Wireless Systems

Future wireless communication systems will require significantly higher data rates and significantly reduced costs per transmitted bit. The demands on data rate, link quality, spectral efficiency, mobility, flexibility and complexity cannot be met with conventional wireless systems. This is accompanied by a paradigm shift from person-to-person (voice) to person-to-person interaction (multimedia) and person-to-machine communication (data services). It has been observed that service bandwidth, both offered to and required by the user, is doubling about every 12 months.

On the other hand, spectrum resources are limited and expensive. Since beyond 3G air interfaces need to be significantly better in terms of mobility versus data rate (peak data rates up to 100Mbps for wide area coverage and 1Gbps for local area coverage) and others, novel approaches will most likely not be backward compatible to the emerging 3G air interfaces. The latter will open room for new innovations, which could significantly increase system capacity.

The Wireless Systems Research Laboratory will address these demands by wireless signal processing and radio access techniques taking system imperfections and implementation complexity issues into account. The underpinning research areas involve channel measurements and modelling, channel estimation and channel prediction, optimised signal set and sequence designs, joint equalisation and decoding algorithms, modulation and coding, link adaptation techniques, and radio resource allocation.

The Institute will achieve this by:
          - Conducting fundamental research to enhance understanding of wireless signal processing and radio access
            with a focus on broadband wireless systems.
          - Performing cross layer design based on physical layer methodologies in a bottom up approach.
          - Implementing of selected hardware and software prototype functions.
          - Contributing technical papers about the research finding to international journals and conferences.
          - Delivering short courses and consulting to industry.
          - Contributing to pre-standardisation and standardisation bodies.

The R&D program will be directed towards the core problems of future broadband wireless systems and will address the broad generic areas of:
          - Channel measurement, modelling, estimation, and prediction;
          - Optimised signal and sequence set designs;
          - Joint design of transceiver functionality at both physical and data link layers;
          - Link adaptation for multimedia over wireless channels;
          - Multidimensional adaptation and radio resource allocation.

Major topics of our current research include:
Modulation, Coding, and Link Adaptation
          - Design of complex spreading codes using global and other optimisation methods.
          - Novel adaptive coding and modulation schemes along with ARQ protocols, code-combining, and
            soft-combining techniques for multimedia applications such as JPEG, MP3 audio, and video.
          - Link adaptation techniques using objective quality characterisation for selected multimedia
            applications. These are based on estimators for measures such as signal-to-noise ratio (SNR),
            segmental SNR, PSNR, ceptral distance, Bark spectral distortion, PSQM.
          - Intelligent transceiver structures and implementations of selected hardware and software
            prototype functions.

Joint Equalisation and Decoding
          - Novel theoretical concepts for joint equalization and decoding with application to broadband
            wireless indoor channels.
          - Channel estimation algorithms and synchronization techniques necessary for a successful operation
            of the proposed joint equalization and decoding.
          - Performance comparison and classification of the developed concepts in terms of link quality versus
            complexity of implementation.
          - Implements of selected hardware and software prototype functions.

Influence of Phase and Timing Offset on Soft Decoding Algorithms
          - Timing and phase offsets models for use with soft decoding algorithms.
          - Understanding of the influence of timing and phase synchronisation errors on trellis-based decoding
            algorithms.
          - Methods, which can improve performance of trellis-based soft decoding algorithms in the presence
            of imperfect synchronisation.
          - Generalise theory to systems, which can be modelled by factor graphs and described by the
            sum-product algorithm.
          - Implementations of selected hardware and software prototype functions.

Channel Estimation and Prediction Algorithms
          - Channel estimation and long term channel prediction algorithms, which can be utilized for adaptive
            coding and modulation schemes.
          - Hidden Markov models for time-varying channels with memory for estimation and prediction
            techniques within an iterative decoding framework.
          - Efficient trellis based channel estimators and predictors, e.g. sequential or decision directed
            techniques, which would offer easy implementation.
          - Implementations of selected hardware and software prototype functions.

Multidimensional Adaptation and Radio Resource Allocation
          - Contributions to the extension of adaptation in time domain to frequency and spatial domains.
          - Contributions to combining channel encoding with various multidimensional adaptation techniques.
          - Contributions to algorithms for radio resource allocation based on multidimensional adaptation
            techniques.
          - Proof of concept of selected software and/or hardware prototype functions such as combining
            encoding with multidimensional adaptation techniques.

Channel Measurements and Modelling

The focus of future wireless communication systems will be on universal coverage (in-door, out-door), complex arrangements of macro, micro and pico cells, higher density of users, and higher data rates. The measurement, analysis, and characterisation of the various wireless communication channels are a complex but necessary task that provides helpful information enabling realistic cell planning.

Activities

The work below shows that in rural Western Australia the new ITU-R P.1546 model provides better overall prediction of path loss compared to the traditional models such as the Hata model. Future work would be to take vegetation and foliage conditions into account when predicting path loss using ITU-R P.1546.

Measurement Equipment

Multidimensional Adaptation and Radio Resource Allocation
          - Simultaneous tracking of up to 8 pilot signals.
          - High signal sample rates enables tracking of fast signal fades.
          - Display of selected parameters in real-time.
          - Display of map with basestation and mobile antennao optional logging of all data parameters.
          - GPS and weather stationo Advanced propagation prediction tool.

Transmitter System.	DSP FPGA Board.	Wideband CDMA Radio.
Graphical User Interface.	Advanced Prediction Tool.

Measurement Campaign

Terrain Example, Kellerberrin Base Station.

Channel Modelling

Digital Elevation Data.	Cumulative Distribution Function.	Semi-Terrain based ITU-R Model.	Empirical Models and Linear Regression.

^ to top of page

Optimised Non-Binary Sequence Design

Non-binary sequences have gained increased attention with respect to applications in wireless communication systems. This is mainly because large sets of non-binary sequences with good correlation properties exist. This finding is in particular useful for direct sequence code-division multiple-access (DS-CDMA) systems where several users simultaneously share the available bandwidth of a common transmission channel. Therefore, sets of sequences with good cross-correlation properties are desired to limit mutual interference among the users. This in turn will improve the error performance and increase the overall capacity of the system. On the other hand, sequences with good auto-correlation properties are desired for synchronisation purposes. Since good cross-correlation properties come on the expense of the auto-correlation properties, a suitable trade-off has to be found depending on the demands of a respective application.

Activities

A modified bridging method is proposed to minimise average mean-square aperiodic cross-correlation with out-of-phase average mean-square aperiodic auto-correlation being allowed to lie within a fixed region. This becomes a complex optimisation problem with a non-linear cost function and a non-linear constraint. This approach is applied to the design of complex spreading sequences.

Design for Optimum Cross-Correlation.	Design for Optimum Auto-Correlation.	Design for Optimum Cross-Correlation.	Design for Optimum Auto-Correlation.

Symbol Error Rate

Design for Optimum Cross-Correlation.	Design for Optimum Auto-Correlation.

^ to top of page

Joint Equalisation and Decoding

The use of digital transmission techniques in future broadband wireless systems demand for use of adaptive equalization in order to suppress the severe intersymbol interference (ISI), which is caused by the frequency-selective behavior of the radio link. In addition, error control coding is a necessity in a broadband wireless system as it is with mobile radio systems in general and this is to support a certain level of quality of service in terms of reliability.

First ideas regarding a combination of equalization with error control coding range back as early as the mid 1980's. Joint adaptive equaliser and decoder structures, for example, may use two sets of decision feedback equalisers (DFE's) along with a forward error control (FEC) scheme. The FEC function processes channel state information and provides reliability estimates of the received data, which then can be used for the adaptation algorithm in the equaliser.

Activities

The current focus in this research area is on single carrier frequency domain equaliser structures and joint equalisation and decoding using block codes.

This includes a frequency domain equivalent implementation of a block modified CMA (FD-MCMA). The proposed FD-MCMA is based on the modified CMA (MCMA) which confines the phase ambiguities of the conventional CMA to a 90° phase shift. Then, a block MCMA which updates equaliser coefficients only once in each block is proposed. The chosen cost function forces all output samples close to a given constant modulus. This allows the derivation of a FD-MCMA, which is equivalent to the block MCMA. The frequency domain approach significantly reduces the computational complexity associated with the conventional MCMA. Simulation results show that a normalized FD-MCMA converges much faster than the FD-MCMA and the MCMA. Especially for long equalisers, the normalized FD-MCMA offers superior convergence speed compared to the MCMA and provides excellent accuracy.

Frequency Domain Modified Constant Modulus Algorithm Equaliser

Left: Hiperlan II Channel. Right: Normalised FD-MCMA Equaliser.

Convergence Comparison for Different Equaliser Length

L = 64 and L = 128.

Joint Equalisation and Decoding Using Block Codes

^ to top of page

Link Adaptation for Multimedia over Wireless Channels

Reliable communication over time varying mobile radio channels has been obtained with the use of link adaptation techniques. Explicitly with link adaptation, the transmission parameters are modified to compensate for the variations in channel conditions, which may be estimated based on feedback from the receiver. Among the many link adaptation techniques, adaptive modulation and coding, power control, selection diversity, multiuser diversity and hybrid ARQ (H-ARQ) are commonly used.

The development of efficient, yet powerful error control coding schemes for the wireless channel is important for the provision of mobile multimedia applications. For such applications, the user perceived quality might be severely degraded by only a few errors in the compressed data stream. Traditional performance metrics for error control coding schemes, such as the post-decoding bit error rate (BER), often bear little relevance to the user perceived quality. The powerful source compression techniques employed for multimedia files often lead to increased error susceptibility. In particular, corruption of the file headers may result in the file being unrecognisable to the application involved. Furthermore, the effects of only a few bit errors in the compressed data stream may propagate through a substantial portion of the decompressed data delivered to the target application.

Activities

Link adaptation to the transmission of Joint Photographic Experts Group (JPEG) images over a correlated Rayleigh flat fading channel is implemented based on hybrid Automatic repeat ReQest (ARQ) scheme using a soft-combining algorithm. With soft-combining techniques, the information obtained with each decoding attempt is preserved and incorporated with the retransmitted copies of the codeword. As far as soft-combining of received words is concerned, it turns out to be beneficial to exploit iterative decoding algorithms based on maximum a-posteriori (MAP) decoding.

Image and Audio Transmission over Fading Channels

Left: Transmission/Receiver Structure. Right: Retransmission Strategy.

Effect of Soft-Combining on Transmitted Image 'Lena' over Rayleigh Fading

Left: One Retransmission. Right: Four Retransmissions.

Effect of Soft-Combining on Transmitted Audio File

Red Hot Chilli Peppers - Get on top.mp3
          - 128 kBit
          - 44,100 Hz
          - MPEG 1-Layer III
          - 10 Seconds
          - 168 kBytes

	Original Sample
	Uncoded over Rayleigh Fading
	Soft-combining over Rayleigh Fading

^ to top of page

Multidimensional Adaptation and Radio Resource Allocation

Adaptation in the context of radio transmission technologies aims at optimising the performance of a single communication link. Additional gains in link and system level performance in terms of reliability, capacity improvements and radio resource usage can be expected by extending conventional adaptation in time domain to the dimensions of code, frequency and spatial domains. This includes the combination of channel encoding with various diversity and multidimensional adaptation techniques. Similar extensions may be applied to provide multidimensional radio resource allocation in the code, time, frequency, and space domain. No application may need always the highest bit rates but the system may need it in order to serve many high bit rate users simultaneously, maximize capacity and minimize latencies.

Thus, there may be an optimum bandwidth, which would maximize the spectral efficiency of a system and may be achieved involving several dimensions.

Activities

A current research area considers a WLAN, which consists of Bluetooth Access Points (APs) that offer services to ad-hoc users. It inherently provides both voice and data connectivity. As Bluetooth offers the establishment of Synchronous Circuit-Oriented (SCO) traffic, the concept under study facilitates a low-cost alternative to the deployment of expensive indoor cellular base stations.

In contrast to standard WLANs, voice connections receive a premium service through guaranteed bandwidth in reserved channels, while data uses the remaining capacity. Since such a WLAN may comprise of a substantial number of APs, the quest for efficient strategies on how to allocate SCO channels to the APs arises.

Furthermore, a single AP can serve a maximum of only three SCO channels and hence may not be capable of satisfying the voice traffic demands in hot spot areas. Therefore, we propose to deploy Multi APs for those particular locations instead of only a single AP.

Multi AP Strategies for SCO Traffic in a Bluetooth based Wireless LAN

Topology of the Considered Network.

A Multi AP Consisting of Voice Units (VU) and Data Units (DU)

	State Diagram (a) and State Reservation Sequence (b) for a Multi AP Strategy.

Performance Results

^ to top of page