CN103929212B - A kind of broadband receiver nonlinear blind identification and compensation method - Google Patents

Abstract

This invention one broadband receiver nonlinear blind identification and compensation method, relate to radio monitoring field, particularly relate to broadband receiver nonlinear blind identification and compensation method.This invention is by compensating distorted signal, obtain mixed signal, calculate the cost function of distortion component short-time energy information in mixed signal again, calculate the compensating parameter making distortion component minimum, then the compensating parameter in compensation model is updated, repeat this step until distortion component is preferably suppressed, thus there is the SFDR of raising receiver and the effect of the ability of extraction weak signal.

Description

A kind of broadband receiver nonlinear blind identification and compensation method

Technical field

The invention belongs to radio monitoring field, particularly relate to broadband receiver nonlinear blind identification and compensation method.

Background technology

Relative to transmitter system strong nonlinearity that may be present, the nonlinear characteristic of receiver is more weak.After receiver receives signal, Through analog devices such as amplifier, wave filter, frequency mixers, signal can produce non-linear distortion through these analog devices.ADC (analog-digital converter), as a kind of mixed processing device, is also the nonlinear important sources of receiver link, and signal passes through ADC Can produce distortion in time domain, frequency domain produces harmonic distortion, intermodulation distortion etc..Reducing receiver one of them method nonlinear is choosing With speed, conversion accuracy is higher, the ADC that SFDR is bigger.But, the price of high performance ADC is very Expensive.Now, research to the linearization technique of receiver both at home and abroad is less.In engineering, in order to ensure the linearity of receiver, One is the thought that can use back-off, will input signal power limited below 1dB compression point, only utilize receiver Linear gain region.The shortcoming of Power back is to have had a strong impact on the dynamic range of receiving of receiver, sacrifices efficiency of transmission simultaneously. Two is to use feedback technique, is extracted by the distortion component in nonlinear output signal, feeds back to input, the most again with input Signal carries out anti-phase cancellation thus obtains the output signal after suppressing nonlinear component, but the method also exists stablizing of reponse system Sex chromosome mosaicism, signal lag limits the maximum bandwidth allowing transmission signal present in transmission system simultaneously.Three is to use predistortion skill Art, is i.e. that input signal carries out certain conversion in advance so that its conversion characteristics is just reciprocal with the nonlinear characteristic of receiver, from And the non-linearity of cancellation receiver.But carrying out self-adapted pre-distortion before radio-frequency front-end, actual realization is upper exists bigger difficulty.

Current a kind of relatively conventional modeling compensation method is to look for table method.The method is to use the excitation of specific input signal to be compensated System, by all possible output code value and it relative to the Difference Storage of preferable input code in look-up table, linearize in reality During, using the instant amplitude value of output signal as index, search and obtain the corresponding margin of error, then by it from output signal Deduct, reach to compensate the purpose of systematic error.But owing to frequency and the amplitude of input signal have bigger scope, compensate simultaneously System be likely to be of memory effect, so for ensure compensate precision, need generate multidimensional look-up table various possible to index Input condition.The most so, to ensure precision, then great memory capacity it is necessarily required to, simultaneously need to address faster and locate Reason speed, the real-time that guarantee compensates.Therefore, the hardware implementation complexity of the method limits its application greatly.

Summary of the invention

It is an object of the invention to just weak point Curve guide impeller to background technology a kind of broadband receiver nonlinear blind identification and benefit Compensation method, thus reach to improve the purpose of the ability of the SFDR of receiver and extraction weak signal.

The solution of the present invention is a kind of broadband receiver nonlinear blind identification and compensation method, and the method includes:

Step 1: receiver receives analog signal, this signal processes through receiver and distortion occurs, obtains distorted signal, and distortion is believed Number it is made up of linear component and distortion component；

Step 2: build compensation model, and predesigned compensation parameter according to distorted signal, compensation model output compensates signal and believes distortion Number compensate, obtain mixed signal；

Step 3: use FFT (Fast Fourier Transform (FFT)) to ask for the power spectral density function of mixed signal, at power spectral density function In, judge linear component and the frequency of distortion component and bandwidth according to default received signal power thresholding；

Step 4: according to receiving signal and the frequency of distortion component and bandwidth, build the spectral response of comb filter, then carry out Inverse FFT obtains filter coefficient, is filtered with compensation signal after filter coefficient windowing, calculates the humorous of linear component Ripple item and intermodulation item；

Step 5: calculated the short-time energy of distortion component by the harmonic term of linear component and intermodulation item, reconstruct cost function, Obtain the compensating parameter when cost function obtains minimum of a value；

Step 6: use adaptive algorithm that the compensating parameter preset in step 2 is replaced with the compensating parameter obtained in step 5, weight Multiple above-mentioned steps, until this adaptive algorithm convergence during Ji Suaning, reaches to suppress the effect of distortion component.

The compensation model wherein built in step 2 is Volterra compensation model.

One broadband receiver nonlinear blind identification of the present invention and compensation method, by compensating distorted signal, obtain mixing letter Number, then calculate the short-time energy cost function of distortion component in mixed signal, calculate the compensating parameter making distortion component minimum, Then update the compensating parameter in compensation model, repeat this step until distortion component is preferably suppressed, thus there is raising and connect The SFDR of receipts machine and the effect of the ability of extraction weak signal.

Accompanying drawing explanation

Fig. 1 is the flow chart of the method that inventive receiver blind discrimination compensates；

Fig. 2 is that the power threshold of linear component judges schematic diagram.

Detailed description of the invention

Distortion is there is after step 1: receiver receives analog signal, the process through receiver radio frequency front end of this signal and AD sampling, Obtaining distorted signal, distorted signal is made up of linear component and distortion component.

Step 2: structure Volterra compensation model, the initial value ω of preset model compensating parameter ω₀If, the rank of Volterra model Number D=3, then compensation model output compensates signal is p (k)=v (k) ω₀,

Wherein v (k)=[y²(k),y(k)y(k-1)…,y³(k),y²(k)y(k-1)…y³(k-1)], ω₀=[ω₁,ω₂,ω₃,ω₄,ω₅,ω₆,ω₇]^T=[1,0,0,0,0,0,0]^TInitial value for model parameter, it may be assumed that

p(k)=ω₁y²(k)+ω₂y(k)y(k-1)+ω₃y²(k-1)+ω₄y³(k) +ω₅y²(k)y(k-1)+ω₆y(k)y²(k-1)+ω₇y³(k-1)

Then obtaining mixed signal s (k)=y (k)-p (k), wherein y (k) is distorted signal, and p (k) is for compensating signal.

Step 3: ask for the power spectral density of mixed signal s (k), and judge linear component and the frequency of distortion component and bandwidth:

Mixed signal s (k) being made FFT computing, is compensated the power spectral density function of signal s (k), previously given judgement is linear Power threshold P of component_threshold, according to linear component judgment criterion P > and P_threshold, then can get linear component frequency in frequency spectrum Rate f₁,f₂…f_NWith bandwidth l₁,l₂…l_N, frequency and bandwidth further according to linear component obtain harmonic component 2f₁,3f₁,…2f_N,3f_N... and intermodulation component f₁+f₂,…2f_N-1+f_N... frequency and bandwidth, as shown in Figure 2.

Step 4: according to frequency and the bandwidth of distortion component, structure comb filter calculates distortion component:

Construct the spectral response of comb filter according to harmonic wave and the frequency of intermodulation component and bandwidth, spectral response is obtained as IFFT To filter coefficient, use FIR filter to be filtered after filter coefficient windowing with mixed signal, calculate linear component Harmonic term and intermodulation item, if FIR filter vector is g=[g₀,g₁…g_B-1]^T, dimension is B × 1, and the most finally extract is mixed The nonlinear component closing signal is expressed as:

s_f(k,ω)=g^Ts(k)=g^T[y(k)-V^T(k)ω]

What wherein y (k) represented is the distorted signal after broadband receiver AD samples, and V (k) is to characterize to compensate mission nonlinear memory The matrix of part.ω is the compensating parameter that the present invention wants identification,

y(k)=[y(k-B+1),y(k-B+2)…y(k)]^T, V (k)=[v (k-B+1), v (k-B+2) ... v (k)], v(k)=[y²(k),y(k)y(k-1)…,y³(k),y²(k)y(k-1)…y³(k-1)]^T。

Step 5: harmonic term and intermodulation item according to linear component construct cost function J (ω), utilize adaptive algorithm to compensation Model parameter is updated.

The core of the present invention is in the case of ensureing that linear component energy is constant, makes nonlinear component energy minimum, due to s_f(k,ω)=g^TNot including linear component part in s (k), therefore making cost function is:

J(ω)=E[s² _f(k,ω)]

Recycling LMS adaptive algorithm (conventional is LMS or RLS algorithm) updates compensation model parameter ω so that cost The value of function is minimum.

LMS adaptive approach:

Cost function J (ω) is launched and seeks gradient, can obtain:

$\▿J\left(\mathrm{\ω}\left(n\right)\right)=-2E\left[V\left(n\right){\mathrm{gg}}^{T}y\left(n\right)\right]+2E\left[V\left(n\right){\mathrm{gg}}^T{V}^T\left(n\right)\right]\mathrm{\ω}\left(n\right)$

According to Wiener filtering steepest descent method thought: $\mathrm{\ω}(n+1)=\mathrm{\ω}\left(n\right)-\frac{1}{2}\mathrm{\μ}\▿J\left(\mathrm{\ω}\left(n\right)\right)$ Can obtain:

ω(n+1)=ω(n)-μ{E[V(n)gg^TV^T(n)]ω(n)-E[V(n)gg^Ty(n)]}

Instantaneous estimation value is utilized to approach assembly average, the LMS Adaptive Identification algorithm in available blind discrimination compensation system:

ω(n+1)=ω(n)-μ{V(n)gg^TV^T(n)ω(n)-V(n)gg^Ty(n)}。

Step 6: the ω substitution compensation model that adaptive algorithm is tried to achieve, thus mixed signal s (k) after being compensated:

The ω tried to achieve after adaptive algorithm being restrained substitutes in step 2, is compensated model output signal p (k), thus is mixed Close signal s (k)=y (k)-p (k).

After blind discrimination compensates output, the distortion component in signal is effectively suppressed, and weak signal can be separated from distortion component Out, improve the dynamic range of receiver, improve the nonlinear characteristic of its front end.

Claims (2)

1. the identification of broadband receiver nonlinear blind and a compensation method, the method includes:

Step 1: receiver receives analog signal, this signal processes through receiver and distortion occurs, obtains distorted signal, and distortion is believed Number it is made up of linear component and distortion component；

Step 2: build compensation model, and predesigned compensation parameter according to distorted signal, compensation model output compensates signal and believes distortion Number compensate, obtain mixed signal；

Step 3: use Fast Fourier Transform (FFT) to ask for the power spectral density function of mixed signal, in power spectral density function, root Linear component and the frequency of distortion component and bandwidth is judged according to default received signal power thresholding；

Step 4: according to receiving signal and the frequency of distortion component and bandwidth, build the spectral response of comb filter, then carry out Inverse FFT obtains filter coefficient, is filtered with mixed signal after filter coefficient windowing, calculates the humorous of linear component Ripple item and intermodulation item；

Step 5: calculated the short-time energy of distortion component by the harmonic term of linear component and intermodulation item, reconstruct cost function, Obtain the compensating parameter when cost function obtains minimum of a value；

Step 6: use adaptive algorithm that the compensating parameter preset in step 2 is replaced with the compensating parameter obtained in step 5, weight Multiple above-mentioned steps, until this adaptive algorithm convergence during Ji Suaning, reaches to suppress the effect of distortion component.

2. a kind of broadband receiver nonlinear blind identification as claimed in claim 1 and compensation method, it is characterised in that in step 2 The compensation model built is Volterra compensation model.