JPH01194614A - Automatic equalizer - Google Patents

Abstract

PURPOSE:To constantly attain the equalization at high-speed convergence even in the various types of line distortion without decreasing equalizing accuracy by switching the convergent coefficient of a switching means according to an equalizing output error from a comparing means, and changing a convergent coefficient alpha in stages according to the size of the equalizing error. CONSTITUTION:A square error signal ek<2> is calculated out of an error signal ek from an error signal calculating part 201 by a signal squaring part 202. A threshold from a threshold part 206 to set the threshold set beforehand is compared with the square error signal ek<2> from a signal squaring part 202 by a comparing part 203. The convergent coefficient alpha is switching by a selector 204 with the use of an output from the comparing part 203. In accordance with the size of the square error signal ek<2>, by switching a convergent speed as alpha[0]>alpha[1],..., >alpha[n], the convergent speed can be made faster. In such a way, by providing the means to successively make the convergent coefficient alphaof the equalizer smaller in stages according to the error signal, the equalization convergent speed and an equalizing performance can be enhanced.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a modem device (hereinafter referred to as a modem) used for data communication.
This is related to other domains, such as automatic machines prepared for this purpose.

[Prior Art] When transmitting digital signal data via a general public line that is an analog line, it is necessary to modulate the digital signal and convert it into a desired analog signal before transmitting it. Further, on the receiving side, it is necessary to demodulate this modulated signal, and a modulation/demodulation device (modem) for this purpose is essential.

The transmission speed during data transmission has also been increased, and currently GM facsimile machines have an information transmission speed of 9.
A type that transmits data at a high speed of 600 bps (bit/sec) is used.

Therefore, the transmission signal modulated by the transmitting modem and sent out to the line is received by the receiving modem with distortion added due to line distortion, jitter, timing error between transmission and reception, carrier error, etc. The receiving modem is equipped with a signal generator to correct this distortion, and when the signal is output from the receiving modem, it is corrected so that it becomes the original transmission signal.

In the equalization operation, prior to data transmission, mutual synchronization processing is normally performed between the transmitting and receiving modems using known synchronization data (training data). As a link to this, the inverse characteristics of the line are created (trained) in advance by other domains. Then, the equalization characteristics are changed to follow the gradual time fluctuations of the line (automatic equalization or adaptive equalization).

The operations of these other domains will be explained below with reference to FIG. 4 and FIGS. 5(A) to 5(C).

In FIG. 4, 10 is a modem on the transmitting side, 20 is a modem on the receiving side, 21 is a built-in modem in the receiving side modem 20, and 30 is a line connecting both modems.

The line 30 has frequency characteristics, and its transmission characteristics are, for example, the characteristics shown in FIG. 5(B). Therefore, the received signal Rk received by the receiving modem 20 is affected by this transmission characteristic. When the receiving side modem 20 is equipped with the domain 21 having the frequency characteristics shown in FIG. The complex characteristic (simple multiplication in the frequency domain) is a flat characteristic within the usage band shown in FIG. 5(A). As a result, distortion-free signal transmission is possible.

Figure 6 shows a typical configuration diagram of the Toto domain 21.

In general, the filter 21 is composed of a transversal filter, and 400 in the figure is a delay element that delays the received data Rm for a certain period of time, and 401 is a tap gain [C-8~ CN] 0 As is well known, this tap gain expressed on the time axis is called the unit impulse response, and the Fourier transform of this is the frequency characteristic of the other domains shown in Figure 5 (C). Become.

Further, 402 is a multiplier that multiplies the received data delayed by the delay element 400 and the tap gain 401;
403 is an adder that takes the total value of the multiplication result of the received data delayed by the delay element 400 from each multiplier 402 and the tap gain 401;

The uniform output signal yk with the above configuration can be expressed by the following equation.

The controller 201 adapts to the inverse characteristics of the line by sequentially calculating each tap gain using the following formula using the MSE method (Mian Sguare Error method) based on the received data.

however, Ce”1=: γTap gain value calculated 11th time Δk = Judgment value (estimated value) It is an estimated value of received data am, During the training period @ = am.

α: Convergence coefficient (generally α (1) yk-δk: Error signal (ek).

Note that the above-mentioned MSE method is an algorithm that minimizes the squared error signal e2k.

Particularly in this modem, speed and precision of this equalization operation are required, and it is no exaggeration to say that this itself determines the performance of the modem.

In general, 2 (in yx--a) on the right side of equation (2) according to the MSE method is called the evaluation function, and the MSE
The method takes advantage of the fact that this evaluation function is a downwardly convex function for each tap gain Ce, performs partial differentiation to bring the function closer to the extreme value, and as a result minimizes and equalizes the function. be.

The method using the recurrence formula (2) is a well-known hill-climbing method, and the amount of correction is determined by the convergence coefficient α.

FIG. 7 shows a graph of this evaluation function and tap gain, and also shows the temporal behavior when the convergence coefficients are different in the graph (however, α, >α).

From this, the relationship between the convergence coefficient α and the equalization (convergence) speed is as follows.

Therefore, the appropriate value of the convergence coefficient α has been determined based on the balance between the final error size, ', and the convergence time (both of which are contradictory to each other).

Recently, there has been a strong desire to shorten this convergence time, especially in order to increase the efficiency of line (transmission path) usage.

[Problem to be solved by the invention] However, in the conventional example described above, if the convergence coefficient α is set large in order to speed up the convergence, the final error (due to the roughness of equalization becomes large)
There is a drawback. For this reason, the convergence coefficient α cannot be increased too much, and there is a limit to increasing the convergence speed.

[Means for Solving the Problems] The present invention has been made for the purpose of solving the above-mentioned problems, and includes the following configuration as one means for solving the above-mentioned problems.

That is, a plurality of delay elements that delay received data at fixed time intervals, a plurality of tap gains that equalize and correct the outputs of the delay elements, and a switching means that switches a plurality of stages of convergence coefficients that correct the tap gain values. , a comparison means for comparing the equalized output error with a plurality of predetermined threshold values, a multiplier for multiplying both the output values of each tap gain and the delay element, and an adder for summing the multiplication results of the multipliers. Equipped with.

Further, the convergence coefficient for correcting the tap gain value is configured to decrease sequentially as the equalization output error increases.

[Operation] In the above configuration, the convergence coefficient α of the switching means is switched according to the equalization output error from the comparison means.
is changed step by step according to the magnitude of the equalization error, and equalization with a high convergence speed can be performed at a constant rate without reducing the precision of equalization even under various types of line distortions. In particular, by reducing the convergence coefficient α step by step according to the size of the equalization error, it is possible to perform equalization with a high convergence speed without reducing the precision of the equalization, even under various types of line distortion. can.

Furthermore, significant effects can be obtained by using a squared error signal as the equalized output error signal.

[Example] Hereinafter, an example according to the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram of a modem equipped with an automatic modem according to an embodiment of the present invention. be.

In FIG. 1, 100 and 118 are transmitting terminals and receiving terminals that are connected to the modem of this embodiment and generate digital signals to be transmitted.

101 is a scrambler that randomizes transmission data in order to prevent continuous output of the same data; 102 is an encoder that assigns a code to the signal from the scrambler 101 for each tribit, guibit, etc.; and 103 is a code amount interference of the signal. Waveform shaping filter (roll-off filter) that prevents
4 is a modulator that performs predetermined modulation processing on the signal from the waveform shaping filter 103. The modulation method in this modulator 104 is a quadrature amplitude modulation (QAM) method that changes the amplitude and phase of a carrier wave.

The signal modulated by this modulator 104 is converted into an analog signal by a D/A converter 105 to be sent to an analog public line, etc., and is further processed by a low-pass filter 106 to match the transmission band of the transmission path. harmonic components are removed and sent to the transmission line.

On the other hand, from the transmission signal from the transmission path, components outside the transmission band are first removed by a bandpass filter 110, then controlled by AGCIII to a signal level that can be handled on the receiving side, and then converted into a digital signal by an A/D converter 112. be converted into After being converted into a digital signal, it is demodulated by the demodulator 113 into the original signal before modulation. Here, reference numeral 114 denotes an equal function, in which distortion components received during transmission are removed from the received signal transmitted here, as described above, and the original transmitted signal is extracted. The output signal of this output signal 114 is sent to a determiner 115, where it is determined to be a code point, and then decoded by a decoder 116 and sent to a descrambler 117.
The signal randomized by the scrambler 101 on the transmitting side is restored to its original state.

In this way, the signal is returned to the same signal as the transmission signal output from the transmitting terminal 100, and is output to the receiving terminal 118 side.

In this way, by using a modem, it is possible to transmit digital signals via a public line, which is a general analog line.

The configuration of the automatic equalizer 114 according to an embodiment of the present invention used in the above modem will be explained below with reference to FIG.

Recently, digital signal processors (DSPs) are often used for high-speed digital processing, especially for medium to
Most high-speed modems are composed of DSP. Therefore, the following description will be made on the premise that the equalizer 114 is constituted by a DSP, but the present invention is not limited to this.

In FIG. 2, the area within the dotted chain line is the equalizer portion of the transversal filter shown in FIG. 6 mentioned above.

In addition to the equalizer section described above, this embodiment includes the following configuration.

200 is No. 4 ak transmitted from the output signal y of the equalizer 114.
201 is a determiner that determines the estimated value ak (ak=ak during training), and 201 is the equalized output yK and the estimated value ak
This is an error signal calculation unit that calculates an error signal (actually yk−ak) from the error signal e from the error signal calculation unit 201.

Therefore, the tap gain of the equalizer 114 is calculated for each equalized output using the above-mentioned equation (2).

Conventionally, the convergence coefficient α was fixed at this time, but in this embodiment, the convergence coefficient α is made variable. That is, the convergence coefficient α is made variable by the following configuration.

A signal squaring unit 202 calculates a squared error signal ek2 from an error signal ek from the error signal calculation unit 201, and a threshold value from a threshold unit 206 that sets a predetermined threshold value and a squared error signal ek2 from the signal squaring unit 202. A comparison section 203 compares the error signal e k2 and a convergence coefficient α based on the output from the comparison section 203.
selector 204 (actually, memory, 20
It has convergence coefficients α[0] to α[calculate the address of nl) stored in 5.

In the above configuration, the squared error signal (the above-mentioned evaluation function)
e By changing the convergence coefficient in equation (2) according to the size of k2 as α[0]〉α[1]...〉α[nl, the convergence speed can be increased and the final error can be reduced. It is something to suppress.

The details of the equalization process by the above equalizer will be explained below with reference to the flowchart of FIG.

The equalizer 114 of the receiving modem monitors the arrival of a transmission request signal from the transmitting modem in step S1, and when the transmission request signal arrives, a well-known training signal for synchronization is subsequently sent. Therefore, the process proceeds to step S2 to prepare for detection (reception) of a training signal. Initialize the isolator when the training signal is detected,
For example, tap gain initialization processing (for example, variable convergence coefficient α and squared error signal ek2 of the threshold unit 206)
A threshold value t h [nl for [setting nl to "0", etc.] is performed.

Next, in step S4, the received signal Rk, which is the input signal, is input to the delay element 130 of the equalizer 114 to calculate the equalized output yk, and in the subsequent step S5, the estimate value ak of the received signal ek is determined by the determiner 200. do. The error signal calculation unit 201 calculates the error signal ek in step S6, and then calculates the error signal ek in step S6.
In step 7, the equalization tap gain is corrected using the following equation based on equation (2).

However, α[nl (n=0.1, 2...) is a convergence coefficient that is switched by the error signal.Next, it is determined in step SIO whether the training is finished or not, and if it is within the training period, it is returned to step Sll. Then, a signal squaring section 202 squares the error signal ek. The comparison unit 203 compares the threshold value t h [nl from the threshold value unit 206 and the squared error signal ek2 in the following step S12.

Here, if (ek"< th[nl), "n" is incremented in step S13 and the process returns to step S4; if (ek"≧th[nl), "n" is not changed and is immediately incremented. Returning to step S4, the process proceeds to the next equalization process.

On the other hand, if the training has been completed in step SIO, the process proceeds to step S15, where it is confirmed whether the error is smaller than the reference value (training end level), and if it is, equalization has been completed and the final convergence coefficient α[ m] is fixed as α, and the process moves on to actual communication of received data.

If it is not smaller, the process advances to step S16, a retraining request is issued, and the process returns to step S2.

Then, the training signal is retransmitted again, and the above-described process is repeated.

As described above, by providing a means for gradually reducing the convergence coefficient α in the basic equalization equation, the following effects can be obtained compared to the conventional case where the convergence coefficient α is fixed.

■The initial convergence coefficient α can be set large and the convergence speed can be increased.

■The final convergence coefficient α can be made small, and the final error can be made small.

(2) Since the switching of the convergence coefficient α is determined by the square error, similar effects can be obtained for various types of lines.

[Other Embodiments] In the above description, an example has been described in which the convergence coefficient α is switched based on an instantaneous squared error, but the present invention is not limited to the above example. 2
02 and the comparison unit 203, the error signal ek from multiple times in the past
It is also possible to add a configuration for accumulating and taking the average value, and to switch the convergence coefficient α based on the average value of the error signals of a plurality of past errors.

In this case, the influence of instantaneous error signal fluctuations can be reduced, and the convergence coefficient α can be switched more accurately.

Further, although this embodiment has been described on the assumption that a DSP is used, the invention is not limited to this, and even if the algorithm is created using hardware, the algorithm is the same and the effect will not change at all.

As explained above, according to the embodiment, the convergence coefficient a
By providing means for reducing the value in stages according to the error signal, the equalization convergence speed and equalization performance can be improved, and the same effect can be obtained in any line.

Therefore, the effective communication speed can be effectively improved, and high-speed data transmission can be performed at a lower cost.

[Effects of the Invention] As explained above, according to the present invention, by changing the convergence coefficient step by step according to the magnitude of the equalization error, equalization with a high convergence speed can be achieved without reducing the precision of the equalization. This can be done consistently even under various types of line distortions.

In particular, by reducing the convergence coefficient step by step according to the size of the equalization error, and by making the equalization output error signal a squared error signal, the convergence speed can be increased without reducing the precision of equalization. , and similar effects can be obtained even with various types of line distortion.

[BRIEF DESCRIPTION OF THE DRAWINGS] Fig. 1 is a block diagram of a modem according to an embodiment of the present invention, Fig. 2 is a diagram showing the detailed structure of the modem of this embodiment, and Fig. 3 is an equalizer of this embodiment. Figure 4 is a diagram for explaining the flow of a general transmission signal, Figure 5 (A) is a diagram showing the frequency characteristics in the line, Figure 5 (B) is a diagram showing the equalization process by Figure 5 (C) is a diagram showing the frequency characteristics of the transmission signal from the transmitting modem and the output signal corrected by the transmission mode. Figure 6 is a diagram showing the configuration of a general transversal type transmission. FIG. 7 is a diagram showing the evaluation function and tap gain in the MSE method. In the figure, 100--Transmission terminal, 101--Scrambler, 102--Encoder, 103--Pulse shaping filter, 104--Modulator, 105--D/A converter, 106 ...Low pass filter, 110...Band pass filter, 111...AGC, 112...A/
D converter, 113... demodulator, 114... etc.,
115... Determiner, 116... Decoder, 117...
・Discrambler, 118...Receiving terminal, 200・
...Determiner, 201...Error signal calculation unit, 202...
- Signal squarer section, 203... Comparison section, 204... Selector, 205... Memory, 206 Threshold section, 400.
...Delay element, 401...Tap gain, 402...
- Multiplier, 403... is an adder.

Claims (3)

[Claims] (1) An automatic equalizer installed in a modulation/demodulation device used for data communication, which includes a plurality of delay elements that delay received data at fixed time intervals, and a plurality of taps that equalize and correct the output of the delay elements. a switching means for switching the gain and a convergence coefficient for modifying the tap gain value in multiple stages; a comparison means for comparing the equalization output error with a plurality of predetermined threshold values;
A multiplier that multiplies the output values of each tap gain and the delay element, and an adder that sums up the multiplication results of the multiplier, and the switching means converges according to the equalization output error from the comparing means. An automatic equalizer capable of successively automatically correcting tap gain values, characterized by switching coefficients. (2) The automatic equalizer according to claim 1, wherein the convergence coefficient sequentially decreases as the equalization output error increases. (3) The automatic equalizer according to claim 1, wherein the equalized output error signal is a squared error signal.