KR20010006101A - System and method for spectrally shaping transmitted data signals - Google Patents

Abstract

The system transmits digital information bits encoded with a predetermined number of encoded symbols per frame from the transmitter 40 to the receiver via the network 46, wherein the transmitted signal symbols have the desired spectral form, and the digital information bits are framed. The first predetermined number of size information bits per second and the second predetermined number of sign information bits. The transmitter includes a size correspondence device 50 for mapping the size information bits to a predetermined number of symbols, a code bit encoder 52 for encoding the code information bits with a predetermined number of encoded symbol code bits per frame, and per frame. And a signal point selector 54 that combines the symbol size and the encoded symbol sign bits in response to the magnitude matching device and the sign bit encoder to form a predetermined number of transmitted encoded symbols.

Description

SYSTEM AND METHOD FOR SPECTRALLY SHAPING TRANSMITTED DATA SIGNALS

[Related application of this application]

This application is a U.S Provisional Application No. entitled "Generalized Spectral Shaping," filed April 8, 1997, filed by the same assignee as the present application. 60 / 042,826 [Inventor: M. Vedat Eyuboglu, Pierre A. Humblet, Daeyoung Kim and David Tung].

This application is based on the above application, the U.S Patent Application No. No. "Device, System and Method for Spectrally shaping Transmitted Data Signals" filed November 13, 1996. 08 / 747,840 [inventors: M. Vadat Eyuboglu and Pierre A. Humblet]; And

U.S patent application entitled "Device and Method for Precoding Data Signals" filed Nov. 29, 1997, Representative Dock No. Priority was claimed by CX096044P04 [Inventors: M. Vedat Eyuboglu, Pierre A. Humblet, and Daeyoung Kim].

Public switched telephone networks (PSTNs) include analog local loops that connect digital backbone networks and end users to these backbone networks. In most telephone calls, the analog signal transmitted by the local user is digitized at the local central office and converted into a 64 Kbps bit stream. This stream is transmitted over a digital backbone network and converted back to analog for transmission over a remote local loop from the remote central telephone station to the user. For example, a dial-up modem, a V.34 modem, modulates digital information into an analog signal for transmission and communicates via a PSTN. The digital-to-analog conversion process at the point of entry into the digital period generates quantization noise that limits the data rate to about 30 Kbps.

For example, a technology has been developed that enables transmission at speeds considerably higher than 30 Kbps, even up to 56 Kbps when the user is directly connected to the digital network via ISDN or T1. Furthermore, there is a standardized protocol for this type of transmission, the International Telecommunications Union (ITU) standard V.90, which is expected to be approved soon. With this technique, random digital information is encoded into μ-law or A-law octets by a digital pulse code modulation (PCM) modem using a channel encoder (depending on region). Octet corresponds directly to a symbol in a digital-to-analog (D / A) converter located at the end user's central telephone station (unless otherwise indicated, all descriptions below follow the law of μ: expansion to law A is easy The correspondence uses all sets or subsets of 255 levels of the D / A converter depending on the regulation constraints on the average power.

Since the information is transmitted in octets over the digital network, the encoded data is first mapped to octets at a rate of 8000 octets per second for transmission. The octet is then converted to the corresponding symbol in the D / A converter at the end user's central telephone station. The generated 8 kHz symbol sequence is passed through a low pass filter (LPF) and sent through an analog loop to the end user's analog PCM modem. The output of the D / A converter may look like a sequence of impulses, each with an amplitude equal to one of the D / A levels. The analog PCM modem first detects which symbol was sent and then reverses it to recover the original information to obtain an estimate of the original information.

When the information is transmitted randomly, the spectral analysis of the signal after the D / A conversion indicates that the spectrum of the sequence output by the D / A converter is completely flat. Therefore, when this sequence is passed through the LPF at the central telephone station, the spectrum of the signal has the spectral form of the LPF. Unfortunately, this spectrum has a significant amount of energy near DC (f = 0) that can drive the system's transducers to saturation and cause unwanted nonlinear distortions in the signal being transmitted. In this application, this type of distortion cannot be tolerated and must be removed.

More generally, PCM requires a technique that can shape the spectrum of the signal transmitted from the D / A converter. Furthermore, there is a need for a spectral shaping technique that is applicable to various transmission techniques other than PCM.

FIELD OF THE INVENTION The present invention relates generally to high speed data communications, and more particularly to systems and methods for spectrally shaping transmitted data signals.

1 is a simplified block diagram of a typical telephone company central office.

FIG. 2 shows the spectral form of the low pass filter of FIG. 1 and the frequency spectrum of the output y _k symbols from the μlaw to linear converter of FIG.

3 shows a portion of two frequency spectra, each of which is null at DC, with one spectrum falling very rapidly to zero at DC and the other falling more gently.

4 is a schematic block diagram of a transmitter of a central digital PCM modem formed in accordance with the present invention.

5 is a schematic block diagram of a receiver of an end user analog PCM modem formed in accordance with the present invention.

6 is a schematic block diagram of the sign bit encoder of the transmitter shown in FIG.

FIG. 7 is a schematic block diagram of the coset representative generator shown in FIG. 6. FIG.

8 is a schematic block diagram of the symbol sign bit selector of FIG.

9 is a grid diagram representing a conventional code.

FIG. 10 is a flow diagram illustrating generalized logic for the symbol sign bit selector shown in FIG. 8; FIG.

FIG. 11 is a schematic block diagram of the sign bit decoder shown in FIG. 5; FIG.

12 is a schematic block diagram of the present invention used as a precoder in an upstream PCM transmitter.

The present invention relates to systems and methods for spectrally shaping transmitted data signals that are generally applicable to various data transmission techniques. For convenience of description, the present invention is described herein in terms of a pulse code modulation (PCM) transmission system. However, one of ordinary skill in the art will appreciate that the present invention can be extended to other transmission techniques and that the PCM implementation described herein can be readily extended to these techniques.

1 and 2 illustrate the presence of energy near DC in a signal transmitted to an analog PCM modem of an end user through an analog loop. In Fig. 1, a typical central telephone station of a PSTN that receives at input 12 μ-law octets transmitted from a central digital PCM modem (not shown) directly coupled to the digital portion of the telephone system. A part of a telephone central office 10 is shown. This octet is converted into a sequence y _k of symbols by a digital to analog (D / A) converter known as μ-law to linear converter 14. Each of the symbols corresponds to one of the 255 μL law levels. This symbol is outputted to the low pass filer (LPF) 18 via lead 16, which is filtered through an analog loop 20 to the receiver of an end user analog PCM modem. Outputs This analog signal is demodulated and decoded by the receiving modem and output as a digital bit stream. The digital bit stream is an estimate of the data originally transmitted.

The sequence y _k of the symbol on the lead 16 from the μ law to linear converter 14 has the flat frequency response 22 of FIG. 2. The spectral form 24 of the LPF 18 contains a significant amount of energy near DC (f = 0) as shown at point 26. Since sequence y _k has a flat frequency response, the spectrum of signal s (t) output by filter 18 has the same spectral form 24 as filter 18, so signal s (t) is also near DC. Contains a significant amount of energy. As mentioned above, this energy near DC tends to saturate the transducers of this system, creating unwanted nonlinear distortion in the signal s (t) transmitted to the receiving modem.

In applications like PCM, this distortion should be reduced. This can be accomplished by reducing the signal energy of the signal transmitted near the DC to produce a DC null. This DC null 28 is shown in FIG. 3. As is known in the state of the art, a running digital sum (RDS) of the transmitted symbol y _k (i.e., all levels of previously transmitted) is needed to generate a spectral null at the DC of the transmitted signal. Mathematical sum) must remain near zero. The spectral shape near the DC null 28 may change from a relatively small slope spectrum 30 to a very rapidly descending spectrum 32 at DC. The sharpness of the null depends on how tightly the RDS is controlled.

As described below, the present invention encodes digital data being transmitted in a manner that keeps the RDS near zero. This creates spectral nulls in DC to reduce nonlinear distortion caused by converter saturation. More generally, the present invention may be used to encode the digital data being transmitted as desired to shape the spectrum of the transmitted signal.

The transmitter 40 of the digital PCM modem receives a serial digital bit stream 42 from a data terminal equipment (not shown), such as a personal computer, and sends the received bit stream to the digital network 46. Encode to octet 44 for transmission via The serial bit stream 42 is converted into parallel form by the serial to parallel converter 48. The transmission / encoding scheme of the present invention is based on an n symbol data frame, where k represents a data frame (time) index. For example, there may be 2, 3, 4, 5, 6 symbols transmitted per data. The transmitted symbol corresponds to the µlaw constellation points selected to represent information bits. For each data frame, serial-to-parallel 48 outputs (n-r) + m information bits, where r is the number of redundancy bits. The number of surplus bits, as specified in the V.90 standard, may be 0, 1, 2, 3.

In the remainder of the description, it should be noted that subscript variables represent scalar quantities and superscript variables represent matrices. Also, the row vector is a bold subscript variable and all indices start at zero. For example, x _k = (x _{k, 0} , x _{k, 1} , ...).

The nr bits, denoted as bits v _k , represent information conveyed through sign bits (signal information bits), and the m bits _denoted as bits u _k are conveyed through magnitude bits (magnitude information bits). Express information. The number of bits, m, can be determined by selecting m that satisfies

Where M ₁ is the number of positive canstellation points for the i th symbol in the data frame. This process is fully described in the V.90 standard.

The m size information bits u _k are provided to a magnitude mapper 50 where the m bits can be provided by a corresponding technique such as shell mapping as described in the ITU V.34 standard or by ITU V. .90 Corresponds to the n symbol size g _k by the modulus conversion described in the standard. The size to which the size information bits correspond is the size of the μ law point used as the canstellation point for transmitting the information bits. This size correspondence technique and the cantilever point selection process can be understood by one skilled in the art and further description is omitted herein. The remaining information bits, sign information bits v _k of the data frame are provided to a sign bit encoder 52 where they generate n sign bits s _k (encoded symbol sign bits) as detailed below. Let's do it. The n symbol size g _k and the n sign bit s _k are provided to the signal point selector and combined to form an n signed symbol y _k . The n encoded symbol y _k is then provided to an octet converter 56 that selects an octet corresponding to each of the signal symbols and transmits the octet to the digital network 46. With other transmission techniques, an octet converter that converts the encoded symbols into a form compatible with the digital portion of the PSTN may not be used and the signal point selector will output the encoded symbols directly to the network.

The octets 44 'leaving digital network 46 (which may have been modified by digital damage in the network) are received by central office CO. The octet 44 'is converted into a symbol by the D / A converter of the CO 60 and transmitted via the analog loop 62 to the receiver 64 of the end user analog PCM modem at the 8 kHz sequence level. The analog level is received by a receiver front end 66, where the analog level is digitized and timing recovery, equalization, and symbol decisions are performed.

Receiver front end 66 and outputs the received symbols y _k the serial symbol n to a serial form to the serial to parallel converter 68 to be converted to a frame of parallel coded symbol y _k. The n parallel coded symbol y _k is provided to a magnitude and sign extractor 70 that extracts the symbol size g _k and sign bit s _k from y _k . The symbol size g _k is provided to a magnitude demapper 72, which is, for example, a coefficient transform inverse corresponder, to recover the magnitude information bit u _k . Backward countermeasures can be understood by those skilled in the art and further explanations are omitted herein. The sign bit s _k is provided to a sign bit decoder 74 to recover the sign information bit v _k as described below. The decoded information bits are further processed and provided to a data terminal equipment such as a personal computer.

Sign bit encoding

The sign bit encoder 52 is shown in more detail in FIG. 6. The sign information bit v _k is provided to a coset representative generator 80 that generates and provides each frame n coset representative sign bit t _k to the symbol sign bit selector 82. The n coset representative sign bit t _k in each frame defines the coset representative element for the defined convolutional code G (D) used by the symbol sign bit selector 82 and the coset representative sign bit t (D). The entire sequence of s defines the corset representation for the convolutional code as a whole. The n coset representative sign bit t _k also identifies a corset of convolutional code that includes candidates of encoded symbol sign bits as described in detail below.

using n koset representing the sign bit t _k symbol sign bit selector 82 in order to form a candidate of the code symbol sign bit for koset representing the sign bit t _k is shown in Figure 9 the grid described in the following Figure (trellis Valid convolutional code sequences and bits defined by a lattice diagram are modified by an exclusive OR operation. This candidate is an element of the corset identified by the corset representative sign bit. Using the symbol size, the symbol sign bit selector 82 selects the encoded symbol sign bit s _k for each frame from candidates of encoded symbol sign bits that produce the desired spectral shape, and selects the sign bit of FIG. To the signal point selector 54. In contrast to a per frame basis, the output of the symbol sign bit selector 82 for the entire sequence is where s (D) is the encoded symbol sign bit and t (D) is the coset representative for the convolutional code and c (D) may be expressed as s (D) = t (D) c (D) when the code sequence is an element of the convolutional code G (D).

It should be noted that all the candidates of the symbol sign bits encoded using this selection process can be used and will be decoded into the encoded sign information v _k as described below. In addition, the spectral shaping technique of the present invention does not affect the symbol size and thus does not affect the transmission power. As a result, it is easy to design a system that performs spectral shaping while satisfying the transmitter power constraints required by the FCC.

The coset representative generator 80 is shown in more detail in FIG. 7 to include a differential encoder 84 and a matrix block 86. Noise on the data channel can affect the transmitted sign bits and cause polarity inversion. For example, a differential encoder 84 that differentially encodes to any bit position that is even positions 0, 2, and 4 sign bits for a given H ^T , H- ^T, and G (D), and the decoding decoder of FIG. By employing 132, it is possible to achieve polarity inversion invariance. The differentially encoded sign information bit v _k 'is integrated (modulo 2 ₎ with the matrix H ^-T _{(nr) xn} to produce the n coset representative sign bit t _k provided to the symbol sign bit selector 82. .

An example of this matrix is when there are 6 symbols and 1 surplus bits per frame as specified in the ITU V.90 standard.

Where D is a frame delay, which is a delay based on frame (time) index k.

As shown in FIG. 8, the symbol sign bit selector 82 includes a selector controller 88 which receives n coset representative sign bits t _k for each frame from the coset representative generator 80 and the magnitude corresponder of FIG. Receive n symbol sizes from 50 and output the encoded symbol sign bit s _k for each frame. The selection controller 88 combines the candidate of the candidate encoded symbol sign bits with the size to form and provide an encoded encoded symbol candidate to the filter 90. Filter 90 calculates the metric referred to herein with the running filter sum (RFS) described below for each candidate to minimize the RFS, resulting in an encoding associated with the encoded coded symbol candidate. This is provided to the selection controller 88 which selects the coded symbol sign bits. The operation of the symbol sign bit selector 82 is described with reference to FIGS. 9 and 10.

The selection controller 88 performs an exclusive OR operation on the corset representative code bits with a valid code sequence of convolutional code to modify n coset representative code bits t _k per frame. Convolutional code is a set of possible sequences defined by grid diagrams and valid code sequences are sequences that do not violate the constraints of grid diagrams. For illustration purposes, the selection controller 88 will use a single redundant bit r and convolutional code G (D) = [1 + D 1 1 + D 1 1 + D 1]. In order to express this in a grid diagram, a two-state grid diagram such as the grid diagram 100 of FIG. 9 is required. The constraint of the grid diagram is explained as follows.

For a given frame k, the selection controller 88 modifies the n coset representative sign bit t _k by an exclusive OR operation with a random convolutional code sequence, subject to the constraints of the grid diagram 100. In this example, the convolutional code sequence is as follows:

A: 000000 (that is, do nothing)

B: 111111 (i.e. invert all sign bits in frame j)

C: 101010 (that is, inverting the even sign bit of frame j)

D: 010101 (that is, inverting the even sign bit of frame j)

Therefore, if the state Q _k of the selection controller 88 at the start of frame k is zero, only convolutional code sequences A 102 and B 104 are valid sequences of frame k. Conversely, if the state Q _k of the selection controller 88 is 1, only convolutional code sequences C 106 and D 108 are valid sequences of frame k. As noted above, the corset representative sign bit is exclusive ORed with each valid code sequence, for example, Form a candidate of encoded symbol sign bits. Each candidate is also a coset element (or element of the coset representative) of the convolutional code identified by the coset representative sign bit. Each candidate is then returned to the selection controller 88 in combination with the symbol size to form an encoded encoded symbol candidate provided to the filter 90 of FIG. 9 in which an RFS for each is calculated. The selection controller 88 outputs the encoded symbol sign bits for frame j that minimizes RFS.

Together with the convolutional code sequence of the encoded symbol sign bits selected for frame k, the current state Q _k is used to determine the next state according to the constraints of grid diagram 100. For example, a candidate If is selected for frame k, then state Q _{k + 1} of selection controller 88 at the start of the frame is one.

Spectral shaping achieved using the present invention can be improved by introducing a look-ahead. That is, instead of selecting the coded symbol sign bits that depend only on the current frame, the symbol sign bit selector 82 uses the magnitude correspondent of FIG. 4 to determine which encoded symbol code achieves the best spectral shaping. The symbol size generated by 50 and the corset representative sign bits for the current frame and the subsequent frame may be used. The V.90 standard specifies that up to three frames may be used subsequently, depending on the amount of look-ahead delay negotiated during startup.

The spectral shaping distance RFS based on the filter transfer function h (D) is calculated by the filter 90 to the look-ahead delay or depth Δ via the grid for all possible paths (or candidate sequences), The selection controller 88 selects the encoded symbol sign bits associated with the candidate sequence for frame k that produces the minimum RFS.

Referring again to the grid diagram 100 of FIG. 9, a possible candidate sequence for a look-ahead having a depth of 1 is described. If the state Q _k of the selection controller 88 at the start of frame k is zero, convolutional code sequences A 102 and B 104 are valid sequences for frame k. However, code sequence A 102 and B 104 for frame k + 1 should also be considered. Since code sequences A 102 and B 104 are valid in frame k, state Q _{k + 1} in frame _{k + 1} is 0 or 1 and therefore code sequences A 102 ', B 104', C ( 106 '), D 108' is valid. As noted above, the corset representative sign bit is exclusive ORed with each valid code sequence to form a candidate of the encoded symbol sign bit. Using the look-ahead, the corset representative sign bits for each frame k and k + 1 are exclusive ORed with the valid code sequence of each path of the grid diagram to form a candidate sequence. In this example, the candidate sequence is the following four sequences { 3) 4) }to be. The RFS for each sequence is determined and a candidate encoded symbol sign bit is selected for frame k in the determined sequence.

In FIG. 10, a flowchart 120 describes the operation of the symbol sign bit selector 82. In step 122, the selection controller 88 modifies the coset representative code bits along the lattice diagram to generate candidates for encoded symbol code bits (or candidate sequences in the case of look-ahead). And, in step 124, the selection controller combines the candidate of the symbol size and the encoded symbol sign bits to form an encoded encoded symbol candidate (or candidate sequence) and provides it to the filter 90. In step 126, filter 90 determines an RFS for each candidate (or candidate sequence) and provides an RFS for each candidate (or candidate sequence) to selection controller 88. Finally, at step 128, the selection controller 88 selects a candidate (or candidate sequence) of encoded symbol sign bits that minimizes RFS and transmits the encoded symbol sign bits.

It should be appreciated that the present invention may use various convolutional codes G (D) represented by different grid diagrams and different convolutional code sequences. Extensions to the various convolutional codes and code sequences outlined herein will be readily apparent to those skilled in the art.

In general, the use of PCM allows the spectral shaping technique of the present invention to filter the spectrum of the analog signal transmitted from the D / A converter of FIG. 5 CO 60 to achieve the desired spectral shape. Set up and mold with minimal RFS. The response h (D) of the filter 90 defining the desired spectral shape can be expressed as follows.

Where A (D) and B (D) are functions and a and b are real numbers chosen to achieve the desired spectral shape. And N _A and N _B are the number of coefficients used for the numerator and denominator, respectively, to represent h (D). The symbol based RFS may be calculated as follows.

The frame-based RFS for the k-th frame may be calculated as follows.

Where j is the symbol (time) index.

When using the sign bit encoder of the present invention to generate spectral nulls in DC, RFS is RDS and RDS and response h (D) of filter 90 is

From the transmitted encoded symbol y _k , the filter 90 calculates the RDS of the transmitted encoded symbol y _k at symbol time j as follows.

Here j is a symbol (time) index and the frame-based RDS for the k-th frame can be calculated as follows.

Where j is the symbol (time) index.

For a look-ahead with a look-ahead depth Δ, the RDS is calculated as follows.

Where LRDS is a look-ahead RDS. In general, a look-ahead can be introduced that minimizes RFS as follows.

Sign bit decoding

5 is shown in FIG. 11 such that the sign bit decoder 74 of the receiver 84 includes the matrix block 110. In matrix block 110, sign bit s _k is the differentially encoded sign information bit. The matrix H ^T _{nx (nr)} and (to modulo 2) are integrated (ie filtered) to recover.

An example of the matrix H ^T when six symbols per frame (n = 6) and one redundant bit (r = 1) is shown in Equation 11 as follows.

The matrix H ^T is designed such that a determination error in v _k due to an error in the received code signal s _k does not propagate more than one frame. This is because H ^T is a matrix in the form of a finite impulse response (FIR) and there is only one single delay.

In order to specify how each of the candidates of the encoded symbol sign bits for each frame of the symbol to be transmitted are generated by the sign bit encoder 52 of FIG. 4, the encoding and decoding process must be represented mathematically. Restored information bits Decoding can be expressed mathematically as follows.

And the sign bit s _k (encoding) can be represented mathematically as follows.

Substituting the right side of equation (13) into equation (12) for s _k , the following equation is derived.

The following conditions (1) H ^T H ^-T = I (I is the identity matrix) (2) If G, H ^T and H ^-T are selected so that GH ^T = 0 is satisfied, regardless of the value of r _k to be.

9 In the grid diagram 100 of FIG. 9, the single surplus bit is r _k and the convolutional code G (D) = (1 + D 1 1 + D 1 1 + D 1). R _k G (D) is equivalent to r _k (111111) + r _{k + 1} (101010) since 1 * r _k = r _k and D * r _k = r _{k + 1} . Here, r _{k + 1} represents the state of the Q _k and r _k are grid also represents the branches (branch) or path (path) taken through the grid also. The convolutional code sequence AD corresponds to the expression r _{k + 1} , r _k as follows.

A: 000000-r _{k + 1} = 0, r _k = 0

B: 111111-r _{k + 1} = 0, r _k = 1

C: 101010-r _{k + 1} = 1, r _k = 0

D: 010101-r _{k + 1} = 1, r _k = 1

Here, the code sequence AD can be thought of as r _k G (D).

Since the r _k value is not affected by how the information bits were decoded, each set of n sign bits generated by another valid code sequence can be used to produce the same decoded information. As a result, a set of n code bits that minimize RFS / RDS can be selected to perform spectral shaping as desired.

Upstream PCM Transmission

The spectral shaping technique according to the present invention can be used in connection with equalization as in the transmitter of an analog PCM modem used for uplink PCM transmission to perform precoding. In this case, the response h (D) represents the channel response between the transmitting modem and the analog-to-digital (A / D) converter of the CO line card and usually includes the filtering effects of the transmitting modem front end, analog local loop, and CO line card. do.

Using the principles of the present invention, a channel output sequence x (n) (z (n) prefiltered) that produces a sequence y (D) at the A / D converter input whose signal point is similar to the A / D quantization level Can be generated. In this case, the goal is to minimize the energy of the transmission signal x (D) = y (D) / h (D) while maintaining a low canstellation extension at the A / D converter input. Larger canstellation extensions are undesirable in this case as well, because of the increased echo-induced quantization noise and other damage.

In this application, usually the channel response h (D) will be determined by the receiving modem based on the channel measurements obtained during modem startup or by both the transmitting and receiving modems and during the data transmission the transmitting modem will transmit the introductory bits x ( D), which will be converted to the channel output sequence y (D) after passing through the channel. The channel response h (D) is usually chosen to be the minimum phase, which is easily achieved by further filtering at the transmitter, for example.

10 is a transmitter of an analog PCM modem capable of uplink PCM transmission. Transmitter 40 'uses the spectral shaping technique of the present invention to precode the incoming data bit stream 42' using precoder 140. One form of PCM upstream precoding (referred to as one-dimensional precoding that precodes per symbol) was filed on December 29, 1997 and the inventors are M.Vedat Eyuboglu, Fierre A. Humblet, and Daeyoung Kim. U.S. Patent Application Attorney Docket No. 2, entitled Device and Method for Precoding Data Signals. It is described in detail in CX096044P02. The precoder 140 performs multidimensional precoding, that is, precodes a symbol for each frame. Although the implementation of the present invention is different from one-dimensional precoding, the concept is similar to the one-dimensional case and further details may be referred to the above application.

Precoder 140 includes a serial-to-parallel converter 48 ', magnitude matcher 950', sign bit encoder 52 ', and signal point selector 54'. These means are formed and operate in the same way as the number in Fig. 4 except for minor modifications. For example, the operation of the sign bit encoder is modified to perform precoding as described below and the signal point selector takes n precoded level x _k per frame corresponding to the encoded symbol y _k instead of the encoded symbol itself. Output The n precoded level x _k is provided to a parallel-to-serial converter 142 that outputs the precoded level to the prefilter 144 in series form. Prefilter 144 filters the levels and outputs the filtered levels to digital-to-analog converter 146 which then transmits the precoded analog levels through analog channel 148. The channel transforms the precoded level x _k and ideally generates at the quantizer of CO 160 a level corresponding to the coded symbol y _k . That is, the precoder selects a precoded level x _k that generates the level corresponding to the desired coded symbol y _k in the quantizer taking into account the response of the analog channel 148 or more precisely the target channel response h (D). do.

The target channel response h (D) is g (n) which is the response of the prefilter 144 convolved to c (n) which is the response of the analog channel 148 when n is the symbol time index and h (0) = 1. Is the same as This relationship can be expressed as

Since h (0) is designed to be equal to 1, Equation 15 is simplified as follows.

The value for h (n) at a given time is known and the past value of x (n) is known. The filter 90 of FIG. 8 calculates the summation term of Equation 16 and provides it to the selection controller 88. The past value of x (n) is determined from the previous symbol y (n) by the known relationship x (D) = y (D) / h (D) and stored in the filter 90 and the selection controller 88 is x An encoded symbol sign bit that minimizes (n) is selected and operates according to the flowchart 120 of FIG. Instead of transmitting the coded symbol y (n), the precoded level is transmitted.

In addition to the above countermeasures, a means for echo cancellation to separate bidirectional transmissions, and a timing interpolation filter to ensure that the symbols are transmitted in synchronism with the network clock. Should be. The timing interpolation filter will usually be driven by clock recovery circuitry used in downstream receivers. The transmitter is primarily responsible for limiting the transmission bandwidth to 4 kHz and may also include a linear filter to provide the necessary prefiltering to minimize the overall channel response h (D).

Furthermore, in real systems, it may include a trellis coding form to enhance noise immunity. See, for example, the application referred to above, filed Nov. 14, 1996, and entitled `` System and Device for and Method of, Communicating According to a Trellis code of baseband Signals Chosen from a Fixed Set of Baseband Signal Points ''. Lattice coding techniques described in Patent 08/749040 (Attorney Docket No CX096050) can be used. The application is incorporated herein by reference in its entirety. The operation of this system is not affected at all by the lattice coding.

It should be appreciated that the present invention may be practiced with software and / or firmware that may be stored on a computer usable medium such as a computer disk or memory chip. The invention may also be in the form of a computer data signal implemented in a carrier wave transmitted electrically over the Internet, such as when the invention is implemented in software / firmware.

Although the present invention has been described and illustrated by way of specific example, the present disclosure has been presented by way of example only, and one skilled in the art will recognize that many changes in the combination and arrangement of parts may exist without departing from the spirit and scope of the claimed invention. You will know.

The present invention can be used in various ways, such as computer communication, signal processing apparatus, modem.

Claims (44)

Transmitting digital information bits encoded in a predetermined number of signed symbols per frame, frame by frame, from transmitter to receiver on a network, wherein the transmitted encoded symbols are in desired spectral form. having a spectral shape, the digital information bits having a first predetermined number of magnitude information bits and a second predetermined number of sign information bits per frame. The transmitter is A magnitude mapping device for mapping the size information bits to a predetermined number of symbol sizes per frame; A sign bit encoder for encoding the sign information bits into a predetermined number of encoded symbol sign bits per frame, and A signal point selector combining the symbol size and the encoded symbol sign bits to form a predetermined number of transmitted coded symbols per frame in response to the magnitude corresponding device and the sign bit encoder. Including, The sign bit encoder is A coset representative sign bit for the sign information bit for each frame, the coset representative sign bit identifying the coset of convolutional code comprising candidates of encoded symbol sign bits A coset representative generator for generating a coset representative element for convolutional code, and A symbol for selecting the encoded symbol code bit among candidates of the encoded symbol code bit that generates the transmitted encoded symbol having the desired spectral shape in response to the coset representative code bit and the symbol size Sign bit selector Transmission system comprising a. The predetermined first number of size information bits when n corresponds to a predetermined number of symbols per frame and r is the number of redundancy bits used by the sign bit encoder. Is m bits and the predetermined second number of sign information bits are nr bits. 3. The differential encoder of claim 2 wherein the coset representative generator is for differentially encoding predetermined bit positions of the nr bits provided to the sign bit encoder to achieve polarity inversion invariance. Transmission system comprising a. 4. The transmission system of claim 3, wherein the magnitude corresponder maps the m size information bits to n symbols per frame using a modulus conversion mapping scheme. 5. The apparatus of claim 4, wherein the coset representative generator further comprises a matrix block integrating the nr code information bits with a matrix H ^-T to generate the n coset representative code bits per frame. Transmission system. 6. The apparatus of claim 5, wherein the symbol sign bit selector comprises a select controller and a filter, wherein the select controller A logic circuit for generating a candidate of an encoded symbol code bit from the corset representative code bit candidate; and A logic circuit that combines the symbol size with the candidate of the encoded symbol sign bits to form an encoded encoded symbol candidate, The filter is Logic circuitry for determining an RFS for each of the encoded encoded symbol candidates in response to the encoded encoded symbol candidate; Logic circuitry for providing the selection controller with the RFS for each of the encoded encoded symbol candidates; The selection controller Logic circuitry for selecting the encoded symbol sign bit associated with the encoded encoded symbol candidate with the minimum determined RFS in response to the determined RFS for each of the encoded encoded symbol candidates; Transmission system, characterized in that. 6. The apparatus of claim 5, wherein the symbol sign bit selector uses a look-ahead and includes a select controller and a filter, wherein the select controller A logic circuit for generating a candidate sequence of encoded symbol code bits from the coset representative code bits, and Logic circuitry for combining the candidate sequence with the symbol size to form an encoded encoded symbol candidate sequence; The filter is Logic circuitry for determining the RFS for each of the encoded encoded symbol candidate sequences in response to the encoded encoded symbol candidate sequence; A logic circuit for providing the selection controller with the RFS for each of the encoded encoded symbol candidate sequences, The selection controller In response to the determined RFS for each of the encoded encoded symbol candidate sequences, further comprising a logic circuit for selecting the encoded symbol sign bit from the encoded encoded symbol candidate having the minimum determined RFS. Containing Transmission system, characterized in that. 8. The method of claim 7, wherein the RFS for the i th symbol is generated by the filter. The transmission system, characterized in that determined as. 10. The apparatus of claim 8, wherein the RFS is performed by the filter on a frame basis for a kth frame. Where j is the symbol (time) index The transmission system, characterized in that determined as. 10. The method of claim 9, wherein the RFS uses a look-ahead when Δ is a look-ahead depth. The transmission system, characterized in that determined as. 12. The apparatus of claim 10, wherein the receiver divides the transmitted encoded symbol into the encoded symbol sign bit and symbol size and a magnitude and sign extractor and the encoded symbol sign bit. And a sign bit decoder for decoding into information bits. 12. The transmission system of claim 11, wherein the sign bit decoder comprises an excitation matrix block that integrates the encoded symbol sign bits with a matrix H ^T to recover the sign information bits. 13. The transmission system of claim 12, further comprising a differential decoder for differentially decoding a predefined bit position of the recovered coded information bits. The transmission system of claim 13, further comprising an octet converter for transmitting an octet corresponding to the encoded symbol through a network in response to the signal point selector. The method of claim 13, wherein n is 6, r is 1, and the matrix H ^T is Transmission system, characterized in that as defined. The method of claim 15, wherein the weaving code is Transmission system, characterized in that as defined. The method of claim 16, wherein the matrix H ^T is Transmission system, characterized in that as defined. 8. The method of claim 7, wherein the RFS is a running digital sum (RDS) and the RDS for the i th symbol is generated by a filter. Where j is the symbol (time) index The transmission system, characterized in that determined as. 19. The apparatus of claim 18, wherein the RDS is selected by the equivalence class selector for a kth frame on a frame basis. [Where j is the symbol (time) index] The transmission system, characterized in that determined as. 20. The system of claim 19, wherein said RDS uses look-ahead [Δ is look-ahead depth] The transmission system, characterized in that determined as. Transmitting digital information bits encoded with a predetermined number of encoded symbols per frame from frame to frame to a receiver on a network, wherein the symbols have a desired spectral form, the digital information bits having a first predetermined number of size information bits per frame. And having a second predetermined number of sign information bits; Mapping the size information bits to a predetermined number of symbol sizes per frame; Encoding the sign information bits into a predetermined number of encoded symbol sign bits per frame, and Combining the symbol size and the encoded symbol sign bits to form a predetermined number of transmitted encoded symbols per frame, The encoding step A corset representative code bit for the sign information bit for each frame—the coset representative code bit defines a coset representative element for the convolutional code that identifies a coset of convolutional code that includes candidates of encoded symbol sign bits Generating-, and Selecting the encoded symbol sign bit from among the candidates of the encoded symbol sign bit producing the transmitted encoded symbol having the desired spectral shape using the coset representative code bit and the symbol size. Transmission method comprising a. 22. The method of claim 21, wherein when n is a predetermined number of symbols per frame and r is a number of redundancy bits used in the encoding step, the first predetermined number of size information bits is m bits and the And the second predetermined number of sign information bits are nr bits. 23. The method of claim 22 wherein the generating step comprises differentially encoding a predetermined bit position of the n-r bits to achieve polarity inversion invariant. 24. The transmission method according to claim 23, wherein the correspondence step corresponds to mapping the m size information bits into n symbols per frame using a coefficient transformation correspondence technique. 25. The method of claim 24, wherein the generating step further comprises integrating the nr code information bits with a matrix H ^T to generate the n coset representative code bits per frame. The method of claim 25, wherein said selecting step Generating an encoded symbol code bit candidate from the coset representative code bits; Combining the symbol size with the candidate of the encoded symbol sign bits to form an encoded encoded symbol candidate; Determining an RFS for each of the encoded encoded symbol candidates; Providing the RFS for each of the encoded encoded symbol candidates, and Selecting the encoded symbol sign bit associated with the encoded encoded symbol candidate having the minimum determined RFS. Transmission method comprising a. The method of claim 25, wherein said selecting step uses a look-ahead Generating a candidate sequence of encoded symbol code bits from the coset representative code bits, Combining the candidate sequence and the symbol size to form an encoded encoded symbol candidate sequence, Determining the RFS for each of the encoded encoded symbol candidate sequences; Providing the RFS for each of the encoded encoded symbol candidate sequences, and Selecting the encoded symbol sign bit from the encoded encoded symbol candidate with the minimum determined RFS Transmission method comprising a. 29. The apparatus of claim 27, wherein the RFS for the i th symbol is generated by the filter. The transmission method, characterized in that determined as. 29. The apparatus of claim 28, wherein the RFS is performed by the filter on a frame basis for a kth frame. Where j is the symbol (time) index The transmission method, characterized in that determined as. 30. The method of claim 29, wherein the RFS uses a look-ahead when Δ is a look-ahead depth. The transmission method, characterized in that determined as. 31. The method of claim 30, further comprising separating at the receiver the encoded coded symbol into the encoded symbol sign bit and symbol size and decoding the encoded symbol sign bit into the sign information bit. The transmission method characterized by the above-mentioned. 32. The method of claim 31 wherein the decoding step includes integrating the encoded symbol sign bits with a matrix H ^T to recover the sign information bits. 33. The method of claim 32, further comprising differentially decoding a predefined bit position of the recovered coded information bits. 34. The method of claim 33, further comprising transmitting an octet corresponding to the encoded symbol through a network. 34. The method of claim 33 wherein n is 6, r is 1 and the matrix H- ^T is The transmission method characterized in that it is defined as. 36. The method of claim 35 wherein the weaving code is The transmission method characterized in that it is defined as. The method of claim 36, wherein the matrix H ^T is The transmission method characterized in that it is defined as. 28. The method of claim 27, wherein the RFS is a running digital sum and the RDS for the i th symbol is: Where j is the symbol (time) index The transmission method characterized in that is determined as The apparatus of claim 38, wherein the RDS is for a k-th frame on a frame basis. [Where j is the symbol (time) index] The transmission method, characterized in that determined as. 40. The method of claim 39, wherein the RDS uses a look-ahead Where Δ is the look-ahead depth The transmission method characterized in that is determined as Frame digital information bits at a predetermined number of precoded levels per frame, the digital information bits having a first predetermined number of size information bits and a second predetermined number of sign information bits per frame. In the system for precoding (transcoding) at the transmitter in units, A size correspondence device for mapping the size information bits to a predetermined number of symbol sizes per frame; A sign bit encoder for encoding the sign information bits into a predetermined number of encoded symbol sign bits per frame, and Combining the symbol size and the encoded symbol sign bits and forming the precoded symbols corresponding to the encoded symbols to form a predetermined number of encoded symbols per frame in response to the magnitude corresponding device and the code bit encoder. Include a signal point selector to output the level, The sign bit encoder is A corset representative code bit for the sign information bit for each frame—the coset representative code bit defines a coset representative element for the convolutional code that identifies a coset of convolutional code that includes candidates of encoded symbol sign bits A representative coset generator for generating-, and A symbol code bit selector for selecting the encoded symbol sign bit from a candidate of the encoded symbol code in response to the coset representative code bit and the symbol size A system for precoding, comprising a. The digital information bits at a predetermined number of precoded levels per frame, the digital information bits having a first predetermined number of size information bits and a second predetermined number of sign information bits per frame, in the frame unit. In the method for precoding, Mapping the size information bits to a predetermined number of symbol sizes per frame; Encoding the sign information bits into a predetermined number of encoded symbol sign bits per frame, and Combining the symbol size and encoded symbol sign bits to form a predetermined number of coded symbols per frame and outputting the precoded levels corresponding to the coded symbols, The encoding step A corset representative code bit for the sign information bit for each frame—the coset representative code bit defines a coset representative element for the convolutional code that identifies a coset of convolutional code that includes candidates of encoded symbol sign bits Generating-, and Selecting the encoded symbol sign bit from a candidate of the encoded symbol code using the corset representative sign bit and the symbol size A method for precoding, characterized in that it comprises a. Digital information bits encoded by a transmitter into a predetermined number of coded symbols per frame, wherein the transmitted coded symbols have a desired spectral form and the digital information bits comprise a first predetermined number of magnitude information bits per frame and in advance. A receiver for receiving from a network on a frame-by-frame basis a fixed number of sign information bits; A size and code extractor that separates the transmitted coded symbols into encoded symbol sign bits and encoded symbol sizes; A magnitude counterpart that decodes the encoded symbol size into the magnitude information bits, and A code bit decoder for decoding the encoded symbol code bits into the code information bits, The sign bit decoder includes a matrix block for integrating the encoded symbol sign bits with a matrix H ^T to recover the sign information bits. Receiver characterized in that. The method of claim 43, wherein H ^T is Receiver, characterized in that defined as.