TELECOMMUNICATION SYSTEM WITH SIMPLIFIED RECEIVER

Telecommunication receivers, for receiving digital data, frequently employ circuits or systems for synchronizing the local carrier signal and symbol clock to the transmitter, estimating and correcting for any reflections or other convolutional effects in the communication channel, and filtering to shape the received symbol pulses so as to avoid inter-symbol-interference (ISI). The circuits or systems that carry out this processing are often computationally demanding, leading to implementations that are power-hungry, large and costly. Furthermore, the processing is usually carried out in the digital domain, leading to the need for one or more analogue-to-digital converters (ADCs) preceding the digital processing stage.

Other telecommunication systems seek to reduce some of the receiver complexity by employing non-coherent modulation schemes, such as frequency shift keying (FSK) or differential phase shift keying (DPSK). These obviate the need for recovery of the transmitter's carrier phase, but they do so to the detriment of other system characteristics. For example, FSK is spectrally inefficient, leading to longer packet lengths in packet-based systems, and hence greater power consumption in the radio frequency (RF) circuitry. DPSK can be more spectrally efficient than FSK, but is less robust to noise and interference than non-differential techniques.

According to a first main aspect of the present invention, there is provided a method of operation of a communication system, the method comprising: transmitting a signal from a transmitter to a receiver, the signal comprising a carrier having data modulated thereon; estimating in the receiver a waveform synchronization error in the signal; sending from the receiver to the transmitter an error signal indicative of the estimated waveform synchronization error; and in the transmitter, correcting the signal transmitted to the receiver, based on the error signal.

According to other aspects of the invention, there are provided transmitters and receivers for operating in accordance with the method.

According to a second main aspect of the present invention, there is provided a method of transmitting a data signal from a data transmitter to a data receiver, wherein the data transmitter and data receiver operate according to a time division duplex system, the method comprising: periodically transmitting a sounding signal from the data receiver to the data transmitter; in the data transmitter, receiving the sounding signal, and forming a channel estimate therefrom; and in the data transmitter, pre-equalizing the data signal based on said channel estimate.

According to other aspects of the invention, there are provided transmitters and receivers for operating in accordance with the method.

The invention thus provides a novel telecommunication system that massively simplifies the receiver at one end of the communication link (the remote station, RS), commensurately reducing its power-consumption, size and cost. In some embodiments, the system is able to avoid completely the use of ADCs and any subsystems for correcting carrier and symbol clock offsets, channel convolution or ISI at the RS. Instead, the corrections are calculated and applied at the other end of the link (the base station, BS), with the RS receiver requiring only a very small amount of circuitry. In some embodiments, the system can offer much higher spectral efficiency than FSK, allowing receiver power savings in the RF circuitry, and it avoids the robustness problems of DPSK.

The application of primary interest is the wireless transmission of audio data to battery- powered loudspeakers or earphones, in which extended battery operating lifetime, and hence low power consumption, is desirable. In this application, a BS resides in the audio source, and one RS resides in each loudspeaker or earphone unit of the audio system. Clearly, however, the communication system presented here has utility in many other power-, cost- or size-sensitive data telecommunication applications.

To show how the present invention may be put into effect reference will now be made, by way of example, to the accompanying drawings, in which:-

Figure 1 is a block schematic diagram of a telecommunication system in accordance with an aspect of the present invention.

Figure 2 illustrates a quadrature constellation map, for explanation of an aspect of the present invention.

Figure 3 illustrates a symbol sequence, for explanation of a further aspect of the present invention.

Figure 4 illustrates a quadrature constellation map, for explanation of a further aspect of the present invention.

Figure 1 illustrates a telecommunication system 10 in accordance with an embodiment of the present invention. A base station (BS) 20 communicates bi-directionally with a remote station (RS) 30 by radio. The BS 20 comprises a transmitter 22 and a receiver 24. Likewise, the RS 30 comprises a transmitter 32 and a receiver 34. The particular system shown employs time division duplexing (TDD), i.e. all the transmitters and receivers 22, 24, 32, 34 operate at the same radio frequency, and in such a way that, at any one time, only transmitter 22 or 32 may transmit (to receiver 34 or 24, respectively). However, most of the features of this telecommunication system can also be applied to a frequency division duplex (FDD) system.

