Exploiting MIMO Technology for Optimal Performance

André Bourdoux and Koen Snoeckx
Mar 31, 2005 (5:14 AM)

Multiple antenna schemes are becoming a key part in almost every new wireless standard. For example, the UMTS/W-CDMA standard uses transmit diversity to improve the link quality in outdoor environments. Next-generation 802.11n wireless local area networks (WLANs) will likely include spatial division multiplex to increase the peak throughput and cell capacity. Emerging 60-GHz WPANs will probably use transmit and receive beamforming to boost link quality. Although many recent and emerging standards include, or will likely include, multiple-input multiple-output (MIMO) techniques (Figure 1 below), it is important to recognize that there is a wide variety of means to exploit the advantages of multiple transmit and/or receive antennas for indoor systems. Overview of General MIMO Concepts
The performance of wireless communications can be dramatically improved when using multiple antennas [1], commonly referred to as MIMO or multiple-input, multiple-output. For instance, multiple antennas can be used to increase the antenna gain by beamforming, to provide diversity gain through some form of antenna combining, to increase the data rate by spatial multiplexing, or to suppress spatial interference by null steering. Two of these techniques—spatial multiplexing and spatial diversity—can be applied in an indoor environment (Figure 2 below).

Spatial Multiplexing: Higher Throughput
Spatial division multiplexing (SDM) occurs when several streams are transmitted simultaneously from the transmitter to the receiver, both equipped with multiple antennas. It is important to note that the multiple streams are transmitted simultaneously (i.e. at the same time and frequency). Transmitter and/or receiver processing make sure that they are separated in the spatial (or antenna) dimension.

Click here for Figure 1

Figure 1: Overview of emerging wireless standards and the incorporation of MIMO techniques.

Figure 2: Spatial multiplexing and spatial diversity are MIMO principles that can be used to improve an indoor wireless system. Top: Spatial division multiplexing (2 streams are transmitted, receiver processing shown) Middle: Spatial diversity by space-time block coding (Transmit diversity, 1 stream is transmitted, multiple antennas at the receiver are preferable, but not necessary) Bottom: Spatial diversity by diversity combining (Receive diversity shown, 1 stream is transmitted).

The net advantage of SDM is to boost the spectral efficiency of the transmission. Indeed, spatial multiplexing allows increasing the bit rate without consuming more time or frequency resources and without increasing the total transmit power. The capacity of a wireless link increases linearly with the minimum amount of transmit or receive antennas. For example, a system with three antennas at both sides can transmit roughly three times as much data than a single antenna system. This is a tremendous capacity increase, especially given the scarce spectral resources below 10 GHz.

It is clear that SDM can only be fully exploited if the number of antennas is, at both sides, equal to or greater than the number of parallel streams. SDM is usually employed in situation where the signal-to-noise ratio is relatively high. To increase the spectral efficiency, SDM relies on low correlation between the antennas of the arrays, which occurs when rich scattering is experienced near the antenna array. Because of this, it converts what was in the past perceived as a source of degradation into an advantage.

In addition to achieving higher throughput by sending several streams to one multi-antenna terminal, parallel spatial transmission of streams can also be used to increase cell capacity by transmitting several streams to different single antenna terminals (known as spatial division multiple access or SDMA).

In terms of signal processing, the simplest form of SDM is the linear spatial filter, which is basically a multiplication of a matrix (the spatial filter) with the vector of transmitted symbols or received signals. The spatial filter is then typically the inverse of the channel matrix or a minimum mean-squared error (MMSE) filter. The dimensions of this matrix are small and are equal to the number of transmit and receive antennas. Without discussing them in detail, it must be said that more complex forms of spatial processing exist, such as successive interference cancellation, sphere detection, and maximum likelihood detection.

Spatial Diversity: Better Signal Quality
The basic principle of diversity is to use different "channels" to convey the same information unit from the transmitter to the receiver. This means that, at the end, only one information stream is exchanged, but with better signal quality. The application of diversity is especially useful when the different channels that are used fade in a statistically independent fashion, or, in other words, when the probability that all channels are bad at the same time is low. In this way, the information can be recovered from the channel(s) where the signal-to-noise ratio is the best.

Spatial diversity can be accomplished by transmitting the same symbols on different transmit antennas or receiving the same signal with several receive antennas. Hence, spatial diversity is useful when the link budget must be improved in order to increase the communication range or in order to reduce the transmit power.

Two techniques can be used to accomplish spatial diversity. The first is transmitter coding over both the time and spatial dimensions (e.g. space-time block coding or STBC). In this case, multiple antennas are needed at the transmitter. At the receiver, both single and multiple antennas are possible. Another technique, known as diversity combining, requires combining in the antenna dimension. This approach uses multiple transmit or receive antennas, providing transmit or receive diversity, respectively.

The signal processing involved in spatial diversity is relatively simple. For space-time block coding, it consists of a special mapping of the symbols on the transmit antennas and a simple matrix multiplication at the receive side. For diversity combining, the signals on the various antennas are multiplied with a complex coefficient. Hence, diversity combining can actually be seen as a form of beamforming where the signal transmitted or received by the antenna array is "weighted" in phase (and sometimes also in amplitude).

An important asset of spatial processing is that it can be combined with orthogonal frequency division multiplexing (OFDM). OFDM, or, more generally, multi-carrier transmission, is used in many new standards (e.g. WiFi, WiMAX). With the same caution of use as for single antenna OFDM transmission (i.e. correct synchronization and cycle prefix longer than the channel delay spread), the spatial processing can be applied per sub-carrier (frequency-domain processing), thereby decoupling the spatial processing from the equalization of the frequency-selective channel and leading to low complexity solutions.

Smart-MIMO in High-Throughput WLANs
In both concepts described above, the MIMO processing can be performed at either the transmitter or the receiver side, or at both sides. Naturally, the power consumption and complexity are the highest where the spatial processing is carried out. This consideration led the Belgian research center IMEC to propose an asymmetric concept for MIMO techniques in WLANs. Since the spatial processing is preferably performed at the access point, where complexity and power consumption pose fewer challenges, transmit processing is used in the downlink and receive processing is used in the uplink. The approach supports SDM, SDMA, maximum ratio combining (MRC) and STBC in both directions, while minimizing the impact on the terminal.

Figure 3: Various combinations of MIMO techniques are shown with different transmit or receive processing schemes to minimize terminal complexity and power consumption.

This "smart-MIMO" approach is a flexible concept whereby the transmission scheme is chosen to obtain an optimal performance in a given situation, taking into account parameters such as user quality-of-service (QoS) requirements, user profile (e.g. autonomy), signal-to-noise ratio, and channel quality.

Smart MIMO systems offer various opportunities to drastically upgrade the performance of current and future WLAN systems. Uniquely, it offers benefits in scenarios whereby the terminals, the access point, or ultimately both, are upgraded with multiple antennas:

Flexible MIMO processing enables upgrading the system performance under dynamic environments and requirements (rate, range, robustness, multi-user capacity). The MIMO processing can be optimally distributed between the transmitter and receiver (e.g. it can be concentrated at the access point if the latter is power plugged).
The combination of receive and transmit processing solutions can improve communication on links where only one side is 'MIMO aware.' This includes future scenarios where single-antenna user terminals could benefit from a multiple-antenna enhanced access point, and scenarios where new multiple antenna user terminals achieve better performances even when communicating with older single antenna access points or terminals.
The SDM(A) feature enables creating 'parallel spatial pipes' next to 'higher capacity pipes' through spatial multiplexing and/or diversity. The parallel pipes offer clear advantages in situations where one or more users are slowing down the total cell traffic dramatically (because they experience bad channel characteristics and/or do not have MIMO possibilities).

One of the key and distinctive characteristics of a flexible MIMO scheme is that it includes transmitter spatial processing (for the downlink SDM, SDMA and MRC schemes). The required channel knowledge at the transmit side is based on channel estimation on the reverse link, as such minimizing the elapsed time between the estimation and the actual pre-compensation. In this case, channel reciprocity is mandatory.

A detailed study has shown that the propagation channel is reciprocal but the non-reciprocity of the access point transceivers has a detrimental effect on downlink performance. A special calibration procedure has been developed that compensates for non-reciprocity in the front-ends [2]. The degradation caused by non-reciprocity has been assessed by simulation. In addition, several solutions that allow alleviating transceiver non-reciprocity have been conceived and analyzed. Among these, a scheme involving an auxiliary transceiver has been implemented. It allows measuring separately the transmitter and receiver frequency-responses in the access point with high accuracy (better than 1deg.), so the downlink pre-filter performances are not degraded by unknown phase or amplitude errors (Figure 4). The calibration procedure has been shown to achieve excellent performances on a wireless prototype, while having the practical advantages of working at the access point side only.

Figure 4: Analyzing a real-time 5-GHz WLAN MIMO-OFDM prototype.

Cross Layer Optimization
Exploiting the advantages of multiple antennas is one thing, but avoiding the possible drawbacks might appear equally important in order to successfully introduce MIMO techniques to the market. When the number of antennas is increased, it is important to avoid an increase in the overall energy budget. In an intelligent environment, the "anything, anywhere, anytime" paradigm requires a dynamic trade-off between performance factors and energy consumption. This is where traditional approaches and most modern procedures for cross-layer optimization reach their limitation: they do not take into account the user level, namely the application and the terminal platform.

Performance metrics for quality of experience (QoE) are derived from the complete communication stack, going up to the application layer, while the actual energy drained from the battery embodies the energy requirement. Both, however, are dependent on many factors, such as propagation conditions, physical layer, protocol stack, application, and implementation aspects, and this overtaxes traditional approaches to system optimization.

In addition, traditional optimization strategies fail to cope with the dynamics of wireless communication. Because of their constantly changing requirements, profiles and channel condition, wireless links no longer have a single optimal working point. Rather, an energy-efficient system must constantly adapt its characteristics to the changing usage conditions in order to be able to provide the performance required by the user at the minimum energy level.

A good approach for global QoE can be based on a controller that chooses a set of system parameters to achieve a trade-off between transmission and application performance and energy consumption. System parameters that can be influenced include those used in the setting of the physical layer (such as front-ends, physical layer mode including the MIMO mode, and number of decoder iterations), the MAC layer (such as burst length) and multimedia functionality (such as resolution, frames per second, and interactivity). The solution for this kind of QoE control consists in finding a cross-layer trade-off between energy and performance at design-time, from which the information is used for system adaptation at runtime. This approach is implemented in more detail in three steps [3] Figure 5:

Modeling the system:Define performance data, establish relevant manipulated variables, and then model performance/energy behavior as a function of different configurations.
Reducing the model (at design-time): Prune the configuration space in a way that only configurations that lead to minimum energy consumption remain.
Configuring the system (at runtime): Identify channel characteristics and analyze user requirements, then select the appropriate hyper-surface, on which an optimum point is found for these requirements, and finally set the parameters according to the assigned configuration vector.

Figure 5: Three-step design-time run-time cross layer optimization approach.

When applied to wireless local networks, this approach's initial simulation results demonstrated up to seven times better energy utilization.

MIMO-OFDM Wireless Prototype on the Air To implement the concepts and acquire experimental performance data on the techniques described above, a smart-MIMO prototype was built [4]. It comprises an access point with two antennas, two user terminals with one antenna each and a terminal with two antennas. The baseband and MAC processors are implemented in field-programmable gate arrays (FPGAs). A special front-end for MIMO application operates in the 5-GHz range and includes the calibration techniques to achieve reciprocity.

The prototype demonstrated that cell capacity could be increased from 54 Mbits/s to more than 108 Mbits/s due to spatial pre-processing and that the reciprocity assumption in slowly varying time-domain-duplex (TDD) systems is valid. So, users with conventional terminals with only one antenna can also simultaneously receive a signal in the same frequency band and time slot. For an N-antenna access point, this multiplies the cell capacity N-fold. The fact that this concept does not require changes in the user terminal protects existing investments, which is an important asset in the context of fast implementation on the market. Furthermore, the same flexible procedure is suitable for transmission to terminals with two antennas in order to increase the data rate to 108 Mbits/s. Finally, the coverage range can increase by factor 2 to 5 thanks to spatial diversity (STBC or MRC).

This prototype is conceptually very close to the IEEE 802.11n specifications, which target high-throughput WLANs with more than 100 Mbits/s, thanks to the joint use of OFDM and SDM. For further experimental research, the prototype set-up is currently being upgraded to also incorporate SDM, STBC, and MRC at the receive side. A new miniaturized 5-GHz front-end, which is tailored for MIMO operations, will be further integrated in a system-in-a-package.

Future Possibilities, Research Focus
This smart MIMO system shows how multiple antennas can be used to meet the actual user need in an OFDM-based WLAN context. Further performance upgrades are possible through combined processing on the transmit and receive side, as well as with the SDMA separation of MIMO terminals, whereby transmitters with multiple antennas communicate with various multiple-antenna receivers.

It is generally accepted that MIMO systems will play a crucial role in future wireless systems. The combination of smart MIMO technology and a flexible air interface will allow multi-mode, multimedia terminals that can seamlessly and power-efficiently roam between different networks. Ultimately, multi-hop networks are envisaged where terminals can be used as relays between an access point and distant terminals, thereby globally reducing the power consumption and increasing ranges. MIMO techniques might play a key role in these future networks where the conventional distinction between access point and terminal will blur.

In the end, multiple antenna techniques are key to boosting the performances of modern wireless systems. Spatial multiplexing increases system throughput without consuming scarce frequency spectrum, and spatial diversity makes the link more robust. An optimal approach is to flexibly exploit the various MIMO modes by taking advantage of transmit or receive processing in order to concentrate the complexity and power consumption at the access point. Once combined with optimization techniques in other layers of future wireless multi-mode multimedia terminals (cross-layer optimization), this flexible approach will globally take into account the user requirements, the platform capabilities and the various physical layer modes.

References

Patrick Vandenameele, Liesbet Van der Perre, Marc Engels, Bert Gyselinckx, Hugo De Man, 'A Combined OFDM/SDMA Approach, Journal on Selected Areas in Communications, Vol. 18, No, 11, November 2000
Bourdoux Andr, Come Boris. 'Non-reciprocal Transceivers in OFDM/SDMA Systems: Impact and Mitigation', RAWCON 2003, Boston, August 2003
Bougard, B.; Lenoir, G.; Eberle, W.; Catthoor, F. and Dehaene, W., 'A New Approach to Dynamically Trade-off Performance and Energy Consumption in Wireless Communication Systems', IEEE Symposium on Signal Processing Systems (SIPS03), Seoul, Korea, August
www.imec.be/wireless/picard

About the Authors
André Bourdoux is senior researcher at the Wireless Group of IMEC's division of Design Technology for Integrated Information and Communication Systems (DESICS). He can be reached at andre.bourdoux@imec.be .

Koen Snoeckx is scientific editor at IMEC and can be reached at koen.snoeckx@imec.be.

Click here to read more ...