METHOD AND DEVICE FOR TRANSMITTING OR RECEIVING AT LEAST ONE HIGH-FREQUENCY SIGNAL USING PARALLEL AND UNDERSAMPLED BASEBAND SIGNAL PROCESSING

Abstract

The method according to the invention and the device according to the invention for receiving at least one high-frequency signal (x(t)) using parallel and undersampled baseband signal processing generate a plurality of filtered signals (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)) through parallel filtering of the high-frequency signal (x(t)), wherein each individual filtering is performed in each case by means of a different filter frequency response. An associated digitized filtered signal (y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A)) is then generated in each case for each filtered signal (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)) through analog-to-digital conversion of the respective filtered signal (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)), wherein each analog-to-digital conversion is performed in each case by means of undersampling. Finally, the signal components (x.sub.1(n.Math.T.sub.A), . . . , x.sub.M(n.Math.T.sub.A); X.sub.1(k.Math.f), . . . , X.sub.M(k.Math.f)) of the high-frequency signal (x(t)) in the digital baseband are determined by means of equalization of baseband signal components (l.sub.1(n.Math.T), l.sub.2(n.Math.T.sub.A), . . . , l.sub.N(n.Math.T.sub.A); L.sub.1(k.Math.f), L.sub.2(k.Math.f), . . . , L.sub.N(k.Math.f)) of the associated digitized filtered signals (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)). A complementary method and device for transmitting at least one high-frequency signal (z(t)) using parallel and undersampled baseband signal processing, and also a system for transmitting at least one high-frequency signal using parallel and undersampled baseband signaling processing are also encompassed by the invention.

Claims

1. A method for receiving at least one high-frequency signal (x(t)) using parallel and undersampled baseband signal processing with the following method steps: generating a plurality of filtered signals (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)) through parallel filtering of the respective high-frequency signal (x(t)), whereby each individual filtering is performed by means of a different filter frequency response, generating a digitized filtered signal (y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A)) associated in each case with each filtered signal (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)) through analog-to-digital conversion of the respective filtered signal (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)), wherein each analog-to-digital conversion is performed in each case by means of undersampling, determining signal components (x.sub.1(n.Math.T.sub.A), . . . , x.sub.M(n.Math.T.sub.A); X.sub.1(k.Math.f), . . . , X.sub.M(k.Math.f)) of the respective high-frequency signal (x(t)) in the digital baseband by means of equalization of baseband signal components (l.sub.1(n.Math.T.sub.A), l.sub.2(n.Math.T.sub.A), . . . , i.sub.N(n.Math.T.sub.A); L.sub.1(k.Math.f), L.sub.2(k.Math.f), . . . , L.sub.N(k.Math.f)) of the associated digitized filtered signals (y.sub.1(t), y.sub.2(t) . . . , y.sub.N(t)).

2. The method as claimed in claim 1, wherein the analog-to-digital conversion of all filtered signals (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)) is performed with an identical sampling frequency $\left(\frac{1}{T_A}\right).$

3. The method as claimed in claim 2, wherein each analog-to-digital conversion of all filtered signals (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)) is performed coherently.

4. The method as claimed in claim 1, wherein each high-frequency signal (x(t)) is in each case a DMT signal or an OFDM signal.

5. The method as claimed in claim 1, wherein the equalization takes account of distortions in the filtering, in the analog-to-digital conversion, in a transmission of the respective high-frequency signal in a transmission channel and/or in a generation of the respective high-frequency signal.

6. The method as claimed in claim 1, wherein the baseband signal components of all digitized filtered signals (y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A)) are equalized in each case by means of equalization filters.

7. The method as claimed in claim 1, wherein the determination of baseband signal components associated in each case with each digitized filtered signal (y.sub.1(n.Math.T), y.sub.2(n.Math.T), . . . , y.sub.N(n.Math.T.sub.A)) is in each case a spectral transformation which determines a spectrum (Y.sub.1(k.Math.f), Y.sub.2(k.Math.f), . . . , Y.sub.N(k.Math.f)) of the respective digitized filtered signal (y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A)), and a further examination of spectral components of the respective spectrum (Y.sub.1(k.Math.f), Y.sub.2(k.Math.f), . . . , Y.sub.N(k.Math.f)) contained only in the baseband spectrum.

8. The method as claimed in claim 7, wherein the equalization of baseband signal components of the individual digitized filtered signals (y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A)) is performed by means of multiplication of the baseband signal components associated in each case with the respective digitized filtered signal (y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A)) by associated spectral equalization coefficients ({circumflex over (F)}.sub.1.sup.1(k.Math.f), {circumflex over (F)}.sub.2.sup.2(k.Math.f), . . . , {circumflex over (F)}.sub.1.sup.N(k.Math.f) . . . , {circumflex over (F)}.sub.1.sup.M(k.Math.f), {circumflex over (F)}.sub.M.sup.2(k.Math.f) . . . , {circumflex over (F)}.sub.M.sup.N(k.Math.f)).

9. The method as claimed in claim 1, wherein the filter frequency responses of the parallel filtering extend each case over a plurality of Nyquist zones.

10. The method as claimed in claim 1, wherein the filter frequency responses of the parallel filtering completely or partially overlap one another.

11. A method for transmitting at least one high-frequency signal (z(t)) using parallel and undersampled baseband signal processing with the following method steps for each individual high-frequency signal (z(t)): generating analog signals (v.sub.1(t), v.sub.2(t), . . . , v.sub.N(t)) by means of digital-to-analog conversion of a respective associated digital signal (u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A)), wherein each digital signal (u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A)) is present in the baseband, generating signals (z.sub.1(t), z.sub.2(t), . . . , z.sub.N(t)) by means of filtering of each associated analog signal (v.sub.1(t), v.sub.2(t), . . . , v.sub.N(t)), wherein each individual filtering is performed in each case by means of a different frequency response, and generating the high-frequency signal y(t)) by means of summation of all filter signals (z.sub.1(t), z.sub.2(t), . . . , z.sub.N(t)).

12. The method as claimed in claim 11, wherein each digital signal (u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A)) is held in the digital-to-analog conversion in each case only over a time period (T.sub.A*) reduced in comparison with the sampling period (T.sub.A) of the digital signal (u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A)).

13. The method as claimed in claim 11, wherein the frequency responses of the parallel filtering completely or partially overlap one another.

14. The method as claimed in claim 11, wherein the frequency responses of the parallel filtering extend over a plurality of Nyquist zones.

15. The method as claimed in claim 11, wherein each high-frequency signal (y(t)) is in each case a DMT signal or an OFDM signal.

16. A device for receiving at least one high-frequency signal (x(t)) using parallel and undersampled baseband signal processing in each case with a plurality of parallel-connected filters (1.sub.1, 1.sub.2, . . . , 1.sub.N) in each case for generating a filtered signal (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)) at whose inputs the respective high-frequency signal (x(t)) is in each case present and whose filter frequency responses differ in each case from one another, an analog-to-digital converter (3.sub.1, 3.sub.2, . . . , 3.sub.N) connected in each case downstream of each filter (1.sub.1, 1.sub.2, . . . , 1.sub.N) in each case for generating a digitized filtered signal (y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A)), wherein an undersampling is performed in each case in each analog-to-digital converter (3.sub.1, 3.sub.2, . . . , 3.sub.N), and an equalization and decoupling unit (8; 8; 8; 8) connected in each case downstream of the analog-to-digital converters (3.sub.1, 3.sub.2, . . . , 3.sub.N).

17. The device as claimed in claim 16, wherein the filter frequency responses of all (1.sub.1, 1.sub.2, . . . , 1.sub.N) together cover the entire frequency spectrum of the high-frequency signal (x(t)).

18. The device as claimed in claim 16, wherein each analog-to-digital converter (3.sub.1, 3.sub.2, . . . , 3.sub.N) is connected to an identical clock source (6).

19. The device as claimed in claim 16, wherein the equalization and decoupling unit (8; 8; 8; 8) contains at least a number (N) of equalization filters (10.sub.11, 10.sub.12, . . . , 10.sub.1N, . . . , 10.sub.21, 10.sub.22, . . . , 10.sub.2N, . . . , 10.sub.M1, 10.sub.M2, . . . , 10.sub.MN) corresponding to the number (N) of filtered signals (y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t)).

20. The device as claimed in claim 16, wherein a spectral transformer (13.sub.1, 13.sub.2, . . . , 13.sub.N) for generating a spectrum (Y.sub.1(k.Math.f), Y.sub.2(k.Math.f), . . . , Y.sub.N(k.Math.f)) associated in each case with the respective digitized filtered signal (y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A)) is connected in each case upstream between each analog-to-digital converter (3.sub.1, 3.sub.2, . . . , 3.sub.N) and the equalization and decoupling unit (8; 8).

21. The device as claimed in claim 20, wherein the equalization and decoupling unit (8; 8; 8; 8) contains at least a number (N) of multiplier elements (14.sub.11, 14.sub.12, . . . , 14.sub.1N, . . . , 14.sub.21, 14.sub.22, . . . , 14.sub.2N, . . . , 14.sub.M1, 14.sub.M2, . . . , 14.sub.MN) corresponding to the number (N) of filtered signals (y.sub.1(t), y.sub.2(t), . . . , Y.sub.N(t)).

22. A device for transmitting at least one high-frequency signal (z(t)) using parallel and undersampled baseband signal processing in each case with a number of digital-to-analog converters (19.sub.1, 19.sub.2, . . . , 19.sub.N) in each case for generating an analog signal (v.sub.1(t), v.sub.2(t), . . . , v.sub.N(t)) from an associated digital baseband signal (u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A)), wherein the number (N) of digital-to-analog converters (19.sub.1, 19.sub.2, . . . , 19.sub.N) corresponds to the number of digital baseband signals (u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T), . . . , u.sub.N(n.Math.T.sub.A)) to be transmitted in each case in the respective high-frequency signal (z(t)), a filter (21.sub.1, 21.sub.2, . . . , 21.sub.N) connected in each case downstream of each digital-to-analog converter (19.sub.1, 19.sub.2, . . . , 19.sub.N) in each case for generating an associated filtered signal (z.sub.1(t), z.sub.2(t), . . . , z.sub.N(t)), wherein the filter frequency response of the filters (21.sub.1, 21.sub.2, . . . , 21.sub.N) extends in total over all relevant Nyquist zones, and a summing element (22) connected in each case downstream of the filters (21.sub.1, 21.sub.2, . . . , 21.sub.N) to generate the respective high-frequency signal z(t)).

23. The device as claimed in claim 22, wherein each digital-to-analog converter (19.sub.1, 19.sub.2, . . . , 19.sub.N) is connected to an identical clock source (20).

24. (canceled)

Description

[0069] Before the individual embodiments of the method according to the invention and the device according to the invention for transmitting or receiving at least one high-frequency signal using parallel and undersampled baseband signal processing are explained in detail with reference to the figures of the drawing, the mathematical foundations necessary for the understanding of the invention are presented in advance: A high-frequency signal x(t) is convoluted in an analog filter i with the latter's pulse response g.sub.Fi(t). The filter i is preferably a bandpass filter. Alternatively, however, a low-pass filter, a high-pass filter, an all-pass filter or a filter with any given transmission function can also be used.

[0070] Following a multiplication by the filter transmission function G.sub.Fi(f), the associated frequency spectrum X(f) of the high-frequency signal x(t) provides the frequency spectrum Y.sub.i(f) of the filtered signal present at the output of the filter according to equation (1).

Y.sub.i(f)=G.sub.Fi(f).Math.X(f)(1)

[0071] A discrete frequency spectrum is present following the analog-to-digital conversion and a subsequent discrete Fourier transform. The frequency spacing f of the individual spectral components of a discrete frequency spectrum of this type corresponds to the frequency spacing between the individual carrier signals of the multicarrier method (DMT/OFDM) used in the high-frequency signal on which the individual modulated data are in each case present.

[0072] Due to the sampling, the spectrum of the filtered signal is repeated with a periodicity in the amount of the sampling frequency f.sub.A. Ignoring the distortion of the filtered signal in the analog-to-digital conversion, the spectrum Y.sub.i(k.Math.f) of a spectral component of the high-frequency signal is obtained at the frequency k.Math.f at the output of the analog-to-digital converter connected downstream of the filter i according to equation (2). The undersampling takes place in this step. A prerequisite here is that this spectral component is positioned within the pass bandwidth of the filter i. The parameters m and k represent integer run parameters.

Y.sub.i(k.Math.f)=.sub.m=.sup.+G.sub.Fi(k.Math.fm.Math.f.sub.A).Math.X(k.Math.fm.Math.f.sub.A)(2)

[0073] For a real signal y.sub.i(n.Math.T.sub.A), the associated spectrum Y.sub.i(k.Math.f) according to equation (3) has complex-conjugate symmetry.

Y.sub.i(k.Math.f)=Y.sub.i*(k.Math.f)(3)

[0074] As shown in FIG. 1, due to the periodicity of the spectrum in the amount of the sampling frequency f.sub.A and due to the complex-conjugate symmetry of the spectrum, the ranges of the spectrum represented by dotted lines and continuing periodically which have a spectral width in each case of a half sampling frequency f.sub.A in each case have an identical spectrum. Equivalently, the ranges of the spectrum not represented by dotted lines and continuing periodically which similarly have a spectral width in each case of a half sampling frequency f.sub.A in each case have an identical spectrum.

[0075] Each of these ranges of the spectrum with or without dotted lines in each case represents a Nyquist zone. The order of the individual Nyquist zone increases starting from the axis of symmetry at the spectral frequency of zero.

[0076] Each Nyquist zone of an odd order in the positive spectral range, i.e. the Nyquist zones I, III and V, etc., in FIG. 1, and each Nyquist zone of an even order in the negative spectral range, i.e. the Nyquist zones II*, IV*, etc., in FIG. 1, since the spectra of these Nyquist zones are identical, contains not only the spectral components of its own associated analog spectrum, but also the spectral components of the analog spectra of the respective other assigned Nyquist zones. Equivalently, each Nyquist zone of an even order in the positive spectral range, i.e. the Nyquist zones II, IV, etc., in FIG. 1, and each Nyquist zone of an odd order in the negative spectral range, i.e. the Nyquist zones I*, III*, V*, etc., in FIG. 1, since the spectra of these Nyquist zones are identical, contains not only the spectral components of its own associated analog spectrum, but also the spectral components of the analog spectra of the respective other assigned Nyquist zones.

[0077] It is thus possible according to the invention, through a spectral evaluation in the spectral range between the negative half sampling frequency and the positive half sampling frequency, i.e. in the baseband of the sampled signal, to determine the spectral components of the high-frequency signal in the Nyquist zones over which the filter frequency responses of all filters extend. If the filter covers a plurality of Nyquist zones, the baseband contains a linear combination of all contained Nyquist zones. In the case of real signals, only the positive or the negative half of the baseband spectrum has to be considered, since both are redundant in relation to one another.

[0078] The invention can also be applied to the I and Q signals at the output of an IQ mixer. The I and Q signal paths of the IQ mixer are in each case filtered in parallel and are undersampled by means of analog-to-digital conversion.Math.The sampled values of the ADCs in the I signal path are interpreted as the real component, the sampling values of the ADCs in the Q signal path as the imaginary component. All further steps in a downstream equalization and decoupling unit remain the same.

[0079] The spectral component L.sub.i(k.Math.f) of the digitized filtered signal in the digital baseband at the spectral frequency k.Math.f in which the spectral components of the digitized filtered signal at the spectral frequency k.Math.fin all Nyquist zones of the filter frequency response of the respective filter are contained is obtained, taking account of equation (3), according to equation (4).

L.sub.i(k.Math.f)=.sub.m=.sup.+G.sub.Fi(k.Math.f+m.Math.f.sub.A).Math.X(k.Math.f+m.Math.f.sub.A)(4)

[0080] Only a limited number M of Nyquist zones are typically occupied in the receive signal. The infinite sum in equation (4) therefore becomes a finite sum according to equation (5).

[00001]$\begin{array}{cc}{L}_i\left(k.Math..Math..Math.f\right)=\underset{m=-\frac{M}{2}}{\overset{+\frac{M}{2}-1}{.Math.}}.Math.G_{\mathrm{Fi}}\left(k.Math..Math..Math.f+m.Math.f_A\right).Math.X\left(k.Math..Math..Math.f+m.Math.f_A\right)& \left(5\right)\end{array}$

[0081] Equivalently, equation (5) can also be represented vectorially according to equation (6).

[00002]$\begin{array}{cc}{L}_i\left(k.Math..Math..Math.f\right)=\left[G_{F.Math.i}.Math.\left(k.Math.f-\frac{M}{2}.Math.f_A\right).Math..Math.\mathrm{.Math.}.Math..Math.G_{\mathrm{Fi}}\left(k.Math..Math..Math.f+\left(\frac{M}{2}-1\right).Math.f_A\right)\right].Math.{\left[X\left(k.Math..Math..Math.f-\frac{M}{2}.Math.f_A\right).Math..Math.\mathrm{.Math.}.Math..Math.X\left(k.Math..Math..Math.f+\left(\frac{M}{2}-1\right).Math.f_A\right)\right]}^T& \left(6\right)\end{array}$

[0082] The individual coefficients G.sub.Fi of the first vector containing the filter transmission coefficient of the filter i at the spectral frequency k.Math.f in the respective Nyquist zone. If the distortion due to the transmitter filters, due to the transmission channel and/or the associated analog-to-digital-converter is to be equalized in the same step, the coefficient G.sub.Fi in equation (6) contains both the transmission function of the filter i and the filter frequency responses of the transmitter filters, the channel transmission function and/or the transmission function of the associated analog-to-digital converter.

[0083] Equivalently, a vector equation according to equation (6) can be constructed for the remaining N1 filters.

[0084] The combination of the vector equations of all used filters results in the matrix equation (7).

L(k.Math.f)={tilde over (F)}(k.Math.f).Math.X(k.Math.f)(7)

[0085] Both the undersampling and the filtering by the N filters are contained in equation (7).

[0086] where

[00003]$.Math.\stackrel{\_}{L}\left(k.Math..Math..Math.f\right)=\left[\begin{array}{c}{L}_1\left(k.Math..Math..Math.f\right)\\ \mathrm{.Math.}\\ L_N\left(k.Math..Math..Math.f\right)\end{array}\right]$$\stackrel{}{F}\left(k,.Math..Math.f\right)=\left[\begin{array}{ccc}{G}_{F.Math..Math.1}\left(k.Math..Math..Math.f-\frac{M}{2}.Math.f_A\right)& \mathrm{.Math.}& G_{\mathrm{Fi}}\left(k.Math..Math..Math.f+\left(\frac{M}{2}-1\right).Math.f_A\right)\\ \mathrm{.Math.}& \mathrm{.Math.}& \mathrm{.Math.}\\ G_{\mathrm{FN}}\left(k.Math..Math..Math.f-\frac{M}{2}.Math.f_A\right)& \mathrm{.Math.}& G_{\mathrm{FN}}\left(k.Math..Math..Math.f+\left(\frac{M}{2}-1\right).Math.f_A\right)\end{array}\right]$$.Math.\stackrel{\_}{X}\left(k.Math..Math..Math.f\right)={\left[X\left(k.Math..Math..Math.f-\frac{M}{2}.Math.f_A\right).Math..Math.\mathrm{.Math.}.Math..Math.X\left(k.Math..Math..Math.f+\left(\frac{M}{2}-1\right).Math.f_A\right)\right]}^T$

[0087] If a plurality of transmission channels (MIMO) are used, the vector X increases according to the number of transmission channels that are used.

[0088] In a first variant of an equalization in the case of a low-noise useful signal, the frequency response of the individual digital equalization filter is defined as the inverse of the filter frequency response of the respective preceding filter (zero-forcing equalizer). The estimation vector {circumflex over (X)}(k.Math.f) of the spectral component of the high-frequency signal at the spectral frequency k.Math.f within the individual Nyquist zones is obtained here by means of matrix inversion of the matrix {tilde over (F)}(k.Math.f) and subsequent multiplication of the inverse matrix {tilde over (F)}.sup.1(k.Math.f) by the vector L(k.Math.f) which contains the Fourier transforms of the output signals of the individual analog-to-digital converters at the spectral frequency k.Math.f, according to the matrix equation (8).

[00004]$\begin{array}{cc}\stackrel{}{\hat{X}}\left(k.Math..Math..Math.f\right)={\tilde{F}}^{-1}\left(k.Math..Math..Math.f\right).Math.\stackrel{}{L}(.Math.k.Math..Math..Math.f)=.Math..Math.[.Math.\begin{array}{cccc}{\hat{F}}_1^1\left(k.Math..Math..Math.f\right)& {\hat{F}}_1^2\left(k.Math..Math..Math.f\right)& \mathrm{.Math.}& {\hat{F}}_1^N\left(k.Math..Math..Math.f\right)\\ \mathrm{.Math.}& \mathrm{.Math.}& \mathrm{.Math.}& \mathrm{.Math.}\\ {\hat{F}}_M^1\left(k.Math..Math..Math.f\right)& {\hat{F}}_M^2\left(k.Math..Math..Math.f\right)& \mathrm{.Math.}& {\hat{F}}_M^N\left(k.Math..Math..Math.f\right)\end{array}].Math..Math..Math.[.Math.L_1(.Math.k.Math..Math..Math.f).Math..Math.L_2\left(k.Math..Math..Math.f\right).Math..Math.\mathrm{.Math.}.Math..Math.L_N\left(k.Math..Math..Math.f\right)].Math.T& \left(8\right)\end{array}$

[0089] In the Nyquist zones represented by shading in FIG. 1, the spectral frequency k.Math.f is in each case the spectral frequency which has a spacing in the amount of k.Math.f from the edge of the respective Nyquist zone shown on the left side in FIG. 1, and, in the Nyquist zones shown in each case without shading in FIG. 1, in each case the spectral frequency which has a spacing in the amount of k.Math.f from the edge of the respective Nyquist zone shown on the right in FIG. 1.

[0090] In this way, it is possible to define consecutively the spectral components of the high-frequency signal in the spectral range in each case on the individual spectral frequencies of the carriers of the high-frequency signal.

[0091] In a second equalization variant for a useful signal with a higher noise component, the individual equalization coefficients are defined in such a way that the noise component is minimized in the equalization. A Minimum Mean Square Error (MMSE) equalization, for example, is carried out for this purpose. According to the prior art, the approach used for the equalization can also be any other approach which appears advantageous for achieving the object. MMSE and ZF represent two known and frequently used equalizer approaches.

[0092] Along with a definition of the spectral components of the high-frequency signal on the individual carrier frequencies in the spectral range, a definition in the time domain is essentially also possible.

[0093] A convolution of the pulse responses in each case associated with the individual filters with the high-frequency signal is to be performed instead of a multiplication of the filter transmission functions in each case associated with the individual filters by the Fourier transform for the mathematical derivation of a technical solution in the time domain. A mathematical derivation is foregone in this context.

[0094] On the transmitter side, the information contents of a specific number N of digital signals u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . u.sub.i(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A) are transmitted in a high-frequency signal z(t) which is to be transmitted. Corresponding analog signals v.sub.1(t), v.sub.2(t), . . . , v.sub.i(t), . . . , v.sub.N(t) are generated in each case via an analog-to-digital conversion.

[0095] Each digital sampling value u.sub.i(n.Math.T.sub.A) generates a pulse having the form rect(t/T.sub.A). The digital-to-analog converter thus has a transmission function having the form sinc(f/f.sub.A) in which spectral components above the half sampling frequency are substantially attenuated. Due to the shortening of the hold time to T*<T.sub.A of the digital-to-analog converter, the zero points of the sinc-shaped transmission function can be shifted to higher frequencies so that the spectral components of the individual analog signals v.sub.1(t), v.sub.2(t), . . . , v.sub.i(t), . . . , v.sub.N(t) remain undistorted in terms of amplitude over a further frequency range. A shortening of the hold time is achieved according to the prior art, for example, by oversampling.

[0096] In the individual filters in each case downstream of the respective digital-to-analog converters, a filtered signal z.sub.1(t), z.sub.2(t), . . . , z.sub.N(t) is convoluted by means of convolution of the associated analog signal v.sub.1(t), v.sub.2(t), . . . , v.sub.N(t) with a pulse response g.sub.BP1(t), g.sub.BP2(t), . . . , g.sub.BPN(t) of the associated filter. The filter may be a bandpass filter, high-pass filter, low-pass filter or all-pass filter or a filter with any given filter transmission function. The passband of a filter can in each case completely or partially cover one or more Nyquist zones. It should be noted that the filter transmission functions of the individual filters must differ from one another and are intended in total to cover the entire spectral range of the high-frequency signal to be transmitted.

[0097] In one possible embodiment, the filter transmission function G.sub.BP(f) of a bandpass filter i extends over the frequency range of a specific Nyquist zone, for example over the frequency range of the i-th Nyquist zone.

[0098] Finally, the individual filtered signals z.sub.1(t), z.sub.2(t), . . . , z.sub.N(t) are added to the high-frequency signal z(t).

[0099] The two subvariants of the first embodiment of the device according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signal processing are described in detail below on the basis of the described mathematical foundations with reference to the block diagrams in FIGS. 2A and 2B in combination with the associated method according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signaling process with reference to the flow diagram in FIG. 4:

[0100] In the first method step S10, at least one high-frequency signal x(t) is fed in each case to a specific number N of parallel-connected filters 1.sub.1, 1.sub.2, . . . 1.sub.N. The filters 1.sub.1, 1.sub.2, . . . 1.sub.N are implemented in each case in analog form and in each case have a different filter frequency response. The filters here are preferably bandpass filters. Alternatively, however, low-pass filters, high-pass filters or all-pass filters or filters with any given filter transmission function can also be used.

[0101] FIG. 6 shows examples of filter frequency responses 2.sub.1, 2.sub.2, 2.sub.3 and 2.sub.4 of four parallel-connected bandpass filters for a given frequency response 2 of the high-frequency signal. It is evident here that the passband of some filters extends over a plurality of Nyquist zones. In addition, the stop band can have a significant attenuation. This gives rise to further superimpositions in the baseband.

[0102] The filter frequency responses of the individual filters 1.sub.1, 1.sub.2, . . . , 1.sub.N in combination cover the spectral ranges 2 in FIG. 6 of the high-frequency signal which are to be taken into account and examined in the further signal processing. If the entire spectral range of the high-frequency signal is relevant to the further signal processing, the filter frequency responses of the individual filters 1.sub.1, 1.sub.2, . . . 1.sub.N cover the entire bandwidth of the high-frequency signal. If only specific frequency bands of the high-frequency signal are examined in the further signal processing, the combination of the filter frequency responses of the individual filters 1.sub.1, 1.sub.2, . . . 1.sub.N is restricted to the spectral ranges of these frequency bands.

[0103] The filter frequency responses of the individual filters 1.sub.1, 1.sub.2, . . . l.sub.N can overlap one another completely, partially or not at all. It is crucial here, in the case of complete or partial overlap, that they in each case have a different filter frequency response in the spectral overlap range.

[0104] In the following method step S20, the filtered signals y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t) are converted at the output of the individual filters 1.sub.1, 1.sub.2, . . . , 1.sub.N in each case in a downstream analog-to-digital converter 3.sub.1, 3.sub.2, . . . , 3.sub.N into a corresponding digitized filtered signal y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A). The analog-to-digital conversion is performed here in undersampling. In order to implement the undersampling, the individual analog-to-digital converter 3.sub.1, 3.sub.2, . . . , 3.sub.N is preferably implemented by way of a serial connection of a sample and hold (SH) element 4.sub.1, 4.sub.2, . . . , 4.sub.N and a downstream quantizer 5.sub.1, 5.sub.2, . . . , 5.sub.N.

[0105] In the individual sample and hold elements 4.sub.1, 4.sub.2, . . . , 4.sub.N, in each case a sampling of the analog-filtered signal y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t) and a holding of the respective sampling value of the analog-filtered signal y.sub.1(t), y.sub.2(t), . . . , y.sub.N(t) are performed in each case over the same sampling period T.sub.A.

[0106] For the sampling, each sample and hold element 4.sub.1, 4.sub.2, . . . , 4.sub.N receives a clock having the same sampling period T.sub.A from a clock source 6, for example a clock generator. The clock fed in each case to each sample and hold element 4.sub.1, 4.sub.2, . . . , 4.sub.N is (phase-) coherent. The clock fed in each case to the individual sample and hold elements 4.sub.1, 4.sub.2, . . . , 4.sub.N is phase-coherent if the phase of the individual clocks changes in each case over time in an identical manner and in each case said phases differ from one another only in a time-invariant phase difference. The clock source 6 is preferably to be implemented with minimal jitter in order to achieve the highest possible constancy in the sampling frequency

[00005]$\frac{1}{T_A},$

since the phase error caused by jitter increases with the order of the Nyquist bands.

[0107] An amplitude quantization of the sampled filtered signal is performed in each case in the quantizer 5.sub.1, 5.sub.2, . . . , 5.sub.N.

[0108] Only the baseband signal components l.sub.1(n.Math.T.sub.A), l.sub.2(n.Math.T.sub.A), . . . , l.sub.N(n.Math.T.sub.A) of the digitized filtered signals

[0109] y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A) are taken into account below. These baseband signal components l.sub.1(n.Math.T.sub.A), l.sub.2(n.Math.T.sub.A), . . . , l.sub.N(n.Math.T.sub.A) of the digitized filtered signals y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A) contain, in each case superimposed, all spectral components of the high-frequency signal x(t) which lie within the passband of the filter frequency response of the respective filter 1.sub.1, 1.sub.2, . . . l.sub.N.

[0110] In order to separate the digitized spectral components of the high-frequency signal x(t) located in each case in the individual Nyquist zones from the individual baseband signal components l.sub.1(n.Math.T.sub.A), l.sub.2(n.Math.T.sub.A), . . . , 1.sub.N(n.Math.T.sub.A) of the digitized filtered signals y.sub.1(n.Math.T), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A), the baseband signal components l.sub.1(n.Math.T.sub.A), l.sub.2(n.Math.T.sub.A), . . . , l.sub.N(n.Math.T.sub.A) of the individual digitized filtered signals y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A) are fed in the following method step S30 to an equalization and decoupling in an equalization and decoupling unit 8.

[0111] This equalization and decoupling unit 8 contains a number M of equalization and decoupling channels 9.sub.1, 9.sub.2, . . . , 9.sub.M, which corresponds to the number M of Nyquist zones contained in the bandwidth of the digitized high-frequency signal x(t).

[0112] Each individual equalization and decoupling channel 9.sub.1, 9.sub.2, . . . , 9.sub.M in turn contains a number of parallel-connected equalization filters 10.sub.11, 10.sub.12, . . . , 10.sub.1N or 10.sub.21, 10.sub.22, . . . , 10.sub.2N or 10.sub.M1, 10.sub.M2, . . . , 10.sub.MN corresponding to the number N of parallel-connected analog-to-digital converters 3.sub.1, 3.sub.2, . . . , 3.sub.N. The inputs of the individual equalization filters of a respective equalization and decoupling channel are connected in each case to the output of an analog-to-digital-converter 3.sub.1, 3.sub.2, . . . , 3.sub.N.

[0113] Each individual equalization filter of an equalization and decoupling channel in each case equalizes the supplied digitized filtered signal in such a way that, following a summation of all signals at the outputs of the equalization filters associated in each case with an equalization and decoupling channel, a signal is produced which contains only the signal components of the high-frequency signal x(t) within a Nyquist zone associated with the equalization and decoupling channel. Here, the respective equalization filter makes a contribution not only to the decoupling of the signal components contained in each case in the individual Nyquist zones, but also to the equalization of the distortion caused by the filter frequency response of the respective upstream filter.

[0114] In addition, the respective equalization filter can furthermore preferably also perform an equalization of the supplied signal in terms of the distortion caused in each case in the individual transmitter filters, the distortion caused in the transmission channel and/or the distortion caused in each case in the respective upstream analog-to-digital converter 3.sub.1, 3.sub.2, . . . , 3.sub.N.

[0115] The individual equalization filters 10.sub.11, 10.sub.12, . . . , 10.sub.1N, 10.sub.21, 10.sub.22, . . . , 10.sub.2N, . . . , 10.sub.M1, 10.sub.M2, . . . , 10.sub.MN are implemented in each case as digital filters, preferably as digital filters with a finite pulse length (FIR filters). The individual equalization filters are preferably structured and parameterized in an initialization phase of the device according to the invention. By supplying the device according to the invention with specific test signals and by measuring the associated response signals of the device according to the invention, the structure and the associated parameters of the equalization filter implemented as an adaptive digital filter are defined via optimization methods according to the prior art. Alternatively, the structure and the associated parameters of the individual equalization filters can also be determined in a simulation-based manner. In rare exceptional cases, a deterministic definition of structures and parameters of the individual equalization filters is also possible.

[0116] In each individual equalization and decoupling channel 9.sub.1, 9.sub.2, . . . , 9.sub.M, the outputs of the associated equalization filters are connected to a common summing element 11.sub.1, 11.sub.2, . . . , 11.sub.M. The signal components x.sub.1(n.Math.T.sub.A), . . . , x.sub.M(n.Math.T.sub.A) of the high-frequency signal x(t) which are located in a specific Nyquist zone of the digitized high-frequency signal processed by the respective equalization and decoupling channel and therefore in a spectral range of the high-frequency signal associated with the respective Nyquist zone are present in each case at the output of the respective summing element 11.sub.1, 11.sub.2, . . . , 11.sub.M.

[0117] The high-frequency signal x(t) is typically implemented as a multicarrier signal. A DMT(Discrete Multitone Transmission) signal is preferably used as the multicarrier signal. Alternatively, the high-frequency signal x(t) can also be implemented, for example, as an OFDM (Orthogonal Frequency Division Multiplexing) signal. Depending on the used sampling frequency

[00006]$\frac{1}{T_A}$

of the undersampling, either the frequency band of one carrier or the frequency bands of a plurality of carriers of the OFDM signal can be placed in each case in a Nyquist zone. In order to guarantee the orthogonality of the individual frequency carriers, the clocks which are fed to the individual analog-to-digital converters 3.sub.1, 3.sub.2, . . . , 3.sub.N not only have an identical frequency

[00007]$\frac{1}{T_A}$

of the clock, but are also to be designed as (phase-) coherent in relation to one another.

[0118] Along with this first subvariant of the first embodiment of the device according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signal processing according to FIG. 2A, a second subvariant also exists as a special form according to FIG. 2B:

[0119] In the case of the second subvariant of the first embodiment of the device according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signal processing, the filter frequency responses of the individual filters 1.sub.1, 1.sub.2, . . . , 1.sub.N are parameterized in such a way that their respective passband covers only the spectral range of a single Nyquist zone of the digitized high-frequency signal x(t).

[0120] The equalization and decoupling unit is simplified in this special case. The equalization and decoupling unit 8 of this second subvariant in each case contains only one single equalization filter 10.sub.11, 10.sub.22, . . . , 10.sub.MM in its individual equalization and decoupling channels 9.sub.1, 9.sub.2, . . . , 9.sub.M. Each of these equalization filters 10.sub.11, 10.sub.22, . . . , 10.sub.MM in each case equalizes the distortion caused in the preceding filter 1.sub.1, 1.sub.2, . . . , 1.sub.N. In addition, as already explained above for the first subvariant, the individual equalization filter 10.sub.11, 10.sub.22, . . . , 10.sub.MM can also equalize the distortion caused in each case in the individual transmitter filters, the distortion caused in the transmission channel and/or the distortion caused in each case in the preceding analog-to-digital converter 3.sub.1, 3.sub.2, . . . , 3.sub.N. A summing element is not required in the individual equalization and decoupling channels 9.sub.1, 9.sub.2, . . . , 9.sub.M.

[0121] The remaining functional units of the second subvariant of the first embodiment of the device according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signal processing correspond to the first subvariant of the first embodiment and the description thereof is not therefore repeated at this juncture. With regard to the mode of operation of these functional units, reference is made to the associated description of the first subvariant of the first embodiment. The same applies to the second subvariant of the first embodiment of the method according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signal processing.

[0122] The two subvariants of the second embodiment of the device according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signal processing are explained in detail below with reference to the block diagrams in FIGS. 3A and 3B in combination with the associated method according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signal processing with reference to the flow diagram in FIG. 5: The first two method steps S100 and S110 of the second embodiment of the method according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signal processing correspond to the first two method steps S10 and S20 of the first embodiment of the method according to the invention. The same applies to the associated functional units of the second embodiment of the device according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signal processing. The description is not therefore repeated at this juncture and reference is made to the associated descriptions of the first embodiment.

[0123] If DMT or OFDM is used as the transmission method, an optional filter, not shown in FIG. 3A, for shortening the channel pulse response and a unit for removing the guard interval are connected downstream of the respective analog-to-digital converter 3.sub.1, 3.sub.2, . . . , 3.sub.N in step S210. Finally, a Spectral Transformer (ST) 13.sub.1, 13.sub.2, . . . , 13.sub.N, transforms the digitized filtered signal y.sub.1(n.Math.T.sub.A), y.sub.2(n.Math.T.sub.A), . . . , y.sub.N(n.Math.T.sub.A) into the respective associated discrete spectrum Y.sub.1(k.Math.f), Y.sub.2(k.Math.f), . . . , Y.sub.N(k.Math.f). This step is preferably carried out by means of a Fourier transform, by means of a Fast Fourier Transformer (FFT) or by means of a discrete Fourier Transformer (DFT).

[0124] In the following and concluding method step S130, only the baseband spectral components L.sub.1(k.Math.f), L.sub.2(k.Math.f), . . . , L.sub.N(k.Math.f) of the digitized spectra Y.sub.1(k.Math.f), Y.sub.2(k.Math.f), . . . , Y.sub.N(k.Math.f) are taken into account in a downstream equalization and decoupling unit 8.

[0125] Here, on the one hand, the spectral components contained in each case in the baseband spectral components L.sub.1(k.Math.f), L.sub.2(k.Math.f), . . . , L.sub.N(k.Math.f) of the individual digitized spectra Y.sub.1(k.Math.f), Y.sub.2(k.Math.f), . . . , Y.sub.N(k.Math.f) and associated in each case with the individual Nyquist zones of the digitized high-frequency signal are decoupled and combined with the spectral components of the digitized high-frequency signal which are associated in each case with one of the Nyquist zones of the digitized high-frequency signal. On the other hand, the distortion due to the filter frequency responses of the individual filters 1.sub.1, 1.sub.2, . . . , 1.sub.N which is present in the individual baseband spectral components L.sub.1(k.Math.f), L.sub.2(k.Math.f), . . . , L.sub.N(k.Math.f) of the individual digitized spectra Y.sub.1(k.Math.f), Y.sub.2(k.Math.f), . . . , Y.sub.N(k.Math.f) is equalized.

[0126] In addition, similar to the first embodiment, an equalization of the distortion caused in each case in the individual transmitter filters, the distortion caused by the transmission channel and/or the distortions caused in each case by the transmission behavior of the analog-to-digital converters 3.sub.1, 3.sub.2, . . . , 3.sub.N and the downstream Fourier transformers 13.sub.1, 13.sub.2, . . . , 13.sub.N can be implemented in the equalization and decoupling unit 8.

[0127] The equalization and decoupling unit 8 is in turn made up of a number of parallel-connected equalization and decoupling channels 9.sub.1, 9.sub.2, . . . , 9.sub.M corresponding to the number M of Nyquist zones within the bandwidth of the digitized high-frequency signal.

[0128] Each equalization and decoupling channel 9.sub.1, 9.sub.2, . . . , 9.sub.M has a number of parallel multiplier elements 14.sub.11, 14.sub.12, . . . , 14.sub.1N or 14.sub.21, 14.sub.22, . . . , 14.sub.2N or 14.sub.M1, 14.sub.M2, . . . , 14.sub.M corresponding to the number N of parallel-connected analog-to-digital converters 3.sub.1, 3.sub.2, . . . , 3.sub.N.

[0129] A number of multiplier elements 14.sub.11, 14.sub.12, . . . , 14.sub.1N or 14.sub.21, 14.sub.22, . . . , 14.sub.2N or 14.sub.M1, 14.sub.M2, . . . , 14.sub.MN corresponding to the number N of Nyquist zones contained in the bandwidth of the digitized high-frequency signal are in each case present in each equalization and decoupling channel 9.sub.1, 9.sub.2, . . . , 9.sub.M. The input of each multiplier element of an equalization and decoupling channel 9.sub.1, 9.sub.2, . . . , 9.sub.M is connected in each case to the output of an associated Fourier transformer 13.sub.1, 13.sub.2, . . . , 13.sub.N.

[0130] Each individual multiplier element of an equalization and decoupling channel equalizes the respectively supplied discrete spectrum of the associated digitized filtered signal in such a way that, following a summation of all discrete spectra, a spectrum containing only the spectral components of the high-frequency signal x(t) within a Nyquist zone associated in each case with the equalization and decoupling channel is present at the outputs of the multiplier elements associated with an equalization and decoupling channel. For this purpose, each multiplier element 14.sub.ij (where i{1 . . . M} and j{1 . . . N}) in each case multiplies the baseband spectral component L.sub.j(k.Math.f) of the respective spectrum Y.sub.j(k.Math.f) for each spectral frequency k.Math.f by the associated spectral equalization coefficient {circumflex over (F)}.sub.i.sup.j(k.Math.f) which is calculated, for example, according to equation (8).

[0131] The outputs of all multiplier elements of an equalization and decoupling channel 9.sub.1, 9.sub.2, . . . , 9.sub.M are connected to the inputs of a summing element 15.sub.1, 15.sub.2, . . . , 15.sub.Massociated in each case with the respective equalization and decoupling channel 9.sub.1, 9.sub.2, . . . , 9.sub.M. The summing element 15.sub.1, 15.sub.2, . . . , 15.sub.M of a respective equalization and decoupling channel 9.sub.1, 9.sub.2, . . . , 9.sub.M in each case supplies at its output all digitized spectral components X.sub.1(k.Math.f), . . . , X.sub.M(k.Math.f) of the high-frequency signal within the Nyquist zone associated in each case with the respective equalization and decoupling channel 9.sub.1, 9.sub.2, . . . , 9.sub.M and therefore within the spectral range of the high-frequency signal associated with the respective Nyquist zone.

[0132] In the second subvariant of the second embodiment of the device according to the invention for receiving at least one high-frequency signal using parallel and undersampled baseband signal processing according to FIG. 3B, the spectral passband of the filter frequency responses associated in each case with the individual filters 1.sub.1, 1.sub.2, . . . , 1.sub.N corresponds in each case to only one single Nyquist zone of the digitized high-frequency signal. In this case, the individual equalization and decoupling channel 9.sub.1, 9.sub.2, . . . , 9.sub.M is reduced in each case to a single multiplier element 14.sub.11, 14.sub.22, . . . , 14.sub.NN.

[0133] In each multiplier element 14.sub.11, 14.sub.22, . . . , 14.sub.NN, the baseband spectral component L.sub.1(k.Math.f), L.sub.2(k.Math.f), . . . , L.sub.N(k.Math.f) of the respectively supplied digitized spectrum Y.sub.1(k.Math.f), Y.sub.2(k.Math.f), . . . , Y.sub.N(k.Math.f) is multiplied in each case for each spectral frequency k.Math.f by the associated spectral equalization coefficients {circumflex over (F)}.sub.1.sup.1(k.Math.f), {circumflex over (F)}.sub.2.sup.2(k.Math.f), . . . , {circumflex over (F)}.sub.N.sup.N(k.Math.f). Since the number of parallel-connected analog-to-digital converters 3.sub.1, 3.sub.2, . . . , 3.sub.N corresponds in this special case of the second subvariant to the number of Nyquist zones within the bandwidth of the digitized high-frequency signal, N=M and therefore {circumflex over (F)}.sub.N.sup.N(k.Math.f)={circumflex over (F)}.sub.M.sup.M(k.Math.f) apply.

[0134] The spectral equalization coefficients {circumflex over (F)}.sub.1.sup.1(k.Math.f), {circumflex over (F)}.sub.2.sup.2(k.Math.f), . . . , {circumflex over (F)}.sub.N.sup.N(k.Math.f) at the individual spectral frequencies k.Math.f are obtained in the first equalizer variant, for example, in each case from the inverse filter frequency response of the respective upstream filter 1.sub.1, 1.sub.2, . . . , l.sub.N at the same spectral frequencies k.Math.f.

[0135] At the outputs of the individual multiplier elements 14.sub.11, 14.sub.22, . . . , 14.sub.NN, which simultaneously also represent the outputs of the equalization and decoupling unit 8, the digitized spectral components X.sub.1(k.Math.f), . . . , X.sub.N(k.Math.f) of the high-frequency signal are thus present in the individual Nyquist zones and therefore in the individual spectral ranges of the Nyquist zones associated with the digitized high-frequency signal.

[0136] Finally, the device according to the invention for transmitting at least one high-frequency signal using parallel and undersampled baseband signaling processing is explained in detail below with reference to the block diagram in FIG. 7 in combination with the method according to the invention for transmitting at least one high-frequency signal using parallel and undersampled baseband signaling processing with reference to the flow diagram in FIG. 8:

[0137] If a multicarrier system (DMT/OFDM) is used for the transmission, the symbols, present in the spectral range, of all digital baseband signals u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A) to be transmitted in each case in a high-frequency signal z(t) are read in the first method step S200 from an internal memory of a digital signal processing unit 16. The symbols present in the spectral range may, for example, be OFDM or DMT symbols.

[0138] The serially read symbols of each individual digital signal u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A) are converted in the same method step S200 in an associated unit for serial-to-parallel conversion 17.sub.1, 17.sub.2, . . . , 17.sub.N into individual units of in each case parallel symbols. The number of parallel symbols corresponds to the spectral length of the inverse Spectral Transformer (iST) 18.sub.1, 18.sub.2, . . . , 18.sub.N connected in each case downstream of the individual unit for the serial-to-parallel conversion 17.sub.1, 17.sub.2, . . . , 17.sub.N. The inverse spectral transformers 18.sub.1, 18.sub.2, . . . , 18.sub.N are in each case preferably an inverse Fourier transformer which may, for example, be an Inverse Fast Fourier Transformer (IFFT) or an Inverse Discrete Fourier Transformer (IDFT).

[0139] In the individual inverse spectral transformers 18.sub.1, 18.sub.2, . . . , 18.sub.N, the units of parallel symbols associated in each case with a respective digital signal u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A) are converted successively in the next method step S210 into associated units of parallel sampling values of the digital signals u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A).

[0140] Finally, in the same method step S210, a guard interval is added in each case in a unit (not shown in FIG. 7) for inserting a guard interval.

[0141] The functions of the serial-to-parallel conversion, the inverse spectral transformation and the insertion of a guard interval are preferably already carried out in the digital signal processing unit 16.

[0142] In the following method step S220, the sampling values associated in each case with the individual digital signals u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A) are fed successively to a digital-to-analog converter 19.sub.1, 19.sub.2, . . . , 19.sub.N in each case downstream of the associated inverse spectral transformer 18.sub.1, 18.sub.2, . . . , 18.sub.N and are converted into an associated analog signal v.sub.1(t), v.sub.2(t), . . . , v.sub.N(t).

[0143] If a different transmission method is used instead of DMT/OFDM, the digital baseband data u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A) are fed in the method step S220 to the digital to-analog converter 19.sub.1, 19.sub.2, . . . , 19.sub.N and are converted into an associated analog signal v.sub.1(t), v.sub.2(t), . . . , v.sub.N(t).

[0144] It should be noted here that the sampling values of the individual digital signals u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A) must be present in the baseband in the preceding processing steps at the inputs of the individual digital-to-analog converters 19.sub.1, 19.sub.2, . . . , 19.sub.N, i.e. in a sampling rate which is less, in particular significantly less, than double the highest spectral frequency present in each case in the associated analog signal v.sub.1(t), v.sub.2(t), . . . , v.sub.N(t).

[0145] The timing of the individual digital-to-analog converters 19.sub.1, 19.sub.2, . . . , 19.sub.N is provided here by a common clock

[00008]$\frac{1}{T_A}$

which is supplied by a common clock source 20.

[0146] The sampling values of the individual digital signals u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A) are preferably not held in the individual digital-to-analog converters 19.sub.1, 19.sub.2, . . . , 19.sub.Nover the entire sampling period T.sub.A, but only over a sampling period T.sub.A* significantly reduced in comparison with the sampling period T.sub.A.

[0147] In the following method step S230, the spectral components of the respective analog signal v.sub.1(t), v.sub.2(t), . . . , v.sub.N(t) which lie essentially in the passband of the filter frequency response associated with the respective filter 21.sub.1, 21.sub.2, . . . , 21.sub.N are filtered in the filters 21.sub.1, 21.sub.2, . . . , 21.sub.N located in each case downstream of the individual digital-to-analog converters 19.sub.1, 19.sub.2, . . . , 19.sub.N. The filter frequency responses of the individual filters 21.sub.1, 21.sub.2, . . . , 21.sub.N differ from one another and in total cover the entire spectral range of the high-frequency signal z(t) to be transmitted. The filter frequency responses of each individual filter 21.sub.1, 21.sub.2, . . . , 21.sub.N may extend in each case completely or partially over one Nyquist zone or a plurality of Nyquist zones. A filtered signal z.sub.1(t), z.sub.2(t), . . . , z.sub.N(t) is present in each case at the output of each individual filter 21.sub.1, 21.sub.2, . . . , 21.sub.N.

[0148] In one special case, only the spectral components in the frequency range of a respectively different Nyquist zone assigned in each case to the respective analog signal v.sub.1(t), v.sub.2(t), . . . , v.sub.N(t) are filtered in each case in each individual filter 21.sub.1, 21.sub.2, . . . , 21.sub.N. In the case described, the filtered signals z.sub.1(t), z.sub.2(t), . . . , z.sub.N(t) designed as bandpass signals at the outputs of the individual filters 21.sub.1, 21.sub.2, . . . , 21.sub.N implemented as bandpass filters in each case consequently contain only the spectral components of a single Nyquist zone, said Nyquist zones in each case differing from one another.

[0149] In the concluding method step S240, the signals z.sub.1(t), z.sub.2(t), . . . , z.sub.N(t) are added in a summing element 22 downstream of the filters 21.sub.1, 21.sub.2, . . . , 21.sub.N to a high-frequency signal z(t) to be transmitted. Whereas the high-frequency signal x(t) received in the receiver is superimposed with a transmission channel interference and is distorted, the high-frequency signal z(t) to be transmitted in the transmitter is free from transmission channel interference.

[0150] According to the invention, the information contained or transmitted in each case in the individual digital signals u.sub.1(n.Math.T.sub.A), u.sub.2(n.Math.T.sub.A), . . . , u.sub.N(n.Math.T.sub.A) is contained in the high-frequency signal to be transmitted z(t) without the performance of a mixing into the high-frequency band.

[0151] A further essential technical advantage of the method according to the invention and the device according to the invention is that the technical structure is adaptable to the technical characteristics of the high-frequency signal to be processed and to the precision in the capture and further processing or in the generation of the high-frequency signal, and is therefore scalable. It is thus possible to adapt the number of parallel analog-to-digital converters or parallel digital-to-analog converters on the one hand to the bandwidth of the high-frequency signal or to the spectral ranges of the high-frequency signal which are to be examined, and also to the used sampling frequency of the undersampling.

[0152] The method according to the invention and the device according to the invention are not restricted to the embodiments, subvariants and variants shown. In particular, all combinations of the features claimed in each case in the individual patent claims, the features disclosed in the description and the features presented in each case in the figures of the drawing are also encompassed by the invention, insofar as they are technically appropriate.