The transmitter 22 and receiver 24 of the BS 20 are connected through a RF switch 26 to a BS antenna 28, while the transmitter 32 and receiver 34 of the RS 30 are connected through a RF switch 36 to a BS antenna 38.

The signalling between the BS transmitter 22 and RS receiver 34 (termed the downlink signal) is based on single carrier modulation. Although the invention is described herein with reference to a wireless communication system 10, it will be apparent that the invention can also be implemented in a wired communication system, in particular where data is encoded by modulating a carrier signal, and especially where data is encoded in the phase of a sinusoidal single carrier signal. We start by considering quadrature phase shift keying (QPSK).

The RS transmitter 32 and the BS receiver 24 may be conventional digital telecommunication elements. For example, they may employ single carrier, spread spectrum or orthogonal frequency division multiplexing techniques. Thus, they will not be described in detail herein. The signal from the RS transmitter 32 to the BS receiver 24 is termed the uplink signal.

The BS transmitter 22 receives the payload data, and passes this to a probe insert block 222, in which probe signals are inserted, as will be described in more detail below. In-phase and quadrature components I and Q of the resulting signal are passed to a resampler 224, having a particular sampling phase. As will be described in more detail below, the sampling phase can be adjusted to adjust the sample phase of the resampled signal. The resampled signal is passed to an FIR filter 226, and the filtered signal is applied to a mixer 228. As will be discussed in more detail below, the other input of the mixer 228 receives a phase adjustment signal, and the output of the mixer 228 has an adjusted carrier phase.

The I and Q components of the phase adjusted digital signal are passed to respective digital-to-analogue converters 230, 232, and then to baseband filters 234, 236.

The filtered signals are passed to the transmitter RF circuitry 238, where they are upconverted to the intended RF frequency for transmission.

The sampling phase of the resampler 224 is determined by a signal from a phase accumulator 240, which receives a signal from a loop filter 242, which in turn receives a signal via a switch 244 from the BS receiver circuit 24.

The tap values of the FIR filter 226 are determined by a tap calculator 246, which in turn receives a signal from a channel estimator 248 in the BS receiver circuit 24.

The phase adjustment input signal to the mixer 228 is received from a numerically controlled oscillator (NCO) 250, which receives a signal from a loop filter 252, which in turn receives a signal via the switch 244 from the BS receiver circuit 24.

As already noted, the RS receiver 34 has very minimal circuitry, with no capability to change or synchronize its local carrier or symbol clock references. Instead, the BS transmitter 22 computes and corrects the transmitted carrier and symbol phases so

that they are correctly synchronized when they are received at the RS receiver 34. The RS receiver 34 has only a very small amount of circuitry to support this, in the form of a probe measurement unit 40. Measures from this unit are delivered back to the BS 20 via the RS transmitter 32. We now describe this scheme in detail.

The antenna 38 at the RS receives the radio signal transmitted from the BS 20, and the received signal is fed to receiver RF circuitry 42, which amplifies and frequency-shifts the signal down to complex analogue baseband. There is no means within the RS 34 for synchronizing the local mix-down oscillator, or oscillators, in the receiver RF circuitry 42 to the received carrier phase.

The complex output from receiver RF circuitry 42 is applied to analogue baseband filters 44, 46, for the in-phase and quadrature components I and Q respectively. These filters are low-pass, with a nominal cut-off frequency at half the symbol rate. They are identical, within manufacturing tolerances, to each other, and to the baseband filters 234, 236 in the BS 20. They form, to a good approximation, matched filters for the purpose of optimal symbol detection in noise, and they also provide frequency selectivity, extracting the wanted frequency channel from unwanted frequencies. For pulse shaping purposes, using practical analogue filters, however, the cascade response of filters 234 and 236, feeding into filters 44 and 46, may lead to some ISI, but this can be corrected for in the BS transmitter 22, as described later.

The outputs of baseband filters 44, 46 are fed to thresholders 48 and 50, respectively, which output a high binary logic value when their input signal is greater than zero; otherwise, they output a low binary logic value. The outputs of the thresholders 48 and 50 are fed into registers 52 and 54, respectively, which sample the thresholder outputs on the active (e.g. rising) edge of a clock signal. The clock signal is simply a regular square wave at the intended symbol frequency, for example derived from the main digital system clock. There is no means in the RS 30 to synchronize the clock signal to the received symbol phase.

If the BS transmitter 22 were conventional and the transmitted carrier and symbol phases were determined using just the transmitter's own local references, we would have no system synchronization with the RS receiver 34 just described, and data communication would not be possible. Therefore, a measurement method and

feedback chain are provided to form measures of the synchronization errors in the carrier phase and symbol phase (which we refer to generically as "waveform" synchronization errors) of the signal received at the RS receiver 34, and to correct these synchronization errors at the BS transmitter 22.

One issue for measurements at the RS receiver 34 is that no ADC is present, only the binary thresholders 48 and 50, sampled at symbol rate by the registers 52 and 54. To obtain useful measurements from the outputs of the registers 52 and 54, some special symbols are inserted into the main data-bearing symbol stream in the probe insert block 222 of the BS transmitter 22. These symbols constitute what are termed here phase probes, and two types of phase probe are employed in this illustrated embodiment, namely carrier phase probes, used to estimate carrier phase offsets; and symbol phase probes, used to estimate symbol phase offsets.

A carrier phase probe is a symbol inserted in the transmitted signal, to allow the receiver to form a measure of carrier phase synchronization errors. Specifically, in this embodiment, the carrier phase probe comprises a special symbol that is transmitted by the BS transmitter 22 at the normal symbol timing phase, but, rather than using one of the four conventional QPSK constellation points for its complex amplitude, the transmitted symbol amplitude is placed exactly on the one of the decision boundaries between the normal QPSK constellation points. Figure 2 shows how carrier phase probe symbols compare to standard QPSK symbols.

Thus, in a QPSK system, data is transmitted by modulating the carrier signal with a signal of known magnitude a, where the data is carried by the phase of the signal during any particular symbol period. In particular, the phase can be equal to 45°, 135°, 225° or 315°. The I and Q components therefore have the same magnitudes (i.e. a/V2), and the data is carried by the signs of the I and Q components.

In this illustrated embodiment of the invention, the possible carrier phase probes also have the known magnitude a, but the phase can be equal to 0°, 90°, 180° or 270°. In each case, one or other of the I and Q components has zero magnitude.

Carrier phase probes are inserted periodically into the main data-bearing symbol stream, according to some predetermined insertion pattern, by the carrier/symbol phase probe insert unit 222.

The same insertion pattern is output by a probe timing control unit 60 in the probe measurement block 40, and fed to the enable input of an accumulator 62.

For the purposes of initial illustration, we will now assume that every carrier phase probe uses the complex amplitude 80 shown in Figure 2. In this restricted case, the outputs of a probe quadrant control unit 64 are fixed such that a multiplexer 66 always selects the output of the register 52, and the line from the probe quadrant control unit 64 to an XOR gate 68 is zero. Thus, in this restricted case, the I component of the sampled symbols are fed uninverted to the accumulator 62. The accumulator 62 accumulates a signed value according to the output of the XOR gate 68, such that it decrements by one for a zero from the XOR gate 68, and increments by one for a one from the XOR gate 68; but it increments or decrements only when its enable input is active, i.e. only for phase probe symbols. The accumulation is performed across a predetermined number of carrier phase probe symbols, termed the accumulation period, and the probe timing control unit 60 asserts reset to zero the accumulator 62 at the start of each such period.

The probe timing control unit 60 also outputs a data_valid line 70 to designate when symbols are main payload data, rather than phase probe symbols.

To aid understanding of how carrier phase probes are used to estimate carrier phase error, first consider the telecommunication system of Figure 1 to be in a state such that there is no carrier or symbol phase error at the RS receiver 34. In this state, at the sampling instant for a received carrier phase probe, the output of the baseband filter 44 would ideally be zero (reflecting the complex amplitude 80 in Figure 2), but in reality it will be either slightly positive or negative, due to additive zero-mean noise, from the communication channel, thermal noise in the receiver analogue circuitry, etc. Therefore, the output of the register 52 for carrier phase probes will be random, with equal probability of a low or high output, and the accumulator 62 will accumulate a value close to zero over the accumulation period.

However, if we now consider a small carrier phase error at the RS receiver 34, the output of the baseband filter 44 will no longer be zero-mean at the sampling instants for received carrier phase probes. In turn, the output of the register 52 will now have a greater probability of either a low or a high output, so the accumulator 62 will

accumulate a value whose sign and magnitude reflect the sign and magnitude of the carrier phase error. In effect, this scheme uses the naturally occurring noise of the channel and receiver circuitry to dither the input to the one-bit quantizer (the thresholder 48), and then averages (with the accumulator 62) to give a value dependent on the input. Beyond a certain magnitude of carrier phase error, there may be insufficient noise to dither the error value, i.e. the output of thresholder 48 will be high for all the carrier phase probes, or low for all the carrier phase probes. This is effectively a form of "clipping" of the measure, and is still of use for steering control loops, or similar systems, so as to correct the error, as discussed below. Thus, we have a system and method that gives a measure of carrier phase error.

The measure from the accumulator 62 is fed back to the BS transmitter 22 via the RS transmitter 32 and the BS receiver 24. To save power, the RS transmitter 32 is switched off except for the short periods required to transmit the measure. At the BS 20, the data switch 244 directs the measure to a loop filter 252, which drives the frequency-control input of the numerically controlled oscillator 250, and which in turn outputs to the complex multiplier 228, acting as a mixer, that performs the required carrier phase correction. The loop filter 252, numerically controlled oscillator 250, and complex multiplier 228 are conventional elements of a carrier phase recovery control loop, as frequently seen in telecommunication receivers and familiar to those skilled in the art, so they are not described further herein.

We have considered the case where every carrier phase probe uses the complex amplitude 80 shown in Figure 2. However, regular insertion of a fixed amplitude pulse leads to a signal frequency spectrum with greater energy around DC. It is often desirable, for a number of technical or regulatory compliance reasons, to have a flat signal frequency spectrum. To flatten the spectrum, the probe insert unit 222 can alternate the carrier phase probe transmitted between the probe values 80, 82, 84 and 86 shown in Figure 2, according to a pseudo-random pattern. The probe quadrant control unit 64 must then use the same pseudo-random pattern to select either the I value or the Q value, by controlling which of the outputs of the registers 52, 54 is passed by the multiplexer 66, and to determine whether or not this output value should be inverted by the XOR gate 68, by controlling the binary value supplied to the second input of the XOR gate.

In a simple scheme, both the density of the carrier phase probes in the downlink symbol stream, and the rate of carrier phase error measurements from the probe measurement unit 40 may be constant. Alternatively, the system may detect when it has reached a stable state, in which the carrier phase error is small and varying only slowly, and reduce both the density of carrier phase probes, and the rate of carrier phase error measurements. This reduces the average power consumed by the RS transmitter 32, and makes more efficient use of the downlink channel, allowing further power savings in the RS receiver 34.

A similar technique is to employ an "alarm" strategy at the RS 30, such that the RS 30 only transmits the carrier phase error measure when its magnitude exceeds a predetermined threshold. The BS receiver 24 detects whether a transmission was sent, and if so receives the value of the measure. Again, this reduces the average power consumed by the RS transmitter 32.

A somewhat similar method can also be used to achieve symbol phase synchronization at the RS receiver 34. The method uses symbol phase probes, rather than carrier phase probes.

A symbol phase probe is a signal inserted in the transmitted signal, to allow the receiver to form a measure of symbol phase synchronization errors. Specifically, in this embodiment, the symbol phase probe comprises two special symbols that are inserted by the probe insert block 222, and transmitted by the BS transmitter 22. Figure 3 shows one possible arrangement. Symbols are usually transmitted at symbol timing points to, ti, ... etc. The two symbols 90, 92 forming the symbol phase probe are adjacent, with one being the inverse of the other, but they are not transmitted at a normal symbol timing phase. Rather, they are transmitted either side of a normal symbol timing point t ₄ , by an equal interval; for example, in Figure 3 the two symbols 90 and 92 are placed half a symbol period either side of the normal symbol timing point U-

The principle of operation, and the options for implementation, are very similar to the carrier phase probe method just described, so the discussion is not repeated in full. In addition to the carrier phase probes, the probe measurement unit 40 may also be used to measure the symbol phase probes. Alternatively, a second probe measurement unit

may be added, allowing carrier and symbol phase probes to be interleaved, and simultaneously accumulated.

The symbol phase probe method is based on the fact that the transmitted symbol phase probe signal crosses zero at point U in Figure 3. This corresponds to the ideal sampling instant for the symbol phase probe at the RS receiver 34. So, with no symbol phase error at the RS receiver 34, the output of the baseband filter 52 would ideally be zero, but in practice it will be zero-mean noise (from the communication channel, analogue circuitry etc.) that is averaged to a value close to zero by the accumulator 62. If the symbol phase is perturbed one way or the other, we effectively move up or down the curve 94 either side of the point t ₄ in Figure 3. Again, the noise effectively dithers the value on the curve, so, even after the one-bit quantization of the thresholder 52, a measure of the value (and hence the symbol phase offset) is averaged by the accumulator 62. Beyond a certain magnitude of symbol phase error, likewise, there may be insufficient noise to dither the error value and "clipping" will occur.

It will be noted than an equivalent alternative would be to transmit the two symbols forming the symbol phase probe at normal transmit timing points, for example t ₃ and t ₄ , or t ₄ and t ₅ , and to sample the received signal at a time midway between these two transmit timing points.

Similarly, the measure is fed back to the BS transmitter 22 via the RS transmitter 32 and the BS receiver 24. The data switch 244 directs the measure to a loop filter 242, which feeds the phase accumulator 240, and which in turn drives a sample phase input of the resampler 224 that performs the required symbol phase correction. The loop filter 242, phase accumulator 240, and resampler 224 are conventional elements of a symbol phase recovery control loop, as frequently seen in telecommunication receivers and familiar to those skilled in the art, so they are not described further herein.

As with the carrier phase probe method, the system may reduce both the density of the symbol phase probes, and the rate of symbol phase error measurements, when the symbol phase error is small and varying only slowly. Again, the "alarm" method may be used to transmit the symbol phase error measure only when it exceeds a predetermined threshold.

Since the symbol phase probes are not transmitted on the normal symbol timing points, ISI from the probes can affect adjoining data symbols. Therefore, it may be desirable to leave a short "guard" period before and after each symbol phase probe, containing no data symbols, to let the probe ISI settle to a low level. For example, in Figure 3, no data symbols are transmitted at the two normal symbol timing points t ₂ and t ₃ immediately preceding the symbol phase probe or at the two normal symbol timing points t ₅ and t ₆ immediately following the symbol phase probe.

Advantage may be gained from synchronizing the mix-down oscillator, or oscillators, in the receiver RF circuitry 42, and the clock signal driving the registers 52 and 54, by deriving them from the same source oscillator (not shown) in the RS receiver 34 for example. This allows just one type of phase probe, either carrier or symbol, to be used once the phases of both carrier and phase have initially been established. The BS transmitter 22 can use the measured changes in one to infer the other, since the ratio between the frequencies of the RS mix-down oscillator, or oscillators, and the clock signal driving the registers 52 and 54 is fixed and known. In particular, it is advantageous to use just the carrier phase probes, and associated measurements, for both the carrier and symbol phase tracking. In typical telecommunication systems, the carrier frequency is much higher than the symbol frequency, so carrier phase offset estimates infer very high accuracy symbol phase offset estimates. Also, each carrier phase probe takes only one downlink symbol, whereas each symbol phase probe takes several downlink symbol periods, including the ISI guard periods.

One issue with the carrier and symbol phase probe method described above is that no measure is made of the amount of noise added by the communication channel, analogue circuitry etc. As the level of noise decreases, the sensitivity of the reported measures increases (but the range of errors measurable, before "clipping", reduces). This is an issue for many control systems, as it effectively changes the gain of the error measure, which can affect the settling time, damping factor, or other properties of the control system. Therefore, it may be desirable to derive not just an average of the dithered and quantised offset, but also a measure of the noise magnitude. The system can then use this to adjust the gain of the error measure fed to the control system, mitigating these problems.

One technique for deriving the noise measure is to use a method we term double phase probing. We explain this method using carrier phase probes, but it is also readily extendible to symbol phase probes, as will be seen. Figure 4 shows the complex amplitudes of the carrier phase probes for this method. Rather than transmitting probe signals exactly on the decision boundaries between the standard constellation points, as in Figure 2, the probes are placed a small amount either side of, and equidistant to, the decision boundaries. Probes 100, 102, 104 and 106 in Figure 4 are referred to as positive probes, as their phase angle has a positive offset from the decision boundary; and similarly, probes 110, 112, 114 and 116 are referred to as negative probes.

