GB2434065A

GB2434065A - Variable bandwidth transmitter and receiver

Info

Publication number: GB2434065A
Application number: GB0600346A
Authority: GB
Inventors: Steve Carl Jamieson Parker; Justin Coon
Original assignee: Toshiba Research Europe Ltd
Current assignee: Toshiba Europe Ltd
Priority date: 2006-01-09
Filing date: 2006-01-09
Publication date: 2007-07-11
Anticipated expiration: 2026-01-09
Also published as: GB0600346D0; GB2434065B

Abstract

The present invention provides a wireless transmitter for transmitting time divided multiple carrier frequency symbols, the transmitter comprising: means for modulating one or more carrier frequencies in a time divided multiple carrier frequency symbol dependent on a respective received information symbol; wherein the number of carrier frequencies modulated in the same time divided multiple carrier frequency symbol by the respective information symbol is dynamically adjustable. The transmitter may comprise means for replicating the information symbols according to a replication factor. The number of carrier frequencies modulated may be dependent on a signal quality parameter (SNR, BER, SJNR etc). The time divided multiple carrier frequency symbols may be OFDM symbols and the modulating means may be an Inverse Fast Fourier Transform (IFFT) processor.

Description

1 2434065 Variable Bandwidth Transmitter and Receiver The present

invention relates to wireless communications, and especially though not exclusively to ultra wideband (UWB) communications.

A well known problem with multiple access wireless communications is the so-called near-far problem that refers to the situation in which a receiver is simultaneously receiving signals from both a near-by transmitter and a far-off transmitter. The sum or combination of the received signals is amplified before being digitised and processed.

However, the level of amplification or gain is dependent on the combined signal power.

Hence, because it is likely that the automatic gain control (AGC) of a receiver is set according to the power of the strongest signal, the weaker signal will be poorly resolved after quantisation by the analogue-to-digital converter (ADC) of the receiver. This problem is typically alleviated in cellular networks by power control: the near device may be instructed to reduce its power level and the far device may be instructed to increase its power, so that balanced receive power is achieved for the two devices. A similar problem of poor resolution of a weaker signal can also occur when interference is received superimposed on a wanted signal. In this case, the transmitter of the weaker signal may be instructed to increase its transmit power.

Ultra-wideband (UWB) is a relatively newly approved form of commercial wireless communications. It is characterised by a signal's transmission power being spread over a wide bandwidth such that the power spectral density (PSD) is very low. For the IEEE 802.1 5.3a standard, it is planned to use UWB for short range and very high data rate wireless communications amongst suitably capable wireless communications devices, for example to transfer data between "personal" devices such as PDA's and cameras, headsets, cell phones, and media centres. An example of a high data rate application is transmitting HDTV signals between a DVD player/recorder and a TV.

Ultra-wideband (UWB) has been approved by the Federal Communications Commission (FCC) for commercial applications in the unlicensed 3.1 to 10.6 GHz spectrum, provided emissions are maintained below -41.3 dBmIMHz and the bandwidth used at any given time exceeds 500 MHz. Two proposals (multi-band OFDM (MB-OFDM) and direct sequence UWB (DS-UWB)) have been submitted to Task Group 3a of IEEE 802.15 for providing a short-range, high-data-rate, UWB extension to the WPAN standard for commercial exploitation of the FCC ruling.

UWB offers great potential for increasing the data rate of transmissions by virtue of the huge available bandwidth (7.5 0Hz). However, the technology is particularly vulnerable to interference because of the low transmit powers that are allowed (e.g. only 39 j.tW average transmit power for the 528 MHz bands used by the MB-OFDM proposal).

Unless the distance between the transmitter and receiver is very short, so that received power is not too low, then there is a danger that the receiver will detect another significant signal from either a neighbouring piconet or from an unknown interferer.

Furthermore, whilst at present, UWB piconets are arranged such that devices contend for a shared channel, in the future it is envisioned that multiple access versions of UWB piconets may be proposed, and therefore, in addition to the interferer problem, the technology may well have to contend with the near-far problem of multiple access.

However, in UWB transmitters it is less attractive to implement the above mentioned power control method to overcome the near-far and strong interferer problems. The transmit PSD cannot be increased due to the highly restrictive FCC spectral mask. In addition, it is undesirable to reduce the PSD of near devices, as their data rate would inevitably deteriorate as a result of the reduced SNR at the receiver (unlike in WLAN or cellular systems, where it is relatively easy to achieve the SNRs required for operation in the highest rate mode). With UWB devices, the deterioration of performance with range is much more pronounced than for WLAN or cellular systems and therefore it is more critical to use all of the available transmit power to maintain the quality-of-service (QoS) requirements of high data rate devices.

In general terms in one aspect the present invention provides a system for dynamically allocating a variable number of carriers, sub-carriers or tones to each information symbol (e.g. QPSK) in a time divided multiple carrier frequency symbol (e.g. OFDM).

Thus, for example, each QPSK information symbol from a modulator may be replicated and allocated two, three, four or more sub-carriers in the same OFDM symbol (normally each QPSK symbol resides on a single sub-carrier). Replicating and modulating symbols onto several sub- carriers increases the amplitude of the time domain OFDM signal that encodes that information component and ensures that it occupies a larger proportion of the dynamic (ADC) range of the receiver. Consequently, performance loss due to quantisation is reduced, because the information symbol's energy compared with that of a hostile interferer or "near" transmitter is increased. The number of tones or sub-carriers allocated per information symbol (e.g. QPSK) is dynamically variable, and can be made dependent, for example, on the received signal-to-noise ratio (SNR) or the signal-to-interference noise ratio (SINR).

