EA001872B1 - System and method for spectrally shaping transmitted data signals - Google Patents

Abstract

1. A system for transmitting from a transmitter, on a per frame basis, digital information bits which are encoded into a predefined number of signed symbols per frame for transmission over a network to a receiver, wherein the transmitted signed symbols have a desired spectral shape; the digital information bits being divided into a first predetermined number of magnitude information bits and a second predetermined number of sign information bits per frame, the transmitter comprising: magnitude mapping device for mapping the magnitude information bits to the predefined number of symbol magnitudes per frame; a sign bit encoder for encoding the sign information bits into the predefined number of encoded symbol sign bits per frame; and a signal point selector, responsive to the magnitude mapping device and the sign bit encoder, which combines the symbol magnitudes and the encoded symbol sign bits to form the predefined number of transmitted signed symbols per frame; the sign bit encoder comprising: a coset representative generator which generates for each frame, coset representative sign bits for the sign information bits, defining a coset representative element for a convolutional code which identifies a coset of the convolutional code containing candidates of encoded symbol sign bits; and a symbol sign bit selector, responsive to the coset representative sign bits and the symbol magnitudes, which selects the encoded symbol sign bits from the candidates of encoded symbol sign bits that produce the transmitted signed symbols with the desired spectral shape. 2. The system of claim 1 wherein the first predetermined number of magnitude information bits is (r-n) bits and the second predetermined number of sign information bits is (n-r) bits, where n corresponds to the predefined number of symbols per frame and r corresponds to a number of redundancy bits used by the sign bit encoder. 3. The system of claim 2 wherein the coset representative generator includes a differential encoder to differentially encode predetermined bit positions of the (n-r) bits provided to the sign bit encoder to achieve polarity inversion invariance. 4. The system of claim 3 wherein the magnitude mapper maps the m magnitude information bits to n symbols per frame using a modulus conversion mapping scheme. 5. The system of claim 4 wherein the coset representative generator further includes a matrix block, which multiplies the (n-r) sign information bits by a matrix, H<-T>, to produce the n coset representative sign bits per frame. 6. The system of claim 5 wherein the symbol sign bit selector includes a selection controller and a filter; wherein the selection controller includes: logic for generating from the coset representative sign bits candidates of encoded symbol sign bits; logic for combing the candidates of encoded symbol sign bits with the symbol magnitudes to form encoded signed symbol candidates; wherein the filter includes: logic, responsive to the encoded signed symbol candidates, for determining the (RFS) for each of the encoded signed symbol candidates and logic for providing the RFS for each of the encoded signed symbol candidates to the selection controller; and wherein the selection controller further includes: logic, responsive to the determined RFS for each of the encoded signed symbol candidates, for choosing the encoded symbol sign bits associated with the encoded signed symbol candidate with the minimum determined RFS. 7. The system of claim 5 wherein the symbol sign bit selector utilizes lookahead and includes a selection controller and a filter; wherein the selection controller includes: logic for generating from the coset representative sign bits candidate sequences of encoded symbol sign bits; logic for combing the candidate sequences with the symbol magnitudes to form encoded signed symbol candidate sequences; wherein the filter includes: logic, response to the encoded signed symbol candidate sequences, for determining the RFS for each of the encoded signed symbol candidate sequences and logic for providing the RFS for each of the encoded signed symbol candidate sequences to the selection controller; and wherein the selection controller further includes; logic, responsive to the determined HFS for each of the encoded signed symbol candidate sequences, for choosing the encoded symbol sign bits from the encoded signed symbol candidate with the minimum determined RFS. 8. The system of claim 7 wherein the RFS for the j-th symbol is determined by the filter as follows: where NA - the first number of coefficients, NB - the first number of coefficients, y - is a signal symbol, a and b real numbers selected so as to have a desired spectral shape. 9. The system of claim 8 wherein the RFS is determined on a frame basis for the k-th frame by the filter as follows: where j is the symbol (time) index. 10. The system of claim 9 wherein the RFS is determined using look-ahead as follows: where Δ is the look-ahead depth. 11. The system of claim 10 wherein the receiver includes a magnitude and sign extractor which separates the transmitted signed symbols into the encoded symbol sign bits and symbol magnitudes and a sign bit decoder which decodes the encoded symbol sign bits into the sign information bits. 12. The system of claim 11 wherein the sign bit decoder includes a matrix block having a matrix H<T> by which the encoded symbol sign bits are multiplied to recover the sign information bits. 13. The system of claim 12 further including a differential decoder which differentially decodes the predefined bit positions of the recovered sign information bits. 14, The system of claim 13 further including an octet converter, responsive to the signal point selector, for transmitting over the network octets corresponding to the signed symbols. 15. The system of claim 13 wherein n is equal to six, r is equal to one and matrix H<-T> is defined as follows: where D is frame delay. 16. The system of claim 15 wherein convolutional code is defined as follows: G(D) = [1+D 1 1+D 1 1+D 1]. 17. The system of claim 16 wherein the matrix H' is defined as follows: 18 The system of claim 7 wherein the RFS is the running digital sum (RDS) and the RDS for the i-th simbol is determined by the filter as follows: where j is the symbol (time) index. 19. The system of claim 18 wherein the RDS is determined on a frame basis for the k-th frame by the equivalence class selector as follows: where j is the symbol (time) index. 20. The system of claim 19 wherein the RDS is determined using lookahead as follows: where Δ is the look-ahead depth. 21. A method for transmitting from a transmitter, on a per frame basis, digital information bits which are encoded into a predefined number of signed symbols per frame for transmission over a network to a receiver, the symbols having a desired spectral shape, the digital information bits being divided into a first predetermined number of magnitude information bits and a second predetermined number of sign information bits per frame, the method comprising: mapping the magnitude information bits to the predefined number of symbol magnitudes per frame: encoding the sign information bits into the predefined number of encoded symbol sign bits per frame; and combining the symbol magnitudes and the encoded symbol sign bits to form the predefined number of transmitted signed symbols per frame; the step of encoding including: generating for each frame coset representative sign bits for the sign information bits, defining a coset representative element for a convolutional code which identifies a coset of the convolutional code containing candidates of encoded symbol sign bits; and selecting, using the coset representative sign bits and the symbol magnitudes, the encoded symbol sign bits from the candidates of encoded symbol sign bits that produce the transmitted signed symbols with the desired spectral shape. 22. The method of claim 21 wherein the first predetermined number of magnitude information bits is m bits and the second predetermined number of sign information bits is (n-r) bits, where n corresponds to the predefined number of symbols per frame and r corresponds to a number of redundancy bits used in the encoding step. 23. The method of claim 22 wherein the step of generating includes differentially encoding predetermined bit positions of the (n-r) bits to achieve polarity inversion invariance. 24. The method of claim 23 wherein the step of mapping maps the m magnitude Information bits to n symbols per frame using a modulus conversion mapping scheme. 25. The method of claim 24 wherein the step of generating further includes multiplying the (n-r) sign information bits by a matrix, H<-T>, to produce the n coset representative sign per frame. 26. The method of claim 25 wherein the step of selecting includes: generating from the coset representative sign bits candidates of encoded symbol sign bits; combing the candidates of encoded symbol sign bits with the symbol magnitudes to form encoded signed symbol candidates; determining the RFS for each of the encoded signed symbol candidates and logic for providing the RFS for each of the encoded signed symbol candidates to the selection controller, and choosing the encoded symbol sign bits associated with the encoded signed symbol candidate with the minimum determined RFS. 27. The method of claim 25 wherein the step of selecting utilizes lookahead and includes: generating from the coset representative sign bits candidate sequences of encoded symbol sign bits; combing the candidate sequences with the symbol magnitudes to form encoded signed symbol candidate sequences; determining the FlFS for each of the encoded signed symbol candidate sequences and logic for providing the RFS for each of the encoded signed symbol candidate sequences to the selection controller; and choosing the encoded symbol sign bits from the encoded signed symbol candidate with the minimum deter

