FREQUENCY HOPPING

Technical Field

The present invention relates to improving the block error rate associated with channel estimation in radio communication systems, particularly cellular-based systems such as Long Term Evolution (LTE) systems suited for Internet of Things applications.

Background

Throughout the course of the past few decades, the extent and technical capabilities of cellular-based radio communication systems have expanded dramatically. A number of different cellular-based networks have been developed over the years, including the Global System for Mobile Communications (GSM), General Packet Radio Services (GPRS), Enhanced Data rates for GSM Evolution (EDGE), and Universal Mobile Telecommunications System (UMTS), where GSM, GPRS, and EDGE are often referred to as second generation (or "2G") networks and UMTS is referred to as a third generation (or "3G") network.

More recently, the Long Term Evolution (LTE) network, a fourth generation (or "4G") network standard specified by the 3 ^rd Generation Partnership Project (3GPP), has gained popularity due to its relatively high uplink and downlink speeds and larger network capacity compared to earlier 2G and 3G networks. More accurately, LTE is the access part of the Evolved Packet System (EPS), a purely Internet Protocol (IP) based communication technology in which both real-time services (e.g. voice) and data services are carried by the IP protocol. However, while "classic" LTE connections are becoming increasingly prevalent in the telecommunications industry, further developments to the communication standard are being made in order to facilitate the so-called "Internet of Things" (loT), a common name for the inter-networking of physical devices, sometimes called "smart devices", providing physical objects that may not have been connected to any network in the past with the ability to communicate with other physical and/or virtual objects. Such smart devices include: vehicles; buildings; household appliances, lighting, and heating (e.g. for home automation); and medical devices. These smart devices are typically real-world objects with embedded electronics, software, sensors, actuators, and network connectivity, thus allowing them to collect, share, and act upon data. These devices may

communicate with user devices (e.g. interfacing with a user's smartphone) and/or with other smart devices, thus providing "machine-to-machine" (or "machine type") communication. However, the development of the LTE standards makes it more practical for them to connect directly to the cellular network.

3GPP have specified two versions of LTE for such purposes in Release 13 of the LTE standard. The first of these is called "NarrowBand loT" (NB-loT), sometimes referred to as "LTE Cat N B1 ", and the second is called "enhanced Machine Type Communication" (eMTC), sometimes referred to as "LTE Cat M 1 ". It is envisaged that the number of devices that utilise at least one of these standards for loT purposes will grow dramatically in the near future.

From a communications perspective, LTE standards (including NB-loT and eMTC) use orthogonal frequency division multiple access (OFDMA) as the basis for allocating network resources. This allows the available bandwidth to be shared between user equipment (UE) that accesses the network in a given cell, provided by a base station, referred to in LTE as an "enhanced node B", "eNodeB", or simply "eNB". OFDMA is a multi-user variant of orthogonal division multiplexing (OFDM), a multiplexing scheme in which the total bandwidth is divided into a number of non- overlapping sub-bands, each having its own sub-carrier frequency. In OFDM, unlike other frequency division multiplexing (FDM) schemes, each of these sub- carriers are orthogonal to one another such that cross-talk between sub-bands is ideally eliminated and removing the need for inter-carrier guard bands. In order to achieve this orthogonality, the spacing Af between the sub-carriers is set such that Af =— , where Τ _υ is the "useful symbol duration" (the receiver-side window size) and k is a positive integer (and is usually set to 1 ). Therefore with N sub-carriers, the total bandwidth B can be expressed as B = NAf. These sub- carriers are then shared between multiple users, thus providing multiple access. At the physical layer, in the downlink of an LTE connection, each data frame is 10 ms long and is constructed from ten sub-frames, each of 1 ms duration. Each sub-frame contains two slots of equal length, i.e. two 0.5 ms slots. Each slot (and by extension, each sub-frame and each frame) will typically contain a certain number of "resource blocks" (where each sub-frame has twice as many resource- blocks as a slot and each frame has ten times as many resource blocks as a sub- frame). A resource block is 0.5 ms long in the time domain and is twelve sub- carriers wide in the frequency domain. Generally speaking, there are seven OFDM symbols per slot and thus fourteen OFDM symbols per sub-frame. These resource blocks can be visualised as a grid of "resource elements", where each resource element is 1/14 ms long and one sub-carrier wide, such that there are eighty-four resource elements per resource block (i.e. seven multiplied by twelve) and one hundred and sixty-eight resource elements per sub-frame.

The exact number of resource blocks that exist in each slot (and by extension, each sub-frame and each frame) depends on the bandwidth configuration of the radio communication system. For example, in LTE eMTC Release 13, the LTE radio channel may have a bandwidth of 1.4 MHz, 3 MHz, 5 MHz, 10 MHz, or 20 MHz, where the channel is divided up into narrowbands of 1.4 MHz width.

As LTE eMTC and NB-loT are intended to operate in extremely low signal-to-noise ratio (SNR) conditions, the OFDM data symbols are typically repeated in a number of sub-frames. For example, in eMTC, each sub-frame may be repeated identically up to 2048 times.

As explained above, each resource element may be used to carry OFDM symbols over the Evolved UMTS Terrestrial Radio Access (E-UTRA) air interface. Typically this is to carry data, however this is not the only purpose served by these resource elements. In addition to conveying data, these resource elements may be used to carry "common reference symbols" (CRS), where these CRS are used to aid demodulation. In noisy conditions it may be necessary to average or filter a number of these CRS in order to obtain a suitable "channel estimate", i.e. an estimate of the characteristics of the channel such as attenuation, phase shifts, and noise. These CRS, or averaged CRS, are then used to aid the demodulation of the OFDM data symbols. In general, the averaging of the CRS should be carried out using a symmetrical sliding window in both time and frequency, i.e. CRS symbols from the past and the future relative to a given data symbol should be used to demodulate the data symbol (e.g. an average of the CRS symbols received in the previous five and following five resource elements relative to a resource element containing the data to be demodulated).

Typically, an LTE eNodeB is arranged to transmit CRS continually (e.g.

periodically), regardless of whether or not it has any data to send at any given time and regardless of which UE any such data is intended for.

In accordance with the standards set by 3GPP, many networks employ frequency hopping, where the transmission narrowband is changed periodically, typically following a pattern known in advance by the UEs. In the extreme case, the narrowband could be changed as often as every sub-frame. When the narrowband is changed, the UE typically uses a series of OFDM control symbols (that contain the CRS described above) reserved under LTE to tune its receiver (e.g. to tune a receiver frequency synthesizer) to the new narrowband. These control symbols (or "control channel symbols") are used by the network for the physical downlink control channel (or "PDCCH"). However, it should be noted that when the frequency changes, the eNodeB continues to transmit the CRS on the same frequencies as before.

However, the Applicant has appreciated that, by changing channel in synchrony with the control symbols, the performance of the receiver is less than optimal, particularly for the demodulation of OFDM data symbols that are close to the control symbols within the resource block. More specifically, when the radio changes between radio channels, the control symbols (or, more specifically, the CRS within the control symbols) immediately following the change in radio channel cannot be used to aid in the demodulation of data symbols occurring before the change in channel. The reverse is true for data symbols occurring before the radio change in channel; these cannot rely on control symbols from after the change in radio channel. The Applicant has appreciated that this results in the probability of errors in data symbols proximate in time to the radio channel changeover time being relatively high compared to data symbols elsewhere within the sub-frame, resulting in a higher block error rate (BLER) than is desirable.

Summary of the Invention

When viewed from a first aspect, the present invention provides a method of operating a radio receiver tunable to a plurality of radio channels and arranged to receive a plurality of sub-frames, wherein each sub-frame comprises a plurality of data symbols and a plurality of control symbols, said method comprising:

a) tuning the radio receiver to a first radio channel at an initial time;

b) receiving a first sub-frame;

c) tuning the radio receiver to a different radio channel after receiving at least one control symbol from a subsequent sub-frame but before receiving the remainder of the subsequent sub-frame;

d) receiving the remainder of the subsequent sub-frame;

e) tuning the radio receiver to a further different radio channel before receiving any control symbols or data symbols from a further subsequent sub- frame; and

f) receiving the further subsequent sub-frame.

This first aspect of the invention extends to a radio receiver tunable to a plurality of radio channels and arranged to receive a plurality of sub-frames, wherein each sub- frame comprises a plurality of data symbols and a plurality of control symbols, said radio receiver being arranged to:

a) tune the radio receiver to a first radio channel at an initial time;

b) receive a first sub-frame;

c) tune the radio receiver to a different radio channel after receiving at least one control symbol from a subsequent sub-frame but before receiving the remainder of the subsequent sub-frame;

d) receive the remainder of the subsequent sub-frame;

e) tune the radio receiver to a further different radio channel before receiving any control symbols or data symbols from a further subsequent sub-frame; and f) receive the further subsequent sub-frame. This first aspect of the invention further extends to a radio communication system comprising at least one radio transmitter and at least one radio receiver, wherein the radio transmitter is arranged to change between a plurality of radio channels and to transmit a plurality of sub-frames, and the radio receiver is tunable to a plurality of radio channels and arranged to receive a plurality of sub-frames, wherein the radio receiver is further arranged to:

a) tune the radio receiver to a first radio channel at an initial time;

b) receive a first sub-frame from the radio transmitter;

c) tune the radio receiver to a different radio channel after receiving from the radio transmitter at least one control symbol from a subsequent sub-frame but before receiving the remainder of the subsequent sub-frame;

d) receive the remainder of the subsequent sub-frame from the radio transmitter;

e) tune the radio receiver to a further different radio channel before receiving from the radio transmitter any control symbols or data symbols from a further subsequent sub-frame; and

f) receive the further subsequent sub-frame from the radio transmitter.

Thus it will be appreciated by those skilled in the art that the present invention provides, at least in preferred embodiments, an improved method for responding to changes in the transmission channel (e.g. through frequency hopping) that may reduce the impact on the demodulation of data symbols close to the control symbols. The approach described herein alternates between: i) sub-frames where the data symbols may be demodulated by using not only control symbol(s) from that sub-frame but also by using control symbol(s) from the next sub-frame; and ii) sub- frames where the data symbols are demodulated using only control symbols from that sub-frame.

As a result of the approach described, one or more data symbols in the sub-frames received in steps b) and f) that are close to the control symbol(s) that occurred before the change in radio channel may be demodulated with greater accuracy. This is because there are more control symbols (and thus, in LTE, more CRS) within the sliding window that can be averaged in order to characterise the channel and demodulate the data. It has been appreciated that this may mean that one or more data symbols in the sub-frame received in step d) that are close to the control symbol(s) that occurred before the change in radio channel may be more prone to errors. This is because there are fewer control symbols (and thus, in LTE, fewer CRS) on which to base the demodulation. However, the Applicant has appreciated that, as data symbols are typically repeated across a plurality of sub-frames, embodiments of the present invention may 'smooth out' the impact on the data symbols so that, in general, in some sub-frames the first one or more data symbols are more error prone while in other sub-frames the last one or more data symbols are more error prone. This may avoid the disadvantageous effect of the same symbols in each frame being affected.

The Applicant has appreciated that, in general, tuning the radio receiver to different frequencies (e.g. by re-tuning a frequency synthesiser within the receiver) takes some time to settle. For example, a typical frequency synthesiser may take the duration of one to two OFDM symbols after changing frequency to fully settle. As such, the CRS shortly after changing radio channel may actually be unusable in practice. In some preferred embodiments, step e) comprises tuning the radio receiver to a further different radio channel before receiving all of the control symbols and data symbols from the subsequent sub-frame. Those skilled in the art will appreciate that, in accordance with such embodiments, the method involves alternating between performing the frequency hopping early and performing the frequency hopping late compared to when it would be carried out conventionally.

If there are three or four control symbols (e.g. containing CRS) then there may not be a cost associated with doing this because the synthesiser can be tuned, for example, during the second and third symbols (or the third and fourth symbols) such that no data symbols are lost. Alternatively, if there are two control symbols and the synthesiser is capable of settling in the duration of a single symbol, there may also be no cost, because the synthesiser can be set to settle during the second symbol. However, if the synthesiser requires the duration of two symbols to settle, there may be a cost in that the data in the first symbol of the selection would be lost. Generally, if there is only a single control symbol, there is usually a cost as one or two data symbols are lost in the selection due to the settling of the synthesiser. The Applicant has also appreciated that more complex frequency hopping procedures where the method comprises alternating between performing the frequency hopping early, on time, and late compared to when it would be carried out conventionally could be carried out. Additionally or alternatively, the degree to which the frequency hopping is carried out early and/or late as appropriate could be varied.

The size of the sliding window used for demodulation will determine how many symbols are affected (positively or negatively) in the corresponding sub-frames.

Viewed another way, in one sub-frame, certain resource element positions may be more error-prone due to the timing of the frequency hopping but, in a subsequent sub-frame, those resource element positions are more reliable and different resource element positions are more error-prone instead. The net effect of this is that the BLER is improved.

In a preferred set of embodiments, at least one data symbol is identical in at least two of the sub-frames. The Applicant has appreciated that the invention may be particularly advantageous where sub-frames are repeated. This is because, if the frequency hopping (i.e. the change in radio channel) always occurs at the same time relative to a given data symbol and the frequency hopping occurs every sub-frame, the likelihood of an error occurring in the demodulation of that data symbol will not change, no matter how many times it is repeated. By alternating the timing of the change in radio channel within the receiver, different data symbols are affected in different repetitions of the sub-frame, so that, on average, there should be validly

demodulated copies of every data symbol, lessening the effect of the frequency hopping on any one individual data symbol.

As outlined above, the frequency hopping process is typically carried out continuously throughout operation and thus, at least in some embodiments, steps c) to f) are carried out periodically, i.e. step c) is carried out again after step f). It will of course be understood that when the steps are repeated, the "subsequent sub- frame", "further subsequent sub-frame", and "control symbol" are new instances and do not necessarily refer to the same sub-frames and control symbols as in the previous cycle of the steps.

While the rate at which the frequency hopping occurs could take any value, i.e. there may be any number of sub-frames between the subsequent and further subsequent sub-frames, in some embodiments the subsequent and further subsequent sub-frames are consecutive. In some potentially overlapping embodiments, the first and the subsequent sub-frames are consecutive. Such embodiments are preferred when the transmitter (typically a base station such as an LTE eNodeB) is arranged to perform frequency hopping every sub-frame.

However, in an alternative set of embodiments, the first sub-frame and the subsequent sub-frame are separated by one or more further sub-frames. Similarly, in embodiments where steps c) to f) are repeated periodically, there may be one or more sub-frames between the further subsequent sub-frame of step f) and the new subsequent sub-frame of step c) in the following cycle, or alternatively these sub- frames may be consecutive. The number of sub-frames separating the subsequent and further subsequent sub-frames may be the same between steps c) and f) and between steps f) and c) in embodiments where the frequency hopping is carried out periodically. Alternatively, the number of sub-frames separating the subsequent and further subsequent sub-frames may be asymmetrical such that a different number of sub-frames are received between steps c) and f) than between steps f) and c) in embodiments where the frequency hopping is carried out periodically. It will be understood by those skilled in the art that, depending on the design, there are a number of ways in which a radio receiver can be tuned to different radio channels. However, in at least some embodiments the radio receiver comprises a frequency synthesiser and step a) and/or step c) and/or step e) comprise tuning the frequency synthesiser.

Those skilled in the art will of course appreciate that, in practical systems, incoming signals of interest may actually be multiple resource blocks wide. For example, an incoming signal might be between one and six resource blocks wide, and these resource blocks need not necessarily be located adjacent one another. The principles of the claimed invention may be readily applied to radio receivers arranged to receive signals that are multiple resource blocks wide.

Brief Description of Drawings

Certain embodiments of the invention will now be described with reference to the accompanying drawings in which:

Fig. 1 is a block diagram of a typical LTE receiver;

Fig. 2 is a diagram of two typical, consecutive LTE sub-frames;

Fig. 3 is a diagram showing a prior art frequency hopping process;

Fig. 4 is a diagram showing a frequency hopping process in which the radio channel is changed after a control symbol from the next sub-frame is received in accordance with an embodiment of the present invention;

Fig. 5 is a diagram showing a frequency hopping process in which the radio channel is changed before a control symbol from the next sub-frame is received in accordance with an embodiment of the present invention;

Fig. 6 is a diagram showing a frequency hopping process in which frequency hopping is carried out every sub-frame with alternating channel change timing.

Detailed Description

Fig. 1 is a block diagram of a typical LTE receiver 100, known in the art per se. The receiver 100 comprises: an antenna 102; a low-noise amplifier (LNA) 104; a frequency synthesiser 106; a mixer 108; and a demodulation module 1 10. Those skilled in the art will appreciate that this is a highly simplified overview of a practical system, which would typically contain many subsystems.

Signals received via the antenna 102 are passed through the LNA 104 which amplifies the signals before inputting them to the mixer 108. The mixer 108 also takes as an input a signal produced by the frequency synthesiser 106 in order to down-mix the received signals from the transmission frequency to baseband frequency suited to further processing by the demodulation module 110. In order to change between different radio channels, for example during a frequency hopping process, the frequency synthesiser 106 can be tuned to change the frequency of the signal it produces that is input to the mixer 108 to match the frequency being used by the transmitter (e.g. an eNodeB).

Fig. 2 is a diagram of two typical, consecutive LTE sub-frames as used in LTE communications. As explained previously, in accordance with the LTE standard, each LTE data frame is 10 ms long and is constructed from ten sub-frames, each of 1 ms duration. Fig. 1 shows two consecutive sub-frames 2, 4, where each sub- frame 2, 4 contains two respective slots 2a, 2b; 4a, 4b of equal length, i.e. each sub-frame has two 0.5 ms slots. Thus the first slot 2a of the first sub-frame 2 starts at t ₀ and ends at t ₀ + 0.5 ms, the second slot 2b of the first sub-frame 2 starts at t ₀ + 0.5 ms and ends at t ₀ + 1 ms, the first slot 4a of the second sub-frame 4 starts at t ₀ + 1 ms and ends at t ₀ + 1.5 ms, and the second slot 4b of the second sub-frame 4 starts at t ₀ + 1.5 ms and ends at t ₀ + 2 ms.

Each slot 2a, 2b, 4a, 4b contains a certain number of "resource blocks". A resource block is 0.5 ms long in the time domain and is twelve sub-carriers 6 wide in the frequency domain. Generally speaking, there are seven OFDM symbols 7 per time- domain slot 2a, 2b; 4a, 4b and thus fourteen OFDM symbols 7 per sub-frame 2, 4. These resource blocks can be visualised as a grid of "resource elements", where each resource element is 1/14 ms long and one sub-carrier wide, such that there are one hundred and sixty-eight resource elements per sub-frame (i.e. fourteen multiplied by twelve).

Each frame 2, 4 is further split into a control channel portion 8 and a payload portion 10. The control channel portion 8, which has a duration of the first two resource elements (i.e. the first 2/14 ms), is typically used by the UE to tune its receiver frequency synthesiser to the current narrowband frequency. It will of course be appreciated that, in practice, the control channel portion 8 may have a duration of between one and four OFDM symbols.

The payload portion 10 contains the data itself such that a number of the resource elements in each frame 2, 4 contain only data symbols - i.e. the data that has been converted into OFDM symbols and transmitted by the eNodeB, over-the-air, to the UE. These data symbols are depicted in the accompanying drawings as white blocks, and an exemplary resource element carrying data is labelled 12 in Fig. 2.

A number of common reference symbols (CRS) are distributed throughout both the control channel portion 8 and payload portion 10. These CRS are used to aid demodulation of the OFDM data symbols. In general, NB-loT and eMTC are expected to operate in very low SNR conditions and so it may be necessary to average or filter a number of these CRS in order to obtain a suitable channel estimate, i.e. an estimate of the characteristics of the channel such as attenuation, phase shifts, and noise. These CRS are depicted in the accompanying drawings as both black blocks and as blocks with vertical stripes (the difference between these depictions is explained in detail below), and exemplary CRS resource elements are labelled 14 in Fig. 2. These CRS 14 are used to demodulate the OFDM data symbols. In general, the averaging of the CRS 14 would be carried out using a symmetrical sliding window in both time and frequency, i.e. CRS 14 from the past and the future relative to a given data symbol are used to demodulate the data symbol (e.g. an average of the CRS 14 received in the previous five and following five resource elements relative to a resource element containing the data to be demodulated). In the examples given here, a sliding window is selected such that, for a given selection 16 two data symbols wide, all CRS 14 in a window 18 covering the previous five symbols and any CRS 14 in a window 20 covering the next five symbols are used to demodulate the selection 16 of data symbols. For ease of reference, the CRS 14 that are used to demodulate the selected data 16 are those that are shown in solid black, while the CRS 14 that are not used are shown with vertical stripes. It can be seen therefore that, for the selection of data 16, there are a total of eight CRS 14 that are used during demodulation. Fig. 3 is a diagram showing a prior art frequency hopping process. In this example, the transmission channel (i.e. the narrowband) changes every sub-frame. At a first time t ₀ , LTE communication is carried out over a first narrowband 20 (twelve sub- carriers wide as before). After 1 ms (i.e. the duration of a single sub-frame in LTE), once a first sub-frame 22 has been transmitted, the eNodeB changes its

transmission frequency according to a predetermined pattern and the UE retunes its frequency synthesiser to a second narrowband 24 at t ₀ + 1 ms. Those skilled in the art will of course appreciate that this is only a single example, and in practical systems the signal of interest may actually be multiple resource blocks wide. For example, an incoming signal might be between one and six resource blocks wide, and these resource blocks need not necessarily be located adjacent one another. For the sake of simplicity, the examples shown here are only a single resource block wide, however it should be understood that the principles outlined herein apply equally to signals of any width in terms of resource blocks. Similarly, after a further 1 ms, once a second sub-frame 26 has been transmitted, the eNodeB changes its transmission frequency again and the UE retunes its frequency synthesiser to a third narrowband 28 at t ₀ + 2 ms before receiving a third sub-frame 30. However, the Applicant has appreciated that this typical frequency hopping process is detrimental to some of the data symbols, in particular the last few data symbols in each sub-frame 22, 26, 30. For example, taking a selection 32 of the last two data symbols in the first sub-frame 22, the channel is changed immediately after receiving the last symbol in the sub-frame and so there are no CRS 14 after this selection of data symbols 32. Therefore this selection of data symbols 32 can only benefit from the CRS 14 within the window 34 preceding this selection 32 for demodulation. This results in the data symbols within this selection 32 being demodulated based on reduced channel estimate information, as only four CRS 14 are within the window 34, causing the average of these to be more susceptible to errors.

Meanwhile, a selection 36 of the first two data symbols in the payload portion of the second sub-frame 26 have access to CRS 14 within both a window 38 before and a window 40 after this selection of data symbols 36. However, as the frequency synthesiser 106 takes some time to settle (for example the duration of one to two OFDM symbols) after hopping frequency, the CRS 14 within the window 38 may actually be unusable in practice. In this particular example, the symbols in the window 38 are unusable because the synthesiser 106 has a one symbol long settling time, where the symbols during the settling time are depicted in the accompanying drawings by wavy lines. Therefore while these data symbols could, in theory, be demodulated using six CRS 14, in practice these data symbols are typically restricted to using only the four CRS 14 in the later window 40 for demodulation. This problem equally applies for the last few data symbols in each sub-frame 22, 26, 30 - changing the channel between sub-frames necessarily results in a degradation in quality for the demodulation of the last few symbols, which has a negative impact on the UE's ability to deal with low SNR conditions. However, as it can be seen, there is a disparity between the confidence in the channel estimate used for demodulating symbols that are positioned earlier in the sub-frame compared to those that come later.

Fig. 4 is a diagram showing a frequency hopping process in accordance with an embodiment of the present invention, in which the radio channel is changed after a control symbol from the next sub-frame is received.

Similarly to the process described with reference to Fig. 3, at time t ₀ , LTE communication is initially carried out over the first narrowband 20. After 1 ms (i.e. the duration of a single sub-frame in LTE), once a first sub-frame 22 has been transmitted, the eNodeB changes its transmission frequency according to a predetermined pattern. However, in contrast to the conventional process described above, the UE does not retune its frequency synthesiser 106 at this time. Instead, the UE waits before switching to the second narrowband 24 at t _SW itc _h i which, in this case, is 1/14 ms later than the conventional switching time. As explained previously, the CRS 14 are transmitted on the same frequencies as before, even after the frequency hopping process takes place (i.e. after the network changes the frequency on which data signals are transmitted).

As a result of this, a selection of the last two data symbols 42 may be demodulated using channel estimate information obtained by averaging CRS 14 from both a window 44 before and a window 46 after the selection 42. This results in this selection of the last two data symbols 42 having access to a total of six CRS 14 for the demodulation, where two of these have been taken from the following sub- frame 26. Furthermore, it should be noted that demodulation of the data symbols within the window 44 is also improved as these can use an average of eight CRS 14 rather than only six.

Theoretically, doing this comes at a cost - by taking two of the CRS 14 from the following sub-frame 26, a selection 48 of the first two data symbols of the second sub-frame 26 now only have access to four CRS 14 for demodulation. However, as outlined above, the frequency synthesiser 106 takes some time to settle (e.g. the duration of one to two OFDM symbols) after hopping frequency. Those skilled in the art will appreciate therefore that, in practice, the first two data symbols of the second sub-frame 26 may in reality only have had access to four CRS 14 for demodulation anyway and so there may actually be no perceivable cost in reality. This will of course depend on the number of CRS 14 and the settling time of the frequency synthesiser as explained below. If there are three or four control symbols (i.e. containing CRS 14) then there is no cost because the synthesiser 106 can be tuned during the second and third symbols (or the third and fourth symbols) such that no data symbols are lost.

Alternatively, if there are two control symbols and the synthesiser 106 is capable of settling in the duration of a single symbol, there is also no cost, because the synthesiser 106 can be set to settle during the second symbol. However, if the synthesiser 106 requires the duration of two symbols to settle, there is a cost in that the data in the first symbol of selection 48 would be lost. If there is only a single control symbol, there is always a cost as one or two data symbols are lost in the selection 48 due to synthesiser 106 settling.

In this particular example, the settling time 47 of the frequency synthesiser 106 is that of a single symbol. As such, the data symbols in the selection 48 would only ever have had access to the CRS 14 in the window 50 after the selection 48 as the earliest two CRS (the resource elements labelled 57) would have normally been received during the settling time (i.e. if the timing of the switch t _SW itc _h had occurred at the normal time t ₀ + 1 ms). Thus these first two CRS 57 may not have been useful for demodulation of the data symbols in the selection 48 anyway, and thus there is no cost associated with performing the frequency hopping one symbol late as shown in Fig. 4. Fig. 5 is a diagram showing a later part of the frequency hopping process. Now, however, similarly to the conventional process described with reference to Fig. 3, here the change in narrowband is carried out by the UE before the next CRS 14 (i.e. from the next sub-frame) are received. Unlike the conventional process of Fig. 3 however, the change in narrowband is carried out one symbol earlier than it would have been using the conventional process. It should be noted that the terms 'early' and 'late' are used in relation to when the conventional process would have instigated the change in narrowband, and this reference point (i.e. the 'normal' time for changing narrowband) is not altered by the timing of the previous change being early or late.

By changing narrowbands at t _SW itch2, which is set to be t ₀ + 2ms - 1/14 ms (i.e. by starting the switch one symbol early), the selection of data symbols 50 at the end of the second sub-frame 26 only have access to the four CRS 14 in the window 52 preceding the selection 50. This comes at the cost of losing any data symbols within the settlement duration 55. While this may appear to be a significant loss, the Applicant has appreciated that this is not an issue in practical communication protocols such as LTE where data symbols are repeated many times over different sub-frames. The data symbols may be "turbo encoded", where those skilled in the art will appreciate that turbo codes are a type of forward error correction codes that provide a mechanism to 'fill in the blanks' during demodulation.

However, performing the switch one symbol early (i.e. at t ₀ + 2ms - 1/14 ms) advantageously ensures that the settlement duration 55 has passed before the beginning of the second sub-frame 30, a selection 54 of the first two data symbols in the payload portion of the third sub-frame 30 have access to CRS 14 within both a window 56 before and a window 58 after this selection 54 of data symbols. These data symbols may therefore be demodulated using channel estimate information obtained from six CRS 14.

Fig. 6 is a diagram showing a frequency hopping process in which frequency hopping is carried out every sub-frame with alternating channel change timing. Fig. 6 combines the frequency hopping steps shown in Figs. 4 and 5 and illustrates how the process results in alternation between the first and last symbols as to which is impacted by the change in narrowband. Over the course of time, this averages out the detrimental effects such that the last data symbols are not as adversely affected as they would be with the conventional frequency hopping process described previously with reference to Fig. 3. It will of course be appreciated that the alternation between changing before or after receiving CRS from the next sub-frame does not necessarily need to be performed every sub-frame. For example, a frequency hopping scheme could involve switching narrowband early (i.e. before receiving any CRS from the next sub-frame as described with reference to Fig. 5) N out of every M sub-frames and switching narrowband late (i.e after receiving the first few CRS from the next sub-frame as described with reference to Fig. 4) in the other (M-N) out of every M sub-frames.

The selection of the values of N and M could be based on the modulation and coding scheme (CMS) being used, where the CMS is typically signalled to the UE by the eNodeB prior to the eNodeB transmitting the sub-frames that contain data. When the network is using 'weaker' coding and/or modulation schemes (e.g. a network opting to use 16-QAM instead of QPSK), the Applicant has appreciated that, in general, it may be safer to change narrowband late more often than early (i.e. by reducing N and thus increasing M-N) so as to reduce the loss rate of data symbols where the coding scheme is less tolerant of such losses. However, with 'stronger' coding and/or modulation schemes that are more tolerant of data symbol losses, the relative value of N can be increased, thus decreasing M-N.

The Applicant has appreciated that it is possible to have a more complex arrangement that involves alternating between performing the frequency hopping early, on time, and late compared to when it would be carried out conventionally, with different weightings for each relative timing.

Thus it can be seen that this results in the position of the data symbols that are negatively impacted being "spread out", such that sometimes data symbols toward the beginning of the sub-frame are negatively impacted and at other times the data symbols toward the end of the sub-frame are negatively impacted. As data symbols are repeated across many sub-frames, an overall improvement to the reliability of the demodulation can be achieved. Thus it will be appreciated by those skilled in the art that embodiments of the present invention provide an improved method for responding to changes in the transmission channel that reduces the impact on the demodulation of data symbols close to the control symbols. Those skilled in the art will appreciate that the specific embodiments described herein are merely exemplary and that many variants within the scope of the invention are envisaged.