JP2001196978A - Adaptive equalization system, diversity reception system, and adaptive equalizer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an effective adaptive equalization system that can enhance transmission efficiency and reduce an arithmetic amount by equalization processing. SOLUTION: A tap coefficient update control circuit 12 updates tap coefficients F0(t)-F-j(t), B1(t)-Bk(t) of an adaptive equalizer. The initial value of each coefficient at reception of a training signal is adjusted by using the Kalman algorithm having a fast converging speed. The LMS algorithm requiring less arithmetic amount is employed for the reception of an information signal to update each tap coefficient by using the difference between the received signal and its discrimination value for an estimated error so as to track fluctuations in the transmission line.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multi-level digital modulated wave received by a diversity receiver and transmitted by an adaptive equalizer in a case where data transmission is performed mainly on a short-wave line through a transmission line with much fading. The present invention relates to an adaptive equalization method for demodulating a signal and a diversity reception method.

[0002]

2. Description of the Related Art Hitherto, diversity has been known as a technique for securing good reception quality in a short-wave line or the like in which transmission lines are often affected by fading due to ionospheric reflection and the like. Specific methods for realizing diversity include frequency diversity and space diversity. Such a diversity technique is effective when it is necessary to appropriately select and receive a signal obtained by modulating a large number of transmission waves in a broadcasting format in the short wave band with the same data according to the current line condition.

On the other hand, various digital data transmission systems have been put into practical use, and recently, in order to improve transmission efficiency, a multi-level digital modulation system starting with the 16QAM system has been developed and put into practical use. The multi-level digital modulation system is capable of high-efficiency information transmission, but is extremely vulnerable to various types of distortion due to fading in the transmission system.
Countermeasures against fading in the transmission path become serious.

As a countermeasure, adaptive equalization processing by an adaptive equalizer (transversal type equalizer or the like) is known as an effective means. At present, high-efficiency transmission systems such as various digital modulation systems are used. It is getting ready.

FIG. 1 is a block diagram of a receiving system using a diversity receiving apparatus. In the drawing, reference numerals 1a and 1b denote antennas, 2a and 2b denote receivers connected to the antennas, and 3 denotes a receiver from both receivers 2a and 2b. The comparison circuit 4 compares the output of the electric field strength of the receiver 2.
A switch circuit for selecting one of the output signals a and 2b, 5 is an adaptive equalizer, and 6 is a demodulator for digital modulation.

In the above, the reception waves received by the antennas 1a and 1b are subjected to reception processing such as high-frequency amplification, frequency conversion and intermediate frequency amplification by the receivers 2a and 2b, respectively, and the output signal is sent to the switch circuit 4. Sent. The reception electric field strength of each reception wave is detected by the receivers 2a and 2b, and the detection signal is sent to the comparison circuit 3. The comparison circuit 3 compares and determines the electric field strength detection signals of the respective received waves, and sends a signal corresponding to the result of the determination to the switch 4. The switch 4 selects the received wave having the stronger electric field strength according to the output signal of the comparison circuit 3 and sends the selected received wave to the adaptive equalizer 5.

In general, a transmission path is accompanied by distortion due to fading or the like in a short wave or the like, and contains a large delay distortion for data transmission. This causes intersymbol interference in a demodulated baseband signal. To compensate for this, it is necessary to perform equalization using an adaptive equalizer. A typical example of the adaptive equalizer is a transversal type equalizer applied to a baseband signal on the receiving side. This adaptive equalizer can automatically adjust the coefficient of the equalizer using the received signal when the fluctuation of the characteristics of the transmission path is relatively small, but when the fluctuation of the transmission path is large, Requires that the tap gain be readjusted using a training signal or the like.

FIG. 2 shows an example of a decision feedback adaptive equalizer (DFE) using two trans-bazaar type equalizers.
This decision feedback adaptive equalizer includes an equalizer unit and a control unit. As shown in the figure, the registers 7-1f to 7-jf synthesize future data as viewed from the center tap Fo (t). 7-1a to 7-ka are registers for synthesizing past data, and multipliers 8-1f to 8-jf
Is a tap coefficient F-1 for synthesizing future data.
(T) to Fj (t) and input signals y (t + T) to y (t
+ JT), multipliers 8-1a to 8-1
−ka is a tap coefficient B ₁ for synthesizing past data.
A multiplier for multiplying (t) to B _k (t) by the number of outputs 1 of the determination circuit 10 or the number of outputs 2 from the training signal generator 13, ie, the number of reference signals 3, and the adder 9 is a multiplier 8- The equalizer section is configured by an addition amplifier that adds and amplifies the outputs of 1f to 8-jf and 8-1a to 8-ka to obtain an equalized output. Multipliers 8-1f to 8-jf, 8-
The control unit automatically controls the tap coefficients of 1a to 8-ka. In the configuration of the control unit, a determination circuit 10 determines the equalized output Z (t) based on a determination value of an ideal value. The adder (subtractor) 11 calculates an estimation error e (t), and subtracts the output Z (t) from the equalizer from the number of outputs 1 of the decision circuit 10,
The result 4 of the subtraction result is output. Tap coefficient update control circuit 1
2 are tap coefficients Fo (t) to Fj (t), B such that the mean square number 5 of the estimation error e (t) is minimized.
₁ (t) to B _k (t) are updated.

[0009]

(Equation 1)

[0010]

(Equation 2)

[0011]

(Equation 3)

[0012]

(Equation 4)

[0013]

(Equation 5)

In the adaptive equalization processing, for example, when delay distortion occurs due to a transmission path, when the direct wave is larger than the delay wave (minimum phase condition), the equalizer performs the direct wave component of the input signal y (t). It operates so that only the direct wave component of y (t−T) cancels out the delayed wave component of y (t). Hereinafter, the equalization operation is performed by increasing the number of feedback taps and sequentially canceling them. Conversely, when the delayed wave is larger than the direct wave (non-minimum phase condition), only the delayed wave component of y (t + T) is extracted and the direct wave component of y (t−T) is extracted by the delayed wave component of y (t + 2T). It works to cancel out.
In the same manner, the equalization operation is performed by increasing the number of feed forward taps and sequentially canceling them.

In the tap coefficient update control circuit 12, the Kalman algorithm or the recursive least squares algorithm (RL)
The tap coefficient is automatically updated by the S algorithm or the like. The equalizing operation using such an algorithm can be divided into an initial pull-in process for initializing tap gains and a tracking process for updating the initialized tap coefficients in accordance with channel fluctuations.

Here, a method based on the Kalman algorithm will be described. Equalization output Z (t) at time t = hTs (Ts; symbol rate), estimation error e (t),
Tap input vector number 6 is Zn, en, number 7,
Equation 8 and tap coefficient vector equation 9

[0017]

(Equation 6)

[0018]

(Equation 7)

[0019]

(Equation 8)

[0020]

(Equation 9)

Assuming Equations 10, 11, and 12,

[0022]

(Equation 10)

[0023]

[Equation 11]

[0024]

(Equation 12)

The equalization output Zn and the estimation error en are given by
3, given by Equation 14.

[0026]

(Equation 13)

[0027]

[Equation 14]

Here, Equation 15 is obtained,

[0029]

(Equation 15)

Updating of the tap coefficient is performed according to equations (16), (17) and (18).

[0031]

(Equation 16)

[0032]

[Equation 17]

[0033]

(Equation 18)

Here, Equation 19 is the transposed conjugate of Equation 20, Equation 21
Is the Kalman gain, Equation 22 is the error covariance matrix of Equation 23, u
Is the variance of Equation 23, and λ is the forgetting factor (0 <λ ≦ 1).

[0035]

[Equation 19]

[0036]

(Equation 20)

[0037]

(Equation 21)

[0038]

(Equation 22)

[0039]

(Equation 23)

At the beginning of each burst in a TDMA system or the like, a training sequence is received, and the tap sequence is converged to an appropriate value using the training sequence. That is, at the start of the burst, the difference between the received training sequence and the known training sequence from the training signal generator 13 is used as the estimation error en to converge the tap coefficient to an appropriate value. Thereafter, while reproducing the data, the difference between the received signal and the determination value is used as an estimation error en so that the tap coefficient follows the variation of the transmission path.

FIG. 7 shows a conventional method 1 for estimating the optimum tap coefficient, and FIG. 8 shows a conventional method 2.

In the method according to the conventional method 1, equalization is performed by the Kalman algorithm both during training and during tracking. According to this method, the data section uses the Kalman algorithm which requires a large amount of calculation.

In the method according to the conventional method 2, the Kalman algorithm is used at the time of training (preamble and postamble), and linear interpolation is performed without performing equalization in the data section. That is, the gain c (k) at the k-th symbol in the data portion is given by Expression 24.

[0044]

(Equation 24)

In the conventional method 2, the amount of calculation is the same as that of the conventional method 1.
The transmission efficiency is reduced due to the redundancy of the postamble.

FIG. 9 shows a main routine for equalization in the second conventional example.

[0047]

As described above, conventionally, equalization is performed by the Kalman algorithm at the time of training and at the time of tracking. However, at the time of training operation, the transmission path is unknown or the fluctuation of the transmission path is large. Therefore, it is necessary to set the initial value of the tap coefficient in a short convergence time.

In the PLS algorithm or the Kalman algorithm, for example, there is a difference in the number of repetitions about 10 times that the root mean value of the tap coefficient converges as compared with the least mean square (LMS) algorithm. The algorithm has a slower convergence speed. For this reason,
It is preferable to perform estimation (equalization processing) for setting the initial value of the tap coefficient by the Kalman algorithm during training.

However, in the conventional adaptive equalization method,
Even when a data (information) signal is received, since the equalizing operation is performed by the RLS algorithm or the Kalman algorithm, which requires a large amount of calculation, the amount of calculation is always large.

Assuming that the number of tap coefficients to be estimated is n,
The number of multiplications required to obtain the estimated value is 25 times by the Kalman algorithm and 26 times by the RLS algorithm. On the other hand, the number is 27 times in the LMS algorithm, and the difference is large.

[0051]

(Equation 25)

[0052]

(Equation 26)

[0053]

[Equation 27]

Conventionally, there have been proposed about three methods for reducing the amount of calculation in the data portion.

The first is a method for removing the redundancy of the operation in the Kalman algorithm. As an example, there is a high-speed Kalman algorithm. According to this method, when the number of taps is a fractional interval of 20, the amount of calculation can be reduced to about 1/2.

Second, the burst signal is composed of a preamble portion, a data portion, and a postamble portion, and forward (forward) equalization, reverse (rearward) equalization, and twice equalization are performed. Is determined by judging which is smaller. In this case, since the equalization direction is reversed, the non-minimum phase condition can be set as the minimum phase condition, so that the number of feedforward taps can be reduced. When the required number of taps becomes 1 / L, the calculation amount is about 2 / L.
L ² become.

The third is a method of obtaining an optimal tap coefficient using a preamble signal and a postamble signal, and performing linear interpolation without performing equalization processing on a data portion. In other words, the tap coefficients subjected to the equalization processing at the time of the preamble signal and the tap coefficients subjected to the equalization processing at the time of the postamble signal are interpolated by a linear function. According to this method, the amount of calculation can be reduced to about 1/7.

However, in the first method, the effect of reducing the amount of calculation is as low as about 1/2, and in the second and third methods, a preamble or postamble portion must be provided for each burst. There is a problem that transmission efficiency is reduced. In addition, in the second method, equalization is performed twice and the smaller equalization error is determined, so that the processing becomes slightly complicated. In the third method, linear interpolation is performed.
A disadvantage arises in that the fading pitch, such as a short-wave band line, cannot follow the fluctuation of the transmission line of about several Hz.

As described above, conventionally, when an information signal is received, a large amount of calculation is required by the Kalman algorithm or the like in the equalization processing for causing the tap coefficients of the equalizer to follow the fluctuation of the transmission path, and the training sequence is not required. Therefore, providing a preamble portion and a postamble portion has a problem that transmission efficiency is reduced.

It is an object of the present invention to provide an effective adaptive equalization method that achieves an improvement in transmission efficiency and a reduction in the amount of computation by equalization processing in view of the above-mentioned problems of the prior art.

[0061]

SUMMARY OF THE INVENTION It is an object of the present invention to automatically adjust an initial value of each coefficient of an adaptive equalizer by a Kalman algorithm for each fixed code period when receiving a training signal of a received wave. Sometimes, this is achieved by an adaptive equalization method in which each coefficient of the adaptive equalizer is automatically adjusted by a least mean square algorithm.

[0062] The above-mentioned object is also aimed at receiving waves which are different from each other at two or more frequencies in the short wave band or spatially different from each other and which are constantly transmitted as data transmission signals modulated by the same data. A diversity receiving apparatus that selects and receives the optimal one, and automatically adjusts an initial value of each coefficient by a Kalman algorithm for each fixed code period when receiving a training signal of the received wave, and when receiving an information signal, An adaptive equalizer that performs adaptive equalization processing by automatically adjusting coefficients by a least mean square algorithm, and demodulates a multi-level digital modulation wave of the adaptive equalized output and outputs a demodulated output including an information signal and a training signal. This is achieved by a diversity receiving system including a demodulation circuit obtained.

The above object is also achieved by estimating the characteristics of the transmission path of the received wave and estimating the transmission signal using the Kalman algorithm, and performing the estimation of the characteristics of the transmission path and the estimation of the transmission signal alternately. This is achieved by an adaptive equalization scheme that performs equalization and information signal estimation.

The object of the present invention is to provide a receiving wave that is different from two or more frequencies in the short wave band or spatially different from each other, and the different receiving waves are modulated by the same data and are always transmitted as data transmission signals. A diversity receiving apparatus that selects and receives the best one, and estimates the characteristics of the transmission path of the received wave and estimates the transmission signal using the Kalman algorithm, and estimates the characteristics of the transmission path and estimates the transmission signal. And a demodulation circuit for demodulating the multi-level digital modulated wave of the adaptive equalized output to obtain a demodulated output of the information signal. Achieved by the receiving scheme.

According to the above means, when the training signal of the received wave is received, the convergence speed is reduced by using the Kalman algorithm to set the initial value of the tap coefficient as the estimation error using the difference between the received signal and the known training sequence. It is faster and transmission efficiency can be increased. Also, when receiving an information signal, the amount of calculation can be reduced by the LMS algorithm, and the tap coefficient can be updated while following the fluctuation of the transmission path while reproducing data.

Further, according to the other means, equalization can be performed without using a training signal, a transmission signal can be estimated, and transmission efficiency can be improved.

[0067]

Embodiments of the present invention will be described below in detail.

According to the first embodiment of the present invention, in a receiver which performs spatial or frequency diversity processing in a short-wave band and receives a signal, an equalizer is used by using a training signal to estimate a transmission path. At the time of initial pull-in (training mode), equalization is performed by the Kalman algorithm or the RLS algorithm for converging tap values as quickly as possible from the viewpoint of transmission efficiency.

In the tracking mode, the difference between the received signal and its determination value is set as an estimation error en by the LMS algorithm while the data (information bits) are reproduced, and the tap coefficient follows the fluctuation of the transmission path.

An LSM algorithm used as an adaptive algorithm in such a tracking mode will be described below.

As in the case of the Kalman algorithm, the time t =
equalized output Z at nT _s (T _s ; symbol rate)
(T), the estimation error e (t), and the number of tap input vectors 6 are Zn, en, Equation 7, Equation 8, and the number of tap coefficient vectors 9 are Equation 10, Equation 11, and Equation 12, respectively.
The n and the estimation error en are calculated by Expressions 13, 14, and 15, and the update of the tap coefficient is performed by Expression 28.

[0072]

[Equation 28]

Assuming that the number of tap coefficients is n,
The number of times of multiplication necessary to obtain the estimated value is as shown in Expression 27, which is extremely smaller than that of Expression 25 of the Kalman algorithm.

FIG. 3 shows a method according to the invention for estimating the optimal tap coefficients. In the method according to the present invention, the transmission efficiency is the same as that of the conventional method 1 shown in FIG. Variations can be followed, and good convergence can be achieved even when the maximum multipath delay is about 3 ms (data length is about 100 ms) as in HF propagation.

FIG. 4 shows a main routine for equalization according to the present invention.

The second embodiment of the present invention uses a Kalman algorithm without using a training signal to estimate a transmission path in a receiver that performs spatial or frequency diversity processing in a short wave band and receives a signal. The estimation of the characteristics of the transmission line and the estimation of the transmission signal are performed, and the estimation of the characteristics of the transmission line and the estimation of the transmission signal are performed alternately, thereby performing the equalization and the estimation of the information signal.

FIG. 5 is a block diagram of a decision feedback adaptive equalizer (DFE) that does not use a training signal. The training signal generator 13 in the configuration of FIG. 1 is unnecessary. FIG. 6 shows a detailed configuration diagram of the determination circuit 14. The tap coefficient update control circuit 15 performs update control of the tap coefficient based on the estimation of the characteristics of the transmission path by the determination circuit 14.

Expression 16 in Expression 10 is rewritten as y (t), and the order of feedback taps 1 to k is changed to 1 to n.
, The impulse response of the transmission path is defined as hi (i = 1, 2,...,
n), the transmission signal and the reception signal are u (t) and y, respectively.
Assuming (t), Equation 29 is obtained.

[0079]

(Equation 29)

U (t) is generally random noise independent of u (t).

Estimation of the transmission path is performed by estimating the number of transmission signals by 30.
Is the number 31,

[0082]

[Equation 30]

[0083]

(Equation 31)

The new vector numbers 32 and 33 are converted to
Assuming Equation 35,

[0085]

(Equation 32)

[0086]

[Equation 33]

[0087]

(Equation 34)

[0088]

(Equation 35)

The estimation of the characteristics of the transmission path is performed by the parameter estimating section 17 of the decision circuit. The estimated value number 36 of the transmission signal output from the state estimation unit 16 is delayed by the delay unit 19 and input.

[0090]

[Equation 36]

On the other hand, the input signal y (t) to the determination circuit is delayed by the delay unit 18 and input to the parameter estimation unit 17. Based on these inputs, the characteristics of the transmission path are estimated in the following order. Number 37

[0092]

(37)

The estimated value of the tap coefficient is given by Expression 38, and the tap coefficient is updated accordingly.

[0094]

(38)

In this case, the estimation error is given by Equation 39.

[0096]

[Equation 39]

The above equation (40) is estimated and input by the state estimating unit 16.

Further, Equation 41 of Equation 38 is the Kalman gain, and becomes Equation 42, Equation 43, and Equation 44.

[0099]

(Equation 40)

[0100]

[Equation 41]

[0101]

(Equation 42)

[0102]

[Equation 43]

[0103]

[Equation 44]

Next, the estimation of the transmission signal is calculated by the state estimator 16 as Equations 45, 46, 47 and 48.

[0105]

[Equation 45]

[0106]

[Equation 46]

[0107]

[Equation 47]

[0108]

[Equation 48]

However, Equation 49 is the Kalman gain, which is Equation 50, Equation 51, Equation 52, Equation 53, and Equation 54.

[0110]

[Equation 49]

[0111]

[Equation 50]

[0112]

(Equation 51)

[0113]

(Equation 52)

[0114]

(Equation 53)

[0115]

(Equation 54)

The decision circuit 14 alternately repeats the above estimation of the characteristics of the transmission path and the estimation of the transmission signal to perform equalization and information signal estimation.

The impulse response of the transmission path is given an initial value in advance. As the initial value, an abbreviated value determined by the position, time, season, frequency, and the like of the transmitting station and the receiving station is stored in the memory 20 in advance, selected, read out at regular intervals, and set as the initial value. By inputting, the efficiency and accuracy of estimation can be improved.

[0118]

As described above, according to the present invention, in the training mode, the tap value can be quickly converged by the Kalman algorithm to improve the transmission efficiency. In the tracking mode, the calculation amount can be reduced by the LMS algorithm. It is relatively easy to follow the fluctuation of the transmission path,
Equalization that can withstand fluctuations and delays in the F band can be performed. Further, equalization can be performed without using a training signal, and a transmission signal can be estimated using a Kalman algorithm, so that transmission efficiency can be improved.

[Brief description of the drawings]

FIG. 1 is a block diagram of a diversity receiver,

FIG. 2 is a configuration diagram of a decision feedback adaptive equalizer.

FIG. 3 is an explanatory diagram of estimating tap coefficients according to the first embodiment of the present invention.

FIG. 4 is a processing diagram of a main routine for equalization according to the first embodiment of the present invention.

FIG. 5 is a configuration diagram of a decision feedback adaptive equalizer according to a second embodiment of the present invention.

FIG. 6 is a detailed configuration diagram of a part of the circuit in FIG. 5;

FIG. 7 is an explanatory diagram of tap coefficient estimation in Conventional Method 1.

FIG. 8 is an explanatory diagram of tap coefficient estimation in Conventional Method 2.

FIG. 9 is a processing diagram of a main routine of conventional equalization.

[Explanation of symbols]

1a, 1b antenna, 2a, 2b receiver, 3 comparison circuit, 4 switch, 5 adaptive automatic equalizer, 6 demodulator for digital modulation, 7-jf to 7-1f, 7-1a to 7-
ka... register, 8-jf to 8-1f, 8-1a to 8-
ka: Multiplier, 9: Adder, 10: Judgment circuit, 11: Adder, 12: Tap coefficient update control circuit, 13: Training signal generator, 14: Judgment circuit, 15: Tap coefficient update control circuit, 16 ... State estimator (transmission signal estimation), 1
7, a parameter estimating unit (estimating the characteristics of the transmission path), 18,
19 ... Delay section.

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) H03H 21/00 H03H 21/00 H04B 7/005 H04B 7/005 7/08 7/08 A 7/12 7 / 12 H04L 27/38 H04L 27/00 G 27/01 K F term (reference) 5J021 AA02 AA03 AA04 AA05 AA06 CA06 DB02 DB03 DB04 EA04 FA17 FA20 FA26 FA32 GA08 HA05 HA06 5J023 DA03 DB03 DC06 DD09 5K004 AA08 JH02 5K046 A05 EE56 EF02 EF13 EF15 EF23 EF46 5K059 AA08 CC03 CC06 DD01 DD39 EE02

Claims (5)

[Claims] In an adaptive equalization system for adaptively equalizing a received wave using an adaptive equalizer, an initial value of each coefficient of the adaptive equalizer is set to a predetermined value when a training signal of the received wave is received. An adaptive equalization method characterized by automatically adjusting by a Kalman algorithm for each period and automatically adjusting each coefficient of the adaptive equalizer by a least mean square algorithm when an information signal is received. 2. An optimal radio wave which is received at two or more frequencies in a short wave band or spatially different from each other, and wherein the different received waves are modulated with the same data and constantly transmitted as a data transmission signal. And a diversity receiving apparatus for receiving the training signal of the received wave, automatically adjusting the initial value of each coefficient by the Kalman algorithm for each fixed code period, and setting each coefficient to at least 2 when receiving the information signal. An adaptive equalizer that performs an adaptive equalization process by automatically adjusting according to a root-mean-square algorithm; and a demodulation circuit that demodulates a multilevel digital modulation wave of the adaptive equalized output to obtain a demodulated output including an information signal and a training signal. A diversity receiving method characterized by comprising: 3. An adaptive equalization method for performing an adaptive equalization process on a received wave using an adaptive equalizer, wherein the adaptive equalization estimates a transmission path characteristic of the received wave using a Kalman algorithm and transmits the transmission signal. And estimating the information signal by alternately performing the estimation of the characteristics of the transmission path and the estimation of the transmission signal. 4. An optimal one of radio waves which are received at two or more frequencies in the short wave band or spatially different from each other and are always transmitted as data transmission signals after the different received waves are modulated with the same data. And a diversity receiver for selecting and receiving, and using the Kalman algorithm to estimate the characteristics of the transmission path of the received wave and estimate the transmission signal, and alternately perform the estimation of the characteristics of the transmission path and the estimation of the transmission signal. An adaptive equalizer that performs equalization and information signal estimation by performing
And a demodulation circuit for demodulating the multi-level digital modulated wave of the adaptive equalized output to obtain a demodulated output of the information signal. 5. An adaptive equalizer for estimating the characteristics of a transmission path of a received wave using a Kalman algorithm, wherein a coefficient initial value of a characteristic of a transmission path determined by a positional relationship between a transmitting station and a receiving station, date and time. An adaptive equalizer characterized by comprising a memory for storing in advance and providing the stored initial value as an initial value of a readout coefficient at regular intervals.