To handle double phase probes, the system of Figure 1 is modified, as follows. A second probe measurement unit is added, in addition to the probe measurement unit 40. One probe measurement unit measures the positive probes, call the measure m _p ; the other measures the negative probes, call the measure m _n . An equal number of positive and negative probes are inserted by the probe insert unit 222 and transmitted by the BS transmitter 22. The accumulation periods for both positive and negative probes are chosen to be equal and coincident. In normal operation, it is seen that m _p > m _n , due to the phase offsets added to the positive and negative probes shown in Figure 4.

The two measures, m _p and m _n , are passed back to the BS 20, via the RS transmitter 32 and the BS receiver 24. The BS performs the addition m _p + m _n , yielding a single measure of phase offset, suitable to be passed into the loop filter 252, as before. It can be seen that, due to our formulation of the probe complex amplitudes in Figure 4, the carrier phase offset added to the positive phase probes cancels the carrier phase offset subtracted from the negative phase probes in this addition.

Additionally, the BS 20 performs the subtraction m _p - m _n , to give a measure that decreases with increasing noise amplitude. The noise measure m _p - m _n is then used to adjust the gain of the phase offset measure m _p + m _n entering the loop filter 252, such that, the lower the value of m _p - m _m the greater the gain applied.

One enhancement of the double phase probing method is for the probe insert unit 222 to dynamically increase the distance between the probes and the decision boundaries (shown in Figure 4) for lower values of the noise measure m _p - m _n . With greater

amounts of noise, a larger distance gives a more accurate measure of noise amplitude from the subtraction m _p - m _n . With less noise, a smaller distance avoids the "clipping" regions where there is insufficient noise to perform dithering.

The "alarm" method of reporting from the RS may be extended to accommodate double phase probing, such that the RS 30 only transmits the two measures m _p and m _n when one or more of the following occur:

(a) \m _p + m _n \ > k _a , (b) \m _p \ < k _b and \m _n \ < k _b> or

(c) \m _p \ > /c _c or \m _n \ > k _c

where k _g , k _b , and k _c are appropriately chosen thresholds, as is now explained. Case

(a) deals with the situation where the phase offset measure exceeds a low threshold, so the control system must react, as already described for single phase probing. Case

(b) deals with the situation where there is too much noise compared to the wanted signal, so both m _p and m _n are close to zero, giving a poor noise amplitude measure. Case (c) deals with the situation where either m _p or m _n are "clipping", or close to that condition.

The telecommunication system described so far has used QPSK. However, it can be seen that the system could alternatively be configured to use a lower-order modulation scheme, such as binary phase shift keying (BPSK), by removing the Q quadrature arm of the RS receiver 34. On-off keying (OOK) is another possibility, with symbol detection at either the thresholder 52 or 54. Another option is to use a higher-order modulation scheme, such as quadrature amplitude modulation (QAM) or m phase shift keying (m-PSK). In the case of QAM, the simple binary thresholders 52 and 54 must be replaced with multi-level quantizers. These may be ADCs, for example. Advantage is gained, compared to existing QAM receivers, in that lower resolution ADCs may be used, with as few quantization levels as the number of discrete amplitude signalling levels on I and Q. This allows reduction in the size, power-consumption and cost of the ADCs.

Another possibility is to employ offset quadrature modulation, such as offset QPSK (OQPSK) or offset QAM (OQAM). In these schemes, the Q component of the

transmitted signal is delayed by half a symbol, yielding a lower peak-to-average-power- ratio (PAPR) signal, which may be advantageous to the transmitter RF circuitry 238. This may be accommodated at the RS receiver 34 simply by delaying the symbol clock to the thresholder 54 in the quadrature detection arm by half a symbol.

It will be readily apparent to a person skilled in the art that other constellations and modulation schemes may be employed, embodying the same essential inventive ideas.

One issue for the RS receiver 34 is that it contains no means for equalizing the received signal to compensate for multipath or other convolutional effects in the communication channel. We now outline a method, suitable for a TDD telecommunication system, in which the equalization is performed at the BS transmitter 22 instead.

A predetermined signal, termed the channel sounding signal, is generated by a channel sounding sequence unit 72, and periodically transmitted by the RS transmitter 32. The channel sounding signal is received by the RS receiver 24, and the convolutional response of the channel is estimated by the channel estimator 248. Means for estimating such a channel response, in the time-domain or the frequency-domain, from a predetermined sounding signal, are well known, and thus will not be discussed further herein.

Conventionally, such a channel estimate might then be used to correct the uplink signal received at the BS receiver 24 for the convolutional effects in the channel. One option is, indeed, to carry out such channel correction in the BS receiver 24. For example, taps may be calculated for a finite impulse response (FIR) filter that is applied to the received signal, to correct the channel response. Again, methods for calculating the tap values are known, and are not detailed here. However, in embodiments of the invention, steps can also be taken to pre-correct the downlink signal transmitted by the BS transmitter 22. This is effected by the FIR filter 226, which takes tap values from the tap calculator 246, calculated as just described. Rather than being applied to the received signal, this filter is applied to the transmitted signal. This has the effect of pre- correcting the signal transmitted by the BS transmitter 22 so that the signal received at the RS receiver 34 is devoid of, or substantially devoid of, channel convolutional effects. The method exploits the principal of reciprocity, i.e. the fact that the

convolutional effect of the channel when transmitting from a first antenna to a second antenna is the same as the convolutional effect of the channel when transmitting from the second antenna to the first antenna. Therefore, in the present TDD telecommunication system, provided there is little change in the channel's convolutional response between the BS 20 receiving the last uplink channel sounding signal and transmitting the pre-corrected downlink, the desired channel correction is achieved.

Rather than, or in addition to, using a predetermined channel sounding signal, a channel estimate may be determined by the BS receiver 24 on the basis of received informational (non-predetermined) signals from the RS transmitter 32. Techniques for this are well known, for example, decision-directed algorithms. Again, the channel estimate may be used to pre-correct the signal transmitted by the BS transmitter 22, as above.

The FIR filter 226 can also be used to correct for any ISI due to the cascade response of the baseband filters 234 and 236, feeding into the filters 44 and 46, as mentioned earlier. Two methods are now outlined. The first method uses a fixed characterization of the cascade response. In this method, a corrective time-domain response, for a notional third filter in the cascade, is pre-calculated such that the total response for the cascade of the three filters meets the Nyquist criterion for zero ISI. Methods of performing this calculation will be apparent to persons skilled in the art. To effect the correction in the telecommunication system of Figure 1 , an additional FIR filter, whose response is that of the notional third filter, is inserted between the tap calculator 246 and the FIR filter 226.

In the second method, the cascade response of the baseband filters 234 and 44, or 236 and 46, is measured and corrected-for on an ongoing basis, as part of the channel response measurements and corrections described above. In this method, the RS transmitter 32 re-uses the baseband filters 44 and 46 as its transmit baseband filters, and the BS receiver 24 re-uses the baseband filters 234 and 236 as its receive baseband filters. (This is possible since TDD operation ensures that transmission and reception occur at different times at each end of the link.) Thus, the channel sounding signal is transmitted through the cascade of both baseband filters, as well as the communication channel. Therefore, the convolutional effect of the channel and transmit/receive baseband filters is seen as one combined convolutional effect by the

channel estimator 248, and the taps calculated by the tap calculator 246 correct for both types of convolutional effect together.

The telecommunication system of Figure 1 is shown as having just one BS and one RS. However, it can be seen that it is straightforward to extend the ideas presented so that one BS communicates with multiple RSs, with the BS maintaining a copy of the necessary state (loop filter state etc.) for each RS. Conventional time division multiple access (TDMA) or frequency division multiple access (FDMA) techniques may be used to accommodate signalling that shares a common communication medium between the multiple units. Similarly, the system may be extended to allow multiple BSs to use the communication medium.

The telecommunication system of Figure 1 shows just the sub-systems necessary to enable communication with the simplified RS receiver 34 presented. Clearly, other sub-systems may be added to the BS transmitter 22 and RS receiver 34, such as forward error correction (FEC), or automatic repeat request (ARQ).

There is thus described a telecommunication system with simplified receiver, which nevertheless allows for accurate synchronization of the receiver with the transmitted signals.