Background technology
In mobile communication system, for transmitting wireless signals without distortion, the linearity to transmitting terminal requires very high, PAPR (Peak to Average Power Ratio when the signal of system emission, when peak-to-average force ratio) too big, at first excessive power can cause very big waste, and secondly too big peak-to-average force ratio makes the performance rapid deterioration of system, too big peak-to-average force ratio will directly influence the operating cost and the efficient of whole system, therefore need to seek the method that reduces the peak-to-average power ratio.In addition, when signal is made up of a plurality of subcarriers, if the too high meeting of the peak-to-average force ratio of subcarrier causes the nonlinear distortion of signal, destroy the orthogonality between the subcarrier, increase the outer interference of band, make system performance degradation, therefore how eliminating the peak signal power point rationally and effectively becomes key issue.
OFDM (Orthogonal Frequency Division Multiple, OFDM) multicarrier system has adopted the orthogonal frequency channel, so can under the situation that does not need complicated balancing technique, support the high speed wireless data transmission, and has a very strong anti-fading and anti-ISI (Inter-SymbolInterference, intersymbol interference) ability, but the topmost shortcoming of ofdm system is to have bigger PAPR, and it directly affects the operating cost and the efficient of whole system.The peak-to-average force ratio problem is ubiquitous problem among the MCM (Multi-carrier Modulation, multi-carrier modulation).
It is 1/2 overlapping that the ofdm system subcarrier has, but keep mutually orthogonal.Isolate information at receiving terminal by the correlation demodulation technology.The sine and the cosine of the sine of any one subcarrier and cosine and another subcarrier all are quadratures.Promptly when n ≠ m
This point can guarantee that each subcarrier frequency of oscillation sinusoidal and cosine all is the integral multiple of fundamental frequency.As shown in Figure 1, be the additive process of observing 4 carrier waves from time domain.The subcarrier stack, each is the carrier wave stack of 4 different frequencies constantly.The passing time of a symbol has become NT, definitely is over-sampling for low-limit frequency.But the signal for highest frequency is to satisfy the Nyquist sampling thheorem.
Along with the continuous increase of number of carriers, when for example subcarrier was increased to 1200, the PAR of signal sharply increased.
OFDM-TDD frame structure relative complex, a radio frames is 10ms, is made up of 2 fields, and each field length is 5mS, and field is divided into 5 subframes (subframes), and 4 conventional subframes and a special subframe are arranged.Each conventional subframe is divided into 2 time slots, so conventional slot length 0.5mS, special subframe is made of DwPTS, GP, these 3 special time slots of UpPTS.DwPTS and UpPTS length are configurable, are 1mS but require DwPTS, GP, UpPTS total length.Each time slot has 15360 sampled points.A time slot of down link has
Individual symbol, the CP length of first symbol is 160, and remaining is 144, and the length of each symbol is 2048, thus a time slots (2048x7+144x6+160) * Ts=15360*Ts=5ms, as shown in Figure 2.
A subframe has 2 time slots, and each time slot has N
Symb DL(7) individual symbol, so a subframe has 14 symbols, each symbol has N
RB DLIndividual RB.Another time slot of subcarrier under symbol down by (k, l) unique identification is RE (Resource Element, a Resource Unit), k=0 wherein ..., N
RB DLN
Sc RB-1 l=0 ..., N
Symb DL-1 is designated as time domain and frequency domain respectively.Can be write as a if consider many antennas again
K, l (p), wherein which antenna indicates be to p.As shown in Figure 3, time slot finishes because channel estimating is based on, so the distribution of reference signal designs according to time slot, transverse axis is exactly a symbolic number, and it is N that a subframe has 14 symbols
Symb DL* 2.The longitudinal axis is the subcarrier distribution situation of a RB, and each RB has N
Sc RB(12) individual subcarrier.So each module in figure below has 12 * 14 RE.And the reference symbol pattern of each RB all is the same on the time slot.
Come as can be seen from the reference signal of downlink map, the distribution difference of distinct symbols reference signal, the reference signal of different antennae distributes different.Owing on the not all symbol reference signal is arranged all, and the RB of traffic assignments meeting Random assignment, carry out IFFT (Inverse FastFourier Transform respectively based on symbol like this, inverse fast Fourier transform) after the processing computing, the peak point that some symbol surpasses thresholding seldom, some symbol surpasses the peak point of thresholding can be a lot.Cause the probability statistics vibration meeting of signal amplitude bigger like this.Each peak point all has a large amount of sampled points that surpass thresholding in the few symbol of peak point.Each peak point can be fewer above the sampled point of thresholding in the peak point symbol how.
PC-CFR (Peak Cancellation Crest Factor Reduction, peak value offset crest factor to be reduced) algorithm is by seeking peak power point, peak power point being carried out pulse to cutting.As shown in Figure 4, be to find peak power point, then peak power point is carried out peak value and eliminate schematic diagram, wherein:
CFR is the digital despicking module, mainly finishes the signal that inside is higher than threshold value and carries out peak clipping and handle, and reduces the average power of signal, further hot-tempered sound is suppressed.The CFR inside modules mainly is divided into six submodules, and its function introduction is as follows:
The DELAY module: owing to produce the link of peak clipping sequence a fixing time-delay is arranged, this module is just finished the constant time lag of input data, keeps identical with the needed time of peak clipping sequence that produces, the last subtraction block of input synchronously.
The PEAK_DETECT module: the detection module of peak point, all are higher than the part of threshold value and all carry out the peak value differentiation.
PEAK_SCALING module: α=(| x|-γ) exp (j θ) | X| is detected peak value sequence; γ is a threshold value, θ=angle (x).
The PEAK_ALLOCATOR module: the peak allocation module, with a detected peak clipping pulse generator of each peak allocation (CPG) module.
CPG module: peak clipping pulse generation module, mainly storing the peak clipping sequence of regular length, each CPG module once can only be handled a peak point, in case when being in idle condition, the PEAK_ALLOCATOR module will distribute a peak value to give at leisure CPG resume module, if all CPG modules all do not have the free time, handle for the peak that can not disappear of detected peak point next so.A CPG needs both 4 multipliers of a complex multiplier, and clock rate is if 4 quilts of signal processing speed, these 4 multiplier time-sharing multiplexs so.So a complex multiplier just can be finished the operation of 4 CPG.
SUM module:, form peak clipping sequence with all additions of output of all CPG modules.
The SUBTRACT module: through after the DELAY module, deduct the peak clipping sequence that the SUM module forms with primary signal, structure be exactly the final output result of CFR module.
Above-mentioned technology must find each time the peak power point that surpasses thresholding, by to the eliminating an of peak power point, makes signal power all drop under the peak clipping thresholding.But a kind of defective of this algorithm maximum be when the OFDM subcarrier allocation is fewer-this situation appears on the symbol that does not have reference signal the professional simultaneously RB that uses also seldom, this time just occurred existing in continuous a period of time many sampled points that surpass thresholding, surpass the sampled point quantity of thresholding this time in continuous a period of time numerous.Because PC-CFR is fit to eliminate sharp-pointed peak point, will lose efficacy for continuous numerous high-power signal PC-CFR, as shown in Figure 5.Therefore directly using PC-CFR is not well suited for for the LTE-TDD system.
The another one algorithm is called NS-CFR (Noise Shaping Crest Factor Reduction, noise moulding peak clipping) algorithm, and algorithm principle is as follows:
The The general frame that FPGA realizes as shown in Figure 6, wherein:
Amplitude and angle calculation module: calculate (| X|-A) e
J θ, wherein X is an input signal amplitude, and A is the peak clipping threshold value, and θ is the angle of input signal.Because the sample rate of input signal is for generally being lower than the FPGA work clock, so I, the Q two-way can multiplier of time division multiplexing.
Peak value is searched module: detect the signal that whether has above the peak value window.
Filter carries out the noise moulding: the signal in the peak value window is carried out the filtering moulding, and this part I, filter of Q two-way time division multiplexing because filter generally can be designed as balanced-filter, in order to save multiplier, adopts the symmetrical expression systolic structures.
The data time delay module: the data to input are delayed time, to guarantee that input signal and peak clipping signal keep synchronously.
The limit filtration algorithm that circulates in addition in addition, circulation limit filtration algorithm is divided into four parts: amplitude limit, FFT (Fast Fourier Transform, fast Fourier transform) conversion, remove frequency domain external noise, IFFT.
The problem of NS-CFR is that the exponent number of this filter generally all needs 120 rank because a noise formed filter is arranged, and so just needs 120 multipliers, so many multiplier FPGA to be difficult to realize.The problem of peak value regeneration can appear in NS-CFR in addition.The amplitude comparison diagram as shown in Figure 7 before and after the NS-CFR peak clipping.
In above-mentioned prior art, above-mentioned algorithm all has defective separately, is difficult to good realization, and for example: the PC-CFR algorithm is difficult to effective peak clipping for the situation of the discontinuous distribution of subcarrier; The resource that the NS-CFR algorithm expends is many, and peak value recurrence easily; Circulation limit filtration algorithm resource is big, and time-delay is long, is difficult to FPGA and realizes.
So, be necessary to propose a kind of scheme that eliminates of effective signal wave crest, can eliminate the peak power under any sub-carrier configuration, processing delay is less simultaneously, and the employed resource of scheme is moderate, the also compliance with system requirement of system performance index such as ACPR, EVM.
Embodiment
Describe embodiments of the invention below in detail, the example of described embodiment is shown in the drawings, and wherein identical from start to finish or similar label is represented identical or similar elements or the element with identical or similar functions.Below by the embodiment that is described with reference to the drawings is exemplary, only is used to explain the present invention, and can not be interpreted as limitation of the present invention.
The present invention proposes a kind of method of signal crest trimming, may further comprise the steps:
Physical layer signal is carried out after the IFFT conversion, carry out secondary lobe compression peak clipping RC1 operation; Described signal through secondary lobe compression peak clipping RC1 operation is carried out digital up converter DUC intermediate frequency filtering; Carry out secondary lobe compression peak clipping RC2 operation to described through the signal behind the DUC intermediate frequency filtering.
As shown in Figure 8, the method flow diagram for embodiment of the invention realization signal crest trimming may further comprise the steps:
S101: physical layer signal is carried out after the IFFT conversion, carry out secondary lobe compression peak clipping RC1 operation.
In step S101, shown in Fig. 9 the first half, be the algorithm The general frame of embodiment of the invention RC-CFR (ReducePeak and PC-CFR reduces peak value and peak value and offsets the crest factor reduction) two steps realization.Particularly, the core concept of RC algorithm is to adopt multistage secondary lobe compression peak clipping algorithm, and the design philosophy of secondary lobe compression peak clipping algorithm is as follows:
Determine the length N of core peak clipping vector according to system bandwidth, while is according to the interpolation multiple L of intermediate frequency, the sampling number of physical layer FFT and a PRB (Physical Resource Block, Physical Resource Block) Dui Ying subcarrier number is determined PRB corresponding sampling points number in the core peak clipping vector.And then determine the frequency domain response of the core peak clipping vector of this time slot according to the PRB index that physical layer sends over, carry out IFFT according to the frequency domain response that obtains and obtain core peak clipping vector, for realizability, the consideration physical layer is not transmitted the PRB index, uses a unified core peak clipping vector to carry out the peak clipping Filtering Processing.The time domain expression formula that just obtain core peak clipping vector this moment is as follows, and the frequency domain of this core peak clipping vector is that the subcarrier full configuration is put:
p=IFFT(I
k)
The complexity that taking into account system calculates, a core peak clipping vector is fixed in the not self adaptation adjustment along with the configuration of subcarrier of core peak clipping vector, and different sub-carrier configuration is all used this same core peak clipping vector.The Sidelobe Suppression algorithm is that main lobe remains unchanged, contiguous secondary lobe compression, and main lobe is at the center of peak clipping vector generally speaking, and the secondary lobe squeeze operation on main lobe both sides is:
left_lob=P(1:left_point)*compress_gain;
right_lob=P(right_point:end)*compress_gain;
The value of main lobe is as follows:
main_lob=P(left_point:right_point)
Wherein 0≤compress_gain≤1 is a compressibility factor, the positive number less than 1.This compression algorithm complexity is still than higher, and the design difficulty of core peak clipping vector is big, has certain defective of cutting of leaking.
In order to obtain a practicable peak clipping algorithm, it is as follows that secondary lobe compression algorithm itself can take to develop algorithm:
Main lobe peak clipping algorithm: despiker only stays the peak signal in the main lobe, promptly compresses all secondary lobes, compress_gain=0, and the peak clipping filter just was equivalent to homophase to the peak clipping of signal and cut operation firmly this moment.This moment, despiker had only a main lobe, so as long as detect peak signal above thresholding, can eliminate one by one.Neither can exist during advantage to leak and cut, and can not influence adjacent signals when eliminating this peak value, so that the EVM that peak clipping this moment causes worsens is minimum, the peak clipping effect is the most obvious.This point is similar to the beam shaping of smart antenna, and the signal behind many antennas excipient is all received by this user, the power maximum that the user receives, and to other user's interference minimum.
Full lobe peak clipping algorithm: despiker has kept all core peak clipping vectors, secondary lobe is not carried out any inhibition, and compress_gain=1 is not so the ACPR of system has deterioration after the peak clipping.
Adopt the advantage of this two-stage peak clipping to be, the peak value of signal can eliminate to greatest extent, and the deterioration of EVM and ACPR is all relative little.
As shown in drawings, wherein, Figure 10 is the full lobe intention of a signal schematic diagram, does not as seen compress side-lobe signal and overlaps substantially; Figure 11 is signal main lobe compression schematic diagram, has only kept the signal of a point after the visible secondary lobe compression.
Shown in Fig. 9 the latter half, secondary lobe compression peak clipping RC1 operation comprises main lobe peak clipping operation, and the main lobe peak clipping is operating as homophase and cuts the peak operation firmly, the peak signal in the main lobe peak clipping operation stick signal main lobe.
Particularly, the main lobe peak clipping algorithm after the IFFT is that the physical layer homophase cuts the peak operation firmly, and carrying out DUC, to carry out homophase peak clipping algorithm before handling as follows:
The groundwork of above-mentioned algorithm is the calculating of the amplitude and the phase place of signal, and preferably, the correction of above-mentioned algorithm is as follows:
Wherein, x (n) carries out IFFT conversion signal afterwards for physical layer signal, and A is predetermined peak clipping threshold value.
Finish normal filtering at intermediate frequency subsequently, the normal DUC after the peak clipping handles, effectively inhibition zone peak clipping outward cause spuious.So no longer need extra IFFT/FFT computing this moment, makes operand reduce greatly, it is very short that time delay also becomes.
A sinusoidal computing is removed in the computing that needs that eliminates of a peak value, outside the cos operation, only needs two multiplyings just can finish.The peak clipping of homophase amplitude limit is a non-linear process, and it will cause certain in-band noise and the outer interference of band.The outer quilt intermediate frequency filtering subsequently that disturbs of band eliminates.
Because carry out the peak clipping operation in physical layer, physical layer rate is very low, handle so can carry out peak clipping by time division multiplexing mode, as shown in figure 12, for schematic diagram is handled in multi-antenna multiplexed peak clipping.The data of many antennas merge before peak clipping, in the present embodiment 4 antennas, as long as the I/Q two-way of each root antenna is strict alignment on time domain, the signal of many antennas just can time-multiplexed mode be imported physical layer and is cut module firmly, the resource of many antenna peak clipping uses is the same with the resource that an antenna peak clipping consumes like this, the use of having saved hardware resource.
S102: the signal through secondary lobe compression peak clipping RC1 operation is carried out digital up converter DUC intermediate frequency filtering.
Though can effectively eliminate peak value though the secondary lobe compression reduces the peak value algorithm, frequency spectrum has certain deterioration.So intermediate-frequency filter needs precipitous transition band, eliminate the influence of sideband distortion.Preferably, intermediate-frequency filter is that the multiple filter combination realizes.As shown in figure 13, be the intermediate-frequency filter spectrogram, the influence that precipitous transition band can well cooperate the physical layer peak clipping to cause.
The peak clipping of homophase amplitude limit is a non-linear process, and it will cause certain in-band noise and the outer interference of band.Band is outer to disturb quilt filtering subsequently to eliminate, and as shown in figure 14, is the spectrogram of handling through DUC after the peak clipping of PRB random arrangement, does not then see the outer interference of band after filtering is finished.
Physical layer is finished after the peak clipping, after handling by DUC filtering interpolation, arrives intermediate frequency.Intermediate-freuqncy signal is carried out the peak clipping operation.The eliminating operation and can use intermediate frequency secondary lobe compression peak clipping algorithm, for example PC-CFR algorithm of intermediate frequency.
Reason by the regeneration of DUC peak value is: physical layer signal is through after the interpolation filtering, because the digital filter exponent number that uses on the hardware is limited now, it not ideal filter, so deposit certain passband ripple shake, stopband attenuation is also limited simultaneously, also have the data volume of physical layer to be multiplied, so physical layer signal through after the intermediate frequency, has the peak value recurrence.If in like manner after PC-CFR, also have interpolation filter can occur the problem of peak value regeneration equally, because the signal after the peak clipping is by DAC (Digit to Analog Converter, number is touched converter), DAC inside is in order to improve signal at the SNR that counts when touching conversion, can carry out interpolation filtering once more to the peak clipping signal of input, to the input signal over-sampling.Sample with the speed that is higher than the signal bandwidth twice and to be called over-sampling.Over-sampling is a very important function, because it can bring gain in numeric field the signal to noise ratio of the signal that receives.Touch in the process of conversion at number, sampling rate is fast more, and quantization noise level will be low more.Because distribution of quantization noise is in wideer frequency band, promptly in Cai Yang the clock frequency, and total noise is certain, and the over-sampling schematic diagram as shown in figure 15.
The over-sampling noise power spectral density:
Wherein, Df
sBe the speed of sampled signal, so as long as improve Df
s, the signal to noise ratio of sampling will be raised.
Specific to A/D, the signal to noise ratio of D/A converter, can calculate by following formula:
SNR=6.02·N+1.76+10log
10(F
S/2f
BW),
Wherein, SNR is a signal to noise ratio, and N is the figure place of ADC (Analog to Digit Converter touches A/D converter)/DAC, and Fs is the frequency of ADC/DAC sampling clock, f
BWBe signal bandwidth.From formula as can be seen, if sampling rate doubles, the SNR of ADC/DAC will improve 3dB.
So the signal that the later signal of peak clipping enters after the DAC generally can carry out interpolation filtering once more, generally be to use multistage HB (Half Band, half band) filter to finish.For the verification msg process intermediate frequency peak clipping interpolation variation of PAR later on more later on, verify by following emulation platform, as shown in figure 16, be the emulation link schematic diagram of checking RC-CFR peak value recurrence.
Handle A point behind the observation CFR, and the peak-to-average force ratio that B, C, D are ordered behind the process HB filter after the multi-carrier signal process PC-CFR peak clipping of employing different IF speed fs again through three Cascade H B filters (HB1, HB2, HB3).
DAC is to the influence (unit: dB) of peak-to-average force ratio
In the present invention, the signal rate f behind the intermediate frequency filtering
sMeet f
s〉=f
S0, wherein, speed f
sFor at f
S0The PAR lifting of carrying out signal behind the interpolation filtering under the speed again is smaller or equal to predetermined threshold LdBc, makes PAR (f
s)-PAR (f
S0)≤LdBc.
As shown in figure 17, be the schematic diagram of different digital intermediate frequency rate signal through peak-to-average force ratio lifting behind the DAC.Can obtain from the The above results analysis:
As can be seen from Figure 17, digital medium-frequency signal through three HB after PAR certain lifting is all arranged really, but data rate is high more, the PAR lifting is more little, stipulates a minimum speed limit f
S0, the PAR lifting of signal is less than 0.3dBc under this speed: Δ PAR (f
S0)≤0.3dBc, the signal rate f behind the PC-CFR
s〉=f
S0Get final product, i.e. PAR (f
s)-PAR (f
S0)≤0.3dBc.Come as can be seen from this emulation, as long as the signal rate of PC-CFR output is greater than f
s〉=92.16MHZ gets final product, this moment enter DAC again after, the peak-to-average force ratio lifting under 0.2dBc, is satisfied system requirements again.Promptly in the above-described embodiments, the LdBc value is 0.3dBc.
As shown in figure 18, be the schematic diagram of peak value recurrence behind the physical layer interpolation filtering, full lobe peak clipping restriction back recurrence this moment peak value.
S103: to carry out secondary lobe compression peak clipping RC2 operation through the signal behind the DUC intermediate frequency filtering.
Shown in Fig. 9 the latter half, secondary lobe compression peak clipping RC2 operates and comprises full lobe peak clipping operation, and full lobe peak clipping operation is not compressed secondary lobe, and full lobe peak clipping is operating as peak value and offsets crest factor reduction PC-CFR peak clipping operation.
By DUC the peak value recurrence is arranged later on, this moment, intermediate frequency adopted the secondary lobe compression algorithm to eliminate remaining peak power.Only to the secondary lobe compression of far-end, the secondary lobe of near-end remains unchanged the secondary lobe compression algorithm, can guarantee that so whole ACPR does not have too big deterioration.If consider that intermediate frequency ACPR is big inadequately, do not compress secondary lobe this moment, and what adopted this moment is exactly that full lobe peak clipping algorithm eliminates the residue peak value, because most of peak value all eliminates before entering CFR, the full lobe despiker of this moment can simple designs.
With respect to single PC-CFR, in the full lobe peak clipping of physical layer main lobe peak clipping+intermediate frequency following benefit is arranged: no matter how the physical layer subcarrier disposes, and peak value can effectivelyly eliminate, though peak value recurrence, the peak value general PAR that recurs can not surpass 7.0dBc.
Wherein important setting of PC-CFR is the peak clipping prototype filter, the frequency spectrum of ofdm signal coupling behind this prototype filter and the DUC,
If prototype filter is flter_f (n), n=0 ..., N-1, N are peak clipping coefficient length, it can generate storage in advance.By the firls function of Matlab, and use the Kaiser windowing to obtain.Its generation method is shown below, f in the formula
1Be pass band width, f
2Be stopband attenuation, f
sBe sample frequency, beta:
cfr_ntaps=255;
fs_mhz=92.16;
beta=5;
f1=18.0/fs_mhz;
f2=19.2/fs_mhz;
filter_f=firls(cfr_ntaps-1,[0 f1 f2 1],[1 1 0 0],[1 10])...
.*kaiser(cfr_ntaps,beta)′;
Wherein parameter b eta has determined the decay of secondary lobe, and is as follows:
In order to make the frequency spectrum after the peak clipping not worsen, the bandwidth chahnel of prototype filter general slightly less than or the signal spectrum bandwidth of coming out near DUC, as shown in figure 19, the prototype filter spectrogram of different peak clipping bandwidth.As can be seen from the figure, select pass band width 16MHZ proper, f1=16.5/fs_MHz, f2=18.5/fs_MHz.
As shown in figure 20, be filtering schematic diagram after the 100RB configuration peak clipping, as shown in figure 21, be amplitude comparison diagram before and after the 100PRB peak clipping.
As shown in figure 22, be amplitude comparison diagram before and after the 40PRB peak clipping, as shown in figure 23, be amplitude comparison diagram before and after the 6PRB peak clipping.
Purpose and difference that RC-CFR peak clipping algorithm carries out the secondary lobe compression before and after the DUC be, the main lobe peak clipping before the DUC is in order to eliminate most peak power points, and the full lobe peak clipping after the DUC filtering mainly is to eliminate the recurrence peak value that DUC filtering causes.
The method of above-mentioned signal crest trimming disclosed by the invention can eliminate the peak power under any sub-carrier configuration; Processing delay can guarantee that processing delay is little in 20us simultaneously.In addition, the method for the above-mentioned signal crest trimming that the present invention proposes consumes the less hardware device resource, and especially the hardware multiplier resource is less than 30, and the degree that system performance index such as ACPR, EVM worsen will be lower than existing algorithm.The method of the above-mentioned signal crest trimming that the present invention proposes, very little to the change of existing system, can not influence the compatibility of system, and realize simple, efficient.
Corresponding to said method, as shown in figure 24, the structural representation for the device 100 of embodiment of the invention signal crest trimming comprises RC1 peak-clipping module 110, intermediate frequency filtering module 120 and RC2 peak-clipping module 130.
Wherein, RC1 peak-clipping module 110 is used for physical layer signal is carried out after the IFFT conversion, carries out secondary lobe compression peak clipping RC1 operation.
The secondary lobe compression peak clipping RC1 operation that RC1 peak-clipping module 110 carries out comprises main lobe peak clipping operation, and the main lobe peak clipping is operating as homophase and cuts the peak operation firmly, the peak signal in the main lobe peak clipping operation stick signal main lobe.
Particularly, the homophase that RC1 peak-clipping module 110 carries out cuts the peak operation firmly and comprises, 110 pairs of input signals of RC1 peak-clipping module cut the output signal that obtains after the peak operation firmly through homophase and are:
Wherein, x (n) carries out IFFT conversion signal afterwards for physical layer signal, and A is predetermined peak clipping threshold value.
Particularly, the data of the many antennas of x (n) signal before peak clipping merge, and the signal of many antennas is imported physical layer in time-multiplexed mode and cut module firmly, finish homophase and cut the peak operation firmly.
Intermediate frequency filtering module 120 is used for the signal through secondary lobe compression peak clipping RC1 operation is carried out digital up converter DUC intermediate frequency filtering.
Particularly, the DUC intermediate frequency filtering in the intermediate frequency filtering module 120 is that the multiple filter combination realizes.
Particularly, the signal rate f in the intermediate frequency filtering module 120
sMeet f
s〉=f
S0, wherein, speed f
sFor at f
S0The PAR lifting of carrying out signal behind the interpolation filtering under the speed again is smaller or equal to predetermined threshold LdBc, makes PAR (f
s)-PAR (f
S0)≤LdBc.
RC2 peak-clipping module 130 is used for carry out secondary lobe compression peak clipping RC2 operation through the signal behind the DUC intermediate frequency filtering.
Particularly, the secondary lobe compression peak clipping RC2 that RC2 peak-clipping module 130 carries out operates and comprises full lobe peak clipping operation, and full lobe peak clipping operation is not compressed secondary lobe, and full lobe peak clipping is operating as peak value and offsets crest factor reduction PC-CFR peak clipping operation.
Particularly, the PC-CFR peak clipping that RC2 peak-clipping module 130 carries out is operating as by seeking the peak power point of signal, and peak power point is carried out pulse to cutting.
Particularly, RC2 peak-clipping module 130 in carrying out PC-CFR peak clipping operation, the signal band of the peak clipping prototype coefficient pass band width of PC-CFR after smaller or equal to the DUC intermediate frequency filtering.
The device of above-mentioned signal crest trimming disclosed by the invention can eliminate the peak power under any sub-carrier configuration; Processing delay can guarantee that processing delay is little in 20us simultaneously.In addition, the device of the above-mentioned signal crest trimming that the present invention proposes consumes the less hardware device resource, and especially the hardware multiplier resource is less than 30, and the degree that system performance index such as ACPR, EVM worsen will be lower than existing algorithm.The device of the above-mentioned signal crest trimming that the present invention proposes, very little to the change of existing system, can not influence the compatibility of system, and realize simple, efficient.
For the technique effect of such scheme disclosed by the invention is described, scheme and existing other scheme that the present invention is proposed carried out test simulation below, and comparative illustration is as follows:
(1) different peak clipping algorithm PAR contrasts, peak value eliminates about 6.0:
The peak clipping algorithm sub-carrier configuration |
100(PRB) dB |
60(PRB) dB |
40(PRB) dB |
12(PRB) dB |
RC-CFR |
6.3391 |
6.0137 |
6.0179 |
6.4152 |
PC-CFR |
6.4234 |
6.1894 |
6.5752 |
7.1542 |
TNS-CFR |
6.7167 |
6.6647 |
6.6847 |
6.7937 |
NS-CFR |
6.5266 |
6.5337 |
6.5141 |
6.7198 |
Ideal-CFR |
6.1136 |
6.0934 |
6.1000 |
6.2389 |
Primary signal |
9.9344 |
9.9697 |
10.5617 |
10.7982 |
(2) the relevant EVM contrast of different peak clipping algorithms, peak value eliminates about 6.0:
The peak clipping algorithm sub-carrier configuration |
100(PRB) % |
60(PRB) % |
40(PRB) % |
12(PRB) % |
RC-CFR |
6.9779 |
6.0637 |
6.6353 |
11.3224 |
PC-CFR |
6.9431 |
6.1901 |
6.8754 |
11.1831 |
TNS-CFR |
7.4430 |
5.2905 |
5.9825 |
12.5794 |
NS-CFR |
13.5287 |
11.6844 |
12.7764 |
21.5308 |
Ideal-CFR |
5.2624 |
4.5799 |
5.2132 |
8.2820 |
Primary signal |
1.4447 |
1.2540 |
1.6309 |
1.7419 |
Therefore, from the link simulation performance, the RC-CFR effect is best, and what PAR can be stable is suppressed to below the 6.5dBc, and EVM can be controlled at below 7%, as shown in figure 25, for RC-CFR to the peak clipping effect performance curve synoptic diagram under the various RB configuration.
Find out from top simulation result: the peak clipping effect of RC-CFR is that EVM worsens minimum, and PAR peak clipping effect is the most obvious.
RC-CFR not only performance is good, and resource is few, and time delay is short, is easy to FPGA hardware and realizes
(3) different peak clipping algorithm resources relatively
|
NS-CFR |
PC-CFR |
RC-CFR |
FFT/IFFT |
Memory block (BRAMs) |
5 |
8 |
10 |
24 |
Logic (SLICEs) |
3379 |
2084 |
3300 |
5193 |
Hardware multiplier (Emb.MULTS) |
130 |
10 |
16 |
83 |
Duration (time delay) |
4.70us |
4.74us |
5us |
84us |
Sampling number (time delay) |
180 |
182 |
211 |
8444 |
The RC-CFR algorithm is very near the PC-CFR algorithm in view of resource, and the resource of increase seldom.But the hardware multiplier that NS-CFR takies is a lot.From the peak clipping effect, to remove beyond the peak value, the influence that the signal of other position is subjected to is very little, and this all has benefited from physical layer and has finished most peak value and eliminate.
The amplitude limit algorithm that except that above-mentioned 3 algorithms, circulates in addition, i.e. IFFT/FFT, the resource consumption of circulation amplitude limit algorithm is big, and in time, prolong, and the peak clipping effect is not too obvious.
The NS-CFR algorithm phenomenon of peak clipping deficiency often occurs except the big shortcoming of resource consumption.The threshold value of peak value window all is located at the 7.5dB place, changes the peak clipping threshold value respectively, emulation three groups of situations, wherein PRB is assigned as 100 and at 4 o'clock, signal can both well converge to about 7.5dB after the peak clipping; PRB is assigned as 40 and at 12 o'clock, is subjected to the influence of peak clipping threshold value bigger, and exists the probability of peak value recurrence also very big.The disadvantage peak clipping effect instability of PC-CFR peak clipping algorithm is especially under sub-carrier configuration situation seldom.
(4) different peak clipping algorithm synthesis relatively
|
NS-CFR |
PC-CFR |
RC-CFR |
EVM worsens |
Greatly |
Generally |
Generally |
The FPGA resource consumption |
Greatly |
Little |
Little |
PAR reduces effect (subcarrier is many) |
Generally |
Good |
Good |
PAR reduces effect (subcarrier is few) |
Generally |
Difference |
Good |
The stability that PAR reduces |
Generally |
Difference |
Good |
So in sum, no matter the RC-CFR algorithm is saved in resource, the validity that time delay still eliminates, and the deterioration degree of EVM all is optimum or suboptimum.So LTE-CFR selects RC-CFR.The PAR of signal can be stable be controlled at below the 7dBC.
One of ordinary skill in the art will appreciate that and realize that all or part of step that the foregoing description method is carried is to instruct relevant hardware to finish by program, described program can be stored in a kind of computer-readable recording medium, this program comprises one of step or its combination of method embodiment when carrying out.
In addition, each functional unit in each embodiment of the present invention can be integrated in the processing module, also can be that the independent physics in each unit exists, and also can be integrated in the module two or more unit.Above-mentioned integrated module both can adopt the form of hardware to realize, also can adopt the form of software function module to realize.If described integrated module realizes with the form of software function module and during as independently production marketing or use, also can be stored in the computer read/write memory medium.
The above-mentioned storage medium of mentioning can be a read-only memory, disk or CD etc.
The above only is a preferred implementation of the present invention; should be pointed out that for those skilled in the art, under the prerequisite that does not break away from the principle of the invention; can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.