CN101621486B - Estimation of data transmission channel - Google Patents

Abstract

The invention discloses a simple and effective method for estimating a data transmission channel, which belongs to the technical field of digital information transmission. The method adopts the technical scheme that a sending end sends a training sequence and a receiving end calculates received data of the training sequence so as to obtain a channel parameter. Specifically, the training sequence consisting of a plurality of data segments is sent from the sending end, and each data segment consists of simple alternate positive and negative code groups; simple calculations of pairwise adding adjacent received data in the training sequence, and the like are carried out at the receiving end so as to accurately obtain a channel pulse response; and a tapping coefficient of a time-domain equalizer can be conveniently calculated according to a channel parameter so as to realize the equilibrium so that intersymbol interferences caused by channel transmission is effectively reduced. The tapping coefficient of the equalizer is once set without gradual and iterative adjustments, thus equalization is easy to realize. The method has strong equalization capacity without requirements on initial distortion, simple algorithm, less operand and easy realization and can also conveniently calculate residual distortion output by the equalizer and guarantee the accuracy of the received data in a mode of carrying out series connection and then equalization, thus the equalization quality is high. The method can also be applied to echo counteraction.

Description

The data transmission channel method of estimation

Technical field

The present invention relates to the transmission technique field of digital information.Particularly relate to channel estimating in the data communication system.

Background technology

In the data communications equipment of extensive use, because the existence of transmission medium receives data and all will inevitably produce intersymbol interference (ISI).Internetwork device (like television network) from the modulator-demodulator of the voice band of low speed to high-speed wideband all need possess the equalizer that transmission medium is compensated and reduce the error rate to improve data transmission quality.In order to realize just need knowing the characteristic of channel, the i.e. impulse response of channel to the equilibrium of transmission channel.Time-domain signal in the data transmission system can be represented as follows

$S\left(t\right)\⊗g\left(t\right)=X\left(t\right)$

S (t) is that transmission signal g (t) is that channel impulse response X (t) receives signal

to send signal and reception signal though the expression convolution can be foreseen; But can not simply obtain channel parameter, because

g(t)≠X(t)/S(t)

This just makes knows that in the receiving terminal desire channel parameter becomes a very complicated problems.So far still have many experts and scholars to carry out deep discussion research to it as research topic.The traditional solution is to adopt the delay equalizer structure, and classic methods is the tap coefficient that makes input signal and decision error associated adjustment equalizer.It utilizes training sequence to use successive iteration method to make its convergence at the beginning of the transmission data.This technology has been used for a long time in the modulator-demodulator of voice band.In extensive use at present and the TV broadcast net greatly developed, utilize training sequence or pilot tone to carry out channel estimating and equilibrium.Wherein the known signal of handlebar training sequence becomes frequency domain from time domain, also becomes frequency-region signal to the time-domain signal of the distortion that receives again, and both become time domain again after frequency domain is divided by, thereby obtains the time domain channel parameter at last.In frequency domain

S(f)×g(f)＝X(f)

The multiplication of following formula in frequency domain is arithmetical operation, can use the division of arithmetic to carry out inverse here.

g(f)＝X(f)/S(f)

In order to realize that equilibrium also need become time-domain signal to frequency domain again.Be to use delay equalizer because transmission channel compensated.

There are a lot of problem and shortage in these technology and the method that adopt at present.At first be that convergence is slow, the training time is long, especially the adjusting step by step of the method for successive iteration needs could restrain, can not be too big in order to guarantee that precision is once regulated.Next is that algorithm routine is complicated, realizes that scheme, the mathematical expression of channel estimating is all more loaded down with trivial details, and particularly time domain and frequency domain are changed each other and seemed that more step is many.The 3rd is that degree of regulation is poor, and eye pattern degree of opening is undesirable.With the successive iteration method is example, and its degree of regulation can not be fine, otherwise regulate too slow and can not adapting to.The 4th is condition of convergence limitation, exceeds this limitation and will produce the channel estimation error equalizer and disperse and make system's cisco unity malfunction.Perhaps exceed this limitation and make judgement generate misjudgement, and then produce diffusion final formation of moving in circles and disperse.In addition also have neural net balanced, this not only needs a large amount of reception data promptly to take very long time of reception but also need the complex calculations process.Useless again worker when this method is promptly useless, long-acting rate of time are low, are not considered in ordinary circumstance.In sum, the method that these technology are used is complicated, operand is big, convergence is slow, efficient is low, and its reason just is can not directly simple and efficient acquisition channel parameter.Research that hence one can see that is a kind of promptly simply, fast, the channel estimation methods of accurate and effective acquisition channel parameter is the important topic that presses for very much solution again.

Summary of the invention

The present invention provides a kind of training sequence that utilizes that data transmission channel is carried out estimation approach.It can obtain channel parameter fast and accurately, solves that the long convergence of training time that more existing relevant art exist is slow, some problems of complex algorithm, low precision.

The present invention uses a kind of training sequence of specific code type.It is generated by the training sequence generator of transmitting terminal, can also can transmit the sequence of sending this specific code type between the data before in the formal data that transmit, and also can receive this yard type of extracting data if having ready conditions certainly.For the fundamental block diagram of band modulation demodulator data transmission system shown in figure one.Training sequence is produced by transmitting terminal, arrives receiving terminal through Channel Transmission, in channel estimator, channel parameter is estimated.Thereby channel estimator obtains to regulate the channel equalizer realization to the balanced minimizing of transmission channel intersymbol interference behind the channel parameter.The characteristic that the training sequence of this specific code type has is: at first this specific code type is made up of the plurality of sections code character; The hop count that full sequence comprises is decided by channel conditions; Each in theory estimation only needs two segment encode groups under the situation that does not have interference, and consideration actual conditions channel disturbance seriously then needs more multistage code character.Secondly this specific code type is made up of the code element of continuous alternate positive and negative, the alternate positive and negative of the code element strictness in every section, appearance that can not be continuous on the occasion of or negative value.The 3rd the absolute value of symbol amplitude in every section is identical, and promptly the range value of code element equally is positive and negative different, promptly maybe be identical also possible different between identical each section of amplitude in each section.The 4th the first code element in every section is that the sign symbol of first code element is identical with the sign symbol of last code element of neighbour the last period.This just means, the continuous alternate positive and negative of symbol in every section does not produce transition, and positive and negative essential between section and section produces transition.The 5th code element number that comprise in every section is decided by channel conditions, and more seriously then code element should be many more for channel distortions, and the code element number should be equal to or greater than the number of channel impulse response.If the code element number that comprises in every section is less than the number of channel impulse response, channel estimating will produce mistake.The 6th training sequence is to form in base band, and its speed is that modulation rate or character rate are baud rate.Its each code element is all the same with normal other application datas of sending of transmitting terminal to have same power spectrum or waveform shape through same signal processing.To do following selection more suitable with convenience of calculation in order to operate in application; It is the same that the code element number of each section is all set, and the number of every segment encode unit all gets odd number, and with on the occasion of initial with on the occasion of ending; All symbol amplitude is all identical in the training sequence; If all symbol amplitude gets 1 in the training sequences, promptly be in each section of whole sequence ± 1 alternate, be made in the receiving terminal calculating operation of just being more convenient for like this.The structure of training sequence is shown in figure two.If transmitting terminal is a two-dimensional structure has homophase and a quadrature component, each segment encode unit amplitude of in-phase component is ± and a is alternate, each amplitude ± b of segment encode unit is alternate for quadrature component.Other requirement of pair of orthogonal component is identical with in-phase component as stated, and this moment, the transmission data of transmitting terminal should be

A _K+JB _K K＝1、2、3、… ------(1)

Training sequence arrives receiving terminal by transmitting terminal through Channel Transmission, receives data and is just comprising channel parameter information.If channel parameter is

g _i+Jh _i i＝1、2、3、…

The reception data of receiving terminal are expressed as

X _K+JY _K

Receiving data then should be the convolution of sending data and channel parameter with the relation of sending data and channel parameter

$X_K=\underset{i}{\mathrm{\Σ}}{g}_i{A}_{K-i}-\underset{i}{\mathrm{\Σ}}{h}_i{B}_{K-i}$ $Y_K=\underset{i}{\mathrm{\Σ}}{g}_i{B}_{k-i}+\underset{i}{\mathrm{\Σ}}{h}_i{A}_{k-i}----\left(2\right)$

Shown in (1) formula, data are after transmission channel arrives receiving terminal in the transmission data of training sequence duration of work transmitting terminal, and the reception data of receiving terminal are shown in (2) formula.For the convenience of channel parameter estimation, normally set B=0 then (2) formula become

$X_K=\underset{i}{\mathrm{\Σ}}{g}_i{A}_{K-i}$ $Y_K=\underset{i}{\mathrm{\Σ}}{h}_i{A}_{k-i}-----\left(3\right)$

Because sent specific training sequence at transmitting terminal; They arrive the information that receiving terminal is just carrying channel parameter from transmitting terminal through transmission channel; The special properties of training sequence makes the channel impulse response that comprises in the reception data of receiving terminal be in special position, and the respective value overwhelming majority of entrained channel impulse response is on the position of opposite in sign and can positive and negatively offsets in the adjacent reception data in front and back.Just because the above-mentioned technical characterictic of training sequence can take very shirtsleeve operation just can calculate channel impulse response at receiving terminal.Reception data addition in twos adjacent during the training sequence just can be obtained channel parameter divided by 2a more accurately.If same receiving terminal is a two-dimensional structure, under the situation of B=0, the adjacent reception data addition in twos of homophase and quadrature branch just can be obtained accurately again the channel parameter of homophase and quadrature divided by 2a.If the estimated value table of channel parameter is shown

$\hat{g}+J\hat{h}$

Computes is used in the estimation of channel parameter

${\hat{g}}_i=(X_{n+i}+X_{n+i+1})\frac{1}{2a}$ ${\hat{h}}_i=(Y_{n+i}+Y_{n+i+1})\frac{1}{2a}----\left(4\right)$

The number of supposing the code element of each section of training sequence be odd number, every section all with on the occasion of beginning being ± the amplitude b=0 of a, B that the position, end of certain section sequence number is the n position on the occasion of ending, wherein the amplitude of A.The impulse response of channel is g _-i～g _i, its length-i～i is an odd number.Then the receiving data sequence X of receiving terminal is

…

X _n-1＝ag _-i-…+ag _-2+ag _-1-ag ₀+ag ₁-ag ₂+ag ₃-ag ₄+…-ag _i

X _n＝-ag _-i+…-ag _-2+ag _-1+ag ₀-ag ₁+ag ₂-ag ₃+ag ₄-…+ag _i

X _n+1＝ag _-i-…+ag _-2-ag _-1+ag ₀+ag ₁-ag ₂+ag ₃-ag ₄+…-ag _i

X _n+2＝-ag _-i+…-ag _-2+ag _-1-ag ₀+ag ₁+ag ₂-ag ₃+ag ₄-…+ag _i

…

The receiving data sequence Y of receiving terminal is

…

Y _n-1＝ah _-i-…+ah _-2+ah _-1-ah ₀+ah ₁-ah ₂+ah ₃-ah ₄+…-ah _i

Y _n＝-ah _-i+…-ah _-2+ah _-1+ah ₀-ah ₁+ah ₂-ah ₃+ah ₄-…+ah _i

Y _n+1＝ah _-i-…+ah _-2-ah _-1+ah ₀+ah ₁-ah ₂+ah ₃-ah ₄+…-ah _i

Y _n+2＝-ah _-i+…-ah _-2+ah _-1-ah ₀+ah ₁+ah ₂-ah ₃+ah ₄-…+ah _i

…

Receiving terminal receives that above-mentioned receiving data sequence just can calculate channel impulse response very easily, just receives data addition in twos to these and can obtain divided by 2a, and wherein the g sequence of channel impulse response is

…

$(X_{n-1}+X_n)\frac{1}{2a}=g_{-1}$

$(X_n+X_{n+1})\frac{1}{2a}=g_0$

$(X_{n+1}+X_{n+2})\frac{1}{2a}={\mathrm{<figure-callout\; id="1"\; label="g"\; filenames="H2009101624980E00013.png"\; state="\{\{state\}\}">g</figure-callout>}}_1$

…

Wherein the h sequence of channel impulse response is

…

$(Y_{n-1}+Y_n)\frac{1}{2a}=h_{-1}$

$(Y_n+Y_{n+1})\frac{1}{2a}=h_0$

$(Y_{n+1}+Y_{n+2})\frac{1}{2a}=h_1$

…

The value that just can accurately obtain whole channel impulse responses through the simple operation that receives data addition in twos can clearly be described as stated.

Obtain the initial distortion that channel impulse response just can calculating channel

$R_0=\underset{i\&NotEqual;0}{\mathrm{\&Sigma;}}{g}_i^2+\underset{i}{\mathrm{\&Sigma;}}{h}_i^2+{(1-g_0)}^2------\left(5\right)$

After receiving terminal obtains accurately channel parameter, can there be several different methods to find the solution the tap coefficient of equalizer, for example for the effective tap coefficient that people's equalizer equation just can calculate time-domain equalizer immediately equalizer is restrained fully.The block representation of linear time equalizer is at figure three, and the block representation of DFF is at figure four.If the dateout of equalizer is expressed as

I _K+JQ _K

The equalizer dateout that its each tap addition of equalizer input data process obtains later on is

$I_K=\underset{m}{\mathrm{\&Sigma;}}{C}_m{X}_{k-m}-\underset{m}{\mathrm{\&Sigma;}}{D}_m{Y}_{k-m}$ $Q_K=\underset{m}{\mathrm{\&Sigma;}}{C}_m{Y}_{k-m}+\underset{m}{\mathrm{\&Sigma;}}{D}_m{X}_{K-m}----\left(6\right)$

The distortion of adopting training sequence of the present invention can also calculate the equalizer dateout supposes that they are

$\hat{G}+J\hat{H}$

During training sequence, can obtain the distortion of equalizer dateout easily, they can pass through computes

${\hat{G}}_i=(I_{n+i}+I_{n+i+1})\frac{1}{2a}$ ${\hat{H}}_i=(Q_{n+i}+Q_{n+i+1})\frac{1}{2b}----\left(7\right)$

The residual distortion of equalizer dateout

$R=\underset{i\&NotEqual;0}{\mathrm{\&Sigma;}}{\hat{G}}_i^2+\underset{i}{\mathrm{\&Sigma;}}{\hat{H}}_i^2+{(1-{\hat{G}}_0)}^2------\left(8\right)$

Technical scheme of the present invention and traditional classics and present compared with techniques following characteristic is arranged, at first the present invention is to the estimation of channel and compare the present invention with the successive iteration equalizer of classics with the equalizer that estimated value constitutes and do not have restrictive condition successive iteration equalizer in order to guarantee convergence strict restriction to be arranged then to the initial distortion of channel.The present invention does not exist the initial distortion to channel that any requirement is arranged in the calculating process of channel parameter estimation, and the configuration of equalizer (tap coefficient) has nothing to do with judgement but directly calculated by the impulse response of channel.And the successive iteration equalizer tap coefficient is adjusted according to decision error.It need be adjudicated when each adjusting and be in the main true, and decision value just possibly exceed the scope of correct judgement and declare mistake if initial distortion is too big, and just possibility wrong direction adjustment of the tap coefficient of equalizer at this moment can make equalizer disperse when serious.Secondly the present invention on channel parameter estimation and equalizer convergence speed be successive iteration equalizer and other any equalizers all can't than.The present invention only can preset equalizer tap correctly with doubling the shared time of channel distortion in theory, considers that other factors maybe the longer slightly time of needs, but both to differ greatly be incomparable.The very simple required operand of the 3rd channel parameter estimation process of the present invention seldom.The channel parameter estimation process complicacy of other method of using at present needs a large amount of sequential operations, and particularly time domain and the present invention such as the mutual conversion of frequency domain, matrix operation of some method of estimation use can save fully.The 4th the present invention can calculate the residual distortion of the initial distortion of channel, equalizer output, and this just can clearly know the quality condition of the situation and the local terminal receiver dateout of transmission channel very easily.What should cause concern especially here is that the present invention can allow to adopt equilibrium again, i.e. equalizer series connection.If owing to delay equalizer distortion diffusion makes the residual distortion of first equalizer too big; Utilize the distortion value of first equalizer dateout can calculate second equalizer continuation of second parametric equalizer composition equilibrium; Calculate the residual distortion of second equalizer then; If also not meeting the demands, this numerical value can also not continue equilibrium, till residual distortion to the last meets the demands.So just can guarantee the quality of last dateout fully.The 5th the present invention can also be used as the Echo Cancellation device.Because its principle of Echo Cancellation device is similar with equalizer.Comparing with other Echo Cancellation devices as the Echo Cancellation device is simple and practical effective too.

Description of drawings

Figure one shows typical data transmission system calcspar, in order to explanation training sequence and equalizer residing position and effect in data transmission system.Wherein modulator and demodulator if existence then is the passband transmission, then is a baseband transmission if do not exist in system.Because the present invention is in Base-Band Processing, so modulator and demodulator is not at the row of analysis.

Figure two shows structure, the composition of training sequence.All training sequence is made up of the plurality of sections code character, and every segment encode group is made up of the code element of some alternate positive and negatives.

Figure three shows the time-domain equalizer sketch.Here only shown the one dimension time domain equalization, two-dimensional case is self-evident, and just two groups of one dimensions repeats to add mutual intersection.The continuous square frame in top is a line of time delay among the figure, and the roundlet of line of time delay below is represented multiplier.All the output sum of the multiplier of line of time delay is as the output of equalizer.

Figure four shows the DFF sketch.Equally here also only shown the one dimension time domain equalization, two-dimensional case can be drawn equally and repeated no more.The latter half that the data of equalizer judgement output feed back to line of time delay simultaneously is in order to improve the performance of equalizer.

Embodiment

Transmission system can be that one dimension also can be a two dimension, for easy be that example specifies execution mode here with the one dimension, as long as two-dimensional case is for the two-dimentional formula of narrating more than the people.

Step 1 is sent training sequence in transmitting terminal design, arrangement.The most important condition that data transmission channel is estimated is at the correct training sequence of transmitting terminal design.It must go up form the transmission data of alternate positive and negative at character rate (or modulation rate), all sets equally in order to be convenient to calculate the code element number of being convenient to operate each section at receiving terminal, and the number of every segment encode unit all gets odd number, and with on the occasion of initial with on the occasion of ending.All symbol amplitude gets 1 in the training sequences, promptly each section of whole sequence be ± 1 alternate, in fact this just forms the sequence of each segment data repetitive cycling.Its structure and composition are shown in figure two.The length of every section sequence is equal to or greater than the total length of channel impulse response, and for example setting channel impulse response length here is 7, and then the data length of every section of training sequence is equal to or greater than 7, supposes that every section code element number is 9.Training sequence according to every section of such setting is all identical, and they should be

1、-1、1、-1、1、-1、1、-1、1、

Form by several sections as for whole training sequences, then will be by whole transmission system and application scenario decision.If how noise jamming greatly then should make it repetitive cycling with several sections, so that can utilize the channel response data of several sections acquisitions average in addition more.Here only narrate transmission channel estimation and equalization methods factors such as noise jamming are not discussed, therefore only analyze as representative with two segment datas, other data segment repetitive cycling are just the same.

The step 2 channel makes the transmission data produce distortion.Automatically generate when this step is actually data through Channel Transmission, also simultaneously enumerate out its generation rule and result for overall process is described.In the data transmission system of figure one, the data that are transmitted will inevitably produce distortion, promptly produce intersymbol interference.Degree for different transmission mediums and conveying circuit length generation distortion is different, and this can represent with different channels impulse response.Suppose that channel impulse response length is 7, then it can be represented as follows

g _-2g _-1g ₀g ₁g ₂g ₃g ₄

In the data of channel, after channel generation distortion, the reception data that just arrive receiving terminal can be expressed as

$X_N=\underset{i}{\mathrm{\&Sigma;}}{g}_i{A}_{N-i}$ i＝-2～4， N＝1、2、3、4、…

Wherein N is the sequence number that sends data.I is the sequence number of channel impulse response.Because training sequence is by the section cycle repeats, therefore the reception data at receiving terminal also should be the cycle repeats sequences.If certain section position, end is the n position, list as follows the corresponding channel parameter g of reception data substitution in N=n and n ± 7 scopes with after sending data (g=1) now

X _n-7＝-g _-2+g _-1-g ₀+g ₁+g ₂-g ₃+g ₄

X _n-6＝g _-2-g _-1+g ₀-g ₁+g ₂+g ₃-g ₄

X _n-5＝-g _-2+g _-1-g ₀+g ₁-g ₂+g ₃+g ₄

X _n-4＝g _-2-g _-1+g ₀-g ₁+g ₂-g ₃+g ₄

X _n-3＝-g _-2+g _-1-g ₀+g ₁-g ₂+g ₃-g ₄

X _n-2＝g _-2-g _-1+g ₀-g ₁+g ₂-g ₃+g ₄

X _n-1＝g _-2+g _-1-g ₀+g ₁-g ₂+g ₃-g ₄

X _n＝-g _-2+g _-1+g ₀-g ₁+g ₂-g ₃+g ₄

X _n+1＝g _-2-g _-1+g ₀+g ₁-g ₂+g ₃-g ₄

X _n+2＝-g _-2+g _-1-g ₀+g ₁+g ₂-g ₃+g ₄

X _n+3＝g _-2-g _-1+g ₀-g ₁+g ₂+g ₃-g ₄

X _n+4＝-g _-2+g _-1-g ₀+g ₁-g ₂+g ₃+g ₄

X _n+5＝g _-2-g _-1+g ₀-g ₁+g ₂-g ₃+g ₄

X _n+6＝-g _-2+g _-1-g ₀+g ₁-g ₂+g ₃-g ₄

X _n+7＝g _-2-g _-1+g ₀-g ₁+g ₂-g ₃+g ₄

More than be receiving terminal serial number bit during the training sequence in n-7～n+7 scope receiving data sequence.

The impulse response of step 3 calculating channel.According to receiving terminal during training sequence receiving data sequence can use very simple method to obtain channel parameter.Just can obtain channel impulse response divided by 2 again to continuous two reception data addition in twos in the receiving data sequence

$(X_{n-6}+X_{n-7})\&times;\frac{1}{2}=g_2--------(3-1)$

$(X_{n-5}+X_{n-6})\&times;\frac{1}{2}=g_3--------(3-2)$

$(X_{n-4}+X_{n-5})\&times;\frac{1}{2}=g_4--------(3-3)$

$(X_{n-3}+X_{n-4})\&times;\frac{1}{2}=0--------(3-4)$

$(X_{n-2}+X_{n-3})\&times;\frac{1}{2}=0--------(3-5)$

$(X_{n-1}+X_{n-2})\&times;\frac{1}{2}=g_{-2}--------(3-6)$

$(X_n+X_{n-1})\&times;\frac{1}{2}=g_{-1}--------(3-7)$

$(X_n+X_{n+1})\&times;\frac{1}{2}=g_0-------(3-8)$

$(X_{n+1}+X_{n+2})\&times;\frac{1}{2}=g_1-------(3-9)$

$(X_{n+2}+X_{n+3})\&times;\frac{1}{2}{=g}_2-------(3-10)$

$(X_{n+3}+X_{n+4})\&times;\frac{1}{2}=g_3-------(3-11)$

$(X_{n+4}+X_{n+5})\&times;\frac{1}{2}=g_4-------(3-12)$

$(X_{n+5}+X_{n+6})\&times;\frac{1}{2}=0-------(3-13)$

$(X_{n+6}+X_{n+7})\&times;\frac{1}{2}=0-------(3-14)$

Can know that from above result of calculation 3-5 formula to 3-13 formula is an one-period, it has comprised the impulse response of whole channels.What can find out that initial data and the end data addition of the last period of each section of training sequence obtain is the channel impulse response central value, if the training sequence that repeats all cycles all calculates, then can repeat to occur the numerical value in this cycle.Thereby can utilize the numerical value in a plurality of cycles on average to improve the reliability that obtains channel parameter, for example reduce interference of noise.

Step 4 is calculated equalizer tap coefficient according to channel parameter.Later just can adopt several different methods to calculate the tap coefficient of linear equalizer or DFF obtaining channel parameter.Adopt the tap coefficient of equalizer Equation for Calculating DFF in the instance here.Have only linear segment just can be readily solved for linear equalizer.DFF has two parts, first linear segment, and it two is feedback fractions.If equalizer has seven grades, tap coefficient 0 ,-1, the-the 2nd wherein, linear segment, tap coefficient 1～4th, feedback fraction, its equation is following

g ₀C _-2+g _-1C _-1+g _-2C ₀＝0 ------(4-1)

g ₁C _-2+g ₀C _-1+g _-1C ₀＝0 ------(4-2)

g ₂C _-2+g ₁C _-1+g ₀C ₀＝1 ------(4-3)

g ₃C _-2+g ₂C _-1+g ₁C ₀+C ₁＝0 ------(4-4)

g ₄C _-2+g ₃C _-1+g ₂C ₀+C ₂＝0 ------(4-5)

g ₄C _-1+g ₃C ₀+C ₃＝0 ------(4-6)

g ₄C ₀+C ₄＝0 ------(4-7)

The value of g just can solve the value of each tap coefficient C of DFF easily in the known above-mentioned simultaneous equations, thereby accomplishes the configuration of equalizer fast and accurately, makes it quick convergence.

Step 5 output receives data and calculates residual distortion.After the tap coefficient of equalizer is once given correct value, just can obtain the dateout of equalizer during training series immediately.Its formula

$\begin{array}{cc}{I}_K=\underset{m}{\mathrm{\&Sigma;}}{C}_m{X}_{K-m}& m=-2~4\end{array}$

Can also calculate the residual distortion of the distortion and the equalizer output of channel.Channel distortion can be with the above channel parameter g substitution computes that obtains.

$R_0=\underset{L\&NotEqual;0}{\mathrm{\&Sigma;}}{g}_L^2+{(1-g_0)}^2$

The residual distortion of equalizer then should be with the distortion computation of equalizer output.This distortion value also can obtain very easily, gets the numerical value addition in twos of the sequence continuous adjacent of the output I of equalizer again during training sequence promptly that it is half the, and this also with to obtain channel distortion the same, just can obtain the distortion of equalizer dateout accurately.

${\hat{G}}_L=(I_{n+L}+I_{n+1+L})\frac{1}{2}$ L＝-4、-3、…0、…7、8、

The formula of people's calculated distortion can be calculated the residual distortion of equalizer with its generation

$R=\underset{L\&NotEqual;0}{\mathrm{\&Sigma;}}{\hat{G}}_L^2+{(1-{\hat{G}}_0)}^2$

It should be noted that its length of distortion after the equalizer before nearly doubles than equalizer, so will there be enough length in the cycle of training sequence, otherwise that the numerical value that calculates has is folding, thus the generation mistake.Just can calculate the residual distortion of whole receiving system, the quality of reception of the self-evident whole receiving system of understanding that at the same time also just can be clear and definite to the distortion substitution distortion formula of the last dateout of system.

Claims (3)

1. method of utilizing training sequence to carry out channel estimating; It is characterized in that transmitting terminal sends a kind of training sequence of particular design; The specific requirement of this training sequence is: training sequence must be made up of the multistage sign indicating number, the code element in each segment encode must the strict code element by alternate positive and negative be formed, the absolute value of the symbol amplitude in each section must be identical, last code element sign symbol of first code element sign symbol in each section and the last period is identical, the code element number in each section should form and experiences same signal processing with other application datas in base band more than or equal to the number of channel impulse response, each code element; Continuous the carrying out of adjacent reception data in receiving terminal reply training sequence phase adduction in twos so just can obtain channel impulse response accurately divided by 2 times of the range value of training sequence code element.

2. a kind of method of utilizing training sequence to carry out channel estimating as claimed in claim 1 is characterized in that this training sequence must be made up of the multistage sign indicating number, and the hop count that training sequence comprises is looked channel conditions and decided, and the bigger hop count that then comprises of channel disturbance is many more.

3. a kind of method of utilizing training sequence to carry out channel estimating as claimed in claim 1; It is characterized in that each code element experiences same signal processing in base band formation and with other application datas; Have same power spectrum or waveform shape, its chip rate is modulation rate or character rate.

Images