CN101547178B - Linearization device and linearization method in broad-band communication system - Google Patents

Abstract

The invention discloses a linearization device with a memory nonlinear amplifier and a linearization method in a broad-band communication system. The linearization procedure is as follows: using a frequency domain dresser, a time domain dresser and a self-adaptive filter to respectively carry out frequency domain dressing, time domain dressing and filtering processing for transmitting signals and amplifying the dressed signals: attenuating the amplified signals to the electrical level before amplification and feeding back the signals and using feedback signals and input signals to respectively update filter tap coefficient, a time domain checking table and a frequency domain checking table; repeating the above steps to continually update the frequency domain checking table, the time domain checking table and the filter tap coefficient and then carrying out frequency domain dressing, time domain dressing and filtering processing for generated transmission signals by utilizing the updating value of the filter tap coefiicient, the dressing value of the time domain dresser and the dressing value of the frequency domain signal dresser; repeating the operation again and again till realizing the linearization function of an amplifier in a broad-band OFDM system. The invention has the advantages of high convergence speed, high convergence precision and high degree of linearization.

Description

Linearization device in the wide-band communication system and linearization technique

Technical field

The invention belongs to communication technical field, relate to the amplifier linearization devices in the broadband orthogonal frequency division multiplexing ofdm system, can be used for amplifier in the wide-band communication system the elimination of memory nonlinear distortion is arranged.

Background technology

OFDM is one of key technology in the Modern Communication System.It through serial to parallel conversion, is assigned to data flow at a high speed in the relatively low experimental process carrier channel of speed and transmits.Advantages such as ofdm system has the spectrum efficiency height, the anti-multipath decline and receiving much concern become one of core technology in the next generation wireless communication system, and a lot of agreements such as DVB, IEEE 802.11a have adopted OFDM as main modulation-transmission technology.The basic principle of OFDM is to pass through serial to parallel conversion to data flow at a high speed, is assigned in the relatively low experimental process channel of transmission rate and transmits, and can use quick inverse Fourier transform to simplify modulated process simultaneously.In order to distinguish the interval of OFDM modulation front and back, before inverse Fourier transform, the interval before the OFDM modulation just is called the frequency domain interval, and with after the inverse Fourier transform, the interval after the OFDM modulation just is called the time domain interval.Signal is carried out the OFDM modulation can be regarded as signal by the conversion process of frequency domain transform to time domain.

Ofdm signal is made up of a plurality of subcarrier stacks, and according to central-limit theorem, when number of sub carrier wave N was bigger, the OFDM baseband signal can be similar to thought Gaussian process, and its phase place is obeyed evenly and distributed, and envelope presents rayleigh distributed.Therefore ofdm signal faces high peak-to-average force ratio, and the ofdm signal after just modulating peak power possibly occur and compare very big problem with average power.Because it is unpredictable that a big peak value occurs, therefore require follow-up power amplifier to have very high dynamic range to come the amplitude of variation of adaptation signal envelope.

Power amplifier is one of critical piece of Modern Communication System, and amplifier commonly used comprises solid-state power amplifier and travelling-wave tube amplifier.Solid-state power amplifier is the less amplifier of a kind of distortion, but its efficient is lower; And travelling-wave tube amplifier is the high amplifier of a kind of efficient, nonlinear distortion suitable significantly.Usually, we more pay attention to the efficient of amplifier for system, so the significant amplifier of nonlinear distortion often is used, and have just put the table top of research to the handling problem of amplifier nonlinearity distortion.

Usually, the amplifier input-output characteristic exists between a linear zone.In case input signal exceeds between linear zone, amplifier can present characteristic of nonlinear, signal is caused the influence of nonlinear distortion.When the bigger signal of ofdm system excursion passes through amplifier, can produce the severe nonlinear distortion.Simultaneously, because there is an input saturation point in amplifier.Surpass the input signal of saturation point, can't carry out processing and amplifying.Because the uncertainty of ofdm signal peak value, its size possibly surpass the input saturation point of amplifier, causes the saturability distortion.So in ofdm system, the variation that amplifier needs very high dynamic range to come adaptation signal.Because other most of signals all are far smaller than this peak signal, so use the great dynamic range amplifier can cause the waste of efficient.

Being conceived to method for processing signals is the focus direction of early stage research.This method directly reduces the peak-to-average force ratio of signal through transmission signals is handled, and makes most of signal fall into a kind of technology in linearity of amplifier district.Based on this theory, a lot of methods has been proposed at present, comprise amplitude limit, windowing, coding.

So-called amplitude limit is limited in predefined maximum level to peak amplitude exactly.This method is simple and easy to realize once being widely used because of it.But the shortcoming of this method is also very tangible.Because the distortion of large-signal, the error rate of receiving terminal and error sign ratio all can increase greatly, and the frequency spectrum hyperplasia of exporting signal simultaneously is very serious.In order to resist this distortion, on the basis of amplitude limit, the method for windowing has been proposed again.Signal after windowing is to use window function to amplitude limit carries out a kind of method of smoothing processing.Windowing can reduce the outer hyperplasia of band of frequency spectrum to a certain extent, but still has the error rate and the too high problem of error sign ratio.

Coding is a kind of in addition to the Signal Processing method.Coding techniques is an a kind of method of the satisfactory signal of peak-to-average force ratio being delivered to transmitting terminal.Typical code character has block code, Reed-Muller sign indicating number etc.Block code is only applicable to subcarrier channel seldom, and practicality is not strong; The Reed-Muller sign indicating number is a kind of encoding scheme efficiently, but this coding method has restriction to the constellation kind, and practicality has limitation.

(selected mapping SLM) is a kind of common technology in the current practical applications to select figure.The SLM technology is used D statistical independent vector Y _dRepresent identical information, select the minimum symbol y of peak-to-average force ratio then _dTransmit.The individual independently vector Y of this D _dProduce by the rotation of pseudorandom vector.The initial D that proposes of this method is merely 2, develops into D afterwards and can get arbitrary value.Though SLM does not produce any distortion to signal, and along with the number of D increases, peak-to-average force ratio can improve gradually, and cost is exactly cost and the complexity that realizes, because it needs D inverse Fourier transform to handle.Simultaneously, transmitting terminal must transmit side information, notifies receiving terminal to use which group vector Y actually _d

In addition, also have other one type peak-to-average force ratio reduction technology, be called the predistortion linear technology.Pre-distortion technology is not to be conceived to directly reduce signal peak-to-average ratio, but through the predistortion transmission signals, offsets the influence of amplifier to signal, and amplify in the linearisation interval of pair amplifier, reaches the effect of covert reduction peak-to-average force ratio.Predistortion is to improve the effective technology of efficiency of amplitude, and pre-distortion technology commonly used has polynomial predistortion distortion, curve fit predistortion and digital look-up table predistortion.Wherein digital look-up table pre-distortion technology becomes the mainstream technology of pre-distortion technology owing to its effective flexibly, as to be easy to realize characteristics.In many pieces of articles, relate to digital look-up table Study on Predistortion Technology; Andrew S.Wright; Willem G.Durtler; " Experimentalperformance of an adaptive digital linearized power amplifier " IEEE Transactions onVehicular Technology, 1992,41 (4): 395-400; The Rascal algorithm of the single order feedback loop linear convergence that proposes is wherein more representational a kind of, in memoryless system, is proved to be very effective.

Traditional digital pre-distortion research all is based on memoryless amplifier model.The memoryless of amplifier is meant that its output is only relevant with current input, and irrelevant with input before.When signal bandwidth was far smaller than the input bandwidth of amplifier, signal was equivalent to a narrow band signal, and think that system is a memoryless system this moment.But in wide-band communication system, along with the increase of signal bandwidth, the amplifier nonlinearity distortion demonstrates Memorability.Compare with memoryless amplifier, have the output of amplifier of memory not only relevant, also relevant with input before with current input.Joel H.K.Vuolevi; Timo Rahkonen and Jani P.A.Mannine, " measurementtechnique for characterizing memory effects in RF power amplifier. " IEEE Trans.Microw.Theory Tech.; Vol.49; No.8.pp.1383-1389, aug.2001, the Memorability of pair amplifier is studied; Prove that this Memorability distortion can be regarded as the frequency selectivity distortion, this means that the distortion that the different frequency component of signal causes is different.Therefore traditional time domain look-up table pre-distortion technology can't be applicable to broadband system, needs improved method to remedy the influence that amplifier nonlinearity and Memorability distortion cause signal.Proposed a lot at present about the pre-distortion technology of amplifier under the memory state is arranged; Pascale Jardin, Genevieve Baudoin, " Filter lookup table method for power amplifier linearization " .IEEE Transactions on Vehicular Technology; 2007; 56 (3): 1076-1087, be called for short FLUT, the filter look-up table of proposition is comparatively outstanding a kind of of performance wherein.Fig. 1 is a structured flowchart of this method, and this structure between serial to parallel conversion module and amplifier module, connect time domain predistorter and sef-adapting filter use direct feedback signal to carry out the time domain question blank and upgrade; Use non-direct feedback signal to carry out the renewal of filter tap coefficients simultaneously.But because this technology has been used the tap coefficient of counting on the same group mutually with the question blank number, the tap group number that needs to upgrade increases greatly; Use non-Direct Learning structure to carry out the renewal of filter tap coefficients simultaneously, caused the shortcoming that system's convergence rate is slow, convergence precision is not high.

Summary of the invention

The objective of the invention is to overcome the shortcoming of above-mentioned prior art, linearization device and linearization technique thereof in a kind of wide-band communication system are provided, with raising convergence rate and convergence precision, thus the performance of raising whole system.

For realizing above-mentioned purpose, linearization device of the present invention comprises:

Constellation mapping; Serial to parallel conversion module, inverse Fourier transform module, parallel serial conversion module, amplifier module, attenuation module, serial to parallel conversion module, Fourier transform module, time domain look-up table module and sef-adapting filter module; Wherein: be connected the frequency-region signal trimmer between serial to parallel conversion and the inverse Fourier transform; After Fourier transform, connect frequency domain question blank renovator; This renovator module provides the trimming value of a signal for the frequency-region signal trimmer, and the output signal of serial to parallel conversion is repaired; Between frequency domain question blank renovator and the time domain question blank renovator; Connect serial to parallel conversion and Fourier transform; This serial to parallel conversion is changed into signal parallel by serial, for Fourier transform provides a parallel signal, Fourier transform is the feedback signal that frequency domain question blank renovator provides frequency domain.

Described frequency-region signal trimmer comprises allocation index, amplitude question blank, phase place question blank, signal trimmer; The output signal of serial to parallel conversion is through the inquire address of allocation index output amplitude question blank and phase place question blank, and amplitude question blank and phase place question blank use this inquire address to be respectively the signal trimmer range value and phase value in order to trim signals are provided.

Described frequency domain question blank renovator uses the output signal of serial to parallel conversion and the output signal of Fourier transform, exports the entry updating value of corresponding inquire address respectively to amplitude question blank and phase place question blank.

For realizing above-mentioned purpose, linearization technique of the present invention comprises the steps:

(1) frequency domain question blank, time domain question blank and filter tap coefficients are carried out initialization;

(2) utilize frequency domain question blank, time domain question blank and filter tap coefficients, respectively transmission signals is carried out frequency domain finishing, time domain finishing and Filtering Processing, the signal after obtaining repairing;

(3) signal after the finishing is divided into two-way after amplifying; One the tunnel sends into channels transmit; Feed back after decaying to the level before amplifying on another road; Use input signal and feedback signal respectively filter tap coefficients, time domain question blank and frequency domain question blank to be upgraded, obtain the trimming value of filter tap coefficients updating value, time-domain signal trimmer and the trimming value of frequency-region signal trimmer;

(4) generate transmission signals once more; Use frequency domain question blank, time domain question blank and filter tap coefficients after upgrading; Respectively signal is carried out frequency domain finishing, time domain finishing and Filtering Processing; Make signal after the finishing better present the contrary characteristic of amplifier memory nonlinear distortion, further offset the memory nonlinear distortion of amplifier;

(5) repeating step (4); Bring in constant renewal in frequency domain question blank, time domain question blank and filter tap coefficients; And the transmission signals that generates carried out frequency domain finishing, time domain finishing and Filtering Processing again; Mean square error realizes having in the wide-band communication system linearization function of memory nonlinear amplifier less than 10-3 between feedback signal constellation point after the decay and transmission signal constellation point.

The present invention is owing to adopted the associating method for using of frequency domain predistorter, time domain predistorter and sef-adapting filter; Memory nonlinear distortion that can erase amplifier; Thereby when channel is a noisy channel; When channel signal to noise ratio size was identical, it is better that its receiving terminal receives the bit error rate performance of signal; Because the present invention has increased predistorter at frequency domain, suppressed the outer hyperplasia of transmission signal spectrum band of transmitting terminal simultaneously, made the secondary lobe that sends signal spectrum lower, littler to the transmission signals influence of adjacent channel; Because the present invention has simplified time domain filtering tap coefficient update method and structure, reduced the renewal number of tap coefficient in addition, therefore whole device is by the convergence that brings into operation, and the speed that amplifier distortion is eliminated improves greatly.

Description of drawings

Fig. 1 is that existing erase amplifier has memory nonlinear distortion apparatus structure block diagram;

Fig. 2 is the structured flowchart of apparatus of the present invention;

Fig. 3 is the structure chart of apparatus of the present invention frequency domain predistorter;

Fig. 4 is the flow chart of the inventive method;

Fig. 5 is the inventive method and existing methods receiving terminal planisphere comparison diagram;

Fig. 6 is the inventive method and existing methods transmitting terminal signal spectrum comparison diagram;

Fig. 7 is that the inventive method and existing method study curve ratio are schemed;

Fig. 8 is the inventive method and existing methods ber curve comparison diagram.

Embodiment

With reference to Fig. 2, apparatus of the present invention are made up of constellation mapper 111, serial-parallel converter 101, frequency-region signal trimmer 102A, frequency domain question blank renovator 102B, inverse Fourier transform device 103, parallel to serial converter 104, time-domain signal trimmer 105A, time domain question blank renovator 105B, time-domain adaptive filter 106A, tap coefficient renovator 106B, amplifier 109, attenuator 110, parallel to serial converter 107 and Fourier transformer 108.Wherein:

Connect serial-parallel converters 101 in constellation mapper 111 back, this serial-parallel converter 101 changes serial signal into parallel signal.Frequency-region signal trimmer 102A is connected between serial-parallel converter 101 and the inverse Fourier transform device 103; This frequency-region signal trimmer 102A receives the parallel signal of serial-parallel converter 101 outputs; In amplitude question blank and phase place question blank, search amplitude trimming value and phase place trimming value that should the input signal address according to input signal; Obtain after the trimming value input signal being carried out the frequency domain finishing, be the signal after the 103 output process frequency domain finishings of inverse Fourier transform device; After Fourier transformer 103, connect frequency domain question blank renovator 102B; The output signal of Fourier transformer 103 and the output signal of serial-parallel converter 101 are sent into this frequency domain question blank renovator 102B, are amplitude question blank and phase place question blank difference output amplitude updating value and the phase place updating value of frequency-region signal trimmer 102A.

Time-domain signal trimmer 105A is connected between parallel to serial converter 104 and the sef-adapting filter 106A, and this time-domain signal trimmer 105A carries out time domain finishing processing to the signal of parallel to serial converter 104 outputs; Connect time domain question blank renovator 105B in attenuator 110 backs, the output signal of attenuator 110 and the output signal of parallel serial conversion 104 are sent into renovator, are the updating value of time-domain signal trimmer 105A output question blank.

Sef-adapting filter 106A is connected between time-domain signal trimmer 105A and the amplifier 109, and this sef-adapting filter 106A carries out Filtering Processing to the output signal of time-domain signal trimmer 105A, and the signal that will pass through after the Filtering Processing outputs to amplifier 109; Connect tap coefficient renovator 106B in attenuator 110 backs, the output signal of the output signal of attenuator 110 and time-domain signal trimmer 105A is sent into tap coefficient renovator 106B, is the updating value of sef-adapting filter 106A output tap coefficient.

Signal through after the Filtering Processing outputs to amplifier 109.109 pairs of these output signals of amplifier carry out processing and amplifying, and the signal after will amplifying is divided into two-way, and one the tunnel sends into channel transmits, and attenuator 110 is sent on another road simultaneously.Signal level after attenuator 110 will amplify is reduced to the level before amplifying, and the signal that will reduce behind the level is transferred to tap coefficient renovator 106B, time domain question blank renovator 105B and serial-parallel converter 107 simultaneously as feedback signal.Serial-parallel converter 107 becomes parallel signal with the feedback signal of input, and through Fourier transformer 108, output obtains the feedback signal of frequency domain.Frequency domain question blank renovator 102B uses frequency domain inverse feedback signal and input signal, respectively frequency domain amplitude question blank and phase place question blank is upgraded.Time domain question blank renovator 105B uses input signal and the feedback signal that is transferred to this time domain question blank renovator; The time domain question blank is upgraded; Tap coefficient renovator 106B uses input signal and the feedback signal that is transferred to this tap coefficient renovator, and the tap coefficient of filter is upgraded.

After generating transmission signals once more, after frequency domain finishing processing, time domain finishing processing and the Filtering Processing after this signal process is upgraded, can better present the contrary characteristic of amplifier memory nonlinear distortion, the distortion of erase amplifier is to the influence of signal.Along with transmission signals constantly generates; The tap coefficient of frequency domain question blank, time domain question blank and filter finally tends towards stability constant; Demonstrate the contrary characteristic of amplifier memory nonlinear distortion at this moment through the signal after frequency domain finishing processing, time domain finishing processing and the Filtering Processing, the distortion of erase amplifier.

With reference to Fig. 4, the inventive method comprises the steps:

Step 1 is carried out the frequency domain finishing to input signal, and the frequency domain question blank is upgraded.

1.1 the frequency domain question blank is carried out initialization process, and wherein the amplitude question blank is initialized as entirely 1, the phase place question blank is initialized as complete 0;

1.2 be transmitted after the signal generation,, carry out the frequency-region signal finishing and handle through being transformed to parallel signal after the digital modulation and being assigned on each number of sub-carrier.To input signal, confirm its subcarrier sequence number k and mapped constellation points numbering i, [k, i] as inquire address, searched the signal amplitude trimming value and the phase place trimming value of corresponding input signal address [k, i] in amplitude question blank and phase place question blank; After the trimming value that obtains amplitude and phase place, the signal Sf ' after adopting following formula to obtain repairing _{K, i}

Sf′ _k，i[Ha _k，i·exp(j·Hp _k，i)]·Sf _k，i

Wherein, Sf _{K, i}Be the input signal of [k, i] position, Ha _{K, i}Be the amplitude inquiry tabular value of corresponding inquire address [k, i], Hp _{K, i}Be the phase place inquiry tabular value of corresponding inquire address [k, i], j is an imaginary symbols;

1.3 the feedback signal after the decay obtains the feedback signal of frequency domain through after the Fourier transform, frequency domain question blank renovator uses this feedback signal Ff and input signal Sf, and the frequency domain question blank is upgraded as follows:

1.3.1 definition range error Function e _{K, i} ^ScaleWith phase error function e _{K, i} ^Rotate:

$\left\{\begin{array}{c}{e}_{k,i}^{\mathrm{scale}}=\left|{\mathrm{Ff}}_{k,i}\right|-\left|{\mathrm{Sf}}_{k,i}\right|\\ e_{k,i}^{\mathrm{rotate}}=\arg \left({\mathrm{Ff}}_{k,i}\right)-\arg \left({\mathrm{Sf}}_{k,i}\right)\end{array}\right.$

Sf wherein _{K, i}The expression input signal, k is the sequence number of subcarrier, i is the numbering of mapped constellation points, Ff _{K, i}Be corresponding Sf _{K, i}The frequency domain inverse feedback signal;

1.3.2 utilize the range error function and the phase error function of definition, amplitude question blank and phase place question blank upgraded by following formula:

$\left\{\begin{array}{c}{\mathrm{Ha}}_{k,i}^{\mathrm{mew}}={\mathrm{Ha}}_{k,i}^{\mathrm{old}}+\mathrm{\μ}\·\mathrm{sign}\left(e_{k,i}^{\mathrm{scale}}\right)\·{\left(e_{k,i}^{\mathrm{scale}}\right)}^2\\ {\mathrm{Hp}}_{k,i}^{\mathrm{new}}={\mathrm{Hp}}_{k,i}^{\mathrm{old}}+\mathrm{\μ}\·\mathrm{sign}\left(e_{k,i}^{\mathrm{rotate}}\right)\·{\left(e_{k,i}^{\mathrm{rotate}}\right)}^2\end{array}\right.$

Ha wherein _{K, i}Be the amplitude inquiry tabular value of [k, i] position, Hp _{K, i}Be the phase place inquiry tabular value of [k, i] position, the value before subscript old representes to upgrade, the value after new representes to upgrade, μ representes to upgrade step-length, and sign () is-symbol function is specially:

$\mathrm{sign}\left(x\right)=\left\{\begin{array}{cc}1& x\&GreaterEqual;0\\ -1& x<0\end{array}\right.;$

1.3.3 utilize the amplitude question blank updating value Ha that obtains _{K, i} ^NewWith phase place question blank updating value Hp _{K, i} ^New, the updating value that obtains the corresponding inquiring position of frequency domain question blank [k, i] is:

$H_{k,i}={\mathrm{Ha}}_{k,i}^{\mathrm{new}}\&CenterDot;\exp (j\&CenterDot;{\mathrm{Hp}}_{k,i}^{\mathrm{new}})$

Wherein j is an imaginary symbols.

Step 2 is done the time domain finishing to the modulated signal of exporting behind the parallel serial conversion, and the time domain question blank is upgraded.

2.1 the time domain question blank is initialized as complete 1;

2.2 to input signal, the amplitude of basis signal adopts following formula to confirm inquire address add:

$\mathrm{add}=\mathrm{ceil}\left(\frac{\left|\mathrm{St}\right|\&CenterDot;{\mathrm{tab}}_{\mathrm{num}}}{{\max }_S}\right)$

Wherein, | St| representes the amplitude of input signal St, tab _NumThe size of the expression question blank that uses, max _SExpression institute allows the maximum of the signal of input, and this is worth normally, and saturation point by amplifier determines, ceil () function is a function that the infinitely great direction of court rounds, and the address that this function assurance obtains is an integer.

2.3 confirm after the inquire address, the trimming value of in the time domain question blank, searching correspondence position, and, adopt following formula to repair to input signal:

St′ _add＝Rt _add·St _add

St wherein _AddThe input signal of expression corresponding address add, and Rt _AddBe illustrated in that sought after address is the signal trimming value of add in the question blank, St ' _AddExpression input signal St _AddThrough the output signal after the time domain finishing.

2.4 the feedback signal Ft after time domain question blank renovator uses input signal St and decays adopts following formula that the time domain question blank is upgraded:

${\mathrm{Rt}}_{\mathrm{add}}^{\mathrm{new}}={\mathrm{Rt}}_{\mathrm{add}}^{\mathrm{old}}(1-\mathrm{\&sigma;}\frac{{\mathrm{Ft}}_{\mathrm{add}}-{\mathrm{St}}_{\mathrm{add}}}{{\mathrm{Ft}}_{\mathrm{add}}})$

Wherein σ is the step-length of upgrading, St _AddThe input signal of representing corresponding inquire address add, Ft _AddThe corresponding input of expression St _AddFeedback signal, Rt _AddThe question blank trimming value of representing corresponding inquire address add, subscript old and new represent before the renewal respectively and the value after upgrading.

Step 3 is carried out Filtering Processing to the signal after finishing is handled through time domain, and the tap coefficient of filter is upgraded.

3.1 with the tap coefficient initialization of filter, first tap w ₁Be initialized as 1, all the other taps all are initialized as 0;

3.2 sef-adapting filter receives the output signal of time-domain signal trimmer, and adopts following formula to carry out filtering to this signal:

${\mathrm{Sw}}_n^{\&prime;}=\underset{i=1}{\overset{N}{\mathrm{\&Sigma;}}}{\mathrm{Sw}}_{n-i+1}{w}_i$

Wherein, Sw ' _nBe filter n output signal constantly, Sw _N-i+1Represent n-i+1 input signal constantly, w _iThe tap coefficient of i tap of expression filter, N representes the total number of filter tap;

3.3 the tap coefficient renovator uses feedback signal Fw and input signal Sw after the decay, adopts following formula that the tap coefficient of filter is upgraded:

$w_i^{\mathrm{new}}=w_i^{\mathrm{old}}+\mathrm{\&delta;}({\mathrm{Sw}}_n-{\mathrm{Fw}}_n){({\mathrm{Sw}}_{n-\mathrm{<figure-callout\; id="1"\; label="i\; +"\; filenames="H2009100222877E00031.png,H2009100222877E00032.png"\; state="\{\{state\}\}">i</figure-callout>}+1}+\mathrm{\&lambda;}\&CenterDot;{\mathrm{Sw}}_{n-i})}^{*}$

Sw wherein _nRepresent the n input signal of renovator constantly, Fw _nThe corresponding input of expression Sw _mFeedback signal, w _iValue after the tap coefficient of i tap of expression filter, subscript old and new are represented respectively and upgraded preceding value and upgrade, δ upgrades step-length, and λ is the estimated value of pair amplifier memory effect, ^*Computing is the conjugate operation symbol.

Step 4; Generate transmission signals once more; Use frequency domain question blank, time domain question blank and filter tap coefficients after upgrading; Respectively signal is carried out frequency domain finishing, time domain finishing and Filtering Processing, make signal after the finishing better present the contrary characteristic of amplifier memory nonlinear distortion, the influence that further erase amplifier distortion brings;

Step 5; Repeating step 4; Continuous generation along with transmission signals; Bring in constant renewal in frequency domain question blank, time domain question blank and filter tap coefficients, and the transmission signals that generates carried out frequency domain finishing, time domain finishing and Filtering Processing again, up to mean square error between the feedback signal constellation point after the decay and the transmission signal constellation point less than 10 ^-3Frequency domain question blank, time domain question blank and filter tap coefficients are finally tended towards stability; The transmission signals that generates carries out frequency domain finishing, time domain finishing and Filtering Processing again; Utilize this contrary characteristic of the amplifier that signal appeared memory nonlinear distortion after handling, offset the memory nonlinear distortion of amplifier, realize having in the wide-band communication system linearization function of memory nonlinear amplifier.

Effect of the present invention can further specify through following emulation experiment condition:

1. emulation experiment condition

Adopt 2048 idea carrier OFDM systems of 8 times of over-samplings in the emulation, the constellation mapping mode of signal is the square mapping of 16QAM of power normalization, upgrades step size mu=10 in the frequency domain question blank renovator ^-2The time domain question blank adopts and the identical question blank of number of sub carrier wave size, renewal step-length σ=10 in the time domain question blank renovator ^-3Sef-adapting filter adopts the FIR filter construction of 5 taps, renewal step-length δ=0.4 in the tap coefficient renovator, memory effect estimated value λ=0.2.

2. simulation result

Under desirable channel condition, as shown in Figure 5 to the result that the receiving terminal planisphere carries out emulation, wherein Fig. 5 a is the planisphere of literature method receiving terminal under desirable channel condition, and Fig. 5 b is the planisphere of the inventive method receiving terminal under desirable channel condition.Can find out that from the contrast of Fig. 5 a and Fig. 5 b existing FLUT method is after having eliminated amplifier memory nonlinear distortion, the receiving end signal constellation point still has certain diffusion and distortion; And the inventive method is after erase amplifier memory nonlinear distortion, and the receiving end signal constellation point is converged on the ideal position.

Under desirable channel condition, the result who the transmitting terminal signal spectrum is carried out emulation is as shown in Figure 6.As can beappreciated from fig. 6, the inventive method than the high 2dB of existing FLUT method attenuation outside a channel about, littler to the influence of adjacent channel;

For comparing convergence rate and convergence precision; Learning curve to the inventive method and existing FLUT method is carried out emulation; This learning curve is meant that mean square error between feedback signal and the transmission signals constellation point after the decay increases along with iterations and the curve that changes is as shown in Figure 7.As can beappreciated from fig. 7, the error after the inventive method learning curve is stable is littler, has improved convergence precision, and has reached stable quickly than existing FLUT method, has improved convergence rate.

Under the white Gaussian noise channel condition, the result who bit error rate performance is carried out emulation is as shown in Figure 8, and the curve of top is the signal to noise ratio-ber curve of existing FLUT method among Fig. 8, and the curve of below is the signal to noise ratio-ber curve figure of the inventive method.As can beappreciated from fig. 8, the bit error rate performance of the inventive method is higher than the bit error rate performance of existing FLUT method, is 10 in the error rate ^-5The time, signal to noise ratio improves about 0.6dB.

Claims (7)

1. the linearization device in the wide-band communication system; Comprise constellation mapping; The first serial to parallel conversion module, inverse Fourier transform module, parallel serial conversion module, amplifier module, attenuator, the second serial to parallel conversion module, Fourier transform module, time domain look-up table module and sef-adapting filter module is characterized in that:

Be connected the frequency-region signal trimmer between the first serial to parallel conversion module and the inverse Fourier transform module; After the Fourier transform module, connect frequency domain question blank renovator; This renovator provides the trimming value of a signal for the frequency-region signal trimmer, and the output signal of the first serial to parallel conversion module is repaired;

Between frequency domain question blank renovator and the time domain question blank renovator; Connect the second serial to parallel conversion module and Fourier transform module; This second serial to parallel conversion module is changed into signal parallel by serial; For the Fourier transform module provides a parallel signal, the Fourier transform module is the feedback signal that frequency domain question blank renovator provides frequency domain;

Connect the tap coefficient renovator between attenuator and the time-domain signal trimmer, this tap coefficient renovator uses the feedback signal of attenuator output and the output signal of time-domain signal trimmer, for sef-adapting filter provides one group of tap coefficient after the renewal.

2. linearization device according to claim 1; It is characterized in that: the frequency-region signal trimmer comprises allocation index, amplitude question blank, phase place question blank, signal trimmer; The output signal of the first serial to parallel conversion module is through the inquire address of allocation index output amplitude question blank and phase place question blank, and amplitude question blank and phase place question blank use this inquire address to be respectively the signal trimmer range value and phase value in order to trim signals are provided.

3. linearization device according to claim 1; It is characterized in that: the output signal of the first serial to parallel conversion module and the output signal of Fourier transform module be through frequency domain question blank renovator, exports the entry updating value of corresponding inquire address respectively to amplitude question blank and phase place question blank.

4. the linearization technique in the wide-band communication system comprises the steps:

4a. frequency domain question blank, time domain question blank and filter tap coefficients are carried out initialization;

4b. utilize frequency domain question blank, time domain question blank and filter tap coefficients, respectively transmission signals carried out frequency domain finishing, time domain finishing and Filtering Processing, the signal after obtaining repairing;

4c. the signal after the finishing is divided into two-way after amplifying; One the tunnel sends into channels transmit; Feed back after decaying to the level before amplifying on another road; Use input signal and feedback signal respectively filter tap coefficients, time domain question blank and frequency domain question blank to be upgraded, obtain the trimming value of filter tap coefficients updating value, time-domain signal trimmer and the trimming value of frequency-region signal trimmer;

4d. generate transmission signals once more; Use frequency domain question blank, time domain question blank and filter tap coefficients after upgrading; Respectively signal is carried out frequency domain finishing, time domain finishing and Filtering Processing; Make signal after the finishing better present the contrary characteristic of amplifier memory nonlinear distortion, further offset the memory nonlinear distortion of amplifier;

4e. repeating step 4d; Bring in constant renewal in frequency domain question blank, time domain question blank and filter tap coefficients; And the transmission signals that generates carried out frequency domain finishing, time domain finishing and Filtering Processing again, up to mean square error between the feedback signal constellation point after the decay and the transmission signal constellation point less than 10 ^-3, realize that the linearization function of remembering nonlinear amplifier is arranged in the wide-band communication system;

Described to the renewal of frequency domain question blank, carry out as follows:

4c1. Define the magnitude of the error function?

and the phase error function?

Sf wherein _{K, i}The expression input signal, k is the sequence number of subcarrier, i is the numbering of mapped constellation points, Ff _{K, i}Be corresponding Sf _{K, i}The frequency domain inverse feedback signal;

4c2. utilize the range error function and the phase error function of 4c1 definition, amplitude question blank and phase place question blank upgraded by following formula:

Ha wherein _{K, i}Be the amplitude inquiry tabular value of [k, i] position, Hp _{K, i}Be the phase place inquiry tabular value of [k, i] position, the value before subscript old representes to upgrade, the value after new representes to upgrade, μ representes to upgrade step-length, and sign () is-symbol function is specially:

4c3. the amplitude question blank updating value of utilizing 4c2 to obtain

With phase place question blank updating value Obtain frequency-region signal trimming value H _{K, i}For:

Wherein j is an imaginary symbols.

5. linearization technique according to claim 4 wherein saidly carries out Filtering Processing to signal, adopts following formula to carry out:

Wherein, Sw ' _nBe filter n output signal constantly, Sw _N-i+1Expression filter n-i+1 input signal constantly, w _iThe tap coefficient of i tap of expression filter, N representes the total number of filter tap.

6. linearization technique according to claim 4, wherein said frequency-region signal finishing, adopt following formula to carry out:

Sf′ _k，i＝H _k，i·Sf _k，i，

Sf ' wherein _{K, i}Be through the signal after the finishing, Sf _{K, i}Be the input signal of [k, i] position, H _{K, i}It is the frequency-region signal trimming value of [k, i] position.

7. linearization technique according to claim 4, wherein said to the filter tap coefficients renewal, adopt following formula to carry out:

Sw wherein _nRepresent the n input signal of renovator constantly, Fw _nThe corresponding input of expression Sw _nFeedback signal, w _iValue after the tap coefficient of i tap of expression filter, subscript old and new are represented respectively and upgraded preceding value and upgrade, δ upgrades step-length, and λ is the estimated value of pair amplifier memory effect, and the * computing is the conjugate operation symbol.

Images