In another aspect there is provided a method of transmitting comprising: modulating (e.g. by an inverse fast Fourier transform (IFFT)) one or more carrier frequencies in a time divided multiple carrier frequency symbol (e.g. OFDM) dependent on a respective received information symbol (e.g. QPSK symbol); wherein the number of carrier frequencies modulated in the same time divided multiple carrier frequency symbol by the respective information symbol is dynamically adjustable. In one embodiment, this adjustable number of carriers assigned to each information symbol may be achieved by replicating the information symbols and modulating carrier frequencies in the same time divided multiple carrier frequency OFDM symbol with both the "original" information symbols and their respective replicated information symbols. In an alternative embodiment, this can be achieved by replicating the modulated time divided multiple carrier frequency symbol. The or each replicated modulated time divided multiple carrier frequency symbol is superimposed on the original to increase the power of each temporal sample. In order to separate the superimposed OFDM symbols at the receiver, the tones are made to be orthogonal; and the replicated symbols are up-converted to a different frequency (with all tones still orthogonal) in order to avoid violating the spectral mask.

The number of carrier frequencies modulated by each respective information symbol can be made dependent on a signal quality parameter derived from the received signal quality of previously transmitted information or time divided multiple carrier frequency symbols. Examples include SNR, SINR, and BER.

In another aspect there is provided a wireless transmitter for transmitting time divided multiple carrier frequency symbols, the transmitter comprising means for modulating one or more carrier frequencies in a time divided multiple carrier frequency symbol dependent on a respective received information symbol, wherein the number of carrier frequencies modulated in the same time divided multiple carrier frequency symbol by the respective information symbol is dynamically adjustable.

In another aspect there is provided a wireless receiver for receiving time divided multiple carrier frequency symbols, the receiver comprising means for demodulating one or more carrier frequencies in a time divided multiple carrier frequency symbol into one or more respective received estimated information symbols, wherein the number of carrier frequencies demodulated in the same time divided multiple carrier frequency symbol into the respective received estimated information symbols is dynamically

adjustable.

The receiver may further comprise a combiner arranged to combine the respective received estimated information symbols.

In another aspect there is provided a method of receiving comprising demodulating one or more carrier frequencies in a time divided multiple carrier frequency symbol into one or more respective received estimated information symbols, wherein the number of carrier frequencies demodulated in the same time divided multiple carrier frequency symbol into the respective received estimated information symbols is dynamically

adjustable.

The transmitter may further comprise means for replicating the information symbols and modulating carrier frequencies in the same time divided multiple carrier frequency symbol with said information symbols and their respective replicated information symbols.

Alternatively the transmitter may further comprise means for replicating the modulated time divided multiple carrier frequency symbol.

In an embodiment the replicas of the time divided multiple carrier frequency or OFDM symbols are up-converted onto different, sufficiently spaced, carrier frequencies, so that no tones overlap In an embodiment the time divided multiple carrier frequency symbols are orthogonal frequency division multiplexed (OFDM) symbols and the modulating means comprises an IFFT processor.

The modulating means may further comprise means for replicating each received information symbol before application to the JFFT processor such that corresponding replicated symbols modulate respective carrier frequencies in the same OFDM symbol.

In one embodiment the same number of carrier frequencies are provided by the IFFT processor for each time divided multiple carrier frequency symbol irrespective of the amount of replication of the information symbols.

In another embodiment the number of carrier frequencies provided by the IFFT processor for each time divided multiple carrier frequency symbol is dependent on the amount of replication of the information symbols.

The IFFT processor may comprise a plurality of IFFT processing blocks, each having the same set of multiple carrier frequency outputs, the output frequencies of respective IFFT processing blocks being frequency translated in order to maintain orthogonality between the sets of multiple carrier frequencies. More generally the IFFT processing blocks need not all be of the same size.

In an embodiment, the allocation of a variable number of OFDM sub-carriers to each information symbol (e.g. QPSK symbol) is achieved by replicating each information symbol one or more times prior to modulating a group of information symbols, including the replicated ones, onto the sub-carriers assigned to an OFDM symbol.

Typically this conversion is achieved using an IFFT digital processor, which for UWB applications pertinent to IEEE 802.15.3a has usually approximately 100 tones or sub-carriers available in each OFDM symbol for information symbols (out of a total of 128 tones). In the case of a QPSK information symbol, one sub-carrier of an OFDM symbol is allocated for each QPSK symbol, thus normally 100 unique QPSK information symbols can be accommodated for each OFDM symbol. However when the QPSK information symbols are replicated (once) then 50 unique QPSK symbols can be accommodated for each OFDM symbol, as each QPSK symbol now modulates two sub-carriers of each OFDM symbol. Thus the information symbol energy for each information symbol (QPSK) in a corresponding OFDM symbol is increased at the expense of a slower rate of transmission, as extra tones are used for sending the replicated information symbols and so correspondingly fewer tones are available for subsequent information symbols. Similarly for higher replication numbers, more of the OFDM's sub-carriers are used up for each information symbol, providing even greater energy or OFDM signal amplitude per information symbol (QPSK), again at the expense of sending progressively fewer information symbols with each OFDM symbol.

In another embodiment, the JFFT has additional, for example 200 or 400, sub-carriers available (out of 256 or 512 sub-carriers respectively). Under normal circumstances only the typical 100 sub-carriers are used, one for each information symbol (e.g. QPSK information symbols). However when the SINR at the receiver has fallen, due to the presence of a hostile interferer, or the signal is badly quantised due to the near-far problem, then the signal level of the transmitted information symbols can be increased by allocating more OFDM sub-carriers to each information symbol. In this embodiment, the additional sub-carriers are used to carry the replicated information symbols. This provides the increased signal level per information symbol, but at the same rate of transmission as previously. Data rate is maintained because information symbols are not displaced by the replicated information symbols which are instead accommodated on new sub-carriers. The additional information symbol energy is carried using a wider bandwidth -the additional new sub-carriers.

Another embodiment uses FFT/1FFT processing blocks that have a reconfigurable size.

This could potentially reduce hardware complexity, reduce power consumption when operating with only a small number of sub-carriers and allow migration between different IFFT sizes and system bandwidths.

In a further embodiment, replicated information symbols are carried on additional sub-carriers in order to maintain the same transmission rate with increased information symbol energy or signal amplitude as with the above mentioned embodiment. However the additional sub-carriers are implemented using separate "standard" IFFT digital processors operating in parallel, for example 128 sub-carriers, which are appropriately linked together. Typically this is achieved by applying a suitable frequency offset for each successive IFFT processor so that over the required bandwidth provided by all of the IFFT circuits, the sub-carriers are appropriately orthogonal (guard bands may be necessary between the separate spectra to avoid interference if the oscillators that set the offset clock frequencies drift). The manner in which information symbols (eg QPSK) are mapped to the available (eg OFDM) tones is configurable, for example replicated QPSK symbols may be used to modulate two or more tones on the same or multiple IFFT processors for example.

Comment: this method could be used if the OFDM symbols are not strictly replicas that are up-converted onto different carriers (see earlier comments), but if the replicated tones are interleaved with new unique data to increase data rate as well as improve robustness. Otherwise, it would be simpler to simply replicate symbols.

In a further embodiment, an initial "standard" time divided multiple carrier frequency symbol (e.g. OFDM) having one sub-carrier allocated per information symbol (e.g. QPSK) is replicated, the replicated symbol being frequency translated and added to the original symbol in the time domain in order to provide a transmission symbol having a wider bandwidth in which more than one sub-carriers are allocated to each information symbol.

Embodiments will now be described with reference to the following drawings, by way of example only and without intending to be limiting, in which: Figures la -ic illustrate example interference or near-far situations for a UWB transmitter/receiver system; Figure 2 illustrates a UWB wireless communications system according to an embodiment; Figure 3a and 3b are schematic diagrams showing a transmitter and a receiver respectively according to Task Group 3a of the IEEE 802.15 standards setting body; Figure 4 illustrates a modification to the transmitter of figure 3a according to an embodiment; Figure 5 illustrates the replication of information symbols in the modified transmitter of figure 4; Figure 6 illustrates a modification to the receiver of figure 3b and which corresponds with the modified transmitter of figure 4; Figure 7 illustrates BER performance improvement when using bandwidth expansion (no energy normalisation) according to the embodiment of figures 4 and 6; Figure 8 illustrates the BER performance improvement when using bandwidth expansion (with energy normalisation) according to the embodiment of figures 4 and 6; Figure 9 illustrates BER performance improvement when using bandwidth expansion and random interleaving of duplicated tones (with energy normalisation) according to the embodiment of figures 4 and 6; Figure 10 illustrates a modification to the transmitter of figure 3a according to another embodiment; Figure 11 illustrates the replication of information symbols in the modified transmitter of figure 10; Figure 12 illustrates a modification to the transmitter of figure 3a according to a third embodiment; Figure 13 illustrates a modification to the transmitter of figure 3a according to a further embodiment; Figure 14 illustrates a modification to the receiver of figure 3b according to another embodiment; Figure 15 illustrates a modification to the receiver of figure 3b according to a yet further embodiment; and Figure 16 illustrates the combination of replicated received symbols in the recevier of figure 6.

Figure Ia illustrates multiple concurrent access in a UWB Piconet P1 from a near' device P1_TX1 and a far' device P1_TX2. Typically the automatic gain control (AGC) of a piconet receiver P1_RXI is set according to the power of the strongest signal, in this case it would probably be the "near" device P1_TX! (may not be if there is no line-of-sight). If the signal received from the "far" device P1_TX2 is significantly weaker, then it may not be amplified appropriately. This is because the excursions of the aggregate signal due to the intended (weak) signal may be unresolvable, or if the amplification is increased then the excursions would be resolved, but the signal peaks would be clipped, leading to a loss of orthogonality of the sub-carriers. Thus in practice, this means the signal received from the "far" device P1_TX2 will occupy only a small proportion of the dynamic range of the receiver's analogue-to-digital converter (ADC) as this accepts inputs from all received signals amplified by the same gain according to the AGC. Therefore weaker signals will be poorly resolved after quantisation by the ADC. This problem is particularly severe for UWB as the ADC is usually of limited resolution (4 or 5 bits) because its design is a trade-off between speed, power consumption, bandwidth and cost. This further reduces the signal-to-noise ratio (SNR) of the weaker signal.

Whilst this scenario is not within the current scope of the 15.3a submissions to Task Group 3a of IEEE 802.15 for MB-OFDM UWB (instead each individual piconet uses time sequenced transmissions), it may be considered in future amendments to the standard in an attempt to improve MAC efficiency.

Figure lb shows the problem that arises when a neighbouring piconet P2 jams the receiver P1RX1 in the piconet P1 of interest because the received hostile signal (from P2_TX 1) is much stronger. This can occur, for example, if the competing piconet P2 (and P2_TX1) are nearby to the receiver Pl_RX1 on the other side of a thin partition wall (e.g. neighbour's house). If the piconets P1 and P2 are uncoordinated then the victim receiver P1RX1 will not necessarily set the front end analogue filters to reject this hostile signal and the dynamic range allocated to the intended signal (from Pl_TX1) will be poor and quantisation errors will occur. As noted above, this is because data decoded from more weakly received transmissions will be prone to error because the signal will be coarsely quantised as it only occupies a small proportion of the dynamic range of the ADC.

Figure 1 c shows the situation where an unknown interferer (narrowband or broadband, for example from a Wi-Fi transmitter) interferes with the UWB signal of interest P1_TX1 and leads to a compromised dynamic range and quantisation errors at a receiver.

Proposed 15.3a solutions provide only a limited mechanism for coping with the near-far problem, or jamming, through the use of a set of operating modes that span different modulation and coding schemes (MCSs). However, the dynamic range improvement that can be achieved in this manner is limited and it forces a reduction in data rate, which may not be acceptable for high data rate applications that are sensitive to latency.

Figure 2 illustrates a MB-OFDM UWB wireless communications system according to an embodiment. The system 200 comprises a MB-OFDM transmitter 210 which wirelessly transmits OFDM symbols 230 to a MB-OFDM receiver 250. The transmifter 210 comprises a dynamic information symbol to OFDM symbol mapper 215 which is controlled by a transmit bandwidth controller 220. Known OFDM transmitters map information symbols such as QPSK symbols to OFDM symbols at a fixed ratio or rate.

For example, typically a QPSK symbol is mapped to each OFDM sub-carrier or tone in an OFDM symbol having 100 tones available for mapping information symbols (i.e. QPSK data or information symbols per OFDM or transmission symbol of 128 tones). However the transmitter 210 of the embodiment is able to vary the bandwidth allocated to each information symbol (e.g. QPSK symbol). Similarly the receiver comprises a dynamic OFDM symbol to estimated information symbol mapper 255 which is controlled by a receive bandwidth controller 260. The transmit and receive bandwidth controllers 220 and 260 are coupled by a control link 235 which is used by the transmit controller 220 to inform the receive controller 260 of the bandwidth allocated to each information symbol. However an explicit link is not essential, and an implicit link could alternatively be used. For example if no acknowledgements (ACKs) are received (failed packets) then the transmitter could simply try frequency expansion.

This (explicit) link 235 could be encoded in the PHY header of the physical layer convergence protocol (PLCP) header. The receive bandwidth controller 260 may also use the control link 235 to inform the transmit bandwidth controller 210 of the SNR or S1NR of received OFDM symbols. However an explicit feedback path might not be needed as reception of other signals can be used to estimate SNRISINR of the link (assuming link reciprocity).

The system 200 can then be configured to vary the bandwidth or number of OFDM tones mapped to each information symbol depending on the received SNR or SINR. In this way if a near-far problem exists, or there is a strong interferer affecting the S1NR of the OFDM symbols 230 received from the transmitter 210, the transmitter can be configured to increase the number of OFDM tones used for each information symbol, in the same OFDM symbol. This increases the signal energy in the OFDM symbol for each respective information symbol (e.g. QPSK) and thereby increases its amplitude in the time domain and hence its proportion of the dynamic range of the ADC in the receiver 250.

The use of dynamic bandwidth expansion ameliorates the near-far and interferer problems without exceeding the FCC spectral mask limits for UWB emissions of -41.3 dBm/MHz. This is because the spectral mask only restricts absolute transmit power once the full UWB spectrum (7.5 GHz) is utilised (0.56 mW). By doubling the bandwidth from 528 MHz to 1056 MHz it is possible to double the transmit power, while maintaining data rate, and therefore achieve power control without recourse to backing off' the near' device and compromising this near device's performance (which, for example, may be critical for QoS of an ultra high rate audio/video (A/V) application). Thus the performance gain is realised by transmitting the data (each information symbol) over a larger bandwidth. Therefore the bandwidth or number of OFDM sub-carriers mapped to each information symbol (e.g. QPSK) can be increased or decreased depending on the received SNR of previous transmissions. In the MB-OFDM proposal, the UWB spectrum is divided up into 528 MHz bands, which are further organised into Band Groups that contain three 528 Hz bands (the highest frequency group contains two rather than three bands). In this embodiment, doubling the information energy whilst maintaining the transmission rate can be achieved by simultaneously transmitting over two band groups instead of one band group, for

example.

Figures 3a and 3b show respectively multi-band OFDM transmitter and receiver architectures that have been proposed within Task Group 3a, the body responsible for drafting the 3a amendment to the IEEE 802.15 standard. The transmitter comprises a scrambler 302, a 64-state binary convolutional code (BCC) 304, a puncturer 306, a 3-stage interleaver 308, a QPSK mapper 310, an IFFT block 312, a DAC 314, a time frequency kernel 316, a multiplier 318, and an antenna arrangement 320. Whilst the various components will be known to those skilled in the art, of interest here the QPSK mapper 310 maps incoming information bits to QPSK symbols. Each QPSK symbol is then used to modulate a sub-carrier in an OFDM symbol by the IFFT block 312. For IEEE 802. 15.3a the use of 128 sub-carriers has been proposed, which are allocated to data, pilot tones, guard bands and nulled tones. Typically this leaves 100 sub-carriers for being modulated with the QPSK information symbols. Thus typically 100 QPSK information symbols are mapped to a single OFDM symbol, which is then transmitted to the receiver.

The receiver 350 comprises an antenna 352, a pre-selection filter 354, a low noise amplifier 356, quadrature and in-phase signal paths each having a receive down-converter 358 (i and q), a low pass filter 360, a variable gain amplifier 362, and an ADC 364. The outputs of the ADC's 364i and 364q are input to a Fast Fourier Transform (FFT) block 368, the output of which is coupled to a digital processing block 370 for removing pilots, frequency domain equalised (FEQ), and correction of carrier frequency offset from pilot information 372. The output is de-interleaved by block 374, the forward error correction code is decoded by a Viterbi decoder 376 and the signal is descrambled by block 378. There is also an automatic gain controller (AGC) 366 which adjusts the gain of the variable gain amplifiers 362i and 362q depending on the peak signal at the respective ADC's 364i and 364q. The incoming baseband analogue signals (in-phase and quadrature) are amplified by respective variable gain amplifiers 362 (i and q) at a gain determined by the AGC 366, and digitised by respective ADC's 364. The digitised signals (OFDM symbols) are then fed to the FFT 368 which transforms each OFDM symbol into the frequency domain and, after equalisation, enables estimates to be calculated of the complex constellation values encoded onto each of the sub-carriers (originally from a QPSK alphabet). Subsequent deinterleaving, error correction decoding, descrambling processes are then used to determine the transmitted sequence of bits.

