JP2001053653A - Echo canceler - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve an echo suppress characteristic by interpolating the tap coefficient of a frequency domain echo calculating means and subsequently calculating the tap coefficient of a time domain echo calculating means. SOLUTION: An EC(echo canceler) 80 erases an echo due to a transmission signal made to take a roundabout way through a hybrid transformer 13. Then, the EC 80 is provided with a frequency interpolating part 81 interpolating the tap coefficient of a frequency domain EC 31, and an IFFT part 34 calculates the tap coefficient of a time domain EC 33 by using the interpolated tap coefficient. The part 81 interpolates a tap coefficient corresponding to a Nyquist frequency by taking the average value of two adjacent tap coefficients. Thus, it is possible to calculate the tap coefficient of a correct time domain EC 33 because of updating the tap coefficient corresponding to the Nyquist frequency where a frequency response becomes zero so that the tap coefficient of the EC 31 cannot be updated by interpolating the tap coefficient.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an echo canceller, and more particularly, to an echo canceller for removing an echo signal from which a transmission signal goes to a receiving side.

[0002]

2. Description of the Related Art Asymmetric digital subscriber lines (Asymmetric digital subscriber lines)
Digital Subscriber Line: Hereinafter, ADSL is abbreviated. )
The transmission method provides a data communication system having different transmission speeds in the upstream and downstream directions by using one existing telephone line. Actually, one telephone line can be used for not only data but also both telephone and data communication. Hereinafter, a data communication system using such an ADSL transmission method is referred to as an ADSL system. AD
The SL system is, for example, the Internet (the Intern) in which a plurality of computer networks are connected to each other.
et) or Video On Deman
d: VOD) service, which is suitable for providing a service for receiving a large amount of data on only one terminal side.

The International Telecommunication Union Telecommunication Standardization Sector (In
ternational Telecommunication Union-Telecommunicat
ion Standardization Sector: ITU-T and American National Standards Institute: AN
A DMT (Discrete MultiTone) system is defined as one of the standard specifications of the ADSL transmission system by a standardization organization such as SI). The DMT method is a type of multicarrier transmission method in which data is combined every few bits and assigned to each carrier, and a plurality of carriers are multiplexed for transmission and reception. The transmission data is quadrature amplitude modulated (Quadrtutu
re Amplitude Modulation (QAM) is performed and allocated to each carrier. Multiplexing of the arranged carriers is
Fast Fourier Transform: F
Abbreviated as FT. ) Is used.

[0004] As a full-duplex communication system, there are a frequency division multiplexing (FDM) system and an echo canceller (hereinafter abbreviated as EC) system. In the FDM system, the frequency bands used in the uplink direction and the downlink direction are completely divided to realize full-duplex communication and obtain good transmission characteristics. On the other hand, E
In the C system, the frequency bands used in the uplink direction and the downlink direction are overlapped to realize full-duplex communication, and the band use efficiency is improved. The uplink and downlink signals of the overlapping frequency band are separated by EC. The EC system has a higher band use efficiency than the FDM system, but requires a strict echo suppression characteristic in addition to an increase in the scale of calculation required for separation processing and the like.

Conventionally, as an EC in an ADSL system, for example, US Pat. No. 5,317,596 “METHOD
AND APPARATUS FOR ECHO CANCELLATION WITH DISCRETE
MULTITONE MODULATION ”(May 31,1994, John M. Cioff
i, ACBingham)
An EC using both the time domain EC and the EC has been proposed.

FIG. 3 is a diagram showing A using the EC proposed in the past.
1 shows an outline of a configuration of a transmission / reception device in a DSL system. Here, a transmission / reception device applied to an ADSL system having a wider reception band compared to a transmission band is shown. This transmitting / receiving apparatus is connected to another opposing transmitting / receiving apparatus (not shown) via a line 10. A transmission unit 11 and a reception unit 12 are connected to the line 10 via a hybrid transformer 13. The line 10 is a metallic cable used for an existing telephone line. The hybrid transformer 13 is a transformer that electromagnetically couples the line 10 with the transmission unit 11 and the reception unit 12,
Performs 2-wire to 4-wire conversion. Thereby, the transmission signal from the transmission unit 11 is transmitted to the line 1 via the hybrid transformer 13.
Sent to 0. Further, a reception signal from the line 10 is received by the receiving unit 12 via the hybrid transformer 13. However, at this time, since the echo of the transmission signal from the transmission unit 11 wraps around the reception signal received by the reception unit 12 via the hybrid transformer 13, the reception unit 1
An EC 14 for suppressing this echo is provided in 2.

The transmitting section 11 includes a serial / parallel converting section 16 for converting the serially input transmission data 15 into parallel, a mapper section 17 for arranging the parallel converted transmission data on each carrier every several bits. , An IFFT unit 18 that performs an Inverse Fast Fourier Transform (hereinafter abbreviated as IFFT) after band expansion of the arranged carriers, and a cyclic prefix (Cyclic Prefi) for the output of the IFFT unit 18.
x: hereinafter abbreviated as CP. ) Is added. The transmitting unit 11 further includes a parallel-serial converting unit 20 that converts the signal with the CP added thereto into serial data, a digital-to-analog converting unit 21 that converts the converted serial data into an analog signal, and a high-frequency component of the analogized signal. And a low-pass filter 22 for blocking.

The receiving section 12 includes a hybrid transformer 13
Low-pass filter 23 that cuts off a high-frequency noise component of the reception signal input through the line 10, for example.
And an analog-to-digital converter 24 for digitizing the output, and an EC for the digitized output signal.
And a serial-to-parallel converter 2 for converting the subtraction result of the subtractor 25 into parallel.
6 is provided. The receiving unit 12 further includes an FFT unit 27 that demodulates the converted parallel data by FFT.
A subtractor 28 for canceling an echo caused by the current symbol estimated by the EC 14 with respect to the FFT signal; and a code identification unit 29 for reconstructing the subtraction result of the subtracter 28 into binary data. ing.

