Statistical Analysis of MIMO Beamforming With Co-Channel Unequal-Power MIMO Interferers Under Path-Loss and Rayleigh Fading
Yongzhao Li, Member, IEEE, Lu Zhang, Member, IEEE, Leonard J. Cimini, Jr., Fellow, IEEE, and Hailin Zhang, Member, IEEE

Abstract—Multiple–input multiple-output (MIMO) beamforming (MBF) can greatly increase the signal gain and reduce the effects of multipath fading for cell-edge users. The objective of this work is to investigate the impact of co-channel unequal-power MIMO interferers on an MBF desired receiver under realistic propagation conditions, including path-loss and Rayleigh fading. Two major contributions are made in this work: i) for the co-channel interference term in the signal-to-interference ratio (SIR) expressions, a closed-form expression of the probability density function (PDF) is derived; ii) for the desired signal term, a simple approximation is proposed. Moreover, closed-form PDF expressions for the SIRs are obtained for some special cases. Simulation results verify the validity of the theoretical analyses. Index Terms—Co-channel interferer, MIMO beamforming, path-loss, probability density function, signal-to-interference ratio.



O-CHANNEL interference (CCI) has a tremendous impact on cellular-like systems, especially when the goal is to use the same frequency in every cell. Multiple-input multiple-output (MIMO) based spatial diversity techniques are extensively adopted for cell-edge (or sector-edge) users to combat fading and CCI. However, when multiple-antenna techniques are incorporated, the interference scenario becomes more complicated [1]. Hence, the research related to the impact of CCI on spatial diversity has gained much attention in recent years. Commonly used spatial diversity techniques can be classified into two categories, receive diversity and transmit diversity. Receive diversity techniques include maximal ratio combining

Manuscript received March 03, 2011; revised April 26, 2011; accepted April 27, 2011. Date of publication May 19, 2011; date of current version July 13, 2011. The associate editor coordinating the review of this manuscript and approving it for publication was Prof. Xiang-Gen Xia. The authors acknowledge the support from NSFC (61072069), RCUK for the UK-China Science Bridges Project: R&D on (B)4 G Wireless Mobile Communications, the Fundamental Research Funds for the Central Universities (72101855), the Important National Science & Technology Specific Projects (2011ZX03003-001-04), the Special Grade of China Postdoctoral Science Foundation funded project (200902588), the State Key Laboratory of Integrated Services Networks (ISN090105) and the 111 project (No. B08038). Y. Li and H. Zhang are with the State Key Laboratory of Integrated Services Networks, Xidian University, Xi’an, Shaanxi, 710071, China (e-mail: yzhli@xidian.edu.cn; hlzhang@xidian.edu.cn). L. Zhang is with the Alcatel-Lucent Shanghai Bell Co., Ltd., Shanghai, 201206, China (e-mail: Lu.c.Zhang@alcatel-sbell.com.cn). L. Cimini is with the Department of Electrical and Computer Engineering, University of Delaware, Newark, DE 19716 USA (e-mail: cimini@ece.udel. edu). Digital Object Identifier 10.1109/TSP.2011.2155654

(MRC) and optimal combining (OC) [2], which are implemented for single-input multiple-output (SIMO) systems. Transmit diversity consists of space-time block coding (STBC) [3], [4], maximal ratio transmission (MRT) and MIMO beamforming (MBF) [5]. STBC is an open-loop MIMO scheme; MBF or MRT, on the other hand, utilizes the channel state information (CSI) available at the transmitter to achieve the optimum transmission. It should be noted that MRT can be only applied in multiple-input single output (MISO) environments; so, it is often considered as a special case of MBF. Analyzing the statistical distribution of the postprocessing signal-to-interference ratio (SIR) or signal-to-interference-plus-noise ratio (SINR) is very useful in investigating the performance of cellular-like MIMO systems. For SIMO, the SIR/SINR distribution of MRC with CCI is discussed in [6]–[11]; while the performance of OC with CCI is investigated in [9]–[11]. For a noise-limited system, MRC is well known to be the optimal combining technique in the sense of maximizing the postprocessing signal-to-noise ratio. In an interference-limited environment, OC performs optimally in terms of maximizing the SINR, but is much more complicated than MRC because it requires information about the CCI that might not be available at the receiver [10]. Thus, in practical systems, MRC is usually adopted instead of OC even in an interference-limited environment. The performance analysis of transmit diversity techniques is much more complicated than receive diversity systems. The distribution of the postprocessing SINR for STBC with CCI was first derived in [12] and compared with other MIMO schemes, such as MRC, cyclic delay diversity, and MRT. However, in [12], it is assumed that all the base stations (BS) adopt the same MIMO mode; further, the statistical distribution is derived when equal transmit power is assumed for all interfering BSs and when the wireless channels are modeled as i.i.d. complex Gaussian random variables (RVs) with zero mean and unit variance (i.e., pure Rayleigh fading). The impact of interfering single-antenna and STBC transmissions on a desired STBC receiver is studied in [13]; however, only a simulation-based approach is used. In [14], with the channels modeled as pure Rayleigh fading, the closed-form probability density functions (PDF) of the postprocessing SIR was derived for STBC transmission when all the interfering BSs use either the same or a different MIMO mode. For the cases where the desired BS employs MBF or MRT, which are the focus in this paper, some work has been done in

1053-587X/$26.00 © 2011 IEEE

(of the desired user) from one transmit antenna (of the th BS). see representative work in [15]–[21]) use the terminology MIMO-MRC to represent MBF and assume that each co-channel interfering signal is from a single transmit antenna. ) is available at the desired BS. respectively. complex Gaussian RVs with zero mean and and ii) each antenna at each BS uses unit power. if each interfering BS adopts STBC.: STATISTICAL ANALYSIS OF MBF WITH CO-CHANNEL UNEQUAL-POWER MIMO INTERFERERS UNDER PATH-LOSS AND RAYLEIGH FADING 3739 determining or estimating the statistical distribution of the postprocessing SIR/SINR. the effect of noise will be neglected in the subsequent analysis. Then. In particular. MBF. -Rx) antenna configuration is adopted in the desired link for unequal-power interferers.LI et al. II. of The channel coefficients include the effects of pathloss and quasi-static flat Rayleigh fading.g. The work in [16]–[18] derived the CDF (equivalently. In [15]. the resulting statistical distribution of the postprocessing SIR will be distinct from the case in which each interfering BS adopts MBF/MRT/SIMO. Note that. closed-form PDF expressions for the interference terms in the postprocessing SIRs are provided. It is assumed that the system is located in a rich scattering environment and the antennas are sufficiently can be modeled as indepenseparated. or closed-loop spatial multiplexing (CLSM). we include the transmit powers in the channel power gains. for the cases where each interfering BS performs single-antenna transmission and each interfering BS adopts MBF/MRT to perform multi-antenna transmission. we derive the distributions of the postprocessing SIRs (noise is neglected here since we are considering an interference-limited environment) for MBF with different co-channel MIMO modes and possibly unequal-power interferers. is perfectly known at the desired BS). which adopts MBF (assuming fined in (1). The channel between the th BS and the desired receiver is denoted by (1) represents the inwhere stantaneous channel coefficient between the th receiving antenna at the desired user and the th transmitting antenna at the th BS.e. SYSTEM MODEL AND NOTATIONS A typical interference-limited system using MBF in the deco-channel interfering BSs is considered in sired link with this paper. OLSM. MRT. When MBF is in used in the desired link. respectively. Each transmitter and the desired receiver are assumed to be equipped with and antennas. denote complex Notations: The superscripts . Without loss of generality. Moreover. open-loop spatial multiplexing (OLSM). For MBF transmission with CCI. i. all the existing work (e. This can be considered as a “worst-case” scenario.e.. and conjugate. respectively. the transmit power of each antenna at the th BS is denoted as . -Rx) or ( -Tx. Since path-loss and transmit power may be distinct for different may not be equal to . no work has been done previously. complex Gaussian RVs with zero mean and variance . Based on the analyses in this section. the when SIMO (i. so. as deIn the desired link. The system model and notations are described in Section II. POSTPROCESSING SIR In this section. for any given . The index indicates the desired BS. outage probability) and/or PDF of the postprocessing SIR/SINR when an ( -Tx. Here. and denotes the gamma function. we derive and analyze expressions for the postprocessing SIR considering the special case where all the interfering BSs adopt the same MIMO mode. but. for the case where the desired link adopts MBF. and assuming all the single-antenna interferers use equal transmit power. we consider an equivalent system in which i) the channels are independent. BSs. the channels dent. We assume that full CSI of the desired link (i. . We also assume that the network is de-centralized. no knowledge of the interferers can be utilized by the desired receiver and the transmissions of the different BSs are independent. ) is used for the th interfering BS. The validity of the theoretical analyses is verified by simulation results. channel matrix between the interfering BS and the desired revector. six widely applied MIMO modes: STBC. the coefficients mean channel gain . CLSM. The postprocessing SIR expressions are derived in Section III. transpose and conjugate transpose. is the identity matrix. denotes the norm of a complex valued vector . cularly symmetric complex Gaussian distribution with mean and variance . the work in [19]–[21] analyzed the distribution of the postprocessing SINR with imperfect channel estimation. taking into account the effects of both path-loss and Rayleigh fading. so. the six typical MIMO modes can be effectively classified into two categories of interfering MIMO modes. denotes the circolumn of matrix . variance is the average received power at one receive antenna Here. denotes statistical expectation. Using the same model for the desired and interfering links as in [16]–[18]. However. we present the PDF of the SIR in Section IV. the same statistical distribution of the postprocessing SIR results. is precoded . the subsequent derivation for the SIR distribution can be greatly simplified.. Since we consider an interference-limited environment. in order to provide a comprehensive investigation of the impact of each possible co-channel MIMO mode on an MBF receiver. . i.. closed-form expressions for the PDF and cumulative distribution function (CDF) of the postprocessing SINR are obtained when a (2-Tx. we use the first column ceiver becomes an to denote it. Then. denotes the vector space of all comand denote the th row and the th plex matrices. For simplicity.. . for these cases. It should be noted that both the effects of path-loss and Rayleigh fading are included in . 2-Rx) antenna configuration is adopted in the desired link. since the distances from the antennas of a co-located array to the antennas of another co-located array can be considered have the same equal.e. Under these assumptions.e. III. and SIMO are considered for the co-channel interfering BSs. with the help of this classification. In this paper. respectively. it will be shown in this paper that.. Simulation results and conclusions are provided in Sections V and VI.

with the same definition as (5). the baseband received vector as (2) and are the precoding matrix (or vector) and the where transmitted symbol(s). the preis constructed from the coding unitary vector . which corresponds to the maxunitary singular vector. the intra-cell link precoding vectors for MBF . 4) Interfering Links Use SIMO: Note that. Using (2) and (3). VOL. Desired User Has One Receive Antenna When the desired user has only one receive antenna (i. Using the notation in . can be modeled ceiver. the average received power of the desired signal can be computed as (4) Next. which consists of the interfering links. MRT [5] is adopted at the reduces to a vector and desired BS. 2) Interfering Links Use CLSM: When all the interfering BSs adopt CLSM.e. independent symbols of the form (5) is sent simultaneously from the th interfering BS.e. Assuming that the CSI of the th intra-cell link is available at the th interfering BS. is also distributed as . the precoding vector for the desired link is defined as (3) . At the desired re. we analyze the average interfering power and obtain the SIR expressions for different MIMO modes used in the interfering links. which also follows have the same distribution. the average interference power corrupting can be of computed as (6) is defined by (1) with and . . A. which means (9) and (10) and MRT have the same norm have the same functionality when they act as the precoding vectors. By comparing with in bution . . and depend on the specific The sizes and natures of MIMO mode used in the th interfering BS. 8. the average interfering power can be evaluated as (8) is an SVD-generated unitary matrix and indepenBecause .3740 IEEE TRANSACTIONS ON SIGNAL PROCESSING. AUGUST 2011 by a vector before transmission. the transmit antennas are used to send the same symbol during one symbol period. By integrating the precoding where block into the th CCI channel. as a special case of MBF. In the case of MBF. In is used by the interfering BSs. no precoding is used in . In this scenario. 59. we can establish an equivalent model for the channel vector between the th interfering BS and the desired receiver which has the form (7) Therefore. the average power of (9) and the assumption of the interference on can be written as (11) represents the equivalent channel bewhere tween the th interfering BS and the desired receiver. 3) Interfering Links Use MBF or MRT: When MBF or MRT reduces to a scalar. for the th interfering BS. 1) Interfering Links Use STBC or OLSM: When all the interfering BSs adopt STBC or OLSM. the precoding unitary vector is (10) where is the th intra-cell channel vector. although this assumpFor simplicity. reduce to the scalars and . is precoded before being sent simultaneously from the th interfering BS. to satisfy the same transmit power constraint as for the other MIMO modes. when SIMO and and MRT are used for the interfering BSs and desired BS. which will be explained in what follows.. we find that (8) and (6) (6). Hence. respectively. satisfy the same transmit power constraint as for other MIMO modes. other words. Note that. the precoding unitary matrix is obtained through a singular value decomposition (SVD) of the th intra-cell channel matrix. we define (12) . i. respectively. we assume tion can be easily relaxed. ). To imum singular value of the th intra-cell channel matrix. Moreover.. By assuming and exploiting the independence . MBF and MRT have the same performance when they are used in the interfering BSs. Since (inter-cell link precoding vector) is unitary and independent of (inter-cell link channel vector). . NO. are still independent RVs with distrident of . . we define (9) In the case of MRT.

the corresponding left cally. using the independence of and .d.: STATISTICAL ANALYSIS OF MBF WITH CO-CHANNEL UNEQUAL-POWER MIMO INTERFERERS UNDER PATH-LOSS AND RAYLEIGH FADING 3741 as a normalized power factor. the following postprocessing SIR expression can be obtained: (22) where ( for Category I or for Category II) is an index representing the category of the MIMO mode of the interference. and CLSM.LI et al. Next. it is easy to show that the nuSince in (16) follows a central chi-square merator distribution with PDF (17) Similarly. . At the desired transmitter. So. the six co-channel MIMO modes can also be classified into two categories when MBF is used in the desired link. This implies other [22]. the same category has the same effect on the desired received signal. Without loss of generality. SIR Expression for General and Here. once the distribution of the denominator is determined. complex Since Gaussian entries. we divide the six co-channel MIMO modes into the following two categories when MRT is used in the desired link: • Category I: STBC. is a random matrix with i. and (20) Then.. (for ). Additioncorrespond to ally. it should be noted that the denominators in (16) and (22) have the same statistical distribution. and are independent and follow and . the numerator and the denominator in (16) are obviously independent. Asis perfectly known sume that the channel matrix at the desired BS. and the corresponding SIRs when MBF is adopted for the desired link.d. for the largest singular value and and right singular vectors are denoted by . As a result. and Gaussian distributed and follows . is precoded by the beamforming . the postprocessing received signal at the desired user is (14) for Category I. non-negative real singular values onal matrix with . where (depends on . we can show that is an exponential distributed RV with PDF (18) Moreover. the received signal is multiplied by Hence.i. respectively. ) is i. the numerator and the denominator in (22) and .i. OLSM. respectively. it is isotropically random (i. Note that. follows an exponential distribution with the same PDF in (18). • Category II: MBF. thus. We conclude that the SIRs of the same category have the same statistical distribution. the desired BS implements an SVD on (19) and are the left and right uniwhere is a rectangular diagtary matrices. complex . we rewrite (14) and (15) as (16) Similar to Section III-A (MRT in the desired link). we can obtain the SIR of the received signal at the desired receiver as B. Hence. spherically sym. the PDF of the SIR can be obtained by using the independence feature. we consider general and . and SIMO. by computing and comparing the average interfering power for different interfering MIMO modes. we obtain the average desired signal power in the postprocessing received signal as (21) (15) for Category II. Obviously. we conclude that the numerator and the denominator are mutually independent. and are independent of each metric). In other words. Based on the above analysis. the PDF of the denominator in (22) can also be used to determine the SIR distribution for MRT in the desired link.e. Hence. Specifi. For simplicity. we derive the PDFs of the denominators. Similar to other MIMO modes. To obtain the precoding unitary vector for . Furthermore. . MRT. (13) and (11) Since have the same distribution. MBF. we assume . the average power of the inwe assume is terference on (13) follows the distribution . respectively. respectively. At the desired vector to recover . Therefore. receiver. and then transmitted.

a usable expression for (24) of closed-form expressions for cannot be obtained.g. both STBC and MBF are equivalent to a receive receive antennas. we propose an approximate dis. -Rx) antenna order configuration. for an ( -Tx. we model the largest eigenas a chi-square distributed value of the Wishart matrix ) multiplied by the coefficient . able approximations to the distribution of Using the same classification of the interfering MIMO modes. . without loss of generality. MBF cannot utilize the full channel energy to transmit the signal. By implementing denotes the trace of a matrix. Hence. An alternative approach is to use some empirical formulas. can be evaluated using where symbolic computation software [23]. we need to make some reason(i. In the case of STBC. it is difficult. Also note that depends (28) is symmetric in and . 8. we get the following eigenvalue : decomposition of (23) . Gaussian RVs and . the largest eigenvalue and is . This means that all of the channel energy is and (28). Here. MBF can also achieve the maximum spatial diversity gain of [24]–[26]. Therefore. 59. -Rx) antenna configuration (28) When (MRT in the desired link). a reasonable approximate statistical model for must take the loss of channel power into account. we can see that the distributions of the denominators (co-channel interference terms) in the SIRs for MBF and STBC are the same. For an ( -Tx. Due to the unavailability . -Rx) MIMO channel can be converted independent channels. however. Although STBC and MBF can achieve the same diversity gain. .. the ceive antennas using MRC achieves diversity order distribution for the desired signal term of MRC (the sum of square moduli of independent channel coefficients) is chisquare.d. and the mean of and . The numerators (desired signal terms) are different. exponential RVs with PDF By comparing (22) with (25). we use the fact where are i. Rather. so. the channel power provided by MIMO cannot be fully utilized by MBF. The validity of this approximation will be justidegree fied by Monte Carlo simulations in Section V. the numerfollows a chi-square distribution with degree . ator that the desired rewhich corresponds to the diversity gain ceived signal can obtain. it is reasonable to approximate the distribution (desired signal term for MBF) as a chi-square RV with of . and (23).. we can use the following expression to evaluate the mean of the largest eigenvalue of the Wishart matrix for an ( -Tx. This is an impordiversity system with retant similarity. AUGUST 2011 IV. Different from other into closed-loop MIMO schemes (e. Let is [23] Clearly. with the aid of the introduction of . Hence. (Note that and have identical eigenvalues. part of the channel power is inevitably consumed by channels with gains corresponding to the smaller eigenvalues. -Rx) MIMO channel. and no energy is lost. the PDF of SVD (transmit precoding) at the transmitter and shaping at the receiver. ). (19). Distribution of The distribution of can be obtained by evaluating the distribution of the largest eigenvalue of the Wishart matrix .3742 IEEE TRANSACTIONS ON SIGNAL PROCESSING. and. Using the fact that tribution for the largest eigenvalue MBF can achieve the maximum diversity gain . Using (1). AlThe numerator in (26) requires the expectation of though (24) can be used. Besides this similarity. Therefore. on Based on the above analysis. The reRV (with degree sulting PDF is (29) where values of the Wishart matrix the denominator in (26) is are the nonzero eigen. VOL. where are independent and .) Using (19). we also need to consider the differences between these two MIMO schemes.e. NO. Because the power of every decoupled channel is proportional to its corresponding eigenvalue in the Wishart matrix. Since a receive diversity system with . concentrated in one channel. specifically. complex that . we define the percentage of energy achieved by the channel with gain as (26) (24) . an ( -Tx. (27). MBF can only uti) lize the channel with the largest gain (corresponding to to transmit the signals. the coefficient of . from (26). as shown in [27]. Hence. the postprocessing SIR when STBC is used for the desired link category interferers can be written as [14] with (25) (27) follows a chi-square distribution with degree . CLSM).i. we use as the example here. SIR DISTRIBUTION A.

Because the identity matrix is a semi-positive definite Hermitian matrix. and is a zero-mean complex Gaussian random vector. and Moreover. That is to say. is specified as (38) (32) (33) . we see that the corresponding denominators have the same distribution. Since the value of is definitely non-negtransform of for . the notations in (16) are used in the following derivations. For central quadratic forms in complex Gaussian RVs. which is the inverse Laplace . where plex Gaussian RVs with variances means “rounding to the next largest integer”. Since decomposition of is equal to the eigenvalue decomposition of . Denote its PDF is as independent zero-mean com. Without loss of generwhere as the total number of distinct ality.LI et al. (37) where denotes the number of ways of selecting elements from elements.e.. exponential distribution) and . Further. we can rewrite (31) . Jordan’s lemma. B.e. the moment generating function (MGF) of can be expressed as [28] (35) Denote as the PDF of . which is a semi-positive definite Denote . shown at the bottom of the page. as Then. Let and denote the identity matrix as . as . denote which can the covariance matrix of as be written as (32). . ) of We rewrite the denominator (excluding the coefficient (16) as (30) in which follows a chi-square distribution with degree 2 (i. Distribution of Denominator in SIR By comparing (16) and (22). By utilizing the Laplace inative. for and . we derive version theorem. there exists for which a real symmetric matrix . the eigenvalue Hermitian matrix. although (29) is derived for which corresponds to MBF in the desired link. and the residue theorem [29]. is a central quadratic form in complex Gaussian RVs [28]. Then. denote as the multiplicity of positive eigenvalues and .. Without loss of generality.: STATISTICAL ANALYSIS OF MBF WITH CO-CHANNEL UNEQUAL-POWER MIMO INTERFERERS UNDER PATH-LOSS AND RAYLEIGH FADING 3743 Note that. the eigenvalues of are where (34) . Because is a semi-positive definite real symmetric matrix. . We can easily obtain (33) at the bottom of the page. it agrees with (17) for (i. ) which corresponds to MRT in the desired link. is the order derivative of . we get (36) is the residue of where we calculate at the pole as follows: .

is with . we can easily obtain . . and • Special Case II: (i) The number of interfering BSs.3744 IEEE TRANSACTIONS ON SIGNAL PROCESSING. VOL. is given as [30] Denote the CDF of the postprocessing SIR. . the closed-form expression for the PDF of is given as (42) (47) C. . Then. and for (50) . and . by utilizing (43) Due to the complicated mathematical form for (42). the PDF of the postprocessing SIR. we get the fact that . . . In practice. using (29) and (43). 59. we denote as . clearly. (48) becomes (44) where with . we have of (39) where and is equal to 0 when . AUGUST 2011 Using the logarithmic derivative [28] to evaluate the derivatives . . is further specified as (46) (41) 2) Interfering Links Use Category II MIMO Modes: For the general scenarios where and for with . Then. . • Special Case I: When the number of transmitting antennas for the interfering BSs. when . we get the MGF of as (49) Furthermore. 8. using (38) and (40)–(42). using (38) and (40)–(42). . using (47). Then . pression of (43) can be obtained. . SIR Distributions 1) Interfering Links Use Category I MIMO Modes: For the general scenarios where and for with . obviously. and . we get (45) and. . Then. we get By combining (37)–(41). Then. for complex Gaussian RVs with zero mean and variance . and with . a closed-form expression for (43) is not easy to obtain. is given as and (40) In (40).i. numerical computations can be used to evaluate it.d. then using (38) and (40)–(42). From (35). NO. But in some can be simplified and/or a closed-form exspecific cases. is equal to 2. Because . it is given as is equal to 1 when . as (48) is also a central quadratic form in i. . equal to 1 or (ii) and with multiplicity .

only SIMO mode can be used.. is equal with . A major observation from Figs. Although there is some mismatch between (29) and the Monte Carlo results within a small SIR range. the deviation is very small. we find that the analytical CDFs match the corresponding simulation CDFs very well. in Figs. only one-pair of PDF curves appear in Fig. means that 1. antenna (i. Validity of the PDF of the interference term in the SIR. 2. N =2 (54) V. we get (53) and. to 1 or ii) and with the multiplicity of . using (29) and (43). 4–6 is that MIMO modes from different categories have different impact . (42) and (29). in the legends. we verify the validity of the PDF expressions. The notation interfering BSs. (42). First.: STATISTICAL ANALYSIS OF MBF WITH CO-CHANNEL UNEQUAL-POWER MIMO INTERFERERS UNDER PATH-LOSS AND RAYLEIGH FADING 3745 By replacing (49) with cessing SIR is . STBC ceiver has and MBF/MRT are used to represent Category I and Category is set to II interferers. 1 and 2. (42). 4–6. . Hence. Moreover. we get the PDF of the SIR as (52) • Special Case: i) the number of interfering BSs. Validity of the PDF of the interference term in the SIR. for this case. K=6 Fig. respectively. when an interfering BS has one ). This distinction results in different impact of the two categories on the desired user. capitalized “I” and “II” are used to denote “Category I” and “Category II” interfering MIMO modes. with .e. respectively. 1–3. Besides the cases of . For the desired link. Note that. the th number within the there are braces is the value of for the th interfering BS. In the Monte Carlo simulations. 1 and 2 is the remarkable distinction between the PDFs of Category I and II interfering MIMO modes. 1. respectively. is given as Fig. the CDF of the postpro- (51) By differentiating (51). Validity of the PDF of the desired signal term in the SIR. using (38) and (40)–(42). 4. Then. the validity of the theoretical analyses is verified by comparing the analytically obtained PDFs/CDFs with Monte Carlo simulations. 3. 2. SIMULATION RESULTS In this section. the case of is Fig. Another important observation from Figs.LI et al. we see that the analytical results of (42) match the corresponding Monte Carlo simulation results perfectly. clearly. verified in Fig. (29). From Figs. All the BSs (including the desired BS) are assumed to be equipped with antennas and the desired reantennas. with . “(A)” and “(S)” represents the “Analytical” and “Monte-Carlo simulation” results. From Figs.

the distinction between the variances becomes larger with an increase in . 5. when the desired link degrades. the parameters are and . Both theoretical analysis and simulation results show that the two categories have different impact on the desired MBF receiver. 11. In Fig. Nov. Hence. Driessen. when are fixed (see Fig.” IEEE Commun. at the BSs. a change in the number of receive antennas does not have a significant effect on the gap between the when two categories. Cambridge. Impact of the number of transmit antennas. the gap between the performance of the two categories decreases.” IEEE J. 6. Areas Commun. Mar. at the desired receiver. 1451–1458. “Space-time codes for high data rate wireless communication: Performance criterion and code construction. The validity of the PDF expressions is verified by Monte Carlo simulation results. [2] A. “A simple transmit diversity technique for wireless communications. on the desired user. Moreover. In (22). Greenstein. VI. Although affects both the desired received signal and interfering received signal. Sel. the performance of As expected. . Hence. therefore.” IEEE Trans. 1998. 4. vol. From (22). In Fig. the total equivalent interfering power does not change with . an increase in is fixed will narrow the gap between the variances. F. the performance gap between the two categories is mainly determined by the number of transmit antennas. The performance gap between the two categories arises from the and difference in the variances. AUGUST 2011 Fig. the parameters of the inter) and are fixed. and A.. no. “Simulation results for an interference-limited multiple-input multiple-output cellular system. Moreover. Therefore. the statistical distributions of the postprocessing SIRs for MBF with two categories of unequal-power MIMO interferers are derived. the performance gap between the two categories increases. Alamouti. Oct. 4. From the SIR expressions.K. Calderbank. the Category II MIMO schemes have less impact on the desired link. [3] S. the CDF curves for the two categories under the same condition have a crossover at some SIR . VOL.e. NO. is shown in Fig.. Tarokh. value. In Fig. . 2. and L. vol. The impact of the number of transmit antennas at the BSs. 8. 1998. the mean and variance of the co-channel interference term for Category I are and . 6. we see that only affects the desired received signal. Another focus in this section is the performance gap between the two categories. Goldsmith. Seshadri. Impact of total equivalent interfering signal power. an increase in improves the performance of the desired MBF receiver. Fig. no. no. the antenna configuration is fixed and the impact of the total equivalent interfering signal power at the desired receiver is investigated. For the case in Fig. vol. U. 744–765. Category II has more impact on the desired link. Theory. respectively. 2005. pp. is evaluated. 6.3746 IEEE TRANSACTIONS ON SIGNAL PROCESSING. 4. including the effects of pathloss and Rayleigh fading. 8. Beyond the crossover. pp. it is easy to see that the total equivalent interfering signal power is profor both Category I and II interferers. when increases (i. The impact of the fering signals ( and number of receiving antennas. J. pp.: Cambridge Univ. hence. 334–336. REFERENCES [1] S. P. the diversity gain increases). Press. 5. M. Wireless Communications. Catreux. Lett. N. [4] V. CONCLUSION With a realistic channel model. 44. For Category II interferers. M .. Impact of the number of receive antennas. 2000. the performance of the desired link improves. 5). 4. Fig. 59. N . R. Inf. below this crossover. 16. portional to increases.

Commun. His research interests include MIMO. pp. pp. M. (S’77–M’82–SM’89–F’00) received the Ph. 1–5..” IEEE Trans. [8] J. pp. he received several ComSoc awards. 1999. J.. Complex Variables and Applications. 4. P. China. Commun. pp. Mar. 1314–1315. He received the M. Himayat. 1343–1349. N. co-channel interference. Technol. U. Akay. 2000. Cui and A. and B. first in Bell Labs and then AT&T Labs. “Outage probability of cellular radio systems using maximal ratio combining in Rayleigh fading channel with multiple interferers. 2000. Dec. Mobile Radio (PIMRC).. G. [30] M.. Netw. 2003. and N. degree from the University of Pennsylvania. and A. “Diversity analysis of single and multiple beamforming.” IEEE Commun. Himayat. 6. Ho. Dighe. 694–703. 5. [18] J.. [13] M. 52. since 2002.” in Proc. 36. Lett. DeGroot and M. in 2005. 1936–1941. From April 2002 to January 2004. 1458–1461. Generalized Multivariate Analysis. no. From November 2008 to March 2009. vol.D. [23] P. 4. Commun. no. In 2010.” IEEE Trans. where he is currently an Associate Professor at the State Key Laboratory of Integrated Services Networks. decentralized relay management strategy. 7. 2008. Sep. Geng. 51. Haimovich. degree in signal and information processing at Xidian University. pp. including the Stephen O. Churchill. he received the Best Paper Award of the IEEE ChinaCOM international conference. MA: Addison-Wesley. RCUK for the UK-China Science Bridges Project. J. 2003. [14] Y. Lett. and the Important National Science & Technology Specific Projects. 990–993. China. 2001.. Helin.” IEEE Signal Process. New York: Springer. 14. Beijing Institute of Technology. “Performance comparison of MRC and IC under transmit diversity. McLellan Meritorious Service Award. Talwar. (ICC). P. no.” Electron. T. vol. Aug.: STATISTICAL ANALYSIS OF MBF WITH CO-CHANNEL UNEQUAL-POWER MIMO INTERFERERS UNDER PATH-LOSS AND RAYLEIGH FADING 3747 [5] T. 2484–2493. Wireless Commun. He has been a Professor at the University of Delaware. and R. 55. 4/1–4/6. 801–803. 4. Newark. space-time coding. 15. 2006. Nov. [19] X. [21] W. [26] M. and space-time coding in MIMO systems. J. in 2009. 47. Oberg. he was an R&D Engineer of the Wireless Communications Department in the Beijing Research Center of HuaWei Technologies Limited Company. Commun. 10. pp. Veh. and E. pp. 1283–1286. Rui. and A. 1454–1463. Commun. Choi. 1–5. vol. the Donald W. Z. 2007.. no. multihop routing with minimal control overhead. Romero-Jerez. vol. Lo. no. Chayawan. since 2009. Yang and J. 8. Cimini. vol. As a Research Associate Professor. K. Aguilera. Sep. 49. Conradie and A.” IEEE Trans. J. and A. Nov. M. P. Dec. Andersen. 2.D. vol. J. and the Ph. Choi.” IEEE Trans. Dr. 47. His research interests include interference management for small cells and/or heterogeneous networks in LTE-Advanced systems. 54. pp. Shah and A. pp. Li.” IEEE Trans. G. Apr. Commun. 2002. Apr. [7] V. vol. P. D. Alouini. “Performance analysis of maximal ratio combining and comparison with optimum combining for mobile radio communications with cochannel interference. Jamuar. 11.. 1–5. Jul. decentralized cooperative relaying techniques in ad hoc networks (including decentralized distributed STBC technique. J. E. 245–247. he joined Xidian University. IEEE Personal. Romero-Jerez. vol. [27] J.. theoretical analyses for multi-cell co-channel interference in MIMO-based cellular networks. in 2002 and 1999. “Performance analysis of space time block coding with co-channel MIMO interferers. Jun. no. 767–771. For this pioneering work. Ayanoglu. [22] K. 10. “Average error probability of digital cellular radio systems using MRC diversity in the presence of multiple interferers.” in Proc. pp. Gupta. and M. Qin. B. Xian. Guan. E. he was a System Engineer of the Information and Communication Mobile Networks Department in Siemens Ltd. “The effects of co-channel interference on spatial diversity techniques. Evans Avant Garde Award from the IEEE Vehicular Technology Society and the 2010 Innovators Award from the NJ Inventors Hall of Fame. Newark. “On the optimality and performance of transmit and receive space diversity in MIMO channels. 2006. Oct.. Kang and M. pp. he was a full-time Research Intern of Mitsubishi Electric Research Laboratories (USA). degree from the Department of Electrical and Computer Engineering. Romero-Jerez.. 1999. among other publications-related positions. multi-source communication with minimal control overhead). Xidian University.” IEEE Trans. vol.” Int. Wireless Commun. “A comparative study on the performance of MIMO MRC systems with and without cochannel interference. New York: McGraw-Hill. Symp..” IEEE Trans. H. Wennstrom. 2007. Pena-Martin. Probability and Statistics. Goldsmith. [9] C.E. In 2008. 1990. “Analysis of SER of MIMO-MRC systems with imperfect channel estimation in the presence of non-Rayleigh CCIs. IEEE GLOBECOM. Brown and R. and T. Aalo and C. 1417–1425. and J. pp. and X. Lett. “Performance analysis of MIMO MRC systems in the presence of self-interference and co-channel interferences. he was given the 2007 James R. 2008. in 1982. degrees from the Department of Electronic Engineering. OFDM. pp. 445–448. M. 3663–3667. Since 1996. “Quadratic forms in complex normal variates: Basic results. Aalo. 2004. no. (WCNC). vol. Jr. China. Jun. He has been a Research Scientist at Alcatel-Lucent Bell Labs. efficient power allocation. 2008. Schervish. is the founding Editor-in-Chief of the IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS: Wireless Communications Series. “Performance analysis of the closed-loop transmit antenna diversity system over Rayleigh fading channels. the Special Grade of China Postdoctoral Science Foundation funded project. From January 2004 to August 2004. A.. Rydberg. no. pp. Chayawan and V. Jul. no. 8. de Carvalho. pp. S. [20] K. pp. Sel. [25] J. He has been funded by more than ten projects.. pp.” IEEE Trans. IEEE Int. [6] J. Prasad. A.D. He was elected an IEEE Fellow in 2000 for contributions to the theory and practice of high-speed wireless communications. pp. pp. 15. and. 3. “Outage probability of cellular radio systems using maximal ratio combining in the presence of multiple interferers. etc. Commun.-S. 2006. 11–15.. pp. M. vol. Zhang. [29] J. vol. 47. IEEE GLOBECOM. 2172–2178. Leonard J. 1987. May 2009. “Impact of MIMO co-channel interference. Cimini began his ComSoc activities 25 years ago in the Communication Theory Technical Committee. 2000. Conf. and the Recognition Award from the Wireless Communications Technical Committee. China. [24] E.. “Array gain and capacity for known random channels with multiple element arrays at both ends. IEEE Wireless Commun. Ahn. Goldsmith. Nov. T. vol. 2007. Areas Commun. 51. L.” in Proc. Jin.” in Proc.LI et al. R. 73–84.” IEEE J. Rice Prize. Nov. 2007. Lett. 1121–1124. Aug. [11] A. Goldsmith. pp. he had been working at the University of Delaware from 2007 to 2008. including the National Natural Science Foundation of China. 25–29. He is currently VP—Publications. 18. S. vol. 4. [16] L. 2001. He worked at AT&T. R.E. Fang and Y.” in IEE Seminar on MIMO: Communications Systems from Concept to Implementations (Ref. beamforming and cooperative MIMO communications. Cimini. 3rd ed. Martin. no. Jul. Tech. Jul. respectively. for 20 years. University of Delaware. “Outage probability of MRC with arbitrary power cochannel interferers in Nakagami fading. Conf.” in Proc. no. 2001. IEEE Int. no./175). [17] J. Rahman. [28] W. vol. Rao. Sheikh. “Maximal ratio transmission. no. Sengul. and S. 4. Veh. Reading. 5. 2010. no. Lu Zhang (S’06–M’09) was born in China. vol. Pena-Martin. pp. 860–864. 3–7. pp. A.” IEEE Trans. “Performance of MIMO MRC systems with co-channel interference. 6. Yongzhao Li (M’10) received the Ph.. Mallik. A. . “Statistical performance analysis of optimum combining with co-channel interferers and flat Rayleigh fading. [12] W. [10] Y. Commun.” IEEE Signal Process. “Outage performance of MIMO MRC systems with unequal-power co-channel interference. Indoor. 8. Syst. no. Zhang. Zhou. Akyildiz and B. S. [15] M. “Analysis of transmitreceive diversity in Rayleigh fading.” Statistica. vol.” IEEE Trans. “Performance analysis of MIMO-MRC systems with channel estimation error in the presence of cochannel interferences. 11. Aguilera.

AUGUST 2011 Hailin Zhang (M’98) received the B. and the Ph. He has recently published 78 papers in core journals and conferences. China.D. . VOL.D. Xi’an. He is currently the Dean of this School. China. the Director of Key Laboratory in Wireless Communica- tions Sponsored by China Ministry of Information Technology. form Xidian University.S. respectively. 59. degrees from Northwestern Polytechnic University. a field leader in Telecommunications and Information Systems in Xidian University. His current research interests include key transmission technologies and standards on broadband wireless communications for B3G.S. 4G. he has been with Xidian University as a member of teaching and researching and is currently a senior Professor and a Ph. in 1985 and 1988. one of the state government specially compensated scientists and engineers. all in electronic information engineering. 8. adviser with School of Telecommunications Engineering at Xidian University. and M. a key member of State Key Laboratory of Integrated Services Networks.3748 IEEE TRANSACTIONS ON SIGNAL PROCESSING. Xi’an. NO. and next-generation broadband wireless access systems. Since then. and an Associate Director for National 111 Project. in 1991.

Sign up to vote on this title
UsefulNot useful