METHOD FOR CALCULATING CFO AND I/Q IMBALANCE COMPENSATION COEFFICIENTS, COMPENSATION METHOD USING THE SAME, AND METHOD FOR TRANSMITTING PILOT SIGNAL

Abstract

The OFDM scheme based communication system is currently being put into practical use because of its effective use of frequencies and its enhanced resistance to multipath. However, since the OFDM scheme treats multiplexed signals with overlapped spectra, the orthogonality between carriers are corrupted and the error rate characteristic is degraded in the presence of CFO. Furthermore, since locally oscillated signals different by a phase of π/2 are difficult to obtain in demodulating the I/Q signal, an imbalance is caused between the I/Q signals, resulting in degradation in the error rate characteristic. The invention suggests a novel pilot signal, and a method for analytically determining compensation values for CFO and I/Q imbalance and compensating for those distortions using the resulting values. Furthermore, the invention is applicable not only to the OFDM scheme but also to any protocol that involves pilot signals.

Description

TECHNICAL FIELD

The present invention relates to a method for compensating carrier frequency offsets (CFO) and I/Q imbalances in direct conversion receivers.

BACKGROUND ART

Recently, attention has been focused on direct conversion receivers (DCR) from the viewpoint of supplying low-cost receiver terminals to users. The DCR refers to a receiver which directly converts signals to baseband signals without the intervention of intermediate frequencies. The receiver can be reduced in size, costs, and power consumption as compared with those receivers that employ the conventional superheterodyne scheme.

However, the direct conversion of the RF-band signal to the baseband signal causes new problems such as direct current offsets and I/Q imbalances to occur. The direct current offset is caused by the self-mixing of a signal from the local oscillator (LO) and a signal leaked from the LO to the RF section. On the other hand, the I/Q imbalance arises from the distortion of the I-phase component and the Q-phase component from their respective ideal status.

The DCR requires RF band carrier signals with a phase difference of π/2 in order to decompose a received signal into the I-phase component and the Q-phase component. However, it is difficult to provide an LO having a precisely π/2 phase shift for high-frequency signals, and accordingly, resulting in a non-frequency-selective I/Q imbalance or the I-phase component and the Q-phase component being distorted from their ideal status.

Furthermore, in a wide-band communication system, differences in property of analog components, such as filters disposed in the I-branch and the Q-branch, also causes the frequency selective I/Q imbalance. The I/Q imbalance may cause image interferences and significant degradation in error rate characteristics (Non-Patent Document 1).

On the other hand, the orthogonal frequency division multiplexing (OFDM) scheme is a communication scheme that can make effective use of frequencies and enhance the resistance to multipath interferences. The OFDM scheme is employed as various wireless communication schemes such as the DAB, DVB, and IEEE 802.11a. However, the OFDM signal is a multiplexed signal with overlapped spectra, so that the presence of a carrier frequency offset (CFO) causes the so-called inter-carrier interference (ICI) or the corruption of orthogonality between carriers, resulting in significant degradation of error rate characteristics. Note that as used herein, the “carrier frequency offset (CFO)” refers to such a case where the frequency of the LO of the transceiver is not consistent with that of the LO of the receiver.

To put a low-cost direct conversion OFDM receiver into actual use, it is inevitable to remove the direct current offset caused in the RF band, and compensate for the CFO and I/Q imbalance. In general, the direct current offset can be removed by AC coupling at the preceding stage. Accordingly, what comes very critical should be the CFO compensation and the I/Q imbalance compensation.

The CFO estimation and compensation for the OFDM communication scheme have been intensively studied. These studies have introduced the compensation for the non-frequency-selective I/Q imbalance by considering only the difference in amplitude and phase in the LO, the compensation for the frequency selective I/Q imbalance by considering the difference in characteristics between the analog components such as a filter disposed in the I-branch and the Q-branch, the compensation for the CFO under the non-frequency-selective I/Q imbalance, and the like.

However, an attempt to compensate for the CFO by considering the non-frequency-selective imbalance and the frequency selective imbalance can be seen only in Non-Patent Document 2.

A brief description will now be made below to the method disclosed in Non-Patent Document 2. Note that as used herein, uppercase (lowercase) boldface letters are used to denote a matrix (column vector) in the equations. Furthermore, some sentences may contain the word “vector” in front of a letter such as in a vector X. Furthermore, the superscripts in some equations, such as the “modified H”, the “italicized T”, the asterisk, and the elongated cross (dagger), will be used to denote the Hermitian, transpose, conjugate, and pseudo-inverse matrices, respectively. Furthermore, the subscripts such as “italicized letters I and Q” will be used as the in-phase (I-branch) and the orthogonal (Q-branch) components, respectively. Furthermore, those letters with a symbol placed over them will also be denoted, for example, as hat A (referring to a letter A with a symbol “caret” above it).

Shown at 13 is a DCR mathematical model which takes into account the I/Q imbalance. A received signal bleb r(t) that has been received through the antenna and the amplifier is divided into two channel signals, i.e., the I-branch and Q-branch signals. Note that as used herein, the term “bleb r” represents “r” which has an upwardly opened arc mark placed above it. These signals will undertake a multiplication by the multiplier in the LO and go through a low-pass filter. After that, it is thought that the signals are converted using a switch into digital signals. It is assumed here that the digital signals of the I-branch and the Q-branch are r_I(k) and r_Q(k).

The non-frequency-selective I/Q imbalance caused by the LO is characterized by an amplitude difference α and a phase difference φ. The frequency selective I/Q imbalance is modeled using two real-coefficient low-pass filters which have frequency characteristics G_I(f) and G_Q(f). However, the G_I(f) and G_Q(f) are zero for the absolute value f>B/2, where B is the bandwidth. On the other hand, the CFO can be expressed by the equation below in terms of the received signal bleb r(t) in the RF band, which has been modulated at an intermediate frequency fc with a frequency offset Δf.

[Equation 1]

{hacek over (r)}(t)=2·Re{{tilde over (r)}(t)·e^{j2π(fc+Δf)t}}  (1)

The bleb r(t) is on the left side of Equation (1). Here, the tilde r(t) on the right side of Equation (1) is the received signal that has been down-converted to a baseband, and can be expressed by Equation (2). Note that the “tilde r” denotes an “r” with a mark “˜” placed above it.

[Equation 2]

{tilde over (r)}(t)={tilde over (r)}_I(t)+j·{tilde over (r)}_Q(t)=s(t)h(t)  (2)

Note that s(t) and h(t) in Equation (2) are representations of the transmitted signal and the channel response in terms of the baseband signal. The encircled “x” placed between s(t) and h(t) is a symbol denoting the operation of convolution.

Furthermore, the tilde r_I(t) and the tilde r_Q(t) are baseband signals in the I-branch and the Q-branch, respectively. Furthermore, “j” is an imaginary unit.

Here, by following the derivations in Non-Patent Documents 1 and 2, the down-converted baseband signal r(t) is given the expression below.

$\begin{array}{cc}\left[\mathrm{Equation}3\right]& \\ \begin{array}{c}r\left(t\right)=r_I\left(t\right)+j·r_Q\left(t\right)\\ =\left\{{}^{j2\pi \Delta \mathrm{ft}}·\tilde{r}\left(t\right)\right\}\otimes c_1\left(t\right)+\left\{{}^{-\mathrm{j2}\pi \Delta \mathrm{ft}}·{\tilde{r}}^{*}\left(t\right)\right\}\otimes c_2\left(t\right)\end{array}& \left(3\right)\end{array}$

where the c₁(t) and c₂(t) can be expressed by Equation (4) and Equation (5) as below.

$\begin{array}{cc}\left[\mathrm{Equation}4\right]& \\ c_1\left(t\right)=\frac{1}{2}{ℱ}^{-1}\left\{G_I\left(f\right)+\alpha {}^{-\mathrm{j\phi }}G_Q\left(f\right)\right\}& \left(4\right)\\ \left[\mathrm{Equation}5\right]& \\ c_2\left(t\right)=\frac{1}{2}{ℱ}^{-1}\left\{G_I\left(f\right)-\alpha {}^{-\mathrm{j\phi }}G_Q\left(f\right)\right\}& \left(5\right)\end{array}$

An AD converter (ADC) with a period T_S that satisfies the Nyquist sampling theorem is used to make the above equation discrete. At this time, assuming that c₁(t), c₂(t), and h(t) have an extent L₁T_S, L₂T_S, and L_hT_S, respectively, a discrete-time signal is obtained as in Equation (6) below.

[Equation 6]

r(k)=(k)+{right arrow over (r)}(k)  (6)

In the equation above, the symbols with the right and left arrows above the “r” on the left side are referred to as the “right arrow r(k)” and the “left arrow r(k),” respectively. The “left arrow r(k)” and the “right arrow r(k)” are expressed as in Equation (7) and Equation (8) below, respectively.

[Equation 7]

(k)={e^j2πΔfkTs[s(k)h]}c₁=e^j2πΔfkTss(k)  (7)

In the equation above, the left side is the “left arrow r(k)”, and the rightmost term with a left arrow above the boldface h on the right side is referred to as the “vector left arrow h”. Note that the c₁ has been rewritten as the vector c₁.

[Equation 8]

{right arrow over (r)}(k)={e^{−j2πΔfkTs}[s*(k)h*]}c₂=e^{−j2πΔfkTs}s*(k){right arrow over (h)}  (8)

In the equation above, the left side is the “right arrow r(k)”, and the rightmost term with a right arrow placed above the boldface h on the right side is referred to as the “vector right arrow h”. Note that the c₂ has been rewritten as the vector c₂.

In the equation above, the vector h is the transpose matrix of (h₀, . . . , h_Lh-1), the vector c₁ is the transpose matrix of (c_1,0, . . . , c_1,L1-1), and the vector c₂ is the transpose matrix of (c_2,0, . . . , c_2,L2-1). Note that the vector h, the vector c₁, and the vector c₂ can be explicitly represented as in the equations below.

$\left[\mathrm{Equation}9\right]$ $\begin{array}{cc}h=\left[h_0,\dots ,h_{L_h-1}\right]\left[\mathrm{Equation}10\right]& \left(9\right)\\ c_1=\left[c_{1,0},\dots ,c_{1,L_h-1}\right]\left[\mathrm{Equation}11\right]& \left(10\right)\\ c_2=\left[c_{2,0},\dots ,c_{2,L_h-1}\right]& \left(11\right)\end{array}$

Here, the left arrow r(k) is a desired signal, while the right arrow r(k) is an image interference signal resulting from the I/Q imbalance. Note that the “vector left arrow h” and the “vector right arrow h” represent a combined channel for the left arrow r(k) and the right arrow r(k), respectively. Furthermore, the “vector left arrow r(k)” and the “vector right arrow r(k)” contain the vector c₁ and the vector c₂ which have α and φ relating to the non-frequency-selective I/Q imbalance.

Furthermore, consider here an OFDM system that includes N subcarriers. Note that in the OFDM scheme, since the bandwidth B is divided into N subchannels at intervals of f₀=B/N, the Nyquist sampling period is T_S=1/(Nf₀). Furthermore, the carrier frequency offset CFO represents the normalized CFO expressed by ε=Δf/f₀. Thus, hereinafter, “ΔfkT_S” will be represented as “εk/N” with Δf and T_S replaced accordingly.

With the preparations made as above, a brief description will be made first to a conventionally well-known method. This is to help the understanding of the present invention.

<MPP-Based Compensation Method>

FIG. 14 shows a pilot signal (hereinafter referred to as the “MPP”) which was used in Non-Patent Document 2 for the CFO and I/Q imbalance compensation. The MPP includes the same symbols with its even symbols rotated by a phase of π/4. Note that in the subsequent discussions, the compensation will be carried out in three stages: the estimation of an CFO, the compensation for the I/Q imbalance, and the compensation for the CFO.

Under the perfectly timed synchronization, hat M received pilot samples with the guard interval (GI) of a length N_GI removed are arranged as in the equation below. Note that hat M is the same as “M” above which an angle bracket or caret symbol is placed.

$\left[\mathrm{Equation}12\right]$ $\begin{array}{cc}\hat{R}=\left[\begin{array}{cccc}\hat{r}\left(1,1\right)& \hat{r}\left(1,2\right)& \dots & \hat{r}\left(1,K\right)\\ \hat{r}\left(2,1\right)& \hat{r}\left(2,2\right)& \dots & \hat{r}\left(2,K\right)\\ ⋮& ⋮& \ddots & ⋮\\ \hat{r}\left(\hat{M},1\right)& \hat{r}\left(\hat{M},2\right)& \dots & \hat{r}\left(\hat{M},K\right)\end{array}\right]& \left(12\right)\end{array}$

In the equation above, hat r(m, k)=r((m−1) hat K+k) shows the k-th sample of the m-th received pilot symbol. Here, hat r(k) or the k-th column vector of a matrix hat R is expressed by the following equation.

[Equation 13]

{circumflex over (r)}(k)=E(ε)[a(k)b(k)]^T  (13)

In the equation above,

$\begin{array}{cc}\left[\mathrm{Equation}14\right]& \\ E\left(\varepsilon \right)=\left[\left(\varepsilon \right)e^{*}\left(\varepsilon \right)\right]& \left(14\right)\\ \left[\mathrm{Equation}15\right]& \\ a\left(k\right)={}^{-j\frac{2\mathrm{\pi \varepsilon }k}{N}}p\left(k\right)\otimes \overleftarrow{h}& \left(15\right)\\ \left[\mathrm{Equation}16\right]& \\ b\left(k\right)={}^{-j\frac{2\mathrm{\pi \varepsilon }k}{N}}p^{*}\left(k\right)\otimes \overrightarrow{h}& \left(16\right)\\ \left[\mathrm{Equation}17\right]& \\ \left(\varepsilon \right)={\left[{}^{j\frac{2\mathrm{\pi \varepsilon }\hat{K}}{N}},{}^{j\left(2·\frac{2\mathrm{\pi \varepsilon }\hat{K}}{N}+\frac{\pi }{4}\right)},\dots ,{}^{j\left(\hat{M}·\frac{2\mathrm{\pi \varepsilon }\hat{K}}{N}+\frac{\pi }{4}\right)}\right]}^T& \left(17\right)\end{array}$

Equation (13) shows a problem of estimating a plurality of line spectra, so that the estimation of CFO by NLS can be expressed by the following equation. Note that the NLS technique is found in Non-Patent Document 3.

$\left[\mathrm{Equation}18\right]$ $\begin{array}{cc}\hat{\varepsilon }=\underset{\tilde{\varepsilon }}{\arg \max }J\left(\tilde{\varepsilon }\right)& \left(18\right)\end{array}$

In the equation above,

[Equation 19]

J({tilde over (ε)})=tr{E({tilde over (ε)})(E^H({tilde over (ε)})E({tilde over (ε)}))⁻¹E^H({tilde over (ε)}){circumflex over (R)}{circumflex over (R)}^H}  (19)

On the other hand, the frequency selective imbalance can be compensated for by placing an FIR filter x of a length L in the Q-branch. Here, for ease of understanding, the compensation filter is placed in the Q-branch instead of the I-branch shown in Non-Patent Document 2. Note that the filter x is a vector, and hereinafter also denoted as the vector x.

Taking α as part of the G_Q(f), the frequency response X(f) of the compensation filter is set to αG_Q(f)·X(f)=G_I(f). From the fact that g(t)=Fu⁻¹{G_I(f)}, the vector g is its discrete-time representation, and g(t) is a real coefficient, the signal compensated for by the filter is expressed by the following equation. Note that Fu⁻¹ represents the Fourier inverse transform.

$\left[\mathrm{Equation}20\right]$ $\begin{array}{cc}\stackrel{.}{r}\left(k\right)=\frac{1}{2}\left[\left(1+{}^{-\mathrm{j\phi }}\right)\ddot{r}\left(k\right)+\left(1-{}^{\mathrm{j\phi }}\right){\ddot{r}}^{*}\left(k\right)\right]& \left(20\right)\end{array}$

Here, the left side of Equation (14) is referred to as the dot r(k), while the r having two dots above it on the right side is called the double-dot r(k). The double-dot r(k) is expressed as in Equation (21).

$\left[\mathrm{Equation}21\right]$ $\begin{array}{cc}\ddot{r}\left(k\right)=\left\{{}^{j2\mathrm{\pi \varepsilon }k/N}·\tilde{r}\left(k\right)\right\}\otimes g& \left(21\right)\end{array}$

In the equation above, the double-dot r(k) is a signal affected by CFO. Here, the real part and the imaginary part of the double-dot r(k) are expressed by the following equations.

[Equation 22]

{umlaut over (r)} _I(k)={dot over (r)}_I(k),  (22)

[Equation 23]

{umlaut over (r)} _Q(k)=tan φ·{dot over (r)}_I(k)+secφ·{dot over (r)}_Q(k)  (23)

The aforementioned two equations show an asymmetrical compensation of the non-frequency-selective imbalance signal which is compensated by two gain coefficient elements β and χ corresponding to tan φ and secφ. Note that this is described in Non-Patent Document 4.

FIG. 15 shows the entire structure that takes the CFO to compensation into account, where it is assumed that hat L=(L−1)/2. The digitized signals of r_I(k) and r_Q(k) are compensated as follows. First, the r_Q(k) is acted upon by the compensation filter x to obtain the dot r_Q(k). On the other hand, the r_I(k) is delayed for the duration of (k−hat L) to be a signal dot r_I(k). The dot r_I(k) is multiplied by β to yield a signal, and the dot r_Q(k) is acted upon by the filter χ to produce another signal, so that these signals are summed to give a double-dot r_Q(k).

The complex signal double-dot r(k) with the dot r_I(k) employed as the double-dot r_I(k) and the double-dot r_Q(k) employed as the imaginary part is a signal with the I/Q imbalance having been compensated for. The double-dot r(k) is multiplied by the amount of the CFO compensation, thereby providing a signal with the I/Q imbalance and CFO having been compensated for. On the other hand, obviously, χ is incorporated into the vector x, so that the compensation problem is now turned into the optimization problem of the vectors x and β.

In the absence of the I/Q imbalance, only the CFO is affected in the received pilot. Thus, the optimum compensation filter vector x and the gain coefficient β become optimum when adjacent pilot symbols after the I/Q imbalance has been compensated for coincide with the phase difference caused by the CFO. When a CFO estimation value hat ε is given, the following quantities are defined as in the equations below.

$\left[\mathrm{Equation}24\right]\begin{array}{cc}{\hat{R}}_Q\left(m\right)={\left[\begin{array}{cccc}{\hat{r}}_Q\left(m,1\right)& & & \\ {\hat{r}}_Q\left(m,2\right)& {\hat{r}}_Q\left(m,1\right)& & \\ ⋮& {\hat{r}}_Q\left(m,2\right)& \ddots & \\ {\hat{r}}_Q\left(m,K\right)& ⋮& \ddots & {\hat{r}}_Q\left(m,1\right)\\ & {\hat{r}}_Q\left(m,K\right)& ⋮& {\hat{r}}_Q\left(m,2\right)\\ & & \ddots & ⋮\\ & & & {\hat{r}}_Q\left(m,K\right)\end{array}\right]}_{\left(K+L-1\right)\times L}\left[\mathrm{Equation}25\right]& \left(24\right)\\ {\hat{r}}_I\left(m\right)={\left[\underset{\hat{L}}{\underset{}{0,\dots ,0}},{\hat{r}}_I\left(m,1\right),\dots ,{\hat{r}}_I\left(m,K\right),\underset{\hat{L}}{\underset{}{0,\dots ,0}}\right]}^T\left[\mathrm{Equation}26\right]& \left(25\right)\\ {\omega }_m=\{\begin{array}{cc}\frac{2\pi \hat{\varepsilon }\hat{K}}{N}+\frac{\pi }{4}\text{:}& m=\mathrm{odd}\\ \frac{2\pi \hat{\varepsilon }\hat{K}}{N}-\frac{\pi }{4}\text{:}& m=\mathrm{even}\end{array}& \left(26\right)\end{array}$

In this case, the optimum values of the vectors x and β are expressed by the following equation (Non-Patent Document 2).

$\left[\mathrm{Equation}27\right]$ $\begin{array}{cc}\left[\begin{array}{c}\hat{x}\\ \hat{\beta }\end{array}\right]={\left[\begin{array}{c}A\left(1\right)\\ ⋮\\ A\left(\hat{M}-1\right)\end{array}\right]}^{†}\left[\begin{array}{c}B\left(1\right)\\ ⋮\\ B\left(\hat{M}-1\right)\end{array}\right]=A^{†}B& \left(27\right)\end{array}$

In the equation above,

$\left[\mathrm{Equation}28\right]$ $\begin{array}{cc}A\left(m\right)=\left[\begin{array}{cc}{\hat{R}}_Q\left(m\right)\sin {\omega }_m& {\hat{r}}_I\left(m\right)\sin {\omega }_m\\ {\hat{R}}_Q\left(m+1\right)-{\hat{R}}_Q\left(m\right)\cos {\omega }_m& {\hat{r}}_I\left(m+1\right)-{\hat{r}}_I\left(m\right)\cos {\omega }_m\end{array}\right]\left[\mathrm{Equation}29\right]& \left(28\right)\\ B\left(m\right)=\left[\begin{array}{c}{\hat{r}}_I\left(m\right)\cos {\omega }_m-{\hat{r}}_I\left(m+1\right)\\ {\hat{r}}_I\left(m\right)\sin {\omega }_m\end{array}\right]& \left(29\right)\end{array}$

Obviously, the vectors x and β depend on the CFO estimation value hat ε. Finally, the CFO is corrected by a simple phase rotation.

• [Non-Patent Document 1] M. Valkama, M. Renfors, and V. Koivunen, “Advanced methods for I/Q imbalance compensation in communication receivers,” IEEE Trans. Signal Processing, vol. 49, no. 10, pp. 2335-2344, October 2001
• [Non-Patent Document 2] G. Xing, M. Shen and H. Liu, “Frequency Offset and I/Q Imbalance Compensation for Direct-Conversion Receivers,” IEEE Trans. Wireless Commun, vol. 4, no. 2, pp. 673-680, March 2005.
• [Non-Patent Document 3] P. Sotica and R. Moses, “Introduction to Spectral Analysis” Englewood Cliffs, N.J.: Prentice-Hall, 1997.
• [Non-Patent Document 4] J. K. Cavers, and M. W. Liao, “Adaptive compensation for imbalance and offset losses in direct conversion transceivers,” IEEE Trans. Veh. Technol., vol. 42, no. 4, pp. 581-585, November 1993.

DISCLOSURE OF THE INVENTION

Problems to be Solved by the Invention

Non-Patent Document 2 is based on the modified period pilot (MPP); however, this method cannot provide a compensation coefficient for I/Q imbalance without the CFO estimation. Accordingly, the CFO estimation problem is a very critical challenge.

The CFO estimation requires accurate synchronous timing as well as the solution of a nonlinear least-squares problem (NLS). The accurate synchronous timing cannot be easily implemented from technical viewpoints. Furthermore, the solution of the nonlinear least-squares problem is comparatively easy but requires a one-dimensional search, which practically causes a major impediment.

Furthermore, the MPP-based compensation method is accompanied by the following difficulties when an attempt is made to implement it.

1. The estimation of CFO by the evaluation function (Equation 19) requires the solution of a nonlinear least-squares problem, i.e., a one-dimensional search, thus practically causing a serious impediment.

2. The optimum values of the vectors x and β in (Equation 27) can be calculated only after the CFO has been estimated. Accordingly, parallel processing cannot be performed which is an effective method of calculation.

3. To estimate the CFO with accuracy, it is necessary to know whether the received pilot symbol is an even or odd one. To this end, the accurate synchronous timing is required. However, it is generally difficult to provide accurate synchronous timing due to the loss of pilot symbols in the preceding stage caused by the leakage of a direct current offset that results from automatic gain control (AGC) and AC coupling.

4. Since only even pilot symbols are rotated by a phase of π/4, it is necessary to insert a guard interval into each symbol to prevent inter-block interferences. This causes an increase in pilot interval and a decrease in the range of CFO estimation.

Means for Solving the Problems

The present invention was developed in view of the aforementioned problems. It is therefore an object of the invention to provide a method for suggesting an extended period pilot (GPP) and for simultaneously estimating imbalance coefficients relating to the CFO and the IQ imbalance. To this end, the present invention provides a method for transmitting pilot signals and a method for determining the compensation coefficients of the CFO and the I/Q imbalance.

The present invention was developed to solve the aforementioned problems. The invention provides a method for sequentially acquiring a predetermined number of pieces of digitized output data of the I-branch and the Q-branch in a complex demodulator and analytically determining a CFO estimation value and a compensation coefficient of an I/Q imbalance by the operation of a matrix made up of the data.

More specifically, the invention provides a CFO estimation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then estimating a CFO of the signal. The method includes the steps of: digitizing the I-branch side signal of the received pilot signal into I data; digitizing the Q-branch side signal of the received pilot signal into Q data; forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (34); forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (37); forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (35); forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (38); determining a matrix u being equal, when multiplied by a matrix of Equation (46) obtained from the Equation (34) and the Equation (37), to a matrix of Equation (45) obtained from the Equation (34), the Equation (37), the Equation (35), and the Equation (38); and determining a CFO estimation value ε based on Equation (48) from first and second elements of the matrix u.

Furthermore, the present invention provides an I/Q imbalance compensation coefficient calculation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then calculating a compensation coefficient to compensate for the I/Q imbalance of the signal. The method includes the steps of: digitizing the I-branch side signal of the received pilot signal into I data; digitizing the Q-branch side signal of the received pilot signal into Q data; forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (34); forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (37); forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (35); forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (38); determining a matrix u being equal, when multiplied by a matrix of Equation (46) obtained from the Equation (34) and the Equation (37), to a matrix of Equation (45) obtained from the Equation (34), the Equation (37), the Equation (35), and the Equation (38); determining an I/Q imbalance compensation coefficient β from first and second elements of the matrix u and a CFO value ε based on Equation (49); and determining an I/Q imbalance compensation coefficient vector x from elements other than the first and second elements of the matrix u and the CFO value hat ε based on Equation (50).

Furthermore, the present invention provides an I/Q imbalance compensation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and thereafter compensating the signal. The method includes the steps of: digitizing the I-branch side signal of the received signal into I data; digitizing the Q-branch side signal of the received signal into Q data; multiplying the Q data by the vector x determined according to the method of claim 2; multiplying the I data by β determined according to the method of claim 2; adding data obtained by multiplying the I data by β to the Q data multiplied by the vector x to yield Qc data; and determining a complex number with the I data employed as a real part and the Qc data employed as an imaginary part.

Furthermore, the present invention provides a signal compensation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and thereafter compensating the signal. The method includes the step of compensating the complex number determined in claim 3, based on the CFO estimation value determined by the method according to claim 1.

Furthermore, the present invention provides a CFO estimation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then estimating a CFO of the signal. The method includes the steps of: digitizing the I-branch side signal of the received pilot signal into I data; digitizing the Q-branch side signal of the received pilot signal into Q data; forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (51); forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (53); forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (52); forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (54); determining a matrix u being equal, when multiplied by a matrix of Equation (61) obtained from the Equation (51) and the Equation (53), to a matrix of Equation (60) obtained from the Equation (51), the Equation (53), the Equation (52), and the Equation (54); and determining a CFO estimation value s from first and second elements of the matrix u based on Equation (63).

Furthermore, the present invention provides an I/Q imbalance compensation coefficient calculation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then calculating a compensation coefficient to compensate for the I/Q imbalance of the signal. The method includes the steps of: digitizing the I-branch side signal of the received pilot signal into I data; digitizing the Q-branch side signal of the received pilot signal into Q data; forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (51); forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (53); forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (52); forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (54); determining a matrix u being equal, when multiplied by a matrix of Equation (61) obtained from the Equation (51) and the Equation (53), to a matrix of Equation (60) obtained from the Equation (51), the Equation (53), the Equation (52), and the Equation (54); determining an I/Q imbalance compensation coefficient β from first and second elements of the matrix u and a CFO value ε based on Equation (64); and determining an I/Q imbalance compensation coefficient vector x from elements other than the first and second elements of the matrix u and the CFO value hat ε based on Equation (65).

Furthermore, the present invention provides an I/Q imbalance compensation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and thereafter compensating the signal. The method includes the steps of: digitizing the I-branch side signal of the received signal into I data; digitizing the Q-branch side signal of the received signal into Q data; multiplying the I data by the vector x determined by the method according to claim 6; multiplying the Q data by β determined according to the method of claim 2; adding data obtained by multiplying the Q data by β to the I data multiplied by the vector x to yield Ic data; and determining a complex number with the Q data employed as a real part and the Qc data employed as an imaginary part.

Furthermore, the present invention provides a signal compensation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and thereafter compensating the signal. The method includes the step of compensating the complex number determined in claim 7 based on the CFO estimation value determined by the method according to the claim 5.

Furthermore, the present invention provides a CFO sign determination method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and determining a sign of a CFO of the signal. The method includes the steps of: digitizing the I-branch side signal of the received pilot signal into I data; digitizing the Q-branch side signal of the received pilot signal into Q data; creating a matrix R of Equation (72) with a first row and a second row, the first row having (P−K) pieces of complex data with (P−K) samples from an n-th sample of the I data employed as a real part and (P−K) samples from an nth sample of the Q data employed as an imaginary part, the second row having (P−K) pieces of complex data with (P−K) samples from an (n+K)-th sample of the I data employed as a real part and (P−K) samples from an (n+K)-th sample of the Q data employed as an imaginary part; creating a matrix of Equation (78) based on an absolute value of a CFO estimation value ε whose sign is wanted to be determined; multiplying the Equation (72) by Equation (78); and comparing a norm of a first row of the resulting matrix with a norm of a second row to determine that the sign of ε is positive when the first row norm is greater than the second row norm.

Furthermore, the present invention provides an I/Q imbalance compensation coefficient calculation method for demodulating a signal at a demodulator having an I-branch and a Q-branch, the signal containing a pilot signal with a short TS and a long TS and with no phase difference between adjacent symbols, and for calculating a compensation coefficient to compensate for an I/Q imbalance of the signal. The method includes the steps of: selecting a predetermined subcarrier from the respective short TS and long TS to create a matrix of Equation (82); creating a diagonal matrix of Equation (83) from a subcarrier element of the short TS; creating a diagonal matrix of Equation (84) from a subcarrier element of the long TS; creating a diagonal matrix of Equation (92) from a CFO value whose absolute value is less than a predetermined value; creating Equation (90) from the Equation (82), the Equation (83), and the Equation (89); creating Equation (91) from the Equation (82), the Equation (84), and the Equation (89); forming (P−K) samples from an n-th sample of the I data of the short TS into a matrix of Equation (86); forming (P−K+(L−1)/2) samples into a matrix of Equation (85) from an (n−(L−1)/2)-th sample of the Q data of the short TS; forming (P−K) samples from an n-th sample of the I data of the long TS into a matrix of Equation (88); forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data of the long TS into a matrix of Equation (87); creating Equation (94) from the Equation (85), the Equation (86), the Equation (87), the Equation (88), the Equation (90), and the Equation (91); obtaining Equation (95) from the Equation (86), the Equation (88), the Equation (90), and the Equation (91); and determining a vector being equal, when multiplied by Equation (94), to Equation (95).

Furthermore, the present invention provides a transmission method for time division multiplexing and then transmitting a main signal and a pilot signal. The method includes the steps of: time division multiplexing the main signal and periodic pilot signal; and imparting a predetermined phase difference to the pilot signal during the time division multiplexing.

EFFECTS OF THE INVENTION

A feature of the present invention lies in solving a linear least squares (LLS) algorithm, so that CFOs and the imbalance coefficients can be all analytically obtained, thereby significantly reducing calculation load. This means that compensation load is reduced for the communication device and the reception status can be frequently corrected. It therefore can be said that this is a preferable compensation method for mobile communications whose reception conditions vary. Furthermore, the conventional periodic pilot (PP) is contained in the GPP, allowing countermeasures to be taken against the ambiguity of CFO sign determinations and the problem of compensating for zero CFO. This allows the present invention to be applied, for example, even to such a case in which no phase difference is set at an IEEE 802.11a WLAN pilot signal.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a view illustrating the configuration of a pilot signal of the present invention;

FIG. 2 is a view illustrating an exemplary layout of a device for performing a compensation method of the present invention;

FIG. 3 is a view illustrating an example of producing a signal in a compensation method of the present invention;

FIG. 4 is a view illustrating another exemplary configuration of a device for performing a compensation method of the present invention;

FIG. 5 is a view illustrating another example of producing another signal in a compensation method of the present invention;

FIG. 6 is a view illustrating still another example of producing another signal in a compensation method of the present invention;

FIG. 7 is a view illustrating the structure of data in the IEEE 802.11a WLAN;

FIG. 8 is a view illustrating the flow of a compensation method according to a second embodiment of the present invention;

FIG. 9 is a view illustrating the results of simulation of the relationship between CFO and SNR;

FIG. 10 is a view illustrating the results of simulation of the relationship between BER and SNR;

FIG. 11 is a view illustrating the results of simulation of the relationship between BER and timing;

FIG. 12 is a view illustrating the results of simulation of the relationship between BLER and SNR;

FIG. 13 is a view illustrating a model of a receiver employing the direct conversion scheme at the time of occurrence of the I/Q imbalance;

FIG. 14 is a view illustrating the configuration of a conventional pilot signal; and

FIG. 15 is a view illustrating the configuration of a circuit for compensating for the I/Q imbalance and CFO.

BEST MODE FOR CARRYING OUT THE INVENTION

GPP-Based Compensation Method

First, the rationale of the invention will be described using a number of mathematical equations. Then, the description will be followed by the explanation of specific embodiments.

Those cases 1) and 2) result from the ability of calculating the optimum value of the imbalance compensation coefficient only after the estimation of CFO, while the cases 3) and 4) arise from a special structure of the MPP.

To solve all the aforementioned problems, the present invention is characterized in that the GPP shown in FIG. 1 is employed as a pilot symbol. The GPP is made up of a group of the same symbols that contains no guard intervals, with a common phase rotation θ between two adjacent symbols. Obviously, if no CFO and I/Q imbalance exist after the convolution with the channel, any two received samples spaced apart from each other by KT_s within the pilot interval have the relation given below.

[Equation 30]

r(n+K)=e^jθr(n)  (30)

On the other hand, if the CFO exists but no I/Q imbalance does, then the equation below holds.

[Equation 31]

r(n+K)=e^j(ψ+θ)r(n)  (31)

In the equation above, ψ=2πεK/N is the function of an unknown CFO or ε.

In the (P+2) hat L received samples within the pilot interval, the I/Q imbalance is compensated for according to FIG. 15, and P samples are accommodated in the two hat P×1 vectors shown below.

[Equation 32]

{umlaut over (r)} ₁ =[{umlaut over (r)}(n+{circumflex over (L)}), . . . , {umlaut over (r)}(n+{circumflex over (L)}+P−K−1)]^T  (32)

[Equation 33]

{umlaut over (r)} ₂ =[{umlaut over (r)}(n+K+{circumflex over (L)}), . . . , {umlaut over (r)}(n+{circumflex over (L)}+P−1)]^T  (33)

In the equations above, hat P=P−K. Here, a vector r_1,I and a vector R_1,Q can be defined as in the equations below.

[Equation 34]

r _1,I =[r _I(n), . . . , r_I(n+P−K−1)]^T  (34)

and

$\left[\mathrm{Equation}35\right]$ $\begin{array}{cc}{R}_{1,Q}=\left[\begin{array}{ccccc}{r}_Q\left(n+\hat{L}\right)& \dots & r_Q\left(n\right)& \dots & r_Q\left(n-\hat{L}\right)\\ r_Q\left(n+1+\hat{L}\right)& \dots & r_Q\left(n+1\right)& \dots & r_Q\left(n+1-\hat{L}\right)\\ ⋮& ⋮& ⋮& ⋮& ⋮\\ r_Q\left(n+P-K-1+\hat{L}\right)& \dots & r_Q\left(n+P-K-1\right)& \dots & r_Q\left(n+P-K-1-\hat{L}\right)\end{array}\right]& \left(35\right)\end{array}$

Thus, from FIG. 1, the relational equation is obtained as shown below.

[Equation 36]

{umlaut over (r)} ₁ =r _1,I +j·(R_1,Qx+βr_1,I)  (36)

On the other hand, n can be replaced by (n+K), thereby obtaining the following equations.

$\left[\mathrm{Equation}37\right]$ $\begin{array}{cc}r_{2,I}={\left[r_I\left(n+K\right),\dots r_I\left(n+P-1\right)\right]}^T\left[\mathrm{Equation}38\right]& \left(37\right)\\ R_{2,Q}=\left[\begin{array}{ccccc}{r}_Q\left(n+K+\hat{L}\right)& \dots & r_Q\left(n+K\right)& \dots & r_Q\left(n+K-\hat{L}\right)\\ r_Q\left(n+K+1+\hat{L}\right)& \dots & r_Q\left(n+K+1\right)& \dots & r_Q\left(n+K+1-\hat{L}\right)\\ ⋮& & ⋮& & ⋮\\ r_Q\left(n+P-1+\hat{L}\right)& \dots & r_Q\left(n+P-\hat{L}\right)& \dots & r_Q\left(n+P-1-\hat{L}\right)\end{array}\right]& \left(38\right)\end{array}$

These equations give the relation below.

[Equation 39]

{umlaut over (r)} ₂ =r _2,I +j·(R_2,Qx+βr_2,I)  (39)

Obviously, if the imbalance compensation has been accurately made, then the aforementioned two vectors satisfy the relational expression of Equation (31), and thus the relation of Equation (38) can be said to hold.

[Equation 40]

{umlaut over (r)} ₂ =e ^j(ψ+θ) {umlaut over (r)} ₁  (40)

Substituting Equation (36) and Equation (37) into Equation (38) gives the following equation.

$\left[\mathrm{Equation}41\right]$ $\begin{array}{cc}\left[\begin{array}{cc}{r}_{1,I}& -R_{1,Q}\end{array}\right]\left[\begin{array}{c}\cos \left(\psi +\theta \right)-\beta \sin \left(\psi +\theta \right)\\ x\sin \left(\psi +\theta \right)\end{array}\right]=r_{2,I}& \left(41\right)\end{array}$

From the above equation, it can be seen that cos(ψ+θ)−β sin(ψ+θ) and vector x sin(ψ+θ), that is, the CFO and the imbalance coefficient are simultaneously obtained from the vectors r_1,I, and the −vectors R_1,Q and r_2,I.

To find three unknown parameters ψ, β, and the vector x, it is insufficient to determine only cos(ψ+θ)−β sin(ψ+θ) and vector x sin(ψ+θ). However, fortunately, Equation (40) holds.

[Equation 42]

{umlaut over (r)} ₁ =e ^−j(φ+θ) {umlaut over (r)} ₂  (42)

Accordingly, the relation of Equation (43) that gives cos(ψ+θ)+β sin(ψ+θ) can be found.

$\left[\mathrm{Equation}43\right]$ $\begin{array}{cc}\left[\begin{array}{cc}{r}_{2,I}& R_{2,Q}\end{array}\right]\left[\begin{array}{c}\cos \left(\psi +\theta \right)+\beta \sin \left(\psi +\theta \right)\\ x\sin \left(\psi +\theta \right)\end{array}\right]=r_{1,I}& \left(43\right)\end{array}$

Thus, Equation (41) and Equation (43) can be combined into Equation (44).

$\left[\mathrm{Equation}44\right]$ $\begin{array}{cc}\Lambda \left[\begin{array}{c}\cos \left(\psi +\theta \right)-\beta \sin \left(\psi +\theta \right)\\ \cos \left(\psi +\theta \right)+\beta \sin \left(\psi +\theta \right)\\ x\sin \left(\psi +\theta \right)\end{array}\right]=r_I& \left(44\right)\end{array}$

Here, the vector Λ and the vector r_I can be expressed as in Equation (45) and Equation (46).

$\left[\mathrm{Equation}45\right]$ $\begin{array}{cc}\Lambda =\left[\begin{array}{ccc}{r}_{1,I}& 0& -R_{1,Q}\\ 0& r_{2,I}& R_{2,Q}\end{array}\right]& \left(45\right)\end{array}$

Here, the vector 0 is a zero vector that has the number of elements of hat P×1.

$\left[\mathrm{Equation}46\right]$ $\begin{array}{cc}{r}_I=\left[\begin{array}{c}{r}_{2,I}\\ r_{1,I}\end{array}\right]& \left(46\right)\end{array}$

In the above equation, if P≧(K+hat L+2), the vector Λ is a nonsingular column matrix.

In this context, the LLS algorithm can be used to find a vector u of (L+2)×1 dimensions in Equation (47). The first and second elements of the vector u contain only ε and β which are an CFO, while the third element onward include only the vector x.

[Equation 47]

u=Λ^†r_I  (97)

That is, the CFO estimation and the I/Q imbalance compensation coefficient can be analytically determined using the elements of the matrix u. Note that Equation (47) was expressed so as to find the pseudo-inverse matrix of the matrix Λ in order to determine the matrix u. However, the method for determining the matrix u from Equation (47) is not limited only to this one, and any other well-known method may also be employed. More specifically, the Gauss-Jordan solution method may also be used. Furthermore, as used herein, it will be referred to simply as “determining the vector u” or “the step of determining the vector u” from Equation (47).

Now, the CFO estimation and the I/Q imbalance compensation coefficient are explicitly shown below.

$\begin{array}{cc}\left[\mathrm{Equation}48\right]& \\ \hat{\varepsilon }=\frac{N}{2\pi K}\left[\arccos \left\{\frac{u\left(1\right)+u\left(2\right)}{2}\right\}-\theta \right]& \left(48\right)\\ \left[\mathrm{Equation}49\right]& \\ \hat{\beta }=\frac{u\left(2\right)-u\left(1\right)}{2\sin \left(2\pi \hat{\varepsilon }K/N+\theta \right)}& \left(49\right)\\ \left[\mathrm{Equation}50\right]& \\ \hat{x}={\frac{1}{\sin \left(2\pi \hat{\varepsilon }K/N+\theta \right)}\left[u\left(3\right),\dots ,u\left(L+2\right)\right]}^T& \left(50\right)\end{array}$

As described above, in the present invention, the CFO estimation value and the compensation value for compensating for the I/Q imbalance are analytically determined from the is received pilot signal by compensating for the signal of the Q-branch. However, the I-branch and the Q-branch are fundamentally the same signal only with different phases. Accordingly, the I-branch side signal can be compensated, thereby determining the CFO estimation value and the I/Q imbalance compensation value in the same manner.

FIG. 2 shows a conceptual view for the I-branch side signal being compensated. The theory of compensation can be explained as below. It is the same as mentioned above up to the fact that the relation of Equation (31) holds for any two received samples spaced apart from each other by KTS within the pilot interval. At this point, the I side signal and the Q side signal are exchanged when the P+2 hat L received samples within the pilot interval are placed in the two hat P×1 vectors. That is, the r_1,I and R_1,Q which have been found in Equation (34) onward are replaced with the r_1,Q and R_1,I.

$\begin{array}{cc}\left[\mathrm{Equation}51\right]& \\ r_{1,Q}={\left[r_Q\left(n\right),\dots ,r_Q\left(n+P-K-1\right)\right]}^T,& \left(51\right)\\ \left[\mathrm{Equation}52\right]& \\ R_{1,I}=\left[\begin{array}{ccccc}{r}_I\left(n+\hat{L}\right)& \cdots & r_I\left(n\right)& \cdots & r_I\left(n-\hat{L}\right)\\ r_I\left(n+1+\hat{L}\right)& \cdots & r_I\left(n+1\right)& \cdots & r_I\left(n+1-\hat{L}\right)\\ ⋮& ⋮& ⋮& ⋮& ⋮\\ r_I\left(n+P-K-1+\hat{L}\right)& \cdots & r_I\left(n+P-K-1\right)& \cdots & r_1\left(n+P-K-1-\hat{L}\right)\end{array}\right]& \left(52\right)\end{array}$

Furthermore, by putting n=n+K, Equation (53) and Equation (54) can be obtained as below.

$\begin{array}{cc}\left[\mathrm{Equation}53\right]& \\ r_{2,Q}={\left[r_Q\left(n+K\right),\dots r_Q\left(n+P-1\right)\right]}^T& \left(53\right)\\ \left[\mathrm{Equation}54\right]& \\ R_{2,I}=\left[\begin{array}{ccccc}{r}_I\left(n+K+\hat{L}\right)& \cdots & r_I\left(n+K\right),& \cdots & r_I\left(n+K-\hat{L}\right)\\ r_I\left(n+K+1+\hat{L}\right)& \cdots & r_I\left(n+K+1\right),& \cdots & r_I\left(n+K+1-\hat{L}\right)\\ ⋮& & ⋮& & ⋮\\ r_I\left(n+P-K-1+\hat{L}\right)& \cdots & r_I\left(n+P-1\right),& \cdots & r_I\left(n+P-1-\hat{L}\right)\end{array}\right]& \left(54\right)\end{array}$

Thus, the modified I-branch and Q-branch signals or the vector double-dot r1 and the vector double-dot r2 are expressed as in Equation (55) and Equation (56) below.

[Equation 55]

{umlaut over (r)} ₁=(R_1,Ix+βr_1,Q)+j·r_1,Q.  (55)

[Equation 56]

{umlaut over (r)} ₂=(R_2,Ix+βr_2,Q)+j·r_2,Q.  (55)

If the imbalance compensation has been correctly made, Equation (30) holds true irrespective of the I-branch and the Q-branch. Thus, the aforementioned two equations, i.e., Equation (55) and Equation (56) are substituted into Equation (30). As a result, Equation (57) is obtained as below.

$\begin{array}{cc}\left[\mathrm{Equation}57\right]& \\ \left[r_{1,Q}R_{1,I}\right]\left[\begin{array}{c}\cos \left(\psi +\theta \right)+\beta \sin \left(\psi +\theta \right)\\ x\sin \left(\psi +\theta \right)\end{array}\right]=r_{2,Q}.& \left(57\right)\end{array}$

Furthermore, since the vector double-dot r1 and the vector double-dot r2 have the relation of Equation (40), Equation (58) can also be obtained as below.

$\left[\mathrm{Equation}58\right]$ $\begin{array}{cc}\left[r_{2,Q}-R_{2,I}\right]\left[\begin{array}{c}\cos \left(\psi +\theta \right)-\beta \sin \left(\psi +\theta \right)\\ x\sin \left(\psi +\theta \right)\end{array}\right]=r_{1,Q},& \left(58\right)\end{array}$

These equations can be combined into Equation (59) in the same manner as with Equation (44).

$\left[\mathrm{Equation}59\right]$ $\begin{array}{cc}\Lambda \left[\begin{array}{c}\cos \left(\psi +\theta \right)+\beta \sin \left(\psi +\theta \right)\\ \cos \left(\psi +\theta \right)-\beta \sin \left(\psi +\theta \right)\\ x\sin \left(\psi +\theta \right)\end{array}\right]=r_Q& \left(59\right)\end{array}$

Note that here, the matrix Λ and the matrix r_Q can be expressed by Equation (60) and Equation (61) below.

$\begin{array}{cc}\left[\mathrm{Equation}60\right]& \\ \Lambda =\left[\begin{array}{ccc}{r}_{1,Q}& 0& R_{1,I}\\ 0& r_{2,Q}& -R_{2,I}\end{array}\right];& \left(60\right)\\ \left[\mathrm{Equation}61\right]& \\ r_Q=\left[\begin{array}{c}{r}_{2,Q}\\ r_{1,Q}\end{array}\right];& \left(61\right)\end{array}$

The matrix u is expressed using a pseudo-inverse matrix as with Equation (62) below, and the vector u is determined using a well-known solution.

[Equation 62]

u=Λ^†r_Q,  (62)

Accordingly, the CFO estimation value and the I/Q imbalance compensation coefficient can be explicitly expressed by Equation (63), Equation (64), and Equation (65) as below.

$\begin{array}{cc}\left[\mathrm{Equation}63\right]& \\ \hat{\varepsilon }=\frac{N}{2\pi K}\left\{\arccos \left(\frac{u\left(1\right)+u\left(2\right)}{2}\right)-\theta \right\}.& \left(63\right)\\ \left[\mathrm{Equation}64\right]& \\ \hat{\beta }=\frac{u\left(2\right)-u\left(1\right)}{2\sin \left(2\pi \hat{\varepsilon }K/N+\theta \right)},& \left(64\right)\\ \left[\mathrm{Equation}65\right]& \\ \hat{x}={\frac{1}{\sin \left(2\pi \hat{\varepsilon }K/N+\theta \right)}\left[u\left(3\right),\dots ,u\left(L+2\right)\right]}^T.& \left(65\right)\end{array}$

As can be seen from the discussions above, the information required to simultaneously compensate for the CFO and I/Q imbalance is also obtained by calculating the vector A and the vector r_I from the received pilot. In practice, the calculation of a trigonometric function can be performed by referencing a look-up table. Table 1 shows the calculation load for the GPP-based scheme and the MPP-based scheme. Note that no consideration was given to the amount of calculation for the CFO estimation by the MPP-based scheme. Since the relation holds that rank (vector E(−N/(8 hat K)))=1 for hat M=2, thus it holds that min(hat M)=3. Since it is satisfied that min(hat P)=(L+3)/2 and generally L<K, the present invention requires only a significantly reduced amount of calculation. Without loss of generality, when 0=π/2, the CFO estimation range allows 6, which is a half of that without taking the I/Q imbalance into account, to lie within the range of (−N/4K, N/4K). Furthermore, since Equation (40) holds true for n≧K, it can be seen that the present invention is robust against timing error. This is because that the present invention makes it possible to determine the compensation value of the CFO and the I/Q imbalance only if a phase difference θ is available between packets even when the earlier packet cannot be acquired.

	TABLE 1
	GPP-based	MPP-based
	Pseudoinverse of	2{circumflex over (P)} × (L + 2)	(2{circumflex over (M)}(K + L − 1)) × (L + 1)
	Addition	1	—
	Substraction	1	3{circumflex over (M)}(K + L − 1)
	Multiplication	L + 3	6{circumflex over (M)}(K + L − 1)

Now, a detailed description will be made to the practical aspects of the present invention. FIG. 2 shows the configuration of the present invention. There is a transmitter 1 which transmits signals and may be either a broadcast station or a privately-owned transmitter. In the present invention, the transmitter 1 includes a signal source 2, a pilot signal generator 3, a combiner 4, and a frequency converter 5. The transmitter 1 may also include an output amplifier 6 and an antenna 7. Here, the transmitter 1 transmits pilot signals which have phases different from each other by θ for each symbol. Furthermore, the pilot signal is time division multiplexed with the original signal emitted from the signal source. This is because the present invention requires a period of time in which only the pilot signals are being received on the reception side.

The output from the combiner 4 is transmitted via the frequency converter 5. The frequency converter 5 may include a symbolizing function, so that the format of the signal transmitted is not limited to a particular one. For example, either the OFDM scheme or the FM modulation scheme may be employed. The transmitter of the present invention imparts a predetermined phase difference to each pilot signal. This may be done by either the pilot signal generator 3 or the combiner 4. The interval at which the phase difference is imparted may be fixed or made variable. It is preferable for the receiver side to know the interval of the same phase difference. Furthermore, it is typically preferable to change the phase difference for each symbol; however, the invention is not limited thereto.

On the other hand, there is also provided a receiver 10 which includes an antenna 11, an amplifier 12, a frequency converter and a filter (17 and 18), switching elements (19 and 20), and a controller 30. The frequency converter is a complex frequency converter. The receiver 10 typically includes a local oscillator LO (15), multipliers (13 and 14), and a phase converter 16.

The output of the amplifier 12 is split into the I-branch and the Q-branch. The I-branch side signal is multiplied by a carrier signal from the local oscillator LO 15 at the multiplier 13. Furthermore, the Q-branch side signal is multiplied by a signal with the phase shifted by π/2 from that of the carrier signal of the local oscillator LO at the multiplier 14.

The signals in the I-branch and the Q-branch pass through the low-pass filters (17 and 18), respectively, to be removed of unwanted high-frequency components. Thereafter, the signals are converted into a digital signal at AD converters (19 and 20) that have a sufficient sampling frequency. The respective signals of the I-branch and the Q-branch are supplied to the controller 30.

Now, a description will be made to the processing of the controller 30 to compensate the Q-branch side signal. FIG. 1 shows a processing section associated with a processing step in the controller 30 as if the section actually exists; however, the processing is fundamentally performed by software. As a matter of course, a dedicated hardware section may also be manufactured to perform the processing. Note that the signal digitized on the I-branch side is hereinafter referred to as the I data and the signal digitized on the Q-branch side referred to as the Q data. When having received the Q data and the I data, the controller 30 allows a compensation value calculation section 28 to calculate compensation values from the respective pieces of data. The resulting compensation values are supplied to the filter section 21, a multiplier section 22, and a CFO compensation signal generation section 27, respectively.

The Q data supplied to the controller 30 is acted upon by the filter x based on the compensation value. On the other hand, the I data is multiplied by β and then added at an adder section 23 to the Q data which has been acted upon by the filter x. The resulting signal is imparted an imaginary unit “j” at an imaginary section 24 and then added to the I data at an adder section 25. The signal to which the imaginary unit was imparted is referred to as a Qc signal. The adder section 25 outputs complex numbers. The complex number is data with the I/Q imbalance having been compensated for. Next, the complex number is multiplied at a multiplier 26 by a value, as a complex number, for compensating for ε or the CFO estimation value. The complex number thus determined is a transmitted signal with both the CFO and the I/Q imbalance having been compensated for.

Now, the processing at the compensation value calculation section will be described in more detail.

FIG. 3 shows the arrangement of received pilot signals in the digitized I data and Q data. The pilot signal has a plurality of symbols 50. Assume that one symbol has K samples. The adjacent symbols (50 and 51) have a phase shift of θ. Likewise, the Q data 52 and 53 has a phase shift of θ. The compensation value calculation section 28 starts acquiring data at any position of the pilot signal. Here, the data refers to individual samples.

The timing at which data starts to be acquired is not limited to a particular one. This is because the present invention makes it possible to calculate compensation values if a predetermined number of pieces of data can be acquired from pilot signals having a phase shift of θ.

The process acquires P pieces of data from both the I data and the Q data. The invention is not limited to a particular value of P so long as it is greater than (K+hat L+2). Here, hat L is (L−1)/2 and L is the number of stages of the filter 21. For example, for a pilot signal with one symbol made up of 16 samples (K=16), it may be acceptable that hat L is roughly equal to 2. That is, so long as P is 20 or greater in the number of pieces of data, compensation values can be calculated with sufficient accuracy. Note that L does not have to be always an odd number, and if it is an even number, then the least digit may be incremented or decremented by one.

Next, (P−K) pieces of data are taken from the first portion of the acquired I data as a vector r_1,I, and the (P−K) pieces of data from the (K+1)-th to the end of the I-data are taken as a vector r_2,I.

On the other hand, (P−K) pieces of data is extracted from the first portion of the acquired Q data. Here, hat L pieces of data are added to the respective acquired pieces of data in front and at the end thereof. For example, hat L can be assumed to be two. Furthermore, assume that the acquired data from the Q-branch is arranged as v1, v2, v3, v4, v5, v6, v7, . . . . Here, focusing on data v3, assume that (v1, v3, v3, v4, v5) is a v3-based data set. In the same manner, if the data v4 is focused, then it is (v2, v3, v4, v5, v6, v7).

FIG. 3 shows hat L as an arrow. Note that hat L is (L−1)/2, and L is the number of stages of the filter. The number of stages of the filter x may be 3 to 4, which will allow calculations to be performed with sufficient accuracy.

In this manner, a (2 hat+1)×(P−K) matrix is obtained from the (P−K) pieces of data. Assume that this matrix is a matrix R_1,Q. On the other hand, (P−K) pieces of data from the (K+1)-th of the acquired data will be used to create a matrix R_2,Q in the same manner.

Then, the vector r_1,I, the vector r_2,I, the matrix R_1,Q, and the matrix R_2,Q are used to form the matrix Λ as in Equation (44). Furthermore, the vector r_I is created from the vector r_1,I and the vector r_2,I as in Equation (46). Then, the vector u of (L+2)×1 dimensions is found, for example, as in Equation (47). As already discussed above, the solution method herein may be used to find the pseudo-inverse matrix of the matrix Λ, and a well-known solution may also be used. Note that the Q data here is mathematically an imaginary number, and to find the vector u, complex number calculations have to be performed for calculations for each element of the vector r_1,I, the vector r_2,I, the matrix R_1,Q, and the matrix R_2,Q.

Using the resulting elements of the vector u, the CFO estimation value or hat ε, the I/Q imbalance compensation value or hat β, and the vector hat x are determined in accordance with Equation (48), Equation (49), and Equation (50). Note that θ is the phase difference between pilot signal symbols and thus a known value.

In the manner mentioned above, the compensation value calculation section determines the CFO estimation value and the I/Q imbalance compensation value.

FIG. 4 shows the arrangement for compensating the I-branch side signal. The transmitter 1 and the receiver 10 are fundamentally the same. However, the receiver 10 has a controller 40. The components other than the controller 40 are the same as those for compensating the Q data, and thus will not be explained repeatedly.

The I data supplied to the controller will be subjected to the filter x41 based on the compensation value. On the other hand, the Q data is multiplied by β at a multiplier section 42 and then added at an adder section 43 to the I data that has been acted upon by the filter x41. The resulting data is referred to as Ic data. The resulting signal is added at an adder section 45 as a real number to the Q data that has been imparted the imaginary unit “j” at an imaginary section 44. The output of the adder 45 is a complex number with the I/Q imbalance compensated for. This complex number is multiplied by a complex number to compensate for the CFO estimation value or ε.

The real part of the resulting complex number is a signal with the CFO and the I/Q imbalance having been compensated for.

Now, the processing performed by the compensation value calculation section will be explained in more detail.

As in FIG. 3, FIG. 5 shows the arrangement of the received pilot signal of the digitized I data and Q data. To compensate the I data, the matrix R_1,I and the matrix R_2,I are produced from the I data, while the vector r_1,Q and the vector r_2,Q are prepared from the Q data. These four matrices and vectors are produced exactly in the same manner as in FIG. 2.

Next, using the vector r_1,Q, the vector r_2,Q, the matrix R_1,I, and the matrix R_2,I, the matrix Λ is formed as in Equation (60). Furthermore, the vector r_1,Q and the vector r_2,Q are used to form a vector r_Q as in Equation (61). Then, the vector u of (L+2)×1 dimensions is determined, for example, as in Equation (62). As already discussed above, the solution method herein may be used to find the pseudo-inverse matrix of the matrix Λ, and a well-known solution may also be used. Note that the Q-branch signal here is mathematically an imaginary number, and to find the vector u, complex number calculations have to be performed for calculations for each element of the vector r_1,Q, the vector r_2,Q, the matrix R_1,I, and the matrix R_2,I.

Using the resulting elements of the vector u, the CFO estimation value or hat ε, the I/Q imbalance compensation value or hat β, and the vector hat x are determined based on Equation (63), Equation (64), and Equation (65). Note that θ is the phase difference between pilot signal symbols and thus a known value.

In the manner mentioned above, the compensation value calculation section determines the CFO estimation value and the I/Q imbalance compensation value.

Note that the discussions above have been made to consecutive pilot signals; however, a predetermined length of data may also be contained between pilot signals.

FIG. 6 shows the structure of the received data in such a case. That is, there exists data 63 indicative of communication contents between one-symbol pilot signal 61 and pilot signal 62. However, it is assumed that the relationship between the pilot signal and the data indicative of the communication contents is known. This case is not like the one that has been discussed so far, i.e., the case of consecutive pilot signals; however, the compensation method of the present invention discussed above can be applied even to such a case.

More specifically, a setting is made as P between the start of the pilot signal 61 to the end of the pilot 62. Then, another setting K is made between the beginning of the pilot signal 61 and the beginning of the pilot signal 62. That is, P to be set is larger. This makes it possible to obtain the first (P−K) pieces of data from the pilot signal 61. Furthermore, as for the train of the next (P−K) pieces of data, it can be obtained from the pilot signal 62 by taking the (P−K) pieces of data from the (K+1)-th to the P-th data. Hereinafter, it is possible to make a compensation in the same manner as discussed above.

Note that the present invention allows for determining the CFO and the I/Q imbalance compensation coefficient; however, only the CFO may be determined by another method, so that the resulting CFO value may be used to find the I/Q imbalance compensation coefficient. This is because the I/Q imbalance compensation coefficient can be determined based on Equation (49) and Equation (50) using the CFO estimation value.

Second Embodiment

Compensation Method for Existing Standards

For Wireless standard pilots, periodic pilots (PP) may be generally used, thereby allowing the GPP-based compensation method provided by the present invention to be applicable thereto. In this context, an example is shown here. FIG. 8 shows a preamble in accordance with the IEEE 802.11a WLAN standards, which is made up of two types of training series (TS). In the figure, the short TS includes the same ten pilot symbols, each symbol having 16 samples, and is used for signal detections, AGC, synchronous timing, and rough CFO estimations.

On the other hand, the long TS includes the same two pilot symbols, each symbol having 64 samples, and is used for channel estimation and accurate CFO estimation. Obviously, it can be seen that both the short TS and the long TS belong to PP.

<PP Caused Problems>

The periodic pilot (PP) can be found from GPP at θ=0. More specifically, the CFO estimation value and the imbalance coefficient can be expressed from Equation (48), Equation (49), and Equation (50) as in the following equations.

$\begin{array}{cc}\left[\mathrm{Equation}66\right]& \\ \hat{\varepsilon }=\frac{N}{2\pi K}\left[\arccos \left\{\frac{u\left(1\right)+u\left(2\right)}{2}\right\}\right]& \left(66\right)\\ \left[\mathrm{Equation}67\right]& \\ \hat{\beta }=\frac{u\left(2\right)-u\left(1\right)}{2\sin \left(2\pi \hat{\varepsilon }K/N\right)}& \left(67\right)\\ \left[\mathrm{Equation}68\right]& \\ \hat{x}={\frac{1}{\sin \left(2\pi \hat{\varepsilon }K/N\right)}\left[u\left(3\right),\dots ,u\left(L+2\right)\right]}^T& \left(68\right)\end{array}$

The above equations may pose two critical problems. One is that Equation (66) is an even function, and thus although the absolute value hat ε can be found, the sign of the CFO estimation value cannot be determined. The other is that since the denominator of Equation (67) and Equation (68)is zero at ε=0, the equations cannot be used to find β and the vector x. Although the LLS problem of Equation (47) is not in a pathological condition (which refers to an insoluble status), the terms of β and the vector x of Equation (44) vanish when the CFO is zero.

In practice, the MPP-based scheme may also encounter similar problems. For each pilot symbol of the PP, it can be set that hat K=K from the cycle prefix of the next pilot symbol, and a phase rotation of π/4 is removed from the related equations. This is an MPP-based scheme. Here, Equation (14) and Equation (17) are turned into the following equations.

$\begin{array}{cc}\left[\mathrm{Equation}69\right]& \\ \stackrel{⋓}{E}\left(\varepsilon \right)=\left[\stackrel{⋓}{e}\left(\varepsilon \right){\stackrel{⋓}{e}}^{*}\left(\varepsilon \right)\right]& \left(69\right)\\ \left[\mathrm{Equation}70\right]& \\ \stackrel{⋓}{e}\left(\varepsilon \right)={\left[{}^{j\frac{2\mathrm{\pi \varepsilon }K}{N}},{}^{\mathrm{j2}·\frac{2\mathrm{\pi \varepsilon }K}{N}},\dots ,{}^{j\hat{M}·\frac{2\mathrm{\pi \varepsilon }K}{N}}\right]}^T& \left(70\right)\end{array}$

The evaluation function for CFO estimation is given by the following equation.

[Equation 71]

{hacek over (J)}({tilde over (ε)})=tr{{hacek over (E)}({tilde over (ε)})({hacek over (E)}^H({tilde over (ε)}) {hacek over (E)}({tilde over (ε)}))⁻¹{hacek over (E)}^H({tilde over (ε)}){circumflex over (R)}{circumflex over (R)}^H}  (71)

In the non-Patent Document 2, it is pointed out that tick J (tilde ε) is in a pathological status at tilde ε=0. However, what is more critically problematic lies in that tick J (tilde ε) is an even function of tilde ε (which is proved at the end of the specification), and gives the maximum value at ε and −ε, i.e., yields ambiguity in the sign of CFO.

On the other hand, substituting the relations of ε=0 and the removal of a π/4 phase rotation into Equation (12) and Equation (13) yields hat r(m, k)=a(k)+b(k), resulting in the relational expressions of the vector hat R_Q(m+1)=vector hat R_Q(m) and the hat r_I(m+1)=hat r_I(m).

With Bar ω_m being defined as ω, it holds from ω=2π hat εK/N that bar ω_m=0 for hat ε=0. Accordingly, substituting these relations into Equation (28) and Equation (29) leads to the vector A(m)=0 and the vector B(m)=0, so that the optimum solutions of the vector x and β obtained here from Equation (27) are meaningless.

<Short and Long TS Based Compensation>

From the analytical results above, a simultaneous compensation for the CFO and the I/Q imbalance in accordance with the IEEE 802.11a requires the determination of the sign to of CFO and the algorithm for finding β and the vector x in the absence of the CFO.

The P samples of the short TS are arranged in the matrix of 2×(P−K) dimensions expressed by the following equation.

$\left[\mathrm{Equation}72\right]$ $\begin{array}{cc}R=\left[\begin{array}{ccc}r\left(n\right)& \cdots & r\left(n+P-K-1\right)\\ r\left(n+K\right)& \cdots & r\left(n+P-1\right)\end{array}\right]& \left(72\right)\end{array}$

In the similar manner as with Equation (7), the following equation is obtained from Equation (5).

[Equation 73]

R=E ₁(ε)[a^Tb^T]^T  (73)

In the equation above,

$\begin{array}{cc}\left[\mathrm{Equation}74\right]& \\ E_1\left(\varepsilon \right)=\left[\begin{array}{cc}1& 1\\ {}^{j\frac{2\pi \varepsilon K}{N}}& {}^{-j\frac{2\pi \varepsilon K}{N}}\end{array}\right]& \left(74\right)\\ \left[\mathrm{Equation}75\right]& \\ a=\left[a\left(n\right),\dots ,a\left(n+P-K-1\right)\right]& \left(75\right)\\ \left[\mathrm{Equation}76\right]& \\ b=\left[b\left(n\right),\dots ,b\left(n+P-K-1\right)\right]& \left(76\right)\end{array}$

Obviously, the vector “a” and the vector “b” represent the desired signal and the image interference signal, respectively. In general, the power of the vector “b” given by the I/Q imbalance is less than the power of the vector “a.” It holds that ε is included in the range of (−N/2K, N/2K), and if is not equal to 0, the vector E₁(ε) is a full-rank matrix as expressed in the following equation.

$\left[\mathrm{Equation}77\right]$ $\begin{array}{cc}\left[\begin{array}{c}a\\ b\end{array}\right]=E_1^{-1}\left(\varepsilon \right)R=\frac{1}{-j·2\sin \left(2\pi \varepsilon K/N\right)}E_2R& \left(77\right)\end{array}$

In the equation above, the relation expressed by the following equation is given.

$\left[\mathrm{Equation}78\right]$ $\begin{array}{cc}{E}_2\left(\varepsilon \right)=\left[\begin{array}{cc}{}^{-j\frac{2\pi \varepsilon K}{N}}& -1\\ -{}^{j\frac{2\pi \varepsilon K}{N}}& 1\end{array}\right]& \left(78\right)\end{array}$

Furthermore, since the vector E₁(−ε) is a matrix with the column elements of the vector E₁(ε) replaced and the relation of the vector E₁(−ε)=the vector. E₁*(ε) holds, the relational expression is obtained as below.

$\left[\mathrm{Equation}79\right]$ $\begin{array}{cc}\left[\begin{array}{c}b\\ a\end{array}\right]=E_1^{-1}\left(-\varepsilon \right)R=\frac{1}{j·2\sin \left(2\pi \varepsilon K/N\right)}E_2^{*}R& \left(79\right)\end{array}$

These two relational expressions imply that the sign of the absolute value hat ε obtained from Equation (66) can be determined by comparing the power of the first row and the second row of Equation (80).

[Equation 80]

E₂(|{circumflex over (ε)}|)R  (80)

In other words, if a first row norm is greater than a second row norm, then the CFO estimation value is the absolute value of hat ε, whereas if the first row norm is less than the second row norm, then the CFO estimation value is the “−” absolute value of hat ε.

According to this simple determination of the sign of CFO, the CFO estimation range is extremely important in the absence of the I/Q imbalance, that is, it is critical that c lies in the range of (−N/2K, N/2K). Furthermore, use is made of the vector E₁(0) being a unit matrix to check the existence of CFO in accordance with Equation (81) which is the conventional autocorrelation based scheme.

$\left[\mathrm{Equation}81\right]$ $\begin{array}{cc}{\hat{\varepsilon }}_a=\frac{N}{2\pi K}\arg \left\{\sum _{k=0}^{P-K-1}r^{*}\left(n+k\right)r\left(n+k+K\right)\right\}& \left(81\right)\end{array}$

Note that if the absolute value hat ε_a<Δε, then hat ε=0. However, Δε is a threshold value and is the same as the search resolution of tick J (hat ε) in the MPP-based scheme.

On the other hand, to find β and the vector x, use is made of the frequency domain representation (FDR) of the short TS and the long TS known in the receiver. It should be noted that there exists a transpose matrix of twelve non-zero elements (S_t,1, . . . , S_t,12) in the FDR of the short TS. To find the time domain representation of the short TS, use is made of the matrix expressed by the following equation which includes twelve fixed subcarriers.

[Equation 82]

W_t=[f₆₄⁴, f₆₄⁸, . . . , f₆₄²⁴, f₆₄⁴⁰, f₆₄⁴⁴, . . . , f₆₄⁶⁰]  (82)

In the above equation, the vector f_Nⁱ denotes the (I+1)-th column vector of the N×N IDFT matrix F^H. Furthermore, the diagonal matrix expressed by the following equation is given.

[Equation 83]

S_t=diag{S_t,1, . . . , S_t,12}  (83)

In the same manner, the diagonal matrix vector S_T is constructed of the same twelve subcarrier elements or the FDR of the long TS.

[Equation 84]

S_T=diag{S_T,1, . . . , S_T,12}  (84)

The four pilot symbols from the endmost of the short TS and the pilot symbol on the forefront stage of the long TS are used, in the case of which they are obviously two types of OFDM symbols that are different from each other. First, the following equations are defined.

$\begin{array}{cc}\left[\mathrm{Equation}85\right]& \\ R_{t,Q}=\left[\begin{array}{ccccc}{r}_Q\left(\hat{n}+\hat{L}\right)& \cdots & r_Q\left(\hat{n}\right)& \cdots & {\hat{r}}_Q\left(\hat{n}-\hat{L}\right)\\ r_Q\left(\hat{n}+\widehat{L+1}\right)& \cdots & r_Q\left(\hat{n}+1\right)& \cdots & r_Q\left(\hat{n}-\hat{L}+1\right)\\ ⋮& ⋮& ⋮& ⋮& ⋮\\ r_Q\left(\hat{n}+\hat{L}+N-1\right)& \cdots & r_Q\left(\hat{n}+N-1\right)& \cdots & r_Q\left(\hat{n}-\hat{L}+N-1\right)\end{array}\right]& \left(85\right)\\ \left[\mathrm{Equation}86\right]& \\ r_{t,I}={\left[r_I\left(\hat{n}\right),\dots ,r_I\left(\hat{n}+N-1\right)\right]}^T& \left(86\right)\\ \left[\mathrm{Equation}87\right]& \\ R_{T,Q}=\left[\begin{array}{ccccc}{r}_Q\left(\hat{n}+\hat{L}\right)& \cdots & r_Q\left(\hat{n}\right),& \cdots & r_Q\left(\hat{n}-\hat{L}\right)\\ r_Q\left(\hat{n}+\hat{L}+1\right)& \cdots & r_Q\left(\hat{n}+1\right),& \cdots & r_Q\left(\hat{n}-\hat{L}+1\right)\\ ⋮& & ⋮& & ⋮\\ r_Q\left(\overleftarrow{n}+\hat{L}+N-1\right)& \cdots & r_Q\left(\hat{n}+N-1\right),& \cdots & r_Q\left(\hat{n}-\hat{L}+N-1\right)\end{array}\right]& \left(87\right)\\ \left[\mathrm{Equation}88\right]& \\ r_{T,I}={\left[r_I\left(\hat{n}\right),\dots r_I\left(\hat{n}+N-1\right)\right]}^T& \left(88\right)\end{array}$

In the equations above, hat n is the index of the first sample to of the received t₇. Assuming that tick K is the sample spacing between t₇ and T₁, and putting hat n into (hat n+hat K) will makes it possible to determine the vector r_T,I and the vector R_T,Q for the first pilot symbol of the long TS.

When the channel has an invariable preamble interval and the I/Q imbalance is compensated for, the equation below is given.

[Equation 89]

Z _t(r_t,I+j(R_t,Qx+βr_t,I)=Z_T(r_T,I+j(R_T,Qx+βr_T,I)  (89)

In the equation above, the following relational expressions can be given as below.

$\begin{array}{cc}\left[\mathrm{Equation}90\right]& \\ Z_t={}^{\mathrm{j2\pi \varepsilon }\stackrel{⋓}{K}/N}S_t^{-1}W_t^{\mathscr{H}}{\Gamma }^{\mathscr{H}}\left(\varepsilon \right)& \left(90\right)\\ \left[\mathrm{Equation}91\right]& \\ Z_T=S_T^{-1}W_t^{\mathscr{H}}{\Gamma }^{\mathscr{H}}\left(\varepsilon \right)& \left(91\right)\\ \left[\mathrm{Equation}92\right]& \\ \Gamma \left(\varepsilon \right)=\mathrm{diag}\left\{1,{}^{j\frac{2\mathrm{\pi \varepsilon }}{N}},\dots ,{}^{j\frac{2\mathrm{\pi \varepsilon }\left(N-1\right)}{N}}\right\}& \left(92\right)\end{array}$

Here, using the LLS algorithm, the optimum solution can be found from Equation (81) as in the following equation.

$\begin{array}{cc}\left[\mathrm{Equation}93\right]& \\ \left[\begin{array}{c}\hat{x}\\ \hat{\beta }\end{array}\right]={}^{†}& \left(93\right)\end{array}$

In the equation above,

$\begin{array}{cc}\left[\mathrm{Equation}94\right]& \\ =\left[\begin{array}{cc}{Z}_{t,Q}R_{t,Q}-Z_{T,Q}R_{T,Q}& Z_{t,Q}r_{t,I}-Z_{T,Q}r_{T,I}\\ Z_{T,I}R_{T,Q}-Z_{t,I}R_{t,Q}& Z_{T,I}r_{T,I}-Z_{t,I}r_{t,I}\end{array}\right]& \left(94\right)\\ \left[\mathrm{Equation}95\right]& \\ =\left[\begin{array}{c}{Z}_{t,I}r_{t,I}-Z_{T,I}r_{T,I}\\ Z_{t,Q}r_{t,I}-Z_{T,Q}r_{T,I}\end{array}\right]& \left(95\right)\end{array}$

In the above equations, the vector Z_t,I, the vector Z_t,Q, the vector Z_t,I, and the vector Z_T,Q are the real part and the imaginary part of the vector Z_t and the vector Z_T, respectively. Since the vector S_t and the vector S_T are not equal, the shaded vector A and the shaded vector B are a full-rank matrix even for ε=0. Accordingly, β and the vector x can be determined from Equation (90) even in the absence of CFO.

In practice, this algorithm is applicable even when it does not hold that ε=0; however, it is used only when the CFO estimation value is close to zero because of its heavy calculation load. In practice, the samples of the short TS cannot always constitute an OFDM symbol of N samples.

In such a case, although the CFO has just been estimated, the left side of Equation (86) has to be corrected. Assume that use can be made of only t₉ and t₁₀ without loss of generality. Here, the vector R_{t, Q} and the vector r_T,I have a row of size N/2.

$\begin{array}{cc}\left[\mathrm{Equation}96\right]& \\ {\hat{W}}_t=\left[f_{32}^2,f_{32}^4,\dots ,f_{32}^{12},f_{32}^{20},f_{32}^{22},\dots ,f_{32}^{30}\right]& \left(96\right)\\ \left[\mathrm{Equation}97\right]& \\ \hat{\Gamma }\left(\varepsilon \right)=\mathrm{diag}\left\{1,{}^{j\frac{2\mathrm{\pi \varepsilon }}{N}},\dots ,{}^{j\frac{2\mathrm{\pi \varepsilon }\left(N/2-1\right)}{N}}\right\}& \left(97\right)\end{array}$

The equations above can be replaced by the vector Z_t expressed by the following equation, thereby allowing for obtaining β and the vector x from Equation (64).

$\begin{array}{cc}\left[\mathrm{Equation}98\right]& \\ {\hat{Z}}_t=\frac{1}{\sqrt{2}}{}^{j\frac{2\mathrm{\pi \varepsilon }\left(\stackrel{\_}{K}-2K\right)}{N}}S_t^{-1}{\hat{W}}_t^H{\hat{\Gamma }}^H\left(\varepsilon \right)& \left(98\right)\end{array}$

In summary, using the CFO enhanced estimation method (CEE) involving the detection of CFO signs and the autocorrelation based CFO estimator as well as the algorithm based on the short TS and the long TS, the GPP-based scheme can be applied, for example, to the IEEE 802.11a. Since the structure of PP for synchronization and the pilot for channel estimation are generalized, the present invention is applicable to other wireless standards.

On the other hand, to estimate the sign of CFO, the following determination method can also be employed. The cost function for determining a sign using the matrices Z_t and Z_T, and the vectors double-dot r_t and r_T is given as in Equation (96) under the condition that hat ε is not zero.

[Equation 99]

J(|{circumflex over (ε)}|)=∥Z_t{umlaut over (r)}_t−Z_T{umlaut over (r)}_T∥²  (99)

Here, J (the “−” absolute value hat ε) corresponds to the opposite of CFO. Accordingly, the CFO is positive if J (the “−” absolute value hat ε) is greater than J (the absolute value hat ε), and is negative in the opposite case.

Now, a description will be made to a specific embodiment. This embodiment takes as an example a signal according to the IEEE 802.11a WLAN. Accordingly, there exists no phase difference θ between pilot signals. That is, θ=0. The hardware structure is the same as that of FIG. 2.

FIG. 7 shows the processing flow followed by the control section. When the processing has started (S100), the determination of termination is first made (S102), and then data is acquired (S104) to check if CFO is zero (S106). This determination is made by knowing if hat εa is less than Δε according to Equation (58) for determining CFO. If the CFO is present, the CFO is determined (S108). The CFO is determined basically in the same manner as explained in relation to the first embodiment. Finally, the CFO estimation value, and the compensation coefficients for the I/Q imbalance, i.e., the β and the vector x are determined from Equation (48), Equation (49), and Equation (50).

Now, the sign of CFO is determined (S110). The determination of the sign of CFO is carried out by comparing the first row norm with the second row norm of Equation (80). Alternatively, it may also be acceptable to use the determination cost function J of Equation (96). Then, the sign of CFO is determined and the I/Q imbalance compensation coefficient is found (S112), then performing the compensation processing (S114). The compensation processing is the same as the compensation processing discussed in the first embodiment.

If the CFO has been determined to be zero, the I/Q imbalance compensation coefficients, i.e., β and the vector x are determined based on Equation (90) (S116). Then, the compensation processing is performed with the CFO being zero (S114). The compensation processing can be carried out in the same manner as with the case of the CFO being not zero. Additionally, only the CFO compensation processing may be skipped. This is because the CFO compensation processing is carried out only after the compensation for the I/Q imbalance has been completed.

Now, a detailed description will be made to the contents of each processing.

Upon reception of the I data and the Q data, the controller acquires P samples from the short TS. Then, the controller acquires the (P−K) pieces of data (hereinafter referred to as the “n series data”) from the first piece of data, and the (P−K) pieces of data (hereinafter referred to as the “n+K series data”) from the K-th. Note that these pieces of data are a complex number. More specifically, the complex number has the I data as its real part and the Q data as its imaginary part.

To realize Equation (81) for determining CFO, the process determines the sum of products of the conjugate complex number of the n series data and the complex number of the (n+K) series. The conjugate complex number of the n series data is the complex number obtained by multiplying the sign of the imaginary part by minus one. This is the multiplication of complex numbers, yielding a resulting complex number. Accordingly, its sum is also a complex number. Now, the process determines the argument (arg) of the complex number. More specifically, the process determines the angle between the real part and the imaginary part using the arctan function. The argument is multiplied by N/(2πK) to find hat ε_a.

Now, a description will be made to the determination of the sign of CFO. To make this determination, the matrix R of Equation (72) is found and the matrix product of the matrix E₂ (the absolute value hat ε) is determined. The matrix R is a matrix with the n series data disposed in the first row and the (n+K) series data placed in the second row. Of course, the individual elements are a complex number. The process determines the first row norm and the second row norm of the matrix product. The norm is defined as the result of multiplying the element in the respective rows by the conjugate value and taking the square root of the sum thereof.

The CFO estimation value remains unchanged as the absolute value if the first row norm is greater than the second row norm, and if not, the CFO estimation value will be the absolute value multiplied by −1.

Now, a description will be made to the processing to be performed when the CFO is determined to be zero. The process chooses twelve non-zero elements (S_t,1, . . . , S_t,12) from four symbols, i.e., t₇ to t₁₀ of the short TS. In particular, it is assumed here that twelve subcarriers, i.e., subcarriers 4, 8, 12, 16, 20, 24, 40, 44, 48, 52, 56, and 60 among 64 subcarriers transmit non-zero elements.

Then, the process forms a matrix of the subcarriers of Equation (93). Note that each element is the (i+1)-th column vector of the following IDFT matrix F^H. Note that the IDFT is known from the communication system specification.

$\begin{array}{cc}\left[\mathrm{Equation}100\right]& \\ F_N=\frac{1}{\sqrt{N}}\left[\begin{array}{ccccc}1& {}^{j\frac{2\pi ·1·0}{N}}& \cdots & \cdots & {}^{j\frac{2\pi ·\left(N-1\right)·0}{N}}\\ 1& {}^{j\frac{2\pi ·1·1}{N}}& & & {}^{j\frac{2\pi ·\left(N-1\right)·1}{N}}\\ ⋮& & \ddots & & ⋮\\ ⋮& & & \ddots & ⋮\\ 1& {}^{j\frac{2\pi ·1·\left(N-1\right)}{N}}& \cdots & \cdots & {}^{j\frac{2\pi ·\left(N-1\right)·\left(N-1\right)}{N}}\end{array}\right]& \left(100\right)\end{array}$

Furthermore, the respective element is the diagonal matrix of (83).

The same processing is also carried out for the long TS. The subcarriers to be chosen are the same as those selected for the short TS. The corresponding elements are (S_T,1, . . . , S_T,12). These elements and subcarriers have known values and thus can be determined in advance.

Then, the process acquires N pieces of data from the (hat n)-th of the t7 symbol in the short TS. More precisely, the data is (N+hat L) pieces of data from the (hat n−hat L)-th to the (hat n−hat L+N−1)-th. These pieces of data are acquired for both the I-branch and the Q-branch. Then, the Q-branch side data is used to form the matrix R_t,Q of Equation (84) and the I-branch side data to form the r_t,I.

In the same manner, for the long TS, the process acquires (N+hat L) pieces of data from the (hat n−hat L)-th to the (hat n−hat L+N−1)-th in the I-branch and the Q-branch, to form the matrix R_T,Q in a similar manner and to form the matrix r_T,I based on the I-branch side data.

The matrix Z_t and the matrix Z_T are determined from Equation (87) and Equation (88). To create these matrices, the matrix S, the matrix W, and the matrix Γ are required, but they have already been determined by Equation (83), Equation (83), and Equation (89). The preparation having been made so far makes it possible to create the shaded vector A and the shaded vector B of Equation (91) and Equation (92). Then, based on this, the process can determine β and the vector x from Equation (90). Note that Equation (90) determines the pseudo-inverse matrix of the shaded vector; however, as described in the first embodiment, another well-known method may also be used to determine β and the vector x.

Note that in the discussions above, four symbols of the short TS were used. However, the invention is not limited to these four symbols. For example, two symbols may also be used to provide the same effects as above. For example, in the case where t₉ and t₁₀ of the short TS are used, the subcarriers can be chosen as in Equation (93), so that the matrices Γ and Z_t are substituted into Equation (94) or Equation (95) to determine β and the vector x in the same manner.

Implementation Example

Simulation Results

Simulations were performed to verify the effectiveness of the GPP-based scheme according to the present invention. To this end, 64-QAM data modulated OFDM signals were used. Note that B=20 MHz, N=64, and N_GI=16.

The frequency selective fading channel was the one with the power profile attenuated exponentially, the CFO was 90 KHz, and the I/Q imbalance was adopted from the two cases of Non-Patent Document 2.

Case A) α=1 dB, φ=5°, and the non-frequency-selective and the frequency selective imbalance with the vector g_I being the transpose matrix of (1,0,1) and the vector g_Q being the transpose matrix of (0,1,1).

Case B) α=1 dB, φ=5°, and the non-frequency-selective imbalance with the vector g_I and the vector g_Q being the transpose matrix of (1,0).

Under the conditions of M=10, K=16, and L=5, a comparison was made between the MGPP-based scheme and the MPP-based scheme. As for the MPP-based scheme, the search range of ε is (−0.48, 0.48), and the search interval is Δε=0.01. Taking into account the effects of the AGC and the residual direct current offset, employed were hat M=6 pilot symbols and P=64 samples.

FIG. 9 is a view showing a comparison between the mean square error E ((ε−hat ε)²) of the normalized CFO and SNR. The GPP-based scheme gives good estimation results, whereas the MPP-based scheme exhibits an error flow at high SNR because of the absence of the optimum CFO value at the search points.

FIG. 10 shows the characteristics of the bit error rate (BER) versus SNR. For information, note that the characteristic with no CFO and I/Q imbalance (No CFO I/Q) is also shown. From the figure, it can be seen that the GPP-based scheme perfectly compensates for the CFO and the imbalance.

On the other hand, the MPP-based scheme causes an error flow at high SNR in Case A. This error flow was mainly caused by being incapable of taking the GI removal precisely, i.e., grasping the convolution structure with accuracy. It was also caused by the assumption that the timing error was to occur before compensation and then to be corrected.

FIG. 11 shows that the GPP-based scheme is hardly affected by the timing error. On the other hand, it is shown in Case A that the MPP-based scheme will not be properly followed without perfectly synchronous timing.

Now, FIG. 12 shows simulation results for the IEEE 802.11a WLAN which works in the 36 Mbps mode of operation. This simulation employed the block error rate (BLER) for a 1000-byte block size in order to measure operations in accordance with the MPP-based scheme, the MPP-based scheme having CEE (MPP-CEE), and the extended GPP-based scheme.

It can be seen regarding the MPP-based scheme that the ambiguity of the sign of CFO prevents reaching a BLER of 0.1 even at high SNR. It is seen that the MPP-CEE does not operate properly at CFO=zero but is provided with a fairly improved characteristic. On the other hand, it can also be seen that the extended GPP-based scheme operates properly in any cases.

A new technique is suggested for simultaneous compensation of CFO and I/Q imbalance. The basic mutual relations between pilots are studied and the NLS problem is changed into the LLS problem for estimation of CFO, whereby the present invention can analytically obtain all the CFO and imbalance coefficients. The invention realizes robust synchronous timing and significant reduction in calculation load.

Furthermore, since the periodic pilot (PP) is contained in GPP, ambiguity in determination of the sign of CFO and the problem of compensating for zero CFO can be addressed, thereby allowing the suggested technique to be applicable to the actual wireless standards such as the IEEE 802.11a WLAN. Furthermore, the effectiveness of the suggested technique was shown through the simulations for various CFO and I/Q imbalance compensations.

<Proof of tick J (−tilde ε)=tick J (tilde ε)>

Equation (45) is rewritten into tick J (tilde ε)=tr{vector G(tilde ε) vector hat R vector hat R^H}. Here,

$\begin{array}{cc}\left[\mathrm{Equation}101\right]& \\ \begin{array}{c}G\left(\tilde{\varepsilon }\right)=\check{E}\left(\tilde{\varepsilon }\right){\left({\check{E}}^H\left(\tilde{\varepsilon }\right)\check{E}\left(\tilde{\varepsilon }\right)\right)}^{-1}{\check{E}}^H\left(\tilde{\varepsilon }\right)\\ =\frac{\hat{M}}{{\hat{M}}^2-{q}^2}\left\{\begin{array}{c}\check{e}\left(\tilde{\varepsilon }\right){\check{e}}^H\left(\tilde{\varepsilon }\right)+{\check{e}}^{*}\left(\tilde{\varepsilon }\right){\check{e}}^T\left(\tilde{\varepsilon }\right)-\\ \frac{1}{\hat{M}}\left[\begin{array}{c}q\check{e}\left(\tilde{\varepsilon }\right){\check{e}}^T\left(\tilde{\varepsilon }\right)+\\ q^{*}{\check{e}}^{*}\left(\tilde{\varepsilon }\right){\check{e}}^H\left(\tilde{\varepsilon }\right)\end{array}\right]\end{array}\right\}\end{array}& \left(101\right)\end{array}$

In the equation above, q=tick vector e^H (tilde ε) tick vector e*(tilde ε) is a scalar quantity. It can be seen from Equation (66) that vector G(tilde ε) is constant even when tick vector e (tilde ε) is replaced by tick vector e*(tilde ε). Accordingly, from tick vector e (tilde ε)=tick vector e*(tilde ε), vector G(−tilde ε)=vector G(tilde ε), resulting in tick J (−tilde ε)=tick J (tilde ε).

INDUSTRIAL APPLICABILITY

The present invention is applicable to the OFDM scheme communication method and those transmitters and receivers that implement the method.

Claims

1. A CFO estimation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then estimating a CFO of the signal, the method comprising the steps of: digitizing the I-branch side signal of the received pilot signal into I data;digitizing the Q-branch side signal of the received pilot signal into Q data;forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (34);forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (37);forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (35);forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (38);determining a matrix u being equal, when multiplied by a matrix of Equation (46) obtained from the Equation (34) and the Equation (37), to a matrix of Equation (45) obtained from the Equation (34), the Equation (37), the Equation (35), and the Equation (38); anddetermining a CFO estimation value ε based on Equation (48) from first and second elements of the matrix u,

2. An I/Q imbalance compensation coefficient calculation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then calculating a compensation coefficient to compensate for the I/Q imbalance of the signal, the method comprising the steps of: digitizing the I-branch side signal of the received pilot signal into I data;digitizing the Q-branch side signal of the received pilot signal into Q data;forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (34);forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (37);forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (35);forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (38);determining a matrix u being equal, when multiplied by a matrix of Equation (46) obtained from the Equation (34) and the Equation (37), to a matrix of Equation (45) obtained from the Equation (34), the Equation (37), the Equation (35), and the Equation (38);determining an I/Q imbalance compensation coefficient β from first and second elements of the matrix u and a CFO value ε based on Equation (49); anddetermining an I/Q imbalance compensation coefficient vector x from elements other than the first and second elements of the matrix u and the CFO value hat ε based on Equation (50),

3. An I/Q imbalance compensation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and thereafter compensating the signal, the method comprising the steps of: digitizing the I-branch side signal of the received signal into I data;digitizing the Q-branch side signal of the received signal into Q data;multiplying the Q data by the vector x determined according to the method of claim 2;multiplying the I data by β determined according to the method of claim 2;adding data obtained by multiplying the I data by β to the Q data multiplied by the vector x to yield Qc data; anddetermining a complex number with the I data employed as a real part and the Qc data employed as an imaginary part.

4. A signal compensation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and thereafter compensating the signal, the method comprising the step of compensating the complex number determined in claim 3, based on the CFO estimation value determined by a CFO estimation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then estimating a CFO of the signal, the method comprising the steps of: digitizing the I-branch side signal of the received pilot signal into I data;digitizing the Q-branch side signal of the received pilot signal into Q data;forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (34);forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (37);forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (35);forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (38);determining a matrix u being equal, when multiplied by a matrix of Equation (46) obtained from the Equation (34) and the Equation (37), to a matrix of Equation (45) obtained from the Equation (34), the Equation (37), the Equation (35), and the Equation (38); anddetermining a CFO estimation value ε based on Equation (48) from first and second elements of the matrix u,

5. A CFO estimation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then estimating a CFO of the signal, the method comprising the steps of: digitizing the I-branch side signal of the received pilot signal into I data;digitizing the Q-branch side signal of the received pilot signal into Q data;forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (51);forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (53);forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (52);forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (54);determining a matrix u being equal, when multiplied by a matrix of Equation (61) obtained from the Equation (51) and the Equation (53), to a matrix of Equation (60) obtained from the Equation (51), the Equation (53), the Equation (52), and the Equation (54); anddetermining a CFO estimation value ε from first and second elements of the matrix u based on Equation (63),

6. An I/Q imbalance compensation coefficient calculation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then calculating a compensation coefficient to compensate for the I/Q imbalance of the signal, the method comprising the steps of: digitizing the I-branch side signal of the received pilot signal into I data;digitizing the Q-branch side signal of the received pilot signal into Q data;forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (51);forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (53);forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (52);forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (54);determining a matrix u being equal, when multiplied by a matrix of Equation (61) obtained from the Equation (51) and the Equation (53), to a matrix of Equation (60) obtained from the Equation (51), the Equation (53), the Equation (52), and the Equation (54);determining an I/Q imbalance compensation coefficient β from first and second elements of the matrix u and a CFO value ε based on Equation (64); anddetermining an I/Q imbalance compensation coefficient vector x from elements other than the first and second elements of the matrix u and the CFO value hat ε based on Equation (65),

7. An I/Q imbalance compensation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and thereafter compensating the signal, the method comprising the steps of: digitizing the I-branch side signal of the received signal into I data;digitizing the Q-branch side signal of the received signal into Q data;multiplying the I data by the vector x determined by the method according to claim 6;multiplying the Q data by β determined according to an I/Q imbalance compensation coefficient calculation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then calculating a compensation coefficient to compensate for the I/Q imbalance of the signal, the method comprising the steps of:digitizing the I-branch side signal of the received pilot signal into I data;digitizing the Q-branch side signal of the received pilot signal into Q data;forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (34);forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (37);forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (35);forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (38);determining a matrix u being equal, when multiplied by a matrix of Equation (46) obtained from the Equation (34) and the Equation (37), to a matrix of Equation (45) obtained from the Equation (34), the Equation (37), the Equation (35), and the Equation (38);determining an I/Q imbalance compensation coefficient β from first and second elements of the matrix u and a CFO value ε based on Equation (49); anddetermining an I/Q imbalance compensation coefficient vector x from elements other than the first and second elements of the matrix u and the CEO value hat ε based on Equation (50),

8. A signal compensation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and thereafter compensating the signal, the method comprising the step of compensating the complex number determined in claim 7 based on the CFO estimation value determined by a CFO estimation method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and then estimating a CFO of the signal, the method comprising the steps of: digitizing the I-branch side signal of the received pilot signal into I data;digitizing the Q-branch side signal of the received pilot signal into Q data;forming (P−K) samples from an n-th sample of the I data into a matrix of Equation (51);forming (P−K) samples from an (n+K)-th sample of the I data into a matrix of Equation (53);forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data into a matrix of Equation (52);forming (P−K+(L−1)/2) samples from an (n+K−(L−1)/2)-th sample of the Q data into a matrix of Equation (54),determining a matrix u being equal, when multiplied by a matrix of Equation (61) obtained from the Equation (51) and the Equation (53), to a matrix of Equation (60) obtained from the Equation (51), the Equation (53), the Equation (52), and the Equation (54); anddetermining a CFO estimation value ε from first and second elements of the matrix u based on Equation (63),

9. A CFO sign determination method for receiving a signal having a pilot signal, demodulating the signal at a demodulator having an I-branch and a Q-branch, and determining a sign of a CFO of the signal, the method comprising the steps of: digitizing the I-branch side signal of the received pilot signal into I data;digitizing the Q-branch side signal of the received pilot signal into Q data;creating a matrix R of Equation (72) with a first row and a second row, the first row having (P−K) pieces of complex data with (P−K) samples from an n-th sample of the I data employed as a real part and (P−K) samples from an nth sample of the Q data employed as an imaginary part, the second row having (P−K) pieces of complex data with (P−K) samples from an (n+K)-th sample of the I data employed as a real part and (P−K) samples from an (n+K)-th sample of the Q data employed as an imaginary part;creating a matrix of Equation (78) based on an absolute value of a CFO estimation value ε whose sign is wanted to be determined;multiplying the Equation (72) by Equation (78); andcomparing a norm of a first row of the resulting matrix with a norm of a second row to determine that the sign of ε is positive when the first row norm is greater than the second row norm,

10. An I/Q imbalance compensation coefficient calculation method for demodulating a signal at a demodulator having an I-branch and a Q-branch, the signal containing a pilot signal with a short TS and a long TS and with no phase difference between adjacent symbols, and for calculating a compensation coefficient to compensate for an I/Q imbalance of the signal, the method comprising the steps of: selecting a predetermined subcarrier from the respective short TS and long TS to create a matrix of Equation (82);creating a diagonal matrix of Equation (83) from a subcarrier element of the short TS;creating a diagonal matrix of Equation (84) from a subcarrier element of the long TS;creating a diagonal matrix of Equation (92) from a CFO value whose absolute value is less than a predetermined value;creating Equation (90) from the Equation (82), the Equation (83), and the Equation (89);creating Equation (91) from the Equation (82), the Equation (84), and the Equation (89);forming (P−K) samples from an n-th sample of the I data of the short TS into a matrix of Equation (86);forming (P−K+(L−1)/2) samples into a matrix of Equation (85) from an (n−(L−1)/2)-th sample of the Q data of the short TS;forming (P−K) samples from an n-th sample of the I data of the long TS into a matrix of Equation (88);forming (P−K+(L−1)/2) samples from an (n−(L−1)/2)-th sample of the Q data of the long TS into a matrix of Equation (87);creating Equation (94) from the Equation (85), the Equation (86), the Equation (87), the Equation (88), the Equation (90), and the Equation (91);obtaining Equation (95) from the Equation (86), the Equation (88), the Equation (90), and the Equation (91); anddetermining a vector being equal, when multiplied by Equation (94), to Equation (95),

11. A transmission method for time division multiplexing and then transmitting a main signal and a pilot signal, the method comprising the steps of: time division multiplexing the main signal and periodic pilot signal; andimparting a predetermined phase difference to the pilot signal during the time division multiplexing.