KR100811014B1 - Uplink Burst Equalizing Method In Broad Wide Access System - Google Patents

Abstract

The present invention relates to an uplink burst equalization method of a broadband wireless access (BWA) system. The present invention relates to a method for performing adaptive training on an equalizer by sending a training list before transmitting user data by using a hybrid method that combines prior adaptation and burst equalization. Preadaptation step; A channel tracking step of recording the coefficients after convergence and transmitting user data so that the equalizer performs tracking on the wireless channel; Executing a burst equalization procedure if the channel change results in a bit error rate of 1 or greater than 2; If the channel change results in a bit error rate of 2 or more, the process returns to the pre-adaptation step again and re-enters.

The present invention achieves switching by significantly different thresholds by setting different thresholds in a combined manner combining pre-adaptation and burst equalization, greatly reducing the number of pre-adaptations, improving effective broadband and increasing burst equalization. The system lowered the need for a working environment (reduced gaps between static channels or bursts) and improved the product utilization environment.

Wireless Communications, Broadband, BWA, Burst Equalization, Wireless Connectivity, Uplink

Description

Uplink Burst Equalizing Method In Broad Wide Access System

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the area of time division multiple access (TDMA) wireless communications systems, and more particularly to the uplink burst equalization technology of broadband wide access (BWA) systems of the time division multiple access (TDMA) method.

Following the constant development of communication technology, broadband wireless access (BWA) has attracted users' attention as an alternative to DSL and wired transmission technology. Services provided through broadband wireless access (BWA) can not only widen the competition with services provided via wired transmission, but also provide a wider range of services that cannot be provided via wired transmission.

Typically, high-speed data services that BWA provides over wireless channels include two types: Line of Sight (LOS) and Non Line of Sight (NLOS). .

Early broadband wireless access (BWA) systems operated under line of sight (LOS) conditions, mainly using single-carrier high-efficiency modulation techniques, such as Quad-phase Shift Key (QPSK), Quadrature Amplitude Modulation (QAM) is used to determine the Feedback Decision Feedback Equalizer (DFE) technique and the steering antenna technique to overcome the effects of multipath and interference.

However, the disadvantage of the system is that the cover range is reduced due to the limited eyes. Next-generation broadband wireless access (BWA) systems operate under NLOS conditions and are heavily impacted by multipath time delays and declines due to the absence of LOS, with key technologies being orthogonal frequency division multiplexing (OFDM) Antenna technology overcomes the shortcomings of early broadband wireless access (BWA) systems. Currently, the two types of technical standards are established by the IEEE802.16 Broadband Radio Access Workgroup.

Early broadband wireless access (BWA) systems typically use point-to-multipoint frequency-division duplex (FDD) operation, where the uplink uses time division multiple access (TDMA) and the downlink uses time division multiplexing. (TDM) is used. The uplink modulation is QPSK-16QAM and the downlink modulation is QPSK-16QAM.

Currently, the uplink uses burst QPSK modulation and the simple constellation during QPSK modulation results in a simple equalization technique (SNR: Signal-to-Noise Ratio), which is small compared to QAM. Prior equalization and prior adaptation techniques) or no equalization techniques. On the other hand, a disadvantage of QPSK modulation is that the frequency spectrum efficiency is lowered. In order to meet ever-increasing user demands for uplink broadband, high-frequency spectral efficiency must be increased. Therefore, 16QAM modulation and self-adaptive modulation techniques should be used, and simple equalization techniques have not been able to meet the demand. Therefore, burst equalization techniques of broadband wireless access (BWA) systems have emerged as one of the important problems to be solved urgently.

In conventional broadband wireless access (BWA) systems, burst equalization is mainly solved through two methods: one is pre-equalization and the other is pre-adaptation.

In this case, the pre-equalization technique is as follows: a pre-equalizer (pre-equalizer or precoder) is installed on the user side in the uplink direction, the equalizer coefficient is calculated at the base station in the distance measurement process, and then the corresponding coefficient is calculated. Is sent to the user to perform transpose equalization. The design of this method divides the multipath quantity into a static part and a dynamic part, and the static part has a slow change so that the equalizer coefficient only changes once over a long time, i.e., the first distance measurement Only coarse one realizes one coefficient update; On the other hand, the dynamic part can change quickly, simplifying the realization of the burst equalizer by first updating the coefficients whenever one or more bursts occur. This method can be used for cable modems, and the DOCSIS protocol version 1.1 adds the burst equalizer part.

On the other hand, the above pre-training technique is as follows: When the prior burst data is completed by sending the training sequence before user data transmission in advance so that the equalizer coefficients are sufficiently converged, the data transmission proceeds. Remember the equalizer count to use the next burst equalization and to resume training if channel changes result in equalizer out-of-range operation.

The advantage of the two methods is that the equalizer structure is simple and each burst packet does not contain a training sequence, which improves efficiency. However, it is only applied to a situation where the interval between static channels or bursts is small, and when the two conditions are not satisfied, the burst Equalization should be done. In this case, the front portion of each burst packet contains one training sequence, which is called a leader sequence or leader code, but in the data transmission the training sequence is a kind of consumption. In a self-adaptive equalizer, different coefficient update algorithms require different training sequence lengths, e.g., the training sequence length required for the Recursive Least Square (RLS) algorithm iterative is the minimum mean square error LMS ( It is smaller than the Least Mean Square algorithm, but the amount of computation and complexity required by the former is greater than that of the latter.

Another method of reducing the length of a training sequence is the equalizer coefficient preload technique, which stores the coefficient initial values pre-predicted before the start of the equalizer training process in the equalizer coefficient register, so that the equalizer coefficient initial value prediction is accurate. In this case, the equalizer may enter converged state without training.

US patent US 5970092 “Adaptively equalized burst receiver and method for upstream broadband data” refers to a burst equalized method of uplink broadband data, which method approximates a Newman-Holfman sequencer approximation. A preload is performed on an equalization coefficient using a wireless channel and using an approximation method with respect to an approximation value estimated by the channel. Because the channel estimation approximation is used and the equalizer coefficient is also preloaded by the algorithm, the equalizer in this method cannot be fully converged through one burst, which greatly affects the system performance.

The technical problem to be solved by the present invention is to provide an uplink burst equalization method for a broadband wireless access (BWA) system, thereby overcoming disadvantages such as cumbersome training process, short time interval, and slow convergence speed. have.

The technical design according to the present invention is to use a method of mixing pre-training and burst equalization. That is, the equalizer is trained and the user data is transmitted before the user data is transmitted. The equalizer tracks the changed wireless channel with reference to the determined user data. If the channel change is outside the equalizer tracking range, perform burst equalization if the bit error rate is greater than or equal to the threshold 1 or less. If the channel change exceeds the equalizer range, retrain if the bit error rate is greater than or equal to the threshold 2. .

The present invention utilizes a hybrid scheme that combines prior adaptation and burst equalization for uplink burst equalization.

The present invention relates to a stream burst equalization method of a broadband wireless access (BWA) system in which an equalizer uses a forward filter (FF),

Procedure 1, pre-adaptation step of sending a training sequence prior to transmitting user data to train the equalizer;

Step 2, a channel tracking step of recording the coefficients after convergence of the equalizer and transmitting user data so that the equalizer performs tracking on the wireless channel;

Step 3, a burst equalization step of executing a burst equalization procedure if the channel change results in a bit error rate of 1 or greater than 2;

Step 4, Iterative regression step to enter the pre-adaptation step again if the channel change causes a bit error rate of 2 or more threshold value; Broadband Wireless Access (BWA) uplink burst equalization method characterized in that proceeds.

In the present invention, the burst equalization method first performs channel estimation at a zero related threshold using the leader code of the burst packet as a reference sequence, then calculates an initial value of the equalizer coefficients based on the channel estimate, and then calculates the coefficient. The initial value is preloaded to the equalizer, the equalizer is trained by the leader code and the partial user data as the reference sequence, and the equalizer is sufficiently converged.After the training, the equalizer is determined as the reference sequence for the user data. Proceeds with equalization and outputs a decision signal.

In the present invention, the reader code uses an M sequence, and the channel estimation first generates a bilateral cyclic extension signal in which the local reference signal of the reader code is the M sequence, and the local reference signal consists of 1 and 0, Subsequently, the method further includes sampling a received signal corresponding to the corresponding reader code and performing a related algorithm on the sampling signal corresponding to the corresponding reader code and the corresponding local reference signal.

In the present invention, the procedure for calculating the equalizer coefficient initial value is

The initial value of the tap coefficient for calculating FF is
Iii) when n = 0,

이고,
ⅱ) n ≠ 0 일 경우,
c(n) = 0 이며;

ego,
Ii) if n ≠ 0,
c (n) = 0;

The initial value of the tap coefficient to calculate the FB (Backward Filter) is

;

;

는 채널 임펄스 응답 벡터,

는 결정재입력등화기(DFE)의 FF 계수 벡터, b는 FB의 계수 벡터, B는 FB의 계수 벡터 길이, 폭이 가장 큰 펄스는 h(0), 프론트 임펄스 응답 펄스는 [h(-n) h(-n+1)…h(-1)]，부분열 임펄스 응답 펄스는 [h(1) h(2)…h(n)]; 등화기 중 FF 계수가 부분열 임펄스 응답 펄스와 대응되는 계수는 [c(-n) c(-n+1)…c(-1)], 프론트 임펄스 응답 펄스와 대응되는 계수는 [c(1) c(2)…c(n)]이다. In the above formula,

The channel impulse response vector,

Is the FF coefficient vector of the DFE, b is the coefficient vector of the FB, B is the length of the coefficient vector of the FB, the largest pulse is h (0), and the front impulse response pulse is [h (-n h (−n + 1)... h (-1)], the partial string impulse response pulse is [h (1) h (2)... h (n)]; The coefficients in which the FF coefficients correspond to substring impulse response pulses in the equalizer are [c (-n) c (-n + 1)... c (-1)], the coefficient corresponding to the front impulse response pulse is [c (1) c (2)... c (n)].

The present invention achieves the transition by setting different thresholds in a mixed manner combining pre-adaptation and burst equalization, greatly extending the time interval of adaptive training, reducing the number of prior adaptations and improving effective broadband. The burst equalization process used in the present invention lowers the demands on the system working environment (static channel or interval between bursts), improves the demands on the product utilization environment, and performs a simple equalizer by performing channel estimation using a zero correlation threshold sequence. A method of estimating the coefficient initial value was provided, the channel estimation accuracy was improved, the burst equalizer training time was shortened, and the convergence speed was improved. In addition, by providing a new zero related threshold realization method for M sequences, the present invention greatly expands the selection range for zero related threshold sequences.

1 is a block diagram illustrating the principle of a broadband wireless access (BWA) system;

2 is a schematic diagram of the response of an exemplary radio channel impulse;

3 is a structural diagram of a Decision Feedback Equalizer (DFE);

4 is a flowchart of an uplink burst equalization method;

5 is a schematic of the Newman-Holfman sequence zero related threshold;

6 is a comparison diagram of convergence speeds of the burst equalization method and the RLS method; And,

7 is a bit error rate comparison diagram of the burst equalization method and the RLS method.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings and examples.

1 is an uplink schematic diagram of a BWA system using frequency division duplex (FDD). Among them, the sign shaping filter uses a 0.25 RRC (Root Rising Cosine) filter. The source terminal contains a 16QAM data source and 0.25 RRC. User data is subjected to 16QAM modulation, with a code rate of 5 MHz / s and a corresponding raw data rate of 20 MHz / s. After modulation, data is sent over the 0.25 RRC filter to the wireless channel. Receiver accepts signal superimposed with receiver Additive White Gaussian Noise (AWGN) first through 0.25 RRC filter and performs frequency complement, phase recovery, and optimal sampling time estimation in register and correlator, followed by first order It enters the DFE (DFE) at the sampling / signal sampling rate to eliminate inter-code interference and finally obtains the user data decision result. The crystal reinput equalizer (DFE) consists of nine FFs and seven FBs. The multipath time delay maximum is 1 microsecond, so that the radio channel length is 5 and the filter length can fully cover all the echoes in the channel.

The burst packet consists of a reader code of length 16 and 2000 coded user data. The reader code can be used for frequency / phase estimation as well as for optimal sampling time estimation and channel estimation.

In wireless data transmission, user data must pass through a wireless channel during transmission from the calling terminal to the receiving terminal. The model of the wireless channel can describe the channel characteristics mainly through the following. That is, path loss (including shading decline), multipath delay spread, decay characteristics, Doppler effect, and equivalent channel / adjacent channel interference.

When a signal sent from a transmitter arrives at a receiving terminal through a plurality of paths, a path through which the signal passes may be described as one time-varying channel impulse response.

(1)

(One)

는 펄스 실현 시간;

는 다경로 수량;

는 제k경로의 증익;

는 제k경로의 위상;

는 제k경로의 도달 시간이다. In this case, in the formula (1), t is the observation time;

The pulse realization time;

The multipath quantity;

Is the gain of the kth path;

Is the phase of the kth path;

Is the arrival time of the kth path.

For time-invariant channels, equation (1) can be simplified as follows:

(2)

(2)

The corresponding baseband receive signal is:

(3)

(3)

(4)

(4)

At this time, in the above formula, x ( t ) denotes a firing signal, and n ( t ) denotes additive noise.

The channel impulse response h ( t ) is shown in FIG. 2. The largest part is the main pulse, the front part of the main pulse is the front impulse response pulse, and the part after the main pulse is the substring impulse response pulse. Under NLOS conditions, the front impulse response pulse is strong due to the severe main pulse breakdown of the LOS signal. On the other hand, under LOS conditions, due to the weak front impulse response pulse, the channel pulse response consists of a wide main pulse and a substring pulse.

When the receiving signal passes through the filter and passes the equalizer, the code sampling rate is t = kT, where T is the code rate and the sampling signal of the receiving signal is the sum of the transmission signal x and the channel impulse response h product, plus the additive noise n . .

(5)

(5)

In this case, in Equation (5), L is the length of the channel impulse response after sampling.

If the above expression (5) is developed,

(6)

(6)

In Equation (6), the first term represents a desired code signal at the k-th sampling time, and the second term represents inter-code interference.

Inter-signal interference caused by multipath can be eliminated using the equalizer. Different equalizers have different performance and realization complexity, and in practice the most commonly used equalizers are DFE, simple equalizers such as line equalizers, eg ZF (Zero-Forcing) equalizers or Minimum Mean Squared Error (MMSE) equalizers. You can choose. In general, DFE outperforms line equalizers in wireless environments. That is, the mean squared error (MSE) that can be reached is small and the bit error rate is small. The reason is that wireless channel multipath can result in zero of the channel frequency response, but line equalizers result in large noise plasts at zero. Although the realization is straightforward because the performance of the MMSE equalizer is superior to that of the zero-forcing equalizer, the maximum likelihood sequence estimation (MLSE) equalizer is superior in performance but complex to realize.

The working principle of the DFE is to use FF and FB to eliminate inter-code interference and to determine one code, so that the code can eliminate the inter-code interference for later code before making a decision on the later code.

를 출력한다. DFE 실행은 두가지, 즉 훈련과정과 추적과정을 포함한다. 훈련 시 수신단이 이미 알고 있는 훈련 시퀀스(또는 리더 시퀀스)는 BF에 진입하여 결정기에 대해 훈련을 진행한다. 훈련 종결 후 훈련 시퀀스 입력을 차단하고, 결정기 출력은 피드백 필더에 진입한다. 마찬가지로, 버스트 등화에 대해서도 버스트 시작 후 등화기 계수는 훈련 시퀀스를 통해 조정을 진행하고 훈련 시퀀스 결속 후 등화기 계수는 판정 출력 시퀀스를 통하여 조정을 진행한다. 등화기 계수 업데이트는 MSE(Mean Squared Error) LMS(Least Mean Square) 알고리즘과 반복(iterative) MSE(Mean Squared Error) RLS(Recursive Least Square) 알고리즘을 이용할 수 있다. In the present invention, the structure of the crystal reinput equalizer DFE is as shown in FIG. DFE typically includes FF, decision device and feedback filter (BF), and the receiving signal y (k) is input to FF to determine the code

Outputs DFE implementation involves two things: training and tracking. During training, the training sequence (or leader sequence) already known by the receiver enters the BF and trains the decision maker. After the training is terminated, the training sequence input is blocked, and the determiner output enters the feedback filter. Similarly, for burst equalization, the equalizer coefficients are adjusted through the training sequence after the start of the burst, and the equalizer coefficients are adjusted through the decision output sequence after binding the training sequence. The equalizer coefficient update may use Mean Squared Error (MSE) Least Mean Square (LMS) algorithm and iterative Mean Squared Error (MSE) Recursive Least Square (RLS) algorithm.

The conventional pre-adaptive equalization method trains the equalizer in the distance measurement process using the existing sequence sent to the distance measurement process, and the equalizer coefficient at the end of the training due to the length of the training sequence in the distance measurement process is sufficient. It can ensure that it converges sufficiently. After the distance measurement is completed, the data is sent to the data transmission stage. At this time, the equalizer tracks the wireless channel using the trained coefficients, and once the cumulative change or sudden variation of the wireless channel exceeds the equalizer range, Training is carried out again. However, this method is suitable for the situation where the static channel or the burst interval is small and the training sequence is long.

4 is a flowchart of the uplink burst equalization method of the present invention. First, the training sequence is sent before the user data transmission to use for equalizer training. This is the initial training course. In the training process, the present invention refers to a long known sequence as a reference sequence. For example, the burst packet during the distance measurement process, the representative length is 200 characters. After the equalizer coefficients have sufficiently converged, the initial training process is complete and begins to transmit user data, while at the same time the equalizer begins to follow the radio channel to proceed with burst detection and time frequency synchronization. The equalizer coefficients are stored and used for the next burst equalization. When the bit error rate is greater than or equal to the threshold 1 and less than or equal to 2 due to the cumulative change in radio channels, burst equalization processing is performed. If the bit error rate due to cumulative changes in the radio channel is greater than or equal to the threshold 2, the equalizer cannot operate normally, and the equalizer must be retrained early. The threshold 1 is a critical point at which the channel change rate exceeds the equalizer tracing rate, and the specific expression indicates that the bit error rate continuously increases and the value is twice the maximum absolute value of the change rate of the bit error rate to an average value of five consecutive burst packet bit error rates. It corresponds to addition. The threshold 2 corresponds to the receiver bit error rate (BER), with a representative value of 10 ⁻³ .

Among them, the burst equalization process is as follows. Select one reader code of the burst packet to use for channel estimation, calculate the initial value of the equalizer coefficients using that channel estimate, then load the initial value of the coefficients into the equalizer and use the reader code and partial user data to After the training is completed, equalization is performed on the user information, and finally, a decision signal is output by using a decision maker.

Each procedure of the burst equalization process is described in detail below.

First, the channel estimation technique will be introduced. According to the present invention, channel estimation is performed using a burst packet reader code, and the theoretical basis may refer to a sequence zero related threshold.

In order to obtain an accurate value of the channel impulse response using the zero related threshold of the reference sequence, the recursive related functions of the reference sequence (ie, the reader code) and the corresponding recursive extension code must satisfy the following conditions.

(7)

(7)

는 순환 관련 함수, p는 시퀀스 번호, N은 시퀀스 길이, rr는 참조 시퀀스의 자기 상관이다. 상기 조건을 만족하는 구역 -D< p <D 를 제로 관련 역치라고 부르고, D는 제로 관련 역치의 길이, Newman-Holfman 시퀀스의 제로 관련 역치 길이는 5로서, 도 5와 같다. In the above formula,

Is a recursive function, p is the sequence number, N is the sequence length, and rr is the autocorrelation of the reference sequence. A region -D <p <D satisfying the above condition is called a zero related threshold, D is a length of zero related threshold, and a zero related threshold length of Newman-Holfman sequence is 5, as shown in FIG. 5.

When the zero-related threshold length of the transmitted leader code sequence is D and the reader code uses the QPSK modulation scheme, the burst packet consisting of the reader code and the user data passes through a radio channel with a channel impulse response length of L and reaches the receiver. . The receiver first generates a local reference signal of the reader code locally, and the local reference signal of the reader code generates both side circular extension signals of the reader code; Sampling is performed on the portion of the received signal corresponding to the next reader code, and thus channel estimation is completed by performing a correlation algorithm on the sampling signal corresponding to the reader code and the corresponding local reference signal. Depending on the nature of the zero correlation threshold at the zero related threshold, the correlation algorithm result is established by the following algorithm.

(8)

(8)

는 순환 상관 함수, p는 시퀀스 번호, x는 리더 코드와 대응되는 샘플링 신호, v는 리더 코드의 현지 참조 신호, L는 샘플링 후 채널 임펄스 응답 길이, c는 비례 상수이다. In the above formula,

Is a cyclic correlation function, p is a sequence number, x is a sampling signal corresponding to the reader code, v is a local reference signal of the reader code, L is a channel impulse response length after sampling, and c is a proportional constant.

는 채널 임펄스 응답이다. 따라서 제로 관련 역치에서 순환 관련 값을 이용하여 채널의 임펄스 응답을 정확하게 구할 수 있으며, 채널 임펄스 응답 추정 정확도의 제고는 버스트 등화기의 수렴 속도를 가속화할 수 있다. As you know, cyclic functions

Is the channel impulse response. Therefore, the impulse response of the channel can be accurately calculated using the cyclic related value at the zero related threshold, and the improvement of the channel impulse response estimation accuracy can accelerate the convergence speed of the burst equalizer.

The main lobe / side lobe of the Newman-Holfman sequence related figure is relatively small, 1 / 8th. Thus, the reader code can reduce the probability of false alarms by using the Newman-Holfman sequence. However, the zero-related threshold length of the Newman-Holfman sequence is 5, and thus, accuracy is lowered when channel estimation is performed using the sequence in a wireless channel having a channel impulse response of 5 or more.

In the present invention, channel estimation may be performed using the M sequence, but since the cyclic related value is -1 instead of the desired zero related value within the length range of the M sequence, the channel estimation is performed using the M sequence. In order to calculate the channel impulse response according to the procedure, an M sequence related threshold must be obtained. First, the M sequence is the leader code, and then the local reference signal of the next reader code is constructed, where the local reference signal of the reader code is a bilateral cyclic extension of the M sequence or a sequence consisting of 1 and 0, 1 and -1 It is not a sequence consisting of In this case, the M sequences have zero-related thresholds of the same length, greatly expanding the selection range of zero-related sequences. The other channel estimation procedure is the same as the procedure of performing channel estimation using a Newman-Holfman sequence.

For example, for an M sequence of length 15, the existing sequence is:

-1 -1 -1 1 -1 -1 1 1 -1 1 -1 1 1 1 1

In the case of circulation, the following sequence is called a cyclic extension sequence.

0 0 0 1 0 0 1 1 0 1 0 1 1 1 1

The zero related threshold in the two sequences is 15.

After the accurate estimate of the channel impulse response is obtained, the equalizer coefficient initial value is calculated.

, DFE의 FF의 계수 벡터를

, FB의 계수 벡터를 b로 할 경우, 채널과 FF의 합성 응답은 d = h*c, 이때 "*"은 합성곱이다. Vector of channel impulse responses

, Coefficient vector of FF of DFE

, If the coefficient vector of FB is b , the composite response of the channel and FF is d = h * c, where "*" is the composite product.

If the decision is correct, the input signal to the determiner is:

(9)

(9)

Where x is the dispatch data code and n is AWGN.

The mean square error is:

(10)

10

In 16QAM modulation, data code x ( k ) takes {± 1 ± 3} value with equal probability, and the channel impulse response length is L = 3,

, The mean square error of the sign data

,

Therefore, the mean square error with the input signal

(11)

(11)

는 양측 노이즈 공율 밀도이다. Where

Is the noise power density on both sides.

이므로，If the determinant is completely accurate,

Because of,

Therefore, if you simplify the above formula (11)

(12)

(12)

In the above equation, to minimize the mean square error, the minimum value of the leader pulse of the total channel response, d (0) = 1, and the noise energy minimization must be satisfied.

Considering that the channel leader pulse under LOS condition is small and there is a strong coupling relationship between the FF and FB of the DFE, the inter-code interference left by the FF through the training sequence can be eliminated by the FB, thus equalizing. The coefficients of the group can converge to finally remove inter-code interference.

Therefore, the equalizer coefficient initial value calculation can be calculated according to the following procedure. If the channel impulse response length is L = 2n + 1, the largest pulse among them is h (0), and the leader pulse is [h (-n) h (-n + 1). h (-1)], the rear pulse is [h (1) h (2)... h (n)]. In the equalizer, the FF coefficient and the channel impulse response are reversed, that is, the coefficient corresponding to the rear pulse is [c (-n) c (-n + 1)... c (-1)], the coefficient corresponding to the leader pulse is [c (1) c (2)... c (n)].

The specific calculation procedure is as follows:

First, the initial tap coefficient value of FF is
Iii) when n = 0,

(13)

(13)

Ii) if n ≠ 0,
c (n) = 0.

Next, the initial tap coefficient of FB is

（14）

（14）

B is the length of the FB tap coefficient vector b. At this time, an initial value estimate of the equalizer can be obtained. Training time can be shortened by quickly obtaining equalizer initial estimates, but equalizer training should be performed to achieve final equalizer convergence.

In the burst equalization process, training for the equalizer is done by leader code and partial user data. First, pre-training is performed by using a leader code as a reference sequence, and the purpose is to prepare for converting to decision feedback training with reference to the decision output. Most of the judgment output after the completion of prior adaptation is correct. Due to the limited reader code, the data in the leader code must be adapted for at least two equalizers.

Next, decision feedback training is performed to input partial user data into the equalizer as a reference sequence for training. After the equalizer is completely converged, the equalizer performs equalization processing on the user data again.

In order to verify the effectiveness of the burst equalization method of the present invention, a comparison is made with respect to the convergence speed using the present invention and the RLS algorithm. In the present invention, after the training is performed twice using the same leader code according to the estimated DFE initial value, the determination output training mode is entered.

6 is a graph in which the mean square error is changed according to a sign sample under the condition that SNR = 20 Db. Among them, graph a shows a relationship in which the bit error rate of the present invention is changed according to SNR, graph b is a relationship in which the bit error rate of the RLS algorithm is changed according to SNR, and graph c is an equalizer bit error rate when FB is not initialized. This is a relationship diagram that changes according to signal-to-noise (SNR). As can be seen from the graph a, using the present invention, after passing 115 codes, the MSE can reach 10 ⁻¹ , but when the RLS algorithm is used, 180 codes are required. Therefore, the convergence speed of the present invention is faster than the RLS algorithm.

6 shows a convergence graph when the FB is not initialized. At this time, the training sequence of 70 codes is additionally required for the final convergence of the equalizer. Through this, it can be seen that the equalizer coefficient preload effect of the present invention far exceeds the existing technology.

7 is a graph in which the bit error rate of the present invention and the RLS algorithm is changed according to the SNR, and the bit error rate is obtained through an average statistics of 1000 burst packets, among which the present invention refers to the reader code (secondary overlap). We train the equalizer with the dog determinant output and calculate the bit error rate. The RLS algorithm calculates the bit error rate after training the equalizer with 138 determinant outputs with reference to the reader code (secondary redundancy). As can be seen from the figure, the bit error rates of the present invention and the RLS algorithm are almost equal within the range of SNR = 10-16Db, demonstrating the validity of the present invention.

Finally, the above embodiments specifically describe the technical solutions of the present invention, and the technical spirit of the present invention is not limited to the above embodiments. Although the present invention has been described in detail with the preferred embodiment, those skilled in the art can understand that a certain modification or equivalent replacement for the technical solution is possible, without departing from the scope and scope of the technical solution.

The present invention relates to an uplink burst equalization technique of a broadband wireless access system in a time division multiple access system in the field of wireless communication systems, and is a hybrid scheme combining organic adaptation and burst equalization, and is automatically switched according to different threshold values. By realizing this, the time interval of adaptive training was greatly extended, the number of prior adaptations was reduced, and the range of effective broadband was further increased.

Claims (9)

In the uplink burst equalization method of a BWA system in which the equalizer uses FF, A pre-adaptation step of sending a training sequence prior to transmitting user data to train the equalizer; A channel tracking step of recording the coefficients after convergence of the equalizer and transmitting user data so that the equalizer performs tracking on the wireless channel; A burst equalization step of executing a burst equalization procedure if the channel change results in a bit error rate of 1 or greater than 2; An iterative regression step of entering the pre-adaptation step again if the channel change results in a bit error rate above the threshold 2; The uplink burst equalization method of the BWA system, characterized in that proceeds to. The method of claim 1, The pre-adaptation step of the uplink burst equalization method of a wideband radio access (BWA) system, characterized in that the training is carried out using a reference sequence of a long length. The method of claim 1, The burst equalization method first performs channel estimation at a zero related threshold using the reader code of the burst packet as a reference sequence, then calculates an equalizer coefficient initial value based on the channel estimate, and then frees the calculated coefficient initial value to the equalizer. Load, and perform equalizer training with the reader code and partial user data in the reference sequence so that the equalizer sufficiently converges.After the training, the equalizer performs equalization on the user data with the determined user data as the reference sequence and determines the decision signal. Uplink burst equalization method of a broadband wireless access (BWA) system, characterized in that it comprises a procedure for outputting. The method of claim 3, wherein The reader code consists of a 16-bit Newman-Holfman sequence and the reader code uses a QPSK modulation method. The channel estimation procedure first generates a local reference signal of the reader code locally, the local reference signal of the corresponding reader code generates bilateral cyclic extension signals of the reader code, and then samples the received signal corresponding to the reader code. And continuously performing an associated algorithm on the sampling signal corresponding to the reader code and the corresponding local reference signal to complete channel estimation. The method of claim 3, wherein The leader code uses an M sequence, The channel estimation first generates a bilateral cyclic extension signal in which the local reference signal of the reader code is the M sequence, and configures the local reference signal as 1 and 0, and then performs sampling on the received signal corresponding to the corresponding reader kodo, Uplink burst equalization method of a wideband radio access (BWA) system, characterized in that channel estimation is completed by performing an associated algorithm on a sampling signal corresponding to a corresponding reader code and a corresponding local reference signal. The method of claim 3, wherein The procedure for calculating the equalizer count initial value The initial tap coefficient for calculating FF is   Iii) when n = 0,

  Ii) if n ≠ 0,       c (n) = 0; The initial tap coefficient for calculating FB is

Is; In the above formula,

The channel impulse response vector,

Is the FF coefficient vector of the DFE, b is the coefficient vector of the FB, B is the length of the coefficient vector of the FB, the largest pulse is h (0), and the front impulse response pulse is [h (-n h (−n + 1)... h (-1)], the partial string impulse response pulse is [h (1) h (2)... h (n)]; The coefficients in which the FF coefficients correspond to substring impulse response pulses in the equalizer are [c (-n) c (-n + 1)... c (-1)], the coefficient corresponding to the front impulse response pulse is [c (1) c (2)... c (n)] uplink burst equalization method of a wideband radio access (BWA) system. The method of claim 3, wherein In the equalizer training procedure, first, at least two preliminary adaptive training sessions of the equalizer are performed using a leader code as a reference sequence, followed by decision re-input training and inputting partial user data to the equalizer as a reference sequence. Further comprising a procedure of proceeding with training until it is fully converged. The method of claim 1, The threshold 1 corresponds to the threshold at which the channel change rate is equal to or greater than the equalizer tracking speed, and the threshold 2 corresponds to the receiver reverse bit error rate. The method of claim 8, The value of the threshold value 1 corresponds to the average of five consecutive burst packet bit error rates plus two times the maximum absolute value of the rate of change of the bit error rate, and the representative value of the threshold value 2 is 10 ⁻³ . Uplink Burst Equalization Method for BWA) Systems.

Images