The EC 14 includes a band extending unit 30 for extending the output of the mapper unit 17 to the band on the receiving side, and a frequency band for calculating a replica of an echo generated by each carrier from the carriers band-extended by the band extending unit 30. E
C31. The EC 14 further includes a data interpolation unit 32 for interpolating the data converted into the time domain by the IFFT unit 18 and a finite impulse response filter (Finite
Impulse Response Filter: Abbreviated below as FIR. ), And a time domain EC33 for generating an echo replica from the data interpolated by the data interpolation unit 32, and a time domain EC using the IFFT from the tap coefficients of the frequency domain EC31.
And an IFFT unit 34 for calculating a tap coefficient of 33. The replicas calculated in the frequency domain EC31 and the time domain EC33, respectively, cancel the echoes in the frequency domain and the time domain in the subtractor 28 and the subtracter 25 in the receiving unit 12, respectively.

In the transmission / reception device having such a configuration, first, transmission data 15 which is time-series serial data is converted into parallel data by a serial / parallel conversion unit 16.
The parallel data converted by the serial / parallel converter 16 is arranged on a plurality of carriers by the mapper 17 every several bits. Here, it is assumed that the parallel data is converted into a plurality of pieces of information every several bits, and then is arranged on each carrier by QAM.

FIG. 4 shows an example in which the mapper unit 17 allocates 2-bit data to one carrier. Each combination of the first and second bits of the data is converted into a combination of a real component and an imaginary component that are orthogonal to each other. For example, parallel data in which the first bit is “0” and the second bit is “0”, and the real part is “1”
Indicates that the imaginary part is converted to "1" and arranged on the carrier. Also, for example, it indicates that the parallel data in which the first bit is “1” and the second bit is “1”, the real part is “1”, and the imaginary part is converted to “−1” and arranged on the carrier.

FIG. 5 shows 4QAM in the mapper section 17.
FIG. Taking the real part (Re) on the horizontal axis and the imaginary part (Im) on the vertical axis, by assigning as shown in FIG.
M places 2-bit data on one carrier.

The mapper portion 17 thus arranged
Is subjected to IFFT by the IFFT unit 18 and converted into time-domain data. The data sequence subjected to one IFFT process is called one symbol. By the way, this IF
Before the FT, the output data of the mapper unit 17 is band-extended. If the output data of the mapper section 17 is X _{k, n} ,
This band extension is performed as in the following equations (1) to (3). Here, k is a symbol number, n is a carrier number, N
/ 2 is the number of transmission carriers.

X _{k, n} = 0 (n = 0, N / 2) (1) X _{k, n} = X _{k, n} (0 <n <N / 2) (2) X _{k , n} = X ^* _{k, Nn} (n> N / 2) (3)

Here, X ^* _{k, Nn} is a complex conjugate of _{Xk, Nn} .

FIG. 6 shows a state of band expansion by the above equations (1) to (3). The data arranged in the first area 40 is centered on the “N / 2” -th carrier corresponding to the Nyquist frequency of the transmission, and the real part thereof is copied to the second area 41 symmetrically with respect to the vertical axis. 42), the imaginary part is point-symmetrically band-extended to the third region 44 (copy 4).
4). As a result, the data after IFFT by the IFFT unit 18 becomes only the real part, and the number of samples is twice as large as “N”.
Becomes

A CP is added to the output data of the IFFT unit 18 by a CP insertion unit 19. Usually, each data is
The response characteristic of “0” and the low-frequency cutoff characteristic of the hybrid transformer 13 cause “swinging”. The CP is added to the head of each symbol in order to take in the "end of the data" at the data end that cannot fit in the symbol in the symbol itself. Furthermore, the purpose of the CP is to reproduce the periodicity during FFT. Since the length of the CP to be added is set on the basis of the “isogas” as the solitary wave response length in the received symbol, it is necessary to perform processing in the time domain EC.

FIG. 7 shows how a CP is added.
Things. FIG. 3A shows that each symbol is continuous in time series.
It shows how you are doing. FIG. 13B shows the (K +
1 shows how a CP is added to the symbol of 1).
ing. The output data converted to the time domain by the IFFT unit 18
Is the K-th symbol 45_K, The (K + 1) th symbol
45_{K + 1}, The (K + 2) th symbol 45_{K + 2},···age
Are chronologically continuous. Each symbol 45_K, 45_{K + 1},
45_{K + 2}Is CP46 at the beginning of each_K, CP46 _{K + 1},
CP46_{K + 2}Is added. As shown in FIG.
The above-mentioned “Soshihiki” is, for example, the (K + 1) th symbol
Le 45_{K + 1}Data of only L samples from the data end of 4
7 is the response characteristic of the line 10 and the hybrid transformer 13
Due to the low-frequency cutoff characteristic of the
Cheating. This “Sisohiki” is “Sisohiki” 48_{K + 1}Part of
Equivalent to a minute. So, CP_{K + 1}Has the following equation (4)
Is added as follows. Where k is the symbol number, j
Is the sample number in the symbol, X _{k, j}The CP is added
Data.

X _{k, j} = X _{k, N + J−1} (j = −1 to −L) (4)

That is, X _{k, j} is set to “Kiyoshi” 48 _{K + 1}
The data 47 corresponding to L samples, which is the sample data in which is generated, is copied 49. Then, a “sowing” 50 _{K + 1} portion, which cannot be avoided, and is equivalent to “sowing” 48 _{K + 1} is taken into the symbol 51. As a result, the symbol 51 includes a “knee” 51 _{K + 1 based on} CP _{K + 1} . In this way _{, N from} X _{K + 1,0 to} X _{K + 1, N-1}
From the (K + 1) th symbol 45 _{K + 1 of the} sample, copy L _{K of the} data ends X _{K + 1, NL-1 to} X _{K + 1, N-1 to} the beginning to obtain CP _{K + 1} And X _{K + 1, -L to} X _{K + 1, N-1}
Of “N + L” samples. Furthermore, CP
Is added so that the data sequence demodulated by the FFT maintains continuity at the start point and the end point. By maintaining continuity at the start and end points, periodicity can be realized, and F
An accurate frequency response can be obtained at the time of FT.

In the range of the symbol 50, "Soshihiki" 4
Since the _{8K + 1} part is also included, the minimum value must be used so that the "sounding" of the past symbol does not overlap with the current symbol, or the necessary data can be contained in the symbol when demodulating the current symbol. Protection time for L samples is secured, and good demodulation characteristics are obtained.

The data to which the CP has been added by the CP insertion unit 19 is converted to serial data by the parallel / serial conversion unit 20. The serial data converted by the parallel-to-serial conversion unit 20 is converted into an analog signal by a digital-to-analog conversion unit 21, the harmonic component 22 is removed by a low-pass filter 32, and then transmitted to the line 10 via the hybrid transformer 13.

On the other hand, the received signal input via the line 10 is input to the receiver 12 via the hybrid transformer 13. In the receiving unit 12, first, the low-pass filter 2
3 removes the harmonic noise component added on the line 10, and is digitized by the analog-to-digital converter 24. From the digitized output signal, the echo component generated by the past symbol estimated by the EC 14 is eliminated by the subtractor 25. The result of the subtraction by the subtracter 25 is converted into parallel data by the serial / parallel conversion unit 26, and is converted into data in the frequency domain by the FFT unit 27. From the data in the frequency range converted by the FFT unit 27, the echo component generated by the current symbol estimated by the EC 14 is eliminated by the subtractor 28. Subtractor 28
Is applied to the carrier arrangement shown in FIG. At that time, "shift" at the time of this fitting
Is supplied to the EC 14.
The data applied to the carrier arrangement shown in FIG.
The conversion reverse to the assignment performed in FIG. 4 is performed by a demapper (not shown) to reconstruct the binary data.

FIG. 8 shows a frequency band used in the full-duplex communication system shown in FIG. Here, the vertical axis indicates amplitude, and the horizontal axis indicates frequency. A normal audio signal 55 using an existing metallic cable in a low frequency range,
By a transmission signal 57 having a plurality of carriers 56 from 30 kHz to 138 kHz and a reception signal 58 having a plurality of carriers from 30 kHz to 1.1 MHz,
Frequency bands are used. Arrangement interval 59 of each carrier
Are arranged every 4.3125 kHz. As described above, in the ADSL system shown in FIG. 3, the bandwidth used for the downlink reception signal is wider than that for the uplink transmission signal. For example, on the subscriber side of the ADSL system, transmission uses carrier numbers "0" to "31", and reception uses carrier numbers "0" to "255".

The EC 14 cancels the echo caused by the transmission signal that passes through the hybrid transformer 13 because the bands used by the transmission signal and the reception signal overlap. The band extending unit 30 extends the output data of the mapper unit 17 to the band of the received signal using a wide band as shown in FIG.

The frequency band EC31 is a one-tap complex filter prepared for each carrier.
From the transmission signal band-extended by 0, a replica of an echo generated by each received carrier is calculated. The calculated replica is subtracted from the signal demodulated by the FFT unit 27 in the reception unit 12, and the code identification unit 29 applies the demodulated signal to the carrier arrangement. At that time, in order to improve the error at the time of demodulation,
The 31 tap coefficients are adaptively updated. This is performed based on the identification error signal 52 calculated by the code identification unit 29. The tap coefficient of the frequency range EC is updated by a least square (LMS) algorithm based on the identification error signal 52. Let the tap coefficient of carrier n in symbol k be C
_{Assuming k and n} , the update of the tap coefficient is performed as in the following equation (5).

C _{k + 1, n} = C _{k, n} + μE _{k, n} X ^* _{k, n} (5)

Here, E _{k, n} is an identification error signal, μ is a step size of coefficient update, and X ^* _{k, n} is a complex conjugate of output data X _{k, n} of the mapper unit 17.

As described above, in the frequency domain EC31, the echo component generated by the current symbol can be eliminated by generating a replica from the transmission data based on the identification error signal 52 from the code identification section 29 in the frequency domain. .

The sampling rate differs between an upstream transmission signal and a downstream reception signal. This is because, as shown in FIG. 8, the transmission signal and the reception signal use different bandwidths. Therefore, in order to equalize the time length of one symbol, the transmission signal and the reception signal per symbol have different numbers of samples. It is necessary. For example, the uplink transmission signal has 64 samples excluding CP, and the downlink reception signal has 512 samples excluding CP. Therefore, the data interpolating unit 32 interpolates the time domain data converted into the time domain by the IFFT unit 18 into “0”, thereby interpolating the number of samples of the received signal in the down direction.

Assuming that the data of the sample n in the symbol k converted into the time domain by the IFFT unit 18 is y _{k, j} , data interpolation is performed as in the following equations (6) to (7).

Y ′ _{k, j} = y _{k, j / 8} (mod (j, 8) = 0) (6) y ′ _{k, j} = 0 (mod (j, 8) ≠ 0)・ (7)

Here, y ′ _{k, j} is output data after data interpolation. “Mod (j, 8)” is “j”
Is divided by "8".

FIG. 9 is for explaining the principle of operation of the data interpolation unit 32. FIG. 3A shows a transmission symbol before interpolation. FIG. 3B shows the data after interpolation. That is, as shown in FIG. 9A, 64 data of the transmission symbol 60 is expanded to 512 samples of the interpolated data 61 into the sample shown by the equation (6), and otherwise "0" is interpolated. Is done.

The time domain EC33 is constituted by an FIR filter. The tap coefficient is first set in the frequency range EC31.
Are obtained by band expansion as shown in the following equations (8) to (10).

C _{k, n} = 0 (n = 0, N / 2) (8) C _{k, n} = C _{k, n} (0 <n <N / 2) (9) C _{k , n} = C ^* _{k, Nn} (n> N / 2) (10)

After that, the tap coefficients of the time domain EC 33 are calculated by the IFFT of the IFFT section 34.

The time zone EC33 performs the following two preprocessings. The first pre-processing is performed by the receiving unit 12 in the subsequent frequency EC3.
Echo caused by past symbols that cannot be eliminated only by subtraction of the replica calculated by 1 is eliminated.
In the second preprocessing, an echo generated in a symbol following the current transmission symbol is copied to the head of the FFT section of the current symbol. By these two processes, only echoes generated only by the current symbol remain in the FFT section of the current symbol with periodicity. These residual echoes in the time domain are generated in the subsequent frequency domain EC
33 can be deleted by subtracting the replica calculated by the calculation.

FIG. 10 is a diagram for explaining two pre-processes performed by the time domain EC33. FIG. 3A shows a transmission signal. FIG. 2B shows an echo generated by the transmission signal. FIG. 3C shows a time-domain residual echo generated by two preprocessings. Here, description will be made focusing on the K-th symbol. As shown in FIG. 2A, the transmission signal is the (K-1) th symbol, the Kth
Symbol and the (K + 1) th symbol are chronologically continuous. A CP is added at the beginning of each symbol. As for CP _K−1 , CP _K , and CP _{K + 1} , samples B _K−1 , B _K , and B _{K + 1 corresponding} to L samples after each symbol are copied as described with reference to FIG. However, there is a case where the “swinging” due to the solitary wave response of the normal echo exceeds the sample length L of this CP. In such a case, an echo occurs in the following symbol.

Each symbol is represented by a CP, a sample B corresponding to the number of L samples to be copied of the CP, and a sample A other than the L samples.
Expressed as In FIG. 13B, for example, when focusing on the K-th symbol, CP _K is replaced by the past symbol (K−1).
The echo EA _K-1 by the sample A _K-1 symbols, echo EB _K-1 by the sample B _K-1 indicates the occurrence. Also, echo E due to CP _K of the K-th symbol
CP _K occurs up to sample A _K.

The FFT section to be demodulated is originally the range 62 of the samples A _K and B _K of the _K-th symbol. However, the hatched portion 63 due to EB _K-1 which occurs up to the _K-th symbol A _K is generated by B _K-1 of the (K-1) -th symbol which is a past symbol, so that the subsequent frequency range E
C31 cannot be erased. Therefore, erasing the hatched portion 63 is the first preprocessing described above.

Also, the CP _{K + 1 of} the (K + 1) th symbol
Generates an echo due to the sample B _K of the K-th symbol. Consequently, by performing a second pre-processing for copying the shaded portion 64 of the first (K + 1) EA _K generated in the symbol of the head of the sample A _K, echo EA _K symbols of the K is, FFT section 62 , And can be periodic. Incidentally, an echo ECP _K by CP _K symbols of the K, the echo EB _K by the sample B _K in the symbol of the (K + 1), since already the same is copied, there is no need to particularly perform copying.

Thus, in the time domain, the time domain EC3
In step 3, the echo caused by the past symbol and the echo caused by the current transmission symbol in the subsequent symbol are estimated. Further, in the frequency domain, an echo caused by the current transmission symbol is estimated in the frequency domain EC31. Then, by performing subtraction from the received signal in each area using these, the echo of the received signal caused by the transmission signal wrapping around through the hybrid transformer 13 is eliminated.

[0044]

As described above, in the conventional EC, in the applied ADSL system, FIG.
As shown in (2), the bands used by the uplink transmission signal and the downlink reception signal are different. So, in EC,
In order to make the band of the transmission signal and the band of the reception signal the same, the band is extended to efficiently cancel the echo, and the cutoff characteristic required for the low-pass filter in the reception unit 12 is reduced.

In order to calculate a replica of an echo by the current symbol in the frequency domain using the frequency domain EC31, output data obtained by band spreading the output data of the mapper section 17 in the transmission section 11 is used. Assuming that the data of the carrier number n in the symbol number k is X _{k, n} , the transmission data is band-extended by the following equations (11) to (13).

X ′ _{k, n} = 0 (n = 0, N / 2) (11) X ′ _{k, n} = X _{k, n} (n <N / 2) (12) X ′ _{k, n} = X ^* _{k, N-1-n} (n> N / 2) (13)

Here, “N / 2” is the number of transmission carriers,
X ′ _{k, n} is the data after expansion. In the used frequency band shown in FIG. 8, N is “64” and the number of transmission carriers is “3”.
2 ". Then, the band is extended by the following equations (14) to (17).

X ″ _{k, n} = X ′ _{k, n} (n <N) (14) X ″ _{k, n} = X ′ _{k, nN} (N ≦ n <2N) (15) _{) X'' k, n = X'k} , n-2N (2N ≦ n <3N) ··· (16) X'' k, n = X'k, n-3N (3N ≦ n <M / 2 ) (17)

Here, “M / 2” is the number of carriers of the received symbol, and X ″ _{k, n} is the data after expansion. In the used frequency band shown in FIG. 8, M is “512” and the number of transmission carriers is “256”.

FIG. 11 is a diagram for explaining the operation principle of band extension in the band extension section 30 shown by the equations (14) to (17). Here, the vertical axis shows the amplitude, and the horizontal axis shows the carrier number of X ″ _{k, n} after expansion. That is,
As shown by the equations (11) to (13), the first region 65
Is converted by "N" carriers and copied to the second area 66 according to equation (15) (copy 67). Further, the carrier is moved by “2N” carriers according to the equation (16) and copied to the third area 68 (copy 6
9). Further, it is moved by “3N” carriers according to the equation (17) and is copied to the fourth area 70 (copy 71).

As is apparent from the equations (11) to (17), when the carrier number n is "N / 2", the band-expanded data becomes "0". Therefore,
As defined by the expression (5), C _{k, N / 2} among the tap coefficients of the frequency range EC31 is not updated. That is, in the frequency range EC31, no carrier exists at the Nyquist frequency, and the corresponding tap coefficient is not updated. In the transmitting / receiving apparatus in the conventional ADSL system shown in FIG. 3, "0" is set every 32 taps.
IFFT is performed on the frequency response to calculate the tap coefficient of the time domain EC33. Therefore, the accurate time range EC31
There is a problem that the tap coefficient cannot be calculated. Therefore, in the time domain EC31, the echo generated by the past symbol cannot be accurately eliminated, and in the frequency domain EC31, this echo cannot be eliminated originally, so that the echo suppression characteristic deteriorates.

Accordingly, an object of the present invention is to provide an echo canceller that accurately cancels an echo that cannot be canceled by a transmission signal wrapping around to a receiving side via a hybrid transformer by a simple calculation.

[0053]

According to the first aspect of the present invention, there are provided (a) band extending means for extending a transmission band narrower than a receiving band to a receiving band, and (b) each band extended by the band extending means. Frequency domain echo calculating means for calculating an echo replica of the current symbol using tap coefficients updated based on the code identification result of the received signal from the transmission data of the carrier; and (c) tap coefficients of the frequency domain echo calculating means (D) inverse fast Fourier transform means for converting the tap coefficients of the frequency domain echo calculation means interpolated by the frequency interpolation means into time domain tap coefficients, and (e) band extension of the transmission data. Data interpolation means for interpolating the transmission data in the time domain converted by the post-inverse fast Fourier transform to the number of samples of the received signal; Time for eliminating echoes caused by past symbols from data interpolated by the data interpolating means using tap coefficients converted by the Rier transforming means, and calculating a replica for compensating for echoes occurring in subsequent symbols by the current symbol. The echo canceller is provided with an area echo calculating means.

That is, according to the first aspect of the present invention, the band extending means extends the band of the transmission data allocated to each transmission carrier to a wider receiving band, and the frequency band echo calculating means expands the code identification result of the received signal. The replica of the echo by the current symbol is calculated by using the tap coefficient updated based on. In the conventional case, the time domain echo calculating means uses tap coefficients obtained by converting the tap coefficients of the frequency domain echo calculating means into the time domain, and the data obtained by interpolating the transmission data to the sampling number of the received signal after band expansion. To eliminate echoes caused by past symbols and calculate a replica that compensates for echoes caused by the current symbol in subsequent symbols. However, in the present invention, after the tap coefficients of the frequency domain echo calculation means are interpolated by the frequency interpolation means, they are converted into the time domain tap coefficients by the inverse fast Fourier transform means, and the time domain echo calculation is performed using the converted tap coefficients. The replica of the echo is calculated by the means.

According to the second aspect of the present invention, (a) band extending means for extending a transmission band narrower than the receiving band to the receiving band, and (b) receiving from the transmission data of each carrier band extended by the band extending means. Frequency domain echo calculating means for calculating a replica of the echo by the current symbol using the tap coefficients updated based on the signal identification result of the signal; and (c) frequency domain echo calculating means corresponding to the Nyquist frequency of the transmission data Frequency interpolation means for interpolating tap coefficients; (d) inverse fast Fourier transform means for converting tap coefficients of the frequency domain echo calculation means interpolated by the frequency interpolation means into time domain tap coefficients; Data interpolation means for interpolating the transmission data in the time domain subjected to the inverse fast Fourier transform after the band extension to the sampling number of the received signal; ) Using the tap coefficients converted by the inverse fast Fourier transform means, cancel the echo caused by the past symbol from the data interpolated by the data interpolation means, and calculate the replica which makes the echo generated by the current symbol in the succeeding symbol. The echo canceller is provided with a time-domain echo calculating means.

That is, in the second aspect of the present invention, the frequency interpolation means according to the first aspect of the invention interpolates a tap coefficient of the frequency band echo calculation means corresponding to the Nyquist frequency of the transmission data. This allows
Since the transmission data of the carrier corresponding to the tap coefficient corresponding to the Nyquist frequency of the frequency domain echo calculating means is "0", the frequency response becomes "0", and the tap coefficient is not updated, so that the time domain echo calculating means is not updated. It is possible to solve the problem that the tap coefficient is not calculated correctly and improve the echo suppression characteristics.

According to a third aspect of the present invention, in the echo canceller of the second aspect, the frequency interpolation means interpolates a tap coefficient corresponding to the Nyquist frequency by taking an average value of tap coefficients of adjacent carriers. And

That is, according to the third aspect of the present invention, the average of the tap coefficients of the adjacent carriers is calculated and the tap coefficients corresponding to the Nyquist frequency are interpolated. With a very simple configuration, it is possible to avoid deterioration of the echo suppression characteristics.

According to a fourth aspect of the present invention, in the echo canceller according to any one of the first to third aspects, the frequency interpolation means converts the tap coefficient of the carrier corresponding to the Nyquist frequency of the received signal into the tap coefficient of the adjacent carrier and its complex coefficient. It is characterized in that the tap coefficient corresponding to the Nyquist frequency of the received signal is interpolated by taking the average value of the conjugate.

That is, according to the fourth aspect of the present invention, the carrier corresponding to the Nyquist frequency of the received signal is interpolated by taking the average value with the complex conjugate. When the noise spreads to the Nyquist frequency, the echo suppression characteristics can be further improved.

[0061]

BEST MODE FOR CARRYING OUT THE INVENTION

[0062]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below with reference to embodiments.

First Embodiment

FIG. 1 is a circuit diagram showing a first embodiment of the present invention.
1 shows an outline of a configuration of a transmission / reception device in an ADSL system to which C is applied. However, the same parts as those of the ADSL system to which the conventionally proposed EC shown in FIG. 3 is applied are denoted by the same reference numerals. Here, a transmission / reception device applied to an ADSL system having a wider reception band compared to a transmission band is shown. The transmitting / receiving apparatus to which the EC in the first embodiment is applied is connected to another opposing transmitting / receiving apparatus (not shown) via a line 10. A transmission unit 11 and a reception unit 12 are connected to the line 10 via a hybrid transformer 13. The line 10 is a metallic cable used for an existing telephone line. The hybrid transformer 13 performs two-wire to four-wire conversion by a transformer that couples the line 10 with the transmission unit 11 and the reception unit 12. As a result, a transmission signal from the transmission unit 11 is transmitted to the line 10 via the hybrid transformer 13.
Further, a reception signal from the line 10 is received by the receiving unit 12 via the hybrid transformer 13. However, at this time, since the echo of the transmission signal from the transmission unit 11 wraps around the reception signal input to the reception unit 12 via the hybrid transformer 13, the reception unit 12 is provided with an EC 80 for suppressing the echo. Have been.

The transmitting section 11 includes a serial-to-parallel converting section 16 for converting the serially input transmission data 15 into parallel data, and a mapper section 17 for arranging the parallel-converted transmission data on each carrier every several bits. To perform IFFT after band extension of allocated carrier
It includes a T unit 18 and a CP insertion unit 19 that adds a CP to the output of the IFFT unit 18. Further, the transmission unit 11
Is a parallel-to-serial converter 20 for converting a signal to which CP is added into serial data, and a digital-to-analog converter 21 for converting the converted serial data to analog.
And a low-pass filter 22 for blocking high-frequency components of the analog signal.

The receiving section 12 includes a hybrid transformer 13
Low-pass filter 23 that cuts off a high-frequency noise component of the reception signal input through the line 10, for example.
And an analog-to-digital converter 24 for digitizing the output, and an EC for the digitized output signal.
A subtractor 25 for canceling an echo caused by a past symbol estimated by 80, and a serial / parallel converter 2 for converting the subtraction result of the subtractor 25 into parallel
6 is provided. The receiving unit 12 further includes an FFT unit 27 that demodulates the converted parallel data by FFT.
A subtractor 28 for canceling an echo caused by the current symbol estimated by the EC 80 with respect to the FFT signal; and a code identification unit 29 for reconstructing the subtraction result of the subtracter 28 into binary data. ing.

The EC 80 extends the output of the mapper unit 17 to the band on the transmitting side to the band on the receiving side.
And a frequency band EC31 for calculating a replica of an echo generated by each carrier from the carriers band-extended by the band extending unit 30. Further EC80
Has a data interpolator 32 for interpolating the data converted to the time domain by the IFFT unit 18 and a time domain EC33 comprising an FIR filter and generating an echo replica from the data interpolated by the data interpolator 32. . Furthermore, the EC 80 in the first embodiment is a frequency interpolation unit 8 that interpolates the tap coefficients of the frequency band EC31.
1 and an IFFT unit 34 that calculates the tap coefficients of the time domain EC33 by IFFT of the tap coefficients interpolated by the frequency interpolation unit 81. Replicas calculated in the frequency domain EC31 and the time domain EC33, respectively,
The subtractor 28 and the subtractor 2 in the receiving unit 12 respectively
At 5, the echo is canceled in the frequency domain and the time domain.

In the first embodiment having such a configuration, E
In the transmission / reception device to which C is applied, the transmission data 15 which is time-series serial data is transmitted to the serial / parallel conversion unit 16.
Is converted to parallel data. The parallel data converted by the serial / parallel conversion unit 16 is bundled every several bits by the mapper unit 17 as shown in FIGS. 4 and 5, and is arranged on a plurality of carriers by QAM.

The output data of the mapper unit 17 arranged on each carrier is band-extended as shown in the above-mentioned equations (1) to (3), and then IFFT by the IFFT unit 18 to
It is converted to time domain data for each symbol.

The output data of the IFFT unit 18 is added with a CP by a CP insertion unit 19. Usually, each data is
The response characteristic of “0” and the low-frequency cutoff characteristic of the hybrid transformer 13 cause “swinging”. The CP is added to the head of each symbol in order to take in the "end of the data" that does not fit in the symbol into the own symbol.
At the same time, the CP aims to reproduce the periodicity at the time of FFT. Since the length of the CP to be added is set on the basis of the “semi-swing” in the received symbol, the processing is performed in the time domain EC.

The data to which the CP has been added by the CP insertion unit 19 is converted into serial data by the parallel / serial conversion unit 20. The serial data converted by the parallel-to-serial converter 20 is converted into an analog signal by a digital-to-analog converter 21, and after being removed by a low-pass filter 32, the harmonic component 22 is transmitted to the line 10 via the hybrid transformer 13.

On the other hand, the received signal input via the line 10 is input to the receiver 12 via the hybrid transformer 13. In the receiving unit 12, first, the harmonic noise component added on the line 10 is removed by the low-pass filter 23, and then digitized by the analog-to-digital converter 24. From the digitized output signal, the echo component generated by the past symbol estimated by the EC 80 is eliminated by the subtracter 25. The result of the subtraction by the subtracter 25 is converted into parallel data by the serial / parallel conversion unit 26, and is converted into data in the frequency domain by the FFT unit 27. From the data in the frequency range converted by the FFT unit 27, the echo component generated by the current symbol estimated by the EC 14 is eliminated by the subtractor 28. The subtraction result of the subtracter 28 is applied to the carrier arrangement shown in FIG. At this time, an error occurs due to the demodulation characteristics, and the corresponding identification error signal 52 is set to EC1.
4 is supplied. The data applied to the carrier arrangement shown in FIG. 5 is subjected to a conversion reverse to the assignment performed in FIG. 4 by a demapper (not shown) to reconstruct binary data.

The EC 80 cancels the echo caused by the transmission signal that passes through the hybrid transformer 13 because the transmission signal and the reception signal use the same band as shown in FIG. Hereinafter, the EC 80 will be described.

The band extending section 30 extends the output data of the mapper section 17 to the band of the received signal using a wide band as shown in FIG. This is because, as shown in FIG. 8, the transmission signal and the reception signal use different bands.

FIG. 2 shows an outline of the main components of the frequency range EC31. Frequency range EC81 is 1 and filter 85 ₁ to 85 _P of the first to P taps a tap coefficient provided for each carrier complex, code identification section 29
And a multiplier 86 for multiplying the discrimination error signal 52 generated by the above by the step size μ. First to first
The configurations of the P-th filters 85 _{1 to} 85 _P are the same. Since the operations of the _{second to} P-th filters 85 _{2 to} 85 _P are performed in the same manner, illustration and description are omitted, and only the first filter 85 ₁ will be described below. The transmission data 87 band-extended by the band extension unit 30 is
It is input to the first through filter 85 ₁ to 85 _P of the P for each carrier. The first to P-th filters 85 _{1 to} 85 _P include:
The result of multiplication of the identification error signal 52 multiplied by the multiplier 86 and the step size μ is input.

First filter 85₁1 character entered in
The transmission data for the rear is delayed by one symbol.
Extension element 88₁And multiplier 89₁Is input to To the multiplier 86
Therefore, the multiplied identification error signal 52 and the step size μ
Multiplication result and delay element 88 ₁Delay delayed by
Data and multiplier 90₁Is input to This multiplication result
Is the adder 91₁And the adder 91₁Addition of
The delay element 92₁Delay by one symbol
Is added to the data. Also, this adder 91₁Addition of
The result is a multiplier 89₁To the transmission data for one carrier.
Data. That is, as shown in the above equation (5)
Tap coefficient C_{k, n}To update. And this
Echo by current symbol with updated tap coefficients
Is calculated.

The adders 89 ₁ to 89 _P of the filters 85 _{1 to} 85 _P
The addition result from 89 _P is input to the subtractor 28 of the receiving unit 12. The subtracter 28 has a subtractor 28 ₁ ~ 28 _P for each carrier, the converted data 93 that has been converted into the frequency domain by the FFT unit 27, the addition result is subtracted. The result of this subtraction is output to the code identification unit 29 as received data 94.

The code discriminating section 29 applies the demodulated signal to the carrier arrangement. In order to improve the error due to the demodulation characteristics occurring at that time, the tap coefficient of the frequency band EC31 is adaptively updated. This is because the code identifying unit 29
This is performed based on the identification error signal 52 calculated by the above.

As described above, in the frequency domain EC31, the echo component generated by the current symbol can be eliminated by evaluating the replica from the transmission data in the frequency domain based on the identification error signal 52 from the code identification section 29. .

As shown in FIG. 8, since the transmission signal and the reception signal use different bands, the time length of one symbol is equal, and thus the number of samples per symbol is different. Therefore, the data interpolation unit 32 interpolates the time domain data converted into the time domain by the IFFT unit 18 by "0", thereby obtaining the downlink reception signal as shown in the above-described equations (6) to (7). Interpolate to the number of samples

The time zone EC33 is constituted by an FIR filter. As the tap coefficient, the tap coefficient of the frequency range EC31 is calculated by the frequency interpolation unit 81 in the following (18) to (18).
After the frequency interpolation as shown in the equation (20), the IFFT is performed by the IFFT unit 34 to calculate. The tap coefficient of the frequency band EC is c _{k, n} , and the input data of the IFFT unit 34 is c ′ _{k, n} .

C ′ _{k, n} = 0 (n = 0, M / 2) (18) c ′ _{k, n} = (c _{k, n−1} + c _{k, n + 1} ) / 2 (mod ( (n, N / 2) = 0, n <M / 2, n ≠ 0) (19) c ′ _{k, n} = c _{k, n} (mod (n, N / 2) ≠ 0, n <M / 2, n ≠ 0) (20)

Here, n is the tap number, N / 2 is the number of transmission carriers, and M / 2 is the number of reception carriers. As described above, the frequency interpolation unit 81 interpolates the tap coefficient corresponding to the Nyquist frequency by taking the average value of two adjacent tap coefficients.

The amount of operation involved in the interpolation by the frequency interpolator 81 expressed by the equations (18) to (20)
a, assuming that the shift operation amount is Ds, the following (21), (2)
It can be obtained from equation (2).

Da = 2 · M / N (21) Ds = 2 · M / N (22)

In the ADSL system described above, N is "6
Since “4” and M are “512”, the amount of calculation involved in the interpolation by the frequency interpolation unit 81 is 16 additions and 16 shifts.
Times. It can be said that this increase in the amount of calculation is a negligible amount of calculation when compared with the amount of calculation in the conventional EC.

The tap coefficients interpolated by the frequency interpolation unit 81 are subjected to IFFT, and the time domain EC33 performs the above-described two preprocessings using the calculated tap coefficients of the time domain EC33. That is, in the first pre-processing, the receiving unit 12 eliminates echoes caused by past symbols that cannot be eliminated only by subtraction of the replica calculated by the subsequent frequency EC31. In the second preprocessing, an echo generated in a symbol following the current transmission symbol is copied to the head of the FFT section of the current symbol. As shown in FIG. 10, the FFT of the current symbol is performed by these two processes.
In the section, a residual echo in a time domain in which only echoes currently generated only by symbols have periodicity remains. The residual echo in the time domain is eliminated by subtracting the replica calculated by the frequency domain EC33 in the subsequent stage.

As described above, the EC according to the first embodiment includes the frequency interpolation unit 8 for interpolating the tap coefficients of the frequency band EC31.
1 and IFFT using the interpolated tap coefficients.
The unit 34 calculates the tap coefficient of the time domain EC32. The frequency interpolation unit 81 interpolates the tap coefficient corresponding to the Nyquist frequency by taking an average value of two adjacent tap coefficients. As a result, the frequency response becomes “0”.
Since the tap coefficient corresponding to the Nyquist frequency at which the tap coefficient of the frequency range EC31 is not updated as a result is updated by interpolation, the correct tap coefficient of the time range EC32 can be obtained. become. In addition, since the interpolation is performed by taking the average value of adjacent tap coefficients, the echo suppression characteristic can be improved with a simple configuration.

Second Embodiment

In the EC according to the first embodiment, the tap coefficients in the frequency domain EC31 are interpolated by the frequency interpolation section 81 as shown by the equations (18) to (20), thereby increasing the tap coefficients in the time domain EC32. It was calculated accurately. However, in the EC in the second embodiment,
In equation (18), the M / 2-th carrier corresponding to the Nyquist frequency of the received signal is set to “0”, but this may be interpolated as in the following equation (23).

C ′ _{k, M / 2} = (c _{k, M / 2-1} + c ^* _{k, M / 2 + 1} ) / 2 (23)

Here, c ^* _{k, M / 2-1} is the complex conjugate of c _{k, M / 2-1} . Other configurations and operations are the same as those of the first embodiment.
The description is omitted because it is the same as that of the embodiment.

In the EC according to the second embodiment, the carrier corresponding to the Nyquist frequency of the received signal is also interpolated by taking the average value with the complex conjugate, so that the spectrum of the echo is When the noise spreads to the Nyquist frequency, the echo suppression characteristics can be further improved.

In the first and second embodiments, the tap coefficients corresponding to the Nyquist frequency are interpolated by averaging the tap coefficients. However, the present invention is not limited to this. By a trade-off between the calculation capability and the required calculation amount, it is possible to further improve the echo suppression characteristic by a known precise interpolation process.

[0095]

As described above, according to the present invention, since the tap coefficients of the time domain echo calculating means are obtained after the tap coefficients of the frequency domain echo calculating means are interpolated, for example, the Nyquist frequency of the transmission data is obtained. Since the frequency response of the tap coefficient corresponding to is “0”, the problem that the tap coefficient is not correctly calculated by the time-domain echo calculating means because the tap coefficient is not updated can be solved, and the echo suppression characteristic can be improved. .

Further, according to the third aspect of the present invention, the tap coefficients corresponding to the Nyquist frequency are interpolated by averaging the tap coefficients of the adjacent carriers, so that an increase in the required amount of calculation is suppressed. On the other hand, it is possible to avoid deterioration of the echo suppression characteristic with a very simple configuration.

Further, according to the fourth aspect of the present invention,
The carrier corresponding to the Nyquist frequency of the received signal is also interpolated by taking the average value with the complex conjugate, so if the spectrum of the echo spreads to the Nyquist frequency of the reception band, the echo suppression characteristics will be further increased. Can be improved

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an outline of a configuration of a transmission / reception device in an ADSL system to which EC is applied according to a first embodiment of the present invention.

FIG. 2 is a block diagram illustrating an outline of a main part of a configuration of a frequency range EC.

FIG. 3 is a block diagram illustrating an outline of a configuration of a transmission / reception device in an ADSL system to which a conventionally proposed EC is applied.

FIG. 4 is an explanatory diagram showing an example of carrier assignment in a mapper unit.

FIG. 5 is an explanatory diagram showing an example of a carrier in a mapper portion.

FIG. 6 is an explanatory diagram for explaining the concept of band extension.

FIG. 7 is an explanatory diagram for explaining the concept of a CP.

FIG. 8 is an explanatory diagram showing a used frequency band in a full-duplex communication system.

FIG. 9 is an explanatory diagram for explaining the operation principle of the data interpolation unit.

FIG. 10 is an explanatory diagram for explaining processing content in a time domain EC.

FIG. 11 is an explanatory diagram for explaining the operation principle of the band extension unit.

[Explanation of symbols]

 Reference Signs List 10 line 11 transmission unit 12 reception unit 13 hybrid transformer 14, 80 EC 15 transmission data 16, 26 serial / parallel conversion unit 17 mapper unit 18, 34 IFFT unit 19 CP insertion unit 20 parallel / serial conversion unit 21 digital / analog conversion unit 22, 23 Low-pass filter 24 Analog-to-digital converter 25, 28 Subtractor 27 FFT unit 29 Code identification unit 30 Band extension unit 31 Frequency domain EC 32 Data interpolation unit 33 Time domain EC 81 Frequency interpolation unit

Claims (4)

[Claims] 1. Band extension means for extending a transmission band narrower than a reception band to the reception band, and updated based on a code identification result of a received signal from transmission data of each carrier band extended by the band extension means. Frequency domain echo calculating means for calculating an echo replica of the current symbol using tap coefficients; frequency interpolating means for interpolating tap coefficients of the frequency domain echo calculating means; and the frequency domain echo interpolated by the frequency interpolating means. An inverse fast Fourier transforming means for converting the tap coefficient of the calculating means into a time-domain tap coefficient, and interpolating the time-domain transmission data obtained by converting the transmission data by the inverse fast Fourier transform after band expansion to the sampling number of the reception signal. Using data interpolation means and tap coefficients converted by the inverse fast Fourier transform means And time domain echo calculation means for calculating a replica for compensating for an echo generated in a succeeding symbol by the current symbol while eliminating an echo generated by a past symbol from the data interpolated by the data interpolation means. Echo canceller. 2. A band extending means for extending a transmission band narrower than a reception band to the reception band, and updated based on a code identification result of a received signal from transmission data of each carrier band extended by the band extension means. Frequency domain echo calculating means for calculating an echo replica of the current symbol using tap coefficients; frequency interpolating means for interpolating tap coefficients of the frequency domain echo calculating means corresponding to the Nyquist frequency of the transmission data; Inverse fast Fourier transforming means for converting tap coefficients of the frequency domain echo calculating means interpolated by interpolation means into time domain tap coefficients; and receiving the time domain transmission data obtained by performing inverse fast Fourier transform after band expansion of the transmission data. Data interpolating means for interpolating to the number of signal samplings, and the inverse fast Fourier transform means Time-domain echo calculating means for erasing echoes caused by past symbols from data interpolated by the data interpolating means using the converted tap coefficients and calculating a replica which makes echoes generated by the current symbol in subsequent symbols. An echo canceller comprising: 3. The echo canceller according to claim 2, wherein said frequency interpolation means interpolates a tap coefficient corresponding to the Nyquist frequency by taking an average value of tap coefficients of adjacent carriers. 4. The tap corresponding to the Nyquist frequency of the received signal by taking the average of the tap coefficient of the carrier corresponding to the Nyquist frequency of the received signal and the complex conjugate of the tap coefficient of the adjacent carrier. 4. The echo canceller according to claim 1, wherein coefficients are interpolated.