KR100947724B1 - Device for processing signal about cancellation of RFIradio frequency interference in VDSLvery high bit rate digital subscriber line modem - Google Patents

Abstract

The present invention is an amateur radio signal and AM (Amplitude) in a modem of a CAP (Carrierless Amplitude and Phase modulation) or QAM (Quadrature Amplitude Modulation) method of a very high bit rate Digital Subscriber Line (VDSL). Modulation (RFI) The present invention relates to an RFI cancellation signal processing device of a VDSL modem that predicts and removes multiple RFIs by broadcasting. The fact that an RFI caused by an amateur radio signal occurs irregularly within a specific band and a sine wave of an eigenvalue separation part By using the signal detection characteristics and the bandpass characteristics of the band pass filter, the position and number of RFIs are identified, and the identified information is sent to the adaptive notch filter to remove the multiple RFIs, thereby reducing the transmission error rate and transmitting the VDSL service. It prevents the speed drop and reduces the manufacturing cost by reducing the complexity in hardware implementation.

Description

Device for processing signal about cancellation of radio frequency interference (RFI) in very high bit rate digital subscriber line (VDSL) modem

1 is a diagram showing a signal band of a VDSL and a band of an amateur radio signal.

2 is a configuration diagram of a conventional VDSL modem.

3 is a configuration diagram of a conventional VDSL modem.

4 is a diagram illustrating a form in which a signal of a sine wave having a large signal power is superimposed on white noise.

5 is an example of incorrect RFI tracking of a notch filter due to improper initial value setting.

6 is an example of RFI tracking of a notch filter due to proper initial value setting.

7 is a block diagram of a signal processing apparatus of a VDSL modem according to the present invention.

8 is a detailed configuration diagram of FIG. 7.

9 is an example of a power density spectrum of data input to a receiving end of a VDSL modem.

10 is an example of a power density spectrum after the data of FIG. 9 passes through a bandpass filter according to the present invention.

FIG. 11 is a comparison diagram of distribution of eigenvalues separated after passing a bandpass filter according to the present invention with data input to a receiving end of a VDSL modem. FIG.

The present invention relates to an RFI cancellation signal processing apparatus of a high speed digital subscriber network modem, and more particularly, to a carrierless amplitude and phase modulation (CAP) of a very high bit rate digital subscriber line (VDSL). The present invention relates to an RFI cancellation signal processing apparatus of a VDSL modem that predicts and removes multiple radio frequency interference (RFI) by amateur radio signals and AM (Amplitude Modulation) broadcasting in a QAM (Quadrature Amplitude Modulation) modem.

VDSL, also known as high-speed Digital Subscriber Line (DSL), refers to a method of modulating and demodulating data at more than 10MB per second over copper, regardless of unidirectional or bidirectional.

VDSL is a high-speed wired communication standard proposed to provide real-time video and vivid high-definition video conferencing. It is suitable for telephone line service area of less than 2.74km, and can be up to 51.84Mbps on short line when implemented asymmetrically. have.

In the case of the VDSL, unlike the voice telephone bandwidth, the bandwidth used is wide, so that the AMFI signal and the amateur radio signal overlap with the band, and the RFI is ingressed to the receiving end of the VDSL.

Figure 1 shows the signal band of the VDSL and the band of the amateur radio signal, the band of the amateur radio signal is shown throughout the VDSL signal band.

Since these RFI signals have a large power, they cause a problem of saturating the A / D converter during analog-to-digital (A / D) conversion.

In addition, the equalization performance is degraded in the equalization step after the A / D conversion, thereby increasing the transmission error rate and thus significantly affecting the transmission speed and the length of the subscriber line.

In order to remove such RFI components, an equalizer with a large number of taps is required, and even if an equalizer with a large number of taps is used, the equalization performance to guarantee high-speed transmission is not achieved. It can't happen.

On the other hand, since RFI by AM broadcasting appears only at a fixed frequency, it is easy to predict and is not a big problem.

However, the RFI by amateur radio signal is very irregular when viewed from the receiver because the user of amateur radio signal moves from time to time.

In particular, the shortwave amateur radio signal uses SSB-SC (Single Side Band-Suppressed Carrier) modulation, so there is no signal power when there is no message transmission.

Therefore, the RFI due to the amateur radio signal appears very irregular at the receiving end, so it is very difficult to predict the RFI.

Accordingly, in order to remove or mitigate the effects of RFI, the DDS (Discrete Multitone) type VDSL modem solves this problem by not assigning a subchannel to a channel corresponding to an amateur radio signal band, and using an SCM (such as a QAM or CAP method). In the Single Carrier Modulation (DFE) method, the decision feedback equalizer (DFE) of FIG. 2 is used, or the adaptive notch filter (ANF) 50 of FIG. 3 is used together with the DFE. This eliminates RFI.

The DFE is an equalizer composed of two linear filters 10 and 30 and a crystal element 20 as shown in FIG. 3, and the forward filter (FFF) 10 removes the RFI, resulting in an RFI frequency. A deep null is generated, and the feedback filter (FBF) 30 relaxes the inter-symbol interface (ISI).

However, removing the RFI with such a DFE is somewhat effective when the number of RFIs is small, but when multiple irregular RFIs occur at the same time, the equalization process becomes very difficult.

As shown in FIG. 3, the analog notch filter 40 and the adaptive notch filter 50 are further provided in FIG. 3 to remove the RFI with the analog notch filter 40 in the analog domain, and the adaptive notch filter 50 in the digital domain. Remove RFI with).

The purpose of removing the RFI from the analog domain is to prevent overloading of the A / D converter during A / D conversion, and since the RFI of the power is large enough to affect the equalization performance even after the RFI is removed from the analog domain. Removing RFI from the digital domain.

 The adaptive notch filter 50 adaptively finds the position of the RFI and attenuates only the vicinity of the predicted RFI frequency to remove the RFI.

This improves the performance of the downstream DFE and reduces the transmission error rate.

However, this notch filter-equalizer structure is also effective when a few RFIs are generated, but it is difficult to cope with a large number of RFIs.

When multiple RFIs occur, the number and location must be known accurately.

If the predicted number of RFIs is smaller than the actual number, the influence of the unremoved RFIs is greatly affected. If the predicted number of RFIs is larger than the actual number, the signal is attenuated unnecessarily by the redundant adaptive notch filter. In addition to accurately predicting numbers, it is important to know their location.

In the case of the adaptive notch filter, it is possible to adaptively track the position of the RFI.

However, as shown in FIG. 4, a sinusoidal signal with good signal power is superimposed on white noise, but a signal of the sinusoid is well traced, but RFI signals of various sizes overlap on a relatively large signal force such as a VDSL signal. If not, the adaptive notch filters track the RFI with the highest power, or if the power is similar, then track the RFI located in the center of the VDSL signal, making it difficult to track the other RFIs correctly.

In particular, when the RFIs are located very close together, it is difficult for the adaptive notch filter to accurately track and remove all RFI signals.

FIG. 5 is an example of simulation results of RFI tracking performance of the adaptive notch filter when the initial RFI prediction value of the adaptive notch filter is arbitrarily provided. FIG. 5 shows amateur radio waves of 7.0 to 7.1 MHz, 10.1 to 10.15 MHz, and 14.0 to 14.35 MHz. When an amateur radio signal with a center frequency of 7.1 MHz, 10.11 MHz, 14.10 MHz and 14.25 MHz in the signal band and a bandwidth of 4 KHz enters the receiver of the VDSL modem, the initial RFI prediction frequency of the adaptive notch filter can be arbitrarily set to 12.96 MHz. The result is when set to.

The RFI of 14.10 MHz and RFI of 14.25 MHz were tracked by the adaptive notch filter but not exactly as shown in the figure. Also, the three adaptive notch filters tracked two RFIs in the 14.0 to 14.35 MHz band. RFIs corresponding to MHz and 7.1 MHz show no convergence and misconverge near 14.0 to 14.35 MHz or near the center frequency.

FIG. 6 shows an example of simulation results of tracking performance when the initial predicted value of the RFI of the adaptive notch filter is set to the frequency at which the RFI actually occurs.

In FIG. 5, the adaptive notch filters randomly set the initial RFI prediction frequency and fail to adaptively track the RFI, while in FIG. 6, the adaptive notch is known after knowing the band in which the RFI occurred and how many RFIs occurred in the band. Adaptive notch filters track RFIs correctly when the initial prediction frequency of the filter is set to the center frequency of the band where the RFI occurs.

Here, the RFI frequency tracking convergence time is also fast, and it can be seen that exactly four RFIs are tracked.

That is, in the case of the VDSL signal, it is difficult to track and remove the RFI unless an appropriate RFI initial prediction value is set in the adaptive notch filter.

On the other hand, other methods for determining the location and number of RFIs include a method of finding spectral peak values using a periodgram method and a high-resolution spectral estimation method using or without a parameter.

First, the method of finding the spectral peak is to use Fast Fourier Transform (FFF) .In case of spectral prediction using Fast Fourier Transform, two sine waves of the same size are located closer than the inverse of the data length. Difficult to do

Therefore, since a large number of data is required to increase the resolution, there is a problem of time delay in determining the location and number of RFIs.

The high resolution spectrum prediction method requires more than five to six stages of prediction processing such as data observation, formation of autocorrelation matrix, inverse of autocorrelation matrix, and linear prediction algorithm.

Therefore, there is a problem of real time processing and hardware complexity.

In order to solve the above problems, an object of the present invention is to predict and remove multiple RFIs by amateur radio signals and AM broadcast in a CAP or QAM type modem of VDSL.

Another object of the present invention is to predict and remove the RFI of a VDSL modem by using a bandpass filter (BPF) and singular value decomposition (SVD).

To this end, the RFI removal signal processing apparatus of the ultra-high speed digital subscriber network modem according to the present invention, after converting the channel signal input to the ultra-high speed digital subscriber network modem to digital and using the independent sine wave component included in the converted signal sequence An RFI predictor for predicting whether an RFI exists for each frequency band of the RFI; An adaptive notch filter for removing the RFI predicted by the RFI predictor; And a decision feedback equalizer for equalizing the distortion of the signal at the output of the adaptive notch filter, wherein the RFI predictor comprises: a bandpass filter for passing only the band of a specific frequency at which the RFI occurs; A matrix constructing unit configured to convert the signal data output from the bandpass filter into parallel data to form a matrix; An eigenvalue separation unit for separating eigenvalues by separating sine wave characteristics from a matrix constructed by the matrix configuration unit; An eigenvalue alignment unit for aligning the eigenvalues separated by the eigenvalue separation unit in size order; And a comparator for activating the adaptive notch filter by comparing the eigenvalues aligned in the eigenvalue alignment unit with a threshold value.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

7 is an overall configuration diagram of a signal processing apparatus of a VDSL modem for predicting and removing the location and number of multiple RFIs according to the present invention.

The channel signal s (t) input to the VDSL modem is mixed with the RFI signal i (t) and the white noise n (t) and input to the A / D converter as r (t).

The signal r (t) is converted into a digital signal by the A / D converter and input to the adaptive notch filter stage 60 and the RFI predictor 70 by x (n) 80.

The number of predicted columns of the RFI predictor 70 matches the number of bands of the amateur radio signal.

The band pass filter (BPF) 71 of each predictive column filters the amateur radio signal with the same frequency as the center frequency of the amateur radio signal band and with a slightly larger bandwidth.

The signal 72 passing through the bandpass filter 71 is separated from the eigenvalue separator (SVD) 74, and the separated eigenvalues are arranged in the order of magnitude in the eigenvalue sorter 75 to the comparator 76. Is entered.

The comparator 76 is compared with a predetermined threshold value, which will be described later, to determine whether the adaptive notch filter stage 60 is activated according to the presence or absence of an eigenvalue exceeding the threshold value.

FIG. 8 illustrates a portion of the RFI predictor 70 and the adaptive notch filter stage 60 of FIG. 7 in more detail.

The band pass filter 71 used for the band pass is preferably a low order filter in the form of Infinite Impulse Response (IIR) in order to simplify hardware.

This is because a phase distortion, a round off error, a quantization error, and the like are not a problem.

The input signal sequence of x (n) input to the bandpass filter 71 at time n is {x (n), x (n-1), x (n-2),... }, A vector X consisting of N signals is defined as follows.

X_n ≡ (x (n), x (n-1), x (n-2),…, x (n-N + 1))^T                          <Equation 1>                     

Where T is the pre-operator.

In addition, the present invention relates to S / P & M (Serial to Parallel Conversion and Matrix) (73) in order to minimize the number of operations instead of the autocorrelation matrix Y = E {X (n) X (n) ^T } of Equation 1. ) Constructs a data matrix like

<식 2>

<Equation 2>

That is, S / P & M 73 converts serial data into parallel data to form a matrix.

행렬을 이용하여 식 3과 같이 고유치를 분리한다. The eigenvalue separator (SVD) 74 then becomes

Using the matrix, separate the eigenvalues as shown in Equation 3.

<식 3>

<Equation 3>

의 왼쪽 고유 벡터(Left Singular Vector)라고 하고, V는

의 오른쪽 고유 벡터(Right Singular Vector)라고 하며, ∑는 식 4와 같이

의 고유치를 대각 행렬로 갖는 정방행렬이다.U in Equation 3

Is called the Left Singular Vector, and V is

Is called the right singular vector of, and ∑ is

Square matrix with eigenvalues of in diagonal matrix.

<식 4>

<Equation 4>

Eigenvalues may represent independent sinusoidal components included in input data x (n).

If the real input data x (n) is composed of p sine wave signals in white noise, the eigenvalues are aligned as shown in Equation 5 in the eigenvalue alignment unit 75.

<식 5>

<Equation 5>

은 백색 잡음에 해당하는 고유치로 크기가 0에 가깝다.double

Is an eigenvalue that corresponds to white noise and is close to zero in magnitude.

That is, as described above, the RFI predictor 70 constructs a matrix of data (x (n)) input from the S / P & M 73 and uses the same to separate the eigenvalues from the eigenvalue separator (SVD) 74. By doing this, the characteristics of the spectrum of the entire input signal can be known.

However, since the band pass filter 71 is not passed, the number of RFI inlets included in the input signal can be known, but it is not known in which position the RFI has occurred.

Therefore, the present invention includes a bandpass filter 71 corresponding to a band of an amateur radio signal in each prediction column of the RFI predictor 70, thereby inputting the input data x (n) to the bandpass filter 71. After passing through, S / P & M (73) constructs a matrix and separates the eigenvalues to find out how many RFIs occur in the band.

In other words, FIG. 9 is an example of the case where four RFI incomings 100 to 130 are generated, which are signals input to the bandpass filter 71, and the signal passing through the bandpass filter 71 is as shown in FIG. 10. You can see that there is only a signal of that bandwidth.

FIG. 11 is a diagram illustrating an eigenvalue distribution of eigenvalue separation results after band-passing the input signal of FIG. 9 as shown in FIG. 10.                     

If there are p independent real number signals in the input signal, 2p eigenvalues appear when the eigenvalues of the signals are separated. In FIG. 11, at least eight eigenvalues are separated when four RFI incomings occur. Eigen values greater than zero appear.

However, there are also more than 16 eigenvalues greater than zero, which are represented by VDSL signals with band limited large power.

Referring to the drawings in detail, the eigenvalue 140 represents the eigenvalue of the signal of FIG. 9 input to the bandpass filter 71. As a result of four RFI ingresses 100 to 130 occurring, eight of the eigenvalue index (Index of Singular Values), namely, Up to λ ₈ may be determined as the RFI entry, but since the RFI predictor 70 does not know the number of RFI entries, for example, λ ₉ at eigenvalue 140 is significantly different from λ ₇ and λ ₈ as eigenvalues of the fourth RFI. It is difficult to predict how many RFIs have occurred.

On the other hand, the eigenvalue 150 is the eigenvalue distribution after passing through the bandpass filter 71 when two RFIs 120 and 130 are generated at intervals of about 100 KHz as shown in FIG. 9 in the signal band 14.0 to 14.35 MHz. The eigenvalue 160 represents the eigenvalue distribution after passing through the bandpass filter 71 when one RFI 100 is generated as shown in FIG. 9 in the signal band 7.0 to 7.3 MHz, and the eigenvalue 170 represents the signal. 9 shows the eigenvalue distribution after passing through the bandpass filter 71 when one RFI 110 occurs in the band 10.1 to 10.35 MHz.                     

In addition, the eigenvalues 180 and 190 show distributions of eigenvalues when RFI does not occur in the signal bands 3.5 to 4.0 MHz and 21.0 to 21.45 MHz, respectively. It can be seen that is not large.

Thus, looking at the eigenvalues 150 to 190 of the band pass signal, it is easy to know whether RFI occurs as follows.

First, it can be seen that the magnitude of the eigenvalue corresponding to the signal component of the VDSL corresponding to λ ₉ or more that is not the RFI component at the eigenvalue 140 is greatly reduced as shown in the eigenvalues 150 to 190 after the band pass.

In the case of the eigenvalues 160 and 170 of the band-passed signal, the difference between the eigenvalues λ ₁ and λ ₂ corresponding to one RFI and the eigenvalues λ ₃ or more corresponding to the VDSL transmission signal increases. You can do it easily.

The eigenvalues 180 and 190 are relatively evenly distributed, indicating that the eigenvalues of the bandpassed VDSL signal components and other noise components are distributed.

On the other hand, the eigenvalue 150 is a case where it is ambiguous whether one or two RFIs are introduced. The difference between the eigenvalue λ ₄ of the second RFI and the eigenvalue λ ₅ of the VDSL signal is larger than the eigenvalue 140, but there is a big difference. This can lead to errors in prediction.

To solve this problem, it is necessary to calculate a sophisticated threshold value that determines whether RFI is generated.

The threshold value is selected as the eigenvalue by passing eigenvalue separation after passing through the bandpass filter outside the band of the RFI, or as an average value of eigenvalues according to the eigenvalue separation result in the amateur radio signal band where RFI is not surely generated. .

After aligning the eigenvalues resulting from the eigenvalue separation in the eigenvalue alignment unit 75, the comparator 76 compares the threshold values in turn and counts the number of eigenvalues by the number of eigenvalues exceeding the threshold value. The control activates the adaptive notch filter 61.

In other words, if none of the eigenvalues exceed the threshold value, the function of the adaptive notch filter 61 is disabled through the selection terminal 90. If the number of eigenvalues exceeding the threshold value is Q, the eigenvalues are applied to the real signal. 2 times, the Q / 2 adaptive notch filters 61 are enabled.

At this time, the initial prediction frequency of the adaptive notch filter 61 is determined as the center frequency of the band of each amateur radio signal.

Then, as shown in FIG. 6, the RFI can be tracked and removed without being greatly affected by the parameters of the adaptive notch filter.

As described above, the present invention utilizes the fact that the RFI caused by the amateur radio signal occurs irregularly only within a specific band, the sinusoidal signal detection characteristic of the eigenvalue separation section and the pass characteristic of the specific band of the bandpass filter, By eliminating multiple RFIs by identifying the number and sending the information to the adaptive notch filter, it can reduce the transmission error rate, prevent the transmission speed of VDSL service, and reduce the manufacturing cost by reducing the complexity in hardware implementation. .

In addition, it is possible to estimate the RFI faster than the multiple RFI estimation method by other spectrum prediction methods.

Claims (4)

RFI prediction unit for predicting the existence of RFI for each frequency band of each RFI (Radio Frequency Interference) using independent sinusoidal components included in the converted signal sequence after converting the channel signal inputted to the high speed digital subscriber network modem into digital ; An adaptive notch filter for removing the RFI predicted by the RFI predictor; And Crystal feedback equalizer for equalizing the deformation of the signal at the output of the adaptive notch filter RFI removal signal processing device of a high speed digital subscriber network modem comprising a. The method of claim 1, The RFI predictor, A bandpass filter for passing the channel signal through only a band of a specific frequency at which an RFI occurs; A matrix constructing unit configured to convert the signal data output from the bandpass filter into parallel data to form a matrix; An eigenvalue separation unit for separating eigenvalues by separating sine wave characteristics from a matrix constructed by the matrix configuration unit; An eigenvalue alignment unit for aligning the eigenvalues separated by the eigenvalue separation unit in size order; And A comparator for activating the adaptive notch filter by comparing the eigenvalues aligned in the eigenvalue alignment unit with a threshold value RFI removal signal processing device of a high speed digital subscriber network modem comprising a. The method of claim 2, The RFI predictor, RFI of the high speed digital subscriber network modem including the adaptive notch filter, the bandpass filter, the matrix constructing unit, the eigenvalue separating unit, the eigenvalue sorting unit, and the comparator in a number corresponding to the number of frequency bands of the RFI. Elimination signal processing device. The method of claim 2 or 3, The comparator, And releasing the adaptive notch filter according to the number of eigenvalues exceeding the threshold value by comparing the eigenvalues arranged in the eigenvalue alignment unit with the Munter value.

Images