GB2410859A - Synchronising a receiver to an ultra wideband signal - Google Patents

Abstract

A receiver is synchronised to a received ultra wideband (UWB) signal r(t) having a known hopping code. A parameter of the received signal, such as a correlation 1 of the received signal r(t) with a generated template signal 6, is determined in a number of time slots. The parameter associated with each time slot is added 14 to one of a number of corresponding time and code offset accumulators. The code and time offset accumulator having an accumulated value according to a predetermined accumulator metric, preferably the largest value, is determined to provide the time and code offset required to synchronise the receiver. The invention may be used in various UWB applications, such as radar, position tracking and communications, and specifically in ground penetrating radar (GPR) and wireless personal area networks (WPAN).

Description

1 24 10859 Ultra Wide Band (UWB) Synchronisation Search

Field of the Invention

The present invention relates to ultra wide band (UWB) applications such as radar, position tracking and communications for example Ground Penetrating Radar (GPR) and Wireless Personal Area Networking (WPANs).

Background of the Invention

A popular method for generating Ultra-Wideband (UWB) Signals uses a technique known as Impulse Radio (JR) as described in Moe Z. Win and Robert A. Scholtz, Impulse Radio: How it Works', IEEE Communications Letters, Vol. 2, No. 2, February 1998, and Moe Z. Win and Robert A. Scholtz, 'Ultra-Wide Bandwidth Time- Hopping Spread-Spectrum Impulse Radio for Wireless Multiple-Access Communications', IEEE Transactions on Communications, Vol. 48, No. 4, April 2000. This technique involves using a low duty cycle pseudo randomly time hopped pulse train of short (subnanosecond) mono-cycles to create a signal with an Ultra Wide-Band spectrum. In a UWB communications system the receiver needs to have an accurate idea of when the pulses will arrive in order to receive the signal correctly, i.e. the receiver has to be synchronised to the received pulse train.

The combination of the low duty cycle pulse train and the pseudo random time hopping code means that it can be difficult and time consuming to find the signal. One of the most reliable method of searching for the signal is to choose a time offset resolution which is smaller than the pulse and perform an exhaustive search for all time offsets and all hopping code offsets. However, this search takes the longest to perform and may not be suitable for some applications. Some searches which examine a subset of the search space and use a threshold technique have been proposed but these methods can exhibit low performance in certain scenarios and have a variable search time which can approach that of the exhaustive search in worst case situations.

Current technology employs a single Time-Integrating Correlator (TIC) as described in Burke, B.E.; Smythe, D.L.; 'A CCD time-integrating correlator', Solid-State Circuits, IEEE Journal of, Volume: 18 Issue: 6, Dec 1983 Page(s): 736 -744. Typically the pulse train is found by triggering the TIC for all time and code offset combinations to perform an exhaustive search for the signal as described in Yao Ma; Chin, F.; Kannan, B.; Pasupathy, S.; 'Acquisition performance of an ultra wide-band communications system over a multiple-access fading channel', Ultra Wideband Systems and Technologies, 2002. Digest of Papers. 2002 IEEE Conference on, 2002 Page(s): 99 - 103. Some methods use a reduced search area combined with a threshold operation as described in Homier, E.A.; Scholtz, R.A.; 'Rapid acquisition of ultra-wideband signals in the dense multipath channel', Ultra Wideband Systems and Technologies, 2002.

Digest of Papers. 2002 IEEE Conference on, 2002 Page(s): 105 -109.

A system which aims to acquire fast acquisition of UWB IR time hopped sequences by ensuring that all absolute time differences between adjacent pulses is unique is described in Leanard S. Haynes and Mark D. Roberts, "Method and System for Fast Acquisition of Pulsed Signals", US Patent 2002/0018514, February 2002. However this method seems impractical in the presence of multi-path and multiple access interference. A simple way of reducing the amount of time taken to perform synchronization is to not observe the code for its full length but to only observe a smaller subset as described in Larry W. Fullerton, "Fast locking mechanism for channelized ultrawide-band communications", US Patent 5,832,035, November 1998.

The problem with using an exhaustive search is that, although the performance is likely to be the best, the search times can become unacceptably high for some applications.

Some methods have been proposed to decrease the time by using a reduced search in combination with a threshold function. However, choosing the correct threshold for dynamic, and unknown, signal to noise ratio (SNR) and Multiple Access Interference (MAI) scenarios may be difficult which will compromise performance. Also these methods, although reducing the average search time, can exhibit high instantaneous search times (approaching that of the exhaustive search time) in worst case conditions.

Moe Z. Win and Robert A. Scholtz, "Impulse Radio: How It Works", IEEE Communications Letters, Vol. 2, No. 2, February 1998 describes impulse radio, a form of ultra-wide bandwidth (UWB) spread-spectrum signalling, that has properties that make it a viable candidate for short-range communications in dense multipath environments. This paper describes the characteristics of impulse radio using a modulation format that can be supported by currently available impulse signal technology and gives analytical estimates of its multiple-access capability under ideal multiple-access channel conditions.

Moe Z. Win and Robert A. Scholtz, "Ultra-Wide Bandwidth Time-Hopping Spread- Spectrum Impulse Radio for Wireless Multiple-Access Communications", IEEE Transactions on Communications, Vol. 48, No. 4, April 2000 describes attractive features of time-hopping spread-spectrum multiple-access systems employing impulse signal technology, and emerging design issues are described. Performance of such communications systems in terms of achievable transmission rate and multiple-access capability are estimated for both analog and digital data modulation formats under ideal multiple- access channel conditions.

Yao Ma; Chin, F.; Kannan, B.; Pasupathy, S.; "Acquisition Performance Of An Ultra Wide-Band Communications System Over A Multiple-Access Fading Channel", Ultra Wideband Systems and Technologies, 2002. Digest of Papers. 2002 IEEE Conference on, 2002 Page(s): 99 -103 describes how ultra wideband time-hopping communications are expected to be a practical scheme in the near future, however that the acquisition analysis of the UWB signals over multipath channels has not been adequately addressed.

Homier, E.A.; Scholtz, R.A.; "Rapid Acquisition of Ultra-Wideband Signals in the Dense Multipath Channel", Ultra Wideband Systems and Technologies, 2002. Digest of Papers. 2002 IEEE Conference on, 2002 Page(s): 105 -109 describes efficient serial search strategies which are shown to reduce drastically the mean acquisition time for UWB signals in a dense multipath environment. Inherent in traditional serial search problems is the assumption that only a single bin or a small number of consecutive bins can properly terminate the search. This assumption leads to search strategies which tend to be linear in nature, e.g., a linear sweep of the uncertainty region. Because of the dense multipath channel present in most UWB systems, this assumption is invalid as seen by the channel's relatively large delay spread. A generalized analysis of various search algorithms is presented based on a Markov chain model of a simple singledwell serial search. The results from this analysis reveal that the linear search has a considerably larger mean acquisition time than the more efficient search strategy termed the bit reversal search.

Burke, B.E.; Smythe, D.L.; "A CCD Time-Integrating Correlator", SolidState Circuits, EKE Journal of, Volume: 18 Issue: 6, Dec 1983 Page(s): 736 -744 describes a CCD binary-analog time-integrating correlator that has been designed and operated at 20 MHz clock rate. The 32-channel device is capable of integration periods in excess of 25 /spl mu/s or 500 clock periods, equivalent to a time-bandwidth product of 250. The device architecture is based on charge-domain signal processing for high-speed operation and does not required on-chip logic for storage of the binary reference. The device is tailored for weak signal applications, and a new charge skimming circuit has been devised which allows the small portion of the integrated charge containing the correlation function to be separated from the large register by tenfold. The correlator has a stationary pattern noise which can be eliminated with simple postprocessing, yielding a dynamic range of 67 dB.

Leanard S. Haynes and Mark D. Roberts, "Method and System for Fast Acquisition of Pulsed Signals", US Patent 2002/0018514, February 2002 describes a method and system for fast acquisition of pulsed radio signals where said signals comprised uniquely-spaced pulses or symbols. The receiver searches for and detects two ore more symbols, determines the interval or intervals between the symbols, and because the intervals are unique, the symbols or symbol sequences are identified. Identification of the symbols and their respective timing within the signal allows the receiver to adjust its timing to synchronize with the full signal.

Larry W. Fullerton, "Fast locking mechanism for channelized ultrawideband communications", US Patent 5,832,035, November 1998 describes a receiver for acquisition and lock of an impulse radio signal comprising an adjustable time base to output a sliding periodic timing signal having an adjustable repetition rate, and a decode timing modulator to output a decode signal in response to the periodic timing signal.

The impulse radio signal is cross correlated with the decode signal to output a baseband signal. The receiver integrates T samples of the baseband signal and a threshold detector uses the integration results to detect channel coincidence. A receiver controller stops sliding the time base when channel coincidence is detected. A counter and extra count logic, coupled to the controller, are configured to increment or decrement the address counter by a one or more extra counts after each T pulses is reached in order to shift the PN code module for proper phase alignment of the periodic timing signal and the received impulse radio signal.

Summary of the Invention

In general terms in one aspect the present invention provides a nonconsecutive search and global maximum-selection and threshold crossing (MAX/TC) algorithm for the acquisition of a UWB signal over a multipleaccess multipath fading channel. In one embodiment, correlator output statistics are analyzed in order to determine code and timing offsets of the receiver relative to the received signal, and provides acceptable robustness to multiple interferes, multi-path fading and noise. The method aims to decrease the search time when the absolute time reference (in terms of nominal pulse arrival time and code position) is unknown without significantly reducing the signal acquisition performance.

In particular in one aspect the present invention provides an apparatus for synchronizing a UWB receiving according to claim 1.

In particular in another aspect the present invention provides a method of synchronizing a receiver to a received UWB signal comprising a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to a time-hopping code; the method comprising: observing the signal; determining a parameter of the signal over a number of time slots corresponding to portions of the receivers time frame; for each determined parameter, adding said parameter to an accumulation of said parameters associated with said time slot each accumulation for the time slot corresponding to each possible signal frame offset and associated code value combination for said time slot; outputting estimated frame offset and code values having the highest accumulated parameter as the time frame offset and associated code values of the received signal compared with the receiver time frame.

For the purposes of this specification, a pulse here is a finite length waveform with zero or more zero crossings. This includes for example an impulse, a mono- cycle and a poly- cycle.

The parameter could be a representation of the observed signal, for example a snap-shot of a pulse received in the time slot. This is a capture of the received waveform for the duration of a pulse. This could take the form of an ADC and some memory to save the waveform. The snap shots could then be intergrated with the contents of the appropriate accumulations.

Preferably the parameter is the correlation of the observed signal and a locally generated template signal having a known waveform matching the expected (training or synchronizing) signal from the transmitter; although being offset in time and/or code value from the transmitter signal. This saves the need for a set of samples to be stored per observation, and instead a single value (the correlation) only needs to be saved making it more efficient.

More preferably the template signal is merely the part of this signal at the time bin currently being correlated. Thus the parameter determining means comprises a template signal generator for providing a template signal at a second time offset, and means for correlating said template signal with said received signal in order to provide said parameter. Where the template signal is a pulse, this preferably corresponds to a pulse having a shape corresponding to a received pulse. In the embodiments, the second time will either correspond to a time bin position which the embodiment cycles through over a number of frames, or a time bin position corresponding to one of a number of suitably offset correlators; or a combination.

Preferably correlation is performed using a TIC which will produce a peak or high output if the template signal is a pulse corresponding with the received signal pulse.

Each output of the TIC is added to previous TIC output results for the appropriate combinations of code offset and time offset which correspond to possible code and time offset combinations for the received signal. This will typically be achieved by using a memory location for each code and time offset combination, to which the correlation is added cumulatively. This requires less memory than the snap shot approach.

By adding the parameters of the signal to a number of signal frame offsets and code value combinations rather than just to one, the correct code and time offset combination will be found since the large parameter values (eg pulses or high correlations) will only sum coherently at the correct time and code offset combination. All the other locations will have a large parameter value (eg pulse) integrated into it once in a while, whereas the correct location will consistently be updated. This saves time by interpreting each correlator output in multiple ways, which reduces the search space, by relying on the coherent summation of received pulses at the correct code and time offset location.

As discussed above, known algorithms tend to run through each possible frame offset and code offset in order to determine which most closely matches the received signal.

This is time consuming and therefore lengthens the receivers synchronization time.

Embodiments of the present invention recognise that each received signal pulse or mono-cycle could correspond to one of a number of frame and code offset combinations. Therefore, for each observed time offset, a parameter such as correlation with a locally generated signal is stored and accumulated for each of the possible frame and code offset combinations that the observed time offset could correspond to. After observing a given number of time offsets, the frame and code offset combination having the largest parameter accumulation is said to represent the best offset values for the receiver to operate with in receiving the traffic or data signals.

Preferably the integrating step comprises integrating the correlation at each said offset.

This is easier to implement, however in an alternative implementation the correlation may only be added to the corresponding time and code offset combinations when it is the result of a coincidence of the received and generated pulse. This situation may be deemed to have occurred if the correlator output is above a certain threshold value for

example.

In particular in another aspect the present invention provides a method of synchronizing a receiver to a received UWB signal transmitted over a series of time frames each having a pulse, the relative position of the pulse in each transmitter time frame according to a code applied to the series of frames; the method comprising: observing the signal; determining a parameter of the signal over a number of time slots corresponding to a receiver time frame, the receiver time frame offset from said transmitted time frame by a frame offset value; for each determined parameter, adding said parameter to a plurality of accumulations of said parameters associated with said time slot, each said accumulation for the time slot corresponding to a number of different signal frame offset and associated code value combinations; outputting estimated frame offset and code values having the highest accumulated parameter as the time frame offset and associated code values of the received signal compared with the receiver time frame.

In particular in another aspect the present invention provides a method of synchronizing a receiver to a received UWB signal comprising a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to a time-hoping code applied to the series of frames; the method comprising: observing the signal; determining a parameter of the signal over a number of time slots; for each time slot, adding said parameter to a number of accumulations of said parameters associated with said time slot, each said accumulation for a possible receiver signal frame offset and associated code value offset combination corresponding to said time slot; outputting estimated receiver time frame offset and code offset values having the highest accumulated parameter as the receiver time frame offset and associated code offset values for the receiver compared with the observed signal.

Preferably the parameter is the correlation between the received signal and a corresponding signal locally generated according to the receiver time frame. This may be found using time-integrating correlators (TIC) for example.

In one embodiment there is a single TIC, such that a minimum of hardware is required.

However a number of frames equal to the code length is the minimum theoretical length of time required for synchronization. This does not include any time required for verification logic or the like. A second embodiment uses a higher number of TIC's, on per code, in order to speed up the synchronization process, which can then be finished in one frame with suitable offsets for each TIC. Various intermediate possibilities could now alternatively be implemented by those skilled in the art.

In another aspect there is provided a method of synchronizing a receiver to a received UVVB signal comprising a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to a time-hopping code; the method comprising: receiving the signal; generating a template signal corresponding to the received signal; correlating the received signal and the generated signal over a number of time offsets; determining corresponding code and time offset combinations for each said time offset; integrating the correlation at a said time offset with a value for each said corresponding code and time offset combination; determining the code and time offset combination having the highest accumulated value in order to synchronise the receiver.

There is also provided corresponding apparatus and computer software for implementing the methods defined in this specification.

Brief Description of the Drawings

Embodiments are described with reference to the following drawings, by way of example only and without and without intending to be limiting, in which: Figure 1 shows a Gaussian mono-cycle in the domain; Figure 2 shows a Gaussian mono-cycle in the frequency domain; Figure 3 shows the frequency spectrum of fixed interval Gaussian mono-cycle pulse train; Figure 4 shows an example UWB IR signal; Figure 5 shows the frequency spectrum of the pseudo random time hopping Gaussian mono-cycle hopping pulse train; Figure 6 illustrates a UWB IR front end correlator; Figure 7 shows a UWB IR sync search space; Figure 8 shows example results when the code offset is correct; Figure 9 shows example results when the code offset is incorrect; Figure 10 shows a graph of frame time versus time taken to search all time offsets; Figures 11 and 12 show respectively the second and third derivative of a Gaussian mono-cycle; Figures 13 and 14 show respectively the auto-correlation of the second and third derivatives of a Gaussian mono-cycle shown in Figures 1 1 and 12 respectively; Figure 15 illustrates an observation corresponding to a number of different time bin and code offset combinations; Figure 16- 19 are example codes; Figure 20 is an example of a bad code where the difference between adjacent codes is always S.; Figure 21 shows an implementation architecture for scheme 1 and a single correlator; Figure 22 shows the update memory and compare stage of Figure 21 in more detail; Figure 23 shows an implementation architecture for scheme 1 and multiple correlators; Figure 24 shows the update memory and compare stage of Figure 23 in more detail; Figure 25 shows an implementation architecture for scheme 2 and multiple correlators; Figure 26 shows simulation results for Probability of Detection Pa versus Eb/No; Figure 27 shows simulation results for coverage time taken versus Eb/No; Figure 28 shows simulation results for the average signal size found relative to the maximum value versus Eb/No; and Figure 29 shows simulation results illustrating the degradation in performance using "bad" codes.

Table 1 shows an example section of a and nO table; Table 2 shows the outputs of the observation clock generator and the bin generator of figure 23; and Table 3 shows the outputs of the observation clock generator and the bin generator of figure 25.

Detailed Description

A UWB signal is a signal which has a fractional bandwidth greater than 0. 25, or a signal which occupies 1.5 GHz or more of spectrum. The 1.5 GHz bandwidth floor only applies when the centre frequency is greater than 6 GHz. Fractional bandwidth, f is given by: _ 2(fh fl) f fh + f, where Ah is the highest frequency limit in the signal below lOdBs of the peak emission andf is the lowest frequency limit in the signal below 1 OdB of the peak emission.

The FCC allows UWB communication systems to use the frequency range 3.1 10.6GHz with a minimum 500MHz bandwidth and maximum emission values of 41 dB/MHz.

UWB IR techniques involve using short (sub-nanosecond) mono-cycle pulses to achieve an Ultra Wide-Band spectrum. Figures 1 and 2 show a Gaussian derivative mono-cycle in the time and frequency domains respectively. Figure 2 shows that the single mono-cycle shown in Figure 1 exhibits a fractional bandwidth of around 1.7 and is therefore a UWB signal.

A signal pulse can produce a UWB spectrum but a single pulse is not useful in a communications system. UWB IR signals consist of a continuous train of mono-cycles repeated at a rate that is much greater than the duration of the mono-cycle, i.e. a low duty cycle is employed. For example a duty cycle of 1 in 200 is typical (e.g. 0.5 nanosecond pulse repeated every 1 OOns). If a fixed pulse repetition is used, the spectrum of the signal contains high amplitude spikes at frequencies which are a multiple of the repetition rate, as shown in Figure 3. Here the pulses are repeated every lOOns giving spectral spikes every I OMHz.

To avoid the spectral spikes which lower power efficiency by placing some out of the bandwidth, and also to provide a multiple access mechanism, a pseudo-random time delay is inserted in between each pulse. This is normally achieved by taking the nominal time between pulses of duration Tf, called the frame time, and splitting it into a number of NC sub divisions, called chips, of duration [c. A pulse is transmitted during a randomly assigned chip duration for each frame. Figure 4 shows a simplified version of a UWB IR signal. Note the fixed nominal time between pulses, the frame duration, the chip intervals and the pseudo random time hopping of the pulses. Note that the pulse duration is normally much shorter compared to the frame time and there are more chip intervals (e.g. 128, 256, 512, etc.) than shown in Figure 4. The time hopping code is fixed in length to Nh and continuously repeats during transmissions. The pseudo random hopping code helps to smooth out the signals frequency spectrum which makes it look more random (and hence is more power efficient), as shown in Figure 5. Note that the greater the code length (Nh) is, and the greater the number of chips (Nc) there are, the better the multiple access performance and spectrum smoothing characteristics will be.

Exemplary values for Tf are in the region of 1 Ons to 1 ps.

The pseudo hopped pulse train signal conveys no information. In order for it to bear information the signal needs to be modulated using user data. This can be carried out in a number of ways; bi-phase modulation, Pulse Position Modulation (PPM), On Off Keying (OOK) and so forth. Note that for the purpose of obtaining sync to a UWB IR signal the encoded data is unimportant and the following description assumes that the received pulses have not been affected by data modulation.

The receiver needs to know the pseudo random code, where the start of the frame occurs in time and which code value corresponds to which frame in the received signal (code offset) in order to receive the signal correctly. Obtaining frame and code offset timing is known as synchronization.

Where UWB communication links are used for ad-hoc Wireless Personal Area Networks (WPANs), where connection life times may be in the order of seconds, fast acquisition of UWB timing is clearly desirable.

The UWB IR signal is detected by the "front end" of a receiver using a single Time Integrating Correlator (TIC), as shown in Figure 6. The frontend performs a one-shot correlation operation by using a multiplier 1 and integrator 2 arrangement. The received signal r(t) is fed to the input of the multiplier 1 along with an internally generated version of the received pseudo-random pulse train s(t-r'). r/ represents a variable that can be used to move the internally generated pulse train backwards and forwards in time in order to match the internally generated pulse train with the pulse train in the received signal. The internally generated pulse train needs to contain pulse shapes that are as close as possible to the received pulse shapes from the desired user and antenna characteristics may need to be taken into account. In order for the signal to be properly received, the generated pulse train also needs to be timed to coincide exactly with the received pulse trains pulse positions. If this is achieved a peak will appear at the output of the integrator each time a UWB IR pulse is received.

The output of the integrator 2 is sampled by a sampler 3, for example at the framerate.

Note that multiple pulses can be used to represent one symbol, with an increased processing gain, and hence the output of the integrator may be sampled at a rate less than the frame rate. For UWB 1R signals processing gain is defined as: PG Ts Tf (1) Where TJ is the frame duration and Ts is the symbol duration. Once the integrator 2 output is sampled, the integrators state must be reset to zero. The output of the sampler 3 may be held by a sample and hold circuit 4 for the duration of the symbol to allow the integration of multiple pulses, and hence processing gain, to be achieved.

If PPM is being used then a number of Time Integrating Correlators (1, 2, 3 and 4) are used, one for each pulse position in the modulation scheme. The only difference between the correlators is that the generated pulse train template is offset by a further (positive or negative) amount according to each pulse position in the modulation scheme. Here the correlator that outputs the highest peak value informs the receiver of the symbol being received. The aim of UWB IR synchronisation is to determine the value such that the

internally generated pulse train matches the received pulse train both in terms of time and code offset.

UWB IR synchronisation comprises of determining both frame and code synchronisation. Frame synchronisation is achieved by a receiver when it knows exactly where the start of the frame occurs in time (see Figure 4). Code synchronisation is achieved by a receiver when it knows which code value corresponds to which frame in the received signal. To obtain a useful state of synchronisation both frame and code synchronisation need to be achieved. Frame and code synchronisation are usually obtained simultaneously.

When the receiver is switched on, it has its own arbitrary idea of where the start of the frame and the start of the code is. The receiver needs to synchronise its idea of where the frame and codes start with that of the signal being received. This can be achieved by attempting to receive the signal at various time shifts relative to the receivers frame time start. At each time shift all of the possible shifts in the code need to be examined to. If a particular time and code shift is correct, a peak should be detected at the output of the correlator. The receiver can receive the signal at each of the time and code shifts for a number of frame durations in order to integrate the received pulses and obtain a more reliable result. The time and code shift that produces the maximum value out of the correlator is most likely to correspond to the time and code shift modification that is required to be made to the receivers clock in order to receive the signal correctly.

Conceptually it is convenient to represent this process as a splitting of the receivers frame duration (Tf) into a number of time bins Nb. Every offset in the code, relative to the receivers current code position (arbitrary at switch on), is examined at each bin.

Each time bin has a duration of Tb making Nb = Tf/Tb time bins in total. The number of observations required to perform the search is equal to NbXNh (recall that Nh is the length of the hopping code). If the duration of a pulse is denoted by Tw the condition Tb < Tw has to be met in order to obtain a reasonable chance of matching the front-end correlator template on to the received pulses. The number of bins required, Nb, therefore depends on the frame duration, Tf, and the pulse duration, Tw. A Tb of T,/4 is an

acceptable example.

Figure 7 shows the search results for a UWB IR system using Tf = lOOns, Nh = 128, Tb = 97.7ps (giving Nb = 1024) and a processing gain (as defined above) of 16 (12dB).

There are 128 code offsets. It can be seen that the optimal time and code offset point is a single point in the search space. In this case a Signal to Noise Ratio (SNR) of OdB was used, a Signal to Interference Ratio (SIR) of -12dB was used (with the number of users, K, being 26) and the Processing Gain (PG) was set to 12dB. With a broad peak that straddled multiple time and code offsets it is possible to intelligently home in on the signal over the time and code offset space using the length of the multi-path channel.

However this would be complex to implement and processor intensive.

Figure 8 shows some results for all time bins that could occur when the code offset is correct. It shows that signal components occur within a block of time bins equal to the channel length. Figure 9 shows a corresponding output when the code offset is incorrect. These results contain no signal components. It can be deduced that, in terms of reducing the number of observations made to perform the search, the reduction can only be made based on the knowledge of the block of signals within the time bins in the correct code offset scenario. No reduction in code offset observations can be made since it is correct in one case and incorrect in all others. The search space reducing techniques of the embodiment described use the block of signal components in the time domain.

The time taken to perform the search is application dependent since it depends upon the frame time Tf, the code length Nh, the processing gain PG (the number of frames to integrate over) and the time bin resolution Tb. As an example, Figure 10 shows the time required to perform an exhaustive search, in seconds, against the frame time Tf, in ns, for different code lengths (Ah) when using a processing gain (PG) of 16 (12dB) , a time bin resolution (Tb) of O.lns, and only one observation is allowed per frame. Note that these times do not take into account any synch state machine in charge of signal verification and false alarm rejection. Indeed the final sync time is likely to be higher than the values shown since it may be necessary to search the space a number of times before successfully obtaining synchronization.

The described embodiments search the search space faster by appreciating that a pulse observed at a particular instant in time could be present in this position in the received signal given a number of time bin and code offset combinations.

Figure 15 illustrates the principle and shows a number of super-frames of the expected UWB signal, but with various time bin and code offsets. The top super-frame corresponds to the receivers internally generated template signal which is compared with the incoming signal, and initially has code and time bin offsets of zero (0,0) with respect to an arbitrarily defined receiver starting time. If the incoming signal frame and and code offsets were synchronized to the receivers (arbitrary) frame and code offsets (0,0), then TIC output peaks would occur at each of the pulse positions shown in the generated super-frame (0,0). Each of these peaks would be added to the memory location corresponding to offset (0,0), and when combined would provide a large peak value. This large value will only occur when all of the pulses of the incoming and generated signal coincide, and will be detected as showing the code and time bin offset (0,0) required for synchronizing the receiver's generated signal to the incoming signal.

In practice however, it is unlikely that the receiver's arbitrary timing will correspond exactly with the incoming signal, and a receiver offset will need to be determined.

Figure 15 shows four internally generated signals having the known superframe structure with a pulse in each frame, the pulse position within each frame according to a code. There are four signals shown, each with a different code and time bin offset with respect to the internal receiver time - (0,0) which is not offset, (0,2) which has no code offset but an offset of two time bins, (1,2) which has a code offset of one and a time bin offset of two, and (0,7). The dashed vertical lines indicate TIC peaks, or pulse positions which correspond to an incoming signal pulse. Thus the first TIC pulse is at a pulse position which could correspond to the second pulse in the (0,0) local signal, or the first pulse in the (0, 7) signal. Since the algorithm knows the hopping code structure it knows to add the TIC output pulse to the memory locations corresponding to these two offsets (0,0) and (0,7). Referring now to the TIC peak at A, it can be seen that this could correspond to three possible offset values: (0,0), (1,2), and (0,7). Although they correspond to different peaks in each of these offset local signals, the TIC output is added to the memory locations for each of these offsets. Of course only a small number of offset possibilities have been shown to illustrate the embodiment, and in practice each observation will map to Nh (the hopping code length) time bin and code offset combinations. After a predetermined time or number of frames or super-frames, or after a memory location has exceed a particular threshold, the memory location having the highest value will be the offset at which to operate the receiver's time base in order to properly receive the UWB traffic or unknown signals.

In principle, the locally generated signal can be cycled through only the different time bins of a single frame, each of the time bin offsets corresponding to a number of code offsets having different time bin offsets. The example of figure 15 has 5 codes and 7 time bins per frame. Each time bin in the first frame of the receiver super-frame will correspond to a different time bin in another frame. More generally, each code and time bin offset (0,0) will correspond to a different time bin offset for each code - (1,6), (2,5), (3,4), (4,3). Thus the embodiment can search the 7x5=35 code and time bin offset search space in just 7 observations. This compares favourably with known solutions which cycle through all of the codes including all their time bins, which would require observations. In a more practical example having 128 codes and 1024 time bins per code, the embodiment requires only 1024 observations compared with the 131072 observations of the above known system. This is an improvement in synchronization time of over 100 fold.

Again in practice, it is likely that the embodiment will integrate the observations over a number of frames for increased accuracy, however as known systems also typically require this, the time saving advantage can still be seen to be significant.

Another way of looking at this is that each time bin offset (with respect to the receiver time) that we observe, were we to find a pulse there it would be the result of the transmitted signal transmitting at one of 5 (or 1023 in the above practical example) code and time bin offset combination values which point to possible adjustments that could be made to the receivers (arbitrary) bin and code offset values in order to align with the signal. Which one of these alignments (if any) is correct is initially unknown. However what we do know at this point is that if a pulse was found here (and was a result of the transmitted signal) one of these 5 hypotheses IS correct, or looking the other way the other 30 are definitely incorrect. The TIC outputs are added to the appropriate memory locations. If a pulse was present the TIC output will be significant and therefore the confidence in the 5 hypotheses will be strengthened. Next frame we observe at a different bin position, which produces another 5 (different) hypotheses and we can update another 5 (different) memory locations. So in 7 frames (or seven observations) we have covered the whole space. As we integrate the TIC outputs into our memory the TIC outputs due to received pulses will be added in to various places, however the pulses will only sum coherently at the correct bin and code offset. All the other locations will have a pulse integrated into it once in a while, whereas the correct location will consistently be updated. Therefore, choosing the maximum location at the end of the search will tell us what modification we need to make to the receivers idea of time in order for it to synchronize to the signal.

Whilst in the preferred arrangement the output of the TIC for each frame time bin is added to corresponding memory locations, over methods of incrementing these memory locations can be provided. For example only TIC outputs corresponding to a peak or correlation between the local and incoming pulses may be added to the memory locations; although for simplicity of implementation the TIC output of every time bin is added to the relevant memory locations. Alternatively a set increment may be added to the respective memory locations each time a TIC output pulse is detected.

In practice, the time bin Tb<TC' the chip time, and as a rule of thumb Tb<TW/4, where Tw is the pulse duration. This is because Tc>Tw and Tb<Tw.

A preferred algorithm is provided as follows. The receiver knows the pseudo random code and the frame duration Tf. It follows that the receiver can determine the code length Nh and the chip duration Tc. The receiver can observe up to a time resolution of Tb (time bit duration) and preferably where the pulse duration Tw divided by the time bit duration Tb is TTb > 4 T. Also c is a positive integer > 0 A mathematical treatment of an embodiment is described as follows. A UWB IR signal with multiple users and multi-path propagation can be defined by.

K L Nh-l ( ) ak,lW(t rk,l Ct,nTC (jNh +n)Tf -dk -Zik)+awgn(t) J=0 k=1 1=1 n=0 (2) w(t) represents the waveform of a single pulse in the received waveform.

n represents the index position into the pseudo random code, the code index, this shall be referred to as theframe number.

j represents the repetition number for the pseudo random code sequence, this shall be referred to as the super frame number. Note that equation (2) shows one symbol to be transmitted every super frame. This means that the number of pulses per symbol Ns=Nh.

This is not a restriction but issue is not dealt with here without any loss of generality since it shall be assume that no data modulation is applied to the signal during synchronization.

K is the number of users, one of which will be the desired user.

L is the number of resolvable paths (assumed that same for every user) N. ' is the pseudo random code length.

Tc is the chip duration.

Tf is the frame duration (Tf>TC).

Pk is the average power for the kth user.

ark,/ is the channel tap coefficient for the Ith path of the kth user.

Tk is the channel tap delay for the Ith path of the kth user.

Ck,n is the code value for the nth frame of the kth user.

dkJ) iS the delay value for the Pulse Position Modulation for the jth symbol for the kth user.

k iS delay equal to the difference between the start of the transmission for the kth user relative to the time the receiver is switched on. In theory -a'< k < +=, however, it will be assumed that the transmitter has begun transmitting when the receiver is switched on, i.e. - < ik< awgn(t) represents Gaussian noise at the receiver.

When the receiver is switched on it will define its own arbitrary idea of time which will be relative to the switch on time. Furthermore, the receiver will maintain its ownframe and super frame counts which shall be denoted by nr andjr respectively.

Assuming for the moment that no data modulation is present on the desired signal, and that no Multiple Access Interference, multi-path or noise is present, a simplified more manageable version of equation (2), is shown as equation (3). Nh-

r(t) = w(t - cnTc - (jNh - n)Tr - 5) j=0 n=0 (3) Note the complete removal of k and 1. can be defined as: 3=(jONh+no)Tf+{ (4) where: iojjr and nO=n-nr no represents the frame or code offset, i.e. the difference between the frame number (or code position) of the received signal and the receiver.

O represents the super frame offset, i.e. the difference between the super frame numbers of the received signal and the receiver.

B represents an unknown additional time offset which can be any real number O < < Tf Note that once the transmitter and receiver have been switched on no and JO are constants.

Using the assumption that the receiver can observe up to a time resolution of Tb Al can be approximated to: p=ATb (5) where is an integer O < < Nb and Tb is the time bin duration (Tb<TC). This approximation should not be too detrimental if Tb < Tw So: s=(joNh+nO)Tf +^Tb (6) Equation (3) then becomes: Nh r(t) = I, w(t - cnTc ( jNh - n - joNh - no) Tf - Ads) J=0 n=0 (7) The values of n and are unknown and it is the job of the synchronization algorithm to find them. It is known that n is an integer On<Nh and, using the approximation for 6, is an integer O<<Nb. A more useful parameter for the receiver to know is no which represents the change needed to the receivers frame counter nr so that it matches the value of n in the received signal.

The receiver makes observations of the received signal using the frontend correlator.

The output of the correlator is given by: P(7Jr,nr,b) | -I r(t)wrX(t + 7, n b)dt| (8) where wrx(t) is the template waveform and Y. n b iS a delay value where jr denotes the super frame number of the observation, nr denotes the frame number and b the bin position within a frame. The template waveform is generated at the receiver and fed to the front-end correlator along with the received signal r(t). Example template waveforms are shown in Figure 11 and Figure 12. These are the second and third derivatives of a Gaussian mono-cycle respectively and the corresponding auto- correlation functions of these template waveforms are shown in Figure 13 and Figure 14 respectively.

When y, n b iS chosen such that wring(t) overlaps a pulse in the received signal exactly, a peak is produced at the output of the correlator. During synchronization different values of Y. n b are tested and peak output values from the correlator are noted in an attempt to deduce the necessary timing required to receive the signal. The receiver can only make observations with reference to nr and jr. Also, assuming that the receiver can make observation down to a time resolution of Tb, the receiver can know where within each frame the observation is being made. Hence the delays Y. n b can be given by: yJrnrb = (jr Nh + nr)Tf + bTb (9) Note that O<b<Nb and that jr, nr and b are taken relative to the receivers idea of time.

Which exact ye, n h values are observed by the receiver depends upon its front-end correlation resources. If it is not possible to observe every bin position in every frame, for example because of hardware constraints such as a limited number of TIC's, a scheme which considers multiple frames must be employed. Care must be taken when devising this scheme to make sure that certain observations are not missed. An example suitable multiple frame observation scheme is made by choosing values of Y. n b which satisfy: YJr,n,,b r Tc) Tb) (10) For example if Nh = 16 and TJTb = 8 the useable Y. n 6 values for this scheme are Yooo'Yoo, Yo20, Torso, Y'o, Y,', Ys''Y202'Y22---' Y7,is,7' Ysoo, Ys,o, As already discussed one way of searching for the signal is to perform an exhaustive search by attempting to receive the signal for all the NhXNb combinations of and no, whilst summing the output of the correlator, for an arbitrary number of frames, Ns This can be easily achieved by choosing appropriate values of Y. n b The and no combination that produces the highest summed output from the corrletor is the most likely combination required to receive the signal. This method can be effective, but requires NhXNbxNs observations which can lead to long synchronization times. The search method of the embodiment makes more use of each observation in terms of where it can fit into the and no space in an attempt to decrease the search time without significantly reducing the detection performance.

When making an observation, P(y, n b), it is possible to create a set of hypothesized values for which correspond to every possible value of n (and ultimately no) that this single observation could belong to. There are Nh of these possible hypotheses for each P(Y, n b) observation. Sweeping b from O to Nb gives us Nb sets of the Nh hypothesis which span the whole NhXNb search space. Hence it is possible to gain information about all of the and no combinations within the NhXNb search space in just Nb observations.

The search will be conducted by maintaining a NhXNb store in memory which contains an element for each and no combination. The correlator outputs p(y, n b) are placed in the appropriate elements of memory based on the proposed algorithm. Note that the (7, n b) values are summed into the appropriate memory locations, using a read modify write operations, so that observations can be integrated to achieve the required processing gain.

To determine which Nh memory elements can be updated for each observation one hypothesizes using the question: If a pulse was found here, which and n combinations could have caused a pulse to be found in this position?' Note that this question is asked regardless of whether a peak is found at the output of the correlator. In an alternative arrangement however, just the TIC peaks might be processed. This is possible in low noise environments, however the observations at every and n combination is preferred in order not to miss pulses in noisy environments. Also note that the value n is mentioned in the question rather than no. It is possible to answer the question using n and then to generalize to no afterwards. The Nh hypotheses for deduced from an observation made at receive super frame jr, receive frame nr and bin position b are denoted by /\Jr,nr,h,n- In words, Jr,nr,b,n represents the time difference between the frame boundaries of the received signal and the arbitrary frame boundary maintained at the receiver if a pulse was received at time bin b and the transmitters frame number (code index) was n. Note that the jr and nr subscripts do not say anything about the context of the value but are used to specify when the observation was made.

Using equations (7) and (8) one can deduce that peaks will occur at the output of the correlator when: 1/ b = cnTc + (jNh + n + joNh + nO)Tf + jr,nr,b,n b (1 1) Using equations (9) and (11): (jrNh + nr)Tf + bTb = CnTc + (jNh + n + joNh + nO)Tf + Air Or b nTb Therefore Jo no b n = T (bTb-cnTc-( jNh + n + ioNh + nO + jrNh + n)Tf) and T. J,,n,,b,n b en T (jNh + n+ joNh +nO + jrNh +nr)Nb Since TiTb = Nb We know that 0< A, n b n <Nb so any addition and subtraction by integer multiples of Nb will have no affect. Therefore: A,,,nr,b,n = modes-en T. Nb This gives us a set of Nh values for each possible transmit frame number, given that we are observing the output of the correlator at time bin b. However what would be extremely useful is a set of values, A, n b n which give a set of Nh values, one for each possible code offset value no, given an observation at time bin b. A, n an can be obtained by appreciating that en, the code index (frame number) to code value look up function, requires n and not no, and a conversion is therefore necessary. Now: n0 =n-nr n =nO +nr A, n h n = mode Crn dno+nr Nh' T. N (12) For convenience it is assumed that nO is a positive integer > 0. This assumption can be made without loss of generality due to the cyclic nature of the pseudo random code.

Note that care must be taken so that O < nO+nr < N', hence the mod() operator on the subscript of c.

As an example lets say that an observation is taken at y3 4 7, i.e. jr-3' nr=4 and b=7 and we subsequently wish to obtain the Nh values of A3,4,7, no, i.e. A3,4,7,0, l\3,4,7,, A3,4,7,' ...

A3,4,7,Nh. These are obtained from: T. A3 4 70 = 7-c4 c T. A3471 =7-cs c ^3,4,7,2 = 7 - C6 T ^3.4,7 Nh 7 Cmod(Nt+4 Nh) T. As mentioned previously the search will be conducted by maintaining a NhXNb store in memory which contains an element for each and nO combination. The correlator outputs p(y, n b) are placed in the appropriate elements of memory based on the algorithm. Note that the P(Y, n b) values are summed into the appropriate memory locations, using a read modify write operations, so that observations can be integrated to achieve the required processing gain. Each p(Y, n b) value corresponds to an observation characterized by unique values of jr, nr and b. These values can be plugged straight into equation (12). The values A, n b O to j n b Nh can be obtained by setting nO to 0, 1, 2, ..., Nh. The combinations of nO and the resulting A, n bn are used to address the appropriate memory location in the store and the p(y, n b) observation is summed in to it. This carries on until the search is terminated when the required processing gain has been fulfilled. The position of the maximum value in the finally updated table will correspond to the most likely and nO combination, according to the algorithm, required to receive the signal.

Table 1 shows an example section of a and no store when the observation scheme of equation (10) is being used and the code in Figure 16 is being employed. The section shown is for O < no < 7 and O < < 4 and the scheme has been in progress for 5 super frames, i.e. O < jr <4. Each element shows which p(y; n b) observations are placed where in the table. The search scheme allows single observations to be placed in several places in the table. A few values within Table 1 have been highlighted in bold or underlined to show this.

If we now remove the restriction that there are no other users and no multi-path the algorithm will still work but the performance will be slightly degraded. The algorithm will attempt to pick out the largest multi-path component from within the Multiple Access Interference and noise.

The equation for yj n b' equation (9), can be used to defined every observable time instant based on a particular UWB IR signal structure and receiver time bin size. If every time bin can be observed every frame it is possible to fill up the and no array every frame. However, due to finite resources in the receiver it may not be possible to observe all combinations of Y. n b For example it may not be possible to observe every time bin every frame since it may not be practical to implement the number of correlators necessary to do this in the front-end of the receiver. This means that every element of the and no array can no longer be filled every frame and observations need to be intelligently taken across successive frames so that the observations made are sufficient to be able to fill up every element of the and no array as quickly as possible given the resources, with minimal duplications.

Two schemes for doing this are described below and each scheme can be scaled for different correlator resources which ultimately affect the search time. An equation for the number of frames required to completely fill up the A and no array for each scheme is given. This value assumes no processing gain using frame integration and the value calculated therefore needs to be multiplied by the required processing gain to obtain the number of frames required to actually perform the search.

It may be useful to think of the observations schemes as a way examining chip synchronism. Looking at chip synchronism at the receiver there are TJTb time bin offsets to choose from, only one of which will be chip synchronous with the received signal. The observation schemes aim to ensure that each of the TJTb offsets are examined so that the proposed scheme is capable of providing the correct and no values for acceptable signal reception.

In a first arrangement (Scheme 1), the and no array can be filled up every time there have been observations made for Ny consecutive frames at every bin in a contiguous block of TJTb bins. Ny depends on the code structure and is the maximum distance between two code indexes which have the same chip value (wrapping if necessary).

Figure 17 to Figure 19 illustrate some examples of Nd.

Thus for example, where a duration of a bin is a quarter of a chip (typical), then TJTb is 4 bins per frame over NO frames. Using this scheme any number of bins from 1 up to TJTb can be observed each frame. However in schemes which use less than TJTb observations simultaneously, the search has to be extended appropriately by further amounts of Ny frames so that all observations are made such that the and no array can be completely filled. For example let's say TJTb = 8 and Nb = 64. One could observe bins 12-19 for Ny consecutive frames, or one could observe bins 12-14 for the first Nit frames, bins 15-16 for the second No frames and bins 17-19 for the third Ne frames.

Note that the absolute position of the bins is unimportant as long as observations from all the bins in a TJTb contiguous block of bins are made. Here the relationship between the number of correlations made per frame, NCorr, and the number of frames required to completely fill up the and no array for this scheme, Nf7, is given by: f' d TEN, (13) Where xl denotes the 'ceiling' of x, i.e. the lowest integer greater than x, and Ncorr is an integer 1 < Ncorr < TJTb An advantage of this arrangement is that it can scale down to one just correlator resource. However, all chip values need to be used within the pseudo random code. If not, some values in the and no array cannot be filled. Also it is constrained by the valueof Ny which is a property of the code being used. In the best case Ny = NC and in the worst case N/ = Nh This scheme is rather like testing the TJTb offsets in turn and observing them for Ny frames. A pulse from the desired user will only be received when we are observing the bin which is chip synchronous with the received signal. Waiting for Nd frames ensures that the pulse will be seen (at the correct, chip synchronous, bin) since every chip value is gauranteed to have arrived at least once within the No, duration.

Figure 21 shows an exemplary implementation structure for the embodiment's UWB IR search algorithm. This system is based on the first Observation arrangement (Scheme 1) using one front-end correlator.

The front-end correlator (1, 2) accepts the input analogue signal r(t). The template reference signal is generated by the Clock 5 and Template Generator 6. The Template Generator is configured to produce an analogue pulse, which is as close as possible in shape to the pulses being received, each time it is triggered by the Observation Clock 5.

Each new observation is captured at the output of the Integrator 2 by the Sampler 3, Sample and Hold 4, ADC 7 and Latch 8 modules to create the discrete digital signal c(n).

The exemplary system aims to implement the search algorithm using Scheme 1 utilising one front-end correlator. In this example we wish to observe bin 0 for Ni frames, then bin l for a further Nil frames then bin 2 etc. up to bin (TJTb)-l. Once this has been carried out the and n, array can be filled once. The process can then be repeated, summing in new values, for however many frames of integration are required to achieve a suitable level of accuracy.

In this system this observation scheme is achieved using a clock 5 which has a period of Tf and whose phase can be changed by time increments equal to Tb when triggered from a second clock (9). This block is labelled as the Observation Clock block 5 in Figure 22.

The Observation Clock 5 will trigger the Template Generator 6 every frame so that an observation is made at the same bin position every frame. This carries on until the Observation Clock 5 is triggered by the Bin Clock 9, which has a period of N4Tf - that is the frame duration multiplied by the maximum number of frames between two code indexes which have the same chip value (wrapping if necessary), as explained above.

The triggering signal from the Bin Clock causes a phase change in the Observation Clock 5 which is equivalent to a momentary delay by Job for the Observation Clock 5.

The Observation Clock 5 then continues triggering the Template Generator 6 at regular frame intervals but it is now effectively triggering at the next bin along from before.

The Bin Clock 9 and Frame Clock 10 are used to update counters. The values of these counters are ultimately used to determine which value in the and no array needs to be updated for each observation. The Frame Clock 10 has a period of TJ and this remains fixed. This is used to update the Frame Counter l l which maintains the value of the receivers frame number nr. The Bin Clock 9 triggers the Bin Counter 12 which keeps a count of the bin offset b which is currently under observation. The Generate Bin Positions for all Code Offsets block 13 uses the values for nr and b and carries out the implementation of equation (12) to generate the A, n b n values for the algorithm.

The Update Memory and Compare 14 stage takes the discrete signal c(n) plus the bin addresses ^,n,,b,O to jJ,n,,h,Nh and updates the and no array by adding the new value to the ones stored in the appropriate memory locations. It also keeps a track of the maximum value obtained so far for each code offset plus its bin position (the bin position of the max values can be deduced from the A, n bn values). The update memory and compare module 14 is described in more detail further below.

Once the search has been completed the stored maximum values are used to determine the most likely code offset and bin position, Ac and nOC' which can be used to synchronize to the received signal.

Whilst not shown in Figure 21, there is also a connection from the Observation Clock module 5 to clear the Integrators memory 2 prior to each observation. The system may be implemented using a combination of analogue hardware, FPGA and/or ASIC digital processing, DSP software plus any other possible implementation mechanisms using any practical and feasible partitioning.

Figure 22 shows an exemplary implementation for the Update Memory and Compare stage 14. Every frame there is a requirement to update Nh memory locations in memory.

Figure 22 shows how the structure can be partitioned such that these memory updates can be performed in parallel. The memory can be partitioned into Nh blocks (21(0) - 21(Nb)) containing Nb locations, one block for each code offset. Since one memory location in each code offset block needs to be updated every frame one can guarantee that no memory contention will occur since each block is independently addressable.

Individual Update and Compare Blocks 21(0) - 21(Nh) run every frame and it is their job to update the relevant memory locations. For example the Update and Compare Block for Code 21(0) Offset 0 shall take A, n bO and c(n), and update the memory location pointed to by A, n b 0 by adding c(n) to it. The Update and Compare blocks 21 are also responsible for maintaining the current maximum value for the relevant Code Offset nO, denoted by CnaXn along with its corresponding bin position, rraxn These values form the outputs of the individual Update and Compare Blocks.

The Final Compare module 22 runs at the end of the search, once all observations have been gathered. It compares all of the CnaXn values with each other and finds the maximum value. The Code Offset that produces the maximum output forms the Chosen Code Offset output, nOC' and its corresponding bin position value (given by the corresponding Anaxn value) forms the Chosen Time Bin Offset value Ac nOC and Ac indicate the most likely Code Offset and Time Bin Offset values, according to the algorithm, , that can be used to synchronise the receivers time reference to that of the received signal.

Figure 23 shows a circuit arrangement similar to figure 2 l but in which a number (K) of front end correlators are employed to increase the speed of the search. This is because multiple correlators allow more than one observation per frame.

K front-end correlators 1 are employed along with the associated samplers 3, sample and hold modules 4, ADCs 7 and latches 8.Compared with figure 21, the Observation Clock module 5 becomes the Observation Clocks Generator 25 which produces K trigger signals, one for each front-end correlator, which stimulate K template generators (TG) which feed the K front-end correlators. The Bin Clock module 9 becomes the Bin Generator 29 which now generates K bin values rather than a single value. This module is still triggered at the same rate of NTf. It outputs a T;KTb long sequence of K bin values such that all TJTb bin values in the TJTb contiguous block are included once, and only once, in the sequence. The Generate Bin Positions for all Code Offsets module 33 now takes in K bin values, instead of only one, and outputs KNh bin positions rather than Ah. Thus there are Nh outputs for each input bk.. The module works in exactly the same way as before, there are simply more inputs and more outputs.

The Update Memory and Compare module 34 now takes in K c(n) values rather than one, and KNh bin position values rather than Oh. Figure 24 illustrates the slight difference in the internal workings of the Update Memory and Compare Module 34.

The Update and Compare Blocks now receive a bus of K bin position values and a bus of K latch output values. A, ,n,hk,o represents the values A,, no bO 0 to A,, or bK 0 and C(n)k represents the values c(n)o to C(n)K. Here the Update and Compare Block deals with updating all of the K memory locations in the code offset memory block with the appropriate c(n) values. Note that due to the nature of the algorithm all of the bin positions will be unique so no attempt will ever be made to update the same location twice in one frame duration. This can therefore be realised using parallel operations.

A simple example is illustrated with reference to Table 2. In this example Tc = ins, Tb = 62.5ps and K= 4. For scheme 1, one can search the whole search space one time by observing every bin in a block of contiguous Tc/Tb bins for Nd frames each, where Nd is the maximum distance in frames between two identical code values in the time- hopping code. If one observation is allowed per frame the search will take NdTc/Tb frames to complete, however with K correlators it is possible to make a time saving of a factor of K to NdTc/KTb frames. Assuming that the chosen contiguous block of bins is 0-15, with the K=4 scenario and a design as in figure 23, the search could be conducted as shown in Table 2. Each row of this table is employed for Nd frames before moving on to the next row; this is triggered by the Bin Clock 29. The Observation Clock Generator 25 is required to send the appropriate trigger signals, in accordance with the appropriate row of Table 2, every frame until it is triggered by the Bin Clock 29 in which case it moves to the next row and uses the new trigger times. Note that when the Observation Clocks Generator and Bin Generator modules are using the final row of Table 2 they will revert to using the first row once triggered further by the Bin Clock module 29. With the Observation Clocks Generator 25 and Bin Generator modules 32 behaving as in this table, and the remaining modules functioning as described above a system which implements scheme 1 with multiple correlators is realised. It is a straight forward matter to envisage extending the example design from 4 correlators 1 to K correlators 1.

For convenience Tc, Tb and K are preferably chosen such that Tc/Tb is an integer and Tc/Tb is a multiple of K. A second arrangement (Scheme 2) looks at observing sets of Nc bins simultaneously within a frame. One set in the context of this scheme consists of Nc bins with each bin being separated by a number of bins equating to a Tc duration. Within the frame there are TJTb such unique sets and the and no array can be filled up completely once all sets have been examined. One can examine any number of these sets per frame from 1 to TJTb. When less than TJTb sets are examined simultaneously (perhaps due to a deficiency in correlator resources) the search is extended by the appropriate number of frames required so that all sets have been examined. The relationship between the number of correlations made per frame, Ncorr, and the number of frames required to completely fill the and no array for this scheme, NJ7, is given by: N T,N, (14) Where axe denotes the 'ceiling' of x, i.e. the lowest integer greater than x, and NCorr is an integer which is a multiple of Nc and Nc < NCor, < NC X T;Tb. Note that Nc X TJTb = Nb.

Advantages of this second arrangement are that all chip values do not need to be used within the code; and it is not constrained by Nit, if there are enough resources it is possible to fill up the and no array in one frame. However it is restricted to a minimum number of correlators of NC and the number of correlator must be a multiple of Nc. If NC is large then this may become impractical.

Like the first scheme (1) this scheme (2) is rather like testing the TJTb offsets in turn but instead of waiting for Nil frames to ensure that a pulse arrives this scheme covers all of the possible chip positions simultaneously for each of the TJTb offsets. This ensures that, if the offset being examined is correct for chip synchronism a pulse will be observed in the frame. This technique allows the search time to be drastically reduced but requires more correlator resources to achieve it. Covering all chip values simultaneously also makes the observation scheme independent of the code structure.

Figure 25 shows a circuit arrangement utilising a number of correlators 1 and analogous to that of figure 23, except that this is implemented for scheme 2 rather than scheme 1.

Scheme 2 looks at observing sets of Nc bins simultaneously within a frame. One set in the context of this scheme consists of Nc bins with each bin being separated by a number of bins equating to a To duration. Within the frame there are Tc/Tb such unique sets and the and no array can be filled up completely once all sets have been examined. The main differences between figures 23 (scheme 1) and 25 (scheme 2) are as follows. The Bin Clock 39 period is Tf as Scheme 2 does not require observations for Nd frames. However it does require at least Nc correlators and the number of correlators, K, must be a multiple of Nc. The Bin Generator 42 now generates a NcTc/KTb length sequence of K bin position values. The collections of bin values will be in accordance with scheme 2 (see example below). The Observation Clocks Generator module 35 generates clock signals in accordance with scheme 2 (see example below). The remainder of the design functions as described previously.

An example is illustrated with reference to table 3. In this example Tf= 2ns, To = ins, Tb = 125ps and K = 2xNc = 4. the table shows a suitable template generator triggering times and bin generator output sequences for the Observation Clocks Generator 34 and Bin Generator modules 42 respectively. As before the Bin Clock 39 would trigger a transition through the rows of the table.

From table 3 it can be seen how the first row will trigger the template generators for bin positions 0, 16, 1 and 17. The logic contained within the Generate Bin Positions 42 for all Code Offsets and Update and Compare Memory modules will ensure that the correct observations go in the correct element in the and no array. Progressing through the rows of table 3 shows that every bin is examined and since this is done in such a way to realise scheme 2 the and no array can be fully populated after all rows have been observed.

Again it is best to choose Tc, Tb and K such the Tc/Tb is an integer and that K is a multiple of Nc.

Note that the designs explained by Figures 23 and 25 show multiple correlators with one correlation being performaed by each correlator per frame. However a fast correlator apparatus could be used immediately after it has performed a correlation. In this way more observations can be made per frame for the same number of correlators.

Using this scheme there is a minimum number of correlators that can be employed such that every time bin can be observed on every frame. This number depends on the pulse duration Tw and the bin time Tb. One can fix Tb to equal Tw/M where M is an integer and the minimum number of correlators is given by M. In the implementation the M correlators will be working substantially faster than a frame duration, performing the next correlation immediately after the current one has finished, and at a different time bin offset to all the other correlators. For example if Tb = Tw/4 (four correlators) the first correlator could correlate against bins 0, 4, 8, 12, etc., the second at 1, 5, 9, 13,etc.

the third at 2, 6, 10, 14, etc. and the fourth at 3, 7, 11, 15, etc. All of these correlations could be made every frame making it possible to observe every bin position every frame using a minimum number of correlators.

When the receiver is switched on it will not necessarily be frame synchronous with the received signal, in fact that is extremely unlikely. Hence the receiver's frames will straddle the received signals frames. This straddling effect can cause the algorithm to get 'confused' which it tries to determine the code offset, no. Normally this confusion will not generate errors, however the level of confusion increases as the receivers frames begin to straddle the received signals frames by factors approaching 50%, i.e. when the receiver 'sees' more or less equal contributions from two different frames of the received signal. This effect manifests itself in the results of the and no array by showing similar peaks at consecutive values of no. One of the code offsets will present the correct value, the other will present a 'confused' value. If the peaks are of similar magnitude (which may occur when the recover straddles the received signals frames by 50%) the maximum operator may pick the wrong offset resulting in a detection error.

An easy way to mitigate this effect is for the algorithm to perform a short verification phase. The value for will always be correct but the no value found could be either no, nO+1 or nO-l. The verification phase merely tries the three alternatives, integrating the output of the correlator to obtain the required processing gain, and chooses the combination which produces the peak output.

The search is terminated when observations have been made for the appropriate number of frames such that every element in the and no array has had enough correlator output values summed into it to fulfill the desired processing gain. For example if we are using the first Observation Scheme (1) with Nat = 16 and a processing gain of 5 a decision will be made on the search, i.e. the max operator used on the and no array, after 16 x 5 = frames of observation. The processing gain can thus be used as a variable parameter for the search method. For example it may be desirable to initially make a noise measurement and use the measurement to determine a suitable processing gain value for the search. It may also be desirable to modulate a known data pattern on to the UWB IR signal to aid the search termination process. For example one may wish to employ an iterative approach where an initial process gain value is tried and an attempt is made to read the fixed training pattern after the above-described verification process has been carried out. If the training sequence is not found, a higher processing gain value can be tried and the process repeated until either the training sequence is found, or a maximum processing gain value is reached. In the latter case it is decided that the signal cannot be found. Note that throughout the whole process the values in the and nO array will never be cleared. At each decision phase, if it is decided that a higher processing gain is required, the integration of the values in the and no array is continued where it left off.

Hence the only time penalty here is in the verification and training sequence reception phases.

There are certain code structures that will perform badly during synchronization. These are codes that have a constant difference between adjacent code values. Figure 20 shows such a code, the difference between adjacent code values (module NC (which is 8 here)) is always 5 in this case.

Confusion and ambiguities occur when examining different code offsets, no, since the absolute difference between the code value in the received signal and that of a particular no is always constant. Normally when an incorrect code offset is being examined pulses from the desired user will be found but they will not sum coherently during the integration process since the code offset is incorrect and the code structure does not match. However when a code structure as described above is employed the spurious peaks found in the incorrect code offsets will sum coherently due to the constant fixed difference between adjacent code values causing the peaks to always occur in the same place. These will sum in exactly the same way as the results at the correct and no. This causes the results for all code offsets to 'look' like they are correct since they all have similar peak values in them. The search algorithm cannot tell which is the 'real' code offset values and will therefore pick one randomly and detection perfonnance degrades.

This effect causes problems for the proposed algorithm as well as the conventional thresholding algorithms and the best way to mitigate this effect is to not use codes of this structure during synchronization. Figure 26 shows some results obtained when using such a 'bad' code.

Time is saved by appreciating the multiple interpretations of each pulse observation at the expense of lower performance. However the reduction in performance is acceptable.

The period of time required to perform the proposed search is also fixed and is therefore

predictable.

The embodiment can significantly decrease the time required to search for a UWB IR signal without a significant degradation in signal acquisition performance. A number of conventional search algorithms have a fluctuating search time which is governed by the particular scenario. However the search time for the embodiment is fixed, which is advantageous in wireless reception applications.

In addition, the implementation of the embodiment can be partitioned into analogue and digital sections where the digital section operates at a realistic sampling rate and parallelism can be employed.

It is possible to obtain more reliable performance by integrating the observed values for a number of frames. This has the effect of effectively increasing the SNR of the signal.

In fact with conventional algorithms a certain amount of integration is essential because observations ambiguities are not taken into account. The embodiments appreciate these ambiguities and it is therefore possible to run it using no signal integration (the signal strength must be increased appropriately though). This makes it possible for the search time to be increased when using low SNR signals.

Simulation results are shown in Figures 26-29. The proposed algorithm is compared \ The first of these is the Linear search method observes each time bin sequentially starting at the first bin position. Whilst examining each bin a number of observations are made (in accordance with the required processing gain) for each and every code offset before moving on to the next bin. The search is terminated at the first time bin and code offset position that produces a signal strength (after integration) which is above a calculated threshold, y. Hence the time taken to perform the search is scenario specific.

The threshold y is calculated based on a noise measurement.

The second known algorithm is the Maximum search method observes all of the time bins in tune, for all of the possible code offsets whilst employing the required processing gain (obtained by integrating correlator outputs). The time bin and code offset value which produces the maximum output is chosen to be the most likely time and code offset value required for signal acquisition. Here the search time is fixed and long since all time and code offset values are considered. However the method should provide the best performance for algorithms based on the amplitude of observations and therefore provides a good benchmark.

Parameters used during the simulations were: Tw = 0.5ns Tf= lOOns Tb = 97. 7ps (giving Nb = 1024) SIR = -12dB PG= 12dB (integrate over 16 frames) K = 26 Nh = 16 y= 6.5dB above the noise measurement Number of runs per point = 10,000 Figure 26 shows results for the Probability of Detection (Pit) versus received energy (for each time and code offset observation; could include multiple pulses) (Eb) over noise spectral density (No). Note that the Eb value take into account the processing gain due to integration being employed. Figure 26 shows that the proposed method outperforms the linear method at EkNo values around and above the point where the max method (best case) curve flattens out. Note that the performance of the linear method decreases for high Eb/No values. This is due to the fact that the linear search does not take into account any ambiguities in pulse observations. The effect of the ambiguities increases with E'JNo and hence PI falls. This effect flattens out at high Eb/No values as the signal dominates over the noise.

Figure 27 shows the time taken to perform the search divided by the time taken to perform the exhaustive search. It shows that at high EJNo values the Linear search algorithm can cut the time taken to about 25% of the exhaustive search (but at reduced performance). However although the average time decreases the instantaneous time taken to perform the search can still be above or below this value and is hard to predict.

The proposed algorithm cuts the search time down to 6.25% of the exhaustive search time and the search time is fixed and therefore predictable. Note that the saving in the search time is a factor of the code length, Nh, the longer the code the greater the saving.

Figure 28 shows the amplitude of the signal found (in the cases when the signal was detected) relative to the size of the maximum signal value that could be found. The ability of the signal acquisition scheme to detect the largest signal multi-path component may be advantageous. It shows that the amplitude of the signal found of proposed algorithm hovers around -0.1 5dB of the maximum value when Eb/No is above 20dB and presents over a 2dB advantage over the Linear search algorithms at high EJNo values. Note that the performance of the Linear search algorithm decreases at high EJNo values as explained above.

The graph shown in Figure 29 shows simulation results obtained when using a so called bad' code as discussed previously. Note that for these results Nc=Nh=8 was used and also note that 100 runs were used per point. It can be seen that the performance of the proposed method reaches a maximum PI of 12.5% at high EJNo values, i.e. the maximum PI achievable is 1/Nh since all code offsets present equal maxima. The thresholding algorithms enjoy a slightly better performance since they do not always search the whole space and may therefore 'stumble' upon the correct detection point slightly more often. However it is conclusive that both approaches suffer significantly when using a 'bad' code.

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 programme 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 analog array or similar device in order to configure analog 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: 1. Apparatus for synchronizing a receiver to a received UWB signal having a known hopping code, the apparatus comprising: means for determining a parameter of the signal in a number of time slots; means for adding the parameter associated with each time slot to a number of corresponding of time and code offset accumulators; means for determining the code and time offset accumulator having an accumulated value according to a predetermined accumulator metric as the code and time offset required to synchronise the receiver. 2. An apparatus according to claim 1 wherein the predetermined accumulator

metric is the largest value.

3. An apparatus according to claim 1 or 2 wherein the parameter determining means comprises a template signal generator for providing a template signal at a second time offset, and means for correlating said template signal with said received signal in order to provide said parameter.

4. An apparatus according to claim 1, 2 or 3 wherein the number of accumulations corresponding to each time slot is two or more for all time slots.

5. An apparatus according to claim 4 wherein the number of accumulations corresponding to each time slot is equal to the code length.

6. An apparatus according to any preceding claim wherein the UWB signal comprises a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to the hopping code; and wherein the apparatus comprises two or more parameter determining means corresponding to two or more time slots in a frame.

7. An apparatus according to any preceding claim wherein the apparatus comprises one parameter determining means for each time slot.

8. An apparatus according to any one of claims 1 to 5 wherein the UWB signal comprises a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to the hopping code; and wherein the apparatus comprises one or more parameter determining means corresponding to one or more time slots in a frame, and means for adjusting the time slot each frame for the or each said parameter determining means.

9. An apparatus according to any preceding claim wherein the corresponding accumulators are determined according to: T. J,,nr,b,nO = mod b Crrod(nO+n,Nh) T. b 10. An apparatus according to any preceding claim wherein the accumulators are individually addressed memory locations.

11. An apparatus according to any one preceding claim and further comprising: verification means comprising means for determining the frame offset and code offsets having N accumulated parameter values according to a predetermined verification metric; means for determining said parameter of a signal corresponding to said received signal but generated according to the time and code offset associated with each of said N accumulator values; means for selecting one of said N time and code offsets according to said corresponding parameter.

12. An apparatus according to claim 11 wherein the verification metric is the highest accumulator values, or the one highest accumulator value and the accumulator values associated with a range of time slots on either side of said highest value.

13. An apparatus according to claim 11 or 12 wherein said selection means is arranged to select the highest parameter.

14. An apparatus according to claim 11, 12 or 13 wherein the parameter is the correlation of the generated signal with the received signal.

15. An apparatus according to any preceding claim further comprising termination means for attempting to determine a training sequence modulated onto the received signal following said synchronization, and if unsuccessful further determining said parameter for further time in order to provide an improved estimate.

16. An apparatus according to any preceding claim wherein the UWB signal comprises a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to the hopping code; and wherein one or more said parameter determining means are arranged to determine said parameter for one time slot per frame.

17. An apparatus according to any one of claims 1 to 15 wherein the UWB signal comprises a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to the hopping code; and wherein one or more said parameter determining means are arranged to determine said parameter for two or more time slots per frame.

18. Use of an apparatus for synchronizing a receiver to a received UWB signal having a known hopping code wherein the UWB signal comprises a series of pulses transmitted over a corresponding series of transmitter time frames, the relative position of the pulse in each transmitter time frame is one of Nc time slots relative to the time frame start and is governed according to the hopping code; and wherein the time hopping code ensures that all NC time slots are stimulated one or more times.

19. Use of an apparatus according to claim 18 wherein the time hopping code further ensures that the numerical difference module NC between consecutive hop code values is always unique for the entirety of the code except the numerical difference between the last and first code value.

20. Use of an apparatus according to claim 18 or 19 wherein the maximum distance, in terms of the number of frames, between any numerically identical code values within the code structureis equal to NC frame durations.

21. Use of an apparatus according to any one of claims 18 to 20 wherein the apparatus is according to any one of claims 1 to 17.

22. An apparatus for transmitting a UWB signal having a known hopping code, wherein the UWB signal comprises a series of pulses transmitted over a corresponding series of transmitter time frames, the relative position of the pulse in each transmitter time frame is one of NC time slots relative to the time frame start and is governed according to the hopping code; and wherein the time hopping code ensures that all NC time slots are stimulated one or more times.

23. An apparatus according to claim 22 the time hopping code further ensures that the numerical difference module Nc between consecutive hop code values is always unique for the entirety of the code except the numerical difference between the last and first code value.

24. An apparatus according to claim 22 or 23 wherein the maximum distance, in terms of the number of frames, between any numerically identical code values within the code structureis equal to NC frame durations.

25. A method for synchronising a receiver to a received UWB signal having a known hopping code, the method comprising: determining a parameter of the signal in a number of time slots; adding the parameter associated with each time slot to a number of corresponding of time and code offset accumulators; determining the code and time offset accumulator having an accumulated value according to a predetermined accumulator metric as the code and time offset required to synchronise the receiver.

26. A method according to claim 25 wherein the predetermined accumulator metric is the largest value.

27. A method according to claim 25 or 26 wherein the parameter determining comprises generating a template signal at a second time offset, and correlating said template signal with said received signal in order to provide said parameter.

28. A method according to claim 25, 26 or 27 wherein the number of accumulations corresponding to each time slot is two or more for all time slots.

29. A method according to claim 28 wherein the number of accumulations corresponding to each time slot is equal to the code length.

30. A method according to any one of claims 25 to 29 wherein the corresponding accumulators are determined according to: ir.nr,b, nO modes Clods n + n Nh - Nh) 31. A method according to any one of claims 25 to 30 and further comprising: determining the frame offset and code offsets having N accumulated parameter values according to a predetermined verification metric; determining said parameter of a signal corresponding to said received signal but generated according to the time and code offset associated with each of said N accumulator values; selecting one of said N time and code offsets according to said corresponding parameter.

32. A method according to claim 31 wherein the verification metric is the highest accumulator values, or the one highest accumulator value and the accumulator values associated with a range of time slots on either side of said highest value.

33. A method according to claim 31 or 32 wherein said selection is arranged to select the highest parameter.

34. A method according to claim 31, 32 or 33 wherein the parameter is the correlation of the generated signal with the received signal.

35. A method according to any one of claims 25 to 34 further comprising attempting to determine a training sequence modulated onto the received signal following said synchronization, and if unsuccessful further determining said parameter for further time in order to provide an improved estimate.

36. A carrier medium carrying processor control code for implementing a method/apparatus according to any one of claims 25 to 35.

Claims (36)

1. Amendment to the claims have been filed as follows Document. 1053493

   CLAIMS: 1. Apparatus for synchronizing a receiver to a received UWE IR signal having a known hopping code, the apparatus comprising: means for determining a parameter of the signal in a number of time slots; means for adding the parameter associated with each time slot to a plurality of accumulators associated with different time and code offset combinations corresponding to the time slot; means for determining the code and time offset accumulator having the largest accumulated value as the code and time offset required to synchronise the receiver.
2. 2. An apparatus according to claim 1 wherein the parameter determining means comprises a template signal generator for providing a template signal at a second time offset, and means for correlating said template signal with said received signal in order to provide said parameter.
3. 3. An apparatus according to claim 1 or 2, wherein the number of accumulations <, " corresponding to each time slot is two or more for all time slots.
4. 4. An apparatus according to claim 3 wherein the number of accumulations corresponding to each time slot is equal to the code length.
5. 5. An apparatus according to any preceding claim wherein the UWB signal comprises a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to the hopping code; and wherein the apparatus comprises two or more parameter determining means corresponding to two or more time slots in a frame.
6. 6. An apparatus according to any preceding claim wherein the apparatus comprises one parameter determining means for each time slot.
7. 7. An apparatus according to any one of claims 1 to 4 wherein the UWB signal comprises a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to the hopping code; and wherein the apparatus comprises one or more parameter determining means corresponding to one or more time slots in a frame, and means for adjusting the time slot each frame for the or each said parameter determining means.
8. 8. An apparatus according to any preceding claim wherein the corresponding accumulators are determined according to: J,.n,b,nO =modb- CmOdn +n Nh)-tNb)
9. 9. An apparatus according to any preceding claim wherein the accumulators are individually addressed memory locations. ' . . . ',
10. 10. An apparatus according to any one preceding claim and further comprising: verification means comprising means for determining the frame offset and code offsets having N accumulated parameter values according to a predetermined verification metric; , means for determining said parameter of a signal corresponding to said received signal.

    but generated according to the time and code offset associated with each of said N accumulator values; means for selecting one of said N time and code offsets according to said corresponding parameter.
11. 11. An apparatus according to claim 10 wherein the verification metric is the highest accumulator values, or the one highest accumulator value and the accumulator values associated with a range of time slots on either side of said highest value.
12. 12. An apparatus according to claim 10 or 11 wherein said selection means is arranged to select the highest parameter. f
13. 13. An apparatus according to claim 10, 11 or 12 wherein the parameter is the correlation of the generated signal with the received signal.
14. 14. An apparatus according to any preceding claim further comprising termination means for attempting to determine a training sequence modulated onto the received signal following said synchronization, and if unsuccessful further determining said parameter for further time in order to provide an improved estimate.
15. 15. An apparatus according to any preceding claim wherein the UWB signal comprises a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to the hopping code; and wherein one or more said parameter determining means are arranged to determine said parameter for one time slot per frame.
16. 16. An apparatus according to any one of claims 1 to 14 wherein the UWB signal' ', comprises a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to the hopping code; " ' and wherein one or more said parameter determining means are arranged to determine said ' " " ' parameter for two or more time slots per frame. ,,, .
17. 17. An apparatus according to claim 15 or 16 wherein the relative position of the pulse in each transmitter time frame is one of NC time slots relative to the time frame start and is governed according to the hopping code; and wherein the time hopping code ensures that all NC time slots are stimulated one or more times.
18. 18. An apparatus according to claim 17 wherein the time hopping code further ensures that the numerical difference module Nc between consecutive hop code values is always unique for the entirety of the code except the numerical difference between the last and first code value.
19. 19. An apparatus according to claim 17 or 18 wherein the maximum distance, in terms of the number of frames, between any numerically identical code values within the code structureis equal to Nc frame durations. f
20. 20. A method for synchronizing a receiver to a received UWB IR signal having a known hopping code, the method comprising: determining a parameter of the signal in a number of time slots; adding the parameter associated with each time slot to a plurality of accumulators associated with different time and code offset combinations corresponding to the time slot; determining the code and time offset accumulator having the largest accumulated value the code and time offset required to synchronise the receiver.
21. 21. A method according to claim 20 wherein the parameter determining comprises generating a template signal at a second time offset, and correlating said template signal with said received signal in order to provide said parameter.
22. 22. A method according to claim 20 or 21 wherein the number of accumulations,, A, corresponding to each time slot is two or more for all time slots. .
23. 23. A method according to claim 22 wherein the number of accumulations corresponding " ' ( . to each time slot is equal to the code length. ' " ' '
24. 24. A method according to any one of claims 20 to 23 wherein the corresponding, accumulators are determined according to: T. jr I, b no = mod b Crnod(no+nr,Nh) T ' b
25. 25. A method according to any one of claims 20 to 24 and further comprising: determining the frame offset and code offsets having N accumulated parameter values according to a predetermined verification metric; determining said parameter of a signal corresponding to said received signal but generated according to the time and code offset associated with each of said N accumulator values; selecting one of said N time and code offsets according to said corresponding parameter. t
26. 26. A method according to claim 25 wherein the verification metric is the highest accumulator values, or the one highest accumulator value and the accumulator values associated with a range of time slots on either side of said highest value.
27. 27. A method according to claim 25 or 26 wherein said selection is arranged to select the highest parameter.
28. 28. A method according to claim 25, 26 or 27 wherein the parameter is the correlation of the generated signal with the received signal.
29. 29. A method according to any one of claims 20 to 28 further comprising attempting to determine a training sequence modulated onto the received signal following said synchronization, and if unsuccessful further determining said parameter for further time in order to provide an improved estimate. ,,,
30. 30. A method according to any one of claims 20 to 29 further comprising attempting to determine a training sequence modulated onto the received signal following said "' synchronization, and if unsuccessful further determining said parameter for further time in t order to provide an improved estimate. ,,
31. 31. A method according to one of claims 20 to 30 wherein the UWB signal comprises a " series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to the hopping code; and wherein the parameter determining comprises determining said parameter for one time slot per frame.
32. 32. A method according to any one of claims 20 to 30 wherein the UWB signal comprises a series of pulses transmitted over a corresponding series of time frames, the relative position of the pulse in each transmitter time frame according to the hopping code; and the parameter determining comprises detemmining said parameter for two or more time slots per frame.
33. 33. A method according to claim 31 or 32 wherein the relative position of the pulse in each transmitter time frame is one of Nc time slots relative to the time frame start and is governed according to the hopping code; and wherein the time hopping code ensures that all Nc time slots are stimulated one or more times. 5Y
34. 34. A method apparatus according to claim 33 wherein the time hopping code further ensures that the numerical difference module NC between consecutive hop code values is always unique for the entirety of the code except the numerical difference between the last and first code value.
35. 35. A method according to claim 33 or 34 wherein the maximum distance, in terms of the number of frames, between any numerically identical code values within the code structureis equal to NC frame durations.
36. 36. A carrier medium carrying processor control code for implementing a methodlapparatus according to any one of claims 20 to 35.

    ., i, 1 (, ( ( . ( ( ( ( 1. < (, ( / < (