CN106209716A - A kind of method reducing extensive MU MIMO ofdm system peak-to-average power ratio - Google Patents

Abstract

The invention belongs to radio communication (wireless communication) technical field, particularly relate to a kind of utilization and alternately update (alternating direction method of multiplier) algorithm to reduce the algorithm of the peak-to-average power ratio (peak to average power ratio, PAPR) of extensive MU MIMO OFDM downlink.The present invention provides a kind of and need not the method that side information can reduce extensive MU MIMO ofdm system PAPR, and provides higher systematic function.By the present invention in that interference signal is positioned at channel zero space, make the interference signal of superposition that what systematic function to be affected almost without, by using alternative and iterative algorithm, each parameter of alternating iteration, finally give this interference signal.

Description

A kind of method reducing extensive MU-MIMO-OFDM system peak-to-average power ratio

Technical field

The invention belongs to radio communication (wireless communication) technical field, particularly relate to a kind of utilization and hand over Extensive MU-MIMO-is reduced for updating (alternating direction method of multiplier) algorithm The algorithm of the peak-to-average power ratio (peak-to-average power ratio, PAPR) of OFDM downlink.

Background technology

OFDM technology is that present radio communication is beneficial to one of most widely used technology, has been included into the bases such as LTE WiMax In these all emerging wireless communication standards, it is strong that it has mitigating frequency-selective fading ability, the high and anti-intersymbol of frequency efficiency The advantages such as interference performance is strong can be greatly improved the transfer rate of system.But, owing between OFDM its subcarrier each, phase place is mutual Independent, after superposition, the fluctuation range of signal may be very big, when launching signal and there is bigger PAPR, and the radio frequency mould of base station Block needs to be equipped with linear power amplifier, and this can increase the hardware cost of radio-frequency module, also can reduce the power of system simultaneously Efficiency.When base station is equipped with large-scale transmitting antenna, its hardware cost and power efficiency problem can become more serious.Cause This, in extensive MIMO-OFDM system, the most effective PAPR Restrain measurement of design is particularly important.

For the PAPR problem of ofdm system, existing research work is concentrated mainly on SISO and two kinds of fields of point-to-point MIMO Scape.And for the scene of MU-MIMO, relevant research work is the most few.Under MU-MIMO, the collaborative work between user is very Difficulty, therefore the PAPR suppression at transmitting terminal processes (such as without side information) the most transparent for user side.This also makes Become existing most of algorithm cannot be extended to MU-MIMO system.In the present invention, by under extensive MU-MIMO-OFDM system The problem reducing PAPR is modeled as an optimization problem, by launching increase by interference signal on signal, and utilizes ADMM Algorithm alternately updates parameters, owing to this interference letter is positioned at channel zero space by the present invention, so can disappear after transmitting Except this interference signal, receiving terminal is had little to no effect, affect the performance of system hardly, and need not any limit letter Breath.

Summary of the invention

It is an object of the invention to provide one to need not side information and can reduce extensive MU-MIMO-OFDM system The method of PAPR, and higher systematic function is provided.By the present invention in that interference signal is positioned at channel zero space, make the dry of superposition Disturb signal what systematic function affected almost without, by using alternative and iterative algorithm, each parameter of alternating iteration, finally give This interference signal.

Based on the method reducing extensive MU-MIMO-OFDM system peak-to-average power ratio alternately updated, including walking as follows Rapid:

S1, a given extensive MU-MIMO-OFDM downlink frequency domain signal X, reduce described frequency domain signal X row PAPR process, wherein, described reduction PAPR is processed as: in described frequency domain signal X, signal delta X is disturbed in superposition one repeatedly；

S2, initial parameter is set: make penalty factor ρ=0 in DMM iteration, order utilize two way classification to carry out the multiplication blocked Factor lambda=0, make iterations t=0, make initial superposition disturb signal delta X⁽⁰⁾=0, calculate X and mapping matrix described in S1Wherein, H_nFor channel；

S3, carry out outer loop, utilize two way classification that the time-domain signal Y after superposition interference signal is blocked, thus more New Y, to m=1 ..., M:Wherein, x_m,Table respectively Show X and Δ X^(t)M row,For L × N point Fourier transformation, PROXINF is for utilizing two way classification pairBlocking, λ is Penalty factor in two way classification, thus obtain Y^(t+1)M row；

S4, carry out ADMM interior loop, specifically comprise the following steps that

S41, make D⁽⁰⁾=Δ X^(t), U⁽⁰⁾=0, calculate A^(t+1)=F_LNY^(t+1)-X, wherein, A^(t+1)For the frequency after blocking Difference between territory signal and primary signal X, F_LNFor L × N point inverse Fourier transform, U is that the Lagrange in ADMM iteration is taken advantage of Son, D is the Δ X in ADMM iteration^(t), the D that final iteration produces is Δ X^(t+1)；

S42, for i=1 ..., I_max-1, repeat following recursion cycle:

$Z^{(i+1)}=\frac{(A^{(t+1)}+{\mathrm{\ρD}}^{\left(i\right)}+U^{\left(i\right)})}{1+\mathrm{\ρ}},$

U⁽ⁱ⁺¹⁾=U⁽ⁱ⁾+ρ(D⁽ⁱ⁺¹⁾-Z⁽ⁱ⁺¹⁾), wherein,Representing D respectively, the line n of Z, U, n isIn Some value,For random some subcarrier chosen from W subcarrier for transmitting, W is total subcarrier number, 1 ≤n≤W；

S5, finally return that interference signalAnd make t=t+1, if t is ＜ T_max, then step S2 is returned, otherwise End loop.

The invention has the beneficial effects as follows:

Even if by the iteration of little number of times, still can effectively reduce system PAPR, only need not affect system simultaneously Performance.Compared to the algorithm of traditional super reduction system PAPR, the present invention has higher actual operation.

Accompanying drawing explanation

Fig. 1 is the flow chart using the inventive method to be applied to reduce PAPR.

Fig. 2 is extensive MIMO-OFDM system down link block diagram

Fig. 3 is the amplitude of each algorithm under time-domain sampling, subcarrier in frequency domain.

Fig. 4 is PAPR and the relation of CCDF, the relation of SNR Yu BER.

Detailed description of the invention

Below in conjunction with the accompanying drawings, the present invention is described in further detail.

The present invention implements the PAPR for reducing ofdm signal, and for convenience of description, following embodiment will set up following system Model illustrates.

The system block diagram of MU-MIMO-OFDM system down link is as shown in Figure 2.Wherein, sky is launched equipped with M root in base station Line, the single-antenna subscriber number of synchronization service is K, and K ＜ ＜ M, and the sub-carrier number of OFDM modulation is N.Two ends at frequency band Generally having portion subcarriers to be used as protection interval, the most N number of subcarrier is divided into two parts: be positioned at setLoad Ripple transmits for data, is positioned at setSubcarrier be used as protect bandwidth.In Fig. 1, S_n∈O^K×1It is that the n-th son carries The transmission information of K user, X on ripple_m∈C^N×1For the frequency domain information after precoding, Δ X_m∈C^N×1For being used for reducing the frequency of PAPR Territory superposed signal, Y_m∈C^N×1For time-domain signal corresponding after superposed signal.

The problem reducing PAPR under extensive MU-MIMO-OFDM system is modeled as

$\begin{array}{ccc}\underset{X}{\min }\underset{m}{\Σ}\left|\right|y_m|{|}_{\mathrm{\∞}}& s.t.& \left\{\begin{array}{c}Y=F_{LN}^H(X+\mathrm{\Δ}X)\\ H_n{\left(x_n^r\right)}^T=0,n=1,...,N\end{array}\right.\end{array}$

Wherein, y_m∈C^LN×1For Y ∈ C^LN×MM row, represent on m root antenna the letter of the time domain after L times of over-sampling Number；X∈C^N×MSignal is sent for known frequency domain,For Δ X ∈ C^N×MLine n, represent on the n-th subcarrier for reducing (frequency domain) signal of PAPR institute superposition.Want this problem of direct solution extremely difficult, therefore, use the method alternately updated, successively Update Δ X and Y, thus obtain the solution of this problem

In emulation, if the base station of this mimo system has M=128 root to launch antenna, service K=16 single-antenna subscriber.Examine Considering total N=128 the subcarrier of this OFDM modulation, the frequency spectrum of use mapsIt is appointed as the 40MHz in WI-FI model, and setsIndividual subcarrier transmits for signal.Set initial transmissions signal X (to generate through convolutional encoding as original bit simultaneously Multinomial [5o 7o])), 64QAM modulation after modulated signal, wireless channel is frequency selectivity, and channel multi-path number is 8, Time domain channel response matrixD=1 ..., D obeys zero-mean, variance is the Gauss distribution of 1, and the n-th son of its equivalence carries Frequency response H on ripple_nCan be expressed asViterbi decoding is used to detect letter at receiving terminal Breath stream.In algorithm each parameter initial value be respectively T_max=200, I_max=2, λ=1, ρ=0.5.In addition frequency-region signal is entered The over-sampling of row L=4 times is to calculate PAPR.

Based on above-mentioned constructed model and definition, the invention provides a kind of algorithm based on alternating iteration and parameter is carried out Alternating iteration, thus try to achieve disturbing signal, to reduce the PAPR of system.

Comprise the steps:

S1, a given extensive MU-MIMO-OFDM downlink frequency domain signal X.Superposition one the most repeatedly Interference signal delta X just can reduce the PAPR of signal；

S2, initial parameter ρ, λ, t=0, Δ X are set⁽⁰⁾=0, and calculate initial transmissions signal X and mapping matrix

S3, outer loop:

Update Y, to m=1 ..., M

$q_m^{\left(t\right)}=F_{LN}^H(x_m+{\mathrm{\Δx}}_m^{\left(t\right)})$

$y_m^{(t+1)}=PROXINF(q_m^{\left(t\right)},\mathrm{\λ})$

WhereinFor LN point frequency domain over-sampling, PROXINF is for utilizing two way classification pairBlock.

S4, ADMM interior loop:

Make D⁽⁰⁾=Δ X^(t), U⁽⁰⁾=0, calculate

A^(t+1)=F_LNY^(t+1)-X

For i=1 ..., I_max-1, repeat following recursion cycle

$Z^{(i+1)}=\frac{(A^{(t+1)}+{\mathrm{\ρD}}^{\left(i\right)}+U^{\left(i\right)})}{1+\mathrm{\ρ}}$

U⁽ⁱ⁺¹⁾=U⁽ⁱ⁾+ρ(D⁽ⁱ⁺¹⁾-Z⁽ⁱ⁺¹⁾)

WhereinRepresent D respectively, the line n of Z, U.

S5, finally return that interference signalAnd make t=t+1, if t is ＜ T_max, then step S2 is returned, otherwise End loop.

Through aforesaid operations, just complete the estimation to Δ X and Y.

The related algorithm algorithm performance relative analysis with the inventive method of reduction PAPR will be make use of, with further below The performance of the checking present invention.

Two kinds of measurement indexs are used to carry out the performance of metric algorithm.One PAPR performance being used to weigh time-domain signal, is named Do complementary cumulative distribution function (complementary cumulative distribution function is called for short CCDF)； One is used to weigh the correctness that receiving terminal bit stream recovers, is called the bit error rate (Bit Error Rate, BER).Assume m The time-domain signal that root is launched on antenna is y_m, then on this transmitting antenna, the definition of PAPR is

$PAPR\left(y_m\right)=\frac{2N\left|\right|y_m|{|}_{\mathrm{\∞}}^2}{\left|\right|y_m|{|}_2^2}$

Bit error rate is in data communication, the digital signal received within a certain period of time occurs the bit number of mistake with The ratio of total bit number of digital signal received in the same time.

Fig. 3 (a), (c), (e) and (g) describe in each algorithm the width of time-domain sampling signal on first transmitting antenna Degree.From these figures it can be seen that it is proposed that the solution tried to achieve of algorithm each element more integrated distribution near the upper bound, from And have lower PAPR.B (), (d), (f) and (h) is the amplitude of the equivalence transmitting frequency-region signal of each algorithm.Can from figure Go out, ZF, FITRA and

It is proposed that algorithm all there is no an out-of-band interference, and clipping algorithm can cause serious dry at protection band alleviating distention in middle-JIAO Disturb.Illustrate that this method all shows more preferably than other algorithms on RSNR and success rate.Fig. 4 (a) is institute after 1000 tests There is the average result of the CCDF of antenna.Algorithm proposed by the invention compares ZF algorithm can reduce the PAPR (CCDF of 7dB (PAPR)=10-2).And FITRA algorithm can only reduce 3dB.But algorithm above all can cause the deterioration of BER performance.Fig. 4 (b) In can be seen that the present invention there are about the performance loss of 1dB, this can increase average emitted power mainly due to the present invention.Total comes Saying, the present invention both can be greatly reduced PAPR relative to other algorithms, and don't causes the significantly sacrificing of BER performance, compares other and calculates Method has obvious performance advantage.

To sum up told, it is proposed that invention be based on the reduction extensive MU-MIMO-OFDM downlink alternately updated The algorithm of PAPR, it is superposition one interference signal on initial transmissions frequency-region signal, and this interference signal is positioned at the kernel of channel In, this signal can be eliminated after by dissemination channel.By well-designed iterative algorithm, alternately each ginseng of update algorithm Number, thus obtain the estimated value of interference signal, the PAPR of system can be reduced after superposition and system BER performance will not be had the biggest Impact.Iterations is the most, and the PAPR performance of system is the best, and it is the most to launch antenna, and the performance of system is the best.

Claims (1)

1. the method reducing extensive MU-MIMO-OFDM system peak-to-average power ratio, it is characterised in that comprise the steps:

S1, a given extensive MU-MIMO-OFDM downlink frequency domain signal X, reduce PAPR to described frequency domain signal X row Processing, wherein, described reduction PAPR is processed as: in described frequency domain signal X, signal delta X is disturbed in superposition one repeatedly；

S2, initial parameter is set: make penalty factor ρ=0 in DMM iteration, order utilize two way classification to carry out the multiplication factor λ blocked =0, make iterations t=0, make initial superposition disturb signal delta X⁽⁰⁾=0, calculate X and mapping matrix described in S1Wherein, H_nFor channel；

S3, carry out outer loop, utilize two way classification that the time-domain signal Y after superposition interference signal is blocked, thus update Y, To m=1 ..., M:Wherein, x_m,Respectively represent X and ΔX^(t)M row,For L × N point Fourier transformation, PROXINF is for utilizing two way classification pairBlocking, λ is two way classification In penalty factor, thus obtain Y^(t+1)M row；

S4, carry out ADMM interior loop, specifically comprise the following steps that

S41, make D⁽⁰⁾=Δ X^(t), U⁽⁰⁾=0, calculate A^(t+1)=F_LNY^(t+1)-X, wherein, A^(t+1)Believe for the frequency domain after blocking Number and primary signal X between difference, F_LNFor L × N point inverse Fourier transform, U is the Lagrange multiplier in ADMM iteration, D For the Δ X in ADMM iteration^(t), the D that final iteration produces is Δ X^(t+1)；

S42, for i=1 ..., I_max-1, repeat following recursion cycle:

$Z^{(i+1)}=\frac{(A^{(t+1)}+{\mathrm{\ρD}}^{\left(i\right)}+U^{\left(i\right)})}{1+\mathrm{\ρ}},$

U⁽ⁱ⁺¹⁾=U⁽ⁱ⁾+ρ(D⁽ⁱ⁺¹⁾-Z⁽ⁱ⁺¹⁾), wherein,Representing D respectively, the line n of Z, U, n isIn certain One value,For random some subcarrier chosen from W subcarrier for transmitting, W is total subcarrier number, 1≤n≤ W；

S5, finally return that interference signal delta X^(t+1)=D^(I _max ⁾, and make t=t+1, if t is ＜ T_max, then return step S2, otherwise tie Shu Xunhuan.