M O D U L AT I O N , C O D I N G A N D S I G N A L P R O C E S S I N G F O R W I R E L E S S C O M M U N I C AT I O N S

INTERFERENCE CANCELLATION FOR CELLULAR SYSTEMS: A CONTEMPORARY OVERVIEW

JEFFREY G. ANDREWS, THE UNIVERSITY OF TEXAS AT AUSTIN
Desired cell (K users)

ABSTRACT
Cellular networks today are interference-limited and will only become increasingly so in the future, due to the many users that need to share the spectrum to achieve high-rate multimedia communication. Despite the enormous amount of academic and industrial research in the past 20 years on interference-aware receivers and the large performance improvements promised by these multi-user techniques, todays receivers still generally treat interference as background noise. In this article we enumerate the reasons for this widespread scepticism, and discuss how current and future trends will increase the need for and viability of multi-user receivers for both the uplink, where many asynchronous users will be simultaneously detected, and the downlink, where users will be scheduled and largely orthogonalized, but the mobile handset will still need to cope with a few dominant interfering base stations. New results for interference cancelling receivers that use conventional front-ends are shown to alleviate many of the shortcomings of prior techniques, particularly for the challenging uplink. This article gives an overview of key recent research breakthroughs on interference cancellation and highlights system-level considerations for future multi-user receivers.

Despite the enormous amount of research in the past 20 years on interference-aware receivers and the large performance improvements promised by these multiuser techniques, todays receivers still generally treat interference as background noise.
Unless specifically discussing information theoretic results, capacity in this article should be construed to mean system capacity (i.e., the bits per second achievable in a fixed bandwidth under various requirements such as bit error rate [BER] and outage).
1

INTRODUCTION
The performance of todays cellular networks is limited by interference more than by any other single effect. Interference is distinguished from noise in that it is caused by other humandesigned devices, often mostly from devices designed to use the same network, which makes it particularly interesting and aggravating. Whereas conventional noise can be overcome by increasing transmit power, overall interference is increased by this simple-minded approach, since neighboring devices now have to contend with even more interference than before. In both centralized and ad hoc networks, overall system capacity 1 can be maximized by having each device use the minimum required transmission power so that the interference caused to other devices in the network is also minimized. Currently, achievable data rates in wide area

wireless networks such as cellular systems are fairly limited. Although peak data rates on the order of 110 Mb/s are advertised for third-generation (3G) techniques such as EV-DO [1] (for 3G Partnership Project 2, 3GPP2, i.e., cdma2000) and HSDPA [2] (for 3GPP, i.e., wideband code-division multiple access, WCDMA [3]), the actual data rate experienced by a typical subscriber is generally less than 100 kb/s, and the latency can also be very poor. In order to compete in the long term with wireless LANs such as the IEEE 802.11 family, the reliability and speed of cellular systems will need to be dramatically improved. This will be achieved by a variety of innovative algorithms at the network and physical layers, including advanced signal processing techniques at both the base station and mobile nodes. In order to understand how signal processing can be used to increase the capacity of cellular systems, it should first be recognized that the downlink and uplink have very different characteristics, and are likely to be further differentiated in future cellular systems. In the downlink each receiver only needs to decode a single desired signal from K intracell signals, while suppressing other cell interference from a few dominant sources as shown in Fig. 1. On the other hand, in the uplink the base station receiver must decode all K desired users while suppressing other cell interference from many independent sources, as shown in Fig. 2. Future cellular systems will employ sophisticated scheduling algorithms in the downlink, so the primary function of the mobile unit will be to decode the desired signal in the presence of interference from the neighboring cells. This is fortunate, since the mobile units will still be highly power limited and hence have limited processing power. It is difficult to coordinate and accurately synchronize scheduling algorithms for the uplink, since all users are at different distances from the base station and have rapidly changing multipath channels. Although the emergent time-division synchronous CDMA (TD-SCDMA) standard from China [4] has implemented uplink synchronization control, challenging the conventional wisdom that the uplink is necessarily asynchronous in CDMA systems, it is likely that most future cellular systems

IEEE Wireless Communications April 2005

1536-1284/05/$20.00 2005 IEEE

19

Multiple antenna systems will be especially subject to interference limitations since by increasing the data rate per user and using many transmit antennas, the total interference imposed on neighboring cells is further increased, making multiuser algorithms all the more prescient if MIMO is adopted

Interfering cells Desired cell

User 1

K users/cell

User 2 Multi-user transmitter User K

User 1 User 2

User K

n Figure 1. In the downlink scenario, each receiver only needs to decode its own signal, while suppressing

other-cell interference from just a few dominant neighboring cells. Because all K users signals originate at the base station, the link is synchronous and the K 1 intracell interferers can be orthogonalized at the base station transmitter. Typically, though, some orthogonality is lost in the channel. cancellation subclass of MUD shows the most promise for future implementations of multiuser receivers, and then explain in detail some recent research progress in this area. We conclude with a view toward the role of interference cancellation in future cellular systems.

will still have an asynchronous uplink. Regardless, the base station will be tasked with decoding all K users in the presence of significant inand out-of-cell interference. Although this is a more challenging task, the base station receivers will generally have much higher complexity allowance than their mobile counterparts. For these reasons, downlink receivers at the user terminals will employ relatively simple multi-user receivers that attempt to restore the orthogonality of intracell users via either a chiplevel equalizer (CDMA) or intercarrier interference suppression (multi-user orthogonal frequency-division multiplexing, OFDM) while handling at most a few dominant and unknown other-cell interferers. On the other hand, uplink receivers at the base station will employ multiuser receivers that are capable of robustly decoding all desired and interfering users in the cell in the presence of nontrivial amounts of other-cell interference. Note that utilizing other sophisticated technologies such as multiple-antenna techniques or opportunistic multi-user scheduling will not change this fundamental reality in a multicell system. Multiple-antenna systems will be especially subject to interference limitations since by increasing the data rate per user and using many transmit antennas, the total interference imposed on neighboring cells is further increased, making multi-user algorithms all the more prescient if multiple-input multiple-output (MIMO) is adopted [57]. And although multiuser scheduling may increase throughput and decrease the number of interfering users, at lower spreading factors interference suppression will become even more crucial. The article is organized to give a historical and technical perspective on multi-user receivers, and interference cancellation methods in particular. First, we describe some of the struggles multi-user detection (MUD) has faced, and overview why it has not found widespread acceptance in commercial systems despite the resounding enthusiasm behind it in the academic community. It is noted that some of these historical issues are not as relevant today, and recent research has also overcome many of MUDs barriers to entry. We explain why the interference

MULTI-USER DETECTION: HISTORICAL OVERVIEW AND SHORTCOMINGS

The idea of simultaneously receiving multiple interfering users is not particularly new. Most current wireless communications systems already have to cope with a large degree of multiple access interference. In the important special case of CDMA systems like IS-95 and 3G cellular, an interference-limited system is willfully created in order to achieve capacity benefits deriving from voice activity and universal frequency reuse. The most popular receiver approach from these CDMA systems is to use a matched filter (a.k.a. correlator) for the desired user, which treats the interfering signals as wideband noise after decorrelating the desired signal. While simple, proven, and robust, this technique is decidedly suboptimal in most circumstances from an information theory perspective, particularly when the number of interferers is large. The optimal multi-user receiver for CDMA was discovered and advanced by Verdu in the mid 1980s [8, 9], having been foreshadowed by Van Ettens multi-user sequence estimator in the mid-1970s [10] and Covers work on broadcast channels [11]. Although Verdus maximum likelihood receiver was able to optimally decode multiple users in parallel with dramatic gains, it is extremely complex, with the computational needs increasing as O ( A K ), where A is the alphabet size (2 for binary) and K is the number of users. This has led to a truly staggering volume of research in the past 20 years on suboptimal multi-user receivers with manageable complexity (e.g., [1215, references therein]). The gains made available by multi-user receivers have also been investigated from the standpoint of information theory, with interesting results some quite recent [16, 17] that fur-

20

IEEE Wireless Communications April 2005

Desired cell (K users)

K users/cell

Other cell interference from >> K users User 1 User 2

Interfering cells (~K users each) Multi-user receiver

User K

n Figure 2. In the uplink scenario, the base station receiver must decode all K desired users, while suppressing
other-cell interference from many independent users. Because it is challenging to dynamically synchronize all K desired users, they generally transmit asynchronously with respect to each other, making orthogonal spreading codes unviable.

ther emphasize the promise of advanced receiver architectures. Despite this large body of work, industry has not enthusiastically embraced multiuser receivers. In fact, despite literally hundreds of funded research projects and proposed multiuser receiver architectures, and thousands of academic papers on the subject, industrial CDMA systems still typically employ the same basic matched filter receiver structure, or simple improvements based on it. While there are a variety of explanations for this failure, including politics and inertia of the status quo standards and design methodologies, the key issues are quite clearly technical, due to the strong financial motivation to increase the capacity of CDMA systems by any attainable means. A controversial paper from QUALCOMM [18] identified at an early stage many of the fundamental problems with academic research on MUD, particularly with the widely researched linear projection techniques of the early to mid-1990s. Although this article made a number of mistakes (as partially documented in [19]), it may have helped spur more intensive research on interference cancellation from the late 1990s to the present. We now summarize the key challenges and historical shortcomings in multi-user receiver implementation, some of which were identified in [18], but will revisit some of these issues later when forecasting a bright future for multi-user reception.

with turnover on the order of every seven years, according to industry executives and expensive to maintain. So adding multi-user receivers to the base station, where they generally are more technically viable, is an uphill battle due to the reality of having to swap out an enormous amount of expensive infrastructure. Furthermore, conventional wisdom has been that moving forward, cellular systems will have asymmetric demands on data rate, with downlink demand far exceeding uplink demand due to Web and other downloads. Hence, even the basic motivation for increasing uplink capacity with multi-user receivers is not obvious. Why add expensive new receivers at the base station if that new capacity will not even be used?

While it may be possible to add new network features through the fastchurning phones, the added cost and complexity for advanced signal processing algorithms like multiuser detection or MIMO must be negligible due to the extreme sensitivity to cost and consumed power.

DIMENSIONAL SEPARATION OF USERS AND ERROR CORRECTING CODES

Many proposed multi-user receivers attempt to orthogonalize the users into unique dimensions at the receiver by means of linear filtering (whether decorrelating, minimum mean square error [MMSE], adaptive, blind, etc.) [2022]. For synchronous systems like the cellular downlink, this approach does not make sense; a better approach would be to simply assign the users orthogonal codes at the transmitter and maintain a simple single-user receiver. Although the orthogonality of these codes is badly damaged by the multipath channel, a pseudorandom noise (PN) scrambling code can be used along with a RAKE receiver to improve the autocorrelation properties of the concatenated orthogonal/PN code. Alternatively, a sophisticated receiver could be used to cancel or equalize the multipath interference, with the users themselves orthogonal on each multipath component [23]. For asynchronous systems like the cellular uplink with random user timing offsets, mutually orthogonal codes are not feasible. But again, multi-user receivers that attempt to project the received users into orthogonal dimensions do not stand up to careful scrutiny. In particular, most of these techniques require the construction of a code cross-correlation matrix. But for the long PN codes generally desired for multi-

COMPLEXITY, COST, AND SERVICE PROVIDER MOTIVATION

Personal wireless devices have become a consumer item, often heavily subsidized by the service provider. The cost of both the end-user telephone and the base station are hence very sensitive. The telephones churn rather quickly due to consumer demand for new features and trendy models, with a typical turnover on the order of two years. While it may be possible to add new network features through fast-churning phones, the added cost and complexity for advanced signal processing algorithms like MUD or MIMO must be negligible due to extreme sensitivity to cost and consumed power. Base stations, on the other hand are less cost-sensitive

IEEE Wireless Communications April 2005

21

The methodology and application of future wireless networks (beyond 3G) may change dramatically from the current voicecentric, circuitswitched paradigm, and other recent technical developments give new hope for certain classes of multiuser receivers.

path properties, this matrix must usually be recomputed every symbol, which is impractical.2 Even more important, it is now well established that strong error correction codes (ECCs) should be used in the uplink (and the downlink, for that matter). The spreading gain of these codes is lost on dimensional multi-user detectors. Furthermore, while the goal of these codes is to lower the required received signal-to-interference-plus-noise ratio (SINR), this complicates channel estimation and other aspects of dimensional MUD. Since linear projection receivers require spreading to be performed by a linear operation, there is direct competition between ECCs and dimensional (linear) MUD [24]. However, as discussed next, interference cancelling multi-user receivers not only do not suffer from this trade-off, but in fact need ECCs to attain their high performance, a result supported by information theory [2527].

THE FUTURE OF MULTI-USER DETECTION

Despite the bleak picture for multi-user receivers that unfolded in the previous section, the future appears much brighter. The methodology and application of future wireless networks (beyond 3G) may change dramatically from the current voice-centric circuit-switched paradigm, and other recent technical developments give new hope for certain classes of multi-user receivers. Many researchers have begun to recognize that the important remaining challenges for multi-user receivers are no longer information theoretic, but rather of system design and implementation, as enumerated earlier. Some encouraging recent trends are now highlighted, in rebuttal to the arguments of the previous section.

ROBUSTNESS TO THE REAL-WORLD ENVIRONMENT

The real-world cellular environment is an extremely challenging radio environment. Highly mobile users have rapidly fading and changing channels of widely varying gains: a typical cellular system might have channel gain differences between the strongest and weakest supported users of over five orders of magnitude. Users on the cell boundaries not only face the weakest channels to their desired base station(s), they also experience the largest amount of other-cell interference. Digital receivers generally employ low-resolution A/D convertors, which along with other A/D nonlinearities further distorts the received signal. Add in timing synchronization errors and filtering mismatches between the transmitter and receiver, and there is a catalog of imperfections that must be dealt with, but are usually neglected in the literature.

COMPLEXITY, COST, AND SERVICE PROVIDER MOTIVATION

New end-user behavior is increasing the demand for uplink data rates in all types of communication systems. Peer-to-peer applications, multimedia messaging service (MMS), camera phones, gaming, and other high-data-rate applications have made the assumption of asymmetric data rates questionable. Lower spreading factors are the norm for the higher-rate standards that support these applications (e.g., HSDPA [2]), making multi-user receivers increasingly viable and valuable. Furthermore, while it is widely expected that fourth-generation (4G) cellular systems will be packet-switched and multicarrier-based, it is not at all clear how the uplink will operate. Due to many factors including the asynchronous link, large bandwidth (making per-user guard times very expensive), and other-cell interference, CDMA will remain very attractive for uplink cellular communications involving nontrivial numbers of users. In addition, the uplink is often coverage limiting [28], and interference cancellation can improve coverage considerably [29]. The combination of these important factors makes multi-user receivers quite attractive for cellular networks, relative to 3G systems. Lowcomplexity and robust solutions will obviously be mandatory.

POLITICAL AND OTHER NONTECHNICAL ISSUES

Although the key issues thus far enumerated are technical or implementation issues, there is no question that politics and technical misunderstandings have also contributed to MUDs underwhelming impact on current systems. In order to make use of most new cutting edge technologies, the entire system should be designed in an integrated fashion with the new technologies in mind. For example, no one plans to simply put extra antennas at the transmitter and receiver, expecting MIMO to work: there need to be special pilot signals, coding techniques, receiver architectures, and a specialized medium access control (MAC) layer to effectively take advantage of the multiple antennas. On the other hand, CDMA and other cellular standards have considered MUD to simply be an optional receiver implementation, and have made very minimal effort to accommodate MUD in the standards (e.g., by including an optional provision for short period spreading codes [3]). It is not altogether clear why this has happened, but it is reasonable to believe that a different design approach could have yielded a far more important and profitable role for multi-user receivers in current cellular systems.

DIMENSIONALITY AND ROBUSTNESS

Most researchers have come to recognize that in a highly frequency-selective and rapidly changing channel with many users, dimensional separation techniques at the receiver are not viable. Further, it has been realized that effective MUD designs must integrate error correction coding/ decoding into the receiver structure, rather than treating coding as a separately concatenated block. For this reason, successive [3032], parallel [33, 34], and iterative interference cancellation (turbo MUD) [3538] appear quite attractive. This article focuses on interference cancellation techniques that, relative to other types of multi-user receivers, can be applied much more readily to existing cellular systems. Additionally, as seen in Table 1, interference cancellation techniques also have desirable attributes as far as complexity and latency, while supporting error correction codes and high user densities. In the

Despite this general truth, there has been some work on linear MUD for long PN sequences (e.g., [12, Sec. 5.4]).

22

IEEE Wireless Communications April 2005

MUD type Optimal max. likelihood Linear Turbo Parallel IC Successive IC Nonorth. matched filter Orth. matched filter
1 2

Complexity order 2K K to K3 PK to 2K PK K K K

Latency 1 1 2P P K 1 1

ECCs? Separate Separate1 Integrated Integrated Integrated Separate Separate

K > N allowed? Yes No (ZF), Yes (MMSE) Yes Yes Yes Yes2 No

With some exceptions (e.g., [39]), generally linear receivers cannot seamlessly integrate ECCs. Although allowed in principle, K > N is not likely to be achievable in practice for the MF receiver.

n Table 1. Key general trends of different multiuser receivers, with spreading factor N, number of users K,
and P receiver stages. next section we consider interference cancellation techniques most applicable to the uplink, although we also consider recent joint detection techniques that can be used in the downlink to cancel other cell interference. Table 1 gives a high-level comparison of some of the different types of multi-user receivers. Due to the vast number of different subtypes for each of these receivers in the literature, these values should be interpreted as general trends. The different types of multi-user receivers have dramatically different scalings for complexity and latency, which generally depends on the number of users K and the number of receiver iterations P. They also incorporate error correction codes in different ways: in some systems like turbo MUD and SIC, ECCs are directly integrated into the receiver structure, whereas in other systems they must form a separate block independent of user separation, which generally reduces performance since the coding redundancy competes with the spreading gain, N. well justified from a theoretical point of view as well, starting with the first (to the authors knowledge) such suggestion of the multi-user interference cancelling principle [11]. Despite its apparent simplicity, a simple successive interference cancellation implementation with suboptimal coding was shown in 1990 to nearly achieve the Shannon capacity of multi-user additive white Gaussian noise (AWGN) channels, assuming accurate channel estimation and a large spreading factor [30]. Other work has proven that SIC with single-user decoding in fact achieves the Shannon capacity region boundaries for both the broadcast (downlink) and multiple access (uplink) multi-user channel scenarios [25, 26]. Additionally, a linear MMSE receiver with single-user coding has been shown to achieve the same capacity as the optimum (maximum likelihood, ML) receiver when combined with SIC [27]. A well known and successful example of interference cancellation is the decision feedback equalizer (DFE), which is used to cancel intersymbol interference (ISI) in frequency-selective channels. In the DFE, the desired symbol x[n] at some time n is decoded. Since this symbol will interfere with many future symbols (i.e., from times n + 1; n + 2, ), given knowledge of the channel, this ISI can be cancelled. The DFE is known to work well in practice, and achieve far better performance than linear equalizers, which suffer from noise enhancement. The same reasoning applies to analogous types of interference, such as multi-user interference or spatial interference. The original Bell Labs layered space-time (BLAST) system [40] and industry adaptations of spatial interference cancellation receivers for multi-antenna systems can be used to separate spatially multiplexed streams of data [41]. These types of postprocessing receivers often significantly outperform standalone linear receivers such as MMSE or zero-forcing in noisy environments. Since the cellular environment will invariably have a high

In addition to being a practical approach to multiuser detection, interference cancellation is fundamentally well-justified from a theoretical point of view as well, starting with the first (to the authors knowledge) such suggestion of the multiuser interference cancelling principle.

INTERFERENCE CANCELLATION: A PROVEN APPROACH

Although the application of interference cancellation to multi-user systems is relatively new and unproven, other forms of interference cancellation have been in widespread use for years. Since the phrase has been used quite loosely, interference cancellation should be interpreted to mean the class of techniques that demodulate and/or decode desired information, and then use this information along with channel estimates to cancel received interference from the received signal. In other words, the front-end may have any type of structure, but for the system to be classified as interference cancellation, signal processing is used after detection to reduce the influence of the interference on future decisions. In addition to being a practical approach to MUD, interference cancellation is fundamentally

IEEE Wireless Communications April 2005

23

Matched filter 1 y(t)

IC block

Matched filter 1

Final decision

^ b1

Matched filter K

IC block

Matched filter K

Final decision

^ bK

n Figure 3. Parallel interference cancellation.

ence cancellation (SIC) and parallel interference cancellation (PIC) have the important advantage over other types of multi-user receivers that error correction coding is integrated into the MUD process, as demonstrated for SIC in 1990 [30] and later for PIC [34]. As previously noted, both SIC and PIC are primarily applicable to the uplink in a many-user CDMA system. However, for a non-CDMA system, these same powerful techniques, suitably modified, can help suppress other-cell interference in the downlink. We now describe these two schemes, and then focus on some of their historical shortcomings and recent research that has addressed these issues to show new promise for multi-user interference cancellation. PIC, as shown in Fig. 3, detects all users simultaneously. This initial very coarse estimate can then be used to cancel some interference, and then parallel detection can be repeated. This process can be repeated over several stages; hence, PIC is sometimes called multistage interference cancellation [33, 42]. Since the first stage generally results in very noisy data estimates, soft interference cancellation is necessary [43]. SIC, shown in simplified format in Fig. 4, detects just one user per stage. The strongest received signal is detected first, then the next strongest, and so on. After each users transmitted data is estimated, the received signal for that user can be reconstructed by recreating the transmit signal and applying an estimate of the channel to it. This can be subtracted from the composite received signal, which then allows subsequent users to experience a cleaner signal. All users have improved performance: earlier users because they can have disproportionately high received power, and later users because a large fraction of the total interference has been removed by the time they are detected. There are a variety of trade-offs between SIC and PIC. PIC has decreased latency, but higher overall complexity because K users must be detected in parallel, plus there are P cancellation stages. So the latency is proportional to P, which is generally much smaller than K for cellular systems, but complexity is proportional to PK. SIC, on the other hand, has complexity and latency proportional to K, and this latency may be prohibitive if there are many users with real-time data. Some authors have attempted to provide a

y0(t)

Update composite signal yk+1(t)= yk(t) zk(t)

Linear CDMA receiver

^ bk Decoder

zk(t)

Imperfect channel estimation

Re-encode and modulate

n Figure 4. Successive interference cancellation.

level of noise and background interference (e.g., from other cells), these linear interference suppression techniques are not viable as they amplify this noise when inverting the spatial matrix channel. Analogous logic can be applied to multi-user systems, and it is well established that linear (dimensional) multi-user detectors have a noise enhancement penalty [12]. In short, there are many similarities between the suppression of: ISI (equalizer design) Spatial interference (MIMO receivers) Multi-user interference (multiuser detectors) All three types of interference are best handled by an ML receiver, but this is generally too complex. Linear techniques are frequently considered, usually in the form of zero-forcing or MMSE designs, and are attractive from a complexity point of view but not very robust. Nonlinear interference cancellation techniques tend to be the most practical solution with the best performance, but also the hardest to analyze and grasp intuitively. The next section overviews recent progress in the research of multi-user interference cancelers.

RECENT RESEARCH ON MULTI-USER INTERFERENCE CANCELLATION

Interference cancellation for multi-user systems has generally been broken into two categories, parallel and successive, although recent developments in iterative interference cancellation have blurred the distinction. Both successive interfer-

24

IEEE Wireless Communications April 2005

smoother trade-off between these techniques by introducing multistage SIC [44, 45]: a group of users are detected in parallel, and then has their aggregate interference subtracted from the composite received signal, and then another group is detected in parallel. There is some debate as to whether PIC or SIC has higher overall system capacity in practice, with a slight consensus that SIC performs better if both are implemented in an optimal manner. If all users are received with equal power, PIC is definitely better, whereas if the users have unequal received power levels, SIC is definitely preferable [31]. The debate has been complicated somewhat recently with the introduction of iterative parallel interference cancellers, also referred to as turbo MUD as noted previously, which achieve performance very near the single-user limit [3538]. There have been several historical problems with multi-user interference cancellation, but in recent years researchers have successfully extensively addressed most of them, clearing the path for commercial adoption. Due to their quite different structure, while PIC and SIC share some of the same challenges, they also have several challenges unique to each of them. We now review some of these challenges, and discuss how they can be addressed. Additionally, we overview and compare recent research on iterative interference cancellation and discuss practical implementation issues, with an example of the recently deployed interference cancellation techniques for GSM systems.

2 Spectral efficiency per sector (b/chip or b/s/Hz) 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0 0.05 0.1 0.15 0.2 0.25 0.3

SIC, optimal power control SIC, previous power control Reference system w/o SIC

0.35

0.4

0.45

0.5

Normalized channel estimation error standard deviation ()

n Figure 5. Spectral efficiency comparison using successive interference cancellation. As long as just the statistics of the channel estimation error are known and incorporated in the power control algorithm, the capacity increase from SIC is dramatic and fairly robust to increasing channel estimation error.

CHANNEL ESTIMATION ERROR AND ERROR PROPAGATION

The entire concept of interference cancellation is based on the premise that the received signal can be reliably estimated. Whereas communication systems are by definition designed to allow the transmitted signal to be recovered, reconstructing the received signal requires accurate description of both what was transmitted and what the channel did to that transmission. Inaccurate channel estimation is a problem for both PIC and SIC, especially for SIC since historically the optimal received power distribution is based on the assumption that interference has been completely cancelled [46, 30], which is never fully achieved in practice. This residual interference then causes later users to have unacceptably bad performance, causing a major fairness problem as well as an overall degradation in bit error rate and system capacity. Some recent research has proposed different methods of addressing this problem. Some authors have suggested a multistage approach to SIC [47], with iterative channel estimation to improve this situation. While effective, this causes increased latency and computational complexity. Another approach, first suggested in [48] and then formalized in [49, 50], is to alter the power control algorithm to take into account the channel estimation error. Essentially, later users are given relatively more power as the channel estimation error increases to compensate for this residual interference. Adopting this commonsense approach to power control has a large

capacity payoff. In Fig. 5 [50] it can be seen that if the channel estimation error is on average larger than about 20 percent, the system is better off without SIC, assuming traditional power control for SIC is used that assumed perfect interference cancellation, as in [46, 30]. By using a modified power control algorithm that accounts for the statistics of the channel estimation error, even with dramatic estimation error as high as 50 percent, SIC nearly doubles the system capacity relative to no interference cancellation. Whereas the key to ameliorating the channel estimation error problem for SIC appears to be properly modified power control, for PIC it may be handled through a combination of multistage detection, error correction coding, and more recently iterative detection and channel estimation [51, 47].

POWER CONTROL
Although PIC functions best in the familiar case where all the received powers are equal, we have just seen that SIC works best when a specific and unequal distribution of user powers is maintained, and furthermore when the distribution specifically considers imperfect interference cancellation. Unequal received power distribution has also been shown to be highly preferable for iterative interference cancellation [52]. This apparent complication of CDMA power control has frequently been cited as a major shortcoming of SIC (e.g., [53, 43]). But recently it has been shown in [54], using new results in power control theory [5557], that the optimum SIC power distribution, even with channel estimation error accounted for, can easily be accomplished using binary iterative feedback algorithms. This somewhat surprising result means that commercial CDMA power control algorithms can be directly applied to SIC without any modification.

IEEE Wireless Communications April 2005

25

Received signal vector for K users CDMA detector (MF, MMSE, PIC, etc.)

Note that here MC-CDMA should be interpreted to mean a system with a large number of fairly narrowband subcarriers typically generated by an (inverse) fast Fourier transform not the three-carrier system proposed in conjunction with 3GPP2 that is simply three single-carrier systems in parallel.
Interleaver

Deinterleaver

ITERATIVE INTERFERENCE CANCELLATION

The development of turbo codes in the early 1990s [69] opened up a new methodology for decoding low-signal-to-noise-rate (SNR) signals. By iteratively passing probabilistic estimates between two decoders, dramatic performance improvements over prior decoding techniques were observed. This same philosophy can be applied to decoding multi-user CDMA signals with ECCs, where one decode is any of the CDMA detectors previously mentioned (matched filter, linear MUD, ML), and the other is a standard single-user ECC decoder. This basic framework is illustrated in Fig. 6. Note that this framework is quite general and can also be used to describe SIC and PIC. The early development of iterative interference cancellation for CDMA came in 19981999, with early work by Moher [70], Reed et al. [36], Alexander et al. [37], and Wang and Poor [38] all showing near-single-user performance with various CDMA system configurations and receiver strategies. Related earlier work included turbo equalization [71], and joint detection of ECCs and interfering users [72, 34], but the connection between multi-user iterative receivers and ISI-suppressing iterative receivers was mainly appreciated in retrospect. There has been a large amount of research on this topic since this early pioneering work, with efforts to adapt iterative IC to have low complexity and robust performance in multipath channels [73]. Of particular interest is recent work that has shown that interference cancellation works almost as well as the more complex belief propagation approaches [52] if the received SINR for the K users is optimized by a suitable power control algorithm, as discussed previously in this section.

Single-user ECC decoder Estimated bit vector for K users

n Figure 6. An iterative interference cancelling receiver.

Hence, both PIC and SIC appear to be readily able to solve the near-far problem without altering the existing standards. An additional advantage of schemes with unequal received powers is that the disparate power levels can be assigned to users strategically to reduce other-cell interference (OCI) and total transmit power [58, 50]. This OCI reduction is accomplished by assigning users at the cell boundaries the relatively lower required received power levels so that the interference caused to their neighboring cells is decreased.

MULTIPATH CHANNELS
Multipath channels are challenging for all wireless systems, but particularly for multi-user receivers. The reason is that each multipath component can appear to be a user of this system, so the quantity of perceived users grows not just with the number of users K, but also roughly with the number of multipath components L . Although interference cancellation receivers, which are based on conventional receivers, can easily employ a RAKE receiver to handle multipath, it may be difficult to accurately regenerate the interference for cancellation if there are many multipath components. If each multipath component has an independent amplitude and phase, generally the estimation error for each will be independent. This can cause the capacity to decrease rapidly as the multipath profile worsens [59], since channel estimation errors affect all users signals in every dimension when the multipath interference is regenerated in the time domain by the RAKE encoder. When multipath has been treated in the interference cancellation literature, researchers have often assumed that the channel is perfectly known (e.g. [60]) or have not adapted to the imperfect channel estimation [6163]. In addition to the more recent joint channel estimation/detection approaches discussed above, another promising technique is multicarrier CDMA (MC-CDMA) [64], which then allows channel estimation to be done in the frequency domain, where channel estimation errors do not compound each other as they do in the time domain. When MC-CDMA is used for interference cancellation [6568], large gains are attained in realistic multipath fading channels.

DOWNLINK MULTI-USER RECEIVERS

Uplink multi-user receivers need to suppress interference from a large number of independent transmitters, while downlink receivers should be able to suppress interference from just a few neighboring base stations, but with extreme complexity limitations. Along these lines, recent research has attempted to develop modified ML and linear detectors for the GSM downlink that are able to suppress other-cell interference [7477]. While this is for a single-carrier nonCDMA system, it provides some important lessons for how future cellular systems may be able to cope with other-cell interference in the downlink. Exploiting the characteristics of Gaussian mininum shift keying (GMSK) modulation allows signal reception to be split into two virtual paths, which can then be processed using classical linear processing techniques such as ZF or MMSE [74]. Hybrid schemes that employ a combination of linear and nonlinear processing are

26

IEEE Wireless Communications April 2005

also possible. The hybrid schemes contrast with joint detection as they do not require estimation of the channel response of the interferer, and are thus referred to as blind techniques. An advantage of being blind with respect to the interferers channel response is that these techniques are amenable to asynchronous networks, where the channel training codes of the interferer will not typically overlap with the training codes of the desired signal. However, as shown in [75], joint detection is also possible in asynchronous networks provided the mobile terminal platform can handle the complexity. Collectively, these joint detection and hybrid/linear receivers are referred to as single-antenna interference cancellation (SAIC) receivers. It should be noted that these kinds of interference cancelling receivers may employ either ML detection or predetection processing rather than the postdetection interference cancellation emphasized in this article. SAIC techniques have proven successful in field trials in suppressing other-cell interference in GSM systems [76], and will likely play a major role in future cellular systems as a method of decreasing the spatial reuse distance, which is critical to achieving high overall spectral efficiency. Further research will be needed to modify these techniques for multicarrier systems such as 802.16/WiMax and low-spreading-factor 3G CDMA data systems such as HSDPA and EVDO.

THE ROLE OF INTERFERENCE CANCELLATION IN FUTURE CELLULAR SYSTEMS

As we have attempted to establish, due to evolving applications, future cellular systems are likely to carry high data rates with more burstiness than voice. The synchronous downlink will likely employ opportunistic scheduling across time, frequency, and/or codes, so the majority of the interference at mobile stations will come from a small number of neighboring base stations, rather than the present scenario where much of the interference is intracell. On the other hand, the uplink may require higher data rates than until recently believed, and will still be an at least partially asynchronous link, making orthogonalization of different users codes or subcarriers in either CDMA or OFDM difficult. For these reasons, multi-user receivers will play an important, but different, role in the downlink and uplink of future cellular systems. In the uplink there will be many asynchronous users, although fewer than in a 2G or 3G system as the cells will continue to grow smaller while data will be burstier and higher-rate, and hence the spreading factor smaller. This makes interference cancellation techniques all the more desirable and practical. Recent research on interference cancellation, highlighted in this article, has made significant strides, but more research and development is required to prototype these systems and adapt them to real-world environments. For the downlink, the need will be to attenuate the interference from a small number of neighboring base stations while maintaining the orthogonality of the users within the cell in time, frequency, code, and/or space, as the case may be. This implies that the success of the recent research and implementation of SAIC techniques for single-carrier GSM systems should be extended through future research to higherbandwidth and more complex multicarrier and CDMA systems.

SAIC techniques have proven successful in field trials in suppressing other cell interference in GSM systems, and will likely play a major role in future cellular systems as a method to decrease the spatial reuse distance, which is critical to achieving high overall spectral efficiency.

COMPLEXITY AND IMPLEMENTATION

As the literature on multi-user receivers has become increasingly comprehensive in recent years, the bottleneck for the adoption of multiuser receivers has increasingly become issues relating to complexity and implementation. Many of these implementation issues are not simply a matter of efficient integrated circuit design, but rather require system-level thinking about the crucial features of multi-user receivers. Although not discussed further here, it should be noted as discussed earlier that because the signal processing algorithms for multi-user and multi-antenna suppression are in essence indistinguishable, implementation progress in MIMO receivers can be immediately applied to advance the implementation of multiuser receivers. A preliminary implementation for SIC receivers was considered in [78], while lowpower linear multi-user detectors were implemented in [79]. A power-hungry PIC receiver was considered in [80], while an integrated circuit version was implemented in [81], and important real-time implementation issues and a real-time architecture were presented in [82]. NTT DoCoMo has developed a prototype uplink PIC receiver, with a complexity about five times greater than that of a conventional receiver [42]. A WCDMA-compliant iterative MUD prototype [29] has been shown to increase capacity by a factor of 23 and coverage by around 50 percent. In short, significant steps have been taken recently toward the realization of practical multiuser receivers, but more research and development is needed to make multi-user receivers practical for future standards.

ACKNOWLEDGMENTS
Feedback from Sergio Verdu (Princeton), Jack Holtzman (Qualcomm), Mark Reed (Australia National University), Aris Papasakellariou (TI), Rich Kobylinski (SBC), and Avneesh Agrawal (Qualcomm) have been very helpful in the preparation of this article. The comments and suggestions of the anonymous reviewers were also especially useful.

REFERENCES
[1] S. R0023, High Speed Data Enhancement for CDMA2000 1x-Data Only, 3GPP2, June 200. [2] High Speed Downlink Packet Access (HDSPA), Release 5, 3GPP, 2003. [3] Wideband CDMA Standard, Release 99, 3GPP, 1999. [4] B. Li et al., Recent Advances on TD-SCDMA in China, IEEE Commun. Mag., vol. 43, no. 1, Jan. 2005, pp. 3844. [5] R. Blum, J. Winters, and N. Sollenberger, On the Capacity of Cellular Systems with MIMO, IEEE Commun. Lett., vol. 6, June 2002, pp. 24244. [6] H. Dai, A. Molisch, and H. Poor, Downlink Capacity of Interference-Limited MIMO Systems with Joint Detection, IEEE Trans. Wireless Commun., vol. 3, no. 2, Mar. 2004, pp. 44253. [7] W. Choi and J. G. Andrews, Spatial Multiplexing in Cellular MIMO CDMA Systems with Linear Receivers, Proc. IEEE ICC, Seoul, Korea, May 2005.

IEEE Wireless Communications April 2005

27

Because the signal processing algorithms for multiuser and multiantenna suppression are in essence indistinguishable, implementation progress for MIMO receivers can be immediately applied to advance the implementation of multiuser receivers.

[8] S. Verdu, Minimum Probability of Error for Asynchronous Gaussian Multiple-Access Channels, IEEE Trans. Info. Theory, vol. 32, Jan. 1986, pp. 8596. [9] S. Verdu,, Optimum Multiuser Signal Detection, Ph.D. diss., Univ. of IL, Urbana- Champaign, Aug. 1984. [10] W. Van Etten, Maximum Likelihood Receiver for Multiple Channel Transmission Systems, IEEE Trans. Commun., vol. 24, no. 2, Feb. 1976, pp. 27683. [11] T. Cover, Broadcast Channels, IEEE Trans. Info. Theory, vol. 18, no. 1, Jan. 1972, pp. 214. [12] S. Verdu, Multiuser Detection, Cambridge Univ. Press, 1998. [13] A. Duel-Hallen, J. Holtzman, and Z. Zvonar, MultiUser Detection for CDMA Systems, IEEE Pers. Commun., Apr. 1995, pp. 4658. [14] S. Moshavi, Multi-user Detection for DS-CDMA Communications, IEEE Commun. Mag. , vol. 34, no. 10, Oct. 1996, pp. 12436. [15] X.Wang and H. Poor, Wireless Communication Systems: Advanced Techniques for Signal Reception, Prentice Hall, 2003. [16] S. Verdu and S. Shamai, Spectral Efficiency of CDMA with Random Spreading, IEEE Trans. Info. Theory, vol. 45, Mar. 1999, pp. 62240. [17] D. Tse and S. Hanly, Linear Multiuser Receivers: Effective Interference, Effective Bandwidth and Capacity, IEEE Trans. Info. Theory, vol. 45, Mar. 1999, pp. 64175. [18] S. Vembu and A. Viterbi, Two Different Philosophies in CDMA A Comparison, Proc. IEEE VTC, May 1996, pp. 86973. [19] S. Verdu, Demodulation in The presence of Multiuser Interference: Progress and Misconceptions, Intelligent Methods in Sig. Proc. and Commun. , 1997, pp. 1544. [20] R. Lupas and S. Verdu, Linear Multiuser Detectors for Synchronous Code-Division Multiple Access Channels, IEEE Trans. Info. Theory, vol. 35, no. 1, Jan. 1989, pp. 12336. [21] U. Madhow and M. L. Honig, MMSE Interference Suppression for Direct-Sequence Spread-Spectrum CDMA, IEEE Trans. Commun. , vol. 42, no. 12, Dec. 1994, pp. 317888. [22] M. Honig, U. Madhow, and S. Verdu, Blind Adaptive Multiuser Detection, IEEE Trans. Info. Theory, vol. 41, no. 4, July 1995, pp. 94460. [23] K. Higuchi, A. Fujiwara, and M. Sawahashi, Multipath Interference Canceller for High-Speed Packet Transmission with Adaptive Modulation and Coding Scheme in W-CDMA Forward Link, IEEE JSAC, vol. 20, no. 2, Feb. 2002, pp. 41932. [24] V. V. Veeravalli and A. Mantravadi, The CodingSpreading Tradeoff in CDMA Systems, IEEE JSAC, vol. 20, no. 2, Feb. 2002, pp. 396408. [25] T. Cover and J. Thomas, Elements of Information Theory, Wiley, 1991. [26] B. Rimoldi and R. Urbanke, A Rate Splitting Approach to the Gaussian Multiple Access Channel, IEEE Trans. Info. Theory, vol. 42, Mar. 1996, pp. 36475. [27] M. K. Varanasi and T. Guess, Optimum Decision Feedback Multiuser Equalization with Successive Decoding Achieves the Total Capacity of the gaussian Multipleaccess Channel, Proc. IEEE Asilomar, Pacific Grove, CA, Nov. 1997, pp. 140509. [28] H. Holma and A. T., Eds., WCDMA for UMTS, 2nd ed., Wiley, 2002. [29] M. Reed, P. Hertach, and J. Maucher, An Iterative Multiuser Detection Receiver for 3GPP with Antenna Arrays: Performance in Terms of BER, Cell Size and Capacity, Mobile Commun. Tech. Conf. 3G 2002, London, U.K., May 2002. [30] A. J. Viterbi, Very Low Rate Convolutional Codes for Maximum Theoretical Performance of Spreadspectrum Multiple-Access Channels, IEEE JSAC , vol. 8, May 1990, pp. 64149. [31] P. Patel and J. Holtzman, Analysis of a Simple Successive Interference Cancellation Scheme in a DS/CDMA System, IEEE JSAC, vol. 12, no. 5, June 1994, pp. 796807. [32] P. Frenger, P Orten, and T. Ottosson, Code-Spread CDMA with Interference Cancellation, IEEE JSAC, vol. 17, no. 12, Dec. 1999, pp. 209095. [33] M. K. Varanasi and B. Aazhang, Multistage Detection in Asynchronous Code-Division Multiple Access Communications, IEEE Trans. Commun., vol. 38, no. 4, Apr. 1990, pp. 50919. [34] T. R. Giallorenzi and S. G. Wilson, Suboptimum Multiuser Receivers for Convolutionally Coded Asynchronous DS-CDMA Systems, IEEE Trans. Commun. , vol. 44, no. 9, Sept. 1996, pp. 118396.

[35] P. Jung and M. Nasshan, Results on Turbo-Codes for Speech Transmission in a Joint Detection CDMA Mobile Radio System with Coherent Receiver Antenna Diversity, IEEE TraNs. Vehic. Tech., vol. 46, no. 4, Apr. 1997, pp. 86270. [36] M. C. Reed et al., Iterative Multiuser Detection for CDMA with FEC: Near Single User Performance, IEEE Trans. Commun., vol. 46, no. 12, Dec. 1998, pp. 169399. [37] P. D. Alexander et al., Iterative Multiuser Interference Reduction: Turbo CDMA, IEEE Trans. Commun. , vol. 47, no. 7, July 1999, pp. 100814. [38] X. Wang and H. Poor, Iterative (Turbo) Soft Interference Cancellation and Decoding for Coded CDMA, IEEE Trans. Commun., vol. 47, no. 7, July 1999, pp. 104661. [39] P. D. Alexander, L. K. Rasmussen, and C. B. Schlegel, A Linear Receiver for Coded Multiuser CDMA, IEEE Trans. Commun., vol. 45, May 1997, pp. 60511. [40] G. J. Foschini, Layered space-time Architecture for Wireless Communication in a Fading Environment When Using Multiple Antennas, Bell Labs Tech. J., vol. 1, no. 2, 1996, pp. 4159. [41] Lucent, Increasing MIMO Throughput with Per-Antenna Rate Control, TSGR1#21(01)0879, Turin, Italy, Aug. 2001, pp. 113. [42] M. Sawahashi et al. , Expirements on Pilot-Assisted Coherence Multistage Interference Canceller for DSCDMA Mobile Radio, IEEE JSAC , vol. 20, no. 2, Feb. 2002, pp. 43349. [43] D. Divsalar, M. Simon, and D. Raphaeli, Improved Parallel Interference Cancellation for CDMA, IEEE Trans. Commun., vol. 46, no. 2, Feb. 1998, pp. 25868. [44] M. K. Varanasi, Group Detection for Synchronous Gaussian Code-Division Multiple Access Channels, IEEE Trans. Info. Theory, vol. 41, no. 4, July 1995, pp. 108396. [45] F. Wijk, G. Janssen, and R. Prasad, Groupwise Successive Interference Cancellation in a DS/CDMA System, Proc. IEEE PIMRC, 1995, pp. 74246. [46] D. Warrier and U. Madhow, On the Capacity of Cellular CDMA with Successive Decoding and Controlled Power Disparities, Proc. IEEE VTC, May 1998, pp. 187377. [47] M. Kobayashi, J. Boutros, and G. Caire, Successive Interference Cancellation with SISO Decoding and EM Channel Estimation, IEEE JSAC , vol. 19, no. 8, Aug. 2001, pp. 145060. [48] R. Muller and J. Huber, Capacity of Cellular CDMA Systems Applying Interference Cancellation and Channel Coding, Proc. IEEE GLOBECOM , Nov. 1997, pp. 17984. [49] R. M. Buehrer, Equal BER Performance in Linear Successive Interference Cancellation for CDMA Systems, IEEE Trans. Commun. , vol. 49, no. 7, July 2001, pp. 125058. [50] J. G. Andrews and T. Meng, Optimum Power Control for Successive Interference Cancellation with Imperfect Channel Estimation, IEEE Trans. Wireless Commun. , vol. 2, no. 2, Mar. 2003, pp. 37583. [51] P. Alexander and A. Grant, Iterative Channel and Information Sequence Estimation in CDMA, Proc. IEEE Intl. Symp. Spread Spectrum Techniques and Apps. , Newark, NJ, Sept. 2000, pp. 59397. [52] G. Caire, R. R. Muller, and T. Tanaka, Iterative Multiuser Joint Decoding: Optimal Power Allocation and Low-complexity Implementation, IEEE Trans. Info. Theory, vol. 50, no. 9, Sept. 2004, pp. 195073. [53] S. Verdu, Adaptive Multiuser Detection, Proc. IEEE Intl. Symp. Spread Spectrum Techniques and Apps, , July 1994, pp. 4350. [54] A. Agrawal et al., Iterative Power Control with Successive Interference Cancellation for DS-CDMA Systems, IEEE Trans. Wireless Commun., vol. 4, no. 3, May 2005. [55] R. D. Yates, A Framework for Uplink Power Control in Cellular Radio Systems, IEEE JSAC, vol. 13, no. 7, Sept. 1995, pp. 134147. [56] J. Herdtner and E. Chong, Analysis of a Class of Distributed Asynchronous Power Control Algorithms for Cellular Wireless Systems, IEEE JSAC , vol. 18, no. 3, Mar. 2000. [57] D. Kim, On the Convergence of Fixed-step Power Control Algorithms with Binary Feedback for Mobile Communication Systems, IEEE Trans. Commun. , vol. 49, no. 2, Feb. 2001, pp. 24952. [58] P. Hatrack and J. Holtzman, Reduction of Other-Cell Interference with Integrated Interference Cancellation/ Power Control, Proc. IEEE VTC, May 1997, pp. 184246. [59] J. G. Andrews and T. Meng, Multiple Access Interference Cancellation in Fading Multipath Channels: Progress and Limitations, Proc. IEEE Vehic. Tech. Conf., Rhodes, Greece, May 2001, pp. 180408.

28

IEEE Wireless Communications April 2005

[60] A. Nahler, R. Irmer, and G. Fettweis, Reduced and Differential Parallel Interference Cancellation for CDMA Systems, IEEE JSAC, vol. 20, no. 2, Feb. 2002, pp. 23747. [61] Y. C. Yoon, R. Kohno, and H. Imai, A Spread-Spectrum Multiaccess System with Cochannel Interference Cancellation for Multipath Fading Channels, IEEE JSAC, vol. 11, no. 7, Sept. 1993, pp. 106775. [62] J. Weng, G. Xue, T. Le-Ngoc, and S. Tahar, Multistage Interference Cancellation with Diversity Reception for Asynchronous QPSK DS/CDMA Systems over Multipath Fading Channels, IEEE JSAC , vol. 17, no. 12, Dec. 1999, pp. 216280. [63] A. Hui and K. Letaief, Successive Interference Cancellation for Multiuser Asynchronous DS/CDMA Detectors in Multipath Fading Links, IEEE Trans. Commun., vol. 46, no. 3, Mar. 1998, pp. 38491. [64] S. Hara and R. Prasad, Overview of Multicarrier CDMA, IEEE Commun. Mag. , vol. 35, no. 12, Dec. 1997, pp. 12633. [65] S. Kaiser and J. Hagenauer, Multicarrier CDMA with Iterative Decoding and Soft-Interference Cancellation, Proc. IEEE GLOBECOM, Nov. 1997, pp. 610. [66] A. McCormick, P. Grant, and J. Thompson, A Novel Interference Cancellation Receiver for Uplink Multi-Carrier CDMA, Proc. IEEE Intl. Symp. Spread Spectrum Techniques and Apps., Sept. 2000, pp. 15558. [67] L. Fang and L. Milstein, Performance of Successive Interference Cancellation in Convolutionally Coded Multicarrier DS/CDMA Systems, IEEE Trans. Commun., vol. 49, Dec. 2001, pp. 206267. [68] J. G. Andrews and T. Meng, Performance of MCCDMA with Successive Interference Cancellation in a Multipath Fading Channel, IEEE Trans. Commun., vol. 52, no. 5, May 2004, pp. 81122. [69] C. Berrou, A. Glavieux, and P. Thitmajshima, Near Shannon Limit Error-correcting Coding and Decoding: Turbo-codes, Proc. IEEE ICC, Geneva, Switzerland, May 1993, pp. 106470. [70] M. L. Moher, An Iterative Multiuser Decoder for Nearcapacity Communications, IEEE Trans. Commun., vol. 46, July 1998, pp. 87080. [71] C. Douillard et al., Iterative Correction of Intersymbol Interference: Turbo-Equalisation, Euro. Trans. Telecommun., vol. 6, no. 5, Sept. 1995, pp. 50711. [72] J. Hagenauer, Forward Error Correcting for CDMA Systems, Proc. IEEE Intl. Symp. Spread Spectrum Techniques and Apps. , Mainz, Germany, Sept. 1996, pp. 56669.

[73] A. Nordio, M. A. Hernandez, and G. Caire, Low-complexity Turbo Equalization and Multiuser Decoding for TD-CDMA, IEEE Trans. Wireless Commun., vol. 3, no., Mar. 20042, pp. 45465. [74] H. Trigui and D. Slock, Cochannel Interference Cancellation within the Current GSM Standard, Proc. IEEE Intl. Conf. Univ. Pers. Commun., Oct. 1998. [75] H. Schoeneich and P. A. Hoeher, Single Antenna Interference Cancellation: Iterative Semi-Blind Algorithm and Performance Bound for Joint Maximum-Likelihood Interference Cancellation, Proc. IEEE GLOBECOM, San Francisco, CA, Dec. 2003, pp. 171620. [76] A. Mostafa et al., Single Antenna Interference Cancellation (SAIC) for GSM Networks, Proc. IEEE VTC, Oct. 2003. [77] A. Hafeez, D. Hui, and H. Arslan, Interference Cancellation for EDGE via Two-User Joint Demodulation, Proc. IEEE VTC, Oct. 2003. [78] K. I. Pedersen et al., Practical Implementation of Successive Interference Cancellation in DS/CDMA Systems, IEEE Conf. Univ. Pers. Commun., Sept. 1996, pp. 32125. [79] C. Teuscher, Low Power Receiver Design for Portable RF Applications: Design and Implementation of an Adaptive Multiuser Detector for an Indoor, Wideband CDMA Application, Ph.D. dissertation, UC Berkeley, Oct. 1998. [80] N. Correal, R. Buehrer, and B. Woerner, A DSP-based DS-CDMA Multiuser Receiver Employing Partial Parallel Interference Cancellation, IEEE JSAC , vol. 17, no. 4, Apr. 1999, pp. 61330. [81] G. Xu et al. , VLSI Implementation of the Multistage Detector for Next Generation Wideband CDMA Receivers, J. VLSI Sig. Proc., Kluwer, vol. 30, 2002, pp. 2133. [82] S. Rajagopal et al., Real-time Algorithms and Architectures for Multiuser Channel Estimation and Detection in Wireless Base-Station Receivers, IEEE Trans. Wireless Commun., vol. 1, no. 3, pp. 46879, July 2002.

Significant steps have been taken recently toward the realization of practical multiuser receivers, but more research and development is needed to make multiuser receivers practical for future standards.

BIOGRAPHIES
JEFFREY G. ANDREWS (jandrews@ece.utexas.edu) is an assistant professor of electrical and computer engineering at the University of Texas at Austin in the Wireless Networking and Communications Group (WNCG). He received a B.S. with high distinction from Harvey Mudd College in 1995, and his M.S. and Ph.D. in electrical engineering from Stanford University in 1999 and 2002, respectively. He helped develop CDMA systems as an engineer at Qualcomm from 1995 to 1997, and has served as a frequent consultant on communication systems. He is an associate editor for IEEE Transactions on Wireless Communications.

IEEE Wireless Communications April 2005

29