Method And Apparatus For Generating Mary CPM Waveforms From A Superposition Of PAM Waveforms
To reflect advantages of a constant phase modulation waveform, the invention provides a pulse amplitude modulated PAM waveform that is a superposition of Q0≦2L−1 PAM component pulses in each symbol interval such that a significant portion of signal energy over each symbol interval is within the Q0 PAM component pulses. The present invention distributes most signal energy in one pulse and progressively lower energies in the remaining Q0−1 pulses of a symbol interval. The Laurent Decomposition is a special case of the present invention, but the present invention exhibits the energy distribution of the Laurent Decomposition in nonbinary CPM waveforms and in multih (binary and nonbinary) CPM waveforms, where h is a modulating index. All energy is distributed among only Q=2L−1 pulses in each symbol interval, though only Q0<Q pulses may in fact be transmitted in certain embodiments. A method, transmitter, receiver, and computer program are disclosed. Embodiments may exactly duplicate a CPM waveform, or approximate a CPM waveform.
Latest Patents:
 Method and apparatus for operating a combine harvester
 Display device comprising remaining portion of inspection line with cut edge
 Method and apparatus for allocating access and backhaul resources
 Packaged semiconductor device having a shielding against electromagnetic interference and manufacturing process thereof
 Method, apparatus, and computer program for a mobile device for reducing interference between a first mobile communication system and a second mobile communication system
The present application is a divisional of a parent application having Ser. No. 11/174,263, filed Jun. 30, 2005, now U.S. Pat. No. ______.
TECHNICAL FIELDThe present invention relates to an Mary constant phase modulated signal, and is particularly directed to decomposing that signal into a superposition of pulse amplitude modulation signals with high accuracy, and communicating with the decomposed signal.
BACKGROUNDMary signaling can be regarded as a waveform coding procedure, and refers specifically to signal processing where the processor accepts k data bits at a time and instructs the modulator to produce one of M=2^{k }waveforms. Binary signaling (M=2) is the special case where k=1. Typically, Mary refers to nonbinary and that convention is continued throughout this disclosure, but the distinction is also made explicit in certain instances. For pulse amplitude modulation (PAM), the signaling order M refers to the number of unique discrete amplitude values over which the pulse is allowed to vary. Instead of transmitting a pulse waveform for each bit (where the rate would be R bits per second), parse the data into kbit groups and then use M=2^{k}level pulses for transmission. Each pulse waveform then represents a kbit symbol moving at the rate of R/k symbols per second, which reduces the required bandwidth as compared to pulsemodulating each bit because symbols are transmitted as opposed to bits, though at a rate slower by a factor of k. As M (and k) increases, the receiver finds it more difficult to distinguish between, for example, octal pulses (M=8) versus binary pulses (M=2). The result in the prior art is that, as k increases (higher order Mary signaling) with orthogonal signaling, error performance increases or the required signal to noise ratio SNR (technically E_{b}/N_{0}) is reduced, at the expense of bandwidth. For nonorthogonal signaling, the tradeoff is reversed in that increasing k improves bandwidth at the expense of lower error performance or increased SNR.
Constant phase modulation (CPM, also known as continuous phase modulation due to its smooth phase transitions between symbols) is a signal modulation technique that increases bandwidth efficiency by smoothing the waveforms in the time domain. Bandwidth efficiency is gained by concentrating the signal's energy in a narrower bandwidth, enabling adjacent signals to be packed closer together. Inherent in this smoothness is the fact that symbol transition features are muted, and many symbol synchronization schemes depend on those transition features being definite. To smooth the time domain signal, various CPM techniques generally rely on one or more of the following features: using signal pulses with several orders of continuous derivatives; intentionally injecting some intersymbol interference so that individual pulses occupy more than one signal time interval; and reducing the maximum allowed phase change per symbol interval.
As a cursory description of CPM, a binary singleh CPM waveform can be expressed over the nth symbol interval as
where t denotes time, T denotes the symbol duration, a_{i}∈ {±1} are the binary data bits and h is the modulation index. The modulation index h is the ratio of the frequency deviation to the frequency of the modulating wave, when using a sinusoidal wave as in CPM. The phase function, q(t), is the integral of the frequency function, f(t), which is zero outside of the time interval (0,LT) and which is scaled such that
An Mary singleh CPM waveform is the logical extension of the binary singleh case in which the information symbols are now multilevel: i.e., a_{i}∈ {±1, ±3, . . . , ±(M−1)}.
Finally, an Mary multih CPM waveform can be written as
where a_{i}∈ {±1, ±3, . . . , ±(M−1)} and the modulation index, h_{n }assumes it value over the set: {h(1) . . . , h(N_{h})}. In one implementation, for example, the modulation index may cycle over the set of permitted values.
In his seminal work entitled “Exact and Approximate Construction of Digital Phase Modulations by Superposition of Amplitude Modulation Pulses (AMP)”, IEEE Transactions on Communications, vol. COM34, No. 2, February 1986, pp. 150160, P. A. Laurent has shown that any binary singleh CPM signal can be exactly represented by the superposition of pulseamplitude modulation (PAM) waveforms.
This is termed the Laurent Decomposition, and a_{i}∈ {±1}, and {b_{k,n}} represents the pseudodata symbols, which are obtained in a nonlinear fashion from the binary data symbols. Laurent lays the theoretical groundwork in the above paper for representing any constant amplitude binary phase modulation as the sum of a finite number of timelimited pulse amplitudemodulated pulses. Hence, Laurent shows that binary singleh CPM, which may be appear to be rather complex in its classical representation (equation [1]), can be replaced by a much simpler notation by using what has become known as the Laurent Decomposition.
The Laurent Decomposition of equation [4] expresses a binary singleh CPM signal as the sum of 2^{L−1 }PAM waveforms (where L denotes the number of symbol intervals over which its frequency function is defined). The Laurent pulses, c_{k}(t), are obtained from the phase response of the CPM signal. An important characteristic of these pulses is that the signal energy is unevenly distributed amongst them and that the pulses are distinctively ordered. Thus, c_{0}(t) is usually the “main pulse”, which carries most of the signal energy (often upwards of 95%), c_{1}(t) contributes much less energy and the last pulse c_{Q−1}(t) contributes the least amount of energy. Thus, in many cases of practical interest, the CPM waveform can be approximated using only the PAM construction of the “main pulse”
Because the pulses of Laurent's approach are defined in order of decreasing energy, equation [5] can be broadened somewhat to sum the energy over the first pulse or over the first few pulses of the decomposition in order to synthesize an “almost binary singleh CPM” signal. Thus, for many cases of practical interest, the binary singleh CPM signal is well approximated as
The Laurent Decomposition is important because it linearizes the binary CPM waveform, which greatly simplifies receiver algorithms for binary CPM by enabling them to use Laurent's linear approximation of the received CPM signal as a single pulse as in equation [5]. Equation [6] enables simplified design of “almost binary CPM” transmission schemes as well as for the simplification of receiver design algorithms by using only a few of the leading Laurent pulses rather than the true CPM waveform itself.
However, with increasing k (and thus increasing M in an Mary waveform), further extensions of the Laurent Decomposition do not appear to preserve the mathematically elegant PAM signal structure that makes this decomposition so useful for generating approximations of binary singleh CPM waveforms.
Specifically, in a paper entitled “Decomposition of Mary CPM Signals Into PAM Waveforms”, IEEE Transactions on Information Theory, vol. 41, No. 5, September 1995, pp. 12651275, U. Mengali and M. Morelli extend the Laurent Decomposition to include multilevel (singleh) CPM signaling and show that Mary singleh CPM waveforms have the following PAM decomposition
where a now denotes the Mary data symbols a_{i}∈ {±1, ±3, . . . , ±(M−1)}, Q=2^{L−1}, and P is an integer satisfying the conditions
2^{P−1}<M≦2^{P}. [8]
The Mengali and Morelli approach is seen to view an Mary CPM signal as the product of P binary CPM waveforms, apply the Laurent Decomposition to each individual factor, and then write the final expression as the sum of PAM components. In general, this approach yields 2^{P−1 }PAM component pulses of significant energy. Furthermore, unlike Laurent's solution for the binary case, their approach does not result in a PAM decomposition in which the component pulses are naturally defined in terms of decreasing signal energy.
E. Perrins and M. Rice also extend the Laurent Decomposition in two papers: “Optimal and Reduced Complexity Receivers for Mary Multih CPM”, Wireless Communications and Networking Conference 2004, pp. 11651170; and “PAM Decomposition of Mary multih CPM”, believed to be submitted to IEEE Transactions on Communications for future publication. The work of Perrins and Rice generalize the Laurent Decomposition by applying it to Mary multih CPM waveforms. In their approach, Perrins and Rice first derive the PAM decomposition for the binary multih case and then extend this result to the general Mary multih case in order to show that
The notation i=i mod N_{h}, where N_{h }denotes the number of modulation indexes (i.e. h={h(1), h(2), . . . , h(N_{h})} and “mod” indicates modulo addition ⊕. An important distinction between equation [9] and the above work of Mengali and Morelli is that {g_{k,n}(t)} is a now set of N_{h}·N component pulses. Note that the PerrinsRice derivation results in the generation of N_{h}·2^{P−1 }“main pulses” that carry the most significant proportion of the total signal energy.
Each of the above extensions of the Laurent Decomposition is seen to not preserve its mathematical simplicity, which makes it so valuable in mirroring or approximating a binary CPM waveform. What is needed in the art is a method and apparatus to linearly decompose an Mary CPM signal, with exact or reasonable approximation, so that efficient algorithms and hardware may be designed. This need is seen for both singleh and multih modulation. The Laurent Decomposition itself is seen as adequate for binary singleh CPM, so the need lies primarily in the area of nonbinary singleh and all multih CPM decompositions.
SUMMARYThe foregoing and other problems are overcome, and other advantages are realized, in accordance with the presently preferred embodiments of these teachings.
In accordance with one aspect of the invention is a method of transmitting a signal. In this method, a signal is modulated onto an Mary pulseamplitude modulated PAM waveform that is a superposition of Q_{0}≦2^{L−1 }PAM component pulses in each symbol interval such that a significant portion of the signal energy over each symbol interval of a burst is within the Q_{0 }PAM component pulses. The modulated signal is then transmitted. The value M is greater than 2 to indicate nonbinary PAM, and L represents a number of symbol intervals. A significant portion is generally more than about 95% of signal energy, and preferably is about 98% or more of the signal energy. Where Q_{0}=2^{L−1 }component pulses, essentially all signal energy is within the superposition.
Another aspect of the invention is similar to the method immediately above, but the PAM waveform uses more than one modulating index and may be binary with M=2.
In accordance with another aspect of the invention is a transmitter that includes a signal source, a modulator configured to output an Mary pulseamplitude modulated PAM waveform, a multiplier, an antenna, and a processor coupled to a memory. The nonbinary PAM waveform is a superposition of Q_{0}≦2^{L−1 }PAM component pulses in each symbol interval. The processor is configured to drive the modulator to shape the Q_{0 }PAM pulses in each symbol interval such that a significant portion of the signal energy for each symbol interval of a transmission burst is carried in that symbol interval's Q_{0 }PAM pulses.
Another embodiment of the invention is similar to the transmitter immediately above, but the modulator is configured to output a binary, multih pulseamplitude modulated PAM waveform, where h>1 is a number of modulating indices.
In accordance with yet another embodiment, the present invention is a memory embodying program of machinereadable instructions, executable by a digital data processor, to perform actions directed toward modulating an input signal. The actions include determining an Mary pulseamplitude modulated PAM waveform that is a superposition of Q_{0}≦2^{L−1 }PAM component pulses in each symbol interval such that a significant portion of the signal energy over each symbol interval of a burst is within the Q_{0 }PAM component pulses, and combining the Mary PAM waveform with a signal to be transmitted, and transmitting the modulated signal. The term M is greater than two, so the PAM waveform is nonbinary. The term L represents the number of symbol intervals over which the frequency function is defined (see Equation [2] for an example).
Another aspect of the invention is similar to the program immediately above, except the PAM waveform uses more than one modulating index and may be binary with M=2.
In accordance with another aspect, the invention is an apparatus for modulating a signal. The apparatus has signal source means, modulating means, and transmitting means. The modulating means is for determining, for each symbol interval of a transmission burst in which symbols are modulated over M>2 discrete amplitudes, a plurality of Q_{0 }pulses that are shaped such that the pulses are ordered by an amount of signal energy they carry, the modulator means further for superimposing the plurality of Q_{0 }pulses into a combined waveform with continuous phase that modulates an output of the signal means for transmission. The transmitting means is for transmitting the signal modulated on the combined waveform.
Another aspect of the invention is similar to the apparatus immediately above, except the PAM waveform uses more than one modulating index and may be binary with M=2.
Another aspect of the invention is an apparatus for receiving a signal. This apparatus has means for receiving a nonbinary pulse amplitude modulated PAM signal that is characterized in that, for each symbol interval, a plurality of pulses that exhibit descending levels of signal energy. The apparatus further has demodulating means for demodulating symbols from the pulses.
Another aspect of the invention is similar to the apparatus immediately above, except the PAM waveform uses more than one modulating index and may be binary with M=2.
Yet another aspect of the invention is a memory embodying a program of machinereadable instructions, executable by a digital data processor, to perform actions directed toward demodulating a received signal. These actions include determining a generalized phase function of a received nonbinary pulse amplitude modulated signal, deriving a function g_{k,n}(tnT) from the generalized phase function, where T is a symbol duration. The actions further include, for each n^{th }symbol interval, resolving pseudo symbols b_{k,n }using the function g_{k,n}(tnT), and then determining symbols from the pseudo symbols.
Another aspect of the invention is a method to construct a signal x(t, a, h). In the method, a function g_{k,n}(t) is constructed and shifted by nT. The shifted function is multiplied by a pseudo symbol b_{k,n}, summed over Q_{0 }pulses, and the Q_{0 }pulses are summed over n symbols. In the above notation, t is a time index, a={a_{i}} represents a phase of a complex data, h is a modulating index, and T is symbol duration. An exemplary mathematical representation of this method is shown at equation [29].
Further aspects and implementation details are given below.
The foregoing and other aspects of these teachings are made more evident in the following Detailed Description of the Preferred Embodiments, when read in conjunction with the attached Drawing Figures, wherein:
An analysis of the prior art extensions of Laurent's work detailed above indicates that they lose certain of the Laurent Decomposition's mathematical elegance and simplicity for the following reasons:

 The Mary CPM decomposition yields several pulses of significant energy, so that the concept of a single “main pulse” for Mary CPM does not exist. For example, the MengaliMorelli decomposition of Mary singleh CPM generates 2^{P−1 }“main pulses” and the PerrinsRice decomposition of Mary multih CPM yields N_{h}·2^{P−1 }“main pulses”, where P˜log_{2}M and N_{h }denotes the number of modulation indices.
 The PAM component pulses in the Mary CPM decomposition are not distinctively ordered as in the binary CPM case, which implies that it is not trivial to determine which of the component pulses contribute the most significant proportion of energy;
 The number of PAM component pulses in the decomposition of Mary CPM is exponentially increased over the number of terms that are required in the expansion of binary singleh CPM.
The present invention can be used to decompose any CPM signal—whether binary CPM, Mary singleh CPM or Mary multih CPM—into an equivalent PAM waveform in a manner that preserves the mathematically elegant structure of the Laurent Decomposition for binary CPM waveforms.
T
where b_{k,n }denotes the pseudodata symbols, which are derived from the Mary symbols in a nonlinear fashion, and g_{k,n}(t) denotes the PAM component pulses. Proof of equation [10] for all CPM waveforms is predicated upon the expression of any complexbaseband CPM waveform as a binary multih waveform.
P
where T denotes the symbol duration, a={a_{i}} are the phases of the complex data symbol. Hence, they are binary random variables: a_{i}∈{ ±1}. A pseudomodulation index, h_{i}, may take on any one of the definitions found in Table 1, or may easily assume other definitions which are apart from those listed in table 1, but which are consistent with the upcoming derivation).
The phase pulse, q(t), is defined as the integral of the frequency pulse, f(t). The frequency pulse is defined to be zero outside of the time interval (0,LT) and is scaled so that
as in equation [2] above. Over the nth symbol interval then, the CPM waveform is given by
Laurent has shown that when a_{n}∈ {±1}
We note that this expression is only meaningful for noninteger values of the pseudomodulation index, h_{n}, (i.e. when sin(πh_{n})≠0).
Now, rewriting the signal
A generalized phase function is now introduced, which is nonzero over the interval (0,2LT) and which it is defined as follows
The time variable, τ=t mod T.
Using this definition, the signal may be expressed as
The product actually yields a total of 2^{L }terms. However, Laurent has shown that many of these terms are similar and that they can be grouped into 2^{L−1 }pulses of various lengths. For example, when L=3, we obtain an expression that has eight terms:
(where the dependence on τ is understood). The pseudosymbol is defined as
and α_{k,i }(i=1, . . . , L−1) is the ith bit in the radix2 representation of k.
α_{k,0}=0 for all k. From expression [18], we observe that the pseudosymbol b_{0,n }modulates the signal over four symbol intervals.
Close inspection of expression [16] reveals that we can define the following four functions
Consequently, for L=3, we may write
This formulation generalizes nicely to all of the other cases, so that any CPM signal, s(t,a,h), can be decomposed into the superposition of PAM waveforms
where the component pulse is defined as
Note that the same definition of v(k,j,t) also appears in Laurent's work. In addition, when N_{h}=1, then w(n,j,t)=0 and the expression for the signal pulses simplifies to the equivalent expression for binary CPM that is found in the Laurent Decomposition of binary CPM waveforms.
As in the Laurent Decomposition of binary singleh CPM, it is also observed that the durations of the component pulses for Mary CPM are defined as follows
In general, the kth component pulse, g_{k,n}(t), is nonzero over the interval 0≦t≦T×min_{i=1,2, . . . , L1}[L(2−α_{k,i})−i].
There are three important differences between the decomposition of equation [26] and those obtained in the work detailed above by Laurent; Mengali and Morelli; and Perrins and Rice.

 The number of PAM components that are required to represent the signal over each symbol interval is equal to Q=2^{L−1}, as in the original Laurent Decomposition.
 We have observed g_{0,n}(t) to be the pulse of longest duration (L+1)T and it also happens to be the pulse that contributes most significantly to the total signal energy. This finding suggests the possibility of approximating the PAM representation of the signal using one pulse of length (L+1)T.
 The shape of the main pulse is dependent on the transmitted data. Hence, this shape can change from one interval to the next.
Based on the proof outlined in this section, we now propose synthesis of the following waveform at the transmitter
where x(t,a,h)=s(t,a,h) when Q_{0}=Q.
Equation [29] is the mathematical embodiment of this invention upon which a transmitter architecture may be developed, and upon which software code to drive the modulation may be based. As software, it may be a program of machine readable instructions tangibly embodied on a computer readable storage medium such as a volatile or nonvolatile memory of a mobile station or other wireless communications device. The instructions are executable by a processor such as a digital signal processor,
The appropriate value of Q_{0 }can be selected according to a particular performance criterion. If, for example, transmitter complexity is the primary concern, then selecting Q_{0}=1 yields a signal that closely approximates CPM with minimum complexity. As a second example, let us suppose that the transmitted waveform should contain T % of the total CPM signal energy. Then, Q_{0 }should be the smallest value such that the approximation error is within (1−T) % of the total signal energy. As a final example, Q_{0 }can also be selected such that the resulting waveform satisfies an upper bound constraint on the PeaktoAveragePower Ratio (PAPR).
A quantitative analysis of the accuracy of the present invention is now presented with reference to the drawing figures.
Contrast
One important aspect of this invention is that it is the first extension of Laurent's Decomposition which allows any CPM waveform—whether binary singleh, Mary singleh or Mary multih—to be described as the sum of Q=2^{L−1 }PAM pulses over each symbol interval. The following observations are noteworthy:

 This invention can be used in order to synthesize “almost Mary CPM” and “exact Mary” CPM waveforms with reduced transmitter complexity visàvis the previous solutions for Mary CPM.
 This invention is general in its application to CPM and therefore represents a unification of results for binary, higher order and multih CPM waveforms.
 Implementation of this invention may reduce transmitter cost visàvis the previously known solutions.
Specific embodiments of the present invention are now detailed. A mobile station MS is a handheld portable device that is capable of wirelessly accessing a communication network, such as a mobile telephony network of base stations that are coupled to a publicly switched telephone network. A cellular telephone, a Blackberry® device, and a personal digital assistant (PDA) with interne or other twoway communication capability are examples of a MS. A portable wireless device includes mobile stations as well as additional handheld devices such as walkie talkies and devices that may access only local networks such as a wireless localized area network (WLAN) or a WIFI network.
Voice or other aural inputs are received at a microphone 30 that may be coupled to the processor 28 through a buffer memory 32. Computer programs such as drivers for the display 22, algorithms to modulate, encode and decode, data arrays such as lookup tables, and computer programs to decompose a CPM signal in accordance with the present invention are stored in a main memory storage media 34 which may be an electronic, optical, or magnetic memory storage media as is known in the art for storing computer readable instructions and programs and data. The main memory 34 is typically partitioned into volatile and nonvolatile portions, and is commonly dispersed among different storage units, some of which may be removable such as a subscriber identity module (SIM). The MS 20 communicates over a network link such as a mobile telephony link via one or more antennas 36 that may be selectively coupled via a T/R switch 38, or a dipole filter, to a transmitter 40 and a receiver 42. The MS 20 may additionally have secondary transmitters and receivers for communicating over additional networks, such as a WLAN, WIFI, Bluetooth®, or to receive digital video broadcasts. Known antenna types include monopole, dipole, planar inverted folded antenna PIFA, and others. The various antennas may be mounted primarily externally (e.g., whip) or completely internally of the MS 20 housing. Audible output from the MS 20 is transduced at a speaker 44. Most of the abovedescribed components, and especially the processor 28, are disposed on a main wiring board, which typically includes a ground plane to which the antenna(s) 36 are electrically coupled.
In traditional CPM receivers, maximum likelihood estimation of the symbol based on Bayesian theory has been a dominant method for estimating an unknown parameter of the received signal in order to separate the symbols, whose transitions are somewhat obscured as compared to other waveforms. This results in a branch metric computation and generally either a Viterbi algorithm feedback or known symbols inserted into the data stream (e.g., in message headers or training sequences). Nondata aided approaches are also known, but generally less reliable. The present invention dispenses with the above computationally complex realizations because the receiver receives a PAM signal that mimics a CPM signal, and decomposes it as a PAM signal whose phase and timing uncertainties are more separable, and better resolved, than CPM. That the present invention does so in a manner that enables a high degree of accuracy using the energy of only one or two pulses (exact decomposition with Q=2^{L−1 }pulses) is a fundamental advantage over prior art PAM decompositions.
The processor synchronizes other components such as the sampler 60 with a local clock 72, and accesses various computer programs, data storage tables, and algorithms that are stored in the memory 34. Certain of these pertain to controlling the modulator to decompose the received signal r(t), which mirrors a CPM signal and which is input into the multiplier 56, as a PAM signal s(t) that is then detected and decoded. In accordance with equation [26], the computer program controlling the modulator causes the processor, for each n^{th }symbol interval, to determine the product of a pseudo symbol b_{k,n }and a function g_{k,n}(tnT), and to sum those products over Q=2^{L }symbol intervals, where the function g_{k,n}(tnT) derives from a generalized phase function and where L is a number of symbol intervals over which a frequency function is defined. If the energy of more than one pulse is to be accumulated, then add the summed products for all k pulses between the 0 and Q−1 pulses.
Similar architecture is present at the transmitter, where some components such as the modulator are duplicated in transmit and receive side of the transceiver for clarity, though in practice one component may operate for both the transmit and receive modes. The desired information signal is converted to analog at a converter 72, and is modulated at the modulator 58b with the PAM decomposed waveform to result in the signal x(t) as in equation [29]. If the transmitter is configurable for different modulation indices and/or different Mary signaling, the computer program at the memory 34 first determines singleh or multih and binary or nonbinary Mary signaling. The operable computer program then determines how many pulses are to be resolved at the receiver. As above, if transmitter complexity is a limiting factor, then set Q_{0}=1 so that there is only one value for the index k. For each n^{th }symbol interval, determine a pseudo symbol b_{k,n }and a function g_{k,n}(tnT), and sum those products over Q_{0 }symbol intervals, where the function g_{k,n}(tnT) derives from a generalized phase function and where L is a number of symbol intervals over which a frequency function is defined. If the energy of more than one pulse is to be accumulated, then add the summed products for all k pulses between the 0 and Q−1 pulses. Modulate the signal as above and transmit. The PAM modulated signal x(t) is then amplified 52b and transmitted by one or more transmit antennas 36b.
The various functions and parameters may be stored in the memory as a lookup table, as algorithms, or a combination of both embodied as hardware, software, or both, and readable/executable by a digital processor.
Embodiments of the present invention preserve certain advantages of the Laurent Decomposition in nonbinary Mary and multih modulations as follows:

 The PAM component pulses in the decomposition are distinctively ordered, so that the first component pulse contributes the most significant energy to the waveform and the last component pulse is negligible—just as in the binary CPM case.
 Over each symbol interval, there is clearly one “main pulse” that may be used to synthesize (or evaluate) the Mary CPM waveform with a high degree of accuracy—just as in the binary CPM case.
 Over each symbol interval, the total number of PAM component pulses is equal to Q=2^{L−1}—just as in the binary CPM case.
 It is possible to generate an “almost Mary CPM” or “exact Mary CPM” waveform using a minimal number of PAM component pulses (when compared to the prior art)—just as in the binary CPM case.
As detailed above, the present invention solves the complexity problem in expressing an Mary CPM waveform as the superposition of PAM waveforms. In addition, this invention also results in the following advantages:

 A reduction in transmitter complexity (and possibly cost) for the synthesis of “almost Mary CPM” waveforms;
 A reduction in the level of difficulty required to evaluate the performance or characteristics of Mary CPM waveforms;
 A reduction in receiver complexity (and possibly cost) for the reception of “almost Mary CPM” or “exact Mary CPM” waveforms.
As a review, the present invention embodies a new decomposition that generalizes Laurent's work on binary CPM to Mary singleh, binary multih and Mary multih CPM waveforms. Importantly, it retains many of the useful properties of the Laurent Decomposition for binary singleh CPM. The significance of this invention is that one can use this new PAM decomposition in order to define an optimized “almost Mary CPM” waveform that constructs the desired signal using the smallest number of component PAM terms in the decomposition for a certain metric of performance (such as an upper bound on the PeaktoAveragePowerRatio) (when compared to the prior art).
Specifically, the present invention poses an alternate, exact and concise formulation of Mary singleh and multih CPM waveforms as the sum of a finite number PAM signals. This formulation facilitates the synthesis of “almost Mary CPM” or “exact Mary CPM” signals with reduced transmitter complexity. The present invention may be used to linearize GMSK waveforms and design simpler receivers, for example by using it to develop simplified techniques for correlating the received signal against a training portion of a burst. The greatest savings over prior art is seen to be in higher order CPM waveforms.
Further, since the present invention embodies a mechanism that simplifies the PAM representation/approximation of higher order CPM waveforms, it overcomes one of the critical shortfalls that has caused CPM to lag behind OFDM for utilization in high capacity wireless communications. Specifically, one of the major drawbacks of OFDM is the occurrence of large envelope fluctuations, which makes linear amplification extremely challenging. The dynamic range of the complex envelope of an OFDM signal can drive a power amplifier to exhibit nonlinear characteristics and lower power efficiency as the signal input power approaches the saturation region. In contrast to OFDM, CPM schemes are efficient in both power and bandwidth, but have not been widely considered as viable due to the nonlinearity of higherorder CPM signals, resulting in high implementation complexity and difficulty of use with certain receiver architectures. The present invention is seen to resolve that complexity/nonlinearity problem for higherorder CPM.
Although described in the context of particular embodiments, it will be apparent to those skilled in the art that a number of modifications and various changes to these teachings may occur. Thus, while the invention has been particularly shown and described with respect to one or more preferred embodiments thereof, it will be understood by those skilled in the art that certain modifications or changes may be made therein without departing from the scope and spirit of the invention as set forth above, or from the scope of the ensuing claims.
Claims
1.40. (canceled)
41. An apparatus comprising:
 a receiver configured to receive an Mary pulse amplitude modulated PAM signal characterized in that, for each symbol interval, a plurality of pulses exhibit descending levels of signal energy, where M>2; and
 a modulator configured demodulate symbols from the pulses.
42. The apparatus of claim 41, wherein each symbol interval comprises Q0=2L−1 PAM component pulses, where L represents a number of symbol intervals.
43. The apparatus of claim 41, wherein the modulator is further configured to resolve pseudo symbols bk,n.
44. The apparatus of claim 41, wherein the receiver comprises at least one antenna, and the modulator comprises a demodulator, a multiplier, a memory, and a processor.
45. An apparatus comprising:
 a receiver configured to receive a binary multih pulse amplitude modulated PAM signal characterized in that, for each symbol interval, a plurality of pulses exhibit descending levels of signal energy, where h>1 is a modulating index; and
 a modulator configured to demodulate symbols from the pulses.
46. The apparatus of claim 45, wherein each symbol interval comprises Q0=2L−1 PAM component pulses, where L represents a number of symbol intervals.
47. The apparatus of claim 45, wherein the modulator is further configured to resolve pseudo symbols bk,n.
48. The apparatus of claim 45, wherein the receiver comprises at least one antenna, and the modulator comprises a demodulator, a multiplier, a memory, and a processor.
49. A memory embodying a program of machinereadable instructions, executable by a digital data processor, to perform actions directed toward demodulating a received signal, the actions comprising:
 determining a generalized phase function of a received pulse amplitude modulated signal;
 deriving a function gk,n(tnT) from the generalized phase function, where T is a symbol duration;
 for each nth symbol interval and each pulse of Q0≦2L−1 pulseamplitude modulated component pulses, resolving pseudo symbols bk,n using the function gk,n(tnT), where L represents a total number of symbol intervals; and
 determining symbols from the pseudo symbols.
50. (canceled)
Type: Application
Filed: Feb 3, 2010
Publication Date: Aug 5, 2010
Applicant:
Inventor: Marilynn P. Green (Pomona, NY)
Application Number: 12/699,307
International Classification: H03D 1/24 (20060101);