JP2544527B2 - Equalizer - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an equalizer for removing distortion generated in a transmission line by forming a filter having an inverse characteristic of the transmission characteristic of the transmission line.

[0002]

2. Description of the Related Art An equalizer removes distortion generated in a transmission line from a received signal. When this distortion dynamically changes, internal parameters that determine the transfer characteristic of the equalizer are changed. It is necessary to follow up and update sequentially. For example, when performing high-speed digital mobile communication of several tens of kb / s or more, an equalizer is used to improve delay distortion due to multipath propagation, that is, deterioration due to frequency selective fading. Since the characteristics greatly change as the radio moves, it is necessary to adaptively follow the internal parameter that determines the transfer characteristics of the equalizer, that is, the tap coefficient.

Here, the structure of the transmission system will be described with reference to FIG. FIG. 7 (a) shows the structure of the transmitter. The transmission signal to be transmitted is configured by the frame configuration circuit 61 as a burst to which a training signal is added, as shown in FIG. 8, and a plurality of N of these bursts are collected and configured as a frame. The output of the frame configuration circuit 61 is band-limited by the baseband modulation signal generation circuit 62, converted to a radio frequency band by the quadrature modulator 63, and transmitted.

FIG. 7B shows the structure of the receiver. The received signal has its level fluctuation suppressed by an automatic gain control amplifier (AGC) 65, and then converted into a baseband analog signal by a quadrature detector 66. This signal is converted into a digital signal by the analog / digital converter (A / D) 67 and input to the equalizer 68.

The equalizer 68 initializes internal parameters (training process) using the training signal read from the received signal memory and the reference signal held inside the equalizer. After that, the signal obtained by removing the distortion from the transmission signal read from the reception signal memory is output, and the internal parameters are sequentially updated according to the variation of the transfer characteristic of the transmission line (equalization process). At this time, a signal whose equalized output is determined is used as the reference signal. In the following description, a method of initializing and updating internal parameters is called an adaptive algorithm.

FIG. 9 shows an RLS (Recurs-ive Least Squares) as an adaptive algorithm for a QPSK modulated signal.
It shows the phase of the equalized output signal when using the algorithm, and the transition of the signal obtained by projecting the equalized output signal on the Q axis, for example, on the time axis. FIG. 9A shows the case where the fluctuation of the transmission path is small. At this time, it can be seen that the signal phase is held at that of QPSK from the training processing section in which the internal parameters are initialized to the equalization processing section, and good tracking characteristics are obtained. Figure 9 (b)
Is when the transmission line fluctuation is large. At this time, although it is good at the beginning of the equalization processing section, it becomes difficult to follow the fluctuation of the transmission line, and the signal phase deviates from that of QPSK. This is because the RLS algorithm has an excellent tracking characteristic in the time domain immediately after the end of the training processing section, but the tracking characteristic deteriorates rapidly with the passage of time.

The RLS algorithm is known,
For example, the document "SIMON HAYKIN:" Adapt-ive Filtering Theo
ry ", Prentice-Hall, 1986, chapter 8," Adaptive Transve
rsalFilter using Recursive Least Squares ", pp.381-
450 ”.

[0008]

As described above, when the equalizer is applied to a transmission line having a large fluctuation, the adaptive algorithm sometimes cannot follow the fluctuation of the transmission line. Therefore, in the conventional equalizer, the burst length is shortened as shown in FIG. 10 (b) or the burst length shown in FIG. 10 (c) is added to the standard burst structure shown in FIG. 10 (a). As described above, there is a method of inserting a plurality of training signals in one burst to shorten the length of continuous use of the adaptive algorithm. However, with this method, although the effect of improving the tracking characteristic can be expected, there is a problem that the transmission efficiency decreases due to the relatively short transmission signal sequence.

In addition to the conventional RLS algorithm, the exponentially weighted RLS algorithm in which a forgetting factor is introduced can improve the tracking characteristic, but there is a drawback that the operation becomes unstable. It is an object of the present invention to provide an equalizer that maintains a predetermined transmission efficiency and has improved tracking characteristics with respect to variations in transmission characteristics of a transmission line.

[0010]

According to the present invention, a received signal memory for storing a received signal, a training signal is read from the received signal stored in the received signal memory, and internal parameters are set to predetermined initial values. An equalizer including an equalization processing unit that subsequently reads out a transmission signal and performs equalization processing, and an equalization output memory that stores the equalization output signal output by the equalization processing unit. When the value representing the quality of the value is below a predetermined value, the equalization process is interrupted, the internal parameter is set to a predetermined initial value, and a reference formed based on the equalized output signal before that time point is set. It is characterized by including a control unit for updating the initial value of the internal parameter to an optimum value by using a signal.

[0011]

The equalizer of the present invention operates according to the following operating procedure. The received signal of one burst is stored in the received signal memory. Initialize internal parameters from the received training signal and the reference signal that is held in the equalizer in advance (training process), and then output the signal with distortion removed while sequentially updating the internal parameters (etc. Process). At the same time, the equalized output signal is stored in the equalized output memory. While performing the equalization process, the reliability of the equalized output signal is monitored. When the reliability of the equalization output signal decreases, the equalization process is temporarily stopped and the internal parameter is preset to a value required to start the training process. The received signal read once is read again, the equalized output signal stored in the equalized output memory is read as a reference signal to reinitialize internal parameters, and the equalization process is subsequently continued. The above operation is repeated until the end of the burst.

As the reference signal, a signal suitable for the training operation is selected from the signals stored in the equalized output memory and used. In the following description, the operation of reinitializing internal parameters according to the present invention is called feedback learning.

[0013]

1 is a block diagram showing the configuration of an embodiment of the equalizer of the present invention. In the figure, the quadrature detector (see FIGS.
The I-channel and Q-channel received signals output from 6) are input as digital signals via an analog / digital converter (67 in FIG. 8). This received signal is input and stored in the received signal memory 11. The equalization processing unit 13 reads a training signal from the signal stored in the reception signal memory 11, operates an adaptive algorithm to set initial values of internal parameters, and subsequently reads a transmission signal from the reception signal memory 11 and the like. The equalization process is performed, and the result is written in the equalization output memory 15. The equalization output memory 15 accumulates a predetermined amount of equalization output signal and outputs it as received data. The reception electric field intensity memory 12 stores the electric field intensity at the moment when the reception signal accumulated in the reception signal memory 11 is received, in association with the reception signal. As a means for measuring the electric field strength, it is not necessary to add a new circuit if the control voltage of the automatic gain control amplifier (FIG. 8, 65) is used.

The control section 17 judges from the received electric field strength read from the received electric field strength memory 12, the equalization error obtained from the equalization processing section 13, and the code sequence read from the equalization output memory 15, and the feedback learning is performed. The control signal for feedback learning is sent to the reception signal memory 11, the equalization processing unit 13, and the equalization output memory 15 only when the above is required. At this time, the equalization output memory 15 sends a known signal sequence for feedback learning to the equalization processing unit 13.

A specific operation example of feedback learning will be described below with reference to FIGS. FIG. 2 shows an example in which an equalization error is used as a means for judging the reliability of the signal stored in the reception signal memory 11. Here, since the equalization error in the equalization processing is smaller than the feedback learning start level, feedback learning is not performed. The specific feedback learning start level may be determined as follows. If the adaptive algorithm cannot follow the variation of the transfer characteristic of the transmission path, no code error occurs first, but the equalization error increases. If it cannot follow further, the phase of the signal exceeds the decision level and a code error occurs. Therefore, the value of the equalization error that causes a code error may be checked in advance under the transmission path condition to which the equalizer is applied, and the magnitude thereof may be set as the feedback learning start level.

FIG. 3 shows an example in which feedback learning is actually performed when the equalization error is used as a means for judging the reliability of the signal stored in the reception signal memory 11. Here, first, internal parameters are set using the training signal included in the transmission signal, and the first equalization process indicated by is performed. The control unit 17 monitors the equalization error, and temporarily suspends the equalization process when the equalization error exceeds the feedback learning start level. Next, after presetting the internal parameter of the equalizer to a predetermined value necessary for starting the training operation, the signal already read from the received signal memory 11 and equalized is read again,
The internal parameters are re-initialized, and the second equalization process indicated by is performed. As the reference signal necessary for this training operation, the signal accumulated in the equalized output memory 15 is used as a correct sequence (Japanese Patent Application No. 2-39491).

The portion shown by (A) in the figure is a section in which the equalization error rapidly increases, and the reliability of the signal is low, so it is not used as a reference signal for feedback learning. In addition, the same symbol may continuously follow the portion indicated by (A) in the figure. In general,
It is known that a signal sequence in which the same code continues is not suitable for a training signal, and the operation of the RLS algorithm may become unstable and diverge. Therefore, also in this case, this portion is not used as a reference signal for feedback learning.

Further, in this example, the feedback learning is performed only once, but if necessary, it may be performed many times in one burst. Furthermore, when the RLS algorithm is used as the adaptive algorithm, the internal parameters include the correlation coefficient, the gain vector, and the like in addition to the tap coefficient. However, when the equalization process is interrupted once and the internal parameters are preset to predetermined values, a method of presetting all of them and a correlation matrix that determines the tracking characteristic are preset, and the remaining internal parameters are equalized. There is a method to leave it as it was when it was interrupted.

