]. In future PMP solutions asymmetric applications are foreseen, which mean high bandwidth demand on the downlink direction, therefore our investigations are presented by downlink interference simulations. Key system parameters of our present investigations are the applied multiple access scheme, antenna system, and receiver algorithm considering the CDMA solution. In Section 2 an overview of LMDS is given with respect to the interference characteristics. CDMA for LMDS is introduced in Section 3, where comparison with the most common TDMA approach is drawn. CDMA-specific interference situations are demonstrated. CDMA multiuser detection is then introduced for LMDS downlink receivers. Section 4 enumerates the system parameters of our investigated network configurations, emphasizing the effect of the applied antenna parameters. The performance of CDMA-based LMDS is examined in Section 5. Various network configurations are compared by bit error rate (BER) distribution maps of an LMDS sector and BER vs. signal to noise ratio (SNR) analyses for a selected user position.

2. Overwiev of LMDS characteristics

LMDS is a cellular networking solution aiming at providing high speed multimedia services to subscribers with fixed terminal stations. LMDS can also serve as a possible transmission network for future mobile communication systems, such as UMTS, connecting the base station of the mobile network to the central stations (CS) of the LMDS network. In this application the base station of the mobile network will be fed by the fixed terminal station (TS) of the LMDS system. The main concept of the LMDS network architecture can be summarized as serving cheap TSs from CSs. In comparison to the TSs higher CS costs are affordable, allowing the implementation of sophisticated antenna, receiver, etc. solutions. To realize broadband distribution services, the system operates at millimetre frequencies, typically in the range of 20-40 GHz (depending on country of licensing) [1]. The cell radius is limited to 2-5 km due to propagation conditions. Because of the applied high frequencies, line-of-sight connection is a requirement. Sectoring solutions are preferred for frequency reuse. The most common cell sectoring solution uses 90ο sectors, therefore CSs can be arranged in a rectangular grid (e.g. in Figure 1). To reduce intercell interference, terminal antennas use accurate narrow-beam focusing on the selected CS. Primary access methods in LMDS are TDMA, FDMA and CDMA, whereas the latter gained less attention. Currently, most system operators and standards activities address the TDMA and FDMA approaches. Therefore the section hereunder discusses the most commonly used TDMA solution in terms of interference.

2.1. Four-frequency sectored TDMA scheme

TDMA assumes time-frame synchronization of the terminal stations (TS), therefore the central controlling of the CSs is necessary. The loss of frame synchronization leads to severe degradation of TDMA system performance. As a large number of TSs has to be served, sectoring is needed. According to the cellular principle, frequency sectoring allows the reuse of spectrum, increasing capacity. Considering the above discussed properties of LMDS, namely 90ο sector CS antennas and narrow-beam TS antennas, the following four-frequency sectored TDMA configuration can be applied as shown in Figure 1 in case of 3×3 CSs. Four different frequency bands represented by different grey levels are used in the sectors of one cell. Downlink interference situations are denoted between sectors which transmit to the same direction operating at the same frequency. In that case the denoted TS in Figure 1 is disturbed by intercell interference. Similar most disturbing situations occur between all of the second nearest neighbours. Downlink interference situations assuming the same arrangement can be observed in the C/I map of the whole coverage area in Figure 2. calculated with antenna system noted as CS0-TS3 given in Section 4. Critical areas with low C/I values (with dark colours in Fig. 2) are aligned along the sector bor-ders and in the corners of the 3×3 network. Avoiding critical interfering places play the major role in downlink network planning. We focused our attention on the most critical sectors at the corners of the service area. Simulation results will be presented for the bottom left corner sector. Please observe the circled sector of Figure 2 with the corresponding magnified C/I map for comparison to our further investigations.

TS

CS9 CS8 CS7

CS6 CS5 CS4

CS3 CS2 CS1

Figure 1 Downlink interference situations in 4-frequency sectored TDMA

LMDS network with 3×3 CSs

Figure 2 Downlink C/I map for LMDS network with 3×3 CSs applying 4-frequency sectored TDMA. The investigated sector

is indicated with circle.

3. Applying CDMA in LMDS networks

The main advantage of CDMA is the capability of eliminating the influence of frequency selective fading mostly caused by multipath propagation. Because line-of-sight connection is a requirement for PMP systems, applying higher frequencies e.g. 40 GHz, fading due to meteorological phenomena i.e. rain attenuation are the dominant degradation factors rather than multipath fading. Therefore, applying CDMA has not yet been preferred in LMDS systems [4]. However, recent studies have shown that CDMA can compete with TDMA [5] and can play a major role in two-layer LMDS architectures [6].

The advantage of applying CDMA against TDMA is that frequency sectoring can be avoided using the whole available frequency band assuming the existence of satisfactory number of codes. In that case the interference will be determined by the used code set. The further advantage of CDMA systems, however, is that the loss of synchronization is not so critical if the periodic and aperiodic cross-correlation functions of the codes in the code set are satisfactory enough. Additionally, CDMA provides inter-system interference suppression determined by the processing gain allowing the co-existence of different systems operating at the same frequency band. We considered the single-frequency approach, i.e. applying only space sectoring by the 90ο CS antennas. The downlink interference situation in single-frequency (i.e. all sectors apply the same frequency) CDMA-based LMDS networks with 3×3 CSs is depicted in Figure 3. Compared with Figure 1, the number of interfering sectors has increased significantly. In contradiction to the worst-case uplink situation of four-frequency TDMA system, the received signal at the CS in single-frequency CDMA is the superposition of the desired signal (from CS1) and all interfering signals from the sectors looking to the same direction. Figure 4 shows C/I conditions of the CDMA-based network with the same antenna parameters as those of TDMA case (Figure 2). High interference level can be observed, even negative C/I values occur in critical positions. However, inter- and intracell interference can be kept low by applying codes with good correlation properties. Using orthogonal codes, inter- and intracell interference can totally be suppressed. In case of using non-orthogonal codes, multiuser detection can lead to the required BER performance.

Code utilization can play an important role in CDMA-based LMDS networks. In addition to antenna sectoring, further sector separation can not only be realized by using different frequency band, but by applying code sectoring in CDMA. In our approach, downlink code utilization is considered, similar to that of UMTS [7]. A unique pseudo-noise (PN) code is assigned to each sector; therefore sector synchronization is not required. Users of a certain sector are separated by orthogonal Walsh codes. The resulting code system therefore is a combination of quasi-orthogonal PN sequences and orthogonal codes. A certain TS receives the desired signal spread by the PN code identifying its serving sector and also multiplied with its unique Walsh code, identifying the desired downlink connection. Interference will consist of inter-sector: intra- and inter-cell interfering signals. As in the downlink, the CS separates the users’ signals by orthogonal Walsh-codes, intra-sector interference at the TS does not occur. Because the sector identifier PN codes are not orthogonal, inter-sector and inter-cell interference must be taken into account, however applying sectored antennas at the CSs and focused antennas at the TSs, intra-cell interference can be reduced.

TS CS1 CS2 CS3

CS4 CS5 CS6

CS7 CS8 CS9

Figure 3 Downlink interference situations in single frequency CDMA

LMDS network with 3×3 CSs

Figure 4 Downlink C/I map for LMDS network with 3×3 CSs applying single frequency CDMA. The investigated sector is

indicated with circle.

Further demonstration of interference situations in single frequency CDMA-based LMDS network can be made by examining the most interfering sectors in a certain position. This way the investigated sector can be divided into separate zones, which have common main interferers. Figure 5 shows five interference zones of the examined corner sector of the network. In Figures 5 a)-e) the dominant interferers are depicted for the five zones respectively. Sectors in these maps marked with darker tones indicate higher interference level received from the certain sector. The effect of these interference zones can be observed in the magnified sector of the network on the C/I map in Figure 4. Please observe the high interference level along the sector borders in Figure 4 and see the interference zones “d” and “e” in Figure 5. The corresponding maps (Figure 5 d, e) show that most of the interference is caused by the neighbour sector, which is due to the not perfect side lobe suppression of the applied sectored CS antenna.

However, critical TS positions of the network show symmetric arrangement, in practical LMDS network configurations the CS arrangement is not necessarily regular, therefore the marked zones will not be symmetric in real LMDS networks.

3.1. Interference suppression with multiuser detection

The application of CDMA produced inter-sector interference which is the multiple access interference (MAI) caused by non-orthogonal PN codes assigned to the sectors. Multiuser detection aims to remove MAI when detecting the desired user. In our contribution parallel interference cancellation (PIC) MUD scheme is investigated. Assuming a CDMA system with K users transmitting continuously the received signal is given by (1)

( )1

( ) ( ) ( ) ,K

k k k ki k

r t A b i s t iT n tτ∞

=−∞ =

= − − +∑∑ (1)

the kth user is identified by spreading waveform sk, bk(i) denotes the sent ith bit with the duration T, Ak is the received amplitude of the kth user and n is the white Gaussian noise. Applying a bank of K matched filters (MF), the output of the kth filter (MFk) for the ith bit is then

( ) ( ) ( ) ( ) ( ) ( ) ,1

1, nlbAibAdttsiTtr

Tiy

T

MAI

l

K

kjj

jjkkjjkkkkkk

k

+−+=−−−= ∫ ∑∑∞

−∞=≠=

44444 344444 21

ττρττ (2)

a) e)

b)

d)

c)

CS1 CS2 CS3

CS4 CS5 CS6

CS7 CS8 CS9

0.0001

0.001

0.01

(CS1-TS3) (CS1-TS0) (CS0-TS3) (CS0-TS0)

aver

age

BER

TDMA

MF

MMSE

PIC a) b)

CS1 CS2 CS3

CS4 CS5 CS6

CS7 CS8 CS9

CS1 CS2 CS3

CS4 CS5 CS6

CS7 CS8 CS9

CS1 CS2 CS3

CS4 CS5 CS6

CS7 CS8 CS9

c) d) e)

Figure 5 Interference zones (zone “a” – zone “e”) of the bottom left sector (hatched sector) of a single-frequency CDMA-based LMDS network with the corresponding maps (a–e) marking the

most dominant interferer sectors with decreasing grey levels, for each zone respectively

where ρj,k is the cross-correlation between users j and k and MAIk is the interference caused by other users.

The PIC receiver [8] detects all users at the same time and then cancels MAI simultaneously as shown in Figure 6.

The output vector of the first MF bank )0(b̂ is used for respreading and subtracting from the delayed received signal, which then will be the input of the second MF bank to perform the final decision corresponding to:

( ) ( ) ( ) ( )( ) ( )

−

−−−= ∫ ∑≠=T

kk

K

kjj

jjjjk dttsibiTtsAtrib ττ1

0ˆsgnˆ (3)

On the downlink direction the knowledge of the applied codes by other users is a question, therefore MUDs with and without the knowledge of other codes were examined.

In the case when only the code of the desired TS is known, MUD with adaptive minimum mean squared error (MMSE) algorithm [8] was simulated. This detector belongs to the family of adaptive equalizers using training sequences for initial channel estimation. The principle of the algorithm is applying an adaptive finite impulse response (FIR) filter so that the adaptation rule should fulfil the MMSE criterion with respect to the decision error. The rule for the (l+1)th iteration step for filter weight vector w results:

( ) ,T1 llllll b rwrww −+=+ µ (4)

where r is the received signal vector; bl is the lth detected bit; µ is the iteration step size, and ( )T means the transpose function.

4. LMDS system model

The described TDMA and CDMA approaches are compared with the following system assumptions. 26 GHz frequency domain was considered. The applied modulation method was BPSK both for TDMA and CDMA. In our CDMA system bit-asynchronous transmission was considered, assuming chip-synchronism.

Propagation assumption: Free-space line of sight propagation conditions were assumed between all of the terminal and central stations, assuming that the signal attenuation is proportional to the square of the distance. All CSs transmit with the same power.

r(t) Σ

Σ

Σ

T AKsK(t)

Aksk(t)

( )1̂b t

( )k̂b t

( )ˆKb t

A1s1(t)

MF1

MFk

MFK

( )(0)1̂b t

( ) ( )0k̂b t

( ) ( )0ˆKb t

MF1

MFk

MFK

Σ

Σ

Σ

Figure 6 Receiver structure of parallel interference cancellation (PIC)

Applied CDMA codes and receivers: PN codes assigned to different sectors are 64 chip long, constructed from 63 chip long Gold codes extended with one further chip. Users of a certain sector are separated by orthogonal 64 chip long Walsh codes, as described in Section 3. Three detectors are compared: single user detector, which means a matched filter (MF), PIC and adaptive MMSE multiuser detectors (see Section 3.1). For the applied PIC detector, however, we used some simplifications. As it could be seen in section 3 (Figure 5), in most of the TS positions only a few sectors cause most of the interference. Interference suppressing capability of the PIC algorithm can maximally be exploited if all of the codes in the systems are utilized in the interference cancelling algorithm as given in equation (3), which can lead to very complex receiver structure. Receiver complexity can be reduced by applying only a few branches of correlators (see Fig. 6), however in this case interference suppression capability will also be reduced. As we will see, this reduction is not significant on the downlink of the examined LMDS system, a detector exploiting the knowledge of the five most interfering sectors performs nearly the same as for ten sectors. Considering PIC detector, we limited the knowledge of the codes in the system to the five most interfering sectors. We assumed that TSs cancel interference exploiting the knowledge of the PN codes of the five most disturbing sectors, which means at least 80% of total interference according to our calculations. Therefore the implemented PIC detector has matched filter banks only for the five most interfering sectors. For comparison PIC detector with the knowledge of 3, 4 and 10 most disturbing sectors are also simulated.

Antenna systems: 90ο sectored antennas are investigated at the CSs and focused narrow-beam TS antennas were considered in our simulations. In our system model a pair of CS and TS antennas were used according to ETSI recommendations (CS1 and TS3, see [9]), and a pair of commercially available antennas with excellent main lobe gain and side lobe suppression, (named CS0 and TS0) were used. ETSI standards lay down minimum requirements for suppliers, therefore simulations with ETSI-recommended antennas mean a worst case scenario. Radiation patterns of the applied CS and TS antennas are given in Tables 1 and 2 respectively and depicted in the corresponding Figures 7 and 8.

Table 1 CS antenna patterns[9] Table 2 TS antenna patterns [9] CS1 CS0 TS3 TS0

Angle (deg)

Rel. gain (dB)

Angle (deg)

Rel. gain (dB)

Angle (deg)

Rel. gain (dB)

Angle (deg)

Rel. gain (dB)

0 0 0 0 0 0 0 -0 50 0 20 0 2 0 1 -15 95 -10 45 -4 8 -17 12 -35

135 -12 60 -10 30 -22 35 -43 155 -15 90 -20 90 -30 60 -50 180 -25 125 -40 100 -35 80 -60

180 -50

180 -37 180 -65

-150 -100 -50 0 50 100 150-50

-45

-40

-35

-30

-25

-20

-15

-10

-5

0

Degree [° ]

Rel

ative

Gai

n [d

B]

CS1BS0

-150 -100 -50 0 50 100 150

-70

-60

-50

-40

-30

-20

-10

0

Degree [° ]

Gai

n [d

B]

TS3TS0

Figure 7 CS antenna radiation patterns [9] Figure 8 TS antenna radiation patterns [9]

5. Results

In this section comparison for different combinations of the described system parameters is given. Investigations are presented by BER distribution maps of the examined sector, sector average BER values, and BER vs. SNR analyses for a certain point on the sector map.

5.1. The effect of applied multiple access method and antenna system

The main goal of this work is to demonstrate the feasibility of applying CDMA in LMDS networks. Furthermore, it will also be shown that according to our previous investigations [3] the effect of the applied antenna system on interference is dominant. BER conditions in the investigated sector (circled area in Figures 2 and 4) are presented by sector maps in Figure 9. A certain point on the map indicates the BER value on the downlink of a TS installed at that point. Darker tones represent poorer BER results. 4×4=16 depicted maps of Figure 9 comprise all combinations of the examined system parameters, such as four access methods: TDMA and three CMDA applying MF, PIC and MMSE detectors, and four antenna configurations. All of our simulations were considered at a constant signal to noise ratio SNR=7 dB.

TDMA (CS1-TS3) TDMA (CS1-TS0) TDMA (CS0-TS3) TDMA (CS0-TS0)

MF (CS1-TS3) MF (CS1-TS0) MF (CS0-TS3) MF (CS0-TS0)

MMSE (CS1-TS3) MMSE (CS1-TS0) MMSE (CS0-TS3) MMSE (CS0-TS0)

PIC (CS1-TS3) PIC (CS1-TS0) PIC (CS0-TS3) PIC (CS0-TS0)

Color scale for BER

Figure 9 Downlink BER conditions of the examined LMDS sector for TDMA and CDMA approaches with MF, MMSE and PIC detectors, applying four combinations of the

CS (CS1, CS0) and TS (TS3, TS0) antennas (SNR=7 dB)

0.0001

0.001

0.01

(CS1-TS3) (CS1-TS0) (CS0-TS3) (CS0-TS0)

aver

age

BER

TDMA

MF

MMSE

PIC Figure 10 Comparison of sector average BER performances on downlink of the examined

LMDS sector for TDMA and CDMA approaches with MF, MMSE PIC and detectors, applying different antenna systems (SNR=7 dB)

In Figure 9 maps are ordered vertically so that antenna performance shows an improving tendency: from minimum ETSI requirements (CS1-TS3) to an antenna system with highest main lobe gain and side lobe suppression (CS0-TS0). Poorer antenna characteristics result in darker maps, meaning poorer BER performance caused by higher interference level. The side lobe suppression of the CS antennas affects CDMA performance significantly. As shown in Section 3, (also see Figures 4 and 5), while downlink interference in 4-frequency sectored TDMA-based LMDS network is received from second nearest neighbour sectors, in single-frequency CMDA approach all sectors transmitting to the same direction cause intercell interference. Therefore CS antenna radiation pattern has greater effect on interference in CDMA case than in TDMA case. The first two columns of Figure 9 representing the ETSI-recommended CS sectored antenna with lower side lobe suppression indicate heavy interference along the sector borders and diagonals. Although applying MUD can significantly reduce the effect of interference, and make the sector area nearly homogeneous in terms of BER performance, CDMA cannot outperform TDMA according to sector average BER values given in Figure 10. Sector average BER performances indicate that TDMA has nearly the same average BER performance for all investigated antenna systems, since it is less sensible to CS and TS antenna characteristics than CDMA. Different behaviour of the examined two multiuser algorithms can be observed in different TS positions. As explained in Figure 5, interference zones near the sector borders have one dominant interferer, namely the neighbour cell, while in the diagonal, interference is composed of several sector’s interfering signals. Because the PIC detector performs better when only a few strong interferers are present, it shows higher interference reduction capability in the borders than MMSE (e.g. see maps MMSE (CS1-TS0) and PIC (CS1-TS0) in Fig. 9). Nevertheless, MMSE shows better performance in case of more but lower interferers, e.g. in the sector diagonal.

Further illustration of the applied antenna system is shown by BER vs. SNR analyses in Figure 11. Simulations for CDMA approach with the PIC detector take the five most disturbing sector into account. A selected TS location at the right bottom corner of the sector was investigated. The difference between the best (CS0-TS0) and the worst (CS1-TS3) antenna systems is revealed by approximately 1 dB gain at 10-3 BER value.

0 2 4 6 8 10

10-5

10-4

10-3

10-2

10-1

SNR [dB]

BE

R

5-PIC (CS1-TS3)5-PIC (CS1-TS0)5-PIC (CS0-TS3)5-PIC (CS0-TS0)BPSK

Figure 11 Comparison of different antenna systems: BER vs. SNR at a fixed TS location for CDMA-based LMDS system with PIC detector taking 5 most interferer sectors into account

5.2. The effect of PIC detectors with different complexity

As discussed in the system model of Section 4, the applied simplified PIC detector takes into account only the five most interfering sectors. The effect of utilizing more and less interferers is demonstrated on the BER maps of Figure 12, proving the feasibility of this simplification. The critical number of the considered interfering sectors is 4 corresponding to Figure 5, above which no significant performance gain can be achieved. Please see the nearly same BER maps of Figure 12 in case of utilizing 4, 5 and 10 interfering sectors’ signals (4-PIC, 5-PIC, 10-PIC) compared to the higher BER values of 3-PIC. The corresponding sector average BER values are also nearly the same for 4-PIC, 5-PIC and 10-PIC (see Figure 13).

Additionally the interference suppression capability of CDMA with PIC detector is investigated by increasing the processing gain from 64 to 128. BER vs. SNR curves in Figure 14 include simulation results for one of the antenna systems (CS0-TS3) in the right bottom location of the sector. The effect of higher processing gain is examined together with the complexity of PIC receiver, i.e. exploiting the knowledge of 3, 4 and 10 interfering sectors (denoted as 3-PIC, 4-PIC and 10-PIC). The nearly equivalent performance of PICs with at least 4 sectors’ utilization can be observed. The interference suppression effect of higher processing gain means nearly 2 dB gain with 3-PIC, while with 4-PIC and 10-PIC only a 0.5 dB gain could be achieved by doubling the code length. Therefore it is worth applying a bit more sophisticated receiver (i.e. 4-PIC instead of 3-PIC) than increasing processing gain.

0 2 4 6 8 10

10-5

10-4

10-3

10-2

10-1

SNR [dB]

BE

R

3-PIC (CS0-TS3, 64)4-PIC (CS0-TS3, 64)10-PIC (CS0-TS3, 64)3-PIC (CS0-TS3, 128)4-PIC (CS0-TS3, 128)10-PIC (CS0-TS3, 128)BPSK

Figure 14 BER vs. SNR analysis at a fixed TS location for CDMA-based LMDS system with PIC detector utilizing 3, 4 and 10 interfering sectors, applying different antenna systems

3-PIC (CS0-TS3) 4-PIC (CS0-TS3) 5-PIC (CS0-TS3) 10-PIC (CS0-TS3)

Figure 12 Downlink BER maps for CDMA-based LMDS network applying PIC-type multiuser detectors utilizing the knowledge of the 3, 4, 5 and 10 most

interferer sectors.

0.0001

0.001

0.01

3-PIC 4-PIC 5-PIC 10-PIC

aver

age

BER

Color scale for BER

Figure 13 Sector average BER performances of downlink PIC detecrors in LMDS systems utilizing 3, 4, 5 and 10 interferer sectors, applying antenna system (CS0-TS3)

6. Conclusions

CDMA-based LMDS networks were investigated in terms of downlink interference. Code sectoring similar to the code system used e.g. in UMTS was examined. Multiuser detection strategies were introduced, examining PIC-type detectors with different complexities. Comparisons are made with a four-frequency sectored TDMA approach, and antenna systems with various side lobe suppressions were considered. CDMA with PIC-type multiuser detector proved to compete with TDMA, even though bit-asynchronous transmission. The comparison of various antenna systems revealed that the applied antenna system affect CDMA interference more seriously than that of TDMA, moreover the side lobe suppression of the sectored CS antennas is dominant. Different interference zones were determined in a sector of the CDMA-based LMDS network, leading to the necessity of using different PIC receiver structures at TSs located in different zones. It has been revealed that in case of the investigated network with 3×3 CSs, most of the interference is received from the four dominant interferer sectors. It has been shown that a PIC detector exploiting the knowledge of only the four most interfering sectors can lead to accessible system performance. Applying a PIC receiver with the proposed complexity proved to suppress interference more effectively than increasing processing gain by doubling the code length of the applied CDMA code system.

In summary, our investigations lead to the experience that system performance is mostly determined by the CS antenna and receiver complexity at the TS. Increasing system performance in CDMA-based LMDS systems can be therefore achieved by applying a bit more sophisticated receiver than increasing processing gain. Because the interference situation in single-frequency CDMA-based LMDS system is mostly determined by the CS antenna, applying cheaper TS antennas together with precisely dimensioned sectored CS antennas; result in an economical future LMDS system.

7. References

[1] A. Nordbotten, “LMDS systems and their application”, IEEE Communications Magazine, June 2000, pp. 150-154

[2] Cs. Novák et al., “Uplink Interference Analysis of LMDS Networks Applying CDMA with Interference Cancellation” submitted to IEEE-ISSSTA 2002, Prague

[3] Cs. Novák et al., “Investigation of Interference in Broadband Wireless Point-to-Multipoint Networks” submitted to VDE Kongress 2002 Networlds, Dresden

[4] B. Friedrichs, K. Fazel, “Efficient multiple access schemes for wireless broadband point-to-multipoint access networks”, Proc. of ECRR 2000, Dresden, Germany, Sept. 12-15, 2000, pp .99-105

[5] H. Sari, “A multimode CDMA with reduced intercell interference for broadband wireless networks”, IEEE Selected Areas in Comm., Vol.19, No.7, July 2001, pp. 1316-1323

[6] Mahonen, T. Saarinen and Z. Shelby, “Wireless Internet over LMDS: architecture and experimental implementation”, IEEE Communications Magazine, May 2001, pp. 126-132

[7] 3GPP Technical Specification 25.213, Spreading and Modulation (FDD)

[8] S. Verdú, “Multiuser detection”, Cambridge University Press, 1998.

[9] Antennas for point-to-multipoint fixed radio systems in the 11 GHz to 60GHz band; Part 2: 24 GHz to 30 GHz, ETSI EN 301 215-2 v1.3.1. (2002-06)