Figure 4 illustrates a modification to the QPSK mapper and IFFT blocks of figure 3a that is used to implement the variable frequency spreading of the QPSK symbols corresponding to the information symbol to OFDM symbol mapper 215 of figure 2. The modified transmitter 400 comprises a QPSK mapper 410 coupled to a replicator 411, which in turn is coupled to the input of an IFFT block 412. The QPSK mapper 410 receives a FEC coded, punctured and interleaved bit stream from preceding components or blocks in the transmitter, for example those illustrated in figure 3a and notionally represented in figure 4 by block 409. The replicator 411 is controlled by the transmit bandwidth controller 220 to replicate each incoming QPSK symbol (S 1, S2, S3...) by a replication factor R. In theexample shown, with R=2, each input QPSK information symbol (e.g. Si) is replicated once such that two identical information symbols (e.g. SI i and SI2) are output to the IFFT block 412. This operation doubles the number of information symbols to be mapped to the OFDM symbols by the IFFT block 412. It can be seen that the effect of this is to double the number of sub-carriers or OFDM tones 425 allocated to each QPSK information symbol. Each QPSK symbol is normally mapped to only one sub-carrier, however when R=2 the same QPSK symbol is mapped to two sub-carriers. This doubles the bandwidth allocated to each information symbol and hence increases the time domain excursions associated with this information.

Consequently, the signal occupies a greater proportion of the receiver's ADC dynamic range and corruption of the signal by quantisation is reduced. The time domain OFDM symbol (OFDMn) is then applied to a DAC, up-converted onto an RF carrier wave, amplified and transmitted to a receiver.

The replication factor R can be any convenient number, and can be varied depending on the quality' of the received OFDM signals. Convenient quality parameters include SNR, SINR, required range, target bit error rate and required data rate. The replication factor R can also be reduced, for example when an interferer or "near" device disappears or stops transmitting.

In this embodiment, the IFFT block 412 is a standard IFFT digital processor, which performs a 128-point discrete IFFT. This means that if the replication factor R is greater than one then the rate of transmission of the transmitter is reduced, because there are now more information symbols to be transmitted per unit time using the same number of OFDM tones. Thus if the bandwidth required for the information symbols is doubled (R=2), then the transmission rate is halved (0.5). This may be acceptable for many applications which are not time critical, and has the advantage of using legacy processing blocks (e.g. IFFT 412) with minimal modification to the transmitter circuitry. In some embodiments, the system could still operate in a higher rate mode to compensate, where the system has several modes to enable performance migration in deteriorating SNR. For example, a low rate mode might have a time spreading factor of 2 and this could be sacrificed for double bandwidth symbols.

Figure 5 illustrates the mapping of information symbols (Si -S 100) to OFDM symbols, where a replication factor of R=3 has been used. Each of the information symbols has been replicated twice, so that for each original input information symbol there are now three successive identical information symbols. Better performance can be obtained if replicated information is not put on successive or adjacent tones but where these tones are interleaved such that tones associated with the same information are randomly allocated across the available tones. As noted above, the chosen number of replicated information symbols (R) can be made to be dependent on the SNR or SINR, with this information conveyed to the receiver via the header in much the same way as the modulation coding scheme used is communicated to the receiver of a traditional system.

Normally (i.e. when R=l) 100 QPSK information symbols (Sl -100) would be mapped to each OFDM symbol. However with the embodiment of figure 4, three OFDM symbols are now needed to transmit the same amount of information, albeit at a higher aggregate transmit power that will result in a higher SNR at the receiver.

The time domain signal resulting from the IFFT will now contain more redundant information and therefore the information symbol energy is increased. Consequently, this information can compete' on a more equal footing' with other received signals for the finite dynamic range of the receiver. Quantisation errors are therefore reduced and the error floor is reduced. At the receiver, the redundant information may be recombined at either the symbol or bit level.

Figure 6 illustrates a modification to part of the receiver of figure 3b and which is used to implement the multiple OFDM symbol to information symbol estimate demapper 255 of figure 2. The modified receiver 600 comprises a variable gain amplifier 662 controlled by an AGC 666, an ADC 664, an FFT block 668, a combiner 669, and a block 670 for removing pilots, frequency domain channel equalisation, and correction of carrier frequency offset -which is analogous to the processing block 370 of figure 3b. For simplicity of explanation, separate in-phase and quadrature analogue baseband signal paths are not shown, however it is assumed that these are as before. Because each information symbol (e.g. QPSK) modulates several OFDM sub-carriers in its respective OFDM symbol due to the replication in the transmitter, the corresponding time domain signal received at the ADC 664 has a greater amplitude than it would otherwise have had, and therefore it occupies a greater proportion of the dynamic range of the ADC 664. The digitised OFDM symbol (OFDMn) is passed to the FFT block 668 which recovers estimated information symbols (S 1,*, S1f, S21*, S22*,...) corresponding to the transmitted information symbols (Si1, SI2, S21, S22, ...), wherein duplicate estimated information symbols are received when the replication factor R=2 is used at the transmitter. The multiple estimated information symbols (Si Sl*; S2 -S2*, are combined by the combiner 669 to recover a combined estimated information symbol (eg S1**) for each information symbol (e.g. 51) from the QPSK symbol mapper 410 in the transmitter. The combiner can be a maximum ratio combiner for example, or any other suitable combiner for example an equal gain combiner or a square law combiner. The combined estimated information symbol (e.g. Sl**) corresponds to a signal with a higher SNR than a single estimated information symbol (eg S i*). This also provides frequency diversity because the channel responses at the two frequencies will be different and one of them may have been in a fade with the information on this faded sub-carrier completely lost. The combined estimated information symbols (Sl**,S2**,S3**...) are then processed as in a conventional system for forward error control, channel equalisation, decoding, and so on, to obtain information bit estimates.

In an alternative arrangement, the signal parts may be combined at the information bit rather than information symbol level, for example after decoding and demodulation from QPSK into information bits.

