Multiple Access, Modulation and Transceivers

Figure 1 : Transmitter diagram of the frequency-hopping multicarrier DS-CDMA system.

The transmitter of the proposed FH/MC DS-CDMA scheme is depicted in Figure 1. Each subcarrier of the K users in the
system is assigned a PN sequence, which produce spread, wideband signals. In the figure, C(Q,Uk) represents a constant-weight code of user k with $Uk$ number of `1's and (Q-Uk) number of `0's, hence, the weight of C(Q,Uk) is $Uk$. This code is read from a so-called constant-weight code book, which represents the frequency-hopping patterns. The constant-weight code plays two different roles. Its first role is that its weight - namely $Uk$ - determines the number of subcarriers involved, while its second function is that the positions of the $Uk$ number of binary `1's determines the selection of a set of $Uk$ number of subcarrier frequencies from the Q number of outputs of the frequency synthesizer. High-rate and variable rate transmissions can be implemented by employing a different number of subcarriers. At the transmitter of the k-th user in Figure 1 the bit stream is first serial-to-parallel (S-P) converted, yielding Uk parallel streams, which is controlled by the constant-weight code C(Q,Uk). After S-P conversion each stream is direct-sequence spread, in order to form the spread, wideband signal. These spread signals then modulate their corresponding subcarriers, and finally are summed, in order to form the transmitted signal sk(t).

The objective of this research is to investigate the relevant techniques in the context of FH/MC DS-CDMA, including modulation/demodulation, error-control, synchronisation, equalisation, multiuser interference suppression, adaptive detection, etc.
[top]

Figure 2 : MC-CDMA and DS-CDMA use the whole bandwidth for a symbol to exploit frequency diversity, while OFDM uses single subcarrier for it. MC-CDMA and OFDM are resilent to inter-symbol-interference and spectrally efficient. MC-CDMA posesses both merits of DS-CDMA and OFDM.

[top]

• 384 kbits/s transmission rate for full area coverage
• 2 Mbits/s transmission rate for local area coverage

In order to support a variety of multimedia applications, variable bit rate and packet transmission capabilities are essential. There
is also a need for flexibility such that different services could be multiplexed within the same environment.

The standardization process for 3G systems commenced with a number of proposals from all over the world, but the standard has
now been narrowed down to four major proposals. The common feature in these proposals is the employment of Wideband CDMA (W-CDMA) as the air interface. In W-CDMA, the increased bandwidth of at least 5 MHz facilitates a variety of performance advantages over previous cellular systems. With the increase in bandwidth, the target bit rate of 384 kbits/s is achievable while ensuring a reasonable user capacity. The paucity of available spectrum calls for a high area spectral efficiency, which can be provided by CDMA. Furthermore, the wide bandwidth enables higher diversity gains to be obtained through the higher-order multipath diversity
schemes facilitated by the widely spread W-CDMA signal. This is provided that all the resolvable multipath components can be
combined due to the higher tolerable receiver complexity. In addition to this, further performance improvements can be obtained
through a variety of techniques, such as the application of antenna arrays, multiuser detection or space-time coding. The
employment of CDMA enables multi-rate services to be provided upon involving various methods, such as multi-code transmission,
multiple spreading factor transmission and modulation mode multiplexing. Some of the key properties emphasized in W-CDMA are:

• Improved performance over second generation systems including improved user capacity and improved quality coverage
• High service flexibility including the support of a wide range of services having bit rates up to 2 Mbits/s
• A fast and efficient packet access scheme
• Support of evolutionary technologies, such as adaptive antenna arrays and multi-user detection

Direct Sequence Code Division Multiple Access (DS-CDMA) is interference-limited due to the multiple access interference (MAI)
generated by the users transmitting within the same bandwidth simultaneously. The signals from the users are separated by means of
spreading sequences that are unique to each user. These spreading sequences are usually non-orthogonal. Even if they are orthogonal, the asynchronous transmission or the time-varying nature of the mobile radio channel may partially destroy this orthogonality. The non-orthogonal nature of the codes results in MAI, which degrades the performance of the system. The frequency selective mobile radio channel also gives rise to inter-symbol interference (ISI) due to multi-path propagation. This is exacerbated by the fact that the mobile radio channel is time-varying. A class of CDMA receivers known as multiuser receivers exploit the available information about the spreading sequences and mobile channel impulse responses of all the CDMA users in order to improve the performance of the CDMA users. These multiuser receivers include joint detection (JD) receivers, interference cancellation (IC) receivers, tree-search type algorithms and iterative receiver schemes. Figure 3 shows a general classification of the various types of multiuser receivers.

Figure 3 : Classification of CDMA detectors

Mobile radio signals are subject to propagation path loss as well as slow fading and fast fading. Due to the nature of the fading
channel, transmission errors occur in bursts when the channel exhibits deep fades or when there is a sudden surge of multiple
access interference (MAI) or inter-symbol interference (ISI). Adaptive-rate CDMA techniques can be used to overcome this
phenomenon, where the information rate is varied in accordance with the channel quality. The information rate is chosen
accordingly, in order to provide the best trade-off between the BER and throughput performance for a given application. The signal
to interference plus noise ratio (SINR) at the output of the JD receiver is derived and this parameter is used as the criterion
for adapting the information rate. There are various methods of varying the information rate are considered, including Adaptive
Quadrature Amplitude Modulated (AQAM) and the Variable Spreading Factor (VSF) scheme. AQAM is an adaptive-rate technique, whereby the data modulation mode is chosen according to some criterion related to the channel quality. On the other hand, in VSF transmission, the information rate is varied by adapting the spreading factor of the CDMA codes used, while keeping the chip rate constant. Figure 4 shows the stylized amplitude variation in a fading channel and the switching of the
modulation modes in a four-mode AQAM system, where the performance degrades but the throughput increases when switching from Mode 1 to 4.

Figure 4 : Basic concept of a four-mode AQAM transmission in a narrowband channel. The variation of the modulation mode follows the fading variation of the channel over time.

[top]

Ubiquitous antenna based systems: According to this concept ahigh number of antennas are distributed over the service area. Each antenna and the base station are connected via optical fibre,conveying the radio signals directly over the optical-fibre cable. The
ubiquitous antennae not only reduce the required number of radio basestations, which reduces the cost of the overall system, but also allow us to employ sophisticated signal processing techniques such as diversity reception, interference cancellation, and multi-user
detection. Currently Minoru is studying simple interference cancellation algorithms for the ubiquitous antenna.

Source matched adaptive transmission: The required bit rate, theacceptable error rate and delay of a wireless communications link
varies depending on the type of information transmitted. Upon varyingthe transmission power, the forward error correction code and the modulation format, we can improve the bandwidth efficiency of wirelesssystems. The optimum algorithm is sought that can be used to control the transmission schemes adaptively according to the type of information.
[top]

Burst-by-burst Adaptive Quadrature Amplitude Modulation (AQAM) is studied in conjunction with adaptive beamforming, diversity and interference cancellation in multiuser environments. The system's performance is further improved upon invoking various error correction techniques, in oarticular space-time coding.
[top]

Figure 5 : Block diagram of an OFDM based transmission system.

The problem of inter-symbol interference can be elegantly addressed by employing OFDM ( Orthogonal Frequency Division Multiplexing ), which is a multi-carrier transmission scheme conceived back in the 60's. With the availability of high-performance signal-processing devices it has been re-discovered since the operation of multi-carrier modulation can be efficiently implemented by means of a Fast Fourier Transform (FFT). As opposed to single-carrier transmission, OFDM seeks to avoid inter-symbol interference by appending a guard interval to each FFT output block - before signal interpolation and up-conversion to the HF ( High Frequency ) stage - at the cost of a slight reduction in bandwidth efficiency. The problem of channel equalization at the receiver is simplified substantially. In multiuser environments of higher mobility, the OFDM scheme suffers from inter-subcarrier interference, which is due to a loss of orthogonality between different sub-channels. OFDM is well established in a variety of services such Digital Audio Broadcasting, Asynchronous Digital Subscriber Lines and the new Hyperlan II standard.

Matthias' research is mainly concerned with the development of signal processing algorithms for tailoring OFDM to the needs of a
multi-cellular environment, of which a very important aspect is the signal separation of simultaneously transmitting users. This
is could be achieved with multiple-antenna assisted reception at the basestation, which requires further investigation in
conjunction with OFDM. Related aspects are channel parameter estimation as well as synchronization mechanisms.
[top]

The SVM approach to the design of decision feedback equalisers (DFE) was found to outperform the minimum mean square error (MMSE) approach, which is broadly used in practice.

Although the advantages of SVM are apparent, its application inCDMA multiuser detection is still in its infancy. Apart from its
conceptual simplicity, SVMs promise reduction in complexity andimprovement in performance compared to the existing
state-of-the-art solutions.
[top]

In most cellular systems, the transmitted information symbols experience both amplitude and phase distortion as a result of the
dispersive and time varying mobile propagation channel. In a system which transmits a sequence of pulse-shaped information
symbols, the time domain full response signalling pulses are smeared by the hostile dispersive mobile channel, resulting in
intersymbol interference. However at the receiver, the distorted signal can be restored by utilizing an equalizer in order
to recover the transmitted information symbols. Consequently the transmission quality can be enhanced by utilizing an equalizer. This process is exemplified by Figure 6 which displays the transmitted, channel distorted and equalized information
symbols.

The transmitted information symbols also experience rapid signal fluctuation as a result of the mobile propagation channel, which is termed as fast fading. Consequently a time varying channel quality is produced, which can be exploited in order to increase the transmission throughput. This can be achieved by using Adaptive Quadrature Amplitude Modulation (AQAM) whereby modulation modes with a higher information throughput can be used, when the channel quality is favourable. Conversely, a more robust modulation mode with a lower information throughput is utilized, when a degraded channel quality is encountered. Consequently, on average, by adapting the modulation modes we can increase the information throughput, while maintaining a certain target transmission quality.

a) Channel distorted signal

b) Equalized signal

Figure 6 : The effect of channel distortion on the transmitted symbols and the corresponding equalized symbols.

[top]

a) One iteration

b) Two iterations

c) Four iterations

d) Six iterations

Figure 7 : Reliability of the decoder's so-called Logarithmic Likelihood Ratio after one, two, four and six turbo equalisation iterations. The reliability is expressed as the product of the LLR values produced by the decoder and its corresponding transmitted source bit.

Brief Overview of the Turbo Equalisation Principles: The basic principle of turbo equalisation is that information, which reflects the reliability of the estimated channel encoded bits, is exchanged between the channel equaliser and channel decoder iteratively. When a certain iteration termination criterion is satisfied, the information related to the source bit --- instead of the encoded bit --- is determined. This information is typically in the form of the Log Likelihood Ratio (LLR), which is the natural logarithm of the probabilities that the information transmitted assumes its two possible values i.e. +1 and $-1$. With each successive iteration, the channel equaliser and channel decoder provides more reliable LLR values, which are associated to the transmitted source bits and the encoded bits. In Figure 7 the reliability of the LLR values --- expressed as the product of the decoder LLR values with the corresponding source bits --- is plotted against the source bit's position. A large positive reliability value indicates a high-confidence LLR value, whereas a negative value represents an erroneous decision. It is observed that in the first iteration of Figure 7.a there was a large number of low reliability LLR values. The negative reliability values in Figure 7.a indicate that decision errors have been made. In the second turbo equalisation iteration of Figure 7.b the number of low-confidence LLR values was reduced as compared to the first iteration. With increasing turbo equalisation iterations --- for example after six turbo equalisation iterations, as shown in Figure7.d --- the number of low-confidence LLR values was significantly reduced, hence justifying the improved BER performance obtained with an increasing number of turbo equalisation iterations.

Current Research: Iterative multi-user interference detection employing the turbo equalisation strategy is an attractive research topic. Recent work utilising turbo equalisation has been successful in mitigating the interference inflicted by other Code Division Multiple Access (CDMA) users, such that the near single-user performance is achieved.

Turbo equalisation for Trellis-Coded Modulation is a further topic investigated in the recent literature in order to improve the
spectral efficiency of the system --- which is one of the key objectives of mobile radio research.
[top]

Mong Suan's research is based on using an ANN structure referred to as the Radial Basis Function (RBF) network for channel equalisation. The RBF based equaliser design is investigated in the context of various Quadrature Amplitude Modulation (QAM) schemes in conjunction with turbo channel coding and iterative decoding / equalisation techniques. Since the schemes investigated may become complex, computational complexity reduction methods are also investigated. The received and transmitted 4QAM signal constellation along with the error-free decided constellation is seen in Figure 9.

a) Anatomy of a typical neuron

b) An artificial neuron

Figure 8 : Comparison between biological and artificial neurons

a) Channel distorted signal

b) Equalized signal

Figure 9 : The effect of channel distortion on the transmitted 4QAM symbols and the corresponding equalised symbols.