JP4569591B2 - Adaptive canceling method of sneak wave in OFDM-based SFN repeater - Google Patents

Description

  The present invention relates to a canceling method for removing sneak waves and multipath interference waves in an OFDM (orthogonal frequency division multiplexing) SFN (single frequency network) relay apparatus. The present invention is characterized by a method for estimating information on dynamic characteristics of a propagation path in real time, which is indispensable when applying an adaptive signal processing technique to an OFDM communication system.

  In recent years, the OFDM system has attracted attention as a multicarrier modulation method in high-speed wireless communication, wireless network, or multimedia communication. OFDM is resistant to multipath interference, and is adopted in terrestrial digital television broadcasting, ADSL (asymmetric digital subscriber line), and 5 GHz band wireless LAN. The OFDM signal may be transmitted through SFN to improve frequency utilization efficiency.

  Although SFN is excellent in terms of frequency utilization efficiency, there is a problem that the quality of a transmission signal is deteriorated due to a wraparound phenomenon between transmission / reception antennas of a relay apparatus (relay station). In particular, when the loop gain of the coupling becomes 1 or more, there is a problem that signal increase and oscillation are likely to occur. Therefore, a system for canceling the sneak wave is required at the relay station.

In order to construct a canceller for a sneak wave, a sneak path characteristic C (z) at the relay station and a multipath characteristic G (z) from the transmission station to the relay station are required. The relay transfer characteristics are greatly influenced by the estimation result at the relay station. Known identification methods for OFDM channel characteristics include a blind identification algorithm based on the subspace method (see Non-Patent Document 1), an identification algorithm using periodic stationarity, and the like.
B. Muquet, M. Courville, and Duhamel, `` Subspace-based blind and semi-blind channel estimation for OFDM systems, '' IEEE Trans.Signal Processing, vol.50, no.7, pp.1699-1712,2002 R. Heath and GB Giannakis, `` Exploiting input cyclostationarity for blind channel identification in OFDM systems '' IEEE Trans.Signal Processing, vol. 47, no. 3, pp. 848-856, 1999

  However, the above-described known means cannot be applied to the estimation of the sneak path characteristic in the actual OFDM digital communication SFN relay station. Actually, the transmission signal in the OFDM transmission system is limited to a finite band so that adjacent transmission channels do not interfere with each other, and the power spectrum outside the band of the input / output signal is zero. This is because it is extremely difficult to identify out-of-band frequency characteristics only from the information of the limited signal.

In order to stabilize the system when the out-of-band transmission line information is not known, a method of reducing the influence of the out-of-band frequency characteristics being unclear using a band-pass filter has been proposed ( Patent Document 1). However, the influence of the phase distortion of the bandpass file on the stability of the system cannot be ignored, and the filter design is not simple.
Japanese Patent Laid-Open No. 2003-060616

  In order to construct a sneak wave canceller at a relay station, sneak path characteristics at the relay station and multipath characteristics from the transmission station to the relay station are required. The performance of the canceller, such as stability and convergence, is greatly affected by the estimation result of the relay transfer characteristics at the relay station. Further, the transmission characteristics of the relay station including the sneak path is a closed loop system, and it is important to stabilize the entire relay transmission characteristic including the sneak wave canceller.

  However, in the presence of band limitations and noise, not only the estimation of the relay transfer characteristics but also the estimation of its uncertainty, the estimation of the out-of-band transfer characteristics and the evaluation of the uncertainties are not limited to the stability of the sneak canceller. This is an important issue in guaranteeing robustness.

  An object of the present invention is to provide an adaptive canceling method for removing a sneak wave at a relay station and a multipath interference wave from a transmission station to the relay station. In particular, a new frequency domain identification algorithm for estimating out-of-band propagation path characteristics, especially for band-limited OFDM propagation path characteristics, has been clarified to suppress sneak waves and multipath interference waves from the transmitting station to the relay station. An object is to provide an adaptive canceling method.

而して、誤差Ｅ（ｅ ^ｊωｎ ）のパワーを表す評価関数Ｑが最小値となるようにキャンセラ更新値が決定される。
４．２：（数７３）によって、時間域の更新量を計算する。
４．３：（数７６）によって、ステップサイズを求める。
４．４：（数７５）によって、キャンセラを更新する。
ステップ５：Ｌ＝Ｌ＋１とし、ステップ１に戻る。
の各ステップを有することを特徴とするＯＦＤＭ系ＳＦＮ中継装置における回り込み波の適応キャンセリング方法である。 The present invention for solving the above-described problems is a canceling method for removing sneak waves and multipath interference in a relay apparatus of a single frequency network (SFN) of an orthogonal frequency division multiplexing (OFDM) transmission system, Next step:
Step 0: An initial value of the canceling characteristic W (z) is given and the number of sequential updates is set to L = 0. Step 1: In the (L + 1) th update, R (L, n) is changed from the received signal r (k). Calculate with FFT. (Since it is difficult to distinguish l (lowercase el) from the Arabic numeral 1, the uppercase el L is used in this specification. The same applies hereinafter.)
Step 2: Estimate transmission path characteristics.
2.1: The propagation path characteristics in the band are estimated by (Equation 40), and the transmission signal D (L, n) is restored by (Equation 41).
2.2: Estimate out-of-band propagation path characteristics by (Equation 44) or (Equation 45).
2.3: The sensitivity characteristic is obtained by (Equation 53) and (Equation 54).
2.4: Multipath characteristics are obtained by (Equation 55).
Step 3: The upper limit of the sensitivity characteristic estimation error is evaluated by (Equation 77).
Step 4: Update canceller
4.1: The canceller update value in the frequency domain is calculated by (Equation 69). Here, E (e ^jωn ) in step 4.1 ( ^Equation 69) is a value when ω = ω _n of the error signal E (e ^jω ) defined in ( ^Equation 57). It can also be rewritten.

Accordingly , the canceller update value is determined so that the evaluation function Q representing the power of the error E (e ^jωn ) becomes the minimum value.
4.2: The amount of update in the time domain is calculated by (Expression 73).
4.3: The step size is obtained by (Equation 76).
4.4: The canceller is updated according to (Equation 75).
Step 5: L = L + 1 and return to Step 1.
This is an adaptive canceling method for sneak waves in an OFDM SFN repeater characterized by comprising the following steps.

  In the present invention, since the OFDM channel characteristics, especially the out-of-band channel characteristics, are estimated by the frequency domain identification algorithm using the band-limited OFDM guard interval characteristics, the stability and convergence of adaptive signal processing are improved. The calculation processing time (calculation load) can be reduced to 1/5 to 1/10 as compared with the case where the out-of-band propagation path characteristics are estimated using the time domain identification algorithm.

The sampling time of the OFDM signal and the normalized discrete time k, the normalized discrete time L th transmission period m, the frequency of the n th sub-carrier and normalized each frequency omega _n. k and m are integers, ω _n = 2πn / M, and M is a window length of Fourier transform. Thus, in the L transmission symbol period, the OFDM baseband signal d (L, m) transmitted from the base station is Fourier-transformed.

Indicated by Where D (L, n) is an information symbol placed on the nth subcarrier,

Given in. _{However, N 1 = (N-1} ) / 2, N is the number of carriers. In multilevel modulation, D (L, n) is _{N D} × _N constellation set of _{D {a 1 + jb 1,} a 2 + jb 2, ---, a ND + jb 1, ---, a ND + jb ND} Elements. Assuming that D (L, n) is a uniformly distributed probability sequence for L, n,

Holds. However, δ (·) is a delta function.

In the OFDM transmission method, a guard interval that is resistant to distortion due to multiple propagation of radio waves is added to the beginning of a symbol interval. If the guard interval length is m _gi , the signal d (L, m) of m = −m _gi , ---, −1 is a signal sequence of the guard interval, and the tail m = M−m _gi , − of the effective signal -, Equal to the signal of M-1. Therefore, the actual transmission symbol period is M + m _gi , and the normal discrete time k = (L−1) (M + m _gi ) + m _gi + m. Also, a pilot signal is usually inserted into the OFDM signal for synchronization and demodulation of the received signal. _P frequency arrangement subcarriers _omega, putting a _n, omega _P, the signal at _n can be used as known information.

In OFDM transmission, N <M is set so as not to interfere with adjacent channels. Effective symbol period in other words m = 0, ---, in M-1 of the interval, the power spectrum of d (L, m) is | n | are limited to ≦ _{N 1} in the region. Since the power spectrum outside the band of the transmission signal d (L, m) is almost zero, the input / output signal related to the transmission signal d (L, m) contains almost no out-of-band signal information and relays signals outside the band. It is very difficult to identify characteristics.

  The transmission / reception signal when there is an interference wave due to multiple propagation of radio waves,

It can be expressed as Here, it is assumed that d (k) is a transmission signal, v (k) is a noise term, and is white Gaussian noise having an average value of 0 independent of the transmission signal d (k). hτ is an attenuation coefficient corresponding to the delayed wave of the delay tap τ , and L _h is an effective delayed wave tap length.

  FIR (finite impulse response) model for propagation path dynamics

In terms of the frequency characteristic R (e ^jω ) from the transmission signal from the base station, d (k) to the reception signal r (k),

Given in. Here, H (e ^jω ) is the signal relay characteristic, D (e ^jω ) is the spectrum of the base station signal in the band, and V (e ^jω ) is the noise frequency characteristic.

When synchronization between the reception signal r (k) and the transmission signal, d (k) is obtained, a Fourier transform R (L, n) of the reception signal r (L, m) having an L symbol period is obtained. Band words, | n | Equation 6 (6) When ≦ _{N 1} is

It can be rewritten as Here, D (L, n) is an information symbol carried on the nth subcarrier, and V (L, n) is a noise term. However, out-of-band words, | n |> is D (L, n) when N ₁ for = a 0, R (L, n) another valid information used for channel dynamics identified other Furthermore, it must be found from the received signal.

  In order to obtain information about the out-of-band signal transmission path, as shown in FIG.

Defined in Also, x (L, m) and y (L, m) are rewritten from k = (L-1) (M + m _gi ) + m _gi + m. Due to the nature of the guard interval,
It can be seen that x (L, m) = 0 when −m _gi ≦ m <0.
Let K be an integer _satisfying m _gi <K + m _gi ≦ M. When x (L, m) is represented by a transmission symbol D (L, n),

Can be written.

Next, consider the nature of the signal y (k). The multi-propagation delayed wave yτ (k) corresponding to the delay tap τ and the attenuation coefficient hτ is

And yτ (L, m) is

And y (L, m) is

Given in. In order to simplify the description, the noise term is omitted here.

On the other hand, the frequency characteristics of the signals x (L, m), yτ (L, m) and y (L, m) in the section 0 ≦ m <K are

, The frequency characteristics of the signals x (k) and y (k) are

Meet. However,

It is.

  The next spectrum is

The spectra XX (L, n), XE _{τ, l} (L, n) and XE _{τ, 2} (L, n) are given by the following equations. | N | spectrum XX (L, n) When ≦ _{N 1} is

And when N ₁ <| n | <M / 2,

Can be approximated by On the other hand, the spectrum XE _{τ, l} (L, n) is

It is. The spectrum XE _{τ, 2} (L, n) when | n | ≦ N ₁ is

When N ₁ <| n | <M / 2,

Given in.

  The spectrum of signals x (k) and y (k) and the frequency characteristics of the propagation path are given by

  Meet. Equation 30 (Equation 30) is obtained from Equation 17 (Equation 17).

In addition,

Is proved to hold.

According to Equation 30 (Equation 30), XX (L, n) and XY (L, n) are obtained from the signals x (k) and y (k). XE _{τ, l} (L, n) and XE _{τ, 2} (L, n) are spectral distortions, and depend on transmission / reception signals and transmission path characteristics according to Equation 27 (Equation 27) and Equation 28 (Equation 28). Can be calculated in advance. Accordingly, if the spectrum XY (L, n) is compensated using the spectra XE _{τ, l} (L, n) and XE _{τ, 2} (L, n), it is possible to capture out-of-band information.

On the other hand, the signal relay characteristic H (e ^jω ) also changes with time as the propagation path characteristic changes. Therefore, the forgetting factor is introduced into Equation 20 (Equation 20) and Equation 22 (Equation 22), and the spectra XX (L, n) and XY (L, n) are

Equation 30 (Equation 30) also holds. Where the forgetting factor λ is
It is a constant of 0 <λ <1.

  According to Equation 30 (Equation 30), the frequency characteristic of the propagation path is

Can be approximated by The right side of Expression 35 (Equation 35) includes the true value of the attenuation coefficient hτ . When the propagation path dynamic characteristics are actually identified by Equation 35 (Equation 35), the estimated value, for example, the following equation is used instead of the attenuation coefficient hτ.

を代入すると、伝搬路特性の推定値は、 Fourier inverse transform sequence of H (e ^jωn ) estimated in

When substituting, the estimated channel characteristics are

Can be estimated. When using the forgetting factor, the right side of Equation 37 (Equation 37) is

And the estimated value of the propagation path characteristic is obtained.

Identification of propagation path dynamic characteristics will be described.
First, identification of propagation path dynamic characteristics within a band will be described. Suppose that the scattered pilot is arranged on subcarriers of frequency ω _{P, n} . When synchronization between the reception signal r (k) and the transmission signal, d (k) is obtained, a Fourier transform R (L, n) of the reception signal r (L, m) having an L symbol period is obtained. In pilot signal subcarriers P _n , D (L, P _n ) is known from pilot symbol information,

Thus, an estimated value of the propagation path characteristic is obtained. Since the scattered pilot signal is uniformly arranged in the signal band | n | ≦ N ₁ , the propagation characteristics within the band are obtained by interpolation within the signal band. That is, the following linear interpolation

Can be interpolated. However, P _{n, 1} and P _{n, 2} are numbers of adjacent pilot carriers, and P _{n, 1} ≦ n ≦ P _{n, 2} . Also, using the estimated channel characteristics, D (L, n) and d (L, m)

In can and restore child.

Next, identification of propagation path dynamic characteristics outside the signal band will be described. The estimated value of H (e ^jωn ) is

の推定値と歪スペクトルＸＥ_ｌ（Ｌ，ｎ）を用いて、 Given in

And the distortion spectrum XE _l (L, n)

Thus, out-of-band propagation path dynamics can be estimated. Also, with forgetting factor,

Can be estimated.

  The configuration of the SFN relay station is shown in the figure. Here, r (k) is a received signal, and s (k) is a retransmission signal from the relay station. G (z) is a propagation path model from the base to the relay station, and has multipath interference. On the other hand, due to buildings and obstacles around the relay station, the retransmitted radio wave from the transmitting antenna reaches the receiving antenna, causing sneak interference mixed with the received signal from the base station, and the signal transmission quality deteriorates. As shown in FIG. 2, a wraparound interference model is C (z).

The sneak wave that has passed through the amplifier causes oscillation when the gain becomes 1 or more due to the coupling between the transmitting and receiving antennas of the relay station, and threatens the stability of the closed loop system including the sneak path. Therefore, it is necessary to suppress the effects of multipath interference and sneak waves. In FIG. 2, W ₁ (z) and W ₂ (z) are FIR type cancellers with tap numbers L ₁ and L ₂ , respectively. At this time, the propagation path model is

It becomes. here,

It is. In order to guarantee the stability of the canceling system, it is necessary to design W ₁ (z) and W ₂ (z) so that all the poles of the sensitivity characteristic Γ (z) are arranged in the unit circle, and the uncertainty The characteristics of the propagation path in the entire frequency range without this are indispensable.

It is assumed that the transmission symbol D (L, n) is 64QAM and the number of subcarriers N = 1405 is an OFDM transmission system. Using scattered pilots, pilot symbols are inserted into 118 subcarriers in each transmission symbol period. Effective symbol length T = 252 μs, guard interval T _cp = T / 4, FFT (fast Fourier transformation) / IFFT (inverse fast Fourier transformation) size M = 2048, K = M / 2, and sampling frequency 8. The frequency is 126984 MHz. The DU ratio between the transmission wave from the master station and the sneak wave

Defined in Here, σ ² _d and σ ² _ci are the powers of the desired signal and the i-th sneak signal, respectively. The multipath DU ratio is defined in the same way. Under the above settings, simulations in the following two cases were performed.

Identification of propagation path characteristics The case where a receiving antenna of a relay station receives five sneak waves is taken up. The delay times are 0.37 μs, 0.62 μs, 0.74 μs, 0.98 μs, and 1.35 μs, respectively, and the DU ratios are 3.3 dB, 1.4 dB, 26.8 dB, 4.7 dB, and 28.2 dB, respectively. Assume that there is. The DU ratio of multipath interference is 3.87 dB, and the delay time is 1.11 μs. The sneak path characteristic and the multipath path characteristic are expressed as follows.
C (z) = 0.6817e ^j0.3243π z ⁻³ + 0.85e ^{jπ / 6} z ⁻⁵ + 0.0457e ^{−j11π / 12} z ⁻⁵ + 0.5795e ^−j0.5091π z ⁻⁸ + 0.0388e ^{jπ / 4} z ^-11
G (z) = 1 + 0.64e ^{jπ / 3} z ⁻⁹

FIG. 3 shows the propagation characteristics estimated sequentially 50 times by the propagation path characteristic identification method ( ^Equation 35) to ( ^Equation 38) based on the spectrum when the frequency characteristic of the propagation characteristic H (e ^jωn ) and the SN ratio = 25 dB. As can be seen from FIG. 3, even if there is a band limitation in the OFDM signal, an estimated value close to the true propagation characteristic can be obtained by the present invention. In addition, as shown in FIG. 4, although the transmission signal d (k) has a severe band limitation in which the power density outside the band is zero, the spectrum of x (k) It can be seen that it has a spectral density of 10 ⁻³ of the density amplitude. Using these pieces of information makes it possible to identify propagation dynamic characteristics outside the signal band.

  Estimate errors in propagation characteristics outside the signal band

The estimation error of the propagation characteristic H (e ^jωn ) in various noise environments is shown in FIG. As is clear from FIG. 5, even if the SN ratio is low and the error rate of the restored signal D (L, n) is high, the estimation error is reduced to 10 ⁻¹ or less by the dozen successive estimations.

Application to adaptive interference canceling The tap length L _{1 of the} FIR type canceller W ₁ (z) is set to 100 and wraparound interference and multipath interference are canceled using W ₂ (z) = 1. Here, since the overall propagation characteristics of the relay station are time-varying due to the update of the canceller, the forgetting factor is set to λ = 0.9. In order to construct a stable canceling system, the closed-loop characteristic equation F (z) is identified. From Figure 2

Holds. Here, v ₁ (k) is a noise term. From the nature of the guard interval, it can be seen that x (L, m) = 0 for k where −m _gi ≦ m <0. then,

AR (auto regression) model as shown below, and the coefficient of the polynomial F (z) can be estimated using y (k). The estimated polynomial is F (z), and the sensitivity function Γ (e ^jωn ) is

Given in. Further, when u (k) = F (z) y (k) is set and U (L, n) and XU (L, n) are calculated in the same manner as in (Equation 16) and (Equation 34), Approximate expression

Is established, and the estimated value G (e ^jωn ) of the out-of-band multipath characteristic can be obtained by the propagation path characteristic identification method using the spectrum. Further, using the estimation results of the relay propagation characteristic H (e ^jωn ), the sensitivity characteristic Γ (e ^jωn ), and the multipath characteristic G (e ^jωn ), the FIR canceller W ₁ (z) is obtained by the following adaptive canceller design method. Can be updated.

The estimation of the relay station model is performed sequentially for every two adjacent signal symbol periods, and the obtained relay transfer characteristics, multipath characteristics, and sensitivity function estimates are respectively H (e ^jωn ) and G (e ^jωn ). , Γ (e ^jωn ), and the upper bound of the estimation error of the estimated value Γ (e ^jωn ) is | ΔΓ (e ^jωn ).

  When the transfer characteristic is the minimum phase, the interference wave can be removed by the canceller shown in FIG. Here, it is considered that the canceller W (z) is updated so as to minimize the canceling error in the entire frequency range. An evaluation function for updating

And set. Here, ω _{P, k} is the frequency of the k-th pilot signal, ω _{H, k} is the frequency outside the signal band, and ^* is the complex conjugate. G If (z) is the minimum phase, frequency range of the error ^{E (e} jω) is restored desired signal ^{D (e} jω) and the retransmission signal ^{S (e} jω) relative error ^{1-D (e} jω) between / S (e ^jω ) and transfer characteristic G (e ^jωn ). When updating the actual canceller, E (e ^jω )

Define in. The frequency characteristic of the closed-loop transfer function from the master station signal d (k) to the retransmit signal s (k) is

Given by. Here, H (z) is called the system transfer characteristic from the master station signal d (k) to the retransmit signal s (k), and Γ (z) is called the sensitivity characteristic.

Defined in W (z) is the canceller characteristics, an FIR filter of a tap number L _omega. According to Expression 57 (Equation 57), an error signal can be calculated with the frequency characteristics of the desired signal and the retransmit signal within the band. From Equation 58 (Equation 58), E (e ^jω ) is

Can be written. On the other hand, since out-of-band signal information cannot be obtained, an out-of-band error signal is calculated from the estimated values of the multipath characteristics and the sensitivity function. From the definition of Γ (e ^jω ), E (e ^jω ) is

It can be expressed. In any case, if W (e ^jω ) satisfies the optimal solution W ^opt (e ^jω ) = C (e ^jω ) K (e ^jω ) + G (e ^jω ) −1, the evaluation function Q is minimized. However, in order to simplify the discussion, the noise component is omitted in Equation 61 (Equation 61) and Equation 62 (Equation 62). From Equation 59 (Equation 59), Equation 60 (Equation 60), Equation 56 (Equation 56), and Equation 57 (Equation 57), when H (z) = 1, Q is the minimum value. That is, W ₁ (z) that satisfies H (z) = 1 is the optimum canceller.

Consider that the canceller W ₁ (z) is adaptively updated so as to minimize the evaluation function Q of Expression 56 (Equation 56). The evaluation function Q is a function of the frequency characteristic W (e ^jωk ) of W (z), and is sequentially updated by a gradient method using a gradient related to W ₁ (e ^jωn ) of Q. If the gradient for W ₁ (e ^jωn ), which is a complex number, is W ₁ (e ^jωn ) = α _n + jβ _n , the partial differentiation for real part α _n and imaginary part β _n is used,

Can be defined as Using this gradient, W ₁ (e ^jωn ) can be sequentially updated by the following gradient method.

Here, μ ^{(L + 1)} is a step size for the ^{(L + 1)} th update. The specific calculation of the gradient ▽ Q (W (e ^jωn )) is performed as follows. From Equation 61 (Equation 61)

に置き換え It becomes. However, since G ^* (e ^jωn ) is unknown, its estimated value

Replace with

Calculate with On the other hand, the gradient at the out-of-band frequency ω _{H, n} is expressed by Equation 62

Given in. From the above, the gradient for sequentially updating W ₁ (e ^jωn ) is

Can be calculated by Now, assuming that the step size is constant, that is, μ ^{(L + 1)} = μ, from an equation 66 (Equation 66) and an equation 68 (Equation 68), under an ideal situation where the estimation error and noise of the path model can be ignored, It can be seen that if 0 <μ <2, the stability is guaranteed, and if μ = 1, the convergence speed of the square error can be optimized. Thus, the absolute value maximum update amount ΔW ^{(L + 1)} (e ^jω ) in the ^{(L + 1)} th time when μ ^{(L + 1)} = 1 is given by the following equation.

However, in practice, it is difficult to realize step size μ ^{(L + 1)} = 1 that maximizes the convergence speed in the presence of an estimation error or noise of frequency characteristics of sensitivity functions in and out of the band. Therefore, a method for ensuring the convergence by selecting the maximum step size within a range where the adaptive algorithm operates stably will be described below.

If the update vector of the canceller weight in the time domain is Δω ^{(L + 1)} , Δω ^{(L + 1)} and ΔW ^{(L + 1)} (e ^jω ) have the following relationship:

Meet. Here, the vector matrix representation at the frequency points ω = ω _{P, 1} , ---, ω _{P, NP} , ω _{H, 1} , ---, ω _{H, NH} is

It can be expressed. However,

It is. Therefore, ΔW ^{(L + 1)} is given by

Can be calculated. Here, the superscript ^H represents Hermitian transpose, and Ω is the following matrix.

Λ (ω) is a diagonal weight matrix, and since the estimated value in the band is more accurate than the estimated value outside the band, the weight element corresponding to ω _{P, n} is 1, and the weight corresponding to ω _{H, n} The element is set to a value smaller than 1, for example 0.5. Ω is able to be calculated in advance, there is no need to one by one calculation at the time of renewal.

Since the column vectors of the matrix Φ (e ^jω ) include frequency components of low and high frequencies, each column vector is close to an orthogonal vector. In particular, when the frequency intervals of ω _{P, 1} , ---, ω _{P, NP} , ω _{H, 1} , ---, ω _{H, NH} are equal, the column vector of Φ (e ^jω ) becomes an orthogonal vector. , Φ ^H (e ^−jω ) Λ (ω) Φ (e ^−jω ) has sufficient regularity, so that Ω can be easily calculated with high accuracy. On the other hand, since the frequency characteristic outside the band cannot be used in the conventional method, Φ (e ^jω ) has ω _{P, 1} , ---, ω _{P, NP} , ω _{H, 1} , ---, ω _{H, NH} It contains only information about frequency points. Therefore, the condition number of the corresponding matrix Φ ^H (e ^−jω ) Λ (ω) Φ (e ^−jω ) becomes very large, and Δω ^{(L + 1)} is strongly influenced by disturbances and estimation errors. From the above, in the present invention, the coefficient vector of the (L + 1) -th canceller W ₁ (z) is

Updated by Here, the step size μ ^{(L + 1)} for stably updating the canceller can be given by the following equation. Assume that the estimated value of the frequency characteristic of the sensitivity function is Γ ^{(L + 1)} (e ^jω ), and the upper limit of the estimation error is | Γ ^{(L + 1)} (e ^jω ) |. The step size μ ^{(L + 1)} for updating the stable canceller W ₁ (z ⁾ is

It is decided by.

FIG. 6 shows the frequency characteristics after compensation by 50 iterations according to the present invention. As is apparent from FIG. 6, characteristics close to 1 are obtained even outside the band. FIG. 7 plots the average DU ratio obtained by 20 simulations. The inventors have published “The Journal of the Institute of Electronics, Information and Communication Engineers” A Vol. J89-A
No. 10 pp. Stable convergence close to the time domain algorithm proposed in 1-13 and a high DU ratio were obtained, and the execution time was about 1/6. In the present invention, since the propagation characteristics are identified in the entire frequency range, stable and good convergence can be obtained because the canceller can be configured adaptively by selecting a large step size within a range that guarantees stability. .

FIG. 8 shows the bit error rate characteristics at the 50th iteration. As is clear from FIG. 8, both the present invention using the frequency domain algorithm and the time domain algorithm by the inventors have good wraparound interference canceling performance.

Summarizing the steps of the method for canceling the sneak wave in the OFDM SFN repeater of the present invention,
Step 0: An initial value of W ₁ (z) is given, and the number of sequential updates is set to L = 0.
Step 1: In the (L + 1) -th update, R (L, n) is calculated from the received signal r (k) by FFT.
Step 2: Estimate propagation path characteristics.
2.1: In-band propagation path characteristics are estimated by Expression 40 (Equation 40), and the transmission signal D (L, n) is restored by Expression 41 (Equation 41).
2.2: Estimate out-of-band propagation path characteristics using Equation 44 (Equation 44) or Equation 45 (Equation 45).
2.3: The sensitivity characteristic is obtained by Expression 53 (Expression 53) and Expression 54 (Expression 54).
2.4: Multipath characteristics are obtained by Equation 55 (Equation 55).
Step 3: The upper limit of the sensitivity characteristic estimation error is evaluated by the following formula.

ステップ４：キャンセラの更新
４．１：式６９（数６９）により周波数域におけるキャンセラ更新値を計算する。
４．２：式７３（数７３）により、時間域の更新量を計算する。
４．３：式７６（数７６）により、ステップサイズを求める。
４．４：式７５（数７５）により、キャンセラを更新する。
ステップ５：Ｌ＝Ｌ＋１とし、ステップ１に戻る。

Step 4: Update canceller
4.1: The canceller update value in the frequency domain is calculated using Equation 69 (Equation 69).
4.2: The amount of update in the time domain is calculated using Equation 73 (Equation 73).
4.3: The step size is obtained by Expression 76 (Equation 76).
4.4: The canceller is updated by Expression 75 (Equation 75).
Step 5: L = L + 1 and return to Step 1.

Schematic diagram illustrating a signal x (k) according to an embodiment of the present invention. The figure which shows the system of the relay station which has a canceller concerning one Example of this invention The graph which shows the propagation characteristic of the relay station estimated 50 times sequentially using this invention The graph which shows the spectrum of the signal x (k) based on one Example of this invention. FIG. 5 is a graph showing an estimation error of propagation characteristics in various noise environments according to an embodiment of the present invention. The graph which shows the frequency characteristic after compensating by 50 iterations of the present invention The graph which plots and shows the average DU ratio obtained by 20 simulations based on one Example of this invention Graph showing the error rate after 50 iterations of the present invention

Claims (1)

A canceling method for removing sneak waves and multipath interference in a single frequency network (SFN) repeater of an orthogonal frequency division multiplexing (OFDM) transmission system, comprising the following steps:
Step 0: An initial value of the canceling characteristic W (z) is given and the number of sequential updates is set to L = 0. Step 1: In the (L + 1) th update, R (L, n) is changed from the received signal r (k). Calculate with FFT.
Step 2: Estimate transmission path characteristics.
2.1: In-band propagation path characteristics

: Estimated values of relay transfer characteristics P _{n, 1} , P _{n, 2} : Estimated by the numbers of adjacent pilot carriers, and the transmission signal D (L, n)

R (L, n): Restoration is performed by Fourier transform of the received signal.
2.2: The out-of-band channel characteristics

: Signal spectrum

: An estimated value of the attenuation coefficient corresponding to the delayed wave of the delay tap τ, or

λ: Estimated by forgetting factor.
2.3: Sensitivity characteristics

F (z): closed-loop characteristic equation

: Determined by the estimated polynomial.
2.4: Multipath characteristics

Ask for.
Step 3: Set the upper limit of sensitivity characteristic estimation error.

: Estimated sensitivity characteristics

: Evaluated by multipath characteristics.
Step 4: Update canceller
4.1:

^{Here, E (e} jωn) is a value when a omega = omega _n of the error signal ^{E (e} jω), in a band, the restored information symbol

And relative error between retransmission signal S (e ^jωn )

^And the estimated value of the multipath characteristic G (e ^jωn ), outside the band, using the estimated value of the multipath characteristic G (e ^jωn ) and the sensitivity characteristic Γ (e ^jωn ),

Calculate by Then , the canceller update value is determined so that the evaluation function Q representing the power of the error signal E (e ^jω ) becomes a minimum value.

: Absolute value maximum update amount

: A canceller update value in the frequency domain is calculated according to the gradient for successive updates.
4.2:

: Update vector of canceller weight

:line; queue; procession; parade

: Calculate the update amount of the time domain by the diagonal weight matrix.
4.3:

To obtain the step size.
4.4:

: Update the canceller according to the step size.
Step 5: L = L + 1 and return to Step 1.
A method for adaptive cancellation of sneak waves in an OFDM SFN repeater, characterized by comprising the steps of:

Images