Figure 16 illustrates an implementation for combining received symbols corresponding to the replicated information symbols transmitted from a transmitter. Here a maximum ratio combiner (MRC) 1669 corresponds with the combiner 669 of Figure 6 and is coupled to the output of the FFT digital signal processor 668. The FFT digital processor produces complex valued received signals (SI 1*, S12*, S2 1*,... in figure 6) corresponding to transmitted information symbols (S 11, S 12, S21,...), multiplied by the effects of the channel for their respective tone or carrier. Estimates for the transmitted information symbols can be obtained by removing the effects of the channel (h).

Combining can be implemented using the complex valued received signals or the estimated symbols (SI 1, S 12...) once the channel has been removed. In a further alternative, combining of replicated information may be done at the bit level.

Figure 16 illustrates combining the complex valued received signals corresponding to replicated information symbols. After using MRC the channel is just a power scaling and the information symbol can be estimated by simple quadrant detection for QPSK (e.g. of SI1 * in figure 6 or more generally y1, in figure 16). A plurality (n) of tones in each received OFDM symbol carry identical information, and are combined to yield a more robust estimate of the transmitted information. Figure 16 illustrates how a maximum ratio combiner may be implemented to combine the ith group of signals from n sub-carriers that share common transmitted information. The received signal from group on sub-carrier k, and the channel response at sub-carrier frequency k relevant to group are denoted by Y,,k and h,.k respectively. In practice, the channel transfer response is estimated by transmission of a known training sequence.

The equation used to combine the signals is given by: z, = Where denotes complex conjugation of the bracketed quantity.

A received signal is related to a transmitted signal X by the following equation, where ng denotes zero mean complex Gaussian noise.

y = h + ng The effective channel response h, (needed for the equaliser) for the new symbol Z1 (weighted average) is therefore given by: =jh1 The resulting averaged' received symbols (z1 or e.g. Si * * in figure 6) are more robust to interference and the effects of ADC quantisation. Combining is also possible at the bit level, further along the receiver chain. However, maximum ratio combining of the symbols at the output of the FFT provides good performance for relatively low complexity.

The performance of the embodiment of figures 4 and 6 for a system that uses a frequency spreading factor of R=3 is illustrated in figure 7, which shows the bit error rate (BER) realised when a receiver decodes jointly detected transmissions from near and far devices (on non-overlapping frequencies, but lying within the same band).

Figure 7 was produced using energy normalisation so that the sub-carrier energy held fixed (E_s plotted on the x-axis), which is what would happen when as much energy as the mask will allow is transmitted. For the simulation, the devices were configured so that the power received from the far device was only 3.7% of that received from the near device. This is consistent with the near device being stationed 1 m away from the receiver and the far device located at 3 m, with a multi-path attenuating environment giving a path loss exponent of three. The abscissa of figure 7 is energy per sub-carrier which in this case is equal to information per symbol (the energy per information symbol is three times higher than the sub-carrier energy for the bandwidth expanded transmission).

The redundant information (i.e. S1f to Slc, S22* to S2*,...) was combined in an optimum manner using maximum ratio combining. Figure 7 shows that when no frequency spreading is employed then at low SNR the near and far results give similar BER performance (energy normalised). However, as the SNR increases the effective noise introduced due to quantisation becomes increasingly detrimental and the performance curves diverge. When frequency spreading is used then there are improvements due to two different mechanisms: the curve is shifted by 1 Ologio(3) due to the additional energy of the information symbols; and the gradient is steeper and the noise floor lower due to reduced quantisation. Hence, spreading increases the robustness of longer range links in the presence of interference. In the above described embodiment this performance increase comes at the expense of a three-fold reduction in data rate.

However, the performance gain is not all due to the increase in information symbol energy and a reduced error floor is observed even when the results are compensated for the increase in symbol energy by plotting against information bit energy Eb/NO -as illustrated in figure 8. This reduced error floor would improve throughput at high SNRs and may enable the operation of applications that have demanding QoS requirements, such as A/V applications. Figure 8 is plotted against bit energy (without energy normalisation), which is why the green and light blue curves overlay at low SNR and don't have a shift like in figure 7. The curves diverge at high SNR because, no matter how high SNR becomes, there is an "effective" noise associated with the quantisation.

The method illustrated uses consecutive symbol replication. In practice, performance can be enhanced further if the replicated symbols are interleaved across the bandwidth so that more frequency diversity is achieved if the coherence bandwidth of the channel exceeds the spacing between OFDM tones (4.125 MHz). Figure 9 shows the result when the duplicated tones are randomly interleaved across each single OFDM symbol.

This shows the effect of spacing the redundant information further apart in frequency to avoid correlated fading. Note that again it is plotted against bit energy/noise (without energy normalisation) and that in a real system there would be an additional shift improvement due to the increased energy provided through bandwidth expansion (the improvement shown on this graph is just due to reduced quantisation noise and not due to any increase in energy). In this result, the error floor is completely eradicated, although, in practice, a deterministic interleaver would have to be used in an open ioop system (so that it is known in advance by the receiver) which may suffer a minor loss in performance. Interleaving could also be performed across several OFDM symbols to extract temporal diversity as well as frequency diversity, although the improvement may be small and not worth the disadvantage of incurring decoding latency at the receiver.

In the above described embodiment, it has been assumed that all of the OFDM tones are populated. However, it is also possible to use the embodiment with a multiple access strategy such as orthogonal frequency division multiple access (OFDMA). In this scenario, for example, a far transmitter may be allowed to transmit using 2/3 of the available tones, whereas a near user would be assigned the remaining 1/3. In this example, the tones from the far device would have two-fold redundancy to help balance power at the receiver. It has also been noticed that a small performance gain (reduced BER) is experienced even when the far device uses unique data on all of the tones and half of them are subsequently ignored (i.e. quantisation errors are reduced if a weak signal is superimposed on a signal whose component is subsequently discarded after the quantiser/ADC).

Another embodiment is illustrated in figure 10, in which the OFDM symbol from the traditional system is simply replicated a number of times (two times is illustrated). The or each replicated symbol is then combined with the original symbol and transmitted on adjacent channels or on separate band groups as a signal OFDM symbol having a wider bandwidth. The sub-carriers from the original and one or more replicated symbols are made appropriately orthogonal as is known in order to generate the combined or wider bandwidth OFDM transmission symbol which is transmitted. If an adjacent channel is used then channel bonding can be used and the time frequency codes used by the IEEE8O2.1 5.3a network can be redesigned accordingly. In the standard, each successive OFDM symbol is transmitted from a different band. A different deterministic hopping pattern is assigned to each new piconet in the network. The hopping patterns ensure that collisions and therefore the collective interference between piconets is minimised and randomised. Alternatively, the replicated symbols can be reassigned as different devices within the piconet.

Figure 11 illustrates the mapping of QPSK information symbols (Si -S 100) to OFDM symbols, where a replication factor of R=3 has been assumed as before. Each of the information symbols have been replicated twice, so that for each original input information symbol there are now three successive identical information symbols.

Traditionally (i.e. when R1) 100 QPSK information symbols would be mapped to each OFDM symbol. With this embodiment as with the embodiment of figure 4, three OFDM symbols are needed to transmit the same amount of information, however the three OFDM symbols are combined and spread over a wider bandwidth rather than over a longer time. Thus instead of three subsequent OFDM symbols in band group 1 for example, a single OFDM symbol having three times the bandwidth (e.g. from three different band groups) is transmitted.. As before, this results in greater energy per information symbol, and hence improved resolution by the ADC of the receiver. This is also achieved at the same transmission data rate.

This scheme may be realised in hardware in a number of ways, for example as shown in figure 10. This uses parallel banks 1018 of OFDM transmitters each having their own individual IFFT's, DACs, mixers (driven from common clock), PAs and co-located antennas (for compactness). Each replicated information symbol is directed to a different OFDM bank 1018, so that as the replication factor R increases, the number of OFDM banks is increased accordingly. Thus in the example shown where R=2, the original information symbols Si, S2... are directed to the first IFFT block 101 2a which forms part of a first OFDM bank 1018a, and the replicated symbols Si2, S22,... are directed to the second IFFT block 101 2b which forms part of a second OFDM bank 101 8b. Identical IFFT processors 1012 could be used in each OFDM bank 1018 for simplicity and cheapness of hardware design, with the time frequency kernel 1016 for each OFDM block translating the respective OFDM symbols by different amounts (Fîa or FTb) in order to ensure orthogonality of the sub-carriers over the entire expanded bandwidth. The OFDM symbols OFDMna and OFDMb are then combined and transmitted as a single combined OFDM symbol having double the bandwidth.

Alternatively one of the OFDM symbols (OFDMflb) may be frequency translated then added to the other OFDM symbol (OFDMna) before both are then further translated by the same time frequency kernel 1016. By running two bloeks (iOl8a and iOl8b) on a common oscillator (and adding a frequency shift to just one of them -1018b), the frequency conversion errors are minimised. This is because the accuracy of crystals is in parts per million and therefore it can be advantageous to keep a common clock for the bulk of the up-conversion and then add a small offset to minimise error in this.

Thus with either of these embodiments, the transmission rate of a standard OFDM transmitter is maintained, by expanding the bandwidth used for transmitting the OFDM symbols. This is useful for applications in which it is undesirable to reduce the transmission rate, for example for latency sensitive applications such as audio or video calls. A guard band may need to be inserted between the two bands to ensure that the tones do not interfere with one another and the receiver would need to estimate and track two frequency offsets rather than the traditional single offset.

Figure 12 illustrates an alternative embodiment in which the hardware is implemented by employing an expanded IFFT block 1212. In this embodiment the original and replicated information symbols are fed to the same IFFT block, which in the example illustrated has 200 available OFDM data tones instead of the more usual 100 available for modulation by the information symbols (S) (256-point IFFT instead of 128-point). It can be seen for example that for R=2, 100 QPSK information symbols can be accommodated using all the available tones (1-200). Thus the transmission rate is maintained whilst increasing the energy per information symbol by increasing the bandwidth used -in this example, two band would be used instead of a single band with known MB-OFDM arrangements. This design simplifies the DAC and mixing stages of the transmitter after the IFFT block 1012, which can now be shared, and additionally all of the sub-carriers will be automatically orthogonal. However, this is achieved at the expense of a more complicated IFFT design. A further disadvantage of this arrangement is that the peak-to-average power ratio PAPR of the individual signals is increased and the clock rate for the digital circuitry is increased.

Figure 13 illustrates a further alternative embodiment in which the hardware implementation is by replicating the transmission (eg OFDM) symbols rather than the information (eg QPSK) symbols. Here only one IFFT block 1368 is used, with the resulting OFDM symbol being fed into two signal paths (a and b), one of the OFDM symbols (OFDMflh) being frequency translated (FT) by a suitable oscillator and mixer block 1316, before both the non-translated (OFDMa) and the translated (OFDMb) symbols being converted to analogue signals, are combined and multiplied by a common carrier frequency. Subsequently, the combined wider bandwidth OFDM symbol is transmitted. This avoids the need for additional banks of IFFT's or an extended IFFT, at the expense of some additional complexity in translating the one of the output OFDM symbol signal paths.

