WO2001003394A1

WO2001003394A1 - Pulse shaping device for mobile communication systems

Info

Publication number: WO2001003394A1
Application number: PCT/IB2000/000978
Authority: WO
Inventors: Natividade Lobo; Dean Saldanha; John Roughley; Simon Furmidge
Original assignee: Nokia Mobile Phones Limited
Priority date: 1999-07-01
Filing date: 2000-06-30
Publication date: 2001-01-11
Also published as: CN1367973A; EP1198937A1; GB9915434D0; AU5837400A; JP2003531507A; GB2351633A

Abstract

A digital telecommunications system is described in which pulse shaping for the up-link and the down link are designed independently in accordance with desired cost parameters (e.g. error functions for amplitude, BER, bandwidth, energy, AFC). A transceiver designed to operate in such a system is also described.

Description

PULSE SHAPING DEVICE FOR MOBILE COMMUNICATION SYSTEMS

In digital radio telephones, serial bit streams of data are transmitted over-the- air. The bit streams are used to modulate a carrier. There are several types of modulation scheme used to transmit data carried by the bit stream. For example, in GSM the modulation scheme used is Gaussian Minimum Shift Keying (GMSK) whereas in CDMA systems the modulation technique used is QPSK.

GMSK is a phase modulation that converts a serial bit stream into a phase shift of a carrier wave. The function of the modulation is to convert the incoming serial bit stream into analog signals that modulate the carrier of the transmitter. In GMSK the outgoing phase shift is filtered. The Gaussian function acts as a filter, removing the sharp edges of the digital pulses. Without this filtering the required bandwidth to transmit the signal would be far greater. Even with the gaussian filter it is acknowledged that the GSM system is spectrally inefficient. The GMSK modulation does, however, provide a constant amplitude signal that is power efficient.

In existing CDMA systems a different phase modulation technique, QPSK, is chosen to provide a higher bit rate. In QPSK, orthogonal signals are transmitted which double the data rate relative to MSK modulation. In QPSK modulation the outgoing phase shift is Nyquist filtered to provide root raised cosine shaped pulses that increase the spectral efficiency and reduced bit error rate by eliminating intersymbol interference. QPSK with root raised cosine pulse shaping is spectrally efficient allowing a high data rate and providing a low BER. According to one aspect of the present invention, there is provided a digital telecommunications system in which first and second communications devices communicate by respectively transmitting pulses indicative of data in accordance with a predetermined modulation scheme, the first and second devices each comprise means for shaping the respective data pulses prior to transmission, the shaping being applied in accordance with respective system criteria.

The first and second devices may be different types of communications device, for example a fixed station and a mobile station.

In prior art modulation schemes the pulse functions used to shape the data streams have had a predefined mathematical relationship.

For example:

root raised cosine

H (f) = 1 | f | < cc

H (f) = -l (1 - cos(2π(f - (T + α) c < I f I < τ+ cc

0 l f l > τ+ oc

for CDMA systems in which QPSK modulation is used and PDC and NADC systems in which DQPSK modulation is used.

Gaussian

for GSM in which an MSK modulation scheme is used.

With pulse shapes according to the conventional predefined mathematical relationships only one parameter is variable for a given energy level. For the gaussian pulse this is 'sigma' that varies the spread of the pulse allowing the bandwidth to alter at the expense of amplitude. For the root raised cosine the variable is 'alpha' that varies the frequency at which the cosine tail begins. This effects the bandwidth and consequently the power efficiency. The relationship between the cost parameters is well defined so as one improves the other declines in a determined fashion. There is no scope for improving both cost parameters.

Because of the severe restrictions placed on the trade-offs achievable by varying the single variable for the predetermined mathematical functions, the pulse shape most appropriate for each modulation scheme is quite clear. The system designer makes a decision on which modulation scheme based on its strengths and weaknesses and selects the appropriate pulse shape. The single variable of the mathematical function is set to provide an acceptable balance in the defined relationship between the cost parameters.

Once, as in the present invention, there is no predetermined mathematical relationship for the pulse shape, the shape of the pulse is defined in order to meet desired cost parameters or system criteria. There is freedom to select new pulse shapes that allow many system criteria to be balanced against each other. The trade-off relationship between two parameters is no longer so restricted.

Only the Gaussian and root raised cosine pulses have been considered for use in modulators of telecommunications systems to date. If other pulse shapes are used, there will be different costs associated with different shaped pulses. By optimising the pulse shape in terms of system criteria, a pulse shape can be determined that best meets system requirements. The modulation scheme can be used as one of the variables in deciding a desired balance of costs. A representative list of system criteria or costs includes Amplitude, Bit Error Rate (BER), Energy and Bandwidth. These are defined in more detail below.

(i) Adjacent Channel Power (ACP) Amplitude variation on a modulated signal. Given a constant amplitude of 1 , the error in amplitude can be given by :

{ absolute value - \ \

(ii) BER Error Function - Spectral Occupancy

To calculate this, the amount of noise needs to be determined. This is given by :

{ absolute value of interfering regions}

(iii) Energy Error Function

Required energy - sum of the square of the sample points.

(iv) Bandwidth Error Function

In order to estimate the bandwidth of the pulses, the derivative of the pulse functions (which at this stage are still unknown) are required. This derivative can be approximated as being proportional to the difference between two adjacent pulse values.

The release from the conventional constraints of pulse shape gives the system designer freedom. In particular, for this invention it can be used to look at each element of a telecommunications system independently to allow the system to work more efficiently overall.

The major operating elements in a telecommunications system are fixed stations and mobile stations. The considerations for each of these are somewhat different. Both have bandwidth constraints imposed by the systems in which they are operating. However, whereas the base station is relatively free from concern about power output, the mobile station is limited in the amount of power it can practically transmit.

The present invention allows the pulse shaping for the up-link to a fixed station and the down-link to a mobile station to be optimised independently to take account of the specific requirements of each.

The respective system criteria may therefore be designed to enhance the performance of the respective devices. Alternatively both criteria can be optimised to suit the weakest device.

The invention may be applied to any modulation scheme.

One method by which the pulse shape can be designed will now be described.

Laurent has suggested that a Gaussian pulse can be approximated by the superposition of AM pulses (Co, Ci ...etc.), these pulses being a fixed family of pulses which are functions of cos and sin.

In embodiments of this invention Laurent's theory that a pulse can be approximated by the superposition of components has been used as the basis of ascertaining a pulse shape which meets the criteria required by a particular communications system. This may be done as follows. Firstly, the fixed function components in Laurent's superposition expansion are replaced by one or more functions representing respective unknown pulse components. Then cost functions are looked at (e.g. BER, bandwidth, amplitude, AFC). That is, the errors from the values that the particular system requires are considered. The weightings of the cost functions can be varied so as to tailor the results. Values for each function are then determined, for example using a (commercially available) optimiser, which minimise these cost functions and thus give a pulse shape which meets the specified system requirements.

Preferably two functions are used to optimise the pulse shape as this provides more optimal pulse shaping than just using one function.

More specifically, the method can be implemented as follows :

Firstly, Laurent's formula is considered. According to Laurent's formulation:

^κ>^N-ⁿ'Cκ,_n<τ₊AT - Equation (1 )

N 1-1 where S(t) is the signal at time t AK_,N ⁼ Σ _n ^~ ∑ a_N-r∞κ_,i n=-∞ z^'=l

L-\

Cκ(t)=$₀(t) xnSi-L-«_K,i(t) ( ≤ K≤ M- ) ι = l Instead of using Laurent's pulses, Q_κ , , we wish to use an alternative pulse,

PULSE ] _κ , which is as yet unknown, but for which we wish to determine an appropriate value depending upon requisite error function requirements. Substituting this in equation 1 gives :

S +ΔΓ- Σ ∑^lJ^AK,N -n'prjlSE _κ,nΥ+AT -Equation (2) A=O «'=0 where J=V-Ϊ

As mentioned above, PULSE is unknown as yet, but, in this embodiment it is read, non zero and of maximum length 8.

In this embodiment we choose to use two components (PULSE [0] and PULSE [1]) to build up S. Hence M = 2. In other embodiments the number of component pulses can itself be used as a variable in determining operating conditions for the respective links.

Expanding equation (2) for M=2 and replacing the function A_κ with a

function of the bit streams o oc₂--- > gives : , (cCA/_4+ccw_3+oc;v_2⁺cCA/^'_l+cc/v)

JΛO,N-5 ( J ^{N N} Pulse [0] [όT] +

_J (cc^+cc^+cc^) _{pulse [0] [2T + Sr] +}

J ^{∞N-4⁺∞N-³⁾ _{pu|se [0 [3τ +} sτ] ₊ J ^∞N~4 Pulse [0][4T+_oT] ₊

Pulse [0] [5T+oT + J ^(~∞N-⁵⁾ Pulse [0][6T+όY] + j ⁽- N-5- N-6 _{pu[se [Q][7T+δτ] +}

J (- N-5- N-6- N-7⁾ _{pulse msτ+δr] +} j _ (θC ^yΛv _4⁴ + OC Λv_ ⁵-l + CC N W_ ⁱ2 + 0C N W- 1 ⁱ+CC NΛ/ ^O - a7 _N A - 1\ — pu .lse [1] [,57] +

j (CC NΛ/-_44+OC JΛV/^'_- 3Ϊ+CC NΛ/_ 12 + OC NW_I 1— CC NΛ/-? V) _ p_uise [1] [r + T] +

j „ (oCΛf-4⁺cCΛ/_τ+oCΛ/_2+cCΛ^'-'?) N ⁱ N ⁵ P ...u .lse [l] [2r+c5 „ ^{N N i} T] +

J ^-4-∞N-3-∞N-4 _{pu|se [l] [3T + sr] +}

J I**-*^'*"- pu|_Se [1] [4Γ ₊ YT] + J ^{~∞N-⁶⁾ Pulse [1] [57+^57] +

J ^(_OC^-^5+OCΛf-⁷⁾ Pulse [l] [6T₊^] ) Equation (3)

Since oc denotes a bit, it must be plus or minus 1. Hence each term in equation (3) can be identified as to whether it is real or imaginary (assuming that the pulse function is real).

eg: Taking the first term of the equation :

J (∞_N-₄ ⁺ ∞N-3 ⁺ ∞N-2 ⁺ ∞N-l ⁺ ∞N^ O _N-4>∞_N_₂' N ^{= odd →} imaginary ∞N-3'∞N-I = even →- real.

Hence it is possible to calculate the absolute value of this expression as a function of the bits (∞s) . A decision to be made is what oc is sent at time N. (In an ideal system this will be the signal received at baseband).

Looking at equation 3 (e.g. for a simple receiver), it can be deduced that the bit cc_N_₄ is transmitted at time(N + 4)r as it is on its own. It is imaginary, and the interfering (ie other imaginary) pulses must be taken into account. The real terms in this expression can be totally ignored both for the interfering terms and the absolute value of the pulses. The interference should be minimised. The BER performance can, for example, be improved by making the terms Pulse [0] at (N + 4)T large compared to the absolute value of all the other terms.

Therefore, given an c sequence of :

{ N N-l,"- °CN-7^ ⁼ {U»U,U,U} . the absolute value of the pulse at time AT can be calculated in terms of the unknown pulses. The absolute value of the interfering terms at time AT can also be calculated in terms of the unknown pulses. This is performed for every possible combination of 1 ,-1 for oc^ to oc₇ (ie all 2 ⁼ 256 possibilities). For each possibility an expression both for interfering terms and absolute value are obtained.

In this embodiment, the pulse is required to meet certain criteria with regard to power, BER, AFC and bandwidth. Hence, error functions for these are determined.

Given an oversampling of 8, AT can take on the following values

. _ ,_ T IT 3T 77Λ

ΔΓ = {o,— , — , — ... — }

8 8 8 8

Clearly, the oversampling rate can be altered depending upon the level of pulse sampling required.

The amplitude and BER costs are calculated for AT taking each of the above values. The total cost for each is the addition of all the 8 expressions obtained over the possible sequences. The pulse can be specifically designed based on system requirements by weighting the above error functions (for example 0.3 for power, 0.3 for BER and 0.4 for bandwidth or if a system requires only, for example, bandwidth considerations, 0 for power and BER and 1 for bandwidth). More weight can be added to whatever is causing a problem. The only restriction is that the total weighting must equal + 1.

Now the total error function is expressed in terms of the unknowns, namely, χ_{0 i}(i = 0 to 71 ) and %. .(i = 0 to 55). To determine appropriate values for the unknowns, and thus deduce the pulse shapes, this expression is minimised using a conventional off-the-shelf optimiser, for example.

This method provides an additional system criteria that can be varied to produce a pulse design and that is the number of pulse components used for shaping. The number of components used for the respective transmission directions can be different.

Further description can be found in UK applications 9805504.9 and 9814300.1 appended to this document.

This method can be used to determine the optimum parameters for the up-link of say a CDMA or GSM system using a determined set of weighted cost parameters and then recalculated using a different set of cost parameters for the down-link.

In accordance with a second aspect of the invention there is provided a transceiver for a digital telecommunications system in which first and second communications devices communicate by transmitting information in accordance with a predetermined modulation scheme, the first and second communications devices transmitting pulses shaped prior to transmission indicative of data, the transceiver comprising a processor for providing a string of data bits, a filter for shaping the transmitted pulse in accordance with a first set of system criteria and a filter operating in accordance with a second set of system criteria predetermined for receiving a string of data bits having a pulse shape determined in accordance with a second set of system criteria different from the first set of system criteria.

Embodiments of the invention will now be described, by way of example, with reference to the accompanying drawings, of which,

Figure 1 is a GSM transmitter according to an embodiment of the present invention,

Figure 2 shows the GSM frame structure,

Figure 3 is a GSM receiver according to an embodiment of the present invention,

Figure 4 is a CDMA transmitter according to an embodiment of the present invention,

Figure 5 is a CDMA receiver according to an embodiment of the present invention,

Figure 6(a) is a CDMA receiver according to a preferred embodiment of the present invention; and

Figure 6(b) is a CDMA receiver demodulator stage according to a preferred embodiment of the present invention. In order to design a transmitter one of the first decisions that has to be made is the duration of the phase pulse. This is practically a decision on how to determine how the pulse shaping is to be achieved. The duration of the phase pulse determines how much history is included in each signal. There are two elements of the transmitter that can be involved in the pulse shaping, a filter and a look-up table. If the duration of the phase modulating pulse is relatively short eg 4T where T is the symbol/bit period. The 'history' ie the impact of previous data bits on the current data bit lasts for 4T. The tails of the previous 3 bits are therefore superimposed onto the current bit and there are 2⁴ possibilities for each bit. If too much history is included it consumes more processor time to map each of the possible histories and therefore makes a look-up table approach to pulse shaping impractical. However with a sufficiently faster adder the output could actually be calculated at a cost in terms of, amongst other things, power consumption.

In most systems, the handset or mobile unit is restricted in terms of power available and one of the most significant considerations is power consumption. An important factor in power consumption, and therefore talk and standby times, is the efficiency of the power amplifier. The handset is a consumer item cost, and therefore complexity, are also issues. In embodiments of this invention, the shaping of the pulses including composition of the transmission output for the up and down links respectively can be used to have a positive effect on handset performance.

Discussion will continue with an example considering an option for the design of a handset for a GSM system.

As the handset has significant power constraints, the up-link (handset to base station) is to be optimised for ACP (amplitude variation on the modulated signal). In a GSM system, the closer the pulse shape is to a pure gaussian, the nearer it is to a constant amplitude signal.

If the handset uses the Laurent-type pulse component solution described earlier to shape the pulse, the more pulses of the Laurent summation that are used, the closer the approximation is to the optimum shape. The Laurent-type pulse shaping for a handset will typically be achieved by calculating the shape of the Laurent pulses by optimising for Amplitude to give a power consumption advantage. For GSM, if amplitude is the prime cost function, this will mean a closer approximation to a gaussian.

When using Laurent-type pulses to construct a desired pulse shape, the higher the BT product (Bandwidth x symbol period), the more power there is in the first pulse. For GSM, the BT product is 0.3 and a phase pulse of duration 4T can be used in practice, the lower the BT product, the longer the duration of pulses.

Other cost functions can be included in the optimisation of the pulse shape for the handset as described above if desired. This would be the case if power amplifier efficiency was only one of the parameters to be considered important. Provided the BT product is high, a single pulse will provide sufficient power for the signal transmission. This is likely to be the case for gaussian or near gaussian shapings which are preferable for transmission in existing GSM systems.

Another consideration outside the listed cost functions for the handset is the computational overhead during demodulation. This is effected by the manner of the pulse transmission to the handset in the down-link (base station to handset). This could be quantified and added to the optimisation routine. For the base station, the power amplifier efficiency is by no means as important as for the handset. This is because the source of power is less restricted. The designer then has a choice. The transmission at the base station could be optimised for power and other costs with weightings determined for the efficient operation of the base station. However, as the power efficiency of the handset is the critical element in the system it is sensible to consider the handset when designing the output of the base station as the handset will be required to demodulate the signal and the type and the composition of the signal will have an impact on the computational overhead of the handset and consequently the power consumption.

The base station has far more available power than the handset. This means that when a gaussian-like pulse is desirable and the pulse is generated by the superposition of pulse components a single pulse component can carry enough power to convey the data. Superposing two pulses to provide the output pulses at the base station although having some benefit in the power efficiency at the base station complicates the transmission process and provides little if any advantage to the handset.

Taking a holistic approach to the system, it is better to reduce the computational overhead of the handset by restricting the number of pulses transmitted by the base station. Ideally in the GSM system the number of pulses used to shape the modulated signal transmitted by the bases station should therefore be kept to one.

GSM conventionally comprises a frame structure as shown in Figure 2. Figure 1 illustrates a GSM transmitter suitable for use in a handset in accordance with an embodiment of the present invention. A bit sequence 101 to be transmitted is input to a frame builder 102 of the transmitter, which puts the bits in the appropriate portion of a burst within a time slot of a TDMA frame. The bit stream is then forwarded to a modulator 104. Conventionally this modulator would be a GMSK modulator. However, in this preferred embodiment the signal is not put through a Gaussian filter. Instead, a lookup table 106 defines a pulse function whose shape has formed from a calculation of the first two pulses of a Laurent-type series. Instead of a look-up table a multi-tap FIR digital filter with characteristics calculated to provide pulse shaping equivalent to the first two Laurent-type pulses of the optimisation could be used. A clock 105 provides the carrier signal as is conventional.

The modulated signal is input to a digital analogue converter 107. This analogue signal is then reconstructed by reconstruction filter 108. This filter might typically comprise a switch capacitor filter for performing some of the spectral shaping and an analogue filter, such as an RC filter, for mainly dealing with residual shaping. Finally, the signal is amplified by a power amplifier 109 and is transmitted via antenna 110.

The transmitter at the base station will look the same, but the look-up table or multi-tap FIR digital filter will hold data that provides a pulse function whose shape is formed from only the first pulse of a Laurent-type series. As a single shaped Laurent pulse has a greater amplitude variation than two combined pulses, the efficiency of the base station transmitter will inevitably be not as great as that of the handset. However, as the power constraints can be tolerated and the cost involved in providing a PA with greater linearity at the system level, is less critical than for the handset that is a consumer product, this trade off is easily acceptable.

Figure 3 shows a GSM receiver in accordance with an embodiment of the invention. Similar receivers can be utilized by both the handset and the base station. In this embodiment, the handset will receive a pulse transmitted by the base station formed by the first pulse of a Laurent-type series. The base station will receive a pulse transmitted by a handset. This pulse will be composed of two Laurent-type pulses to optimise the power efficiency at the handset. Because for a gaussian shape the majority of the power is in the first pulse of the Laurent-type series, there is typically sufficient power in the first Laurent-type pulse to make this the only pulse that needs to be decoded by the base station. If circumstances were such that the was insufficient power in the first pulse, the base station could be designed to decode both the first and second pulses of the series.

For a receiver decoding a single component pulse, the receiver down converts the received signal to the baseband of the receiver 120. This signal is then provided to the complex number sequencer 121 and then to a demodulator 122 to provide a bit sequence 123 in a conventional fashion.

The principles for optimising the up-link and down-link pulse shapings in a spread spectrum system are the same as for GSM although the cost parameters involved are rather more complex. This is because bandwidth has been the prime consideration in determining the modulation scheme for spread spectrum systems leaving more work to be done in terms of power efficiency by shaping the pulse. The other cost parameters then need to be monitored closely to give an overall acceptable solution.

In spread spectrum systems unique digital codes, rather than separate RF frequencies or channels are used to differentiate mobile stations. The codes are shared by both the mobile stations and base stations and are called pseudorandom binary codes, one type of which is the gold code.

In the introduction, the issues for the various systems were discussed. In

CDMA systems, the power output of a handset is very important as the typical, root raised cosine shaping selected for bandwidth efficiency of the output pulses is ACP inefficient for use with non-linear amplifiers as the constant amplitude variation is much greater than for GSM using GMSK shaping. When designing and considering how to produce a pulse shape for

CDMA, a number of criteria can be taken into consideration. Important criteria are the output energy of the signal and the efficiency of the power amplifier.

In this embodiment of the invention, as with the GSM embodiment, an important design factor for the handset is the efficiency of the power amplifier in providing sufficient output power to be received by a base station. The power is limited as the handset is mobile and carries its power supply with it. Because the amplitude variation is significant, the non-linearity of the power amplifier causes the spectrum to spread more widely than is ideal for the typical root raised cosine pulse shape.

By producing a pulse shape that optimises for power efficiency, using Laurent-type pulses as described above the efficiency of the power amplifier can be improved. The pulse shape can be modified intricately to produce ever greater improvements in efficiency. The question arises as to how many Laurent-type pulses it is worthwhile including in the pulse shaping.

In order to apply Laurent-type pulses method to optimise pulse shape the duration of the phase pulse needs to be determined. This has an impact on the BT. For CDMA with a phase pulse of duration 6T the BT is approximately 0.15. The higher the BT product the more power resides in a first pulse of the Laurent-type series. For GSM most of the power is in the first pulse. The BT of 0.15 is low and this means that more than one pulse is necessary to provide the requisite output power for transmission. It has been found that at least two Laurent-type pulses are needed to provide the desired output power or reduction in frequency spread for a BT of 0.15. Figure 4 illustrates a Code Division Multiple Access (CDMA) transmitter according to an embodiment of the invention. CDMA conventionally comprises a frame made up of a dedicated physical data channel (DPDCH) and a dedicated physical control channel (DPCCH). A bit sequence 301 to be transmitted is input to a frame builder 302 of the transmitter, which puts the bits in the appropriate part of the frame (i.e. in the DPDCH).

The bit stream is then spread across the spectrum by the Gold Code Encoder. This Gold Code Encoder 303 operates as follows.

Given {c₀,c ...c_Λf_₁ } bit stream

and {f₀fι....f| i-ι} frame sequence

(i.e. M symbol bits)

the output of the Gold Code Encoder 303 is a sequence with N x M terms having the following elements :

ro^C0 » *0^Cl » ^{■ • ■} *0^CN-1 » *1^C0 "'*1 ^CN-I -" —J M-\ o ""J M-\ N-\ )

Hence, there are MN chips to modulate.

A modulator 304 modulates these MN chips output by the Gold Code Encoder 303 on to a carrier, which is output by clock 305. The modulator 304 generally used in CDMA systems such as IS95 is a linear QPSK modulator. The bandwidth of the signal output by the modulator 304 is directly related to the spectrum of the pulses that are used to make up a look-up table 306. Conventionally, for a CDMA system, this lookup table would comprise data defining a root raised cosine. However, in this embodiment of the present invention, the look-up table defines a different pulse whose shape has been optimised with reference to desired cost functions and is produced using a look-up table holding data representing the first two AM pulses according to Laurent's superposition theory, these pulses being a fixed family of pulses which are functions of cos and sin as described earlier. In one example the

ACP is optimised for a non-linear power amplifier.

The output of the modulator 304 is input to a digital-to-analogue converter 307. The analogue signal is then reconstructed by a reconstruction filter 308. A reconstruction filter might typically comprise a switch capacitor filter for performing some spectral shaping and an analogue filter, such as an RC filter network, for mainly dealing with residual spectral shaping. Once the signal has been reconstructed, it is input to a power amplifier 309, which amplifies the signal for transmission by the antenna 310.

As the power in the signal will be carried in two pulses rather than one it is important that a receiver is able to receive both the first and the second pulse in order to reconstruct the message sent by the handset.

Figure 5 is a block diagram of a spread spectrum receiver with a demodulator. In this embodiment, the receiver complements the CDMA transmitter of Figure 4. It comprises an antenna for receiving a spread signal, frequency downconverting circuitry 401 , analogue to digital converter 402, means for storing the receiver's code and a despreader 404. The despreader comprises means 405 for transforming the receiver's code according to the present invention, a correlator 406 for correlating the received signal and the transformed code, and a comparator 407 for determining the sign of the received signal. The code transformer 405 may solely comprise the transformation for detecting a + 1 bit. In this event, the comparator assumes it is a +1 if the value of the correlated signal output by correlator 406 is above a certain threshold, and -1 if it is below this threshold. However, this type of receiver requires more complex demodulation. Simpler demodulation is possible if the transformer 405 also has a transformation for detecting a -1. In this event the comparator determines the sign which produces the largest value. That value should be well above the noise floor, and hence, no complex demodulation is required to determine whether in fact the received signal is a -1 , intended for that receiver, or noise resulting from signals for other receivers.

Figures 6a and 6b show a CDMA receiver according to a preferred embodiment, which complements a transmitter which transmits a signal constructed using the superposition of a plurality of amplitude modulated pulses. In the case of the present embodiment, two pulses.

As can be seen from figure 6a, the frequency downconverting circuitry 401 comprises at least 1 IF stage 501 , mixers 502a, 502b and low pass filters 503a and 503b. A received signal is put through the IF stage(s) 501 to reduce its frequency to a base band frequency and then the signal is split into its I and Q components and the carrier is removed from the signal, using mixers 502a and 502b and low pass filters 503a and 503b. The signal is then converted from an analogue signal into a digital signal by A/D converters 504a and 504b and forwarded to the demodulator stage 404. Figure 6(b) shows this demodulator stage in more detail.

The code transformer 405 transforms the Gold Code to detect a +1 and a -1 for both of the amplitude modulated pulses which make up the received signal. Exemplary transformations are given below: Transformation 1 (Tl) 505a (to detect +1 symbol, for 1^st AM pulse). y, = (-1)' fori = 0.1,2, ... N-1

Given code (C₀, C,, ... C_N_ gold code where N is the number of elements in the sequence. a, = 1 if Cj = 1 ; for i =0, ...N-1 a, = -1 if C, = 0; for i = 0, ...N-1

b, = b_M+ a, for i = 1,2, ... N-1; d, = y, i^bi for i = 0, 1,2,... N-1 and

There is another transformation that can also be used

Transformation 1b (to detect + 1 symbol, for 1^st AM pulse) using the same notation for d, d, = i^bl for i = 0, 1,2, ...N-1 and

Transformation 2 (T2) 505b (to detect -1 symbol for 1^st AM pulse) y_i = (-1)ⁱfori = 0,1,2, ...N-1

Given code {C₀, C,, ... C_N _,} where N is the number of elements in the sequence a,= -1 if C_f = 1 for i = 0, ...N-1 a, = 1 if C, = 0 for i = 0, ...N-1 b₀ = a₀; b| = b_h1+ a, for i = 1,2, ... N-1; d_i = y_iiⁱfori = 0, 1,2,...N-1 and

There is another transformation than can also be used Transformation 2 b (to detect - 1 symbol, for 1^st AM pulse) d, = i^{" i} for i = 0, 1,2, ...N-1

Transformation 3 (T3) 505c (to detect +1 symbol for 2^nd AM pulse) y_i = (-1)ⁱfori = 0, 1,2, ...N-1

Given code {C₀, C_1f ... C_N _ where N is the number of elements in the sequence a = if C_r = 1 for i = O, ...N-1 a, = -1 if C, = 0 for i = 0, ... N -1 b₀ = a₀ -/+ a_N bi = b_M + a_; - a_M for i = 1 , 2, ... N - 1 ; di = y i^bi for i = 0, 1,2,... N-1 and

There is another transformation than can also be used

Transformation 3 b (to detect - 1 symbol, for 2^nd AM pulse) ci, = ^bi for i = 0,1,2, ...N-1

Transformation 4 (to detect +1 symbol for 2^nd AM pulse) y_i = (-1)ⁱfori = 0, 1,2, ...N-1

Given code {C₀, C,, ... C_N _ where N is the number of elements in the sequence a,= -1 if = 1 for i = 0_t ... N-1 a, = 1 if Ci = 0 for i = 0, ...N-1 b₀ = a₀ -/+ a_N bj = b_M + a, - a_M for i = 1 , 2, ... N - 1 ; di = y, i^bi for i = 0, 1,2,... N-1 and

There is another transformation than can also be used Transformation 4 b (to detect - 1 symbol, for 2^nd AM pulse) dj = i^"bi for i = 0, 1,2, ...N-1

Likewise, the correlator 406 performs a correlation of each transformed code with the respective pulse of the received signal. For example, the transformed Gold Code associated with the first AM pulse for detecting a + 1 is correlated with the 1^st AM pulse of the received signal (x_h,x_q1) by correlator

506a. The absolute value z of the correlated signal y, is forwarded to the comparator 407. The same stages occur for the first AM pulse for detecting a

- 1 , and for the second AM pulse for detecting a + 1 and for detecting - 1.

The comparator 407 determines whether the received signal is +/- 1. This is achieved by a comparison of the absolute values (z, - z₄) received from the comparator with expected absolute of the values (E., - E₄) assuming the received signal is of the sign being detected. The values E, - E₄ can be precalculated and stored in the receiver. In this embodiment, if the received signal is a + 1 , then the value of z₁ and z ₃ will be close to their associated expected values E, and E₃, so that the values of h₁ and h₃ will be small. In contrast z₂ and z₄ will be much smaller in value than E₂ and E₄, so that the values of h₂ and h₄ will be large. Hence, the comparator determines that a + 1 is received as h₁ + h₃ < h₂ + h₄. Alternatively, if a -1 is received, the value of z₂ and z₄ will be close to their expected values E₂ and E₄, so that the values of h₂ + h₄ will be small, whereas z, and z₃ will be much smaller than E and E₃, so that h₁ and h₃ will be larger values. Hence the comparator determines that a - 1 is received as h₂ + h₄ < h, + h₃. When there is little interference on the channel, the receiver need only perform the correlations for the first pulse.

Similar transformations can be provided for further AM pulses, should they be so desired. However, use of the first two pulses is generally acceptable as most energy is found in these pulses. In the present case because of power inefficiency constraints on the handset, only the first two pulses are sent.

Jointly optimising pulses in general provides more scope for reducing the cost function e.g. In GSM the non-linearity of the PA was better compensated for (with regard to phone noise and spectrum error) by optimising two pulses jointly rather than optimising each pulse separately. In general, two pulses tends to give the bulk of the improvement. In theory N pulses can be used.

In the embodiments of the invention described above the transmitted pulse is another variable in the system that can be used to provide maximum advantage. The pulse shape and the formation of the pulse can be tailored to meet the particular needs of one or more of the elements in the system.

In view of the foregoing description it will be evident to a person skilled in the art that various modifications may be made within the scope of the invention. The cost function weightings can inparticular be varied to meet different priorities.

The present invention includes any novel feature or combination of features disclosed herein either explicitly or any generalisation thereof irrespective of whether or not it relates to the claimed invention or mitigates any or all of the problems addressed.

Claims

1. A digital telecommunications system in which first and second communications devices communicate by respectively transmitting pulses indicative of data in accordance with a predetermined modulation scheme, the first and second devices each comprise means for shaping the respective data pulses prior to transmission, the shaping being applied in accordance with respective system criteria.

2. A digital telecommunications system according to claim 1 wherein the first and second devices are different types of communications device.

3. A digital telecommunications system according to claim 1 or 2 wherein the respective system criteria are criteria designed to enhance perfomance of the respective devices

4. A digital telecommunications system according to any preceding claim wherein the first device is a fixed station.

5. A digital telecommunications system according to any preceding claim wherein the second device is a mobile station.

6. A digital telecommunications system according to any preceding claim wherein the means for shaping the pulses comprises means for superposing one or more pulses to provide the shaped data pulse.

7. A digital telecommunications system according to any preceding claim wherein one of the first and second devices shapes the data pulses by the superposition of a first number of pulse components and the other of the first and second devices shapes the pulses by the superposition of a second number of pulse components in accordance with its respective system criteria.

8. A digital telecommunications system according to any preceding claim wherein the predetermined modulation scheme is Gaussian Minimum Shift

Keying.

9. A digital telecommunications system according to claim 7 or 8 wherein the first and second numbers are different.

10. A digital telecommunications system according to claim 9 wherein the first number is one and the second number is two.

11. A digital telecommunications system according to claim 7 wherein the first and second numbers are two.

12. A digital telecommunications system according to any one of claims 1 to 7 or claim 11 wherein the predetermined modulation scheme is QPSK.

13. A digital telecommunications system according to any preceding claim wherein the system criteria are selected from one or more of the group including power efficiency, spectral efficiency, bit error rate, AFC, Nyquist, and energy efficiency.

14. A digital telecommunications system according to any preceding claim wherein one of the system criteria for the first device is power consumption at the second device.

15. A digital telecommunications system according to any preceding claim wherein one of the system criteria is the complexity of the demodulation of the data pulse.

16. A transceiver for a digital telecommunications system in which first and second communications devices communicate by transmitting information in accordance with a predetermined modulation scheme, the first and second communications devices transmitting pulses shaped prior to transmission indicative of data, the transceiver comprising a processor for providing a string of data bits, a filter for shaping the transmitted pulse in accordance with a first set of system criteria and a filter operating in accordance with a second set of system criteria predetermined for receiving a string of data bits having a pulse shape determined in accordance with a second set of system criteria different from the first set of system criteria.

17. A digital communications system substantially as herein before described with reference to Figures 1 to 3 of the accompanying drawings.

18. A digital communications system substantially as hereinbefore described with reference to Figures 4 to 6 of the accompanying drawings.

19. A transceiver substantially as herein before described with reference to Figures 1 to 3 of the accompanying drawings.

20. A transceiver substantially as herein before described with reference to Figures 4 to 6 of the accompanying drawings.