JP2006245810A - Fraction spacing equalizer and receiver using it - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a fraction spacing equalizer which carries out the over sampling technique of the received signal K times (K≥2), and which can perform least square deviation reference equalizing. <P>SOLUTION: Equalizing is performed by using a formula of a least square deviation references weight based on the received signal by which the oversampling technique is carried out at rate K times (K≥2) of the density of symbol. λ<SB>i</SB>is the transfer function of the transmission path here, σ<SB>n</SB>is the variance of a noise, σ<SB>s</SB>is the variance of the signal, and "*" represents about a complex conjugate. Consequently, in fraction spacing equalizing, the characteristic of the least square deviation references wait which is not influenced of the noise reinforcement is acquired. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a fractional interval equalizer that can correct transmission path characteristics by oversampling a received signal and equalizing using all of these sample values.

  Block transmission using a cyclic prefix (CP) such as an OFDM (Orthogonal Frequency Division Multiplex) scheme or a single carrier block transmission scheme (SCBT) has attracted attention. In the block transmission scheme using CP, efficient equalization can be performed using a frequency domain equalizer for frequency selective fading.

The equalizer weight when performing frequency domain equalization can be represented by a diagonal matrix Γ. A diagonal component of the diagonal matrix Γ is represented by {γ ₁ ,..., Γ _M }. Here, M represents the number of discrete frequencies, that is, the FFT size.
Conventional equalizer weights include zero forcing (ZF) based weights. The ZF equalizer weight is
γ _i = 1 / λ _i (i = 1, ..., M)
It is represented by Here, λ _i is a frequency transfer function of the transmission line.

The ZF equalization has a problem of noise enhancement.
Noise enhancement is a phenomenon in which when a transfer function λ _i of a transmission line takes a value of 0 or a value close to 0 at a certain frequency, the weight γ _{i at} that frequency takes a very large value and the noise is amplified. It is. As a result, the receiver cannot accurately restore the received signal, and the BER characteristics are deteriorated.

On the other hand, a fractional interval equalization method has been proposed in which oversampling is performed twice on the receiving side and all sample values are equalized (see Non-Patent Document 5). This makes it possible to obtain receiver performance that is strong against frequency selective fading.
US Patent Publication No. 2004/0076239 A1 K. Hayashi and S. Hara, “A New Spatio-Temporal Equalization Method Based on Estimated Channel Response,” IEEE Transactoins on Vehicular Technology, Vol. 50, No.5, p.1250-1259, 2001. Mitsuo Yokoyama "Spread Spectrum Communication System" Science and Technology Publishers p.393, 6.3 PN series J. Cioffi and JAC Bingham, “A Data-Drien Multitone Echo Canceller,” IEEE Transactions on Communications, Vol.42, No.10, p.2853-2869, 1994. Kazunori Hayashi “Fundamentals of Modulation / Demodulation and Equalization Technologies” Proc. MWE2004, pp.523-532, 2004 PP Vaidyanathan and B. Vrcelj, “Theory of fractionally spaced cyclic-prefix equalizer,” Proc. ICASSP, vol.2, pp.1277-1280, 2002.

As an improvement of the ZF equalizer, a weight of a minimum mean-square-error (MMSE) is known. The least square error criterion equalizer weight is
γ _i = λ ^* _i / (| λ _i ² | + σn ² / σs ² ) (i = 1, ..., M)
It is represented by Here, σ _n represents noise variance, σ _s represents signal variance, and * represents complex conjugate.
According to this least square error reference equalizer, although the circuit configuration is complicated, the problem of noise enhancement is alleviated and an excellent BER (bit error rate) characteristic of the receiver can be obtained.

Therefore, when this least square error reference equalization is used and oversampling is performed on the receiving side, it is considered that a very excellent receiver can be configured that suppresses the influence of noise enhancement.
However, it is not known how to set the least square error reference equalizer weight when oversampling.
SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a fractional interval equalizer capable of performing effective least square error reference equalization when a received signal is generally oversampled K times (K ≧ 2). .

The fractional interval equalizer of the present invention includes means for oversampling a received signal at a density K times (K ≧ 2) a symbol rate, means for Fourier transforming the oversampled received signal, and the received signal. Means for estimating the transfer function λ _i (i = 1,2, ..., KM) of the transmission line based on the known signals contained in the signal, and the variance of the received signal power and the noise power of the receiver Means for estimating dispersion, and frequency domain equalization means for obtaining an estimated value of the transmission signal, the frequency domain equalization means comprising: a transfer function λ _{i of the} transmission line; a dispersion of received signal power; and a receiver. The equalization is performed using the least square error reference weight including the variance of the noise power.

According to this configuration, by performing equalization using the least square error reference weight based on the received signal oversampled at a density K times the symbol rate (K ≧ 2), the fractional interval equalization is performed. The characteristic of the least square error reference weight that is not affected by noise enhancement can be applied.
The principle of the fractional interval equalizer of the present invention in the case of oversampling twice will be described with reference to FIG. Time is represented by t and frequency is represented by f.

  FIG. 2A shows how the received signal is sampled at the symbol rate. FIG. 2B shows the signal after Fourier transform. The signal after the Fourier transform is actually a complex number, but is drawn like a real number in the graph. FIG. 2 (c) shows a conventional equalizer weight. This is actually a complex number, but it is drawn like a real number in the graph. FIG. 2D shows a signal obtained by multiplying the signal after the Fourier transform by the equalizer weight. FIG. 2E shows a time signal waveform obtained by performing inverse Fourier transform on this signal.

  On the other hand, in the present invention, as shown in FIG. 1A, the received signal is oversampled at twice the symbol rate. The signal is Fourier transformed (FIG. 1 (b)), and a weight based on the least square error is applied (FIG. 1 (c)). FIG. 1 (d) shows a signal obtained by multiplying the signal after the Fourier transform by the equalizer weight. Since the size of the Fourier transform is doubled, each signal spreads twice on the frequency axis. These signals are added to obtain a signal as shown in FIG. If this signal is subjected to inverse Fourier transform to obtain a time signal waveform, the result is as shown in FIG.

In the above example, the oversampling has been performed twice. However, in general, the case of oversampling K times (K ≧ 2) can be similarly indicated.
As described above, according to the present invention, it is possible to perform equalization which is excellent in BER characteristics and hardly causes noise enhancement.
The receiver of the present invention further includes an inverse Fourier transform unit that restores a signal that has been frequency domain equalized by the fractional interval equalizer to a time domain signal. Thereby, in single carrier block transmission or the like, it is possible to configure a receiver that suppresses the influence of noise enhancement and is less susceptible to the influence of transmission channel fading.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
FIG. 3 is a block diagram of a single carrier block transmission system (hereinafter referred to as an SCBT system). This transmission method includes a transmitter side, a transmission line, and a receiver side.
First, the processing on the transmitter side will be described in mathematical form.

  The transmission signal symbol s (n) is blocked every M. n is a discrete time.

A pilot signal is inserted into each block of the transmission signal s _e (n) that has been blocked.
For example, FIG. 3 of Patent Document 1 shows an example in which a Pilot signal is inserted between CP and data. In FIG. 3 of Non-Patent Document 1, an example of superimposing a suppressed pilot channel on a data channel is shown. In this document, a PN sequence is used as a signal transmitted through a pilot channel.

This method is also applied to the single carrier block transmission of this embodiment.
Next, a cyclic prefix (CP) copied from the last part of the block at the head is added.
Next, a preamble (Preamble) is generated. Examples of the preamble include a PN (Psuedorandom Noise) signal series and a chirp signal. Here, see Non-Patent Document 2 for details of the PN signal. The chirp signal is a “sine wave whose frequency increases linearly”, and the generation method is described in Non-Patent Document 3, p.2866.

The preamble generated in this way is added to the block. The preamble and block added in this way are called “frame”.
FIG. 4 shows a specific example of the frame configuration. In FIG. 4, one preamble is attached to a plurality of blocks to form one frame. However, one preamble may be attached to one block to form one frame.

This frame is modulated to a radio frequency and transmitted to the transmission path.
Next, processing on the receiver side will be described.
FIG. 5 is a flowchart showing the flow (outline) of reception processing.
The received signal is A / D converted to detect the beginning of the frame (step S1). Next, a block (including a preamble) is cut out from the frame (step S2).

It is determined whether the extracted block is a data block or a preamble (step S3). If it is a preamble, a discrete Fourier transform is performed on the preamble signal (step S5). Based on the signal obtained by the discrete Fourier transform, the frequency transfer function λ _i (i = 1,..., KM) of the transmission path is estimated (step S6).
On the other hand, if it is a data block, the data part is extracted from the block, and the cyclic prefix CP is removed from the data (step S8).

Next, a pilot signal is extracted from the data (step S9). The pilot signal is subjected to discrete Fourier transform (step S10), and the frequency transfer function λ _i is updated using the result of the discrete Fourier transform (step S7). This is because the transmission path is expected to change even after estimating the frequency transfer function of the transmission path when receiving the preamble.
Next, the received signal block after cyclic prefix removal is subjected to discrete Fourier transform (step S11).

Then, a frequency domain equalization process is performed based on the frequency transfer function (if updated, the updated frequency transfer function) (step S12). Frequency domain equalization is a process of multiplying a weight Γ for each frequency component in the transform domain and returning the signal to the time domain signal again by discrete inverse Fourier transform. Details of this will be described later.
Finally, discrete inverse Fourier transform is performed on the equalized signal (step S13) to convert it into a time function. Then, this time signal is determined (step S14).

The contents of the reception process described above will be described in detail using mathematical expressions.
Assuming that the cyclic prefix length is _LCP , the transmission signal block s _e (n) to which the cyclic prefix is added is as follows.

However,

Matrix size: (KM + L _CP ) × KM
In and, 0 _{LCPx (KM-LCP)} is the zero matrix of _{L CP × (KM-L CP} ), I LCP represents respectively a unit matrix of L _CP × L _CP.
The impulse response of the transmission line is [h ₀ , h ₁ , ..., h _Lh ], and the receiver noise is n (n) = [n ₀ (n), ..., n _KM-1 (n)] ^T Then the received signal block is

It is represented by However, the matrix H representing the convolution with the impulse response of the transmission path is as follows.

The matrix size of H is (KM + L _CP ) × 2 (KM + L _CP ).
Further, H is a submatrix of (KM + L _CP ) × (KM + L _CP )

The received signal block is

Matrix size: (KM + L _CP ) × 1
Can be written. Here, the first term on the right side is a signal component from the (n−1) -th transmission signal block, and represents a component of inter-block interference (IBI).
On the receiving side, the cyclic prefix is removed (step S8). This is expressed as follows:

The matrix size is KM × 1. However,

It is. If L _CP ≧ L _h at this time, R _CP H ₁ = 0 _KMxKM regardless of the transmission signal,

It becomes. Now, when R _CP H ₀ T _CP of this equation is expanded, it becomes as follows.

A matrix having such a structure is called a circulant matrix (Circulant Matrix), and unitary similarity transformation is possible using a discrete Fourier transform (DFT) matrix. See Non-Patent Document 4 for details.
Using the nature of the cyclic matrix,

Can be written. Where Λ is a diagonal matrix with λ ₀ , λ ₁ , λ ₂ , ..., λ _KM-1 as diagonal components,
λ ₀ , λ ₁ , λ ₂ ,..., λ _KM-1 are calculated as follows.

Thus, the received signal r (n) after cyclic prefix removal can be written as follows.

The receiver of the present invention includes a fractionally spaced equalizer (FSE) 1.
When performing A / D, the fractional interval equalizer 1 samples (oversamples) the waveform at a rate that is an integral multiple of the transmitted signal (in the case of 2 ⁿ times, the calculation amount can be reduced by using DFT as the FFT). The ISI (intersymbol interference) is reduced by performing the equalization process.

Non-Patent Document 5 discloses a technique for equalizing a fractional interval of a signal using a cyclic prefix.
The case where oversampling is performed at twice the signal speed on the receiving side will be described below. In this case, the received block vector r (n) corresponding to s _e (n) is

It is represented by
The equalization process in the fractional interval equalizer 1 is expressed by the following equation.

Where A is a constant, Γ ₁ and Γ ₂ are equalizer weights, and “H” is Hermitian transpose.

(k = 1, 2)
D _M is a DFT matrix of size M and can be expressed as follows.

  The case where oversampling is performed at twice the signal speed has been described above. In general, the output signal of the equalizer when oversampling at K times is given by the following equation.

In the present invention, as a weight of the equalizer, a method for calculating and deriving a minimum square error reference weight at the time of oversampling K times is newly shown.
The least square error reference weight is expressed by the following equation.

Here, σ _s ² is a variance value of the signal, and σ _n ² is a variance value of the noise.
FIG. 6 is a block diagram showing the flow of processing of the fractional interval equalizer 1 of the present invention.
The fractional interval equalizer 1 includes the equalizer weights Γ ₁ Γ ₂ . . . A weighting processing unit 11 for applying Γ _K and an adding unit 12 are included. The input of the weighting processing unit 11 after Fourier transform that has been oversampled by K times is denoted as X ₁ , X ₂ ,..., X _KM (X _i when represented). The weighting processing unit 11 multiplies the least square error reference weights γ ₁ , γ ₂ ,..., Γ _KM . The adding unit 12 adds the outputs of the weighting processing unit 11. That is, the adder 12 divides the output of the weighting processor 11 into groups of M signals as shown in FIG. 6, and adds the i (i = 1,..., M) -th elements of each group. Process. Accordingly, the output of the adder 12 is a vector of size M.

By performing inverse Fourier transform on this output, the transmission signal s (n) can be accurately restored.
Although the embodiments of the present invention have been described above, the embodiments of the present invention are not limited to the above-described embodiments. For example, the above equation [Equation 20] shows an equation on the assumption that the receiver noise is white noise, but the present invention can also be applied to the case of general receiver noise. The present invention can also be applied to the OFDM system. Unlike the SCBT method, in the case of OFDM, a signal subjected to inverse Fourier transform by a transmitter is transmitted, so that at the receiver, the signal subjected to Fourier transform becomes a transmission signal s (n). Means for inverse Fourier transform at the receiver are not required.

It is a figure explaining the principle of the fractional interval equalizer of this invention in the case of 2 times oversampling. It is a figure explaining the principle of the conventional fractional interval equalizer in the case of 2 times oversampling. It is a block diagram which shows the outline | summary of a SCBT system. It is a frame block diagram of a transmission signal. It is a flowchart which shows the flow of a reception process. It is a block diagram for demonstrating the process of the fractional interval equalizer of this invention.

Explanation of symbols

1 fractional interval equalizer 11 weighting processing unit 12 addition unit

Claims (2)

Means for oversampling the received signal at a density K times the symbol rate (K ≧ 2);
Means for Fourier transforming the oversampled received signal;
Means for estimating the transfer function λ _i (i = 1, 2,..., KM) of the transmission line based on a known signal included in the received signal;
Means for estimating received signal power variance and receiver noise power variance;
Frequency domain equalization means for obtaining an estimated value of the transmission signal,
The frequency domain equalization means performs equalization using a least square error reference weight including a transfer function λ _{i of the} transmission path, variance of received signal power, and variance of noise power of the receiver. A featured fractional interval equalizer. The fractional interval equalizer of claim 1;
A receiver comprising: inverse Fourier transform means for restoring the frequency domain equalized signal to a time domain signal.

Images