KR100913080B1 - Method of channel estimation and equalizer coefficient initialization in digital transmit-receive system - Google Patents

Abstract

The present invention is to initialize the tap coefficients of the equalizer so that tap coefficients of the equalizer can converge even in a poor channel environment using only the training columns used in the existing 8VSB system. A method of initializing an equalizer coefficient, the method comprising: receiving the training sequence and calculating a cross correlation value with a training sequence previously stored in memory; estimating an impulse response of the channel from the calculated cross correlation value; and Obtaining the equalizer tap coefficients from the estimated impulse response of the channel and initializing the equalizer tap coefficients of the receiver. Accordingly, the performance of the equalizer can be improved by initializing the equalizer with more optimal coefficients. .

Feedforward filter, feedback filter, tap coefficient, equalizer

Description

Method of channel estimation and equalizer coefficient initialization in digital transmit-receive system

1 is a block diagram of a VSB receiver in a general digital transmission / reception system

2 is a diagram showing the structure of a decision feedback equalizer in a general digital transmission / reception system

3 is a view showing a VSB standard data frame in a general digital transmission and reception system

4 is a view illustrating a structure of a field simultaneous signal of 8VSB standard in a general digital transmission / reception system

5 is a flowchart illustrating a channel response estimation and equalizer coefficient initialization method of a digital transmission / reception system according to the present invention.

6 is a graph showing an ideal impulse response

7 is a graph showing an estimated impulse response

8 is an impulse response graph shown using the method according to the invention.

* Explanation of symbols for main parts of the drawings

1: antenna 2: tuner                 

3: intermediate frequency filter and synchronization detector 4: NTSC interference cancellation filter

5: equalizer 6: FEC

7: MPEG 8: timing information provider

9: feedforward filter 10: feedback filter

11 signal discrimination unit 12 adder

The present invention relates to a digital transmission and reception system, and more particularly, to a method of estimating a channel using a correlation between a synchronization sequence and a training sequence and initializing an equalizer coefficient using the result.

Unlike conventional analog transmission and reception systems, the digital transmission and reception system converts video and audio signals to digital and transmits them. Therefore, the digital transmission and reception system can not only receive the original signal but also compress the video and audio data without distortion. It is also possible to transmit more data in the same band than the analog transmission method, which enables high quality HDTV broadcasting.

In addition, the digital transmission and reception system also has the advantage that can transmit more than one program on one channel at the same time.

Modulation schemes for fully digital HDTV use mainly 16/32 quadrature amplitude modulation (QAM) for cable broadcasting or 8/16 VSB for terrestrial broadcasting.

This is because the use of multi-level modulation is inevitable in order to transmit a lot of data generated after encoding through the 6MHz band, which is an existing NTSC transmission channel.

The QAM method is one of very excellent transmission efficiency. Since the signal is represented in two dimensions, the distance between signals is relatively long, which increases the transmission efficiency. However, since the signal itself is represented by a complex number, the hardware cost is relatively increased. There is a disadvantage.

On the other hand, unlike the QAM method, the VSB method has a simple one-dimensional constellation, so hardware for processing data is simple, but the signal-to-symbol error rate is increased due to the small distance between signals, and the modulation / demodulation system is relatively complicated. .

On the other hand, the signal transmitted from the transmitter has various distortions as it passes through the transmission channel. The factors causing the distortion include Gaussian thermal noise, fading or multiplication noise due to fading, frequency conversion, nonlinearity, and time dispersion (Time dispersion).

Inter-Symbol Interference (ISI) is the biggest factor that degrades system performance in dual high speed digital communication systems. Multipath channel, non-ideal frequency response ), Group delay, etc.

The equalizer is used to reduce the ISI caused by the linear distortion of the channel. In particular, research is being conducted on the adaptive equalizer that can track the time-varying characteristics of the communication channel.

The adaptive equalizer exchanges a predetermined training sequence between the transmitting and receiving sides during the initial training period and cancels the ISI by canceling the ISI by canceling the channel's distortion characteristics, thereby reducing the internal distortion. It is a kind of inverse filter installed in the network and restores the data sent by the sender without error.

1 is a block diagram illustrating a VSB receiver in a general digital transmission / reception system.

1, a tuner 2 for tuning only a desired channel from a signal received by an antenna 1, and an intermediate frequency for filtering only a middle frequency of a channel tuned by the tuner 2 and detecting segment synchronization and field synchronization. A filter and synchronization detector 3, a timing information provider 8 that provides timing for processing an input signal according to the synchronization signal detected by the intermediate frequency filter and synchronization detector 3, and the intermediate frequency filter And an NTSC interference cancellation filter 4 for excluding channel interference from the channel signal output from the synchronization detector 3, and an equalizer 5 for compensating for linear channel distortion of the signal output from the NTSC interference cancellation filter 4. And an FEC 6 for correctly reconstructing the signal received from the output of the equalizer 5, and an MPEG 7 for reproducing the signal reconstructed in the FEC 6.

When configured in this way, when the channel received through the antenna 1 is a terrestrial channel, since the multipath changes from time to time, the equalizer 5 uses an adaptive equalization algorithm to track the change and restore the original signal. Should be used.                         

Such adaptive equalization algorithms include Least Mean Square (LMS) and Recursive Least Square (RLS) algorithms, and there are various structures such as a trapped delay line (TDL) and a lattice (Lattice) structure. The TDL structure is mainly used.

2 is a diagram illustrating a structure of a decision feedback equalizer in a general digital transmission / reception system.

As shown in Fig. 2, the conventional decision feedback equalizer is composed of a feedforward filter unit 9, a feedback filter unit 10, an adder 12, and a signal discriminator 11.

Here, the feedforward filter unit 9 outputs a convolution of the vector string and the tap coefficient which delay the input signal, and the feedback filter unit 10 delays the signal discrimination value of the feedforward filter unit 9 and the tap coefficient. Outputs the convolution of.

In this case, the feedback filter unit 10, which is a feedback part, is used to remove a part where interference between symbols occurs from a current estimation generated by previously detected symbols.

That is, the training train of a predetermined length known at the receiving end is transmitted to the feedforward filter 9, and the feedback filter 10 receives the training train and sets the tap coefficient to an appropriate value.

Immediately after this training sequence, real data is transmitted, and the adaptive equalizer uses an iterative algorithm to estimate the channel and tap coefficients and compensate the channel.

FIG. 3 is a diagram illustrating a VSB standard data frame in a general digital transmission / reception system, and FIG. 4 is a diagram of a field simultaneous signal structure of an 8 VSB standard in a general digital transmission / reception system.

As shown in Figs. 3 and 4, one segment is composed of 832 symbols, the first four symbols of each segment are segment synchronization signals, and 828 symbols are bit strings such as one PN511 and three PN63 used for frame synchronization. The trained training sequence is being used.

One field is composed of 312 segments of data and one segment of field synchronization signal, and one frame is composed of two fields.

In addition, 4 symbols constitute 1 byte in 8VSB, and 2 symbols constitute 1 byte in 16VSB.

The equalizer should be set to the optimum tap coefficient of the equalizer when reception is completed, but the worse the channel environment, the longer the training sequence is required, resulting in a problem that the actual data rate is lowered.

That is, in the current ATSC type 8VSB system, 832 symbol training sequences used for frame synchronization are used, but equalizer coefficients do not converge within the training sequence.

Accordingly, the present invention has been made to solve the above problems, and initializes the tap coefficients of the equalizer so that the tap coefficients of the equalizer can converge even in a poor channel environment using only the training trains used in the existing 8VSB system. The purpose is.

delete

delete

delete

An equalization method in a digital receiver according to the present invention for achieving the above object comprises the steps of calculating a cross correlation value between a received training sequence and a training sequence previously stored in the receiver; Estimating a channel impulse response of a feedforward filter from the cross-correlation between the training sequence pre-stored in the receiver and the interval where the received training sequence is shifted by a predetermined range of 0 or less by using the calculated cross-correlation value; Estimating a channel impulse response of a feedback filter from a cross correlation between a training sequence previously stored in the receiver and a section in which the received training sequence is shifted by one or more predetermined ranges; And initializing an equalizer tap coefficient of the receiver by obtaining an equalizer tap coefficient from the estimated impulse responses of the channel.
In addition, the equalization method in the digital receiver according to the present invention for achieving the above object comprises the steps of calculating a cross-correlation value between the training train received and the training train pre-stored in the receiver; Extracting a channel impulse response estimate of a feedforward filter and a channel impulse response estimate of a feedback filter using the calculated correlation values, and filtering components except for an impulse response component having an absolute value greater than or equal to a defined reference value; Obtaining an initial coefficient of the feedforward filter whose product with the convolution matrix composed of the channel impulse responses of the estimated feedforward filter becomes a delta function; Convolving the calculated initial coefficients of the feedforward filter and the channel impulse response of the estimated feedback filter to obtain initial coefficients of the feedback filter; And initializing coefficients of the equalizer with the obtained initial coefficients of the feedforward filter and the initial coefficients of the feedback filter.
Other objects, features and advantages of the present invention will become apparent from the following detailed description of embodiments with reference to the accompanying drawings.

A preferred embodiment of the channel response estimation and equalizer coefficient initialization method of the digital transmission / reception system according to the present invention will be described with reference to the accompanying drawings.

5 is a flowchart illustrating a channel response estimation and equalizer coefficient initialization method of a digital transmission / reception system according to the present invention.

Referring to FIG. 5, first, when the equalizer receives the training sequence, the correlation between the signal and the training sequence previously stored in the memory is calculated (S10).

In this case, since auto-correlation of the training sequence transmitted from the VSB transmitter has a property similar to that of the direct delta function, it is mutually calculated using this characteristic.

In embodiments, when the display of a PN sequence of length L to _L PN autocorrelation of a PN generally _L of the amplitude A as is shown in equation (1).

인 지점에서의 값이{-A}^{2}이 아닌 마치 잡음과 같은 값으로 나타날 것이다.However, if the length of the PN sequence used is finite and linear correlation is used,

At this point, the value will appear as noise, rather than {-A} ^ {2}.

Therefore, as shown in FIG. 6, the autocorrelation value of the sequence of six PN511 sequences used in the 8VSB system can be seen that the PN sequence has a periodic delta function. The characteristics are estimated.

와, 피드백 필터 부분의 채널 임펄스 응답 추정

을 추출하고 필터링한다(S20).And channel impulse response estimation of the feedforward filter part from the correlation value calculated by Equation 1

And channel impulse response estimation of the feedback filter portion.

Extract and filter (S20).

In detail, as can be seen from the correlation characteristics of the PN sequence shown above, the channel can be estimated from the cross correlation between the received training sequence u [k] and the training sequence s [k] stored in the receiver.

는 수신기에 저장되어 있는 훈련열 s[k]와 수신된 훈련열 u[k]가 0 ~ -N_ff의 범위만큼 천이된 구간과의 상호 상관으로부터 구할 수 있고, L_tr은 훈련열 길이를 나타내며, 정리하면 다음 수학식 2와 같다.That is, the channel impulse response estimation of the feedforward filter portion

Is obtained from the cross correlation between the training sequence s [k] stored in the receiver and the received training sequence u [k] transitioned by the range of 0 to -N _ff , and L _tr represents the training sequence length. To sum up, the following equation (2).

는 수신기에 저장되어 있는 훈련열 s[k]와 수신된 훈련열 u[k]가 1 ~ N_fb의 범위만큼 천이된 구간과의 상호 상관으로부터 구할 수 있으며, 정리하면 다음 수학식 3과 같다.In addition, channel impulse response estimation of the feedback filter portion

Can be obtained from the cross-correlation between the training sequence s [k] stored in the receiver and the received training sequence u [k] shifted by a range of 1 to N _fb .

라고 하면,

는 수학식 2와 수학식 3의 Z-변환의 합이 되며, 이를 정리하면 수학식 4와 같다.Estimated channel transfer function

Speaking of

Is the sum of the Z-transformations of Equations 2 and 3, which are summarized as Equation 4.

은 수학식 5와 같다.Estimation of the channel impulse response in which Equation 3 is connected to Equation 3 in Equation 4

Is the same as Equation 5.

Looking at the correlation values calculated as described above, not only peak values due to the channel but also noise values are included.                     

In addition, the channel impulse response of Equation 5 includes components due to actual channels and noise components due to thermal noise and channel interference, and these components serve as factors that hinder stability when estimating equalizer coefficients.

6 is a graph showing an ideal impulse response, and FIG. 7 is a graph showing an estimated impulse response.

As shown in FIG. 7, it can be seen that the impulse response estimated in the region having no signal component due to the noise component has an arbitrary value.

In order to prevent performance degradation due to such noise components, only impulse response components having an absolute value above an experimentally determined reference value among the channel impulse responses of Equation 5 are left through filtering, and the components having an absolute value below the reference value are replaced with '0'. By doing so, an impulse response as shown in Equation 6 below can be obtained.

N _th : minimum absolute value of the channel impulse response

When the filtering is performed by Equation 6, the impulse response is smaller than the impulse response of Equation 5 and is closer to the ideal impulse response as shown in the graph shown in FIG.

The initial coefficient of the feedforward filter is obtained using the estimated channel impulse response (S30).                     

Here, the initial coefficient of the feedforward filter to be obtained should be in the form of a direct delta function when convolved with the channel response of the feedforward filter.

Therefore, the initial coefficient of the feedforward filter is obtained by solving the equation by setting the product of the initial coefficient to be obtained and the convolution matrix of the feedforward channel response as a direct delta function.

은 수학식 7과 같은 열백터로 나타낼 수 있다.In detail, the impulse response of the channel corresponding to the impulse response feedforward filter portion of the channel estimated above.

May be represented by a column vector as shown in Equation (7).

From Equation 7, the convolution matrix H _CONV for performing the convolution operation with the feedforward coefficient of the equalizer may be configured as shown in Equation 8.

을 도입하면, 이상적인 피드포워드 등화기 탭 계수 벡터

는 수학식 9의 조건을 만족시킨다. In addition, the ideal response vector is configured for the tap length of the equalizer

Is introduced, the ideal feedforward equalizer tap coefficient vector

Satisfies the condition of Equation (9).

In addition, when all the vector components of the convolution matrix H _CONV are _solved and represented, they have a form of an upper triangular matrix as shown in Equation 10.

Therefore, the feedforward tap coefficient vector {h} _ {ff} to be obtained can be obtained by an iterative operation as shown in Equation 11 from the relationship of Equation 9.

The repetition operation of Equation 11 is represented by Equation 12.

The initial coefficient of the feedback filter is calculated using the initial coefficient of the feedforward filter calculated as described above, and the initial coefficient of the feedback filter is obtained by convolving the channel impulse response of the feedback filter with the initial coefficient of the feedforward filter (S40).

In other words, the tap coefficient of the feedback portion of the equalizer is obtained by the convolution of the channel response estimate of the feedback portion of Equation 3 and the feedforward tap coefficient of the equalizer.

이 되고, 따라서 등화기 출력 전달함수가 1이 되기 위해서는 피드백 등화기의 전달함수가

이 되어야 한다. 그러므로 등화기의 초기계수는 -{h}_{c}*{h}_{ff}[n]이 된다. (*:컨벌루션(convolution))In other words, the feedforward filter output

Therefore, in order for the equalizer output transfer function to be 1, the transfer function of the feedback equalizer

Should be Therefore, the initial coefficient of the equalizer is-{h} _ {c} * {h} _ {ff} [n]. (*: Convolution)

This process is summarized in Equation 13.

The coefficients of the equalizer are initialized with the initial coefficients of the feedforward filter and the feedback filter thus obtained (S50).

In this way, initializing taps of the equalizer using the feedforward tap coefficients obtained in Equation 12 and the feedback tap coefficients obtained in Equation 13 improves the convergence speed of the equalizer and improves the channel equalization capability.

After the coefficients are initialized and operated as a general equalizer (S60), if the system is reset during the operation of the equalizer or a new training sequence is received, the process is restarted (S70).

The channel response estimation and equalizer coefficient initialization methods of the digital transmission / reception system according to the present invention as described above have the following effects.

First, the characteristics of the channel can be estimated and used at the receiving end.

Second, by initializing the equalizer with more optimal coefficients, the performance of the equalizer can be improved.

Third, the present invention can be applied to all digital transmission / reception systems such as ATSC digital TV transmission / reception systems, digital data broadcasting transmission / reception systems, and digital cellular mobile phone systems that use a bit sequence having correlation characteristics as synchronization information or training sequence.

Those skilled in the art will appreciate that various changes and modifications can be made without departing from the spirit of the present invention.

Therefore, the technical scope of the present invention should not be limited to the contents described in the embodiments, but should be defined by the claims.

Claims (4)

delete Calculating a cross-correlation value between the received training sequence and the training sequence previously stored in the receiver; Estimating a channel impulse response of a feedforward filter from the cross-correlation between the training sequence pre-stored in the receiver and the interval where the received training sequence is shifted by a predetermined range of 0 or less by using the calculated cross-correlation value; Estimating a channel impulse response of a feedback filter from a cross correlation between a training sequence previously stored in the receiver and a section in which the received training sequence is shifted by one or more predetermined ranges; And Obtaining an equalizer tap coefficient from the estimated impulse responses of the channel and initializing the equalizer tap coefficient of the receiver. The method of claim 2, The channel impulse response estimate of the feedforward filter is

Using a formula, The channel impulse response estimate of the feedback filter is

Using a formula, each of which is

Is a channel impulse response estimate of the feedforward filter,

Is a channel impulse response estimate of the feedback filter, s [k] is a training sequence stored in the receiver, u [k] is the received training sequence, L _tr is a training sequence length, and N _ff is a feed A forward filter transition section length, wherein N _fb indicates a feedback filter transition section length. Calculating a cross-correlation value between the received training sequence and the training sequence previously stored in the receiver; Extracting a channel impulse response estimate of a feedforward filter and a channel impulse response estimate of a feedback filter using the calculated correlation values, and filtering components except for an impulse response component having an absolute value greater than or equal to a defined reference value; Obtaining an initial coefficient of the feedforward filter whose product with the convolution matrix composed of the channel impulse responses of the estimated feedforward filter becomes a delta function; Convolving the calculated initial coefficients of the feedforward filter and the channel impulse response of the estimated feedback filter to obtain initial coefficients of the feedback filter; And And initializing the coefficients of the equalizer with the obtained initial coefficients of the feedforward filter and the initial coefficients of the feedback filter.

Images