Receiver architectures suitable for use with the transmitters of figures 10, 12 and 13 are similar to those of figure 6. Either an expanded FFT block 1468 corresponding to the IFFT block 1212 of figure 12 could be used, with additional (e.g. 256) OFDM tone capability, or multiple banks of FFT' s 1568 corresponding to the IFFT block 1012 of figure 10 could be used. These are shown respectively in figures 14 and 15. In the embodiment of figure 14, a wider pre-selection filter 1452 is required in order to allow the wider bandwidth of the expanded (e.g. two band) OFDM symbol to pass. In the embodiment of figure 15, banks of receivers 1580 are used, each with their own pre-select filter 1552 bandwidth according to their respective OFDM symbol frequency range (e.g. Band 1, Band 2, Band 3). However in an embodiment where band hopping is used, then either the filter has to be wide enough to cover the whole band group, or it has to be dynamically tuneable. Respective mixers 1558 arc also required to recover the analogue baseband signal from the different band groups. Each OFDM symbol is then converted by a respective "standard" FFT circuit 1568 into information symbol estimates (S1*, S2*, ...). Information symbol estimates from each receiver bank 1580a and 1580b are then directed to a common combiner 1569 to be combined with corresponding information symbol estimates from other receiver and FFT banks.

The various transmitter and receiver architectures described could be used together in different combinations; for example the transmitter of figure 12 with the receiver of figure 15.

Whilst the embodiments have been described predominantly with respect to QPSK information symbols, the skilled person will appreciated that various modulation and coding schemes could be used with the embodiments.

Whilst the embodiments have been described with respect to MB-OFDM wireless communications systems, other wireless systems could alternatively be adapted for use with the embodiments..

The skilled person will recognise that the above-described apparatus and methods may be embodied as processor control code, for example on a carrier medium such as a disk, CD-or DVD-ROM, programmed memory such as read only memory (Firmware), or on a data carrier such as an optical or electrical signal carrier. For many applications embodiments of the invention will be implemented on a DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit) or FPGA (Field Programmable Gate Array). Thus the code may comprise conventional program code or microcode or, for example code for setting up or controlling an ASIC or FPGA. The code may also comprise code for dynamically configuring re-configurable apparatus such as re-programmable logic gate arrays. Similarly the code may comprise code for a hardware description language such as Verilog TM or VHDL (Very high speed integrated circuit Hardware Description Language). As the skilled person will appreciate, the code may be distributed between a plurality of coupled components in communication with one another. Where appropriate, the embodiments may also be implemented using code running on a field-(re)programmable analogue array or similar device in order to configure analogue hardware.

The skilled person will also appreciate that the various embodiments and specific features described with respect to them could be freely combined with the other embodiments or their specifically described features in general accordance with the above teaching. The skilled person will also recognise that various alterations and modifications can be made to specific examples described without departing from the scope of the appended claims.

Claims

CLAIMS: 1. A wireless transmitter for transmitting time divided

multiple carrier frequency symbols, the transmitter comprising: means for modulating one or more carrier frequencies in a time divided multiple carrier frequency symbol dependent on a respective received information symbol; wherein the number of carrier frequencies modulated in the same time divided multiple carrier frequency symbol by the respective information symbol is dynamically

adjustable.

2. A transmitter according to claim 1, further comprising means for replicating the information symbols and modulating carrier frequencies in the same time divided multiple carrier frequency symbol with said information symbols and their respective replicated information symbols.

3. A transmitter according to claim 1, further comprising means for replicating the modulated time divided multiple carrier frequency symbol.

4. A transmitter according to any one preceding claim, wherein the number of carrier frequencies modulated by the respective information symbol is dependent on a signal quality parameter associated with previously transmitted information or time divided multiple carrier frequency symbols.

5. A transmitter according to any one preceding claim, wherein the time divided multiple carrier frequency symbols are orthogonal frequency division multiplexed (OFDM) symbols and the modulating means comprises an Inverse Fast Fourier Transform (IFFT) processor.

6. A transmitter according to claim 5, wherein the modulating means further comprises means for replicating each received information symbol before application to the IFFT processor such that corresponding replicated symbols modulate respective carrier frequencies in the same OFDM symbol.

7. A transmitter according to claim 6, wherein the same number of carrier frequencies is provided by the IFFT processor for each time divided multiple carrier frequency symbol irrespective of the amount of replication of the information symbols.

8. A transmitter according to claim 6, wherein the number of carrier frequencies provided by the IFFT processor for each time divided multiple carrier frequency symbol is dependent on the amount of replication of the information symbols.

9. A transmitter according to claim 8, wherein the IFFT processor comprises a plurality of IFFT processing blocks having the same set of multiple carrier frequency outputs, the output frequencies of respective IFFT processing blocks being frequency translated in order to maintain orthogonality between the multiple carrier frequencies.

10. A wireless receiver for receiving time divided multiple carrier frequency symbols, the receiver comprising: means for demodulating one or more carrier frequencies in a time divided multiple carrier frequency symbol into one or more respective received estimated information symbols; wherein the number of carrier frequencies demodulated in the same time divided multiple carrier frequency symbol into the respective received estimated information symbols is dynamically adjustable.

11. A receiver according to claim 10, further comprising a combiner arranged to combine the respective received estimated information symbols.

12,. A method of transmitting comprising: modulating one or more carrier frequencies in a time divided multiple carrier frequency symbol dependent on a respective received information symbol; wherein the number of carrier frequencies modulated in the same time divided multiple carrier frequency symbol by the respective information symbol is dynamically

adjustable.

13. A method according to claim 12, further comprising replicating the information symbols and modulating carrier frequencies in the same time divided multiple carrier frequency symbol with said information symbols and their respective replicated information symbols.

14. A method according to claim 12, further comprising replicating the modulated time divided multiple carrier frequency symbol.

15. A method according to any one of claims 12 to 14, wherein the number of carrier frequencies modulated by the respective information symbol is dependent on a signal quality parameter associated with previously transmitted information or time divided multiple carrier frequency symbols.

16. A method of receiving comprising: demodulating one or more carrier frequencies in a time divided multiple carrier frequency symbol into one or more respective received estimated information symbols; wherein the number of carrier frequencies demodulated in the same time divided multiple carrier frequency symbol into the respective received estimated information symbols is dynamically adjustable.

17. A method according to claim 17, further comprising combining the received estimated information symbols.

18. A processor code product carrying processor code which when executed on a processor causes the processor to perform a method according to any one of claims 12 to 17.