FIG. 4 shows an example in which the received electric field strength is used as a means for judging the reliability of the signal stored in the received signal memory 11. It is known that under the Rayleigh fading condition, which is a characteristic of mobile communication transmission lines, the state of the transmission line fluctuates greatly at the moment when the received electric field strength greatly decreases. Therefore, the control unit 17 monitors the received electric field strength,
When the received electric field strength falls below the feedback learning start level, the equalization process is temporarily stopped. Next, the feedback learning is operated so that the portion where the received electric field strength is reduced is included in the time region where the tracking characteristic is excellent immediately after the training is completed. Even in this case, the portion indicated by (A) in the figure is not used as a reference signal for feedback learning because the reliability of the signal is low.

FIG. 5 shows a simplified algorithm for feedback learning. When the moving speed of the mobile device is known, the feedback learning operation can be performed at regular intervals without using the means for evaluating the quality of the equalization process. However, this method may use an incorrect equalized output signal as the reference signal. In that case, it is effective to perform error correction on the equalized output signal and then write it to the equalized output memory 15 (Japanese Patent Application No. 2-
39491).

Here, FIG. 6 shows the effect of improving the error rate characteristic according to the present invention by an experiment using a transmission system.
The error rate characteristics with respect to the transmission path state E _b / N ₀ [dB] are the case where the dotted line is the case without feedback learning, the thick solid line is the case where feedback learning is performed, and the thin solid line is the feedback by the simplified algorithm shown in FIG. This is the case when learning is performed.

As shown in the figure, according to the present invention, the total E _b / N _{0 is}
It can be seen that the effect of feedback learning appears in the area. However, it has been shown that when feedback learning using a simplified algorithm is used when the noise in the transmission line is large and there are many errors, the error rate characteristic deteriorates because the wrong equalized output signal is used as the reference signal. .

[0023]

As described above, according to the present invention, when it is detected that the adaptive algorithm cannot follow the variation of the transfer characteristic of the transmission line, the equalization process is temporarily suspended and the internal parameter is reset. repeat. Therefore, it is possible to obtain a good tracking characteristic even in a transmission line in which the variation of the transfer characteristic is large.

Further, the reference signal required for resetting the internal parameters is equalized and the signal sequence stored in the equalized output memory is extracted and used. Therefore, it is not necessary to transmit a training signal separately. However, it does not reduce the transmission efficiency. Furthermore, by monitoring the equalized output error, the received electric field strength, or the code sequence of the transmission signal, the feedback learning is executed only when the internal parameters need to be reset and the effect can be expected. A high improvement effect can be obtained.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of an equalizer of the present invention.

FIG. 2 is a diagram illustrating an operation of feedback learning based on an equalization error.

FIG. 3 is a diagram illustrating an operation of feedback learning based on an equalization error.

FIG. 4 is a diagram illustrating an operation of feedback learning based on a received electric field strength.

FIG. 5 is a diagram illustrating a feedback learning operation using a simplified algorithm.

FIG. 6 is a diagram for explaining the effect of improving the error rate characteristic according to the present invention by an experiment using a transmission system.

FIG. 7 is a block diagram showing a configuration of a transmission system.

FIG. 8 is a diagram showing a frame structure.

FIG. 9 shows a phase of an equalized output signal and a transition of a signal obtained by projecting the equalized output signal on, for example, the Q axis on a time axis when the RLS algorithm is used as an adaptive algorithm for a QPSK modulated signal. It is a figure.

FIG. 10 is a diagram showing an improvement example in the related art.

[Explanation of symbols]

 11 Received Signal Memory 12 Received Electric Field Strength Memory 13 Equalization Processing Unit 15 Equalization Output Memory 17 Control Unit 61 Frame Configuration Circuit 62 Baseband Modulation Signal Generation Circuit 63 Quadrature Modulator 65 Automatic Gain Control Amplifier (AGC) 66 Quadrature Detector 67 Analog / Digital Converter (A / D) 68 Equalizer

 ─────────────────────────────────────────────────── ─── Continuation of front page (56) References JP-A-4-90614 (JP, A) JP-A-4-211525 (JP, A) JP-B 3-74061 (JP, B2) JP-B 4- 49294 (JP, B2) US Patent 5297165 (US, A)

Claims (1)

(57) [Claims] 1. A reception signal memory for storing a reception signal, a training signal is read from the reception signal stored in the reception signal memory to set internal parameters to predetermined initial values, and then a transmission signal is read. In an equalizer including an equalization processing unit that performs equalization processing and an equalization output memory that stores the equalization output signal output by the equalization processing unit, a value representing the quality of the equalization processing is a predetermined value. when below the, the internal parameter set to a predetermined initial value to interrupt the equalizing processing, the time before from the equalization output memory
An equalizer comprising: a control unit that reads out the equalized output signal of 1) as a reference signal and updates the initial value of the internal parameter to an optimum value.