A METHOD OF GENERATING A DIGITAL PHASE AND AMPLITUDE MODULATED SIGNAL
This invention relates to a method of generating a digital phase and amplitude modulated signal, in particular for creating a binary offset carrier signal. In accordance with a first aspect of the present invention, a method of generating a digital phase and amplitude modulated signal, the digital modulated signal carrying a multiplex of binary symbols from information that is calculated in advance and stored, comprises defining the multiplexing technique and to define a conversion of a set of binary symbols at a time of the multiplex into an enumerated state corresponding to an amplitude and phase state of the modulated signal; storing all possible sampled transition trajectories of amplitude and phase of the modulated signal between two consecutive states; using a pair of states sequentially read from the store as a pointer to an entry of the stored sampled transition trajectories, wherein the entry of the stored sampled transition trajectories contains a sampled transition of the modulated signal for a range of times; and sequentially constructing a complete modulated multiplexed waveform as a succession of segments of the modulated signal, read from the stored sampled transition trajectories. In one example, the information is stored in a plurality of successive look up tables (LUT). For example, use of a pair of LUTs can be achieved with less memory than required for a single, all purpose, LUT and fits well conceptually with the signal synthesis method steps of modulation followed by pulse shaping. Alternatively, the information is stored in a single look up table (LUT). Preferably, the pair of states are read sequentially from the first LUT as a pointer to an entry of the second LUT, together with a finite number of states preceding and following the sampled transition to accommodate for inter-symbol interference effects between transmitted symbols. In accordance with a second aspect of the present invention, a method of generating a binary offset carrier signal comprising a digital phase and amplitude modulated signal generated according to the method of the first aspect, further comprises providing first and second input signals; combining each input signal with first and second spreading codes respectively to produce first and second sub-signals
defining a shaped spectral output signal; providing a third sub-signal; determining an address in the store of an initial phase state using the first, second and third sub-signals; determining an address in the store of a stored sampled section of each input signal waveform; retrieving stored values of initial phase state and sampled signal waveform; and generating the shaped spectral output signal from the retrieved stored values. Preferably, the third sub-signal comprises a third spreading code Preferably, the input signals and spreading codes are converted to digital signals. Preferably, spreading is applied via an exclusive OR function. Preferably, a binary level representing each of the three sub-signals provides the address of the initial phase state for the combination of inputs. Preferably, the stored information is stored as separate in-phase and quadrature components. An example of a method of generating a digital phase and amplitude modulated signal according to the present invention will now be described with reference to the accompanying drawings in which: Figure 1 is a simplified view of conventional binary offset carrier (BOC) generation architecture; Figure 2(a) illustrates sub-carrier within one chip duration; Figure 2(b) shows a generated BOC signal and an unwanted output mask; Figure 3 is a block diagram of a system for signal generation according to the present invention, used to generate a BOC with spreading code and sub-carrier shaping, using a pair of LUTs; Figure 4 is a block diagram of a system for signal generation according to the present invention, used to generate a BOC with spreading code and sub-carrier shaping, using a single LUT; Figure 5 illustrates interpretation of phase states stored in a first look-up table (LUT) in the system of Fig. 4; Figure 6(a) illustrates entries in a second LUT for a shaped sub-carrier within one chip duration; Fig. 6(b) shows an example of a generated BOC signal with shaped sub-carrier pulses;
Figure 7 shows examples of Galileo waveform segments stored in quadrature LUT 2 for an LI signal; and, Fig. 8 illustrates a comparison of signals resulting from a conventional signal generation method and the method of waveform generation according to the present invention.
Direct sequence spread spectrum chip and sub-carrier waveform shaping for generation of signals with controlled out-of-band spectral content in binary offset carrier (BOC) modulation scheme is proposed. The method enables generation of a BOC signal that is compliant to adjacent unwanted output spectrum masks, thus effectively reducing or removing the requirement for an adjacent band protection transmit filter. The method has been designed for Galileo satellite navigation signals, but can also be used for other navigation signals, and in other spread spectrum communication systems where pulse shaping has to be implemented for adjacent frequency band protection, or for signals which allow multiple access on the same channel, such as CDMA for UMTS or CDMA 2000. Binary Offset Carrier (BOC) modulation is likely to be extensively used in next generation satellite navigation systems. Conventional methods of generating satellite navigation signals, for example those used in Galileo, or other communication signals, require a modulator. With the modulator there is a lot of high speed processing, then the processed waveforms are filtered to protect adjacent signals. Without filtering the original signal will fall outside the mask and cause interference. The effect of the filter is to limit the extent to which the signal is outside the mask and so reduce interference. Conventional modulator architecture that is used to generate the BOC modulated signal is given in Fig. 1. Figure 1 illustrates typical inputs for a satellite system, NANA 1 and ΝANB 2 which are multiplexed 3, 4, 5, 6 with respective codes A and B 7, 9 and subcarriers A and B 8, 10. Each output eA(t), eB (t) is input to a coherent adaptive subcarrier modulation (CASM) multiplexer and multiplexed with an output ec(t) from code C 11 that has been multiplexed 12 with sub-carrier C 13. Signals eA(t), eB(t) and ec(t) are all combined together to create a composite, multiplexed signal from which they can then be extracted independently, i.e. one can despread and demodulate, say, eA(t) without knowledge of spreading codes B, and C; all that is required is knowledge of spreading
code A. In phase and quadrature components I, Q are filtered in adjacent band protection filters 14, 15, then passed through digital to analogue converters DAC 16, 17 and reconstruction filters 18, 19. In the BOC-modulated signal generation, navigational data, direct sequence spreading code and sub-carrier are all implemented as rectangular pulses. A spreading sequence chip modulated by a rectangular sub-carrier is shown in Figure 2(a) where fs is the sub-carrier frequency and Tcιl is a chip period. Rectangular pulses are chosen in order to preserve the constant amplitude of the modulated signal, which enables better efficiency of output solid-state power amplifiers (SSPA). The second goal is for the signal to occupy the widest bandwidth possible, which is beneficial in navigation applications. The use of rectangular-shaped pulses, however, results in strong unwanted output introduced in adjacent bands, as shown in Fig. 2(b). Radiation in adjacent frequency bands represents a particular problem in radio astronomy, and this is addressed by relevant ITU-R recommendations. Effectively, ITU-R requirements set adjacent unwanted output masks, similar to the one shown in Fig. 2(b), which cut out signals which fall outside the masks limits. For some satellite navigation systems, Galileo in particular, a transition band of the unwanted output mask is very narrow. This results in severe analogue filtering requirements, which impairs generated signal phase characteristics. Alternatively, the filtering is implemented in the digital domain, but this puts heavy requirements on the digital hardware. In all cases this filtering makes the amplitude of the generated signal variable, removing the benefits of rectangular pulse shaping. The rectangular shape of sub-carrier pulses imposes another limitation on the generating hardware, namely, that the sampling frequency has to be an integer multiple of twice the sub-carrier frequency. This may preclude the use of the same hardware for simultaneous generation of navigation signals with different sub-carrier frequencies. For example, the Galileo system is likely to require simultaneous generation of sub- carriers of 14 and 15 MHz. This combination of sub-carrier frequencies requires the generating hardware to run with a sampling frequency of 420 MHz. Since out-of band unwanted output is caused by the use of a rectangular chip and sub-carrier pulse shape, the present invention proposes a solution to the problem of unwanted spectral output which is to design chips and sub-carrier pulses so that they
are shaped, instead of rectangular. This approach is shown in Fig. 3. Spreading code generators, 20, 21 , 22 provide inputs to a first look-up table 23. Signal eA(t) is NANA signal exclusive OR'd with code A, signal eB(t) is ΝANB signal exclusive OR'd with code B and signal ec(t) is from code C generator 22. A chip rate clock 24 provides an input to the first modulation look up table (LUT) 23. In-phase and quadrature components from the LUT 23 pass through delay cells 25 and also directly to the second LUT 26 which contains waveform segments. A sample rate clock 29 at the second LUT 26 also provides an output to conventional DAC's 27 and reconstruction filters 28. The present invention makes use of the linear nature of the system, which allows signals to be filtered first, then stored in a LUT, thereby doing all the high speed processing offline. Prior to providing the input signals, in-phase and quadrature components of a shaped spectral signal waveform are determined and stored in a first store; and definitions of how the components are combined are determined and stored in a second store. In conventional methods, the multiplication must be done in real time and a finite impulse response filter (FIR) is required. Cost in terms of components and processing power can be prohibitive. In this invention, pre-filtering is applied so that the waveform can be designed to have desired spectral properties, giving a better resulting signal than would be expected in a conventional system. The effect is a signal with minimal ripple and sharp slope transitions. It is particularly useful to have a constant amplitude signal for power amplification, but this would not be possible with a conventional method because of the amount of ripple generated. Conventional methods do not suggest the possibility of pre-filtering because the prior art methods are all directed at narrowband signals, where pre-filtering would be assumed to have the effect of removing important signal information. A particular advantage of being able to do the filtering offline is that the results are obtained with greater precision. This is because the absence of time constraint means that floating point calculations can be used. To do the same processing online, as required by the prior art, requires a huge amount of power and large processors. The functionality of this modulator architecture is explained below. The present invention involves a method of generating a digital phase and amplitude modulated signal, x(t) carrying a multiplex of binary symbols, QΝ(n)={sι(n),
s2(n)...,SN(n)}, where N>1 , t=nT, and T is a signalling period. The signal is generated from information that is calculated in advance and stored in two successive look-up tables (LUT), using the first LUT to define the multiplexing technique and a conversion of the set, QN(Π) of N digital symbols at time, nT of the multiplex into the enumerated state m(n), l≤m≤M, corresponding to the amplitude and phase state of the modulated signal x(t) at t=nT. The second LUT is used to store all possible sampled transition trajectories of the amplitude and phase of the modulated signal, x(t) between two consecutive states, m(n) and m(n+l) using a pair of states, (m(n),m(n+l)} sequentially read from the first LUT, as a pointer to an entry of the second LUT, containing the sampled transition of x(t) between the x(nT) and x((n+l)T). The complete modulated multiplexed waveform x(t) is sequentially constructed as a succession of segments, x(t) for nT≤t<(n+l)T, read from the second LUT. A pair of states (m(n),m(n+l)) sequentially read from the first LUT is used, together with a finite number of states preceding and following the transition, {m(n- K),..., m(n-l), m(n), m(n+l)...,m(n+l+K)}, where K>1, as a pointer to an entry of the second LUT, to accommodate for the inter-symbol interference effects between the transmitted symbols. The waveform is split up into a number of elements. In the first LUT there are all possible elements and in the second LUT there is a definition of how each of the elements are combined. In a similar way to the conventional modulator architecture shown in Fig. 1 , navigation signals NANA and ΝANB are combined with spreading codes A and B. However, both NAN messages and spreading codes are expressed as logical levels (0, 1), and spreading is performed as an exclusive-or operation. Together with spreading code C, three binary streams eA(t), eB(t) and ec(t) have to be multiplexed onto a single carrier. Instead of modulating these signals on respective sub-carriers, though, they are used to calculate an address to access a modulation and multiplexing look-up table (LUT 1) 23. Modulation and multiplexing look-up table (LUT 1) 23 contains a definition of a modulation and multiplexing scheme used to generate desired navigational signals. By changing the content of this table, various Galileo modulation and multiplexing schemes can be implemented, for example, CASM, quadrature phase shift keying (QPSK) or binary phase shift keying (BPSK). Effectively, this table contains an initial
phase state of the modulated signal for all combinations of binary sub-carriers present at its input. Alternatively, LUT 1 may be implemented as two look-up tables, with in-phase and quadrature entries, instead of a single table containing locations of the modulation phase in a complex plane. The implementation of LUT 1 as two tables (in-phase and quadrature) offers a possibility of some required memory space trade-off between the LUT 1, 23 and LUT 2, 26. Alternatively, Fig. 4 illustrates a system using a single modulation and phase shaping LUT 32. The delay cells 25 are no longer required, nor is the chip rate clock. Otherwise, the component parts are substantially the same as for
Referring back to Fig. 3, as an illustration, phase states that have to be stored in LUT 1 in order to generate Galileo LI and E6 CASM signals are shown in Fig. 5 for two possible storage methods. Another improvement, illustrated in Fig. 5, is that only one clock is required, a sample rate clock, whereas conventionally, it was necessary to generate clocks for all inputs and get them aligned. Table 1 shows the content of LUT 1 when it is organised as a single table (a "Complex LUT 1 entry" column) or as two tables (the "In-phase" and "Quadrature" columns). The Table also shows the corresponding phase state of the modulated Galileo signal.
Subcarrier shaping and inter-chip interference (ICI) required for generated signal to be compliant with adjacent unwanted output mask is defined in LUT 2, 26. This LUT is again implemented as two "parallel" tables. One contains samples of in- phase signal component, while the other contains samples of the quadrature component.
Both tables (in-phase and quadrature) contain sampled short sections of the Galileo signal. Length of each sampled section is equal to chip duration, i.e. to a time interval that corresponds to a single entry in LUT 1. However, waveforms stored in LUT 2 do not represent Galileo signals for one chip; it is a transition from one chip to the next that is stored. This is done in order to accurately model the inter-chip interference. Consequently, each LUT 2 entry is a second half of the current chip period and a first half of the next chip period. The LUT 1 contains definitions of phase states φ(n) that correspond to a signal that is to be transmitted, while the LUT 2 contains waveform transitions from state φ(n-l) to state φ(n), as shown in Fig. 6a. In Fig. 6b the effects of pulse shaping on the amount of interference introduced into adjacent frequency bands are illustrated. The high value of product between the chip duration and bandwidth occupied by satellite navigation signals, means that the effects of ICI are most prominent near to chip edges, whereas distortion of the central part of each chip is negligible. As a result of this, segments of waveform can be stored as transitions from one chip state to the next one, and all stored waveforms can start from and end at the same value. In case of Galileo waveforms, this common value is zero. Galileo waveform segments stored in quadrature LUT 2 for LI signal are shown in Fig. 7 as an example. As quadrature LUT 1 contains three values (0, 1 and 2, as shown in Table Table 1), there are nine possible transitions between two consecutive states defined by LUT 1. For an assumed chip rate of 2 MHz for LI Open Service subsignals and sampling frequency of 120 MHz, each waveform entry in quadrature LUT 2 contains 60 signal samples. Shaped sub-carrier pulses enable the use of sampling frequency that is not an integer multiple of sub-carrier frequency; i.e. this enables a use of multirate signal generation. With the same example of Galileo sub-carriers of 14 and 15 MHz, the sampling frequency can be reduced from 420 MHz to 80 MHz. The use of the look-up table generated signals for protection of adjacent frequency bands from unwanted emission assumes that pulse shapes are adequately designed. This is done, for example, by minimisation of the peak of chip-subcarrier pulse by linear programming, subject to the constraints that signal power is constant and all harmonic components are below the output spectrum mask. The approach can be used in Galileo and other navigation signals, as well as in communication systems.
Pulse shaping removes LO harmonics and chip sidelobes. Power amplifier pre- distortion can be implemented as a part of the pulse shaping look-up table. The sub- carrier and spreading code pulse shaping can be implemented independently or in combination, which allows multirate processing. It is not just the spectral content which can be improved. A better peak to mean ratio of the signal can be obtained. For amplitude modulation, a very linear power amplifier is required, which is less efficient. The peak to mean ratio can be reduced, so that the power amplifier is less linear and uses less power. In manufacturing terms, the LUTs are filled in the factory and the same LUT can be arranged to carry out various functions, including using a single LUT for storing all of the data types. In addition, the LUT can be remotely updated, e.g. after installation, so that a change to the type of signal to be generated, is taken into account. This is a7nother advantage over prior art filters which are hard wired, so cannot be updated remotely. The pulse shaping enables a generation of a BOC signal with local oscillator harmonics and chip sidelobes suppressed. The generated signal is compliant to adjacent unwanted output spectrum masks, thus effectively removing the requirement for an adjacent band protection transmit filter. At the same time, the generated signal has the peak to mean ratio minimised, in order to reduce signal distortion introduced by an output power amplifier. Additional improvements can be achieved by modifying the pulse shapes stored in the look-up tables to include predistortion for the power amplifier, thus improving the system efficiency. Furthermore, pulse shaping enables multirate generation of navigation signals. In Fig. 8, a snapshot is shown of satellite navigation waveforms generated using a classical approach, where the waveforms are filtered to conform to the spectrum mask 30, and generated using the method of the present invention where the waveform has been synthesized to conform to the same mask 31. In this example, the waveforms are normalised in amplitude and the X-axis marks signal samples, with a sampling frequency of 122.8 MHz. The waveform 31 generated according to the method of the present invention is almost perfectly flat, whereas the other waveform 30 generated by conventional methods exhibits undesirable ringing. Effective bandwidth is roughly the same, the shape of the spectrum in the band is almost the same and outside of the band is suppressed, as originally required. The
top of the correlation peak is effectively equally shaφ in both cases. Ringing results in the peak being skewed and sides of the correlation peak being wobbly, which is highly undesirable in navigation applications. In principle, the closer one makes the correlation peak to an ideal triangular shape, the better. In practice, the peak may be rounded, or leaning to one side, and its sides may be ragged instead of straight. Removal of ringing is useful in its own way on the transmit side, irrespective of the navigation properties (that manifest themselves at the receive side). Less ringing mean less back-off and better utilisation of the transmit power amplifier which is useful when the transmitter is on a satellite. The desired output signal achieved as a result of the waveform optimisation method of the present invention is not possible using the classical approach. The described method of signal generation has been designed with Galileo satellite navigation signals in mind, but it can also be used for other navigation signals and in other spread spectrum communication systems where pulse shaping has to be implemented for adjacent frequency band protection.