Description

This application is associated with the following US patent applications, all of which belong to the applicant of this application, and all of which are used here by reference:

Provisional application for US patent No. 60/042826, entitled Generalized formation of the spectrum, filed April 8, 1997 by the inventors M. Us6a1 Eyuododi, P1etge A. Nitye !, Eeuoipd K1t and Όανίά Tipd; this application is based on this application for invention, and priority is claimed on a common subject; U.S. Patent Application No. 08/747840, entitled Device, System, and Method for Forming a Spectrum of Transmitted Data Signals, filed November 13, 1996 by the inventors Ue6a1, EuBod1i and P1etge A. Thread1 ;; and U.S. Patent Application, Attorney Case Number CX096044P02, entitled Device and Method for Precoding Data Signals, filed December 29, 1997 by the inventors M.Ue6a1 Eyuododi, P1etge A. Thread1e! and Eeeuoopt Ut.

The scope of the invention

This invention relates to high-speed data transmission and, more specifically, to a system and method for forming a spectrum of transmitted data signals.

BACKGROUND OF THE INVENTION

The public switched telephone network (Ρ8ΤΝ) contains a digital backbone network and analog subscriber circuits by which end users connect to the backbone. In a normal telephone call, an analog signal from a local user is digitized at a local telephone exchange and converted into a 64 kbit / s data stream, which is transmitted over a digital trunk network and then converted back to analog at a remote telephone exchange for transmission to an end user via a remote subscriber circuit. Modems for dial-up transmission, such as ν.34 modems, transmit data through Ρ8ΤΝ by modulating digital information into an analog signal for subsequent transmission. The process of digital-to-analog conversion at the point of entry into the digital highway introduces quantization noise, which limits the data rate to about 30 kbit / s.

Techniques have been developed that provide data transfer at significantly higher speeds than 30 kbit / s, potentially up to 56 kbit / s, when one of the users has a direct connection to a digital network, for example, Ι8ΌΝ or T1. Moreover, for the case of such data transfer, there is a standardized protocol, standard ν.90 of the International Telecommunication Union (1TI), which is expected to be approved in the near future. According to this technique, random digital information is encoded in the form of octets according to g law or A-law (depending on the region of the world) by a digital pulse-code modulation (PCM) modem using a channel encoder. Octets are mapped directly to characters in a digital-to-analog converter (DAC) located at the end-user telephone exchange (unless otherwise indicated, all of the descriptions below apply to the law; their distribution to the A-law is obvious). The mapping could use all or any subset of the 255 DAC levels, depending on regulatory constraints on average energy.

Since information is transmitted over the digital network in the form of octets, the encoded data is first converted to octets for transmission at a speed of 8000 octets per second. Then, at the end-user telephone exchange, the octets are converted into DACs into the corresponding characters. The resulting 8 kHz sequence of characters is passed through a low-pass filter (low-pass filter) and transmitted over the analog circuit to the analog PCM modem of the end user. The DAC output can be considered as a sequence of pulses whose amplitudes correspond to one of the D / A conversion levels. An analog PCM modem restores the original information, first establishing which symbols were transmitted, and then performing a reverse mapping of these symbols in order to obtain an estimate of the initial digital information.

When the information is transmitted randomly, the spectral analysis of the signal after the D / A conversion shows that the spectrum of the sequence leaving the DAC is essentially flat. Therefore, when this sequence passes through the low-pass filter to the telephone exchange, the signal spectrum takes the form of the low-pass filter. Unfortunately, this spectrum contains a significant amount of energy near the frequency of the constant component (ES) (frequency 1 = 0), which can put the transformers in the system into saturation mode and introduce unwanted non-linear distortion into the transmitted signal. For the applications under consideration, this type of distortion is unacceptable and, therefore, there is a need to eliminate it.

In a more general case, when using PCM, there is a need for a method that would ensure the formation of the spectrum of the signal transmitted from the DAC. In addition, there is a need for a spectrum shaping method that is applicable, in addition to PCM, in various types of data transmission technologies.

Brief Description of the Drawings

In FIG. 1 is a simplified schematic diagram of a typical telephone company telephone exchange;

in FIG. 2 - a graph of the frequency spectrum _{to the} symbols have issued from inverter gzakona in linear form as shown in FIG. 1, and the spectrum shape of the low-pass filter shown in FIG. one;

in FIG. 3 is a graph of parts of two frequency spectra, each of which has zero at a constant component, while one of the spectra decreases to zero with a constant component very steeply, and the second decreases to zero more smoothly;

in FIG. 4 is a schematic illustration of a transmitter of a digital PCM modem of a central node made in accordance with this invention;

in FIG. 5 is a schematic illustration of a receiver of an analog PCM modem of an end user made in accordance with this invention;

in FIG. 6 is a schematic representation of the sign bit encoder of the transmitter shown in FIG. 4;

in FIG. 7 is a schematic representation of an adjacent class presentation generator shown in FIG. 6;

in FIG. 8 is a schematic representation of a symbol bit selector shown in FIG. 6;

in FIG. 9 is a trellis diagram representing a convolutional code;

in FIG. 10 is a block diagram illustrating the generalized logic of the symbol sign bit selector of FIG. 8;

in FIG. 11 is a schematic representation of the symbol bit decoder of FIG. 5, and in FIG. 12 is a schematic illustration of the present invention used as a precoder in a PCM transmitter of a subscriber.

Description of Preferred Implementation

The present invention includes a system and method for forming a spectrum of transmitted data signals that are applicable to various data transmission technologies. For purposes of clarity, the invention is described based on a PCM data transmission system. However, one of ordinary skill in the art will appreciate that the invention can be extended to other data transfer technologies, and that the PCM implementation described here can easily be extended to these technologies.

FIG. 1 and 2 illustrate the presence of energy in the DC region in signals transmitted over an analog circuit to an end-user analog PCM modem. In FIG. Figure 1 shows a part of a typical central telephone exchange 10 Ρ 8ΤΝ, which receives 12-g-law input octets transmitted by a digital PCM modem of a central node (not shown) connected directly to the digital part of the telephone system.

The octets are converted using the DAC 14, also called a converter from gzakon to linear form, in the sequence of characters u _to . Each of the symbols corresponds to one of the 255 levels of g-law. Symbols are provided on line 16 to a low-pass filter (LPF) 18, which provides the filtered analog signal 8 (1) to the receiver of the analog PCM modem of the end user through the analog circuit 20. The analog signal is demodulated and decoded by the receiving modem, which provides a digital bit stream. This digital bitstream is an estimate of the originally transmitted data.

The sequence of symbols y _k in line 16 from the converter from g-law to linear form 14 has a flat frequency response 22, FIG. 2. The shape of spectrum 24 of the low-pass filter 18 contains a significant amount of energy near the constant component (1 = 0), as shown at point 26. Since the sequence yk has a flat frequency response, the spectrum of signal 8 (1) produced by filter 18 has the same shape 24 as with filter 18, and therefore, signal 8 (1) also contains a significant amount of energy near the DC component. As mentioned above, this energy near the DC component saturates the system transformers, which introduce unwanted nonlinear distortion into signal 8 (1) transmitted to the receiving modem.

In applications like PCM, this distortion should be reduced. This can be achieved by reducing the energy of the transmitted signal near the DC component to zero. Such a zero of the constant component 28 is shown in FIG. 3. As is known from the prior art, in order to obtain a zero signal in the spectrum near the constant component in the transmitted signal, the current difference between the number of zeros and units (QO 8 Ktsiit Οίβίΐαΐ 8it) of the transmitted symbols y _k (namely, the algebraic sum of all previously transmitted levels) must be close to zero. The shape of the spectrum near zero with a constant component 28 can vary from a relatively smoothly decreasing spectrum 30 to a spectrum 32 that sharply decreases near the constant component. The acuity of zero depends on the accuracy of the control of KO8.

As shown below, the present invention encodes transmitted digital data such that KO8 is kept close to zero. This provides a spectral zero with a constant component, thus reducing the nonlinear distortion caused by the saturation of the transformer. More generally, the invention can also be used to encode transmitted digital data in order to form a spectrum of the transmitted signal in the desired manner.

The transmitter 40 (Fig. 4) in a digital PCM modem receives a serial digital bit stream from a terminal (not shown), such as a personal computer, and encodes the received bits into octets 44 for transmission over digital network 46. The serial bit stream 42 is converted to parallel format a serial-parallel converter 48. The transmission / coding scheme of the present invention is based on an η-character data frame, where k denotes the index of the (temporary) data frame. For example, 2, 3, 4, 5, or 6 characters may be transmitted in each data frame. The transmitted symbols correspond to the constellation points of the g-law, selected to represent the information bits. For each data frame, the serial-to-parallel converter 48 outputs (n-g) + t information bits, where g is the number of redundant bits. The number of redundant bits, as stipulated by the U.90 standard, may be 0, 1, 2 or

3.

It should be noted that in the rest of the description, lowercase letters are used to denote scalar quantities, and uppercase letters denote matrices. In addition, row vectors are indicated by lowercase letters and are shown in bold, and all indices begin with 0, for example, x _k = (x _k0 , x _k1 ,).

(η-d) bits denoted as y _k represent information transmitted using signed bits (sign information bits), and t bits denoted like _k represent information transmitted using modular values (meaning information bits). The number of bits m can be chosen so as to satisfy the following relationship:

p-1 where Μΐ is the number of positive points of the constellation for the ιth character of the data frame. This process is more fully described in the U.90 standard.

m significant information bits, and _k , are transmitted to a significant bit mapping unit 50, which maps t bits to η symbol values q _to using a mapping algorithm such as a shell mapping (spherical mapping) described in standard ΙΤϋ U.34, or residual mapping described in standard ΙΤϋ U.90. The values to which significant information bits are mapped are the values of the z-law points used as constellation points when transmitting information bits. These mapping algorithms and the process of selecting constellation points are understandable to specialists in this field, and their further description is omitted. The remaining information bits of the data frame marks stems from the information bits _to served sign bits to encoder 52 which generates sign bits 8 η _k (iconic symbols coded bits) according to the following detailed description. η values of the symbols d _k and η of the sign bits 8 _k are supplied to the selector of the signal point 54 and combined to form η sign characters (i.e., symbols with a sign) at _k . η iconic symbols _yk are then fed to the transmitter 56 octets, octet which selects corresponding to each iconic symbols, and transmits these octets in the digital network 46. In other data transfer technologies octets converter which converts symbolic characters in a format compatible with the digital part Ρ8ΤΝ , may not be used, and the signal point selector provides sign characters directly to the network.

Octets 44 'coming out of digital network 46 (possibly distorted by digital noise on the network) are received by the telephone exchange (TS) 60. The octets 44' in TS 60 are converted to characters using the DAC and transmitted in the form of an 8 kHz sequence of levels along an analog circuit 62 to the receiver 64 of the analog PCM modem of the end user. The analog levels are received by the input stage of the receiver 66, which digitizes the analog levels, performs synchronization restoration, alignment, and symbol extraction.

The input stage of the receiver 66 outputs the received symbols yk in serial format to the serial-parallel converter 68, which converts the serial symbols into frames of η parallel symbolic symbols yk. n parallel iconic symbols yk are supplied to the block extraction module (symbol values) and a sign 70 which extracts, from y _to values of symbols d _k and sign bits _to 8. The values of the symbols q _to are fed into the block for the inverse mapping of the significant bits 72, for example, the block of the inverse residual mapping, in order to restore the significant information bits and _k . Since the reverse mapping process is obvious to a person skilled in the art, it is not described here. The sign bits 8 _k are supplied to the sign bit decoder 74 in order to recover the sign information bits y _k , as described below. The decoded information bits may be further processed and transmitted to a terminal, for example, to a personal computer.

Sign Bit Encoding

The character encoding bits 52 are depicted in more detail in FIG. 6. The bits in _a character information fed to the generator representations coset 80 which generates for each frame n representing smezh7 ny class 1 _{to the} sign bits, and supplies them to the selector iconic symbols representing bits 82. η coset sign bits _to 1 for each the frames define an element representing an adjacent class for a given convolutional code Ο (Ό), which is used by the symbol bit selector of characters 82, and the whole sequence representing an adjacent class of symbol bits ΐ (Ό) together determines tavlenie coset of the convolutional code. η representing an adjacent class of sign bits 1 _k also defines an adjacent class of convolutional code that contains candidate encoded symbol sign bits of characters, as described in detail below.

Using η representing the adjacent class of 1 _k sign bits, the character sign bit selector 82 modifies the adjacent class of 1 _k sign bits by applying the EXCLUSIVE OR operator over these bits and sequences of valid convolutional codes defined by the trellis diagram (trellis diagram) shown in FIG. 9 and described below, with the aim of generating candidate encoded symbol sign bits. These candidates are elements of an adjacent class defined by the sign bits representing the adjacent class. With iconic symbols symbol selector 82 selects the bit values for each frame of the candidate _{to the} candidate of the coded bits iconic symbol which provides the desired spectral shape, and supplies the sign bits in a signal point selector 54, FIG. 4. The output of the symbol sign bit selector 82 for the entire sequence, alternatively to frame-by-frame output, can be represented as δ (Ό) = ΐ (Ό) Θο (Ό), where δ (Ό) is the sequence of encoded symbol sign bits, ΐ (Ό ) is a representation of an adjacent class for a convolutional code, and e (E) are code sequences that are elements of a convolutional code 0 (0).

It should be noted that in this selection process, any of the candidate encoded symbolic sign bits of the symbols can be used, and they can be decoded, as described below, into encoded signed information bits of _k . Thus, the proposed method of spectrum formation does not affect the values of the characters and therefore does not affect the transmitted power. As a result, it is easy to design a system that satisfies the restrictions on transmitted power imposed by the ECC, and at the same time provides the necessary shape of the spectrum.

An adjacency class generator 80 is depicted in more detail in FIG. 7; it contains a differential coding unit 84 and a matrix unit 86. Noise in the data channel can cause a polarity reversal when it affects the transmitted signed bits. By enabling differential encoding with the differential encoding unit 84, and decoding with the differential decoding unit 132 (FIG. 11), the sign bits at certain bit positions, for example, even positions 0, 2 and 4 for N ^t , N- ^t and Ο ( Ό), below, it is possible to achieve invariance to a change in polarity. The differentially encoded sign information bits yk 'are multiplied (modulo 2) (i.e., filtered) in the matrix block 84 by the matrix H- ^T (p-G) _xn in order to obtain η representing an adjacent class of 1 _k sign bits that are transmitted to character symbol bit selector 82.

An example of this matrix is for the case when six characters and one redundant bit are transmitted in each frame, as defined according to the standard! TI Υ.90, is given below:

'0 o

oh oh

0 0 0 1 / (1 + 2))

LLC 1 / (1 + 2)) О

0 1 / (1 + 2)) 0 O

1 / (1 + 2)) 0 00

1 / (1 + 2)) 0 0 00 (2) where Ό is the frame delay, which is the delay determined on the basis of the (temporary) frame index k.

As shown in FIG. 8, the symbol sign bit selector 82 comprises a selection controller 88, which receives n representing the adjacent class of sign bits 1 _to each frame from the representation generator of the adjacent class 80 and η symbol values from the display unit of the significant bits 50 (FIG. 4) and provides encoded sign bits 8 _k characters for each frame. The selection controller 88 combines the candidate encoded symbolic bits of the characters with the values of the characters to form the candidate encoded symbolic characters that are supplied to the filter 90. The filter 90 calculates for each candidate a metric called here the current filter sum (KE8 - Kipspd Yeshe 8ish), described below, and transmits the calculation results to the selection controller 88, which selects the encoded signed character bits corresponding to the candidate encoded signed character that minimizes KE8. The operation of the symbol bit selector 82 is described below using FIG. 9 and 10.

Selection controller 88 modifies n representing an adjacent class of 1 _k sign bits for each frame by performing an EXCLUSIVE OR operation on an adjacent class of sign bits and valid convolutional code sequences. A convolutional code is the set of possible sequences determined by a trellis diagram, and valid code sequences are sequences that do not violate the constraints of the trellis diagram. For reasons of clarity, selection controller 88 will use one redundant bit g and a convolutional code Ο (Ό) = [1 + Ό 1 1 + Ό 1 1 + Ό 1]. To present this in terms of a trellis diagram, it will be necessary to use a trellis diagram with two states, such as trellis diagram 100 (FIG. 9). The limitations of the trellis diagram are defined as follows.

For a given frame k, the selection controller 88 modifies η representing the adjacent class of sign bits 1 _k by performing an EXCLUSIVE OR operation on them and certain convolutional code sequences in accordance with the restrictions of the trellis diagram 100. The convolutional code sequences for this example are as follows:

A: 000000 (i.e. do nothing)

B: 111111 (i.e., invert all the sign bits of the frame.))

C: 101010 (i.e., invert all odd signed bits of the frame.))

Ό: 010101 (i.e., invert all even signed bits of the frame.))

Thus, at the beginning of the frame k, if the state is O |. _: selection controller 88 is 0, valid sequences in frame k are only convolutional code sequences A 102 and B 104. Otherwise, if state O _1: selection controller 88 is 1, valid sequences in frame k are only convolutional code sequences C 106 and Ό 108. As described above, the sign bits representing the adjacent class undergo EXCLUSIVE OR operations on them and each valid convolutional code sequence, thus forming candidate encoded sign s symbol bits, e.g., (1: _1:.. ®A 1 _to ® B} Each candidate is also an element of the coset convolutional code defined representing a coset sign bits (or representation element coset) Then, each of the candidates to be combined. symbol values in order to generate candidate encoded symbolic characters that are supplied to filter 90 (FIG. 8), where KE8 is computed for each candidate and returned to selection controller 88. Selection controller 88 outputs encoded signed bits of the symbol for a frame f that minimize KE8.

The current state Οι, together with the convolutional code sequence of the encoded sign bits of the characters selected for frame k, are used to determine the next state, taking into account the constraints of trellis diagram 100. For example, if candidate 1 is selected for frame k ₍ ®В, state of the selection controller 88 in the beginning of the Oy frame is 1.

Spectrum formation obtained by the present invention can be improved by the introduction of lead. Namely, instead of selecting encoded symbol sign bits of symbols based only on the current frame, the symbol sign bit selector 82, for deciding which encoded symbol sign bits provide optimal spectrum shaping, can use the symbol values generated by the significant bit mapping 50 (FIG. 4), and the sign bits representing the adjacent class, for the current frame and for future frames. The ν.90 standard stipulates that up to three future frames can be used depending on the amount of anticipatory delay set during initialization.

BE8 spectrum formation metric based on the transfer function of the filter 1ι (Ό). is computed for all possible routes (or candidate sequences) of the trellis diagram to a pre-emptive delay or depth Δ by the filter 90, and the selection controller selects the encoded sign bits of the characters corresponding to the candidate sequence for frame k, which provides the smallest value of KE8.

Let us again turn to the trellis diagram 100 (Fig. 9), which shows the candidate sequences for lead 1. At the beginning of frame k, if the state of selection controller 88 is 0, only sequences of convolutional code A 102 and B 104 are valid for frame k. However, the code sequences for frame k + 1 should also be taken into account. Since the code sequences A 102 and B 104 are valid for frame k, the state Oy can be 0 or 1 for frame k + 1, and therefore the code sequences A 102 ', B 104', C 106 ', and Ό 108' are valid . As described above, the adjacent sign bits representing the adjacent class undergo EXCLUSIVE OR operations on them and a valid convolutional code sequence to form candidate encoded signed sign bits of the characters. When using the lead, the sign bits representing the adjacent class for each frame k and k + 1 undergo EXCLUSIVE OR operations on them and valid code sequences for each of the possible routes of the trellis diagram, thus forming candidate sequences. The candidate sequences for this example are the following four sequences: {1) {1 _to ®A, 1 _to + 1®A}; 2) {1 _to ®A, ΐ _Μ ®Β}; 3) {1 _k ®B,

1 _{to +} 1®С} and 4) {1 _to ®В, 1 _{to + 1} ®И}}. The value of КТ8 is determined for each sequence and candidate encoded symbolic sign bits of symbols for frame k are selected.

The operation of the character symbol bit selector 82 is described using flowchart 120 of FIG. 10. At step 122, the selection controller 88 generates candidate encoded symbol sign bits (or candidate sequences if anticipated) by modifying representing an adjacent class of sign bits in accordance with the trellis diagram. Then, in step 124, the selection controller combines the symbol values and the candidate encoded signed character bits to form candidate encoded signed characters (or candidate sequences) and transmits them to the filter 90. At step 126, the filter 90 determines CT8 for each candidate (or candidate sequence) and transmits CT8 for each candidate (or candidate sequence) to the selection controller 88. And finally, at step 128, the selection controller 88 selects the candidate encoded symbol sign bits (or didactic sequences) that minimize CT8, and sends the selected encoded character bits of the characters.

It should be noted that the present invention may use various convolutional codes Ο (Ό), which are represented using various trellis diagrams and various convolutional code sequences. The extension of various convolutional codes and code sequences in the light of the description given here can be easily performed by a person of ordinary skill in the art.

In general, using PCM, the spectrum forming method according to this invention generates a spectrum of an analog signal transmitted from the DAC to the TC 60 (Fig. 5) by setting the filter characteristic 90 (Fig. 8) in order to achieve the desired spectrum shape and by minimizing CT8. The characteristic H (T)) of the filter 90, which determines the desired shape of the spectrum, can be described as follows:

1 / Κΰ) = Β (Ώ) / Αψ) = ^ ----, α ₀ = 1, (3) / = 0 where Α (Ό) and Β (Ό) are some functions, and a and b are real numbers selected to achieve the desired shape of the spectrum. Ν _Α and Ν _Β determine the number of coefficients used in the numerator and denominator, respectively, to represent Η (Ό). CT8 can be calculated character by character as follows:

and KT8 for each frame on the example of frame k is calculated as

where) is the index of the (temporary) symbol.

When using the sign bit encoder of this invention to provide spectral zero with a constant component, CT8 represents the current difference between the number of zeros and ones in the signal (ΚΌ8) and the characteristic Η (Ό) of filter 90 is expressed as th (L) = 1-2) ( 6)

On the basis of the transmitted iconic characters have _to filter 90 calculates ΚΌ8 transmitted iconic characters have _to at the time as a symbol ι ΐ

= Σ ^ ·, (7) ₇ · = 0 where | is the index of the (temporary) symbol, and ΚΌ8 for each frame on the example of frame k is calculated as

where | - index of the (temporary) symbol.

For lead in the lead depth Δ ΚΌ8 is calculated as follows:

Δ (?) / = 0 where LBCO is the prefaced ΚΌ8. In the same way, one can introduce a lead for the generalized minimization of KT8 as

Δ (9) / = 0

Sign bit decoding

The sign decoder 74 at receiver 64 (FIG. 5) shown in FIG. 11, comprises a matrix unit 110. The matrix unit 110 sign bits of _a multiplied (mod 2) (i.e., filtered out) by the matrix H ^t _{nx (nT)} to restore differentially encoded sign information bits Che '.

An example of the matrix H ^t for the case of the transmission of six characters in the frame (n = 6) and one redundant bit (g = 1) is given in formula (11):

'one 1 + Ζ> one 1 + £> one ' 0 0 0 0 1 + Ζ> t 0 0 0 1 + Ζ> 0 n ¹ = (eleven) 0 0 1 + Ζ) 0 0 0 1 + Ζ> 0 0 0 1 + Ζ) 0 0 0 0

The matrix H ^{t is} designed so that the resolution error y _k due to an error in the received sign signal z _k does not extend to more than one frame. This is ensured by the fact that H ^t is a matrix of finite impulse response (CIS), and only a single delay takes place. In order to demonstrate how each of the candidate encoded symbol sign bits generated by the symbol bit encoder 52 (Fig. 4) for each transmitted symbol frame is decoded into the same sign information bits, the encoding and decoding process must be mathematically described. Recovered information bits (decoding) can be expressed mathematically as follows:

H = ^ N ^T (12) and the signed bits s _to (coding) can be expressed mathematically as follows:

If the right-hand side of equation (13) instead of _a substitute in the equation (12), we obtain the following equation:

o * =? _k N ~ ^t N ^t + g _k an ^t , (14)

Choosing O, N ^t and N- ^t , you must ensure that the following conditions are met:

1) N ^t N ^-t = 1 (where I is the identity matrix); and 2) CH ^m = 0, and then ^ = y _k will be satisfied regardless of the value of r _k .

In the above example of trellis diagram 100 (FIG. 9), there is one redundant bit g _k , and a convolutional code Ο (Ώ) = [1 + Ώ 1 1 + Ώ 1 1 + Ώ 1]. Since 1t _k = g _k and Eg _k = g _k-1, y _to G (e) is equivalent _to g (1 1 1 1 1 1) + _{rk-1 (1 0 1 0 1} 0) . Here r _{k − 1} represents the state 0 _{k of the} trellis diagram, and r _k represents the branches or routes of the trellis diagram. Four sequences of convolutional code A-1) can be displayed in g _to , g _to-1 as follows:

A: 000000 - Г £ _1 ⁼ 0, / - ^ = 0

B: 111111 - / - ^ = 1, r _A = 0

C: 101010 - ^ .1 = 0, r _k = 1 ϋ: 010101 - r _k _! = 1, r _k = \ where the code sequences A-1) can be considered as r _k O (E).

Since the value of r _k does not affect the decoding of information bits, each set of η sign bits generated by different valid code sequences can be used to obtain the same decoding information. As a result, to perform the desired spectrum formation, a set of η sign bits can be selected that minimizes выполнения8 / формирования8.

PCM transmission from the subscriber

The spectrum forming method in accordance with this invention can also be used to correct an analog PCM modem used for PCM transmission from a subscriber in a transmitter to perform precoding. In this case, the characteristic Ε (Ώ) is the characteristic of the channel between the transmitting modem and the ADC in the line card of the telephone exchange (TC) and usually reflects the influence of filtering in the input stage of the transmitting modem, the analog circuit, and the circuit board of the TC.

Using the principles of this invention, the output sequence of the channel x (n) (ζ (η) with prefiltration) can be generated so that at the input of the ADC a sequence y (E) is obtained, the signal points of which simulate the levels of A / D conversion. In this case, the goal is to minimize the energy of the transmitted signal x (P) = y (O) / d (P) while maintaining a small spread in the constellations at the ADC input. The spread in the constellations in this case is also undesirable because a larger spread in the constellations can lead to an increase in the echo of quantization noise and other impairments.

In this application, the characteristic of the channel Ε (Ώ) will usually be determined either by the receiving modem, or by jointly receiving and transmitting modems, based on the measurements in the channel made during initialization (start-up) of the modem, and then during the data transfer the transmitting modem will display the incoming bits in the transmitted sequence x (E), which, after passing through the channel, is converted into the output sequence of the channel y (E). The channel characteristic Ε (Ώ) is usually selected as minimally phase, which is easily achieved, for example, by means of additional filtering in the transmitter.

The transmitter 40 '(Fig. 12) is a transmitter in an analog PCM modem that can perform PCM transmission from a subscriber. Transmitter 40 'uses the spectrum shaping method of the present invention for precoding using precoding unit 140, input bitstream 42'. One of the types of precoding of PCM transmission from a subscriber (called one-dimensional precoding, in which precoding is carried out character by character) is described in detail in the US patent application, patent attorney number СХ096044Р02, Device and method for precoding data signals, filed December 29, 1997 M. Ueba! Eeebodi, P1egge A. Niche! and Eeeuoopd K1sh. The precoding unit 140 performs multidimensional precoding, i.e. precoding is performed for each frame as a whole. This implementation differs from univariate encoding, but its concept is similar to the one-dimensional case and additional details can be found in the related application mentioned above.

The precoding unit 140 comprises a series-parallel converter 48 ′, a mapper of significant bits 50 ′, an encoder of the signed bits 52 ′, and a signal point selector 54 ′. These components are configured and function as components with the same numbers in FIG. 4, with only minor changes. For example, the operation of the sign bit encoder is changed to perform the precoding described below, and the signal point selector gives η pre-encoded levels x _k for each frame in accordance with the sign characters y _k instead of the sign characters themselves, η pre-encoded levels x _k are fed to the parallel-serial converter 142, which provides precoded levels in serial form to the prefilter 144. The prefilter 144 filters the levels and provides filtered levels to the DAC 146, which in turn b, transmits the precoded analog levels through the analog channel 148. The channel modifies the precoded levels x _k and outputs the levels ideally corresponding to the sign symbols y _k to a quantizer in the telephone exchange (TS) 150. In other words, the precoding unit selects the precoded levels xk that provide levels corresponding to the desired symbolic symbols yk on the quantizer, taking into account the characteristics of the analog channel 148, or, more precisely, the target characteristics of the channel B (p).

The target characteristic of the channel d (n) is equal to the convolution of the characteristics of the prefilter 144 d (n) and the characteristics of the analog channel 148 s (n), where n denotes the index of the (temporary) symbol, and d (0) = 1. This relationship is reflected as follows:

y (n) = f (0) x (u) + Η (ϊ) χ (η - 1) + ... + λ (Ύ _Α ) χ (η - Ν _Λ ). (fifteen)

Since th (0) must be equal to 1, equation (15) can be simplified as follows:

The value of y (n) at a given time is known and previous values of x (n) are also known. Filter 90 (FIG. 8) calculates the term of equation (16) as CT 8 and passes it to selection controller 88. The previous values of x (n) are determined based on the previous characters y (n) from the known relation x (I) = y (P) fth (P) and are written into the filter 90, and the selection controller 88 operates according to the block diagram 120 (FIG. 10), selecting encoded symbolic bits of the characters minimizing x (p). Instead of sending symbolic characters, y (n) transmits precoded levels.

In addition to the display operations described above, it is also necessary to include echo cancellation modules that separate the two directions of transmission, an interpolation synchronization filter that provides transmission of characters in synchronization with a timer. This interpolation synchronization filter is usually controlled by the synchronization recovery circuitry used in the receiver receiving the data. The transmitter may also contain a linear filter, which is primarily responsible for limiting the transmission bandwidth to about 4 kHz and providing the necessary prefiltration, which would make the overall characteristic of the channel (I) minimally phase.

Further, in a practical system, one of the types of trellis coding can be used to increase noise immunity. For example, the trellis coding methods described in the aforementioned application entitled System, apparatus, and communication method using trellis codes of narrowband signals selected from a fixed set of narrowband signal points filed November 14, 1996, US Patent Application No. 08 / 749040, (Patent Attorney No. CX096050). This application is hereby incorporated by reference in its entirety. The functioning of the system is essentially not affected by trellis coding.

It should be noted that this invention can be implemented in the form of programs or firmware that can be stored on a computer medium, such as a computer disk or memory chip. The invention may take the form of a computer data signal, represented as a carrier wave, as in the case when the invention is implemented in the form of programs / microprograms that are transmitted using electrical signals, for example, via the Internet.

This invention can be implemented in other specific forms without deviating from its essence or basic features. The described implementations can be considered in all respects only as illustrative and not restrictive. Thus, the scope of the invention is indicated by the attached claims, and not the previous description. All changes that are made within the meaning and range within the equivalence of this formula should be considered as being covered by its scope.

Claims (43)

CLAIM 1. System for frame-by-frame transmission from the transmitter of bits of digital information, which are encoded into a predetermined number of sign characters per frame, for transmission over a network to a receiver in which the transmitted sign characters have the desired spectrum shape, said digital information bits being divided into a first one in advance a predetermined number of significant information bits and a second predetermined number of sign information bits per frame, where said transmitter comprises: a significant bit mapping unit for mapping significant information bits to a predetermined number of symbol values per frame; a sign bit encoder for encoding sign information bits in a predetermined number of encoded sign bits of symbols per frame, and a signal point selector that operates in response to the outputs of the significant bits display unit and sign bit encoder, which combines symbol values and encoded sign bits of characters to form a predetermined number of transmitted symbolic characters per frame; and said character bit encoder comprises: an adjacent class representation generator that generates, for each frame, an adjacent class of sign bits for signed information bits, defining an adjacent class element for a convolutional code that defines an adjacent class of convolutional code containing candidate encoded signed character bits, and a character sign bit selector that operates in response to the representative bits representing the adjacent class and the meaning of the characters, which selects from the candidate encoded signed bits the symbol ols such iconic symbols encoded bits that generate iconic symbols transmitted with the desired spectral shape. 2. The system of claim 1, wherein said first predetermined number of significant information bits is t bits, and said second predetermined number of signed information bits is (n-d) bits, where n is said predetermined number of characters in a frame, and g is the number of redundant bits used by the sign bit encoder. 3. The system of claim 2, wherein the adjacent class presentation generator comprises a differential coding unit for differential coding of predetermined bit positions from (η-g) bits supplied to the sign bit encoder to achieve polarity inversion invariance. 4. The system of claim 3, wherein the significant bit mapping unit maps m significant information bits to η characters per frame using a residual mapping algorithm. 5. The system of claim 4, wherein the adjacent class presentation generator also comprises a matrix block that multiplies (η-g) sign information bits by an H- ^t matrix to obtain η representing an adjacent class of sign bits per frame. 6. The system of claim 5, wherein the symbol character bit selector comprises a selection controller and a filter, wherein the selection controller comprises: logic circuits for generating candidate encoded sign symbol bits from representing the adjacent class of sign bits; logic circuits for combining candidate encoded signed character bits with character values to form candidate encoded signed character bits; and where the filter contains: logic circuits that operate in response to candidate encoded symbolic characters to determine a current filter amount (KE8) for each candidate encoded symbolic symbol, and logic circuits for supplying KB8 for each candidate encoded symbolic symbol to a selection controller; and where the selection controller further comprises: logic circuits that operate in response to computed KB8 for each candidate encoded signed symbol to select encoded signed bit bits of characters corresponding to the encoded candidate signed symbol with a minimum computed KB8. 7. The system according to claim 5, in which the selector of the sign bits of the symbols uses the lead and contains a selection controller and a filter, where the selection controller contains: logic circuits for generating candidate sequences of encoded signed symbol bits from representing an adjacent class of sign bits; logic circuits for combining candidate sequences with symbol values in order to form candidate sequences of encoded signed characters; where the filter contains: logic circuits that operate in response to candidate sequences of encoded signed symbols to determine KE8 for each of the candidate sequences of encoded signed symbols, and logic circuits for supplying KB8 for each of the candidate sequences of encoded signed symbols to a selection controller; and where the selection controller further comprises: logic circuits that operate in response to computed KE8 for each of the candidate sequences of encoded symbolic characters, for selecting encoded symbolic bits of symbols corresponding to the encoded candidate symbolic symbol with a minimum computed KB8. 8. The system according to claim 7, in which KB8 for the символаth character is determined by the filter according to the following formula A a 7 = 0 7 = 1 where Ν _α is the first number of coefficients, Ν _{Β is the} first number of coefficients, y is the signal symbol, and b are real numbers chosen so as to provide the desired shape of the spectrum. 9. The system of claim 8, in which KE8 is frameless for the k-th frame is determined by the formula i-1 (5) 7 = 0 where | - index of the (temporary) symbol. 10. The system according to claim 9, in which KE8 is determined using the lead in the following way: Δ LKP8 _k = ^ KE8 _{k ¥ 1} , 7 = 0 where Δ is the lead depth. 11. The system of claim 10, wherein the receiver comprises a symbol and character value extraction unit that separates the transmitted symbol characters into encoded symbol symbol bits and symbol values, and a symbol bit decoder that decodes the encoded symbol symbol bits into information symbol bits. 12. The system of claim 11, wherein the symbol bit decoder comprises a matrix block comprising a matrix H ^t by which the encoded symbol sign bits of the symbols are multiplied to recover the sign information bits. 13. The system of clause 12, which further comprises a differential decoding unit that differentially decodes the predetermined bit positions of the restored sign information bits. 14. The system according to item 13, which further comprises an octet converter, functioning in response to the output of the signal point selector, for transmission over the network of octets corresponding to the symbolic characters. 15. The system according to item 13, in which η is equal to six, r is equal to unity, and the matrix N- ^m is determined as follows: where I) is the delay of the frame. 16. The system of clause 15, in which the convolutional code is defined as follows: C (P) = [1 + P 1! + £ ► 1 1 + Ώ 1]. 17. The system according to clause 16, in which the matrix N ^t is determined as follows: 1 + Ζ) 1 o1 + P 1 + Ζ) oo oo oo 18. The system according to claim 7, in which KP8 is the current difference between the number of zeros and ones in the signal (ΚΏ8), and ΚΏ8 for the 1st character is determined by the filter as follows: = Στ; > 7 = 0 where) is the index of the (temporary) symbol. 19. The system of claim 18, wherein ΚΏ8 is frameless for the k-th frame by the filter as follows: n-1; = 0 where | - index of the (temporary) symbol. 20. The system of claim 19, wherein ΚΏ8 is determined using the lead in the following manner: Δ 7 = о where Δ is the depth of lead. 21. A method for transmitting digital information bits from a transmitter that are encoded into a predetermined number of symbolic characters per frame for transmission over a network to a receiver, wherein the symbols have the desired spectrum shape, said digital information bits are divided by a first predetermined number of significant information bits and a second predetermined number of sign information bits per frame, including: mapping meaningful information bits into a predetermined number of symbol values per frame; encoding the sign information bits into a predetermined number of encoded sign characters bits per frame, and combining the values of the characters and encoded sign bits of the characters to form a predetermined number of transmitted sign characters per frame; and the encoding operation includes: generating, for each frame, an adjacent class of sign bits for signed information bits defining an adjacent class element for a convolutional code that defines an adjacent class of convolutional code containing candidate encoded signed character bits of a symbol, and making selections using the adjacent class of signed bits and character values , of candidate encoded symbolic bits of symbols of such encoded symbolic bits of symbols that generate transmitted symbolic bits e characters with the desired spectrum shape. 22. The method according to item 21, in which said first predetermined number of significant information bits is t bits, and said second predetermined number of sign information bits is (n-d) bits, where n is said predetermined number of characters in a frame , and g is the number of redundant bits used in the encoding operation. 23. The method of claim 22, wherein the generating operation includes differential coding of predetermined bit positions from (η-g) bits to achieve polarity inversion invariance. 24. The method according to item 23, in which the operation of the mapping of significant bits maps w significant information bits in η characters per frame using the residual mapping algorithm. 25. The method according to paragraph 24, in which the generation operation further includes multiplying (η-g) the sign information bits by the matrix H ^_t to obtain η representing an adjacent class of sign bits per frame. 26. The method according A.25, in which the selection operation includes: generating candidate encoded sign symbol bits from representing an adjacent class of signed bits; combining candidate encoded signed character bits with character values to form candidate encoded signed character bits; determining KB 8 for each of the candidate encoded symbolic characters and supplying KB 8 for each of the candidate encoded symbolic characters to the selection controller, and selecting encoded symbolic bits of the symbols corresponding to the encoded candidate symbolic symbol with the minimum calculated KB 8. 27. The method according A.25, in which the selection operation uses the lead and includes: generating candidate sequences of encoded signed symbol bits from representing an adjacent class of sign bits; combining candidate sequences with character values to form candidate sequences of encoded signed characters; determining KB8 for each of the candidate encoded character sequences, and supplying KB8 for each of the candidate encoded sequences of sequences to the selection controller, and selecting encoded character bits of the symbols corresponding to the encoded candidate character with a minimum computed KB of 8. 28. The method according to item 27, in which KB8 for the символа-th character is determined by the following formula Y = o y = 1 29. The method according to p. 28, in which KB8 poframe for the k-th frame is determined by the formula KG8 _k = X ^ 8 ² _{pc +]} 2 = 0 (5) where) is the index of the (temporary) symbol. 30. The method according to clause 29, in which KB8 is determined using the lead in the following way: / = 0 where Δ is the lead depth. ZE The method of claim 30, further comprising, at the receiver, dividing the transmitted symbolic characters into encoded symbolic symbol bits and symbol values, and decoding the encoded symbolic symbol bits into symbolic information bits. 32. The method according to p, in which the decoding operation includes multiplying the encoded symbolic bits of the characters on the matrix N ^t to restore the sign information bits. 33. The method according to p, which further includes differential decoding of predetermined bit positions of the restored sign information bits. 34. The method according to claim 3, which further includes transmitting over the network of octets corresponding to the symbolic characters. 35. The method according to claim 3, in which η is six, g is equal to unity, and the matrix H ^_t is defined as follows: 0 0 0 0 1 / (1 + 2)) 0 0 0 1 / (1 + 2)) 0 0 0 1 / (1 + 2)) 0 0 0 1 / (1 + 2)) 0 0 0 1 / (1 + 2)) 0 0 0 0 36. The method according to clause 35, in which the convolutional code is determined as follows: <? (P) = [1 + 2) 1 1 + Ζ) 1 1 + 2) 1]. 37. The method according to clause 36, in which the matrix N ^t is determined as follows: about 1 + 2) about oh oh 0 1 + 2) 1 + 2) 0 one 1 + 1) one 0 0 1 + 2) 0 1 + 7) 0 1 + 2) 0 0 0 0 0 0 0 0 38. The method according to item 27, in which KB8 is the current difference between the number of zeros and ones in the signal (КЕ) 8), and КЕ) 8 for the символа-th character is determined by the filter as follows: 7 = 0 where) is the index of the (temporary) symbol. 39. The method according to § 38, in which KE) 8 is per-frame for the k-th frame is determined by the filter as follows: 2 ^ = ^ 2) 5 ^, ₇ = 0 where) is the index of the (temporary) symbol. 40. The method according to § 39, in which ΒΌ8 is determined using the lead in the following way: Δ tick, 7 = 0 where Δ is the lead depth. 41. A frame-free precoding system in a digital information bit transmitter for a predetermined number of pre-encoded levels per frame, said digital information bits being divided into a first predetermined number of significant information bits and a second predetermined number of signed information bits per frame, which comprises: a significant bit mapping unit for mapping significant information bits to a predetermined number of symbol values per frame; a sign bit encoder for encoding the sign information bits into a predetermined number of encoded sign bits of symbols per frame, and a signal point selector that operates in response to the outputs of the significant bit display unit and the sign bit encoder that combines the character values and the encoded sign bits of the characters to form in advance a predetermined number of transmitted symbolic characters per frame and provides precoded levels corresponding to these symbolic symbols; and said character bit encoder comprises: an adjacent class representation generator that generates, for each frame, an adjacent class of sign bits for signed information bits, defining an adjacent class element for a convolutional code that defines an adjacent class of convolutional code containing candidate encoded signed character bits, and a character sign bit selector that operates in response to the character bits representing the adjacent class and character values, which selects the encoded character bits of the characters from the candidate GOVERNMENTAL iconic symbols coded bits. 42. A method for framelessly encoding digital information bits in a transmitter into a predetermined number of precoded layers per frame, said digital information bits being divided into a first predetermined number of significant information bits and a second predetermined number of signed information bits per frame, which includes: mapping meaningful information bits into a predetermined number of symbol values per frame; encoding the sign information bits into a predetermined number of encoded sign bits of symbols per frame, and combining the values of the characters and encoded sign bits of symbols to form a predetermined number of transmitted sign bits per frame and outputting precoded levels corresponding to these sign symbols; and the encoding step includes: generating for each frame representing an adjacent class of sign bits for the sign information bits defining an adjacent class element for a convolutional code that defines an adjacent class of convolutional code containing candidate encoded signed character bits of characters; a selection using representing an adjacent class of sign bits and character values, encoded sign character bits from candidate encoded sign character bits. 43. A receiver for receiving frame-by-frame digital information bits encoded by a transmitter into a predetermined number of sign characters per frame, the transmitted sign characters having a desired spectrum shape, said digital information bits being divided into a first predetermined number of significant information bits and a second predetermined the number of sign information bits per frame, which contains: a symbol and character value extraction unit that divides transmitted symbolic characters into encoded symbolic symbol bits and symbol values; a significant bit demapper, which decodes the encoded symbol values into significant information bits, and a sign bit decoder that decodes the encoded symbol sign bits into symbol information bits; where the symbol bit decoder comprises a matrix block comprising a matrix H ^t by which the encoded symbol symbol bits are multiplied to reconstruct the symbol information bits. 44. The receiver according to item 43, in which the matrix N ^t is determined as follows: