JP2002511708A - System, apparatus and method for improving the definition characteristics of a transform domain signal - Google Patents

Abstract

(57) [Summary] A system, apparatus, and method for improving the definition characteristic of a conversion area symbol. The apparatus comprises: a signal mapping device for mapping input data to blocks of symbols in a first region and generating offset bits corresponding to each block of symbols; and for mapping blocks of symbols and corresponding offset bits in a first region. In response, a perturbation / conversion device is provided for generating a block of perturbation conversion domain symbols and improving the definition characteristics of the conversion symbol.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[Related application]

This application is a continuation-in-part of U.S. Ser. No. 09 / 058,671, filed Apr. 10, 1998, which is hereby incorporated by reference in its entirety.

[0002]

[Industrial applications]

The present invention relates to a system, an apparatus and a method for improving a definition characteristic of a transform domain signal, and more particularly to a peak-to-average (PAR) peak-to-average energy ratio of a time domain signal.
The present invention relates to a system, an apparatus, and a method for reducing an energy ratio.

[0003]

[Prior art]

A large time-domain maximum average energy ratio (PAR) of a discrete multitone (DMT) signal is often a major drawback of DMT systems. This problem is addressed by orthogonal frequency division multiplexing (OFDM).
n multiplexing and orthogonal quadrature amplitude (OQAM)
It also exists in a system using another modulation method, such as a system using itude modulation.

When the PAR is large, a high-precision digital-to-analog converter (DAC: digital-to-analo
g converter), or the system must be resistant to signal distortion (clipping) introduced when the input signal exceeds the DAC range. For a given DAC accuracy, scaling (adjusting) the input signal so that the signal value is always within the range can result in excessive quantization noise. On the other hand, insufficient scaling of the signal results in excessive clipping noise.

[0005] Several approaches have been proposed to reduce the time domain maximum amplitude of DMT and OFDM symbols. These methods can be classified into three types. First
In the type, the same data is expressed using a plurality of symbols, and a reserved tone (reservation tone) is used.
The receiver informs the receiver which symbol has been transmitted using the side information transmitted on the ed tone). For example, JS Chow, JAC Bingham and MS Flowers, "Mitigating clipping noise in multicarrier systems" Proc. 1997 Int. Co.
In nf. Commun. (ICC '97), pp. 715-719 (June 1997), if the peak of the DMT symbol is too high, the DMT symbol is scaled and the counting ratio is received using the reserved tone. Relay to the machine. In this method, the signal-to-noise ratio (SNR) of the transmitted symbol is reduced, thereby increasing the bit error rate. T1E1.4 / 97270 submitted for ADSL Standard Version 2 in September 1997
In Djokovic's "PAR reduction without noise enhancement", the transmitter makes a choice between the original DMT symbol and the conjugate formed by scrambling the symbol. "A method to redu" by DJ Mestdagh and PM Spruyt
ce the probability of clipping in DMT-based transceivers ”IEEE Trans. On
Commun., Vol. 44. 1234-1238 (October 1996) adds a pseudorandom phase sequence to the original DMT symbol. The biggest drawback of this type of method is that the transmitter must rely on side information for the symbols to transmit to the receiver. In addition to loss of data rate and increased bandwidth, corruption of side information will destroy the entire DMT symbol.

A second type of PAR reduction method is based on determining a sequence having excellent PAR characteristics. For example, S. Shepherd, J. Orriss and S. Barton "Asymptotic li
mits in peak envelope power reduction by redundant coding in orthogonal
frequency-division multiplex modulation, IEEE Trans.on Commun., vol. 46,
pp. 5-10 (January 1998)). These methods typically result in a loss of data rate because they are performed by removing the "bad" time-domain sequence from the set of possible transmitted symbols. In addition, these methods "good" the data
Need to map to symbols. This map is typically implemented via a look-up table. The required look-up table size becomes impractical for DMT systems having a large number of tones and large constellations.

[0007] In the third type of method, PAR reduction is performed by redundant signal representation. In this way, a particular block of data can be represented by any of a plurality of possible transmitted signals from an equivalence class. The "most desirable" class--in this case, the one with the smaller time-domain peak value--is representative and is selected for transmission. In such a manner, the receiver is designed to operate on a "modulo equivalence class" that produces a data block associated with the equivalence class whenever it detects an element of that class. In this way, the receiver does not need to know the exact algorithm used to select the class representing the transmitter. One way to operate the "modulo equivalence class" in the DMT case is to simply let the receiver ignore the contents of the various frequency bins. A. Gatherer and M
.Polley "Controlling clipping probability in DMT transmission" Conf. R
ecord 31st Asilomar Conf. On Sign. Sys. And Comp. pp. xx-yy, (1997
A. Gatherer and M. Polley, "Proposed PAR Reduction Technique"
s for G. lite Universal ADSL Technical Group Contribution "TG / 98-025 (
Feb. 4, 1998); J. Tellado and JM Cioffi, “PAR reduction in mult.
for the icarrier transmission systems "T1E1.4 standard committee 97-367 (19
Dec. 1997). For any particular data block, the transmitter can place values in these unused bins to minimize (as much as possible) the peak value of the transmitted time domain signal. These methods cause large data rate losses because some bins are not used for transmitting data.

Accordingly, a PAR reduction method for reducing DPAR by using data carrying or complex frequency bins in a DMT modulation method without affecting data rate, and further reducing PAR by using unused frequency bins Is required. Also, other modulation methods are generally applicable, and there is a need for such a method that improves the time domain or other characteristics of the transformed domain signal in general.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention generally relates to a signal after block conversion--hereinafter a transform domain signal (transform-
The present invention relates to a system and a method for improving a definition characteristic of a domain signal. To provide a more easily understandable description of the present invention, the maximum average energy ratio (PAR: peak) of a time-domain signal (referred to more generally herein as a transform-domain signal) is provided herein. A practical application to mitigate the invention in a discrete multitone (DMT) modulation scheme will be described. As will be apparent to those skilled in the art, the present invention is directed to orthogonal frequency division multiplexing (OFDM).
Other modulation methods are generally applicable, such as: orthogonal frequency division multiplexing and orthogonal quadrature amplitude modulation (OQAM). Further, as will be apparent to those skilled in the art, the present invention relates to PAR
Not only can it be used to improve other defined characteristics of the transform domain signal. For example, if the transform domain signal is, for example, a simple conventional telephone service (POTS: plain
Voice band (0) caused by nonlinearity in telephone of old telephone service)
-4 kHz) It can also be used to shape the spectrum of the signal after introducing nonlinearities, such as in a splitterless operation of a DSL system where it is desirable to reduce interference.

[0010] The transmission method of the DMT system is based on a block of N symbols. Each symbol in the block corresponds to a different frequency bin. Therefore, each symbol block X
It is composed of the frequency-domain symbols X ₀ -X _N-1. In an asymmetrical digital subscriber line (ADSL) system using DMT,
N is an even number), zero (X ₀₎ and Nyquist (X _{N / 2)} not symbols transmitted in the frequency bins. There are symbols transmitted in the lower N / 2-1 complex frequency bin (X ₁ −X _{N / 2-1} ) and the upper N / 2-1 complex frequency bin (X _{N / 2 + 1} −X _N-1 ) is the lower N / 2
Since it is selected as a complex conjugate image of -1, the frequency domain signal derived therefrom is a Hermitian symm (Hermitian symm) required for realizing the time domain signal.
etry). Therefore, actually, n blocks per block (where n = N / 2-1
) Complex frequency bins.

As shown in FIG. 1A, DMT transmitter 10 receives a serial digital bit stream on line 12 from a data terminal device (not shown) such as a personal computer. The serial bit stream is converted to a parallel format by a serial-to-parallel converter 14. For each block, the serial-to-parallel converter outputs (kn-r) + m information bits v, u. Here, r is the number of redundant bits, and k is the number of bits required to represent the equivalent class in the extended constellation. The terms r and k and the terms equivalence class and extended conformation are described below.

As noted above, n is the number of complex frequency bin symbols generated per block. The m base information bit u is m = m ₁ + m ₂ +. . . corresponds to _mn . However,
m _i represents the number of base information bits transmitted in complex frequency bin i. for each n-number of complex frequency bins, based conformational mapping device maps the m _i pieces of base information bits in the symbol from the base conformation. The ith base conformation contains 2 ^mi points.

In the base constellation mapping unit 16, the base constellation for each frequency bin is determined by the channel quality for the frequency bin, and n base symbols g are generated for each block. Channel quality is usually
Determined by exploring the channels in the sequence. The size of the constellation, ie, the number of input data bits that can be represented by a symbol selected from the constellation, depends on the quality of the channel within the frequency range of the bin. A good quality channel can use a denser constellation where points are more closely spaced, so more bits can be transmitted with each symbol. Thus, the number of input data bits represented by a block of symbols depends on the quality of the channel.

To illustrate the general idea of the base constellation according to the invention in a specific example, consider a DMT system with a base constellation capable of transmitting 2 bits per symbol. FIG. 2 illustrates a base configuration 30 (assuming that it is in one quadrant) including points A, B, C, and D. From these points, the base conformation mapping device 16
Selects the base conformation symbol. Furthermore the present invention, at least some of the base conformation is expanded include an additional point of greater than 2 ^mi points required to transmit m _i pieces of base information bits per symbol. These extended conformations are divided into equivalence classes.

[0015] The base signal constellation is extended to accommodate the transmission of up to k additional bits per symbol. Some of these kn bits are used to transmit additional information bits, and other bits are used to give the transmitter flexibility in selecting the symbol to be transmitted. This extra degree of freedom can be used to optimize the target function of the resulting signal, for example, the maximum time domain amplitude or spectral shape of the transmitted signal after nonlinear transformation. Such kn additional bits are called "offset bits".

In the example shown in FIG. 2, the base constellation 30 is expanded by a factor of four to
Form 2 Thus, k = 2 offset bits per symbol are required to determine which of the equivalents in the extended constellation 32 will be transmitted. The extended constellation 32 includes a base constellation 30 and extended regions 34, 36, 38 each including four points marked A-D. All points with the same label belong to the same equivalence class.

In the example of FIG. 2, the extended configuration is formed from the base configuration by repeating the base configuration in each of the four quadrants. The minimum distance between the i-th neighboring point in the constellation about the symbols is defined as d _i. This distance depends on the channel quality. This type of conformational expansion is called additive expansion. This is because the equivalent point in the extended conformation is a value in the base conformation (0 or-in each dimension in this example).
2d _i ).

In this example, the base conformation was expanded by a factor of four. Of course, the base configuration can also be extended by multiples greater than four. For example, 0 or + in each dimension
Additional equivalent class points can be generated by adding integer multiples of / −2d _i to points in the base conformation. As shown below, both performance capabilities and system complexity increase with larger constellation extensions. For a general square constellation with c points, extended constellation points can be generated by adding an integer multiple of cd _i / 2 to each dimension. Other methods of generating extended conformations with multiple conformation point equivalence classes will be apparent to those skilled in the art. Hereinafter, the extended configuration generated by the rotation of the base configuration will be described.

Referring again to FIG. 1A, kn-r bits, v are the offset coset representative generator (offs
et coset representative generator) 18. The generator 18 generates kn offset (or, more specifically, coset representative offset) bits. The n base symbols g and the kn coset representative offset bits t are combined by the extended constellation mapping device 20 to form n extended symbols h from the extended constellation. In the above example, when k = 2, the base symbol g corresponds to the base constellation, and the k offset bits select the quadrant of the corresponding symbol. The n extended symbols are converted into N conjugate symmetric symbols X by the Hermite symmetric block generator 22.
It is mapped to the block of (X ₀ −X _N−1 ). Hermite symmetric block generator 2
In 2, n symbols are mapped to symbols X ₁ −X _{N / 2-1} , where symbols X _{N / 2 + 1} −X _N−1 are X ₁ −X _{N / 2-1} complex conjugates . A block of N symbols X transforms the N frequency domain symbols into N time domain symbols x (x ₀ -x _{N -1} ), such as an inverse discrete Fourier transform (IDFT) device 24 Sent to the inverse converter. The perturbation device 26 changes the N time-domain symbols x,
By modifying the coset representative offset bit t to improve the definition characteristics of the N time domain symbols, in this example, by minimizing the peak value as described below, the perturbed time domain block y Generate The perturbed time-domain block y is sent to a parallel-to-serial converter 28, which converts the perturbed time-domain block y from parallel format to serial format and sends it over the channel.

Base conformation mapping device 16, offset coset representative offset generator 18
, The extended constellation mapping device 20 and the Hermitian symmetric block generator 22 collectively form a signal mapping device 23. The mapping device 23 is a serial-
The data input from the parallel converter 14 is mapped to the frequency domain symbol block X. The IDFT device 24 and the perturbation device 26 collectively form a perturbation / conversion device 27.

Referring to FIG. 3, a schematic block diagram of a receiver 40 according to the present invention is shown. The perturbed time-domain symbol y after passing through the channel is received as a symbol w in the serial-parallel converter 44. The converter 44 receives the time domain symbol w in a serial format and converts it into a block of the received time domain symbol w, w ₀ −w _N−1 . The block of the received time domain symbol w, w ₀ −w _N−1 is a discrete Fourier transform (DF
T) fed to the device 46, converts a block of 46 time domain symbol to be received frequency domain symbols W, to W ₀ -W _N-1 blocks. Received frequency domain symbol W
, W ₀ -W _N−1 are sent to the frequency domain equalizer unit 48, which considers the effect of the channel on the transmitted perturbed frequency domain block Y, Y ₀ −Y _N−1 ,
The received symbols W, W ₀ -W _N-1 scaling to the transmitted block Y, Y ₀ -Y _N-1 of which is an estimated value symbol Y ', Y' to produce a ₀ -Y _'N-1.

The estimate of the transmitted perturbed frequency domain block Y ′ is sent to a base symbol and offset extraction unit 50, which extracts a base symbol g and an effective perturbation offset bit s. Since the offset bits have been modified in the perturbation device 26 of FIG. 1A, these bits are exactly the coset representative offset
Does not correspond to bit t. The base symbol and offset extraction device 50
First, each of the n symbols g transmitted in the block Y of N symbols is decoded to a point in the corresponding extended constellation. The offset bits s of these n symbols are sent to the offset decoder 52 to recover the offset information bits v '. This is equivalent to the offset information bit v described below. The n base symbols g are sent to the base conformation demapper (demapper) 54, and m
The base information bits u are recovered. The base symbol g corresponds to a point in the base constellation.

The decrypted information bits u, v are further processed and sent to a data terminal device such as a personal computer. Offset Coset Representative Generator The offset coset representative generator 18 is illustrated in more detail in FIG. 1x (kn-r
The offset information bits v, which are considered row vectors,
It is post-multiplied (modulo 2) (ie, filtered) by a matrix H- ^T having kn-r rows and kn columns to generate a 1xkn row vector of coset representative offset bits t. This bit t is sent to the extended constellation mapping device 20 and the perturbation device 26 of FIG.

An example of a matrix ^HT when n = 3 symbols are transmitted for one block and one redundant bit exists for one symbol (r = 3) is as follows:

[0025]

(Equation 1)

In this example, assuming that the offset information bits are given by v = [v ₀ v ₁ v ₂ ], the output bits are simply input bits sandwiching zero, ie, t = [v ₀ 0 v ₁₀ v ₂₀ ]. Perturbation device The perturbation device 26 is shown in more detail in FIG. The perturbation device 26 operates in the time domain and perturbs blocks of symbols. However, it can be easily modified to operate in the frequency domain as perturbation device 26 'in perturbation / conversion device 27' shown in FIG. 1B. kn residue class representative offset bit t is effective perturbation generator 7
Sent to 0. Effective perturbation generator 70 generates 2 ^r effective perturbation vectors (or a subset thereof for less complexity). Here, r is the number of redundant bits. In general, "perturbations" are not additive, but can be considered additive if the following approach is followed. Let y _{1 be} the perturbation time domain block corresponding to perturbation i. Where i =
0, 1,. . . , 2 ^r -1. Let p _i = y _i −x. However, x = y ₀ . The vector ｛p _i :
i = 0, 1,. . . ^Let 2 ^r -1} be the “effective perturbation vector” and x be the “nominal time domain block”.

The 2 ^r effective perturbation vectors (or a subset thereof)
receiving p _i (p _i , ₀ −p _i , _N−1 ) and adding each of these perturbation vectors to a nominal time-domain block x (x ₀ −x _N−1 ), yielding y _i = x + p _i Generate. The smallest vector y in the time domain peak (or with another objective function) is selected and transmitted.

When an extended constellation is formed by extending 2 ^k times for each of the n symbols, kn offset bits t in addition to m base bits are added to the channel in each block of symbols. Can be sent on. Of these kn offset bits, additional information bits are transmitted using the coset representative offset bits of kn-r, and the flexibility provided by the r redundant bits is used to complement the transform domain signal. Improve the characteristics of The greater the value of r, the greater the flexibility in improving the definition characteristics of the transform domain signal, but the lower the bit rate of information transmission. Effective Perturbation Generator The effective perturbation generator 70 will be described based on a binary linear code. However, as will be apparent to those skilled in the art, this structure can be extended to non-binary group codes.

The effective perturbation generator 70 of FIG. 6 generates an effective perturbation vector corresponding to the offset change of each symbol. The kn cosets given to the effective perturbation generator 70 represent a coset representing the defined linear code C generated by the perturbation codeword generator 80 using a matrix G having r rows and kn columns. Define.

The matrix block H ^-T matrix G and 4 are selected such that GH ^T = 0. However,
H is a matrix having a kn-r rows and kn columns, itself satisfies the characteristics _{^{H -T H T = I kn-}} r.
Here, I _kn-r is a unit matrix of (kn-r) x (kn-r). In other words, the H ^T is a right inverse of H ^-T (right inverse). It is required that G has a row rank r and H has a row rank kn-r. An example of a matrix G when n = 3, r = 3, k = 2 is as follows:

[0031]

(Equation 2)

Here, G can be selected as a generator matrix of any well-known binary linear code, and can be a truncated or terminated convolutional code that is optimized with respect to Hamming distance characteristics or according to other alternative criteria. May correspond.

Using the kn coset representative offset bits t, the effective perturbation generator 70 performs an exclusive OR operation, ie, modulo-2 addition, on the bits by an effective codeword c _i defined by the matrix G. Change the offset bit t. Codewords perturbation codeword generator 80 generates has a characteristic of cH ^T = 0.
If a valid codeword c XORing with coset representative offset bits t, it generates an exclusive OR operation s _i = tc _i effective perturbation offset bits t. Effective perturbation offset bits s _i are mapped to N symbol blocks P _i via perturbation mapping unit 82 and Hermite symmetric block generator 84. This will be described below. This selection process uses any of the effective perturbation offset bits s _i and decodes them into offset information bits v as described below.

Each set of kn effective perturbation offset bits s _i is represented by k offsets per symbol
Corresponds to a bit. Continuing with the above example that assumes k = 2, r = n, offset bits s _i define the quadrant of n symbols. Similarly, for each symbol, k = 2 bits define the displacement of the base-free point required to generate the appropriate point in the equivalence class including the original base-constellation point. In the nominal time domain block x, the offset bit is defined by t. Similarly, in this example, the effective perturbation offset bits s _i corresponds to changes in the quadrant of the transmitted symbols. Recall that t was generated from the offset information bits v. Therefore, the effective perturbation offset bits s _i formed from t is dependent on the information.

If H− ^T is defined in the above example and r = n (one redundant bit per symbol), the unmodified coset representative offset bit t consists of n pairs of bits. At this time, the second bit of each pair is zero. Thus, the coset representative offset bit t will only select from one of the two quadrants represented by 00 and 10. In this example, the valid codeword c _i consists of n pairs of bits, the first of each pair.
The bit becomes 0. If the second bit of the pair is not zero, c _i modifies the quadrant from 00 to 01 or from 10 to 11. Let d be the distance between neighboring points in the base constellation. In this example, if quadrant 00 is defined to indicate the quadrant containing the base configuration,
Quadrant 10 is the quadrant below the base constellation, quadrant 01 is the left quadrant of the base constellation, quadrant 1
One is defined as the remaining quadrant, and the effective perturbation offset bit modifies the coset selected by coset representative offset bit t, which depends on the information by a perturbation of 0 or -2d for each symbol.

The perturbation mapping unit 82 maps each set of effective perturbation offset bits s _i to n symbol perturbations. These n symbols perturbation represents the perturbation results from changing the offset bits from t to s _i. h _i '
It represents the n extended symbol corresponding to the symbol g and offset bits s _i. The n-perturbed symbol q _i is the difference between h _i ′ and h, that is, q _i = h _i ′ -h. In the above example, q _i contains 0 or −2d perturbations in each symbol.

For each set of effective perturbation offset bits s _i , n perturbed symbols (0 or −2d for each symbol in the above example) are Hermitian symmetric block generators 8.
By 4, is mapped to N symbol frequency domain symbols P _i with the complex conjugate symmetry. The operation of the Hermitian symmetric block generator 84 has been described above. The frequency domain symbol P _i is sent to the IDFT device 86 to generate a 2 ^r time domain perturbation vector p _i .

As will be apparent to those skilled in the art, different values of k, r, H, and G result in different perturbation vectors. In addition, different perturbation vectors are obtained by different bit mapping methods for symbols. The generated effective perturbation vector depends on the offset information bits v.

In general, when k, H, G, and r are constant and the mapping method is constant, a set of 2 ^r effective perturbation vectors is obtained from a set of 2 ^kn possible time domain perturbation vectors. Instead of generating these effective perturbation vectors for each incoming t as described above, the effective perturbation vector generator 70 stores all 2 ^kn time-domain perturbation vectors in memory and calculates the The coset representative offset bit t can also be used to determine which ^2r (or a subset thereof) are valid for a given t. Note that if the extended constellation is not an additive extension as described above, the perturbed symbols will depend on the base symbol g as well as the coset representative offset bit t. In this case, there will be 2 ^kn or more possible time domain perturbation vectors. Perturbation selector Perturbation selector 72 is shown in more detail in FIG. For each of the 2 ^r effective time domain perturbation vectors p _i , a perturbation time domain block y _i is calculated by block 90. Here, y _i = x + p _i . Next, all of the calculated y _i perturbation time domain blocks are evaluated at block 92 and y _i with the lowest peak value is selected as the perturbation time domain block of the symbol to be transmitted. Base Symbol and Offset Extractor The base symbol and offset extractor 50 of FIG. 3 maps the frequency domain equivalent block Y ′ to n symbol points in the extended constellation. Each point in the extended conformation is equal to each point in the base conformation (equivalence class representative). The offset indicates which of the 2 ^k equivalents was actually transmitted. The 2 ^k equivalence class points are represented by k offset bits per symbol. The transmitted equivalence class point is represented by kn offset bits s. These offset bits are sent to offset decoder 52, which determines the information bits to be encoded in the offset bits. This will be described below. The n equivalent class representatives in the base constellation are estimates of the transmitted base symbol g and are sent to the base constellation demapper 54. This device demaps these points to the estimate of the transmitted base information bit u. Offset Decoder The offset decoder 52 shown in more detail in FIG.
Is provided. In the matrix block 100, the 1 × kn row vector of offset bits s is post-multiplied (modulo 2) (ie, filtered) by a matrix H ^T having kn rows and kn−r columns to obtain 1 × (kn) of offset information bits v ′. -r) Recover row vector.

In order to show how each of the possible perturbation offset bits s _i is decoded into the same offset information bits, the encoding and decoding processes must be mathematically represented. . The recovered information bit v '(decryption) can be expressed mathematically as follows:

[0041]

[Equation 3]

The effective perturbation offset bit s (encoding) can then be expressed mathematically as:

[0043]

(Equation 4)

Where c = rG is an effective codeword generated by the perturbation codeword generator 80 of FIG. Substituting the right side of equation (4) for equation (3) with respect to s leads to the following equation:

[0045]

(Equation 5)

G, H ^T , and H ^-T are selected so that the following conditions (1) and (2) are satisfied: (1) H ^T H ^-T = I (where I is a unit matrix); 2) GH ^T = 0. Then, v ′ = v regardless of the value of r. In block-based systems, such as frame-based perturbation DMT-based systems, bits are mapped to blocks of N symbols. The invention described above assumes that the offset bits are modified on a block-by-block basis. In other words, all n = N / 2-1 symbols transmitted on the block of N symbols are jointly perturbed. In some cases, it may be convenient to divide the block into frames having a size less than n symbols. For example, if n and r are large, a large set of effective perturbation vectors must be generated, stored, and / or tested.
With a smaller frame size, if the perturbation is performed on a frame-by-frame basis, the number of effective perturbation vectors that must be tested, stored, and / or generated is reduced. The tradeoff for this method is that there is some performance loss. This is because the perturbations are selected on a frame-by-frame basis to optimize the desired properties. These performances can be somewhat restored by using look-ahead described below. This of course complicates the system.

When performing frame-based perturbation, the transmitter of the present invention differs in two ways: 1) the offset coset representative generator operates on kn / f-bit f-frames as follows: 2). A perturbator splits its input and operates on f frames of kn / f bits as shown below. The receiver of the present invention also differs in two respects: the offset decoder divides its input and operates on kn / f-bit f-frames as described below.

For simplicity, it is easiest to assume that n / f is an integer
Note that Otherwise, the offset coset representative generator and the perturbation
Position and offset decoders need to operate on different sized frames
is there. Even then, generalization to the case where n / f is not an integer is straightforward.Frame-based offset coset representative generator 9 receives the kn-r information bit v and outputs the kn-r information.
A frame divider 110 for dividing the bits v into frames of size kn / fr-r / f
You. These frames are v_iWhere n = n / f and r '= r / f. For this reason
, F, the information bits of kn′-r ′ are v_iSent through
The r 'redundant bits are used to improve the desired properties of the transform domain symbol. offset
These 1x (kn'-r ') frames of information bits are₀−112_{f -1} A matrix H having kn′-r ′ rows and kn ′ columns^-TMultiplication (modulo 2
) (Ie, filtered) and the kn 'coset representative offset bit t_j(t_o−t_f-1)of
Generate a 1xkn 'frame. f frame (t_o−t_f-1) In the frame concentrator 114
To form the kn coset representative offset bit t. This bit is
The extended constellation mapping device 20 and the perturbation device 26a in FIG.Frame-based perturbation device 10A includes a frame divider 120. This divider
, Kn coset representative offset bit t, and_o−t_f-1Indicated by
Is divided into f frames of size kn '. Alternatively, a frame of size kn '
It can also be obtained directly from the offset coset representative generator 18a. Coset representative off
Set bit t_jEach frame of the effective perturbation generator 112_j(112₀−112_{f- 1} ), The generator is 2^rGenerate an effective perturbation vector (or a subset of it)
To j-th perturbation selector 124 corresponding to the j-th frame._j(12
4₀−124_f-1).

In general, perturbations are not additive, but can be considered additive in the following manner. Let y _j , _{1 be} the perturbed time domain block corresponding to perturbation i. Where i = 0,
1,. . . , 2 ^r -1. Let p _{j, i} = y _{j, i} −y _{j, 0} and y _{j, o} = y _j−1 . This will be described later. Vector {p _{j, i} : i = 0, 1,. . . 2 ^r -1} is referred to as an “effective perturbation vector” corresponding to the j-th frame of the coset representative offset bit t _j .

The j th perturbation selector is provided with 2 ^r effective perturbation (or a subset thereof) vector p _{j, i} corresponding to the j th frame of the coset representative offset bit t _j .
Further, y _j-1 "and the output of the perturbation selector 124 _j-1, shown is also provided. A first perturbation selector, the perturbation selector 124 _0, nominal time domain block x is given. This y _-1 ". The perturbation selector 124 _j calculates y _{j, i} = y _j-1 "+ p _{j, i} for each of the effective perturbation vectors given to the perturbation selector 124 _j . Also, the perturbation selector 124 _j determines y _j " having the smallest time-domain peak. 124 _{j + 1} . The final perturbation selector 124 _f-1 outputs y = y _i-1 "to the channel.

In the perturbation device 26a, perturbations are sequentially selected for each frame. The performance of this device can be improved by incorporating preemption. That is, the perturbation output vector corresponding to the effective perturbation offset bit s _j based on the current frame only
Without selecting “y _j ”, perturbation selector 124 _j uses the effective perturbation offset bit s for the current frame and future frames to determine which perturbation output vector has the lowest peak time domain power. be able to.

In order to explain this concept, first, a preemption depth 1 is considered. 10B perturbation
The position 26b is a perturbation selector 124_jIs provided. The perturbation selector 124
Kuta 124_{j + 1}Find the effective perturbation offset bit
S_jIs the effective offset bit s_{j + 1}And y_{j + 1}"= Y_j-1"+ S_j+ S_{j + 1}No
To determine whether to minimize the peak power. Then the vector y_{j "}= Y_j-1"+ S_jIs a perturbation cell
Kuta 124_{j + 1}Is output to Similarly, perturbation selector 124_{j + 1}Is a perturbation selector_{j + Two} Anticipate.

If the preemption depth is Δ, the perturbation selector 124 _j preempts the effective perturbation vectors entering the perturbation selectors 124 _{j + 1} to 124 _{j +} Δ and determines which effective perturbation offset bits s _j are effective perturbation offset bits s _{j + 1} −s _{j +} Δ and y _{j + 1} "= y _j-1 " + s _j
+ S _{j + 1} +. . . It is determined whether the peak power of + s _{j +} Δ is reduced most. Note that in an implementation of this apparatus, it is not possible to preempt beyond a block boundary. Thus, the last Δ-1 perturbation selector will have a preemption depth less than Δ. Further, the last Δ-1 perturbation selector is a perturbation selector f−Δ−
1 is completely determined. The perturbation device 26b has a preemption depth Δ = 1.
When trying to improve the peak power of a symbol's time domain block, preemption beyond the block boundary is useless. For other objective functions, prefetching across block boundaries may be useful. As will be apparent to those skilled in the art, the present invention can be modified to provide preemption across block boundaries. If the prefetch depth Δ is equal to the number of frames f in one block, this method reduces the above to the first perturbation selector. That is, this is equivalent to a case where one frame of size n symbols is assumed. Frame-Based Effective Perturbation Generator FIG. 11 shows the structure of the effective perturbation generators 122 ₀ -122 _f-1 . The effective perturbation generator is provided with individual frames of kn 'bits corresponding to frames of n' symbols, and the effective N per symbol used to modify the time domain symbol x to minimize peak power. Generate a perturbation vector.

The effective perturbation generator generates an effective perturbation vector corresponding to a change in the offset bits for the n ′ symbols in each frame. Kn given to the effective perturbation generator
The 'coset representative offset bit t _j defines the coset representative for the defined linear code C generated by the perturbed codeword generator 126 using a matrix G having r' rows and kn 'columns. Matrix G and the matrix block H ^-T (in offset coset representative generator 18 of FIG. 1) is chosen to be GH ^T = 0. Where H is kn'-r
It is a matrix having a 'rows and kn' column, itself satisfies the characteristics of the _{^{H -T H T = I kn'-}} r '. At this time, I _kn'-r 'is a (kn'-r') x (kn'-r ') unit matrix. In other words, HT
Is the right reciprocal of ^HT . We need G to have row rank r 'and H to have row rank kn'-r'.

In the perturbation codeword generator 126, 2 ^{r ′} codewords (or a subset thereof) c _i = r _i G have 2 ^{r ′} possible choices (or their corresponding) for r ′ redundant bits denoted by r _i. ) Is generated by post-multiplying all the subsets) by G.
Codeword generated by the perturbation codeword generator 126 has a characteristic of cH ^T = 0. There is a valid codeword c _i of 2 ^{r '.} Using the kn ′ coset representative bit t _j , the effective perturbation generator performs an exclusive OR operation, ie, modulo-2 addition, on the bits with the effective codeword c _i defined by the perturbation codeword generator 126 to produce a coset to change the representative sign bits t _j. The effective perturbation bit s _{j, i} =
t _j c _i is mapped to N symbol blocks P _{j, i} via perturbation mapping unit 128 and Hermite symmetric block generator 130. With this selection process,
Note that any of the effective perturbation bits s _{j, i} can be used, which are decoded into information bits v _j as described below.

The perturbation mapping unit 128 converts each set of effective perturbation offset bits s _{j, i} to n ′
Maps to the symbol perturbation q _{j, i} '. These n ′ symbol perturbations are t _j or s _{j, i}
Represents the perturbation resulting from changing the offset bits of frame j of FIG. Let h _j denote the j-th frame of the n ′ extension symbol of h. These extended symbol was determined from the base symbol g and offset bits t _j. Base symbol g
And the n ′ extended symbol corresponding to the offset bit s _{j, i} is denoted by h _{j, i} ′. The n ′ perturbation symbol q _{j, i} ′ is the difference between h _{j, i} ′ and h _j , that is, q _{j, i} = h _{j, i} −h _j . These n ′ symbol perturbations q _j , _i ′ are mapped to a set of q _{j, i} , n symbol perturbations. However, only the j-th frame of the n ′ symbol in the set of n-symbol perturbations q _{j, i} is non-zero. For each set of effective perturbation offset bits s _{j, i} , n perturbed symbols are mapped by the Hermitian symmetric block generator 130 to N symbol frequency domain symbols P _{j, i} having complex conjugate symmetry. Hermite symmetric block generator 13
The operation of 0 has been described above. The frequency domain symbol P _{j, i} is sent to the IDFT device 132 to generate a 2 ^{r ′} time domain perturbation vector p _{j, i} .

As will be apparent to those skilled in the art, different values of k, r ', H, G
The following perturbation vector is derived. Also, when mapping bits to symbols
If different methods are used, different perturbation vectors are obtained. Generated effective perturbations
Vector is information bits v_jDepends on. In general, the values of k, H, G, r 'are constant,
If the mapping method is constant, 2^{r '}The set of effective perturbation vectors is 2^{kn '}Possible time
Obtained from a set of region perturbation vectors. As above, each incoming t_jAbout
Without generating these effective perturbation vectors, the effective perturbation vector generator^{kn '}Possible
All time-domain perturbation vectors are stored in memory, and the coset representative offset bit is stored.
T_jWhich of these perturbation vectors^{r '}Individual vectors (or
That subset) is a particular t_jCan be determined as valid for Extension
If the constellation is not an additive extension as described above, the perturbed symbol is the coset representative offset.
Bit t_jNot only depends on the base symbol g but also on
. In this case, 2^{kn '}There are the above possible time domain perturbation vectors.Frame-based perturbation selector FIG. 12 shows a perturbation selector 124._j(124₀−124_f-1) Are shown in more detail. Two ^{r '} Effective time domain perturbation vector p_{j, i}(Or a subset of it)
Dynamic time domain block y_{j, i}Is calculated by block 140. I.e. y_{j, i}= Y_{j -1} "+ P_{j, i}It is. (Note: perturbation selector 124₀The input to is y_-1"= X
. ) Calculated y_{j, i}All of the perturbed time domain blocks are evaluated in block 142.
Value and the lowest peak value y_j"= Y_{j, i}Is the perturbation selector 124_{j + 1}Given to
The symbol is selected as the perturbed time domain block. (Note: Perturbation selector 12
4_f-1Output of channel y_f-1"= Y is output.)Frame-based offset decoder The offset decoder 52a of FIG. 13 uses the kn 'effective perturbation offset bit s
Are divided into frames of kn ′ = kn / f bits, respectively.
You. Each frame s₀−s_f-1Is a matrix block (152₀-152_f-1).
In the j-th matrix block, j-th of 1xkn 'effective perturbation offset bits
A matrix H whose frame has kn 'rows and kn'-r' columns^TMultiplied by (modulo 2)
(Ie, filtered) and 1 × (kn′-r ′) offset information bits v ′_jThe j-th
Recover the frame. Offset information bit v '₀−v '_f-1F frame is a frame
It is sent to the coupler 154, and the coupler 154 concatenates the f-frames and outputs the kn-r offset
Forming an estimate of the report bit v '.

To explain how each of the possible effective perturbation offset bits s _{j, i} is decoded into the same offset information bit v _j , the encoding and decoding process must be mathematically represented. Must. Recovered offset information bits v _j '
(Decryption) can be expressed mathematically by the following equation:

[0059]

(Equation 6)

The effective perturbation offset bit s _j (encoding) can also be expressed mathematically as:

[0061]

(Equation 7)

Here, c = rG is an effective codeword generated by the perturbation codeword generator 126 in FIG. Substituting the right hand side of equation (7) into equation (6) with respect to s leads to the following equation:

[0063]

(Equation 8)

G, HT, and HT are selected so that the following conditions are satisfied: (1) H ^T H ^-T = I (where I is a unit matrix); (2) GH ^T = 0. Then, v _j ′ = v _j regardless of the value of r. Splitterless Operation In an asymmetric digital subscriber line (ADSL) modem operating without a splitter, the transmitted ADSL signal causes interference in the POTS telephone voice band (0-4 kHz). This interference is the result of intermodulation effects due to the non-linear devices of the POTS telephone. This interference can be reduced by using the invention described above to improve the appropriate target function of the transmitted signal. One possible objective function is to form a notch in the signal's spectrum after the transmitted ADSL signal has gone through a 2 kHz nonlinearity.

The time is represented by the subscript k. Let X (k) denote the kth non-perturbed time domain DMT symbol block and let x (k + 1) denote the k + 1th block. Similarly, let y _{k ′} denote the _k- th transmitted perturbed time domain DMT symbol block, and let y (k + 1) denote the k + 1-th symbol block. Let Z (k) represent the output of spectrum calculator 164 of FIG. 14, ie, the spectrum of transmitted signal y transmitted by time k, after the POTS nonlinearity evaluated at 2 kHz.

This objective function can be improved by the invention disclosed above. In practice, this can be achieved by incorporating a new objective function into the selection criteria of the perturbation selector. As shown in FIG. 14, the perturbation selector 72a of this embodiment includes a perturbator 160, a nonlinear device 162, a spectrum calculator 164, and a selector 1
66. Perturber 160 modifies the nominal time-domain block x with each of the effective perturbation vectors to generate candidate transmission blocks _yi . These blocks are POTS
To a nonlinear device 162 that mimics the nonlinearity of The spectrum calculator 164 calculates the power of the nonlinear deformation signal near 2 kHz, and the selector 166 selects a candidate perturbation time domain block y _i that minimizes the output of the spectrum calculator 164.

In the above example, the “additive” conformational extension described with respect to FIG. 2 was assumed. For this particular objective function, further advantages can be obtained by using the alternative conformational extensions described below. Among other things, the implementation complexity can be significantly reduced, as will become apparent below.

The rotational expansion constellation 170 of FIG. 16 moves the base constellation as shown in FIG. 2 (
This is formed by rotating the symbols in the base constellation 172 rather than in the above (referred to as additive constellation extension).

In the example shown in FIG. 15, consider a DMT system with a base constellation 172 capable of transmitting 2 bits per symbol. In the base conformation 172,
Points A, B, C, and D are included, from which base constellation symbols are selected by the base constellation mapping device 16 of FIG. Base constellation 172 is expanded by a factor of four to form a 16-point constellation. Thus, k = 2 bits per symbol are needed to determine which equivalent point in the extended constellation is transmitted. The expansion pocket 170 includes a base constellation 172 and expansion regions 174, 176, 178, each containing four points labeled AD. Extended constellation 170 is formed from base constellation 172 by rotating each point in the base constellation by 0, 90, 180, and 270 degrees.

For each of the n base symbols g, the base constellation mapping unit 16 selects a point in the base constellation 172. The extended conformation mapping device 20 of FIG.
Or 2n (2 bits per symbol) coset representative offset bit t
Is used to rotate n symbols by 0, 90, 180 and 270 degrees. 1
One way to define a two-bit mapping per symbol is that 00 corresponds to 0 degree rotation, 01 corresponds to 90 degree rotation, 11 corresponds to 180 degree rotation, and 10 corresponds to 270 degree rotation.

To reduce runtime complexity, this method uses perturbed codewords that do not require IDFT recalculation. That is, selecting a row of the matrix G perturbation codeword generator 80 of Figure 1 with the r rows and kn or 2r rows, the code word c _i which is generated from this matrix obtained easily from the nominal time domain block x In order to derive a perturbation time domain block y _i that can be used. G, H ^T, H ^-T is, GH ^T = 0 and _H ^-T H ^T = I 2nx2n
Recall that it must be selected to be Where I _2nx2n is 2nx
This is a 2n unit matrix.

The reduced complexity method can be realized by allowing the matrix G to have three rows. This also corresponds to r = 3 redundant bits. n = 8
The three lines are as follows: 1.11 11 11 11 11 11 11 11 11; corresponds to the sign inversion of x. 2.00 00 00 00 00 00 00 00 00; corresponding to rotation of N / 2 samples of x. 3.00 01 11 10 00 01 11 10; corresponds to the rotation of N / 4 samples of x.

In this example, there are r = 3 redundant bits corresponding to 2 ³ = 8 possible perturbations y _i . Selector 166 selects the best of these eight perturbations to minimize the output of spectrum calculator 164. That is, a null is generated at 2 kHz in the spectrum of the transmitted block y after being deformed by the nonlinearity of the POTS. Note that these perturbations do not change the peak of the transmitted symbol, but can be used to shape the non-linearly deformed spectrum of the transmitted symbol.

The perturbation selector 72 a of FIG. 14 can improve performance by incorporating preemption. With one stage of preemption, the perturbation selector selects a perturbation time domain block y (k), where y (k) is combined with the best option for y (k + 1) to reduce the nulls in spectrum Z _{k + 1} Try to be the deepest. In the Δ stage preemption, a perturbation selector operating on the kth block selects the perturbation time domain block y (k) and combines it with the best option for y (k + 1) −y (k + Δ). Try to minimize the output of the spectrum calculator.

Although the above example has been described with reference to a non-coded system, the invention can also be easily extended to a coded system, for example a trellis coded system, in a manner obvious to a person skilled in the art. .

It should be noted that the present invention can be embodied in software and / or firmware that can be stored on a computer-usable medium such as a computer disk or memory chip. The invention may also take the form of a computer data signal embodied in a carrier such as when the invention is implemented in electronically transmitted software / firmware, such as the Internet.

The present invention may be embodied in other specific forms without departing from its spirit or essential characteristics. The described embodiments are in all respects illustrative only and not restrictive. The scope of the invention is, therefore, indicated by the appended claims rather than by the foregoing description. All changes having the meaning and scope equivalent to the claims shall be included in the scope.

[Brief description of the drawings]

FIG. 1A is a schematic block diagram of a DMT transmitter constructed according to the present invention.

FIG. 1B is a schematic block diagram of an alternative DMT transmitter constructed in accordance with the present invention.

FIG. 2 is an extended signal point constellation according to the present invention.

FIG. 3 is a schematic block diagram of a receiver constructed according to the present invention.

FIG. 4 is a schematic block diagram of the offset coset representative generator of FIGS. 1A and 1B.

FIG. 5 is a schematic block diagram of the perturbation device of FIG. 1A.

FIG. 6 is a schematic block diagram of an effective perturbation generator in the perturbation device of FIG. 5;

FIG. 7 is a schematic block diagram of a perturbation selector in the perturbation device of FIG. 5;

FIG. 8 is a schematic block diagram of an offset decoder of the receiver of FIG. 3;

FIG. 9 is a schematic block diagram of an alternative frame-based structure of the offset coset representative generator of FIGS. 1A and 1B.

FIG. 10A is a schematic block diagram of an alternative frame-based structure of the perturbation device of FIG. 1A.

FIG. 10B is a schematic block diagram of the perturbation device of FIG. 10A incorporating preemption.

FIG. 11 is a schematic block diagram of the effective perturbation generator shown in FIG. 10A.

FIG. 12 is a schematic block diagram of the perturbation selector shown in FIG. 10;

FIG. 13 is a schematic block diagram of an alternative frame-based structure of the offset decoder of FIG.

FIG. 14 is a schematic block diagram of an alternative structure of the perturbation selector of FIG. 5 in a DSL splitterless application.

FIG. 15 is an alternative rotation extension signal point constellation according to the present invention.

──────────────────────────────────────────────────続 き Continuation of front page (81) Designated country EP (AT, BE, CH, CY, DE, DK, ES, FI, FR, GB, GR, IE, IT, LU, MC, NL, PT, SE ), OA (BF, BJ, CF, CG, CI, CM, GA, GN, GW, ML, MR, NE, SN, TD, TG), AP (GH, GM, KE, LS, MW, SD, SL, SZ, UG, ZW), EA (AM, AZ, BY, KG, KZ, MD, RU, TJ, TM), AL, AM, AT, AU, AZ, BA, BB, BG, BR, BY , CA, CH, CN, CU, CZ, DE, DK, EE, ES, FI, GB, GD, GE, GH, GM, HR, HU, ID, IL, IN, IS, JP , KE, KG, KP, KR, KZ, LC, LK, LR, LS, LT, LU, LV, MD, MG, MK, MN, MW, MX, NO, NZ, PL, PT, RO, RU, SD, SE, SG, SI, SK, SL, TJ, TM, TR, TT, UA, UG, UZ, VN, YU, ZW ) Inventor Aradana Narla, USA 02166 Auburndale Grove Street No. 7 290 (72) Inventor Frank Robert Kushishang El 5 Jay 1 Em 6 Mississauga Heartland Drive 1998 F-term Ontario, Canada (Reference) 5K004 AA05 AA08 FA01 FD02

Claims (9)

[Claims] 1. An apparatus for improving the definition characteristics of a transform domain symbol, comprising: a signal mapping apparatus for mapping input data to a block of symbols in a first domain; and an offset corresponding to the block of the symbol in the first domain. A perturbation / conversion device that generates blocks of perturbation conversion domain symbols in response to the bits and improves the defined properties of the conversion symbols. 2. The inverting and transforming device, in response to the signal mapping device, transforming each block of symbols in the first domain into a block of transform domain symbols in the first domain; 2. A perturbation device for selectively perturbing each block of a transform domain symbol in response to a corresponding offset bit and forming the block of the perturbation transform domain symbol. Equipment. 3. The perturbation / transformation device includes: a perturbation device for selectively perturbing each block of symbols to form a block of perturbation symbols in response to the signal mapping device; and in response to the perturbation device. 2. The apparatus of claim 1, further comprising: an inversion converter that converts the block of perturbation symbols into the block of a perturbation conversion area in the first area. 4. A receiver for receiving and decoding symbols formed by the perturbation / conversion device according to claim 1. 5. The apparatus according to claim 1, further characterized by a selector that selects a block of perturbation transform domain symbols that satisfy a predetermined criterion. 6. The apparatus of claim 5, wherein said predetermined criterion is a minimum maximum average power ratio. 7. The apparatus of claim 5, wherein the predetermined criterion is a minimum amount of non-linear interference in a voice band. 8. The apparatus of claim 5, further characterized by a transmitter transmitting said selected block of perturbed transform domain symbols. 9. A conversion device for converting the received block of perturbation symbols into frequency domain symbols; an offset extraction device for extracting base symbols and perturbation offset bits, and converting the base symbols into base information bits; Offset that converts perturbation offset bits into offset information bits
A receiver for receiving a block of perturbed symbols characterized by a decoder;