JP3395894B2 - Echo canceller training system and method - Google Patents

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION The present invention relates to an echo canceller. In particular, the present invention relates to ADSL (Asymmetric).
Digital Subscriber Line)
The present invention relates to an echo canceller training system and method for shortening the training time of the echo canceller (EC) of the transmission system and reducing the calculation amount.

[0002]

2. Description of the Related Art An echo canceller which has been conventionally used and its problems will be described by taking a multi-carrier transmission system, such as the ADSL system, in which a downlink use band is wider than an uplink use band. The ADSL method is a type of multi-carrier transmission method, DMT (Discrete Mul).
The tiTone) method is adopted.

A detailed explanation of the DMT system is given in "Start of countdown, ADSL in the country of light (Nikkei Communication, N
o. 283, pp. 92-111, 1998.1.
4) ”. ADSL transmission data is
It is grouped into several bits and placed as amplitude and phase information on each carrier.

The data arranged in the frequency range is IFFT.
It is demodulated as time domain data by (inverse fast Fourier transform). Received data in the ADSL method is modulated into data in the frequency domain by FFT (Fast Fourier Transform). In the ADSL method, as a means for canceling the line echo, there has been a conventional method called "METHOD AND APP".
ARATUS FOR ECHO CANCELLAT
ION WITH DISCREET MULTITIT
NE MODURATION, United Stat
e Patent (5317596) ", a frequency-time domain echo canceller was used.

FIG. 5 is a schematic diagram showing a use band of the ADSL system. As shown in the figure, the carrier is 4.312.
It is arranged every 5 kHz. In the station side device of the ADSL system, the reception band is narrower than the transmission band, and the transmission is 0 to
Up to 255 carriers are used, and reception uses 0 to 31 carriers.

FIG. 6 is a block diagram showing the configuration of an ADSL transmitter / receiver using a training system for an echo canceller, which is the premise of the present invention. As shown in this figure, AD
The transmission unit 100 of the SL transmission / reception unit includes a training data generator 110 that generates training data and an IFF that demodulates the generated training data into time domain data.
T (Inverse Fast Fourier Transform) 120, parallel-to-serial converter 130 that serially converts demodulated data, digital-to-analog converter 140 that converts serially converted data from digital to analog, and signal converted to analog And a low-pass filter 150 that removes unnecessary high-frequency components.

The training data generator 110 converts the known training data X _{k, n} into the IF data as described above.
The training data X _{k, n} is output to the FT 120.
Is created as follows. First, known random binary data of 512 bits is created, the following conversion is performed every 2 bits from the beginning, and 256 complex data S _{k, n} are created.

[0008]   1st bit 2nd bit Real part Imaginary part       0 0 → 1 1       0 1 → -1 1       1 0 → -1 -1       1 1 → 1 -1

Next, the training data X _{k, n} is generated using the following equation (8). X _{k, n} = 0 (n = 0, N / 2) X _{k, n} = S _{k, n} (0 <n <N / 2) X _{k, n} = S ^* _{k, N−n} (n> N / 2) (8) where k is a training number, n is a carrier number, N
/ 2 is the number of transmission carriers (256 in the ADSL method).

Since the training data has a Hamiltonian symmetry, the IFFT result remains only N sample real numbers. The training data X _{k, n} generated by the training data generator 110 is the IFFT 120.
Is converted into time domain data by way of the parallel / serial conversion section 130, the digital / analog conversion section 140, the low-pass filter 150, and further passes through the hybrid transformer 300 that performs 2-line-4 line conversion, and then the path 400.
Is output to.

Next, the receiving section 1000 of the ADSL transmitting / receiving section
Is a low-pass filter 1100 connected to the hybrid transformer 300 for removing unnecessary high-frequency components, an analog-digital conversion unit 1200 for converting an analog signal from which high-frequency components have been removed into a digital signal, and a subtraction connected to the analog-digital conversion unit 1200. Vessel 1300,
A serial-parallel conversion unit 1400 that parallel-converts the output data of the subtractor 1300, an FFT 1500 that demodulates the parallel-converted data into data in the frequency range, and a complex subtractor 1600 that subtracts the demodulated data. .

The received data is the hybrid transformer 30.
0, low-pass filters 1100 and 1200, a serial / parallel converter 1400, and an FFT 1500, and demodulated into frequency domain data. Next, the frequency domain-time domain echo canceller 200 of the ADSL transmitter / receiver is operated by the receiver 1
In order to perform the subtraction processing at 000, cancellation data for canceling the line echo generated in the hybrid transformer 300 is formed based on the transmission data from the transmission unit 100.

Therefore, the frequency range of the ADSL transmitter / receiver is
The time domain echo canceller 200 includes the training data from the training data generator 110 and the complex subtracter 1
Received data from 600 is input, and complex subtractor 1600
Frequency echo canceller 210 for outputting to, and IFFT 220 for demodulating output data of frequency domain echo canceller (EC) 210 into time domain data, and IFFT
The time domain echo canceller (EC) 230 receives the data demodulated by the 220 and the data demodulated by the IFFT 120 and outputs the data to the subtractor 1300.

As described above, in the frequency domain-time domain echo canceller 200, the frequency domain echo canceller 21 is used.
0, the time domain echo canceller 230 is used together, the training data from the training generator 110 is used, and the training for setting the optimum condition for canceling the line echo is performed. The frequency domain echo canceller 210 is configured by a 1-tap filter having a complex number coefficient prepared for each carrier.

The time domain echo canceller 230 uses the FIR
(Finite Impulse Response) filter. Frequency range echo canceller 2
10 is the training data X according to the following equation (9).
Carrier tap coefficient C corresponding to each carrier of _{k and n}
Data Y _multiplied by _{k and n} Calculate ec _{k, n} .

Y ec _{k, n} = C _{k, n} × X _{k, n} (9) Y ec _{k, n} : Internal data of the frequency range echo canceller 210 of the carrier n in the symbol k X _{k, n} : Training data of the carrier n in the symbol k C _{k, n} : Tap of the frequency range echo canceller 210 of the carrier n in the symbol k coefficient

The frequency domain echo canceller 210 transmits
Data Y corresponding to the carrier ec _{k, n}The following formula (1
0) to the reception band for use in the complex subtractor 160
Output to 0.   Y ec ’_{k, n}= Y ec_{k, n}+ Y ec_{k, 64 + n}+   Y ec_{k, 128 + n}+ Y ec_{k, 192 + n}+ Y ec_{k, 256 + n}+   Y ec_{k, 320 + n}+ Y ec_{k, 384 + n}+ Y ec_{k, 448 + n}                                 (0 ≦ n <M) (10)

M: number of carriers received (= 64) Y ec ′ _{k, n} : Output data of frequency domain echo canceller 210 The tap coefficient of frequency domain echo canceller 210 is LMS (Least Mean Sq) shown in the following equation (11).
updated) by the algorithm.

[0019]   C_{k + 1, n}= C_{k, n}+ ΜE_{k, n}X^* _{k, n}      (0 ≦ n <512)                                                             … (11)   E_{k, n}= E '_{k, mod (n, 64)}                        … (12) μ: Step size of coefficient update X^* _{k, n}: Training data X_{k, n}Complex conjugate of E ’_{k, n}: Identification error

At the time of half-duplex communication, the device on the opposite side has stopped transmission, so that the output of the complex subtractor 1600 is only the discrimination error _{E'k, n} . Time domain echo canceller 230
The tap coefficient of is calculated by performing an inverse fast Fourier transform on the tap coefficient obtained by the frequency echo canceller 210 by the IFFT 220.

The time domain echo canceller 230 performs two preprocessing. First, the frequency range echo canceller 210
Intersymbol interference caused by past symbols that cannot be eliminated by is eliminated. Second, the inter-symbol interference that occurs in the next symbol section due to the current transmission symbol is copied to the beginning of the current symbol.

By these two processes, in the FFT section,
Only the echo caused by the current symbol remains with periodicity. This echo is canceled by the frequency echo canceller 210. Output Y of time domain echo canceller 230 ec _n, because the sampling rates are different in transmission and reception, as represented by the following formula (13) are taken out one by one for every eight samples, are converted into 64 samples of data.

Y ec '_n' = Y ec _8n (13) where Y ec ′ _{n ′} is output data of the time domain echo canceller 230 after being thinned out. FIG. 7 is a schematic diagram showing how data is thinned by the equation (13). As shown in the figure, the data of the transmission symbol 512 becomes 64 data after data thinning.

In the case of full-duplex communication, the tap coefficient is updated and the echo is erased as in the case of half-duplex communication. In this way, the echo canceller training system
Set the optimum echo canceller data.

[0025]

By the way, in the above-mentioned echo canceller training system, since the training data is transmitted by arranging the data in all carriers, echoes of a plurality of transmission carriers are superimposed on one reception carrier. Was there. Therefore, in order to obtain sufficient echo suppression characteristics, it is necessary to reduce the step size of the frequency domain echo canceller 210.

However, if the step size is reduced, the convergence time of the frequency echo canceller 210 becomes long, and as a result, the training time of the entire echo canceller becomes long. Furthermore, since the training including the inverse fast Fourier transform is repeated to calculate the tap coefficient of the time domain echo canceller 230, there is a problem that the amount of calculation at the time of training increases.

Therefore, in view of the above problems, it is an object of the present invention to provide an echo canceller training system and method which can shorten the training time and the amount of calculation.

[0028]

In order to solve the above problems, the present invention is arranged in a transmission carrier in a training system of an echo canceller for multicarrier transmission in which a downlink transmission band is wider than an uplink reception band . frequency
Converts the transmission data in several areas to the transmission data in time area and sequentially
Transmitter to transmit and receive time domain received data in frequency
Converted to data, received data in the time domain, the frequency
Occurs when the transmission data is transmitted with respect to the reception data in the area
Line echo time Echo cancellation data, frequency
Receive by subtracting each echo cancellation data
Section and transmission data in the frequency range formed by the transmission section.
Data, the transmission data in the time domain, and the reception unit.
Of the transmission data based on the reception data in the frequency range
Time echo cancellation of the line echo that occurs during transmission
Data, the frequency range echo cancellation data respectively
Frequency domain-time domain echo canceller
Generating the training data and setting the transmission band to the reception band.
Placed on the transmission carrier in one divided transmission band divided into
Training data in the frequency domain
Converted into encoding data and sequentially transmitted to the transmitting unit,
Line eco that occurs when transmitting time-domain training data
Are received by the receiving unit, and training in the time domain is performed.
Data, based on training data in the frequency range
Of the line echo that occurs when the training data is transmitted
Training time echo cancellation data, trainee
Frequency range echo canceling data in the frequency range
− Form each on the time domain echo canceller, and
The training data received by the receiving unit.
For time domain received data and frequency domain received data
Training from the frequency-time domain echo canceller
Time echo cancellation data for training, frequency for training
Subtract the echo cancellation data for each range,
The frequency range received data of the line echo is the frequency range-time.
Training with feedback to the echo canceller
And a training data generator.
Providing training system that echo canceller.

By this means, the transmission band is divided by the reception bandwidth, and data is transmitted for each divided transmission band and transmitted, so that echoes of a plurality of transmission carriers are no longer superimposed as in the conventional case. As a result, the tap coefficient of the echo canceller can be calculated by dividing the received data by the training data in the frequency domain, instead of the method of asymptotically obtaining the tap coefficient like the LMS method, and the training time can be shortened. it can.

Preferably, the multi-carrier transmission method is the ADSL method. By this means, the present invention can be applied to a multicarrier transmission system such as the ADSL system. Preferably, the frequency domain-time domain echo canceller is a frequency domain echo canceller configured by inputting the signal in the frequency domain from the training generator and including a 1-tap filter having a complex number coefficient prepared for each carrier. And a time domain echo canceller which receives the time domain signal from the training generator and is configured by an FIR filter. The tap coefficient of the time domain echo canceller is the tap coefficient obtained by the frequency domain echo canceller. It is calculated by inverse fast Fourier transform.

By this means, intersymbol interference caused by past symbols that cannot be canceled by the frequency domain echo canceller is canceled. Inter-symbol interference caused by the current transmitted symbol in the next symbol period is copied to the beginning of the current symbol. In the FFT interval, only the echo caused by the current symbol remains with periodicity. This echo is canceled by the frequency domain echo canceller.

Preferably, the frequency-time domain echo canceller is provided with an initial value calculation circuit, and the initial value calculation circuit trains received data in the frequency domain.
An initial value of the tap coefficient of the frequency echo canceller is calculated by dividing by the data . By this means, the tap coefficient of the time domain echo canceller can be obtained by performing the inverse fast Fourier transform only once after calculating all the tap coefficients of the frequency domain echo canceller by the initial value calculation circuit. Can be reduced.

Preferably, a time domain correction unit is provided between the frequency domain echo canceller and the time domain echo canceller, and the time domain correction unit corresponds to a reception carrier whose transmission carrier in the divided transmission band is Nyquist. Due to this, the tap coefficient is corrected. By this means, it becomes possible to improve the deterioration of the line echo suppression characteristic.

Furthermore, the present invention provides a echo canceller training system wide multi-carrier transmission than the reception band of the transmission band of the downlink uplink, transmission carrier
The transmission data in the frequency domain placed in the
Process of converting to data and transmitting sequentially, and received data in time domain
To frequency reception data, and
The transmission data for the reception data in the frequency range
The time of the line echo that occurs when transmitting
Data and the frequency echo cancellation data
Arithmetic processing , transmission data in the frequency range, and the time
Inter-band transmission data, the frequency formed by the receiving unit
Based on the received data in the area
Time echo canceling data of the line echo, before
Form each of the frequency range echo cancellation data
A step to produce the training data, the transmission
Signal band divided into the reception band into one divided transmission band
Training data in the frequency range assigned to the transmission carrier
Data into time-domain training data and send them sequentially
And the process of sending training data in the time range
The process of receiving the resulting line echo and the time domain trace
Training data, training data in the frequency range
The line that occurs when transmitting the training data based on
Echo training time echo cancellation data,
Form training frequency range echo cancellation data
Process and line echo of received training data
For time domain received data and frequency domain received data
From frequency domain to time domain echo canceller for training
Time echo cancellation data, training frequency range
Process of subtracting echo cancellation data
For training the line echo frequency range received data
The feedback is used to form the frequency echo cancellation data.
And a step of performing training.
Providing training method of the echo canceller to be.

By this means, similarly to the above-mentioned invention, the transmission band is divided by the reception bandwidth, and data is transmitted for each divided one of the transmission bands to be transmitted. No longer overlap. As a result, the tap coefficient of the echo canceller can be calculated by dividing the received data by the training data in the frequency domain, instead of the method of asymptotically obtaining the tap coefficient like the LMS method, and the training time can be shortened. it can.

[0036]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of an ADSL transmitter / receiver using a training system for an echo canceller according to the present invention. As shown in the figure, what is different from the configuration of FIG. 6 is an initial value calculation circuit 240 of the frequency-time domain echo canceller 200.

The initial value calculation circuit 240 receives the output of the FFT 1500 and outputs it to the frequency domain echo canceller 210. FIG. 2 is a block diagram illustrating details of the initial value calculation circuit 240. As shown in the figure, the initial value calculation circuit 240 includes a training data generator 243, a band expansion unit 241 that receives the output of the FFT 1500 of the reception unit 1000, and an averaging process that receives the output of the band expansion unit 241. It has a circuit 242 and a complex divider 244 which receives the outputs of the training data generator 243 and the averaging circuit 242 as an input and outputs the output to the frequency domain echo canceller 210.

Frequency domain-time domain echo canceller 200
Training is performed by the following procedure. The training data generator 243 of the initial value calculation circuit 240 is the training data generator 1 of the transmitter 100.
The same training data X ′ _{k, n} as the training data X _{k, n} output at 10 is output. Further, N represents twice the number of transmission carriers, and M represents twice the number of reception carriers (N and M are powers of 2, N> M).

The training data generator 110 is
Carriers from kM / 2 to ((k + 1) M / 2) and from (N- (k + 1) M / 2 + 1) to (N-kM / 2) (however, when k = 0, (N-M / 2 + 1) ) To (N
Training data X _{k, n} having power only in the carriers up to -1) is generated, and L symbols are repeatedly transmitted.

Here, k represents a training number for initial value calculation, and the initial value is 0 and takes values up to N / M-1.
FIG. 3 is a diagram showing a model (here, k = 0 to 7) for explaining the training data X _{k, n} to be transmitted. As shown in the figure, the training data X _{k, n} generated by the training data generation unit 110 is arranged in carriers 0 to 31 and 481 to 511 when k = 0. When k = 1, data is arranged on the 32nd to 63rd, 449th to 480th carriers, ..., When k = 7, data is arranged on the 224th to 288th carriers.

As described above, in the conventional echo canceller, data is arranged in all carriers and training data is transmitted, so that echoes of a plurality of transmission carriers are superimposed on one reception carrier. Therefore, in order to obtain a sufficient echo suppression characteristic, it is necessary to reduce the step size of the frequency range echo canceller 210.

When the step size is reduced, the convergence time of the frequency echo canceller 210 becomes longer, and as a result, the training time of the entire echo canceller becomes longer. On the other hand, in the echo canceller of the present invention, the transmission band is divided by the reception bandwidth, and the data of each divided band is transmitted and transmitted, so that the echoes of a plurality of transmission carriers are not superimposed.

Thus, the tap coefficient of the echo canceller can be calculated by dividing the received data by the training data in the frequency domain, instead of the method of asymptotically obtaining the tap coefficient as in the LMS method, and the training time can be shortened. can do. The band extension unit 241 uses the training data X _{k, n.}
The band of the output Y _{k, n} of the FFT 1500 at the time of transmission is extended by the following equation (1).

Y ′ _{k, n} = Y _{k, mod (n, M)} (1) Y _{k, mod (n, M)} : output of FFT 1500 Y ′ _{k, n} : output of band expansion unit 241 (carrier number n
= 0 to N) where mod (n, M) represents the remainder when n is divided by M. The averaging circuit 242 averages the output data of the band extending unit 241 for L symbols, and outputs the data Y ″ _{k, n} to the complex divider 244.

In the complex divider 244, the frequency domain echo canceller 210 is used for the carrier having the power of the training data X ′ _{k, n} by using the following equation (2).
The tap coefficient initial value C _{k, n of} is output. C _{k, n} = Y ″ _{k, n} / X ′ _{k, n} (2) is repeated from k = 1 to N / M−1. When all the tap coefficient initial values have been obtained, the complex divider 244 outputs the values to the frequency domain echo canceller 210 and sets the tap coefficient initial values.

Frequency domain-time domain echo canceller 200
When the tap coefficient of the frequency domain echo canceller 210 is set, the same operates as in the conventional method. The operation of the training system for the echo canceller according to the present invention will be described below. As an example of an ADSL system side device, transmission is performed with 512 carriers and reception is performed with 64 carriers.

The training data generator 110 generates known training data X _{k, n} . Where k
Is a training number (k = 0 to N / M-1), N is twice the number of transmission carriers (here, 512), M is twice the number of reception carriers (here, 64), and n is a carrier number. It represents (n = 0 to N-1). Training data X
_{k and n} are generated as follows, for example.

First, 64 bits of known pseudo-random binary data are generated, and the following conversion is performed by 2 bits from the beginning to generate 32 complex data S _n . 1st bit 2nd bit Real part Imaginary part 0 0 → 1 1 0 1 → -1 1 1 1 0 → -1 -1 1 1 → 1 -1

Next, the training data X _{k, n} is generated using the following equation (3). The training data X _{k, n} generated by the following formula (3) is called Hamiltonian symmetry, and if the training data X _{k, n} having this Hamiltonian symmetry is IFFTed, it has the property of being only real data. . X _{k, n} = 0 (n = 0, N / 2)

X _{k, n} = S _{n−kM / 2} (kM / 2 ≦ n ≦ (k + 1) M / 2−1) X _{k, n} = S ^* _{N−kM / 2−n} (N− (k + 1) M / 2 + 1 ≦ n ≦ N−kM / 2), but when k = 0, (N− (k + 1) M / 2 + 1 ≦ n ≦ N−kM / 2−1. X _{k, n} = 0 (Other) (3)

The training data generator 110 uses k =
The training data is output from 0 to k = 7 for each L symbol. Next, the operation of the initial value calculation circuit 240 will be described. At the time of half-duplex communication, the device on the opposite side has stopped transmission, so the Y demodulated by the FFT 1500
_{k and n} are only the echo component of the transmission data.

In the initial value calculation circuit 240, the frequency component of the echo path is derived by dividing the echo component in that frequency range by the training data. First, the band expansion unit 241 outputs the output data Y _{k, n of the} FFT 1500.
For, the bandwidth extension is performed according to the following equation (4). _{Y'k, n} = _{Yk, mod (n, 64)} (4)

The averaging circuit 242 uses the band-extended Y '.
_{K, n} are averaged over L symbols and data Y ″ _{k, n} is output. The training data generator 243 outputs the same training data X ′ _{k, n} as the data X _{k, n} output by the training data generator 110. The complex divider 244 uses the following equation (5) for the carrier in which the data is arranged and uses the frequency domain echo canceller 210.
The initial value C _{k, n} of the tap coefficient of is calculated.

C _{k, n} = Y ″ _{k, n} / X ′ _{k, n} (5) For example, when k = 0, data is arranged in the 0th to 31st carriers and the 481st to 511th carriers. Therefore, the complex divider 244 uses the output Y ″ _{0, n} of the averaging circuit 242 and the training data X ′ _{0, n} to calculate the initial value C _{0, n} of the tap coefficient according to the following equation (6). calculate.

C _{0, n} = Y ″ _{0, n} / X ′ _{0, n} (n = 0 to 31) (6) The complex divider 244 calculates the tap coefficient value of the frequency band echo canceller 210 obtained. Save it. When the initial value of the tap coefficient for all carriers is obtained, the complex divider 244 outputs the initial value of the tap coefficient to the frequency domain echo canceller 210.

When the initial value of the tap coefficient is set by the initial value calculation circuit 240, the frequency domain echo canceller 210 updates the characteristics of the echo replica and the tap coefficient as in the conventional echo canceller. In the conventional echo canceller, as described above, the training including the inverse fast Fourier transform is repeated to obtain the tap coefficient of the time domain echo canceller 230.

On the other hand, in the echo canceller of the present invention, the initial value calculation circuit 240 is provided to calculate all the tap coefficients of the frequency domain echo canceller 210, and then the inverse fast Fourier transform is performed to tap the time domain echo canceller 230. We are looking for a coefficient. Therefore, the tap coefficient of the time domain echo canceller 230 can be obtained by performing the inverse fast Fourier transform only once, and the amount of calculation during training can be reduced.

Next, the 32 × i (i = 1, 2, ..., 15) th transmission carrier corresponds to the Nyquist reception carrier. The echo of the 32 × i-th carrier has an imaginary number of 0 and is data of only a real number, so that the correct tap coefficient of the frequency range echo canceller 210 cannot be obtained.

Therefore, the correct tap coefficient of the time domain echo canceller 230 cannot be obtained, and the echo suppression characteristic is deteriorated. Therefore, the tap coefficient of the time domain echo canceller 230 is corrected as follows. FIG. 4 is a block diagram showing another configuration of the ADSL transmitter / receiver using the training system of the echo canceller according to the present invention.

As shown in the figure, in the frequency domain-time domain echo canceller 200, a time domain correction section 250 is provided between the IFFT 220 and the time domain echo canceller 230. The time domain correction unit 250 corrects the tap coefficient of the time domain echo canceller 230 according to the following equation (7).

[0061]   C ’_n= C_n-A_{mod (n, 64)}                          … (7)   A_m= C_{m + 448} C ’_n: Touch of the time domain echo canceller 230 after correction
Coefficient C_n: Tap of time domain echo canceller 230 before correction
coefficient A_m: Tap of time domain echo canceller 230 before correction
Coefficient of 64 taps in the latter half of the coefficient

N: tap number of tap coefficient of time domain echo canceller 230 (n = 0 to 511) m: tap number of stored component (m = 0 to 63) mod (n, 64): n is divided by 64 On the other hand, by correcting the equation (7), the tap coefficient of the time domain echo canceller 230 that is closer to the actual echo path can be calculated, and the echo suppression amount can be improved.

[0063]

As described above, according to the present invention,
By dividing the transmission band by the reception bandwidth and transmitting the data by dividing each divided band, echoes of a plurality of transmission carriers are no longer superimposed as in the conventional case. As a result, the tap coefficient of the echo canceller can be calculated by dividing the received data by the training data in the frequency domain, instead of the method of asymptotically obtaining the tap coefficient like the LMS method, and the training time can be shortened. it can.

After the initial value calculation circuit 240 is provided and all the tap coefficients of the frequency range echo canceller 210 are calculated,
The tap coefficient of the time domain echo canceller 230 can be obtained by performing the inverse fast Fourier transform only once, and the amount of calculation at the time of training can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an ADSL transmitter / receiver using a training system of an echo canceller according to the present invention.

FIG. 2 is a block diagram illustrating details of an initial value calculation circuit 240.

FIG. 3 is a diagram showing a model (here, k = 0 to 7) for explaining training data X _{k, n} to be transmitted.

FIG. 4 is a block diagram showing another configuration of an ADSL transceiver unit using the training system for the echo canceller according to the present invention.

FIG. 5 is a schematic diagram showing a use band of the ADSL method.

FIG. 6 is a block diagram showing a configuration of an ADSL transmitter / receiver using a training system of an echo canceller which is a premise of the present invention.

FIG. 7 is a schematic diagram showing how data is thinned out according to equation (13).

[Explanation of symbols]

100 ... Transmitter 110 ... Training data generator 120 ... IFFT 130 ... Parallel-to-serial converter 140 ... Digital-analog converter 150 ... Low-pass filter 200 ... Frequency-time domain echo canceller 210 ... Frequency echo canceller 220 ... IFFT 230 ... Time echo canceller 240 ... Initial value calculation circuit 241 ... Band extension unit 242 ... Averaging circuit 243 ... Training data generator 250 ... Time correction unit 300 ... Hybrid transformer 400 ... Route 1000 ... Receiver 1100 ... Low-pass filter 1200 ... Analog-to-digital converter 1300 ... Subtractor 1400 ... Serial-parallel converter 1500 ... FFT 1600 ... Complex subtractor

─────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-11-186942 (JP, A) Yuichi Fukuchi, Katsutoshi Seki, Yuichi Ito, “A Study of Echo Cancellers in Multi-Carrier Transmission Systems”, 1999 Den Child and Information Communication Society General Conference, Communication 2, Nihon, March 8, 1999, p. 284, (B-8-7) (58) Fields investigated (Int.Cl. ⁷ , DB name) H04B 1/76 H04B 3/00 H04B 7/00 H04J 11/00 INSPEC (DIALOG) JISST file (JOIS)

Claims (6)

(57) [Claims] 1. The downlink transmission band is more than the uplink reception band
Wide Multi-Carrier Echo Canceller Trainee
In theTime the transmission data in the frequency band allocated to the transmission carrier.
A transmission unit for converting the transmission data of the area and transmitting sequentially Convert the received data in the time domain to the received data in the frequency
For the received data in the time domain and the received data in the frequency domain
And the time of the line echo that occurs when the transmission data is transmitted
Echo cancellation data, the frequency echo cancellation
A receiver that subtracts each data, The transmission data in the frequency range formed by the transmitter,
The transmission data in the time domain, the circumference formed by the receiving unit.
When transmitting the transmission data based on the reception data in the wave number range
Time of the line echo occurring in
And the above-mentioned frequency range echo cancellation data
Frequency domain-time domain echo canceller Generate training data and receive the transmission band
The transmission carriers are allocated to one divided transmission band divided into bands.
The training data in the frequency domain
Converted into training data and sent to the sending unit in sequence,
Circuit generated when transmitting training data in the time range
The echo is received by the reception unit, and the trainer in the time domain is received.
Based on the training data in the frequency range
And the line eco that occurs when the training data is sent
Training time echo cancellation data, training
The frequency canceling echo cancellation data for
Formed in the several-time domain echo canceller,
In addition, the line of training data received by the receiving unit
For echo time domain data and frequency domain data
From the frequency-time echo canceller to trainee
Time echo cancellation data for training, training frequency
Subtraction processing is performed for each wave number echo cancellation data.
The line echo frequency band reception data to the frequency band −
Give feedback to the time domain echo canceller and train
And a training data generation unit.
To Echo canceller training system. 2. The method of multi-carrier transmission is ADS
The echo canceller training system according to claim 1, wherein the training system is an L system. 3. The frequency domain-time domain echo canceller receives the signal in the frequency domain from the training generator and has a complex number coefficient 1 prepared for each carrier.
The frequency domain echo canceller is composed of a tap filter, and the time domain echo canceller is composed of an FIR filter that receives the time domain signal from the training generator. The tap coefficient of the time domain echo canceller is The training system for an echo canceller according to claim 1, wherein the tap coefficient obtained by the frequency domain echo canceller is calculated by performing an inverse fast Fourier transform. 4. An initial value calculation circuit is provided in the frequency domain-time domain echo canceller, and the initial value computation circuit comprises:
Divide the received data by the training data in the frequency range
And calculates the initial value of the tap coefficients of the frequency domain echo canceller by Rukoto, echo canceler training system according to claim 3. 5. A time-domain correction unit is provided between the frequency-domain echo canceller and the time-domain echo canceller, and the time-domain correction unit is configured such that a transmission carrier in a divided transmission band corresponds to a Nyquist reception carrier. The training system for the echo canceller according to claim 3, wherein the tap coefficient is corrected due to 6. The downstream transmission band is more than the upstream reception band.
Wide Multi-Carrier Echo Canceller Trainee
In theTime the transmission data in the frequency band allocated to the transmission carrier.
A step of converting into the transmission data of the area and transmitting sequentially, Convert the received data in the time domain to the received data in the frequency
For the received data in the time domain and the received data in the frequency domain
And the time of the line echo that occurs when the transmission data is transmitted
Echo cancellation data, the frequency echo cancellation
A step of subtracting each data, The transmission data in the frequency range, the transmission data in the time range,
Based on the received data in the frequency range formed by the receiving unit
Then, the line echo generated when the transmission data is transmitted
Time echo cancellation data, the frequency range echo key
Forming the cancel data respectively, Generating training data, One divided transmission obtained by dividing the transmission band into the reception band
Training of frequency bands allocated to transmission carriers in the band
Data into time-domain training data and sequentially
The process of sending, Circuit generated when transmitting training data in the time range
Receiving an echo, Training data in the time domain, training in the frequency domain
Sending the training data based on the training data
Time echo key for training line echo
Cancellation data, frequency range echo cancel for training
Forming cell data, Time-domain reception of line echo of received training data
Data, frequency range For received data, the frequency range-time
Inter-zone echo canceller to training time echo key
Cancellation data, frequency range echo cancel for training
A step of subtracting the cell data respectively, Frequency data received in the line echo frequency range for training
Feedback to the formation of the numerical echo cancellation data
And the step of performing training.
Echo canceller training method.