JP2002501693A - System and method for spectrally shaping a transmitted data signal - Google Patents

Abstract

(57) [Summary] A system for transmitting, on a frame basis, digital information bits encoded from a transmitter into a predetermined number of coded symbols for each frame and transmitted to a receiver over a network, comprising: The resulting encoded symbol has a desired spectral shape, and the digital information bits are divided into a first predetermined number of intensity information bits and a second predetermined number of sign information bits for each frame. The transmitter is: an intensity mapping device that maps intensity information bits to a predetermined number of symbol intensities per frame; a sign bit encoder that encodes the sign information bits into a predetermined number of encoded symbol sign bits per frame; A signal point selector responsive to the strength mapping device and the sign bit encoder for combining the symbol strength and the encoded symbol sign bits to form a predetermined number of transmitted coded symbols for each frame. The sign bit encoder generates: for each frame, a corset representative of the sign information bits, and for a convolutional code that identifies the corset of the convolutional code that includes the candidate encoded symbol sign bits. And a coded symbol having a desired spectral shape from the candidate encoded symbol sign bits in response to the corset representative sign bits and the symbol strength. Is generated by a symbol sign bit selector that selects the encoded symbol sign bit that produces

Description

DETAILED DESCRIPTION OF THE INVENTION System and method for spectrally shaping a transmitted data signal                           Cross reference with related application   This specification is related to the following US patent applications: All of these applications Assigned to the assignee of the present invention and incorporated herein by reference:   US Provisional Application No. 60 / 042,826; title "Generalized Spectral S haping "filed on April 8, 1997; inventors M. Vedat Eyuboglu, Pierre A. Humblet Daeyoung Kim, David Tung; The present invention is based on this application and has common subject matter. Its priority with respect to this case is claimed in this case;   U.S. patent application Ser. No. 08 / 747,840; title "Devices, System and Me." thod for Spectrally Shaping Transmitted Data Signal "November 1, 1996 Filed for 3 days; inventors Vedat Eyuboglu, Pierre A. Humblet; and   U.S. Patent Application, Attorney Document No. CX096044P02, Title "Device and Method  for Precoding Data Signals "; filed on December 29, 1997; inventor M. Vedat E yuboglu, Pierre A. Humblet; Daeyoung Kim.                                Field of the invention   The present invention relates to high-speed data communication, and more particularly, to transmitting data signals to be transmitted. A system and method for shaping a vector.                                Background of the Invention   The public switched telephone network (PSTN) is a digital switched telephone network. Backbone network and connecting end users to this backbone And an analog local loop. In a normal telephone call, Analog signals sent by local users are sent to a local central office. Digitized and converted to a 64 kbit / s bit stream, Transported over the digital backbone network and re-opened at a remote central office Converted to analog and transmitted to end user on remote local loop Is done. For example, V. Dial-up modems such as the 34 Information is communicated on the PSTN by modulating the information into analog signals and transmitting them. entry Digital-to-analog conversion process to digital backbone at points Causes quantization noise, which limits the data transmission rate to about 30 kbit / s I do.   User directly to digital network, for example via ISDN or T1 Transmission rate significantly higher than 30 kbit / s, and potentially Techniques have been developed that allow transmission at rates up to 56 kbit / s. Further A standardized protocol for the transmission of this output, namely the International Telecommunication Union (ITU: International Telecommunications Union) Standard V.90 exists and will soon It is to be ratified. In this technology, random digital information is Digital pulse code modulation (PCM) using a channel encoder ulation) Encode μ or A octets (depending on region) by modem Is executed. These octets are stored in a digital location at the end user's central office. -Mapped directly to symbols in the analog (D / A) converter. (Especially light Unless otherwise noted, the following discussion relates to the μ-method; extension to the A-method is straightforward. ) Ma All or any lower limits subject to prescribed limits in average power A set of 255 level D / A converters can be used.   Information is carried over digital networks in the form of octets, The transmitted data is first mapped to the transmitting octet at a rate of 8000 octets per second. Be pinged. Next, in the end user's central office, the octet is Converted to the corresponding symbol in the A converter. Of the 8kHz sequence The symbol is a low-pass filter. End user on an analog loop through a low pass filter (LPF) Sent to an analog PCM modem. The outputs of the D / A converters each have a D / A level of 1. It can be viewed as a sequence of impulses having corresponding amplitudes. ANALO The PCM modem first determines which symbol was transmitted by first detecting Recover information and demap these symbols to estimate initial digital information Find the value.   If the information is transmitted randomly, the spectrum analysis of the signal after D / A conversion The spectrum of the sequence output by the D / A converter is basically flat. I understand. Therefore, when this sequence passes through the LPF at the central office, the signal Spectrum takes the shape of the spectrum of LPF. Unfortunately, this spectrum is D It has a significant amount of energy near C (f = 0), and as a result, Saturates the transformer, causing undesirable nonlinear distortion on the transmitted signal Sometimes. In this application this kind of distortion cannot be tolerated It is necessary to eliminate this.   More generally, PCM shapes the spectrum of the signal transmitted from the D / A converter. There is a need for an approach that can do that. Furthermore, in addition to PCM, various transmission technologies An adaptive spectral shaping method is needed.                             BRIEF DESCRIPTION OF THE FIGURES   FIG. 1 is a simplified block diagram of a typical telephone company central office.   FIG. 2 shows a symbol y output to the μ-method-to-linear converter of FIG._kFrequency spec 2 shows the spectrum and the spectral shape of the low-pass filter of FIG.   FIG. 3 shows parts of two frequency spectra, each having a null at DC. However, at this time, one spectrum drops very sharply to zero at DC, The other spectrum falls more gradually.   FIG. 4 is a central site digital PCM modem transmitter constructed in accordance with the present invention. It is a schematic block diagram of.   FIG. 5 shows the reception of an end-user analog PCM modem constructed according to the present invention. FIG. 3 is a schematic block diagram of a transceiver.   FIG. 6 is a simplified block diagram of the sign bit encoder of the transmitter shown in FIG. FIG.   FIG. 7 shows a coset representative generator shown in FIG. It is a schematic block diagram of.   FIG. 8 is a schematic block diagram of the symbol sign bit selector of FIG. is there.   FIG. 9 is a trellis diagram showing a convolutional code.   FIG. 10 relates to the symbol sign bit selector shown in FIG. 5 is a flowchart showing general-purpose logic.   FIG. 11 is a schematic block diagram of the sign bit decoder shown in FIG. is there.   FIG. 12 shows an outline of the present invention used as a precoder in an upstream PCM transmitter. It is a schematic block diagram.                            Description of the preferred embodiment   The present invention provides a method for transmitting transmitted data signals that is generally adaptable to various data transmission methods. Vector shaping system and method. For illustrative purposes, the present invention provides a PCM transmission system. The stem is described here. However, the invention can be extended to other transmission methods And that the PCM examples described here can be easily extended to those technologies. Those skilled in the art will understand.   FIGS. 1 and 2 show an end-user analog PCM on an analog loop. 2 illustrates the presence of near DC energy in a signal transmitted to a modem. No. FIG. 1 illustrates a portion of a typical telephone central office 10 on the PSTN. This PS TN is directly attached to the digital part of the telephone system at input 12 Central site receives a μ-octet transmitted from a digital PCM modem (not shown). I believe. Octets are converted by a D / A converter, also known as a μ-method-to-linear converter 14. Symbol y_kIs converted to the sequence Each symbol has 255 μ-method levels. Corresponding to one of the These symbols are low-passed on line 16 The signal is output to a filter (LPF) 18, which is an analog loop 20. At the top, the filtered analog signal s (t) is received by an end-user analog PCM modem. Output to the machine. The analog signal is demodulated and decoded by the receiving modem, This modem outputs a digital bit stream. This digital bit Stream is an estimate of the originally transmitted data.   symbol y on line 16 from μ-method-to-linear converter 14_kSequence is flat It has a frequency response 22 (FIG. 2). The spectral shape 24 of the LPF 18 is As shown, a large amount of energy is included near DC (f = 0). Sequence Y_kHas a flat frequency response, the signal s (t) output by the filter 18 Has the same spectral shape 24 as the filter 18. Therefore, the signal s (t) also contains a large amount of energy near DC. As mentioned above, this energy near DC is Tends to saturate the transformers in the system, which Undesirable non-linear parts are created in the signal s (t) transmitted to the system.   In applications such as PCM, this distortion must be reduced. This is DC This is achieved by generating DC nulls by reducing the signal energy of nearby transmitted signals. Can be Such a DC null 28 is shown in FIG. As is well known in advanced technology In order to generate this spectral null at DC in the transmitted signal, Transmission symbol y_kof The running digital sum (RDS) (ie, all previously transmitted Algebraic sum of all levels) must be kept close to zero. Near DC Null 28 The shape of the spectrum falls very sharply at DC from the relatively shallow slope of the spectrum 30. The spectrum may fluctuate up to 32. How strict is RDS for null sharpness Depending on what is controlled.   As shown below, the present invention converts the transmitted digital data to an RDS near zero. Encode to maintain. This creates a spectral null at DC , Thereby reducing nonlinear distortion caused by transformer saturation . More generally, the present invention is used to encode digital data in transit. The spectrum of the transmitted signal can be shaped as desired.   The transmitter 40 (FIG. 4) in the digital PCM modem is a personal computer From a data terminal device (not shown) such as Receive stream 42 and transmit received bits over digital network 46 Into octets 44 for encoding. Serial bit stream 4 2 is converted to a parallel format by a serial-parallel converter 48. Departure The clear transmission / encoding method is based on n symbol data frames. However And k represents a data frame (time) index. For example, one data frame 2,3 per team 4, 5 or 6 symbols are transmitted. The transmitted symbol is an information bit Corresponds to the constellation points selected to represent Each de For the data frame, the serial-to-parallel converter 48 outputs (n−r) + m pieces of information. Output bits, where r is the number of redundant bits. Redundant video specified by the V.90 standard The number of bits is 0, 1, 2, or 3.   In the following text, the subscript variable indicates the scalar quantity and the superscript Variables do not indicate matrices. Also, the row vector is represented by a bold subscripted variable, The index starts from 0. For example, x_k= (X_{k, 0}, X_{k, 1}, ...).   Bit v_kThe n-r bits written as are transmitted through a sign bit (sign information bit). Bit u_kThe m bits described as “intensity (intensity information ) Represents the information conveyed via Select the number of bits m that satisfies the following equation Determined by: Where M_lIs the number of positive constellation points for the ith symbol in the data frame is there. This process is described in detail by the V.90 standard.   m intensity information bits u_kSent to the strength mapper 50 The strength mapper 50 allocates m bits to the shell map described in the ITU V.34 standard. Ping or modulus conversion as described in the ITU V.90 standard yields n symbols Strength g_kTo map. The strength to which the strength information bits are mapped Is the intensity of the μ method point used as a constellation point when transmitting a point. These strength The mapping method and the conformation point selection process are well understood by those skilled in the art and will be discussed further. No clarification will be made herein. The remaining information bits in the data frame Sign information bit v_kIs sent to the sign bit encoder 52 and the encoder 5 2 is n sign bits s_k(Encoded symbol sign bit) To achieve. This will be described later in detail. n symbol strengths g_kAnd n signs ・ Bit s_kIs sent to the signal point selector 54 and synthesized to generate n encoded symbols. y_kIs formed. n coded symbols y_kIs then sent to octet converter 56 Here, the octet corresponding to each of the coded symbols is selected, and the octet is selected. Is transmitted to the digital network 46. Other transmission methods use coding symbol An octet converter is used to convert files into a format compatible with the digital part of the PSTN. Instead, the signal point selector outputs the encoded symbols directly to the network.   Exit digital network 46 (changed due to digital obstacles in the network) Octet 44 'has been received by the central office (CO) 60. You. Octe The bit 44 ′ is converted into a symbol by a D / A converter in the CO 60 and is converted into an analog signal. End user analog PCM as 8kHz sequence of level on loop 62 It is transmitted to the receiver 64 of the modem. Analog level is at the receiver front Received by end 66, front end 66 digitizes analog levels And perform timing recovery, equivalence, and symbol determination.   The receiver front end 66 receives the received code y_kTo serial-parallel conversion And outputs it to the device 68 in a serial format. The serial-parallel converter 68 is Real symbols are converted into n parallel / encoded symbols y_kTo the frame. n parallel and encoded symbols y_kTo the strength and sign extractor 70 The extractor 70 sends the symbol strength g_kAnd the sign bit s_kAnd y_kOr Extract from Symbol strength g_kBut the strength of e.g. modulus conversion demapper Sent to the demapper 72 and the intensity information bits u_kTo recover. Demapping (Mappi The unlocking process is well known to those skilled in the art and will not be described herein. Rhinoceros Bit s_kIs sent to the sign bit decoder 74, and the IN information bit v_kTo recover. The decoded information bits are further processed to It may be sent to a data terminal device such as a sonal computer.Sign bit encoding   The sign bit encoder 52 will be described in more detail with reference to FIG. Signature information Bit v_kSent to coset representative generator 80 Generator 80 provides n corset representative sign bits t_kRaw And sends it to the symbol sign bit selector 82. n courses T representative sign bit t_kIs the symbol sign bit between each frame Define the predefined convolution code G (D) used by selector 82 and The entire sequence of representative sign bits t (D) collectively represents the code for the convolutional code. Define a rep representative. n coset representative sign bits t_kIs encoded A coset of the convolutional code that includes the candidate symbol sign bit is also identified. This will be described in detail below.   n coset representative sign bits t_kUsing the symbol sign bit The selector 82 converts the bits into a trellis diagram such as that shown in FIG. 9 and described below; OR operation on the effective convolutional code sequence defined by the trellis diagram of By doing the corset representative sign bit t_kChange the encoded thin Form a candidate for the bol sign bit. These candidates will be identified by Is the element of the corset identified by the reset bit. Depending on the symbol strength, Symbol sign bit selector 82 is a candidate encoded symbol sign bit for each frame Encoded symbol sign bits to generate the desired spectral shape from s_kAnd sends these sign bits to the signal point selector 54 (FIG. 4). The symbol sum for the entire sequence, as opposed to frame-based In-bit selector 8 A sequence of encoded symbol sign bits, where t (D) is the convolutional code C (D) is a code sequence that is an element of the convolutional code G (D). It is a sequence.   This selection process results in a candidate for the encoded symbol sign bit. Either can be used and the encoded signature information will be Information bit v_kWill be deciphered. Thus, the spectral shaping method of the present case reduces symbol strength. Does not affect the transmission power. As a result, imposed by the FCC A system that satisfies the constraints of the transmitter power and performs spectrum shaping It becomes easier to do.   FIG. 7 shows a corset set including different encoders 84 and matrix blocks 86. The table generator 80 is shown in more detail. The noise on the data channel Polarity reversal may occur by affecting bits. Different encoding Using different encoders 84 to further support The in-bit differential decoder 132 (FIG. 11)^T , H^-TTo the even positions 0, 2 and 4 and to G (D) It is possible to obtain inversion invariance. Difference encoded signature information bits v_k'Is in matrix block 84, matrix H^T _{(nr) xn}Multiplied by (modulo 2 N) (ie, filtered) and the n coset representative sign bits t_kGenerate , Which are sent to the symbol sign bit selector 82.   Six transmitted symbols and one redundant symbol per frame specified in the ITU V.90 standard. An example of this matrix with long bits is: Here, D is a frame delay, which is a delay based on a frame (time) index.   As shown in FIG. 8, the symbol sign bit selector 82 A controller 88 is provided. The controller 88 includes the corset representative generator 8. 0 to n corset representative sign bits t for each frame_kAnd strength ma And the n symbol intensities received from the upper frame 50 (FIG. 4). Encoded symbol sign bit s_kIs output. Selection controller 8 8 combines the encoded symbol sign bit candidate with the strength and Form coded encoded symbol candidates, which are sent to filter 90. H The filter 90 is called a running filter sum (RFS) (described below). Description) Calculate the quantity for each candidate and send it to the selection controller 88. Con Controller 88 relates to encoded encoded symbol candidates that minimize RFS. Select the encoded symbol sign bit to be encoded. Symbol Sign Be The operation of the bit selector 82 will be described below with reference to FIGS. 9 and 10.   The selection controller 88 converts the corset representative sign bit into a valid convolution code. By performing an exclusive OR operation on the code sequence, n Corset representative sign bit t_kTo change. Convolutional code is based on trellis diagram A set of possible sequences that are defined, and a valid code sequence This is a sequence that does not violate the constraints of the source diagram. For illustration purposes, the selection controller 8 8 uses one redundant bit r and a convolution code G (D) = [1 + D11 + D11 + D1]. This To represent them in a trellis diagram, a two-state trellis such as trellis diagram 100 (FIG. 9) is used. It is necessary to use figures. The restrictions on the trellis diagram will be described below.   For a certain frame k, the selection controller 88 determines whether the constraint XORing with a particular convolutional code sequence according to Coset representative sign bit t_kTo change. Convolutional code sequence in this example These are: A: 000000 (do nothing) B: 111111 (all sign bits in frame j are inverted) C: 101010 (inverts odd-numbered sign bits in frame j) d: 010101 (inverts even numbered sign bits in frame j)   As a result, the state Q of the selection controller 88 at the beginning of the frame k is_kIs 0, Only the convolutional code sequences A102 and B104 are valid sequences in frame k. Become. Conversely, the state Q of the selection controller 88_kWhen is 1, convolutional code sequence Only the cells C106 and D108 are valid sequences in the frame k. As mentioned above, The set representative sign bit is exclusive logical with each valid code sequence. Sum operation, and the encoded symbol Is done. Each candidate has a corset representative sign bit (or Or the coset element of the convolutional code identified by There is also. Each candidate is then combined with the symbol strength and the encoded encoded symbol Candidates are formed. These are sent to a filter 90 (FIG. 8) where each Is calculated and returned to the selection controller 88. The selection controller 88 , The encoded symbol sign bit that minimizes RFS for frame j Output   Current Q_kIs the selected encoded symbol sign for frame k Along with the convolutional code sequence of bits, by following the constraints of trellis diagram 100, , To determine the next state. For example, for frame k The state of the selection controller 88 is 1.   Spectral shaping performed using the present invention introduces look-ahead Can improve it. That is, based only on the current frame Instead of selecting the encoded symbol sign bit, the symbol size The bit selector 82 is a thin bit generated by the intensity mapper 50 (FIG. 4). Bol strength and the size of the coset representatives for current and future frames. Which encoded symbol sign bit is the best It can be determined whether to perform spectrum shaping. The V.90 standard was approved at the start. Ahead delay amount Specifies that up to three future frames can be used .   The spectral shaping quantity RFS based on the filter transfer function h (D) is Pre-fetch delay or depth for all possible paths (or candidate sequences) Is calculated by the filter 90. And the selection controller produces the lowest RFS The encoded symbol signature associated with the candidate sequence of the resulting frame k -Select a bit.   Referring again to trellis diagram 100 in FIG. 9, possible candidates for prefetch depth 1 The sequence will be described. State Q of selection controller 88 at the beginning of frame k_kIs 0 , The convolutional code sequences A102 and B104 are valid for frame k. It becomes Kens. However, the code sequence for frame k + 1 must also be considered. No. In frame k, code sequences A102 and B104 are valid. State Q at frame k + 1_{k + 1}Is 0 or h1, so the code sequence A102 ' , B104 ′, c106 ′ and D108 ′ are effective. As mentioned above, the Cosset representative sign The bits are exclusive ORed with each of the valid code sequences to Form candidates for loaded symbol sign bits. Pre-emption, each frame The k and k + 1 corset representative sign bits are the valid codes in each path of the trellis diagram. Exclusive OR operation in the sequence You. The candidate sequence in this example is D)}. The RFS for each sequence is determined and the frames in the determined sequence A candidate encoded symbol sign bit for k is selected.   Referring to FIG. 10, a flowchart 120 illustrates the operation of the symbol sign bit selector 82. Explain the work. In step 122, the selection controller 88 follows the trellis diagram. Symbol by changing the corset representative sign bit Generate a sign bit candidate (or candidate sequence in case of preemption). Next, in step 124, the selection controller determines the symbol strength and encoded sequence. Combined with the candidate for the symbol sign bit, the encoded encoded symbol Form a complement (or candidate sequence) and send it to filter 90. Step 126 In, the filter 90 determines the RFS of each candidate (or candidate sequence) and The RFS of the candidate (or candidate sequence) is sent to the selection controller 88. Last step At floor 128, selection controller 88 encodes an encoded sequence that minimizes RFS. Selects and encodes a candidate (or candidate sequence) for symbol sign bit Send the completed symbol sign bit.   Note that the present invention utilizes various convolutional codes G (D). These convolutions Codes are represented by different trellis diagrams and different convolutional code sequences . Various tatami mats Enlarging embedded codes and code sequences in the context of this specification is well known in the art. It is easy for people.   Generally, in the PCM, the filter 90 (eighth) is applied by the spectrum shaping method according to the present invention. The desired spectral shape is obtained by setting the response of By minimizing, the analog signal transmitted from the D / A converter in the CO60 (Fig. 5) Shape the spectrum of the log signal. The response h (d) of the filter 90 is the desired spectrum Defines the shape, which can be represented by: Where A (D) and B (D) are functions, and a and b are used to obtain a desired spectral shape. Is a real number selected for Also N_A, N_BRepresents h (D) for the numerator and denominator, respectively. Is the number of coefficients used to Calculate RFS for each symbol as follows: You can:   RFS based on the k-th frame is calculated by the following equation. Will be:Here, j is a symbol (time) index.   Spectral nulls at DC using the sign bit encoder of the present invention If generated, the RFS is the current digital sum (RDS) and the response h (D) of the filter 90 is Expressed as:     h (D) = 1-D (6)   The transmitted encoded symbol y_k, The filter 90 determines the transmitted encoding of the symbol time i. Find the RDS of a symbol with the following equation: Here, j is a symbol (time) index, which is based on the k-th frame. RDS is calculated by the following equation: Here, j is a symbol (time) index.   For preemption, if the preemption depth is △, the RDS is given by: However, LDRS is a preemptive RDS. Similarly, introduce preemption and generally minimize RFS can do. This is represented by the following equation; Sign bit decryption   The sign bit decoder 74 in the receiver 64 of FIG. 5 is shown in FIG. A column block 110 is provided. In the matrix block 110, the sign bit s_kThe matrix H^T _{xn (nr)}(In modulo 2) (ie, filtered) and the difference encoder Sign information bit ν_kTo recover.   Six transmitted symbols (n = 6) and one redundant bit (r = 1) per frame Matrix H^TIs shown in equation (11) as follows: The matrix H is the received signal s_kV from the error of_kWithin two or more decision errors Is set not to propagate This is because H is a finite impact response (FIR: finite im pulse response) type matrix because there is only one delay.   The sign bit encoder 52 (FIG. 4) Candidate encoded symbol sign bits generated for each frame In order to demonstrate that each of the Mathematically describe the process of reading and decoding Can be expressed mathematically: Also, the sign bit s_k(Encoding) can be expressed mathematically as follows: You can:       s_k= ν_kH^-t+ r_kG (13) S on the right side of equation (13)_kSubstituting into equation (12) instead of RU: G, H, and H are selected so as to satisfy the following conditions: (1) H, H = I (I is a unit matrix); and And (2) GH = 0. This gives r_kΝ regardless of the value of_k= V_kBecomes   In the above example, trellis diagram 100 (FIG. 9), a single redundant bit r_kIs a convolution Do G (D) = (1 + D11 + D11 + D1). 1 * r_k= R_kAnd D * r_k= R_k-1So r_kG (D) r_k(111111) + r_k-1It becomes equivalent to (101010). Where r_k-1Is training Squirrel diagram state Q_kAnd r_kRepresents a branch or path taken in the trellis diagram. 4 R convolutional code sequences A to D_k-1, R_kMapping to the expression is as follows Become:     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, code sequences A to D are represented by r_kIt can be regarded as G (D).   r_kHas no effect on how the information bits are decoded, so a different valid code sequence is used. Each set of n sign bits generated by the Can be generated. The result is a collection of n-sign bits that minimizes RFS / RDS. The case can be selected to perform the desired spectral shaping.Upstream PCM transmission   The spectrum shaping method according to the present invention performs precoding of upstream PCM transmission. Used in the same equivalence relationship as the analog PCM modem transmitter used for Wear. In this case, the response h (D) is between the transmitting modem and the central office (CO) line card. Represents the channel response to and from an analog-to-digital (A / D) converter, typically Modem front end, analog local loop and CO line card Including the effects of filtering.   By using the principle of the present invention, the A / D conversion in which the signal point is similar to the A / D quantization level is performed. Channel output sequence x (n) (pre If filtering is involved, z (n)) can be generated. In this case, the purpose is , The transmitted signal energy x (D) = y (D) / h (D) is minimized, and To keep conformational expansion low. Echo if conformational extension is large Quantization noise and other disadvantages. Therefore, it is not desirable that the conformation is expanded in this case as well.   In this application, the channel response h (D) is usually determined by the receiving modem, or The transmitter and receiver modems work together to determine the channel measurement performed at modem startup. It is determined based on the rules. Next, during data transmission, the transmitting modem transmits the incoming bit. Map to sequence x (D), which, after passing through the channel, Is converted to y (D). The channel response h (D) is usually determined to be at the minimum phase. But this is easily achieved, for example, by adding filtering in the transmitter. You.   The transmitter 40 'in FIG. 10 is an analog PCM model capable of performing upstream PCM transmission. The transmitter in the system. Transmitter 40 'utilizes the spectral shaping method of the present invention to Precoding incoming data bit stream 42 'using precoder 140 I do. One type of PCM upstream precoding (one-dimensional precoding for each symbol Precoding) is a United States Patent Application, Attorney Docket No. CX096044P02 , Title "Devices and Method for Precoding Data Signals", 1997, 1 Filed February 29, filed with inventors M. Vedat Eyuboglu, Pierre A. Humblet, and Daeyongu Kim It is disclosed in detail. The precoder 140 performs multi-dimensional precoding. That is, the symbols are precoded based on the frame. This execution example is one-dimensional What is precoding? Although different, the concept is similar to the one-dimensional case, and the details See continuation application.   The precoder 140 includes a serial-parallel converter 48 ', an intensity mapper 50', It comprises a sign bit encoder 52 'and a signal point selector 54'. this These components, with some changes, have the same reference numerals as in FIG. Set and operate similarly. For example, the operation of the sign bit encoder is It has been modified to perform precoding as shown below, Is the encoded symbol y instead of the encoded symbol itself._kCorresponding to N precoded levels x per frame_kIs output. n pre-coded records Bell x_kIs sent to a parallel-to-serial converter 142, which converts the pre-filter The pre-coded level is output to 144 in a serial format. Pre-filter 144 filters this level and outputs the filtered level to the digital-to-analog converter 14. 6 is output. This converter 146 precodes on analog channel 148 Level x_kSend Channel is precoded level x_kChange the ideal In general, the coding symbol y in the quantizer in the central office (CO) 150_kTo Generate corresponding levels. In other words, the precoder is on analog channel 1 48 responses, more precisely quantized by canceling the desired channel response h (n) The desired coded symbol y in the device_kThe level corresponding to The precoded level x to generate_kGenerate   The desired channel response h (n) is c (n), the response of analog channel 148. It is equal to the response g (n) of the prefilter 144 that has been subjected to the convolution operation. Where n is a symbol It is an interval index and h (0) = 1. The relationship can be expressed as:       y (n) = h (0) × (n) + h (1) × (n-1) + ... h (N_h) × (n-N_h) (15) Since h (0) is set equal to 1, equation (15) can be simplified as follows: You can simply:   The value of h (n) at a specific time is known, and the past value of x (n) is also known. No. The filter 90 in FIG. 8 calculates the addition term of equation (16) as RFS and selects it Give to the controller 88. The past value of x (n) is based on the known relationship x (D) = y (D) / h (D). Is determined from the previous symbol y (n), stored in the filter 90, and LA 88 is encoded according to flowchart 120 of FIG. 10 to minimize x (n). It operates by selecting the symbol sign bit. Encoded symbol y (n) Instead of sending Sent level is sent.   In addition to the mapping operation described above, an echo cancellation mode that separates two transmission directions Module and symbols are transmitted in sync with the network clock. It is necessary to provide a timing interpolation filter to be realized. This timing interpolation Filters are usually driven by clock recovery circuits used in downstream receivers . The transmitter is primarily responsible for limiting the transmission bandwidth to almost 4kHz, Also has a linear filter that performs the necessary pre-filtering to minimize the entire channel response h (D) Sometimes.   Furthermore, in a real system, a trellis to increase the immunity of noise ・ It may include the form of coding. For example, in the applications cited above "Device for, and Method o" f 、 Communicating According to a Trellis Code of Baseband Signals Chosen November 1, 1996 entitled "From a Fixed Set of Baseband Signal Points" U.S. patent application Ser. No. 08 / 479,040 filed on Apr. 4 (Attorney Docket No. Can be used. This specification is hereby incorporated by reference in its entirety. Cis The operation of the system is essentially unaffected by trellis coding.   Of course, the present invention is not limited to computer disks or memory chips. Software and / or firmware stored on any computer media Realized in a can do. When the present invention is embodied in software / firmware The invention may be implemented in a carrier that is transmitted electrically, for example, via the Internet. It may take the form of a computer data signal to be represented. The present invention may be embodied in other specific forms without departing from its spirit or essential characteristics. Can be The disclosed embodiments are illustrative in all respects. It is nothing more than a thing to impose. Therefore, the scope of the present invention The enclosure is indicated by the appended claims rather than the above description. Equivalent to the claim All changes falling within the meaning and range are intended to be embraced within the scope.

────────────────────────────────────────────────── ─── Continuation of front page    (81) Designated country EP (AT, BE, CH, CY, DE, DK, ES, FI, FR, GB, GR, IE, I T, LU, MC, NL, PT, SE), OA (BF, BJ , CF, CG, CI, CM, GA, GN, ML, MR, NE, SN, TD, TG), AP (GH, GM, KE, L S, MW, SD, SZ, UG, ZW), EA (AM, AZ , BY, KG, KZ, MD, RU, TJ, TM), AL , AM, AT, AU, AZ, BA, BB, BG, BR, BY, CA, CH, CN, CU, CZ, DE, DK, E E, ES, FI, GB, GE, GH, HU, ID, IL , IS, JP, KE, KG, KP, KR, KZ, LC, LK, LR, LS, LT, LU, LV, MD, MG, M K, MN, MW, MX, NO, NZ, PL, PT, RO , RU, SD, SE, SG, SI, SK, SL, TJ, TM, TR, TT, UA, UG, UZ, VN, YU, Z W (72) Inventor Tan, Chien-Cheng             Month, Massachusetts, United States             Field, Francis Avenue 2509 [Continuation of summary] Coset representative to define coset representative elements for A generator; and a corset representative sign bit and The encoded symbol size in response to the From the candidates for the Encoded symbols that generate transmit encoded symbols Symbol sign bit to select sign bit -It is constituted by a selector.

Claims (1)

[Claims]   1. The transmitter encodes each frame into a predetermined number of coded symbols, Digital information bits transmitted over the network to the receiver In this system, the coded symbols to be transmitted have a desired spectrum. And the digital information bits have a first predetermined number of pieces of intensity information per frame. Bits and a second predetermined number of sign information bits, wherein the transmitter:   Strength mapping the strength information bits to a predetermined number of symbol strengths per frame Degree mapping device;   The sign information bits are converted into a predetermined number of encoded symbol sizes per frame. A sign bit encoder that encodes the input bits; and   In response to the intensity mapping device and the sign bit encoder, The symbol strength and the encoded symbol sign bit are combined to form a file. A signal point selector which forms a predetermined number of coded symbols to be transmitted per frame; Wherein the sign bit encoder comprises:     Generates a corset representative sign bit of the sign information bit for each frame. Code for the convolutional code containing the candidate encoded symbol sign bits. A coset that defines a coset for the convolutional code that identifies the set A representative generator; and     In response to the corset representative sign bit and the symbol strength, Have the desired spectral shape from the encoded symbol sign bit candidates The encoded symbol sign bit to generate the transmitted encoded symbol Symbol sign bit selector to select;     A system characterized by comprising:   2. The first predetermined number of intensity information bits is m bits, and the second predetermined number of N information bits are nr bits, and n is the predetermined number of symbols per frame. And r is the number of redundant bits used by the sign bit encoder 2. The system according to claim 1, wherein   3. The corset representative generator is provided to the sign bit encoder. The predetermined bit position of the nr bits is differentially encoded to make the polarity inversion unchanged. The system of claim 2, comprising a realizing differential encoder.   4. The intensity mapper maps the m intensity information bits to a modulus transform mapping 4. The method according to claim 3, wherein mapping is performed on n symbols for each frame by using a method. The described system.   5. The corset representative generator converts the nr sign information bits into a matrix H^-T And a matrix block for generating n representative sign bits for each frame. Specially 5. The system according to claim 4, wherein the system comprises:   6. The symbol sign bit selector comprises a selection controller and a filter. Wherein the selection controller comprises:     From the corset representative sign bit, the encoded symbol sign .Logic for generating bit candidates;     The encoded symbol sign bit candidates and the symbol strength And logic for combining encoded data to form encoded encoded symbol candidates. Ruta is:     Responsive to the encoded encoded symbol candidate, For each of the encoded symbol candidates     Logic for determining RFS, and RFS for each of the encoded encoded symbol candidates To the selection controller, wherein the selection controller:     The RFS determined for each of the encoded encoded symbol candidates In response to the determined RFS is the smallest encoded encoded symbol candidate Logic to select the associated encoded symbol sign bit;   The system of claim 5, further comprising:   7. The symbol sign bit selector uses prefetching and selects Comprising a roller and a filter, wherein the selection controller comprises:     From the corset representative sign bit, encode Logic for generating a candidate sequence of used symbol sign bits;     Combining the candidate sequence with the symbol strength to produce an encoded code Wherein the filter forms a symbolized symbol candidate sequence, wherein the filter comprises:     Responsive to the encoded encoded symbol candidate sequence, Logic for determining an RFS for each of the loaded encoded symbol candidate sequences; Selecting the RFS of the encoded coded symbol candidate sequence into the selection controller And the logic to provide to the rollers, wherein the selection controller comprises:     Determined for each of the encoded coded symbol candidate sequences The determined RFS is the smallest encoded coding symbol in response to the RFS Logic for selecting the encoded symbol sign bit from a candidate file.     The system of claim 5, further comprising:   8. RFS for the i-th symbol is determined by the filter by the following equation: The system of claim 7, wherein:   9. RFS based on frame for kth frame 9. The method according to claim 8, wherein the filter is determined by the following equation. System. Here, j is a symbol (time) index.   10. Wherein said RFS is determined by preemption as: Item 10. The system according to Item 9. Where △ is the preemption depth.   11. The receiver converts the transmitted encoded symbol to the encoded symbol. Intensity and sine-ex A tractor and the encoded symbol sign bit into the sign information And a sign bit decoder for decoding the bits. The system of claim 0.   12. The sign bit decoder is adapted to generate the encoded symbol size. Matrix matrix having a matrix H that is multiplied by The system of claim 11, comprising a lock.   13. Differentially decoding a predetermined bit position of the recovered sign information bit The system of claim 12, further comprising a differential decoder.   14． In response to the signal point selector, Octet converter for transmitting octets corresponding to coded symbols The system according to claim 13, wherein   15. n equals 6, r equals 1, and the matrix H is defined as 14. The system according to claim 13, wherein:   16. The method of claim 15, wherein the convolutional code is defined as: system.     G (D) = [1 + D 11 1 + D 11 1 + D 1]   17． 17. The method of claim 16, wherein the matrix H is defined as: system.   18. The RFS is a current digital sum (RDS), and the RDS of the i-th symbol Is determined by the filter as follows: system. Here, j is a symbol (time) index.   19. The RDS is equivalent to the kth frame with respect to the frame. 19. The method according to claim 18, wherein the value is determined by the class selector as follows: On-board system. Here, j is a symbol (time) index.   20. The RDS is determined using the prefetching as follows: The system of claim 19. Where △ is the preemption depth.   21. From the transmitter, each frame is encoded into a predetermined number of coded symbols, Digital information bits transmitted over a network to a receiver Transmitting as a reference, wherein the coded symbols have a desired spectral shape. Wherein the digital information bits comprise a first predetermined number of intensity information bits for each frame. Divided into a second predetermined number of sign information bits, the method comprising:   A step of mapping the strength information bits to a predetermined number of symbol strengths for each frame Floor;   The sign information bits are converted into a predetermined number of encoded symbol sizes per frame. Encoding the bits into binary bits; and   Combining the symbol strength with the encoded symbol sign bit Forming a predetermined number of coded symbols to be transmitted for each frame, The loading stage is:     The corset representative sign bit of the sign information bit for each frame And generate a convolutional code containing the encoded symbol sign bit candidates. Defining a corset for the convolutional code identifying the corset; and     Utilizing the corset representative sign bit and the symbol strength, Have the desired spectral shape from the encoded symbol sign bit candidates The encoded symbol sign bit to generate the transmitted encoded symbol Selecting;     A method characterized by comprising:   22. The first predetermined number of intensity information bits is m bits, and the second predetermined number of The sign information bits are nr bits, and n is the predetermined number of symbols for each frame. And r is the number of redundant bits used in the encoding step. 22. The method according to claim 21, wherein the method comprises:   23. The generating step differentially encodes a predetermined bit position of the nr bits. 23. The method according to claim 22, further comprising the step of: Method.   24. The step of mapping is performed using a modulus transform mapping method. For each m intensity information bits, The method according to claim 23, wherein the mapping is performed on a bol.   25. The generating step multiplies the nr sign information bits by a matrix H; Generating an n-coset representative signature for each frame. The method according to claim 24, wherein   26. The selection steps include:   From the coset representative sign bit, the encoded symbol sign Generating candidate bits;   The encoded symbol sign bit candidates, the symbol strength and Combining to form encoded encoded symbol candidates;   An RFS for each of said encoded encoded symbol candidates; For sending an RFS for each of the candidate encoded symbols to the selection controller. Determining the nature of the agreement; and   The encoded coded symbol candidate with the smallest determined RFS. Selecting a contiguous encoded symbol sign bit;   26. The method of claim 25, comprising:   27. The selection step utilizes preemption, and further:   From the corset representative sign bit candidates, the encoded symbol size Generates a candidate sequence of bits To do;   Combining the candidate sequence with the symbol strength to produce an encoded code Forming a symbolized symbol candidate sequence;   An RFS for each of said encoded encoded symbol candidate sequences; Selecting said RFS for each of said encoded encoded symbol candidate sequences. Determining logic to send to the controller; and   The determined RFS is encoded from the smallest encoded encoded symbol candidate. Selecting a coded symbol sign bit;   26. The method of claim 25, comprising:   28. The RFS for the ith symbol is determined by the filter using the following equation: 28. The method of claim 27, wherein:   29. The RFS is based on the k-th frame with respect to the frame, 29. The method according to claim 28, wherein the method is determined. Here, j is a symbol (time) index.   30. Wherein said RFS is determined by preemption as: Item 30. The method according to Item 29.Where △ is the preemption depth.   31. In the receiver, the encoded symbol to be transmitted is encoded Separation into symbol sign bit and symbol strength, The method further comprises the step of decoding a bol sign bit into the sign information bit. 31. The method of claim 30, wherein:   32. The decoding step comprises matrixing the encoded symbol sign bits Multiplying by H and recovering the sign information bits. 32. The method of claim 31.   33. A step of differentially decoding a predetermined bit position of the recovered sign information bit The method of claim 32, further comprising a floor.   34. Transmitting an octet corresponding to the coded symbol on the network; 34. The method of claim 33, further comprising the step of:   35. n equals 6, r equals 1, and the matrix H is defined as 34. The method of claim 33, wherein:   36. 36. The method according to claim 35, wherein the convolutional code is defined as: Method.     G (D) = [1 + D 11 1 + D 11 1 + D 1]   37. The method of claim 36, wherein the matrix H is defined as: Method.  38. The RFS is a current digital sum (RDS), and the RDS of the i-th symbol 28. The method of claim 27, wherein is determined as: Here, j is a symbol (time) index.   39. The RDS is calculated for each frame with respect to the k-th frame as follows: The method of claim 38, wherein the method is determined. Here, j is a symbol (time) index.   40. The RDS is determined using the prefetching as follows: 40. The method of claim 39. Where △ is the preemption depth.   41. At the transmitter, the digital information bits are framed on a frame basis. A system for precoding a predetermined number of precoded levels for each system, The digital information bits comprise a first predetermined number of intensity information bits and a second predetermined number of The sign information bits are divided into:   Strength mapping the strength information bits to a predetermined number of symbol strengths per frame Degree mapping device;   The sign information bits are converted into a predetermined number of encoded symbol sizes per frame. Sign bit encoder that encodes the input bits;   In response to the intensity mapping device and the sign bit encoder, The symbol strength is combined with the encoded symbol sign bit to form a frame. A predetermined number of coded symbols are formed for each program, and the programs corresponding to the coded symbols are formed. A signal point selector for outputting a recorded level; -Bit encoder:     Coset cost of sign information bit for each frame Generates table sign bits and suggests encoded symbol sign bits Define a coset for the convolutional code that identifies the coset of the containing convolutional code A corset representative generator; and     In response to the corset representative sign bit and the symbol strength, The encoded symbol / sign from the encoded symbol / signature candidate. A symbol sign bit selector for selecting a bit;     A system characterized by comprising:   42. At the transmitter, the digital information bits are framed on a frame basis. Precoding a predetermined number of precoded levels for each system, The total information bits are a first predetermined number of intensity information bits and a second predetermined number of size bits for each frame. And the information bits are divided into:   A step of mapping the strength information bits to a predetermined number of symbol strengths for each frame Floor;   The sign information bits are converted into a predetermined number of encoded symbol sizes per frame. Encoding to a bit;   The symbol strength is combined with the encoded symbol sign bit to A predetermined number of encoded symbols are formed for each frame, and the symbols corresponding to the encoded symbols are formed. Outputting the pre-recorded level;   Wherein said encoding step comprises:     For each frame, the Generates representative sign bits and provides an indication of the encoded symbol sign bits. Define a coset for the convolutional code that identifies the coset of the convolutional code including the complement. Defining; and     Using the corset representative sign bit and the symbol strength, Encoded symbol sign bit from encoded symbol sign candidate Selecting a project;     A method characterized by comprising:   43. From the network, a predetermined number of coding symbols are Receiving digitally encoded digital information bits on a frame basis The transmitted symbol has a desired spectral shape, and The digital information bits are composed of a first predetermined number of intensity information bits and a second predetermined number of And the receiver information bit is divided into:   The transmitted coded symbol is encoded with an encoded symbol sign bit. Strength and sign extractor separated from coded symbol strength;   An intensity demodulator for decoding the encoded symbol intensity into the intensity information bits. Par: and   Decodes the encoded symbol sign bit into the sign information bit Sign bit decoder to perform;   And the sign bit decoder is The encoded symbol sign bit is multiplied to obtain the sign information bit. A receiver comprising a matrix block having a matrix H to recover.   44. 44. The receiver of claim 43, wherein H is defined as: