Method and system for reduction of peak-to-average power ratio of transmission signals comprising overlapping waveforms

Abstract

The present invention provides a method and system for reducing the peak to average power ratio (PAP) of a signal with low computational complexity. According to one embodiment, the present invention is applied to reduce the PAP of an OFDM signal. According to an alternative embodiment, the present invention, is applied to reduce the PAP of a CDMA signal. Rather than seeking the optimum solution, which involves significant computational complexity, the present invention provides for a number of sub-optimal techniques for reducing the PAP of an OFDM signal but with much lower computational complexity.

Claims

1. A method, comprising: calculating, by a processor, a first peak to average power value of a signal comprising a linear combination of a plurality of vectors; generating, by the processor, a phase vector comprising a sequence of random phase components, each random phase component corresponding to one of the plurality of vectors; calculating, by the processor, a second peak to average power value of a signal comprising a linear combination of the plurality of vectors each multiplied by its corresponding phase component of the phase vector; and if the second peak to average power value is less than the first peak to average power value, storing, by the processor, the phase vector in a memory, and setting the first peak to average power value equal to the second peak to average power value.

2. The method of claim 1, further comprising: repeating the generating, calculating, storing and setting until a predetermined number of phase vectors have been considered.

3. The method of claim 2, wherein the signal is an orthogonal frequency division multiplexed signal.

4. The method of claim 2, wherein the signal is a wireless signal transmitted using code division multiple access.

5. The method of claim 2, wherein the signal is a wireless signal transmitted using multicode code division multiple access.

6. The method of claim 2, wherein the plurality of vectors comprises partial transmit sequences.

7. A transmitter, comprising: a transponder device; a memory; and a processor coupled to the memory and the transponder device, the processor for: calculating a first peak to average power value of a signal comprising a linear combination of a plurality of vectors; generating a phase vector comprising a sequence of random phase components, each random phase component corresponding to one of the plurality of vectors; calculating a second peak to average power value of a signal comprising a linear combination of the plurality of vectors each multiplied by its corresponding phase component of the phase vector; and if the second peak to average power value is less than the first peak to average power value, storing the phase vector in a memory, and setting the first peak to average power value equal to the second peak to average power value.

8. The transmitter of claim 7, further comprising: repeating the generating, calculating, storing and setting until a predetermined number of phase vectors have been considered.

9. The transmitter of claim 8, wherein the signal is an orthogonal frequency division multiplexed signal.

10. The transmitter of claim 8, wherein the signal is a wireless signal transmitted using code division multiple access.

11. The transmitter of claim 8, wherein the signal is a wireless signal transmitted using multicode code division multiple access.

12. The transmitter of claim 8, wherein the plurality of vectors comprises partial transmit sequences.

13. An apparatus, comprising: a processor and a memory, the processor and the memory for performing operations, the operations comprising: calculating a first peak to average power value of a signal comprising a linear combination of a plurality of vectors; generating a phase vector comprising a sequence of random phase components, each random phase component corresponding to one of the plurality of vectors; calculating a second peak to average power value of a signal comprising a linear combination of the plurality of vectors each multiplied by its corresponding phase component of the phase vector; and if the second peak to average power value is less than the first peak to average power value, storing the phase vector in a memory, and setting the first peak to average power value equal to the second peak to average power value.

14. The apparatus of claim 13, further comprising: repeating the generating, calculating, storing and setting until a predetermined number of phase vectors have been considered.

15. The apparatus of claim 14, wherein the signal is an orthogonal frequency division multiplexed signal.

16. The apparatus of claim 14, wherein the signal is a wireless signal transmitted using code division multiple access.

17. The apparatus of claim 14, wherein the signal is a wireless signal transmitted using multicode code division multiple access.

18. The apparatus of claim 14, wherein the plurality of vectors comprises partial transmit sequences.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1, which is prior art, is a block diagram that depicts an implementation of OFDM at a wireless transmitter.

(2) FIG. 2, which is prior art, is a block diagram that depicts an implementation of OFDM at a wireless receiver.

(3) FIG. 3 is a graph showing the complementary cumulative distribution function (CCDF=PR(PAP>PAP.sub.0) of the PAP of a continuous-time, analog, OFDM signal for the particular case of 256 subcarriers.

(4) FIG. 4, which is prior art, is a block diagram depicting the SLM approach for reducing the PAP of an OFDM signal.

(5) FIG. 5, which is prior art, is a block diagram depicting the PTS approach for reducing the PAP of an OFDM signal.

(6) FIG. 6 is a block diagram depicting a wireless network architecture according to one embodiment of the present invention.

(7) FIG. 7 is a flowchart depicting the steps of an iterative algorithm utilizing partial transmit sequences for reducing the PAP of an OFDM signal.

(8) FIG. 8 is a flowchart depicting the steps of an algorithm for reducing the PAP of an OFDM signal using randomly generated sequences of phase factors or structured sequences according to one embodiment of the present invention.

(9) FIG. 9 shows a comparison of CCDF plots utilizing the iterative approach of the present invention and the optimum approach.

(10) FIG. 10 shows a comparison of CCDF plots for the iterative technique demonstrating the effect of varying the number of subblocks.

(11) FIG. 11 shows a comparison of CCDP plots for the iterative technique demonstrating the effect of four-phase weighting factors and 16 QAM signal constellations.

(12) FIG. 12 shows a comparison of CCDP plots for the iterative technique demonstrating the effects of the use of power control.

(13) FIG. 13 shows a comparison of CCDP plots for the iterative technique and the use of random sequences.

(14) FIG. 14 shows a comparison of CCDF plots for the use of iterative technique, random sequences and Walsh sequences.

DETAILED DESCRIPTION

(15) The techniques for performing OFDM transmission are well known. In OFDM transmission, a block of N symbols {X.sub.n, n=0, 1, . . . N1} is formed with each symbol modulating one of a set of N subcarriers {f.sub.n, n=0, 1, . . . N1}. The N subcarriers are chosen to be orthogonal, i.e., f.sub.n=n f, where the subcarrier spacing f=1/NT and where T is the original data symbol period. The original signal after digital-to-analog conversion can be expressed as:

(16) $\begin{array}{cc}x\left(t\right)=\underset{n=0}{\overset{N-1}{.Math.}}{X}_n{}^{2f_{n^1}},0t\mathrm{NT}& \left(1\right)\end{array}$

(17) An important advantage of OFDM is that, in sampled form equation (1) can be implemented using an Inverse Fast Fourier Transform (IFFT).

(18) FIG. 1, which is prior art, is a block diagram that depicts an implementation of OFDM at a wireless transmitter. A block of transmission data (corresponding to a particular symbol interval u) is digitally modulated in modulation block 110 using an appropriate modulation scheme such as quadrature amplitude modulation (QAM). A data vector output from modulation block 110a d.sub.u={d.sub.u,0, . . . , d.sub.u,N-1} is then mapped onto f.sub.N carriers (140a, 140b, etc.) via a serial to parallel converter block 120 to form a modulated sub carrier vector X.sub.u={X.sub.u,0, . . . , Xu.sub.N-1}. The subcarrier vector X.sub.u comprising all carrier amplitudes associated with OFDM symbol interval u is transformed into the time domain, using an N-point IDFT (Inverse Discrete Fourier Transform) (150) or IFFT producing time domain vector x.sub.u. After digital to analog conversion (in D/A converter block 155), the continuous time signal x(t) (157) is transmitted over a wireless channel via RF block 160a.

(19) FIG. 2, which is prior art, is a block diagram that depicts an implementation of OFDM at a wireless receiver. A continuous time signal x(t) (157) is received via RF block 160b. The analog signal is converted to a digital signal via A/D converter 220 producing time domain vector x.sub.u. The time domain vector x.sub.u is transformed to the frequency domain using an N-point DFT (Discrete Fourier Transform) (230) or FFT (Fast Fourier Transform), producing subcarrier vector X.sub.u. After parallel to serial conversion in block 235, the signal is demodulated in QAM block 110b and the transmitted data recovered.

(20) FIG. 3 is a graph showing the complementary cumulative distribution function (CCDF=PR(PAP>PAP.sub.0)) of the PAP of continuous-time, analog, OFDM signal for the particular case of 256 subcarriers. The PAP of a transmitted signal is defined as:

(21) $\begin{array}{cc}\mathrm{PAP}=\frac{\max {.Math.x\left(t\right).Math.}^2}{E\left[{x\left(t\right)}^2\right]}& \left(2\right)\end{array}$
To more accurately approximate the true PAP, the results of FIG. 3 were computed by oversampling (1) by a factor of four (e.g., by zero-padding the data input to the IFFT).

(22) FIG. 4, which is prior art, is a block diagram depicting the SLM approach for reducing the PAP of an OFDM signal. Subcarrier vector X.sub.u.is multiplied by M random sequences r.sub.1-r.sub.M (420). Each multiplied vector is transformed to a time domain vector using an IFFT (430). The PAP of each time domain vector x.sub.u,1-X.sub.u,M. is calculated and the sequence with the lowest PAP is selected for transmission (440).

(23) FIG. 5, which is prior art, is a block diagram depicting the PTS approach for reducing the PAP of an OFDM signal. Subcarrier vector X.sub.u is partitioned in M subblocks (510). Each of the M subblocks is transformed to a partial transmit sequence x.sub.u,1-X.sub.u,M using an IFFT (530). A peak value optimization is then performed on the set of partial transmit sequences by appropriately assigning to each partial transmit sequence an appropriate phase factor b so that the PAP of the combined set of partial transmit sequences, each multiplied by its assigned phase factor 540, is minimized (550). The set of optimized phase factors is obtained by:

(24) $\left\{{\tilde{b}}_u^{\left(1\right)},\mathrm{.Math.},{\tilde{b}}_u^{\left(M\right)}\right\}=\underset{\left\{{\tilde{b}}_u^{\left(1\right)},\mathrm{.Math.},{\tilde{b}}_u^{\left(M\right)}\right\}}{\arg \min }\left(\underset{0kN}{\max }\underset{m=1}{\overset{M}{.Math.}}.Math.b_u^m.Math.x_{u,k}^{\left(m\right)}.Math.\right)$
The partial transmit sequences, each multiplied by its assigned phase factor, are linearly combined 560 and transmitted.

(25) In order to reconstruct the signal at the receiver, the receiver must have knowledge regarding the generation process of the transmitted OFDM signal (i.e., the chosen set of phase factors). The phase factors, therefore, are transmitted as side information resulting in some loss of efficiency. Alternatively, differential encoding can be employed across the subcarriers within a subblock; in this case, the overhead is a single subcarrier per subblock. Using 128 subcarriers with four subblocks and phase factors limited to the set{1,j}, the 1% PAP can be reduced by more than 3 dB.

(26) While the SLM and PTS approaches provide significantly improved PAP statistics for an OFDM transmit signal with little cost in efficiency, a significant issue in implementing these approaches is reducing the computational complexity. In particular, the SLM approach requires the use of M full-length (i.e., N-point) IFFTs at the transmitter. While the PTS approach requires a similar number of N-point IFFTs (one IFFT for each partial transmit sequence), computation complexity in computing these IFFTs is reduced by taking advantage of the fact that a large fraction of the input values are zero (in particular, only N/M values are non-zero). Nevertheless, in the PTS approach, an optimization is required at the transmitter in order to determine the best combination of the partial transmit sequences. In its most direct form, this process requires the PAP to be computed at every step of the optimization algorithm, necessitating numerous trials to achieve the optimum. It is known from C. Tellambura, Phase Optimisation Criterion for Reducing Peak-to-Average Power Ratio in OFDM, Electron. Letts., Vol. 34, No. 2, January 1998, pp. 169-170, that using an alternative performance criterion, less computations are necessary for each trial of the optimization algorithm.

(27) FIG. 6 is a block diagram depicting a wireless network architecture specifically adapted to reduce the PAP of signals transmitted through the network according to one embodiment of the present invention. The network depicted in FIG. 6 may be specifically adapted for the transmission of OFDM signals. In particular, transmitter 105 contains CPU/DSP 110a, which is specifically adapted either through specific hardware design or software components to perform operations upon digital (discrete time) signals to be transmitted through the wireless network. CPU/DSP 110a may also be an ASIC (Application Specific Integrated Circuit Device) specifically adapted to perform OFDM as well as other operations to reduce the PAP of an OFDM signal.

(28) CPU/DSP 110a communicates with memory 120a in order to store data and program instructions. For example, CPU/DSP 110a may communicate with memory 120a to temporarily store intermediate results of DSP operations on signals to be transmitted through the wireless network. Transmitter 105 also contains digital to analog converter 115 for conversion of digital signals for wireless transmission to receiver via transponder 130a and antenna 140a.

(29) Receiver 145 receives wireless signals via antenna 140b and transponder 130b. Analog signals received at receiver 145 are converted to digital format via analog to digital converter 155. Receiver 145 contains CPU/DSP 110b and memory 120b for performing operations on received digital signals. In particular, according to one embodiment, CPU/DSP 110b is specifically adapted to perform demultiplexing of OFDM signals as well as other operations to reconstruct the original signals sent by transmitter 105.

(30) In the PTS approach, a major portion of the computational complexity originates from the need to optimize the phase factors used for combining the subblocks. FIG. 7 is a flowchart depicting the steps of a sub-optimal iterative process for reducing the PAP of an OFDM signal. The procedure is initiated in step 710. In step 720 a set of partial transmit sequences are generated for a particular signal interval u. For example, this may be accomplished by segmenting a subcarrier vector X.sub.u into M subblocks. Then an IFFT is performed on each subblock to produce each partial transmit sequence. In step 730 an initial phase factor b.sub.m from a set of possible phase factors is assigned to each partial transmit sequence. In step 740 the PAP value of a linear combination of the partial transmit sequences, each multiplied by its respective phase factor, is calculated and stored in memory 120a.

(31) In step 745 each partial transmit sequence is analyzed and assigned a final phase factor according to steps 750-765. In particular, in step 750 the current phase factor assigned to the partial transmit sequence under consideration is stored in memory. Then a phase factor from the set of possible phase factors is assigned to the current partial transmit. The PAP value of the linear combination of the partial transmit sequences each multiplied by its respective phase factor is then calculated. In step 755, this calculated PAP value is compared with the PAP value stored in memory. If the calculated PAP value is lower than the PAP value stored in memory (yes branch of step 755), the current PAP value is stored (step 760) and the partial transmit sequence under consideration retains the assigned phase factor. Otherwise, if the current PAP value is greater than the stored PAP value (no branch of step 755), the temporarily stored phase factor from step 750 is re-assigned to the current partial transmit sequence (step 765). In step 770, it is determined whether all phase factors from the set of possible phase factors have been examined for the current partial transmit sequence. If not, (no branch of step 770), step 750 is executed again. If all phase factors have been examined (yes branch of step 770), in step 775 it is determined whether all partial transmit sequences have been examined and assigned a final phase factor. If not (no branch of step 775), step 745 is executed again. If all partial transmit sequences have been examined and assigned a final phase factor, the procedure ends (step 780).

(32) The following pseudo-code defines an embodiment of the present invention:

(33) TABLE-US-00001 Steps: #define SIZE_OF_PARTIAL_TRANSMIT_SEQUENCE 5 #define NUMBER_OF_PHASE_FACTORS 2 int phase_factors[2]= {1,1}; int best_PAP; Struct PTS { int[SIZE_OF_PARTIAL_TRANSMIT_SEQUENCE]; int phase factor; } { assign initial phase factor to each partial transmit sequence; best_PAP=PAP of combine set of partial transmit sequences each multiplied by its corresponding phase factor; for each partial transmit sequence do{ for (i=0;i<=NUMBER_OF_PHASE_FACTORS1; i++) { temp_phase_factor=partial_transmit_sequence.phase_factor; partial_transmit_sequence.phase factor=phase_factors[i]; current_PAP=PAP of combined set of partial transmit sequences each multiplied by its corresponding phase factor; if current_PAP<best_PAP best_PAP=current_PAP; else partial_transmit_sequence.phase_factor=temp_phase_factor; }

(34) According to one embodiment of the present invention, the set of possible phase factors can take on only binary values from the set {1, 1}. Using this example, after dividing the input data block into M subblocks, M N-point PTSs are generated using an IFFT. Each partial transmit sequence is assigned the same phase factor, (i.e., b.sub.m=1 for all m). The PAP of the combined signal is then computed. The first phase factor b.sub.1 is then inverted and the PAP is then recomputed. If the new PAP is lower than in the previous step, b.sub.1 is retained as part of the final phase sequence. Otherwise b.sub.1 is reassigned its previous value. This procedure continues in a sequential fashion until all of the M possibilities for flipping the signs of the phase factors have been explored.

(35) Results of the sub-optimal iterative approach (as discussed below) show a significant improvement in the PAP of an OFDM signal with only a small degradation compared to the optimum. Nevertheless, the iterative approach requires some feedback for implementation. An alternative approach, which avoids feedback, is to approximate the optimum by simply multiplying the desired information sequence by a number of random sequences and choosing the best to transmit.

(36) FIG. 8 is a flowchart depicting the steps of a sub-optimal technique using random sequences of random phase factors for reducing the PAP of an OFDM signal according to one embodiment of the present invention. In step 810 the procedure is initiated. In step 820, for the symbol interval u, a set of partial transmit sequences is generated (i.e., see discussion of step 720 in FIG. 7). In step 830 a PAP value of a linear combination of the partial transmit sequences is calculated and stored in memory 120a. In step 840, a vector of random phase factors r is generated. In step 850, the PAP value of the combined partial transmit sequences each multiplied by a corresponding phase factor in r is calculated. If the PAP value using the current vector r is lower than the stored PAP value (yes branch of step 855), the vector r and the current PAP value are stored (step 860). Otherwise (no branch of step 855), it is determined whether a sufficient number of random sequences have been considered (i.e., the number of random sequences is user determined). If a pre-determined number of sequences have not been considered (no branch of step 865), step 840 is executed again and another random vector is generated. Otherwise, the procedure ends in step 870 and the currently stored vector r is used for transmission.

(37) According to simulation results (discussed in more detail below), it was found that 16 random trials produced statistically the same results as the iterative approach described above. Based upon this observation, according to an alternative embodiment of the present invention, a known set of sequences, which are easily generated, were used instead of random sequences. According to one embodiment, for example, Walsh sequences were used. Walsh functions reduce the number of required additions by a large factor if partial sums are stored. This is similar to the way a FFT reduces the computations required for a DFT. Using structured sequences such as Walsh sequences resulted in degradation of only 0.3 dB. Similar results can be obtained with other well-known sequences such as the Shapiro-Rudin sequences.

(38) Simulation Parameters

(39) The PAP is associated with the continuous-time OFDM transmit signal. Many experimental results compute the PAP based on T or symbol-sampled data in which case overly-optimistic results are produced due to missing peaks in the signal. Simulations, with regard to the present invention, were conducted in which the transmitted symbol was oversampled by a factor of four. Simulations showed that this oversampling was sufficient to capture signal peaks. In the results, described below, 100000 random OFDM blocks were generated to obtain CCDF plots. 256 subcarriers were used as were QPSK data symbols.

(40) Simulation Results

(41) FIG. 9 shows a comparison of CCDF plots utilizing the iterative approach of the present invention and the optimum approach for the case of a single OFDM block and 16 subblocks each composed of 16 subcarriers. The unmodified OFDM signal exhibits a PAP which exceeds 10.4 dB for less than 1% of the blocks. By using the PTS approach with the optimum binary phase sequence for combining, the 1% PAP can be reduced to 6.8 dB. In addition, the slope is improved so that the reduction in the PAP would be even more significant at lower values of the CCDF. Using the iterative technique, a value of 7.8 dB was obtained. While this represents a degradation of 1 dB, the optimization process is reduced to 16 set of 16 additions, a considerable savings over attempting to find the optimum set of phase factors (the PAP is actually computed 17 times, including the case where b.sub.m=1 for all m).

(42) FIG. 10 shows CCDF plots for the iterative technique demonstrating the effect of varying the number of subblocks. FIG. 10 shows CCDF plots using the iterative technique for 4, 8 and 16 subblocks, each increase in the number of subblocks impacting the complexity of the implementation. As expected, the improvement decreases as the number of subblocks decreases. However, FIG. 10 shows that with only 8 subbocks, and therefore, only 8 additional steps in the optimization, a reduction of more than 2 dB in the 1% PAP was achieved.

(43) FIG. 11 shows CCDF plots for the iterative technique demonstrating the effects of allowing the phase factors to be chosen from the set {1, j} instead of the binary set {1, 1}. As shown in FIG. 11, for 16 subblocks the effect is small. With only 16 steps, the additional degrees of freedom were not enough to offset the increased possibility of encountering a poor sequence. For a smaller number of subblocks, the added degree of freedom in choosing the phase factors provided only an addition 0.4 dB reduction. Also, the results showed that using a 16-QAM constellation to modulate each subcarrier resulted in negligible differences from the QPSK (Quadrature Phase Shift Keying) case.

(44) OFDM has been proposed for use in many applications, including multiple-user or multiple-access modes. For such an application, the base station transmits a block of N subchannels in which only a small subset of the subchannels are intended for an individual mobile or portable receiver. In this case, the signals to be transmitted (e.g., 16 subcarriers for each of the 16 users) are combined and transmitted over one antenna at the base station. However, each of the user-clusters is transmitted with a different power level, usually depending on the distance from the base to the individual portable receiver. Given the use of power control, it is necessary to assess its effect on the PAP when the iterative technique is used.

(45) FIG. 12 shows CCDF plots for the iterative technique demonstrating the effect of no power control, power control with the transmit levels chosen uniformly in the interval [10 dB, 0 dB] and power control in which the transmit levels are chosen from a distribution with a wider spread [20 dB, 0 dB]. As shown from the plots of FIG. 12, the additional variations in transmitter power in each subblock resulted in very little degradation for the 10-dB spread case and about 0.5 dB for the 20 dB case. These results depict a worst-case scenario. In more realistic scenarios the distribution would not be uniform and the degradation would be minimal.

(46) FIG. 13 shows a comparison of CCDF plots for the iterative technique and the use of random sequences. FIG. 13 shows results using 5, 16,200 and 2000 random phase sequences (each of length M). Note that 2000 tries results in performance, which is essentially equal to the global optimum. However, even using as few as 5 random tries, it is possible to obtain results within 1.5 dB of the optimum.

(47) FIG. 14 shows a comparison of CCDF plots for the use of iterative technique, random sequences and Walsh sequences. FIG. 14 shows results when the 16 Walsh sequences of length 16 are used. These sequences result in a degradation of only 0.03 dB.