Discrete multi-tone system having DHT-based frequency-domain equalizer

Abstract

A discrete multi-tone (DMT) communication system is provided herein, replacing the conventional inversed discrete Fourier transform (IDFT) and DFT with IDHT (together with the complex-to-real transformation) and a DHT alone respectively at the transmitting end and the receiving end. A DHT-based frequency-domain equalizer (FEQ) at the receiving end equalizes each of the 0-th to (N−1)-th DHT subchannels, where N is the number of point of the DHT. Finally, each of the 0-th to (N/2−1)-th subchannels of the DMT system is obtained by combining the k-th and (N-k)-th subchannels of the DHT-based FEQ for k=0, 1, . . . , (N/2−1).

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention generally relates to discrete multi-tone communication systems, and more particularly to a discrete multi-tone communication system using discrete Hartley transform for modulation and demodulation.

2. The Prior Arts

Discrete multi-tone (DMT) is a multi-carrier modulation (MCM) technique commonly applied in wireline communications such as digital subscriber loops (xDSL) including ADSL, ADSL2, ADSL2+, and VDSL, and power-line communications such as HomePlug. The basic idea of DMT is that a large number of sinusoids (i.e., subcarriers) are modulated by complex-valued quadrature amplitude modulation (QAM) symbols derived from an input bit stream, and transmitted in parallel. Performing a modulation on the complex valued constellation points generates samples of the continuous-time signal to be transmitted during a DMT symbol period. At the receiver, the QAM symbols are recovered by performing a demodulation on the analog-to-digital-converted received signal. A typical DMT system for xDSL is shown in FIG. 1. As illustrated, discrete Fourier transform (DFT) and IDFT (i.e., Inversed DFT) are adopted for baseband demodulation and modulation for the DMT system in the receiving end (Rx) and transmitting end (Tx) (separated by the dashed line in FIG. 1) respectively. A pair of digital-to-analog (D/A) converter and an analog-to-digital (A/D) converter is also provided around the analog communication channel. A time-domain equalizer (TEQ) is used at the receiving end to shorten the channel dispersion and, thereby, minimize intersymbol interference (ISI) caused by channel distortion resulted from, for example, long loop length, gauge variation, and bridge-tap. In addition, to compensate the phase rotation and amplitude distortion existing on each subchannel in the frequency domain, a frequency-domain equalizer (FEQ) is deployed. Also, to guard against the ISI, cyclic prefix (CP) are added and removed at the transmitting end and the receiving end, respectively.

The modulation and demodulation operations performed by the IDFT and DFT can be expressed mathematically as follows: $\begin{array}{cc}\mathrm{IDFT}\text{:}{x}_n=\frac{1}{N}\sum _{k=0}^{N-1}{X}_k·W_N^{-\mathrm{kn}},n=0,1,\cdots ,N-1& \left(1\right)\\ \mathrm{DFT}\text{:}{Y}_k=\sum _{n=0}^{N-1}{y}_n·W_N^{\mathrm{kn}},k=0,1,\cdots ,N-1& \left(2\right)\end{array}$ where $W_N=e^{-j\frac{2\pi }{N}},$ X_k or Y_k is the complex transmission symbol on the k-th subchannel in frequency domain, x_n or y_n is the n-th transmission sample in time domain, and N is the number of points for the IDFT/DFT.

The FEQ or, as it is used behind the DFT, the DFT-based FEQ can be implemented using various algorithms. For example, the MMSE (minimum mean-square error) algorithm tries to minimize the estimated error between the equalized signal and the transmitted signal. Among these algorithms, the LMS (least mean square) algorithm is most popular in terms of VLSI implementation. The LMS algorithm in the DFT-based FEQ consists of three operations: filtering, error estimation, and coefficient updating, expressed mathematically as follows:Filtering: Ŷ_k=Y_k·W_k*=(Y_k,R+jY_k,I)·(W_k,R−jW_k,I)  (3)Error estimation: E_k=X_k−Ŷ_k  (4)Coefficient Updating: W_k(n+1)=W_k(n)+μ_kY_kE_k*  (5) where the subscripts R and I represent the real part and imaginary part of a complex number respectively, the superscript asterisk (*) denotes the complex conjugate operation, Y_k is the DFT output, X_k is the training symbol of the k-th subchannel of the DMT system, W_k is the FEQ coefficient of the k-th subchannel, and μ_k is the updating step-size of the FEQ for the k-th subchannel. An embodiment of the DFT-based FEQ is illustrated in FIG. 2.

Prior arts have suggested replacing the DFT with discrete Hartley transform (DHT) so as to reduce the computing complexity from complex to real multiplication involved in DFT, as DHT itself is real-valued operation. With such a substitution, the DFT-based DMT system shown in FIG. 1 would become a DMT system illustrated in FIG. 3, where the modulation at the transmitting end is realized by the inverse DHT (IDHT) with a preceding complex-to-real transformation (C2RT), and the demodulation at the receiving end is realized by DHT with a succeeding real-to-complex transformation (R2CT). C2RT is required to transform the complex symbol X_k into the real symbol H_k; and R2CT is required to transform the real symbol {tilde over (H)}_k to the complex symbol Y_k. Please note that the DFT-based FEQ still has to be adopted in the DHT-based DMT system of FIG. 3, as basically the FEQ architecture remains the same.

The modulation and demodulation operations performed by the IDHT and DHT can be expressed mathematically as follows: $\begin{array}{cc}\mathrm{IDHT}\text{:}{x}_N=\frac{1}{N}\sum _{k=0}^{N-1}{H}_k·\mathrm{cas}\left(2\pi nk/N\right),n=0,1,\dots ,N-1& \left(6\right)\\ \mathrm{DHT}\text{:}{\tilde{H}}_k=\sum _{n=0}^{N-1}{y}_n·\mathrm{cas}\left(2\pi nk/N\right),k=0,1,\dots ,N-1& \left(7\right)\end{array}$ where cas(·)=cos(·)+sin(·), and N is the number of points for the IDHT/DHT and is the same as that for the IDFT/DFT. Similarly, the C2RT and R2CT can be expressed as follows:C2RT: H_k=X_k,R−X_k,I  (8)R2CT: Y_k=Y_k,R+jY_k,I={tilde over (H)}_k,E−j{tilde over (H)}_k,O  (9) where {tilde over (H)}_k,E and {tilde over (H)}_k,O are the even and odd parts of the {tilde over (H)}_k respectively, which can be obtained by: $\begin{array}{cc}{\tilde{H}}_{k,E}=\frac{{\tilde{H}}_k+{\tilde{H}}_{N-k}}{2}& \left(10\right)\\ {\tilde{H}}_{k,O}=\frac{{\tilde{H}}_k-{\tilde{H}}_{N-k}}{2}& \left(11\right)\end{array}$

SUMMARY OF THE INVENTION

A DMT system is provided herein. As described above, in practice, the IDFT and DFT of a conventional DMT system can be replaced by the IDHT with a preceding complex-to-real transformation (C2RT), and by the DHT with a succeeding real-to-complex transformation (R2CT) at the transmitting end and the receiving end respectively. With the present invention, the DFT and the DFT-based FEQ at the receiving end of a conventional DMT system is replaced by the DHT and a DHT-based FEQ, omitting the use of R2CT.

At the transmitting end, the DMT system of the present invention performs modulation either by IDFT or by the IDHT along with C2RT while, at the receiving end, the demodulation is just realized by the DHT. The DHT-based FEQ directly equalizes each of the 0-th to (N−1)-th, subchannels output from the DHT as R2CT is omitted, where N is the number of points of DHT. Finally, each of the 0-th to (N/2−1)-th subchannels of the DMT system is obtained by combining the k-th and (N-k)-th subchannels of the DHT-based FEQ for k=0, 1, . . . , (N/2−1).

The LMS algorithm can also be adopted for the DHT-based FEQ, which is very suitable for VLSI implementation. The LMS algorithm of the DHT-based FEQ also contains the filtering, error estimation and coefficient updating operations.

The foregoing and other objects, features, aspects and advantages of the present invention will become better understood from a careful reading of a detailed description provided herein below with appropriate reference to the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic diagram showing a conventional DFT-based DMT system for xDSL.

FIG. 2 is a schematic diagram showing an embodiment of the DFT-based FEQ of FIG. 1.

FIG. 3 is a schematic diagram showing a conventional DFT-based DMT system for xDSL realized by DHT.

FIG. 4 a is a schematic diagram showing a DMT system for xDSL according to a first embodiment of the present invention.

FIG. 4 b is a schematic diagram showing a DMT system for xDSL according to a second embodiment of the present invention.

FIG. 5 is a schematic diagram showing an embodiment of the DHT-based FEQ of FIGS. 4a and 4b.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

The following descriptions are exemplary embodiments only, and are not intended to limit the scope, applicability or configuration of the invention in any way. Rather, the following description provides a convenient illustration for implementing exemplary embodiments of the invention. Various changes to the described embodiments may be made in the function and arrangement of the elements described without departing from the scope of the invention as set forth in the appended claims.

FIGS. 4 a and 4b are schematic diagrams showing a DMT system for xDSL according to two embodiments of the present invention. As illustrated, at the transmitting end, the modulation can be realized either by IDFT alone or by IDHT with a preceding C2RT. In the following, for simplicity, the first embodiment of FIG. 4a will be used as an example. Please note that, even though the application of the present invention to xDSL is presented throughout the specification, the present invention can actually be applied to various communication systems, now known or later developed.

As illustrated, the DMT system of the present invention is similar to the prior arts in terms of architecture where, at the transmitting end, the system of the present invention performs modulation by the C2RT plus IDHT while, at the receiving end, the demodulation is now realized by the DHT only, omitting the R2CT. A major characteristic of the present invention is that a DHT-based FEQ is invented and employed to directly equalize each of the real-valued 0-th to (N−1)-th subchannels (i.e., {tilde over (H)}_k of eq. (7)) output from the DHT.

The mathematical model for the DHT-based FEQ has been derived by the present inventor based on the MMSE criteria as the cost function of the equalization's residual error, and is presented in “DHT-Based Frequency Domain Equalizer for DMT Systems” in Proceedings of European Signal Processing Conference (Proc. of EUSIPCO'05). Using the LMS algorithm, the mathematical model of the DHT-based FEQ is as follows, which also contains the filtering, error estimation, and coefficient updating operations: Filtering: $\begin{array}{cc}\left[\begin{array}{c}{\hat{Y}}_{k,R}\\ {\hat{Y}}_{k,I}\end{array}\right]=\left[\begin{array}{cc}{S}_k& C_k\\ -C_k& S_k\end{array}\right]·\left[\begin{array}{c}{\tilde{H}}_k\\ {\tilde{H}}_{N-k}\end{array}\right],k=0,1,\dots ,\frac{N}{2}-1{\hat{Y}}_k={\hat{Y}}_{k,R}+j{\hat{Y}}_{k,I}& \left(12\right)\end{array}$ Error estimation: $\begin{array}{cc}{E}_k=X_k-{\hat{Y}}_k,k=0,1,\cdots ,\frac{N}{2}-1& \left(13\right)\end{array}$ Coefficient Updating: $\begin{array}{cc}\left[\begin{array}{c}{S}_k\left(n+1\right)\\ C_k\left(n+1\right)\end{array}\right]=\left[\begin{array}{c}{S}_k\left(n\right)\\ C_k\left(n\right)\end{array}\right]+{\mu }_k·\left[\begin{array}{c}{\tilde{H}}_k{E}_{k,R}+{\tilde{H}}_{N-k}{E}_{k,I}\\ {\tilde{H}}_{N-k}{E}_{k,R}+{\tilde{H}}_k{E}_{k,I}\end{array}\right],k=0,1,\dots ,\frac{N}{2}-1& \left(14\right)\end{array}$ where the subscripts R and I represent the real part and imaginary part of a complex number respectively, {tilde over (H)}_k and {tilde over (H)}_N-k are k-th and (N-k)-th subchannels output from DHT, X_k is the training symbol of the k-th subchannel of the DMT system, Ŷ_k is the k-th subchannel output from the DHT-based FEQ, μ_k iS the updating step-size of the DHT-based FEQ. In an embodiment of the present invention, the coefficients S_k and C_k can be derived from W_k,R and W_k,I, which are the real and imaginary part of coefficients W_k from the DFT-based FEQ as follows: $S_k=\frac{W_{k,R}-W_{k,I}}{2};$$\mathrm{and}$$C_k=\frac{W_{k,R}+W_{k,I}}{2}.$

In an alternative embodiment, the coefficient updating operation can also be carried out as follows: $\left[\begin{array}{c}{S}_k\left(n+1\right)\\ C_k\left(n+1\right)\end{array}\right]=\left[\begin{array}{c}{S}_k\left(n\right)\\ C_k\left(n\right)\end{array}\right]+{\mu }_k·\left[\begin{array}{c}{\tilde{H}}_k·\mathrm{sgn}\left(E_{k,R}\right)+{\tilde{H}}_{N-k}·\mathrm{sgn}\left(E_{k,I}\right)\\ {\tilde{H}}_{N-k}·\mathrm{sgn}\left(E_{k,R}\right)+{\tilde{H}}_k·\mathrm{sgn}\left(E_{k,I}\right)\end{array}\right],k=0,1,\dots ,\frac{N}{2}-1$ where sgn(x) is the x's quantized value, which represents a +1 if the x≧0 and −1 if x<0.

An embodiment of the DHT-based FEQ is illustrated in FIG. 5. At last, each of the 0-th to $\left(\frac{N}{2}-1\right)-\mathrm{th}$ subchannels of the DMT system is obtained by combining the k-th and (N-k)-th subchannels of the DHT-based FEQ for k=0, 1, . . . , $\left(\frac{N}{2}-1\right).$

A comparison of the computing complexity for a single tone operation between the conventional approach (DHT+R2CT along with DFT-based FEQ) and the present invention (DHT+DHT-based FEQ) along the receive path is listed in Table 1:

TABLE 1
Comparison in Computing Complexity
	DHT + R2CT	DHT
Receive Path Computing	DFT-based	DHT-based
Complexity	FEQ	FEQ
R2CT	+	2	—
LMS-based	Filtering	×	4	4
FEQ		+	2	2
	Error Estimation	+	2	2
	Updating	×	4	4
		+	4	4

In Table 1, the R2CT is considered only requiring 2-addition for each tone since the ½-term in Eq. (10) and (11) can be ignored for VLSI realization. From Table 1, it can be seen that the computing complexity of the DHT-based FEQ is the same as the DFT-based FEQ. In addition, it could be concluded that the present invention requires less computing complexity as it omits the 2N-addition (2 for each tone) of the R2CT.

Simulation results of data rates on a number of test loops under the conventional approach (DHT+R2CT along with DFT-based FEQ) and the present invention (DHT+DHT-based FEQ) are listed in Table 2:

TABLE 2
Simulation Results
	Case I	Case II
	DHT + R2CT	DHT
	DFT-based FEQ	DHT-based FEQ
	Test Loop	REVERB(kbps)
	T1.601 Loop#7	3556	3556
	T1.601 Loop#9	4560	4560
	T1.601 Loop#13	4408	4408
	CSA Loop#4	9864	9864
	CSA Loop#6	9272	9272
	CSA Loop#7	9176	9176
	CSA Loop#8	8496	8496
	mid-CSA Loop	12664	12664

As illustrated, the DHT-based FEQ and the DFT-based FEQ have identical data rates. Also by performance analysis, it can be proved that the DHT-based FEQ and the DFT-based FEQ have identical subchannel signal-to-noise ratios (SNRs).

Although the present invention has been described with reference to the preferred embodiments, it will be understood that the invention is not limited to the details described thereof. Various substitutions and modifications have been suggested in the foregoing description, and others will occur to those of ordinary skill in the art. Therefore, all such substitutions and modifications are intended to be embraced within the scope of the invention as defined in the appended claims.

Claims

1. An N (N>1)-point DMT system, comprising a transmitting end, said transmitting end having an IDFT operator accepting complex-valued symbols Xk (k=0, 1, . . . , N−1) as input for modulation onto xn (n=0, 1, . . . , N−1) subchannels of said DMT system, said IDFT operator being characterized by the equation: xn=1N⁢∑k=0N-1⁢Xk·WN-kn,⁢n=0,1,⋯⁢ ,N-1where WN=ⅇ-j⁢2⁢πN;and a receiving end, said receiving end having a DHT operator for demodulation yn (n=0, 1, . . . , N−1) subchannels of said DMT system, and a DHT-based FEQ operator following said DHT operator for equalization, said DHT operator being characterized by the equation: H~k=∑n=0N-1⁢yn·cas⁡(2⁢π⁢ ⁢nk/N),⁢k=0,1,…⁢ ,N-1,andsaid DHT-based FEQ operator performing three operations: filtering, error estimation, and coefficient updating, said filtering operation being characterized by the equation: [Y^k,RY^k,1]=[SkCk-CkSk]·[H~kH~N-k],⁢k=0,1,…⁢ ,N2-1,⁢Y^k=Y^k,R+j⁢ ⁢Y^k,lsaid error estimation operation being characterized by the equation: Ek=Xk-Y^k,k=0,1,…⁢ ,N2-1,andsaid coefficient updating operation being characterized by the equation: [Sk⁡(n+1)Ck⁡(n+1)]=[Sk⁡(n)Ck⁡(n)]+μk·M⁡(H~k,H~N-k,Ek),⁢k=0,1,…⁢ ,N2-1,where M is a matrix operator involving Ek, {tilde over (H)}k and {tilde over (H)}N-k, {tilde over (H)}k and {tilde over (H)}N-k are the k-th and (N-k)-th subchannels output from said DHT operator, Xk is the training symbol of the k-th subchannel of said DMT system, Ŷk is the k-th subchannel output from said DHT-based FEQ operator, μk iS the updating step-size of said DHT-based FEQ operator, and Sk and Ck are coefficients.

2. The DMT system according to claim 1, wherein said IDFT operator comprises an C2RT operator accepting complex-valued symbols Xk (k=0, 1, . . . , N−1) as input, and an IDHT operator following said C2RT operator for modulation onto xn (n=0, 1, . . . , N−1) subchannels of said DMT system; said C2RT operator is characterized by the equation:

3. The DMT system according to claim 1, wherein said coefficients Sk and Ck are derived from the coefficient Wk of a DFT-based FEQ as follows:

4. The DMT system according to claim 1, wherein said matrix operator M is:

5. The DMT system according to claim 1, wherein said matrix operator M is:

6. A receiver of a N (N>1)-point DMT system whose transmitter performs an IDFT operation for modulation, said receiver comprising a DHT operator for demodulation yn (n=0, 1, . . . , N−1) subchannels of said DMT system, and a DHT-based FEQ operator following said DHT operator for equalization, said DHT operator being characterized by the equation:

7. The receiver according to claim 6, wherein said coefficients Sk and Ck are derived from the coefficient Wk of a DFT-based FEQ as follows:

8. The receiver according to claim 6, wherein said matrix operator M is:

9. The receiver according to claim 6, wherein said matrix operator M is:

10. A DHT-based FEQ of a receiver of a N (N>1)-point DMT system whose transmitter performs an IDFT operation for modulation, said receiver having a DHT operator for demodulation, said DHT-based FEQ following said DHT operator and comprising three operations: filtering, error estimation, and coefficient updating, said filtering operation being characterized by the equation:

11. The DHT-based FEQ according to claim 10, wherein said coefficients Sk and Ck are derived from the coefficient Wk of a DFT-based FEQ as follows:

12. The DHT-based FEQ according to claim 10, wherein said matrix operator M is:

13. The DHT-based FEQ according to claim 10, wherein said matrix operator M is: