KR100349987B1 - High speed communications system for analog subscriber connections - Google Patents

Abstract

The present invention discloses a new data communication system that enables data transmission at a rate higher than is possible using existing methods over existing telephone lines. By using a new asymmetric configuration for the two communication points of the two sides, the current maximum accepted theoretical limit on the data rate is no longer acceptable. Communication One side of the cord is connected directly to the digital telephone network, while the other side uses a common dial-up connection. This reduces transmission problems for a single telephone interface and compensation for a single analog subscriber line. Means for providing such compensation and reduced clock synchronization are also generated, enabling a practical implementation of the system. The new system can achieve rates up to 64,000 bits per second and has wide availability in most active areas, including broadband audio transmission, video transmission, networking, facsimile transmission and remote computer access.

Description

TECHNICAL FIELD [0001] The present invention relates to a method for training a high-speed modem,

Data communication plays an important role in many areas of modern society. Banking transactions, facsimiles, computer networks, remote database access, credit card validation, and many other applications all depend on the ability to move digital information quickly from one point to another. This transmission rate directly affects the quality of such services, and in many cases, the application is impossible to realize without some significant underlying capability.

At the lowest level, most of this digital data traffic is carried through the telephone system. Computers, facsimile machines, and other devices often communicate with each other via dedicated lines that share common dial-up or many similar characteristics. In any case, the data must first be converted into a form compatible with the telephone system originally designed for voice transmission. On the receiving side, the telephone signal must be converted back into a data stream. These two tasks are performed by the modem.

The modem may perform two tasks corresponding to the above requirements; And performs demodulation to reconstruct the data stream by taking the modulation and voice signals that transform the data stream into voice signals that can be carried by the telephone system. A pair of modems, a modem located at each end of the connection, allows bi-directional communication between two points. Restrictions on voice signals cause a restriction on the transmission rate at which data can be transmitted using the modem. Such constraints include limited bandwidth and degradation of data due to noise and cross talk. The telephone system is generally capable of carrying only signals belonging to the frequency range between 300 Hz and 3,400 Hz. Signals out of this range sharply decrease. This range is used in the design of telephone systems because it occupies a significant part of the human voice spectrum. However, the bandwidth of the channel is a factor in determining the maximum achievable data rate. If all the remaining factors are constant, the data rate is directly proportional to the bandwidth.

Another factor is distortion of the speech signal or any other signal that the communication endpoint can not control. This includes electrical pick-ups for noise induced by other signals (crosstalk) carried by the telephone system, electrical noise, and the conversion of signals from one form to another. The final type will be described in the detailed description.

For general use, the modem is designed to be operable through most dial-up connections. Thus, the modem should be designed for the worst case, which includes bandwidth limitation and considerable noise that can not be eliminated. Even so, substantial progress has been made in modem design over the past several years. Devices capable of operating at elevated speeds up to 28,800 bits per second are currently routinely available. See International Telecommunication Union, Telecommunication Standardization Sector (ITU-T), Recommendation V.34, Geneva, Switzerland (1994), which is incorporated herein by reference. However, the theoretical issue based on channel bandwidth and noise level shows that the maximum possible rate is nearly achieved and that no significant further increase is possible under given constraints. This is illustrated in C.E. Shannon, " A Mathematical Theory of Communication, " Bell System Technical Journal 27: 379-423, 623-656 (1948).

Unfortunately, even though speeds approaching 30,000 bits per second (or 3,600 bytes per second) are feasible for many data communication applications, generic modem transmission is still not fast enough for all uses. At such a rate, the transmission of text is fast and acceptable for low-quality speech such as digitized speech. However, while facsimile or still image transmission is slow, high quality audio is limited and full-motion video is not satisfactorily achieved. In summary, increased data transmission capability is required. This is indispensable for new applications and is essential to maximize the performance of many current applications.

Of course, telephone companies, cable TV providers and other companies do not ignore this increased data transmission demand. One method of eliminating the need for additional modems and providing increased speed of data access for businesses and homes is to use an end-to-end (or end-to-end) digital connection . The way to provide such services is the Integrated Services Digital Network (ISDN). (ISDNs), " Recommendation I 120, Geneva, Switzerland (1993) and John Landwehr ", all of which are incorporated herein by reference. The Golden Splice: Beginning a Golbal Digital Phone Network ", by Northwestern University (1992) ISDN replaces existing analogue subscriber lines with 160,000 bits per second digital connections Most inter-office and inter-office calls Such digital subscriber lines may be used for digital voice, computer data or any other type of information transmission between one end and the other end, since the traffic is already digitally performed. However, A particular device must be installed at both ends of the line. A change from a re-voice transmission network to a general data transmission service is performed, and voice is only a specific data form.

Once installed, each basic ISDN link provides two data channels capable of 64,000 bits per second, a control channel with 16,000 bits per second capability, reduced call connection time, and other advantages. At this rate, facsimile and still image transmission is almost instantaneous, high quality voice is feasible, and remote computer access has the benefit of a 5x speed increase. Any advance toward full motion video can also be achieved.

The disadvantage of ISDN is its availability or its lack. In order to use ISDN, the user's central station must be upgraded to provide such services, and the user must replace the existing (such as a telephone) device with a digital device, and each individual line interface at the central office has a digital data stream . This last step, the conversion of numerous analog connections between each phone and central station into a digital link, is vast. The size of such work stipulates that ISDN development is slow and coverage is sporadic for a given amount of time to flow. Rural areas and sparsely populated areas can not afford these services.

Another existing infrastructure that can potentially provide high-speed data communication services is cable television systems. Unlike telephone systems that are connected to users over low bandwidth, cable systems provide high bandwidth for a significant portion of the connection. The unused capacity on these interconnects can provide data rates of tens or even hundreds or millions of bits per second. This is more than adequate for all of the above planned services, including full motion digital video. However, cable systems have serious problems, that is, problems with network structure. The telephone system provides point-to-point connectivity. That is, each user totally uses the full capabilities of the user's access, it is not shared with other users and is not directly disturbed by usage by other users. Cable systems, on the other hand, provide broadband connectivity. Since the same signal appears on each user connection, the entire capability is shared by all users. Thus, even though the overall capability is high, this is divided by the number of user requested services. This structure is effective when all users require the same design data, such as television distribution, for the original design purpose of the cable, but it is not effective for communication of users with different data needs. In a metropolitan area, the data capabilities available to each user are significantly less than with ISDN or modem connections.

In order to provide high-speed data access to a large number of users, cable systems can be modified to separate different segments for a user population that efficiently share cable bandwidth with a small population. However, similar to ISDN, this will be a slow and costly process that only provides partial services for years to come.

The methods used to design the modem are based heavily on models of telephone systems that have remained unchanged for decades. That is, the modem is modeled as an analog channel with a limited bandwidth (400-3400 Hz) and an additional noise component as low as 30 db below the signal level. However, most of the telephone systems currently use digital transmission of sampled representations of analog waveforms for inter-office communication. In each central station, the analog signal is converted to a PCM (pulse width modulation) signal of 64,000 bits per second. The receiving office then reshapes the analog signal before placing the signal on the subscriber line. For the first approximation, the noise induced by this generation is similar to that observed on an analog system, and the noise source is quite different. K.Pahlavan and J.L., et al. Holsinger et al., &Quot; A Model for the Effect of PCM Compandor on the Performance of High Speed Modems ", Golbecom'85, pp. 758-76 (1985). Most of the noise found in dial-up using the digital switching scheme is based on the quantization by the analog-to-digital converter required to convert the analog waveform to a digital waveform.

As noted above, most dial-up connections are now digitally performed between central offices at a rate of about 64,000 bits per second. Moreover, the ISDN service proves to be able to transmit significantly above this rate over the subscriber line. It has been proposed that it is also possible to design a transmission system using these factors. Kalet et al. Assume a system as shown in Fig. 2 where the transmitting side chooses the correct analog level and timing so that the analog-to-digital conversion occurring at the central station of the transmitter can be achieved without quantization errors. I.Kalet, J.E. Mazo, and B.R. Saltzberg, " The Capacity of PCM Voiceband Channels ", IEEE International Conference on Communications'93, pp. 507-511, Geneva, Switzerland (1993). J.E. Using mathematical results of Mazo, it is inferred that it is theoretically possible to reconstruct digital samples using only the available analog levels at the receiving end of the second subscriber line of the communication path. As cited in the present specification, J.E. Mazo's " Fast-Than-Nyquist Signaling ", Bell System Technical Journal, 54: 1451-1462 (1975). The resulting system can achieve a data rate of 56,000 to 64,000 bits / sec. The disadvantage of this system is that it is a theoretical possibility that may or may not be feasible. Kalet et al. "It is a difficult implementation problem and we only guess whether a reasonable solution is possible." Id. On page 510.

Examples of conventional attempts to solve the foregoing problems are found in the invention by Ohta, U.S. Patent Nos. 5,265,125 and 5,166,955, which are incorporated herein by reference. Ohta discloses an apparatus for reconstructing a PCM signal transmitted over a communication channel or reproduced from a recording medium. These patents exemplify many prior arts in the literature to address the general problem of reconstructing multi-valued signals transmitted over a distortion channel. See, for example, "Data Communication Principles" Plenum (1992) by Richard D. Gitlin, Jeremiah F. Hayes, and Stephen B. Weinstein et al., Which are incorporated herein by reference. However, such conventional techniques do not consider the application of the method of handling the output from the non-linear quantizer nor handle specific problems with the decoded digital data transmitted over the telephone subscriber line. Moreover, the problem with reconstructing the sampling rate clock from PCM data is not trivial when the PCM signal takes more than two values. For example, in Ohta's patent, a simple clock regeneration scheme that relies on a binary input signal is used. This type of clock reproduction can not be used with multivalued codes used in telephone systems. Also, compensation for changing drift and line conditions over time requires the use of adaptive systems not included in the prior art for PCM reconstruction.

Thus, there is a critical imbalance between required or preferred data communication capabilities and available data communication capabilities. Existing modems do not provide adequate capabilities, and new digital access solutions are years away from general availability. Modifying an existing infrastructure with ISDN would require a significant amount of work and would take decades until its use spread. The new method for data transfer not only makes many current applications immediately useful, but otherwise makes the service available for a variety of new services that must wait until the infrastructure meets their needs .

Thus, there is a need to provide a new system for data transmission that provides the ability to receive data at a high rate over existing telephone lines.

There is also a need for an improved data transmission system that enables systems, devices and applications designed for digital telephone systems (such as ISDN) to be used with analog connections.

There is a need for an improved data transmission system that makes it possible to use the digital infrastructure of the telephone system without costly replacing all subscriber lines.

It is also desirable to create a high speed communication system that provides a means for distributing high quality digital voice, music, video or other elements to users. Such an improved data transmission system advantageously provides a means for distributing customized (individually ordered) information, data or other digital elements to a large number of users.

There is a need for a high speed communication system that provides increased efficiency for commercial applications such as facsimile, point-of-sale, remote list management, credit card validation, and remote computer networking.

FIELD OF THE INVENTION The present invention relates generally to data communication devices, and more particularly to devices for transmitting digital data over a telephone connection.

1 is a block diagram illustrating a general modem data connection;

2 is a block diagram illustrating an embodiment of a home symmetric digital system;

3 is a block diagram illustrating a fast dispense system in accordance with aspects of the present invention;

4 is a block diagram of a hardware configuration of the encoder 150 of FIG. 3, in accordance with an aspect of the present invention.

FIG. 5 is a block diagram illustrating the functionality of the encoder 150 of FIG. 3, in accordance with an aspect of the present invention.

Figure 6 is a block diagram illustrating the functionality of the DC remover 184 of Figure 5, in accordance with an aspect of the present invention.

7A is a graph of a time-related data stream 100 applied to an encoder 150, in accordance with an aspect of the present invention.

FIG. 7B is a graph of the output of a typical encoder 150, in time, applied to the digital network connection 132 of FIG. 3, in accordance with an aspect of the present invention.

FIG. 7C is a graph of the time-dependent linear value 194 of FIG. 6, which is the output signal from the encoder 150 after conversion to a linear form in accordance with aspects of the present invention.

Figure 8 is a block diagram illustrating the functionality of an existing digital line interface to understand features of the present invention.

9 is a block diagram of a hardware configuration of decoder 156 shown in FIG. 3, in accordance with an aspect of the present invention.

Figure 10 is a block diagram illustrating the functionality for the decoder 156 of Figure 3, in accordance with an aspect of the present invention.

11A is a graph of the analog signal 154 of FIG. 10 relative to time, in accordance with an aspect of the present invention.

FIG. 11B is a graph of the compensation signal 274 of FIG. 10, formed in the decoder 156 and related to time, in accordance with aspects of the present invention.

FIG. 11C is a graph of the estimated code stream 280 of FIG. 10, formed in the decoder 156 and related to time, in accordance with aspects of the present invention.

11D is a graph of the data stream 126 of FIG. 3 generated by the decoder 156 and related to time in accordance with aspects of the present invention.

FIG. 11E is a graph of the error signal 272 of FIG. 10 generated by the decoder 156 and related to time, in accordance with aspects of the present invention.

Figure 12 is a block diagram illustrating an inverse filter 268 of Figure 10, in accordance with an aspect of the present invention.

FIG. 13 is a block diagram illustrating the feed-forward equalizer 300 of FIG. 12, in accordance with an aspect of the present invention.

Figure 14 is a block diagram illustrating the filter tab 300 of Figure 13, in accordance with an aspect of the present invention.

15 is a block diagram illustrating the clock estimator 264 of FIG. 10, in accordance with an aspect of the present invention.

Figure 16 is a block diagram illustrating the functionality of the clock synchronizer 260 of Figure 10, in accordance with an aspect of the present invention.

17 is a block diagram illustrating an end-to-end asymmetric system using an inversion channel in accordance with aspects of the present invention.

18 is a block diagram showing an application of a feature of the present invention using a database server;

19 is a block diagram showing features of the present invention applied to a high-speed facsimile system;

Figure 20 is a block diagram illustrating a digital telephone relay in accordance with a feature of the present invention.

One aspect of the present invention includes a system for transmitting data over existing dial-up connections at an increased rate than data transmissions of known modems or known methods. The present invention provides two important observations:

1. The underlying telephone system is digital using PCM transmission,

2. The high-speed data rate is only required in one direction and the source of it achieves a significant improvement over the conventional method by using observation information indicating direct digital access to the telephone system.

One aspect of the present invention utilizes the observation information described above to achieve a higher data rate than can be achieved using conventional systems. The second observation information described above asserts the maximum use of the modem to access and retrieve information from the central server. In addition, the invention has been found to be particularly useful for applications requiring high data rates, such as database access and on-demand video or audio. Such an application can be realized using a high data rate achievable through the present invention.

An important characteristic of the present invention is that it is simple and extremely effective; That is, allowing the user to directly connect the data provider to the digital telephone network while using his or her existing analog connection without changing the line.

Such a configuration greatly changes the model in which the user's data device must operate. Existing modems must deal with a large number of unequal noise sources that will vary the performance of the signal through bandwidth limitations and the entire transmission path. On the other hand, the characteristics of the present invention carry data digitally through most paths from the central office to the user's home or office, and convert the data to analog form only for the last segment of the path. Because such a conversion is no longer required, useful quantization noise during one of the major noise sources for an existing modem, i. E. Analog-to-digital conversion, is completely eliminated. Moreover, the quantization noise during digital-to-analog conversion is modeled as a deterministic phenomenon and is therefore considerably reduced.

By using the above-described characteristics of the present invention, a data source having direct access to a digital network (e.g., by ISDN) can send accurate data to a central station serving data to the consumer. All of the required devices are devices on the user side of the subscriber line that are due to the filtering performed in the digital-to-analog converter of the central station and compensate for the distortion of the data signal due to the transmission line. These two distortions can be handled appropriately using existing digital signal processing hardware as described herein.

Although this method can not be used to feed back data from a user to a server, existing modems that provide an asymmetric channel with 20,000 to 30,000 bits / second capability with up to 64,000 bits / It should be noted.

The characteristics of the present invention enable a certain type of digital data (such as audio, video, information, etc.) to be transmitted at an improved rate than can be obtained using a conventional modem or conventional data transmission to an individual user It is considered. Moreover, unlike cable television distribution systems, the present invention can serve any number of users simultaneously requesting different data at a full data rate.

In addition to providing increased operating speeds for existing applications such as remote computer access, high speed facsimile transmission, etc., certain features of the present invention enable a number of new applications. This includes high quality audio or music transmission, video on demand, still picture transmission, video phones, teleconferencing or similar applications requiring high data rates.

Another aspect of the present invention is to reconstruct multi-valued PCM data signals from the analog representation of the signal. This is accomplished using a new scheme that combines new clock synchronization techniques with adaptive equalization.

In addition to the foregoing description, other features and advantages of the present invention are (1) the ability to effectively reconstruct a digital pulse-code modulated (PCM) data stream of a telephone system using only analog signals at the subscriber side of a telephone line; (2) the ability to reconfigure the clock frequency and phase of the PCM data using only analog signals at the subscriber side of the telephone line; (3) the ability to increase the effective data rate between the central office and the subscriber without modifying the telephone system or adding additional equipment to the central office; And (4) reconstructing the digital data after the digital data has been modified due to performance changes due to the conversion to one or more analog forms, filtering, distortion, or the addition of noise.

The above-described techniques of the present invention can be easily understood by considering the following detailed description with reference to the drawings.

Conventional modem data connection

A conventional modem data connection is shown in FIG. The operation of such systems is well known and standardized by an organization such as the International Telecommunication Union. Depending on the modem 104 and the modem 124, data is applied at a rate of up to 28,800 bits / second through the first user data stream 100. [ The modem 104 converts the data stream 100 to an analog signal, which is applied to the subscriber line 106, which is connected to the telephone switch 108. The analog signal then travels over the telephone network 114 via the network connection 112 and reaches the telephone exchange 120 via the network connection 118 to eventually serve the second subscriber. The signal is then transmitted in analog form via the subscriber line 122 to the second subscriber modem 124 and the modem 124 converts the signal into a data stream 126 that is a delayed version of the data stream 100 Conversion. The data stream 128 is transmitted over the telephone network to the modem 124, the subscriber line 122, the telephone exchange 120, the network connection 116, the telephone network 114, the network connection 110, (104) that forms a delayed version of the data stream (108), the subscriber line (106), and the data stream (102).

Such a system assumes that the telephone system reproduces the analog signal applied at the one-user telephone connection on the other user side with distortion and delay not greater than the set of standard values prescribed for the telephone system. One method shows that it is impossible to transfer data at a rate of approximately 35,000 bits per second or more based on these values. These systems ignore many detailed distortions that are actually a significant change to the signal, rather than an unexpected change. One such critical change is the quantization noise when the telephone network 114 is implemented digitally. Existing modems do not recognize these important noise sources in removing distortion, and thus limit their data rate. This is a theoretical limit to the key disadvantages of existing modem systems: low data rates and the greatest possible improvement within the boundaries of current homes.

To overcome the above-mentioned disadvantages and disadvantages of the conventional modem data connection shown in FIG. 1, the method for increasing the data transmission rate creates a hypothetical symmetric digital communication system. Such a system is shown in Figure 2 in combination with a digital telephone network.

Such a system, disclosed by Kalet et al. Is similar to an existing modem, except that the underlying infrastructure is the digital telephone network 134. Except that the signals are carried in digital form inside the digital telephone network 134 and in the digital network connection 138, the digital network connection 132, the digital network connection 136 and the digital network connection 138 Is similar to the operation of the above-described conventional modem system. Each user is required to transmit information to the telephone exchange 120 and the telephone exchange 108 via the subscriber line 122 and the subscriber line 106 respectively and to the analog and digital telephone network 134 at the exchange The conversion between the standard digital formats used by the system is performed.

Unlike conventional modems, no theoretical dispute has been found to limit the speed of such systems to less than 56,000 or 64,000 bits per second, generally less than that used internally in the digital telephone network 134. Therefore, it is hypothetically possible for such a system to obtain a data rate of up to 64,000 bits / sec. However, there is no evidence that such a system is actually feasible, nor can it implement such a system. The author of this system says, "This is a real problem, and even if a reasonable solution is possible, we can only guess."

The problem is that the transmission modem has to encode the signal by controlling the digital level selected by the network only through its analog output in order to take advantage of the knowledge that the underlying network is digital and most of the observed signal distortion is based on quantization noise will be. In addition, the receiving modem should only extract these digital levels correctly through its analog inputs. Distortion due to analog-to-digital conversion occurs at both the transmitter and receiver sides, but only the desired signal is added and the combined distortion is directly observable. Moreover, additional distortion due to electrical noise and crosstalk also occurs on subscriber line 122 and on subscriber line 106. It is difficult and possibly impossible to separate these distortion components from desired signals.

A feature of the present invention is a method for eliminating the disadvantages of this method. The method provides a data rate that can be obtained higher than is possible using any other known solution, using the basic digital networking concept of an implementable manner.

Sampling rate conversion

As described below, a system for recovering PCM data from a distorted analog representation requires a method for synchronizing the clock and decode clocks used to transform PCM data from the digital stream to analog values. The digital implementation of this synchronization requires that the digital data sequence be resampled and changes its rate at a rate adjacent to the rate used for conversion from PCM data from the rate used by the analog to digital converter. To get this, the previously known techniques are either strictly limited in performance or computationally intensive. See, for example, R.E. Crochiere and L.R. Rabiner et al., &Quot; Multirate Digital Signal Processing ", Prentice-Hall, Englewood Cliffs, NJ, 1983. Transforming the sampling rate between two separate clocks where the correlation changes as a function of time complicates the operation.

A feature of the present invention is a method for performing such a transformation using a minimum computational cost. The method accepts a continuously variable input / output sampling rate ratio and performs the conversion with high accuracy. The disclosed technique can obtain more values than the 90dB anti-aliasing rejection and can be implemented in real time on existing processors.

Whole system

Figure 3 shows an overview of the proposed system. The method of using the system shown in Fig. 3 is the same as the method for a current data communication circuit or modem. Data applied to the data stream 100 appears in the data stream 126 after a predetermined time. The data stream 100 is applied to an encoder 150 that has the capability to convert the data stream into a format compatible with the telephone system. The converted data is applied to the digital telephone network 134 via the digital network connection 132. The converted data appears at the telephone central office of the client via the digital network connection 138, and the central interface is located at the line interface 140. In this regard, if the client has direct digital access to the digital connection from the digital network connection 138 to the client line interface, the transfer is complete. However, similar to most users, if the client does not have direct digital access to the telephone network, this is not possible and the following additional operations are required.

The line interface 140 converts the digital data on the digital network connection 138 into analog form in a manner consistent with the standard specifications of a digital telephone. The analog form is moved to the client premise equipment on the subscriber line 122, where the hybrid network 152 terminates the line and generates the analog signal 154. The hybrid network 152 is a standard part for converting a two-wire type bidirectional signal into a pair of one-way signals. Decoder 156 uses analog signal 154 to compensate and estimate distortions induced by the conversion to analog form performed by circuit interface 140 and to cause an estimate of digital network connection 138 , It should be assumed that this is the same as the digital data applied to the digital network connection 132. [ The transform performed by the encoder 150 is then inversely transformed and the decoder 156 outputs a data stream 126 that is a delayed estimate for the original data stream 100.

It should be noted that all elements in FIG. 3 are present in the present digital telephone system except for encoder 150 and decoder 156, which are well known and will be described below. A method for initializing and adapting the decoder 156 for the exact conditions that occur in normal operation will also be described below.

Physical implementation of the encoder

FIG. 4 is a block diagram illustrating one possible implementation of the encoder 150 of FIG. The data stream 100 from Figure 3 is applied to the serial data input of the digital signal processor 160, such as the AT & T DSP 32C. The processor uses the processor bus 162 for communication with ISDN interface circuitry 164 such as read only memory 168, ROM, random access memory 166, RAM, and Advanced Micro Devices Am79C30A. The ROM 168 includes a storage program having the functional characteristics described in the following sections. RAM 166 is used for program storage and parameters. The ISDN interface circuitry 164 also includes an ISDN connection 170 that is connected to a network terminating device 172 such as Northern Telecom NT1 and subsequently to the digital network connection 132 shown in FIG.

Additional additional elements such as decoders, oscillators and glue logic are required to be added to the basic block diagram shown in FIG. 4 to form a fully-functional implementation. Such additives are well known and will be apparent to those skilled in the art.

The following description of the encoder 150 relates to a functional aspect rather than a physical component, all of which are implemented as a program or subroutine for the digital signal processor 160 using, for example, known digital signal processing techniques .

Encoder operation

Figure 5 shows a functional block diagram for the encoder of Figure 3; The channel from the server to the client starts with any digital data provided as a data stream. Encoder 150 converts this bit stream into a sequence of 8-bit words sampled at a telephone system clock rate of preferably 8,000 samples / second. This is accomplished by a sequence of operations starting at the S / P converter 180, which categorizes each 8 bits read out in the data stream 100 into a stream of 8-bit parallel values in an 8-bit code stream 182 . This mapping is preferably such that the first of each of the 8 bits read from the data stream 100 is located in the least significant bit of the 8 bit code stream 182 until the output word ends and the subsequent bits occupy a more significant bit position And the process is repeated. The DC remover 184 inserts an additional 8-bit value at uniform intervals, preferably once per 8 samples, so that the analog value associated with the inserted value is the sum of all previous values on the 8-bit code stream 182 Make it negative. This is required because the telephone system often reduces or eliminates certain DC biases on the signal. DC remover 184 is an example of circuit means for reducing the DC component in the received analog signal.

Detailed basic elements of the DC remover 184 of FIG. 5 are shown in FIG. The code stream 186 output from the two-input selector 190 is also converted to a linear value 194 by a μlaw to linear converter 192, Can be implemented as a look-up table of 256-elements using a linear transformation table. The value of the linear value 194 may be adjusted by the adder 196 and the unit delay 200 to produce a DC offset 198 and a previous DC offset 202 corresponding to the unit delay value, negate. The DC offset 198 is applied to a linear approximation μ law transformer 204 which uses the same lookup table as the μ law to linear transformer 192 but performs the inverse mapping. It should be noted that if DC offset 198 is greater than or less than the maximum or minimum value in the table, then each maximum or minimum entry is used. The DC restoration code 206 is generated by the linear approximation rule transformer 204 and applied as one input of the two-input selector 190. 2 input selector 190 preferably reads seven consecutive values from the 8 bit code stream 182 to output these values into the code stream 186 and then reads a single value from the DC decompression code 206 And outputs it. This operation sequence is continuously repeated.

Referring again to FIG. 5, a code stream 186 is applied to the input lead of the ISDN converter 188, which provides a known conversion to an ISDN signal. The function of the ISDN converter 188 is implemented directly by various existing integrated circuits including the Advanced Mcro Devices Am79C30. The output of the ISDN converter 188 also forms a digital network connection 132 that is the output of the encoder 150 of FIG.

For further understanding, certain signals used by the encoder 150 are shown in Figures 7A-7C. FIG. 7A shows a sequence for a sample of a data stream 100. FIG. After processing by the serial-to-parallel converter 180 and DC remover 184, a code stream 186 is shown in FIG. 7B. Within DC remover 184, a linear equivalent, or linear value 194, for code stream 186 is shown in FIG.

Line interface

With reference to the following description, FIG. 8 illustrates a functional model for the line interface 140 of FIG. 3, such as found in a typical telephone system to use features of the present invention. It should be noted that such an interface is known and is currently used in a digital telephone exchange. The digital telephone network 134 of FIG. 3 transmits the 8-bit μ-law-encoded digital data stream per sample over the digital network connection 138 to the μ law-to-linear converter 210 shown in FIG. The μ law-to-linear converter 210 performs a known μ law-to-linear conversion to convert each sample to a linear value 212. The linear value 212 is converted to an analog signal 216 by a digital to analog converter 214 that is sampled using a telephone system clock 236 in a known manner. Although not shown in FIG. 3 for simplicity, the telephone system clock 236 is generated by the digital telephone network 134. The analog signal 216 is smoothed by a low pass filter 218 to form the next filtered signal 220. The main purpose of the low pass filter 218 is to provide a low pass function with a cutoff frequency of approximately 3100 Hz. The International Telecommunication Union has a digital-to-analogue converter 214 and a low-frequency (NTSC) interface 214 in the International Telecommunication Union, Telecommunication Standardization Sector (ITU-T), Transmission Performance Characteristics of Pulse Code Modulation Recommendation G.712, Geneva, Switzerland, And the characteristics for the filter 218 were standardized.

The filtered signal 220 is multiplexed onto the subscriber line 122 by a four-wire to two-wire converter 222. Subscriber line 122 is bidirectional; The signal input on the subscriber line 122 is applied to the four-wire to two-wire converter 222 and is output as the unfiltered signal 234. [ The unfiltered signal 234 is applied to a bandpass filter 232 standardized by ITU-T on the basis of the above-mentioned criteria. The output from the bandpass filter 232, that is, the filtered signal 230, is converted to a linear value 226 by the analog to digital converter 228. The linear value 226 is transformed into a digital network connection 136 by a linear algo- rithm transformer 224 that performs a standard linear algo- rithm transform. In the system shown in FIG. 3, it should be noted that the digital network connection 136 is not used and is omitted for simplicity.

Physical implementation of the decoder

FIG. 9 shows a block diagram of a possible implementation of the decoder 156 of FIG. The analog signal 154 from Figure 3 is sampled by an analog to digital converter 240 present as an integrated circuit, such as Crystal Semiconductor CS5016. Generated by the oscillator 242 to generate a digital input signal 246 connected to a bank 248 of a digital signal processor such as the AT & T DSP 32C via one of the serial digital input leads. 244), preferably using a clock signal of 16 kHz. The processor is also connected to each other and connected to the RAM 252 and the ROM 252 via the processor bus 250. ROM 252 includes a storage program having functional features to be described in the next section. The bank of digital signal processors 248 produces a data stream 126 that is the final output of the decoder 156 of FIG.

Additional additional elements such as decoders, oscillators, and glue logic circuits are required to be added to the basic block diagram shown in FIG. 9 to form a full functional implementation. Such additives are well known and will be apparent to those skilled in the art.

The following description of the decoder 156 relates to a functional aspect rather than a physical component, all of which may be implemented as a program or subroutine for the digital signal processor bank 248 using, for example, do.

Decoder operation

FIG. 10 shows the functional structure for the decoder 156 of FIG. The analog signal 154 of FIG. 3 provides the input data to the decoder 156. The analog signal 154 is supplied to the analog to digital converter 240 and converted into a digital input signal 246, preferably a signal sampled at 16,000 samples per second with 16 bits per sample precision. The analog to digital converter 240 exists as an integrated circuit such as the Crystal Semiconductor CS5016. The digital input signal 246 is processed by a clock synchronizer 260 that interpolates and resamples the digital input signal 246 at discrete intervals by a period estimate 262 to form a synchronized signal 266 . The operation of the clock synchronizer 260 will be described in the next section. The synchronized signal 266 is filtered by an inverse filter 268 to be described below to reconstruct the compensation signal 274. The purpose of the inverse filter 268 is to filter the low pass filter 218 of Figure 8, To the line interface 140 of FIG. 3 having the first component as the first component. 10, the inverse filter 268 also outputs a delay error estimate 270 that provides a timing error that is inherent in the synchronized signal 266 and is used by the period estimator 262 The delay error estimate is used by the clock estimator 264 to be described below. A decision means is used to convert the compensation signal from the discrete set to a sequence value. By way of example, the compensation signal 274 is converted to an approximate equivalent 8-bit μ law word using a linear approximate μ-law converter 276 to provide an estimated code stream 280. As described above, the linear approximation rule transformer 276 can be implemented as a simple look-up table.

In normal operation, the exchange 292 gathers the estimated code stream 280 back to the desired output signal 286, which is converted by the μ law-to-linear converter 278 into a linear signal And the μ law-to-linear converter 278 can be implemented with a simple look-up table as described above. In initialization, the exchanger 292 is set in such a way that a predetermined training pattern 288 (not shown in FIG. 3) is gated to the desired output signal 286. Such use will be described below.

The linear value 284 provides an estimate of the desired value of the compensation signal 274. This is used to adaptively update the inverse filter 274 in such a way that the compensation signal 274 is as close as possible to the linear value 284. [ This adaptation is an example of a training means for adjusting the parameters of the decoder 156, which will be further described below in the description of the inverse filter 268. The subtractor 282 uses the compensation signal 274 and the linear value 284 to calculate the error signal. The error signal 272 is reapplied to the input lead of the inverse filter 268 in the feedback loop. The estimation code stream 280 is also passed through a data extractor 290 which inversely transforms the transform performed by the encoder 150 of Figure 3 to form the final output data stream 126 of the decoder.

For illustrative purposes only, an example for a given signal provided in FIG. 10 is shown in FIGS. 11A-11E. Figure 11 shows a typical input analog signal 154 to a time-dependent decoder 156. During processing of such a signal, the decoder 156 forms the compensation signal 274 shown in FIG. 11B. This signal is further processed to form the estimated code stream 280 shown in FIG. 11C. Finally, the data extractor 290 of FIG. 10 outputs the data stream 126 shown in FIG. 11D. The error signal 272 formed for use within the decoder 156 is shown in FIG.

As noted above, analog-to-digital converter 240, subtractor 282, linear approximation law converter 276, switch 292 and μ law to linear converter 278, all of Figure 10 are known, And can be readily implemented by those skilled in the art. The following description assumes that the remaining blocks; The implementation and operation of inverse filter 268, clock estimator 264, clock synchronizer 260 and data extractor 290 are described.

Inverse filter

12 shows the internal detail of the inverse filter 268 of FIG. The inverse filter 268 is an example of an equalization means that operates by performing a linear filtering operation on the input signal (the synchronization signal 266) to form the output signal (the compensation signal 274). The inverse filter 268 also receives an error signal 272 that indicates a mismatch between the compensation signal 274 and the desired signal. The filter uses the error signal 272 to update its filtering function so that the error signal 272 is minimized. Such an adaptive filter structure is known; See, for example, "Data Communication Principles" Plenum (1992) by Richard D. Gitlin, Jeremiah F. Hayes, and Stephen B. Weinstein, et al. However, for the sake of simplicity, a preferred implementation for the inverse filter 268 is disclosed herein. In addition, the inverse filter 268 forms a delay error estimate 270 used by the clock estimator 264 of FIG.

The synchronized signal 266 is applied to a feed-forward equalizer 300 that generates a partial compensation signal 302 while using the correction signal 324 to perform the adaptive update. The operation of the feed-forward equalizer 300 will be described below. The feed-forward equalizer 300 also outputs a delay error estimate 270 used by the clock estimator 264 of FIG. The partial compensation signal 302 is downsampled by the following downsampler 304 using two factors to form the downsampling signal 306. [ The downsampler 304 operates by repeatedly reading two consecutive values from its input lead and outputting its preceding output to its output lead, and discards the second value. The downsampling signal 306 is then applied to a subtractor 308 to form a compensation signal 274. The compensation signal 274 is used in the successive steps of FIG. 10 and is also applied to the unit delay 310 to form a delayed signal 312. The delay signal 312 is then applied to the input lead of the feedback equalizer 314 to form a distortion estimate 316. The feedback equalizer 314 is similar to the feed-forward equalizer 300 and will be described in detail below. Distortion estimation 316 provides a second input to subtractor 308. [ The error signal 272 of Figure 10 is scaled using a constant factor in the gain element 318 of Figure 12 to form the correction signal 320 which is used as the second input signal of the feedback equalizer 314 . The feedback equalizer 314 performs the adaptive update using the correction signal 320.

The error signal 272 is also upsampled using the factor by the upsampler 326 by inserting zeros between each sample of the error signal 272. [ The upsampler 326 generates an upsampling error signal 328 that is scaled by the gain element 322 to provide a correction signal 324. [ The use of the correction signal 320 and the correction signal 324 by the feedback equalizer 314 and the feed-forward equalizer 300 will be described below. The values of the parameters (kf and kb) of the gain element 322 and the gain element 318 are preferably 10-2 to 10-15, respectively. Optimal values can be readily taken by those skilled in the art without undue experimentation.

Feed-Forward and Feedback Equalizer

FIG. 13 shows the internal structure for the feed-forward equalizer 300 of FIG. The feed-forward equalizer 300 preferably consists of 8-128 identical filter taps 330 connected in a chain. The first filter tap (TAP 1) 330 receives the synchronization signal 266 of FIG. 12 and the final filter tap 330 outputs the partial compensation signal 302 used in FIG. A tab interposed between each of the two input signals; The first input 332 and the target input 336 are taken as two output signals; A first output 334 and a target output 338 are formed. Each filter tap 330 also provides the tap weights 340 used by the delay estimate 342 as an output signal to compute a delay error estimate 270. During operation, using the correction signal 324 as an input signal, each filter tap 330 performs an adaptive update.

Fig. 14 shows in detail the function of each filter tap 330 in Fig. 14, each tap has two inputs, a first input 332 and a target input 336, and two outputs, a first output and a second output, (338). The first input 332 is delayed by one sample by unit delay 350 to form a first output 334. While the first input 332 is also multiplied by the tap weight 340 using the multiplier 352 to provide a weight input 354. The weight input 354 is added to the target input 336 by an adder 356 to form the target output 338. [

The adaptive updating of the tap weights 340 is performed by multiplying the correction signal 324 by the first input signal 332 using a multiplier 366. [ The multiplier output value 364 provides a tap error estimate and is subtracted from the previous value 360 using the subtractor 362 to form the tap weight 340. The previous value 360 is formed by the unit delay 358 using the tap weight 340 as an input. Each filter tap 330 also outputs a tap weight 340.

Referring again to FIG. 13, each filter tap 330 is applied to a delay estimator 342. The delay estimator 342 estimates the delay;

(1)

To calculate a delay error estimate 270 of the overall filter, where w _i is the abbreviation for the i-th tap weight 340. In this way, the delay estimator 342 provides estimation means for determining the degree of error in the period estimate 262 of FIG.

The above-described feed-forward equalizer 300 of FIG. 10 also applies to the feed-back equalizer 314. The structure and operation of the feedback equalizer 314 is identical to the feed-forward equalizer 300 except that the delay estimator 342 is required, so there is no output equal to the delay error estimate 270. The feed back equalizer 314 also uses a different number than the feed-forward equalizer 300, preferably between 1/4 and 1/2 the number of the feed forward equalizers 300. The optimal number of taps for both feed-forward equalizer 300 and feedback equalizer 314 can be readily taken by one skilled in the art without going through the experiment.

Clock estimator

FIG. 15 illustrates the basic components of the clock estimator 264 of FIG. Clock estimator 264 is an example of a circuit means that uses delay error estimate 270 to update period estimate 262. [ Estimated signal, that delays the error input to clock estimator 264, 270 is dependent on the accuracy of the clock used for analog-to-digital converter 240, by a loop gain 370, but preferably from 10 ^-1 to And is scaled using a factor k _{1 in} the range of 10 ^-8 to produce a phase error 374. [ Phase error 374 is filtered using loop filter 376 to form period offset 378. [ The loop filter 376 is a low-pass filter having a design known to those skilled in the art for the design of a phase-locked loop. Period offset 378 is added to nominal period 380 by adder 372 to produce period estimate 262. [ Nominal period 380 is an estimate of the half rate of the sampling rate of analog to digital converter 240 of FIG. 10, which is ahead of the frequency of telephone system clock 236 of FIG. Since the clocks used by telephone system clock 236 and analog to digital converter 240 are not derived from a common source, the exact ratio is slightly different from 1.0 for the preferred choice of parameters. During operation, the period estimate 262 refines and tracks this ratio using an estimate of the current error provided by the inverse filter 268 of FIG.

Clock synchronizer

A functional block diagram of the clock synchronizer of FIG. 10 is shown in FIG. The function of the clock synchronizer 260 should be interpolated and resampling its input signal at intervals separated by a period estimate 262. [ For example, if the period estimate 262 has a value of 2.0, all second samples read from the digital input signal 246 will be output as the synchronization signal 266. If period estimate 262 is not an integer, clock synchronizer 260 is required to properly interpolate between input samples to form an output sample.

The clock synchronizer 260 performs one cycle of operation for each of the required output samples. Each cycle begins with an accumulator 424 that reads the value of the period estimate 262 of FIG. The accumulator 424 forms an execution sum for all input values read and outputs this sum as a sample index 426 of actual values. This is scaled by a factor in the range of a factor (N [ ^mu] ), preferably from about 10 to 400, using the gain element 428 to form the upsampled sample index 430. [ Optimum value (N ^μ) it can be easily obtained by those skilled in the art, without going through an experiment. The integer / fractional splitter 432 decomposes the upsampled sample index 430 into a sample index 422 and a fractional value 414. For example, if the upsampled sample index 430 has a value of 10.7, the integer / fractional splitter 432 sets the sample index 422 to 10.0 and the fractional value 414 to 0.7.

One of the input signals applied to the sample selector 398 is formed by a series of operations beginning at the digital input signal 246. [ The upsampler 390 reads the value from the digital input signal 246 and outputs N [ ^mu] samples of the value read from the digital input signal 246 following the N [ ^mu] -1 zero value. The output stream from upsampler 390, the upsampling input signal 392, is applied to a low pass filter 394 having a passband cutoff frequency such as 4 kHz. The design of the up-sampler 390 and the low-pass filter 394 is known. See, for example, "Multirate Digital Signal Processing" by RE Crochiere and LR Rabiner, Prentice-Hall, Englewood Chffs, NJ, 1983, which is incorporated herein by reference. A lowpass filter 394 forms a filtered upsampled signal 396 that is used as an input to the sample selector 398.

The sample selector 398 is an example of a selection means for reading a value from the sample index 422 and interpreting it as the sample number s ⁿ . The selector includes an internal count indicating the number of samples read from its input lead connected to the filtered upsampling signal 396 since the system was initialized. The selector reads additional samples from the filtered upsampling signal 396 and determines whether the sample 400 becomes an imitation sample s ⁿ read out of the filtered upsampling signal 396 and the sample 402 imitates And an output sample is formed so as to be a sample (s ^{n + 1} ). The sample 400 is then scaled by a fractional value 414 using a multiplier 404 to form a sample component 408. Similarly, the sample 402 is scaled by the fraction value 416 using the multiplier 406 to form the sample component 410. The magnitude of the fractional value 416 is the magnitude of the fractional value 414 minus one as calculated using the subtractor 420 and the unit constant 418. [ The sample component 408 and the sample component 410 are added in an adder 412 to form a synchronized signal 266 that is also the output of the clock synchronizer 260 of FIG. The combination of multiplier 404, multiplier 406 and adder 412 is an example of interpolation means for combining samples selected by sample selector 398. [

Clock synchronizer 260 may also be used in other applications or as a stand-alone sampling rate converter. In general, the synchronization signal 266 has only a different sampling rate, but is equivalent to the digital input signal 246. The ratio for the two ratios is characterized by a varying period estimate 262 with respect to time.

It should be noted that although linear interpolation appears to be a rough approximation to the desired result, in practice this is fairly accurate. By oversampling by the up-sampler 390, the filtered up-sampled signal 396 has a frequency spectrum that is nearly zero anywhere except for a narrow width around the DC. The interpolation operation efficiently generates this narrow bandwidth image in the frequency domain. The function of linear interpolation is to filter and output these images. A typical implementation uses a sharp, computationally expensive low-pass filter to accomplish this. Even though the linear interpolator is a low-pass filter with very weak characteristics, it has a very deep spectral notch at the very frequencies at which unwanted images appear. This is a combination for placement of such a notch with a narrowly elided image which eliminates many calculations in the prior art and at the same time makes the method very accurate.

Data extractor

The final step of the decoder 156 of FIG. 3 is the data extractor 290 of FIG. The function of the data extractor 290 is to invert the transformations performed by the encoder 150 of FIG. This transformation consists of the serial-to-parallel converter 180 and the DC remover 184 shown in Fig.

In order to invert this transformation, the data extractor 290 preferentially removes the values inserted into the data stream by the DC remover 184. This is done by simply discarding all eighth samples read from the input (assuming that DC elimination is performed by DC remover 184 using a preferred ratio of once per 8 samples). When this is done, the remaining stream of 8-bit values can be converted back to the serial data stream 126 by outputting one bit out of each word at a time, which starts at the least significant bit. Such techniques are known to those skilled in the art.

System initialization

First, when a connection is established between the server and the client, both the encoder 150 and the decoder 156 of FIG. 3 start in a state known to each other. The following initialization is performed in the encoder 150.

1. The DC remover 184 of FIG. 5 is initialized using the two-input selector 190 of FIG. 6, with its next output set to be an imitation DC restoration code 206.

2. The output of the unit delay 200 of FIG. 6, i.e., the previous DC offset 202, is initialized to 0.0.

3. The code stream 186 of FIG. 5 is temporarily disconnected from the DC remover 184. Instead of N _c , preferably 16-128 known sequences, the value is repeated N _t times, preferably 100-5000 times. Optimal values for the use of N _c and N _t can be readily obtained by those skilled in the art without experimentation.

The choice of N _c bound to the design of the decoder 156 is preferably half of the number of taps in the feed-forward equalizer 300 of FIG. One possible choice for a sequence for code values repeatedly transmitted by the encoder 150 without loss of generality is shown in Table 1. The same sequence, applied to the training pattern 288 of FIG. 10, is also used by the encoder 150. FIG.

(Table 1)

General training patterns

When N _t times repeated sequences are output, the code stream 186 is reconnected to the DC remover 184 and the next output from the decoder 156 is reconnected to the DC remover 184 and the subsequent output from the decoder 156, 3 < / RTI >

In the decoder 156 of FIG. 3, the following initialization is performed before the first sample is read out from the analog signal 154.

1. The exchange 292 of FIG. 10 sets the gate training pattern 288 for the desired output signal 286.

2. The data extractor 290 of FIG. 10 is set in such a way that the next input value, i.e., the estimated code stream 328, is considered to be a DC equivalent value and thus discarded.

3. Unit delay 310 of FIG. 12 is initialized to output zero as delayed signal 312.

4. The upsampler 326 of FIG. 12 is initialized in such a way that its next output, the upsampling error signal 328, becomes an imitation error signal 272.

5. The downsampler 304 of FIG. 12 is initialized so that its next input value, the partial compensation signal 302, is copied as the downsampling signal 306.

6. Within the feedback equalizer 314 and the feed-forward equalizer 300 of Figure 12, each unit delay 350 of Figure 14 is initialized to have a zero output.

7. Within the feedback equalizer 314 of FIG. 12, each unit delay 358 of FIG. 14 is initialized to zero.

8. Within feed-forward equalizer 300, each unit delay 358 in Fig. 14 is initialized to zero.

9. The accumulator 424 of FIG. 16 is initialized to output a value of zero to the sample index 426 of the actual value.

10. Low pass filter 394 is all initialized to have an internal state of zero input.

11. The up-sampler 390 is initialized so that its output, the up-sampling input signal 392, is the value of the digital input signal 246. [

The decoder 156 then operates as described above until the N _c -N _t value is formed in the estimation code stream 280 of FIG. At this point, the switch 292 moves to channel the estimated code stream 280 to the desired output signal 286. From this point, the data stream 126 should coincide with the data read from the data stream 128, as shown in FIG.

It should also be ensured that the encoder 150 and the decoder 156 enter and exit the initialization mode so that the values on the data stream 100 and the data stream 126 of FIG. 3 are precisely matched. Interrupting the DC recovery performed by the DC remover 184 is an example of a method for achieving such synchronization. To signal (signaling) the beginning of training, the code stream 186 is set to the maximum legal code value for a period longer than the normal DC recovery period, for example, for 16 samples. Followed by setting the code stream 186 to the minimum legal code value for the same number of samples. The training pattern follows this synchronization pattern. Similarly, the final training is signaled by changing the order of the above synchronization pattern, i. E. By repeating the maximum value after the minimum value. This synchronization pattern may be detected by the decoder 156 and used to control the switch 292.

Other techniques for such synchronization are known and used in existing modems, see, for example, ITU-T, V.34, mentioned above.

An alternative delay estimator

In the foregoing description, the delay estimator 342 is formed by consideration of the filter tap weights in the feed-forward equalizer 300. Other delay estimation means are also possible. For example, the error signal 272 and the compensation signal 274 of FIG.

(2)

Where v is the compensated signal 274 and e is the error signal 272 and k is the number of samples that have passed through the experiment And can be readily obtained by those skilled in the art. The value of k depends on the appropriate combination of observed signal noise and clock jitter. Any other method for implementing the delay estimation means that forms the delay error estimate 270 may also be used in the present invention.

Alternative decoder initialization method

As described above, the parameters of the decoder 156 may be set using a fixed initialization value prior to the training period in which the known data sequence is transmitted. The above-described approach uses a training sequence to perform continuous updating of the parameters of the inverse filter 268 and the clock estimator 264 in a basic sample unit.

It is also possible to perform a single block update on all parameters. During the transmission of the training sequence, only the decoder 156 stores the value represented as the digital input signal 246. Once the entire training sequence is transmitted, the decoder 156 performs an analysis of the obtained values and performs calculations on parameters within itself.

The calculations required to perform the parameter estimation are as follows.

1. Calculate the basic digital period (T _u ) of the signal obtained using the velocity estimation means. This is done using any of a variety of signal processing techniques known in the art, such as autocorrelation analysis. Further, assuming the use of the preferred sampling rate for the analog-to-digital converter 240, the period T _U is known to be approximately twice the length N _c of the training sequence. Only the source of the car is due to the difference between the sampling rate of the telephone system clock 236 and the sampling rate of the analog to digital converter 240.

2. The nominal period 380 of FIG.

(3)

.

3. By passing the digital input signal 26 through the clock synchronizer 260 using a delay error estimate 270 set to zero to form the synchronization signal 266, Sampling.

4. Form a matrix (Y) with 2 - N _c rows and N _t rows. The element of Y becomes the value of the synchronization signal 266 as calculated above. These are stored in the matrix by successively populating the first row with the second row, etc., using successive samples of the synchronization signal 266.

5. Compute the average for each row of U to form an element vector of r, 2 · N _c .

6. The following equation:

(4)

, An estimate of the energy (sigma ² ) of the noise component of the input signal is calculated, where _Yij is an element of row I, column J.

7. Calculate N _c , the element vector, c, by passing the training sequence value through a transducer, such as the μ law-to-linear converter 278, as shown in Table 1.

8. Form a matrix A having N _f + N _b columns and N _c rows as follows,

(5)

Where N _f is the number of filter taps in the feed-forward equalizer 300 of FIG. 12, and N _b is the number of filter taps of the feedback equalizer 314. For example, and N _c = 3, and N _f = 4, if N _b = 2,

(6)

.

9. The following equation:

(7)

Finds the N _f + N _b element vector value (x) that minimizes e ² .

This can be solved using techniques known in the art of linear algebra, calculus, and iterative techniques, which will be apparent to those skilled in the art.

10. For each tap of the feed-forward equalizer 300, the previous value 360 of FIG. 14

Initialize each with x _l ... x _Nf .

11. Initialize the previous value 360 for each tap of feedback equalizer 314 using x _Nf +1 ... x _Nf + N _b , respectively.

12. Once these parameters are calculated, normal operation begins. It should be noted that the parameter is substantially changed due to the adaptive update based on the error signal 272 as previously described.

The above-described sequence is shown as an example of another method of performing the initialization of the decoder 156 using the training sequence. Other methods and numerous different methods are available. For example, the received training sequence may be truncated at each end to eliminate the effects of transients upon switching between the normal and training modes, and the linear vs. μ-law converter 276 and the μ-law to linear converter 278, The exact transient levels can be adjusted using the training information, and for each previous value 360,

Modified expressions may be used.

Preferred Alternative Training Procedures

The preferred steps for training the encoder 150 and decoder 156 of FIG. 3 will now be described.

1. The encoder 150 transmits the repeating pattern to the digital telephone network. This pattern consists of M repeated sequences of N PCM codewords to provide a graceful M x N codewords. The term " PCM codeword " as used herein refers to a set of codewords used by the digital telephone network 134. The N PCM codewords are randomly selected from two opposing values. For example, the PCM codewords (0x14 and 0x94) correspond to each other under the companding rule of the telephone network rule. The random selection is also forced so that each of the two PCM code words is used exactly N / 2 times. This ensures that the training sequence does not contain any DC components.

2. Decoder 156 receives analog signal 154, which is an equivalent analog of the PCM codeword pattern, and stores the signal using prior recognition of M and N. [ The analog signal 154 is sampled and stored at a nominal rate of 16,000 samples / second.

3. Stored sequences are analyzed to find their repetition rate using standard signal processing techniques. Even if the period associated with the repetition rate is N / 8000 seconds, there is some inconsistency due to the difference between the correct value of the decoder clock and the clock used by the telephone network 134. This discrepancy in the repetition rate is used to adjust the clock of the decoder and resampled the stored signal using the corrected clock. This technique is repeated many times until the measured period matches the expected period exactly.

4. The noise level is preferably measured by observing the change with M iterations of the test pattern. Each iteration must be the same, and it is only noise that causes this change. Any repetition, including the first or final, should be ignored if significant differences exist in the mean. This eliminates or reduces the end effect, intermittent noise spikes or crosstalk.

5. The mean signal level of the considered iteration is determined to provide an average receive sequence of length 2N. The length 2N results from sampling the input analog signal 154 at 16000 samples / second.

6. The optimal equalizer is designed using known transmission sequence, average received signal level and noise estimate. An equalizer with minimum mean square error is obtained by solving a set of linear equations using a known scheme. In particular, a combination of partially spaced equalizers with 90 taps operating at 16,000 samples / second followed by a 20-tap decision-feedback equalizer operating at 8000 samples / second can be used.

7. The mean squared error estimate and equalizer settings may be sent back to the encoder 150 when a reverse channel, as described below, is used. Encoder 150 uses these values to select a codeword set and encoding method that maximizes the efficiency of the communication link.

After the training is completed, the encoder 150 starts data transmission via the link. Decoder 156 uses the calculated clock adjustment and equalizer settings to compensate for the received signal and then determines which PCM codeword is transmitted by encoder 150. [ The selection is further processed to convert the PCM codeword into a data sequence. In addition, the measured deviation between the compensation signal and the nearest real PCM codeword is used for continuous error measurement. This error measurement is applied again to the equalizer and clock adjustment circuitry in decoder 156 to allow for continuous updating and to prevent any drift. This error measurement is also used to determine whether the quality of the circuit changes significantly when the encoder 150 is aware that re-training should be initiated.

Additional explanation for the reversal channel

FIG. 17 illustrates one aspect of the present invention in which the above-described communication system is combined with an inversion channel. The data stream 100 is applied to the encoder 150 as described with reference to FIG. Which in turn is connected to the digital telephone network 134 via the digital network connection 132. Data is presented over the digital network connection 138 in the network of the client's central office. The digital information is converted to analog form by the line interface 140 and placed on the subscriber line 122 in analog form. The client's preconditioning equipment or hybrid network 152 forms an incoming analog signal 448 and the echo canceller 442 generates an analog signal 444 output from the analog signal 444, Thereby eliminating the contribution to signal 448. The analog signal 154 is applied to a decoder 156 that provides a data stream 126. The data stream 128 from the client is converted to an analog signal 444 output by the modulator 446 according to known techniques such as those currently used in a modem and not only applied to the echo canceller 442, And is provided on the subscriber line 122 via the network 152. At the central office, it is converted to digital network connection 136 by the line interface 140. The digital telephone network 314 transmits data on the digital network connection 136 to the digital network connection 130. [ Demodulator 440 converts this signal into a data stream 102 for the server.

As shown in FIG. 17, for a system using an inverse channel, the decoder 156 is coupled to a conventional modulator 446, such as a V.34 modulator, to provide bidirectional communication. In such a case, preferably, the echo canceller 442 prevents the output of the modulator 446 from appearing as an input to the decoder 156.

Even if a general echo canceller is used, the signal characteristics of the system described herein have a specific purpose that can be usefully used. In particular, the incoming analog signal 154 has a frequency component from approximately DC to 4 kHz, while the signal output from the modulator 446 is more stringently band limited in the 400 Hz to 3400 Hz range. This asymmetry in the bandwidth between the incoming channel and the output channel can be used by using an asymmetric echo canceller. In addition, if the bandwidth of the output channel is further reduced, the asymmetry will be further increased, and the advantage of asymmetric echo cancellation will be increased.

For connection of the end of the encoder 150, a digital echo canceller is preferably used between the encoder 150 and the demodulator 440 shown in Fig. Here again, the asymmetric form of the connection can be used by using an asymmetric echo canceller.

action

The system shown in Figure 17 comprises two telephone subscribers; Duplex communication between one subscriber using a digital connection and another subscriber using an analog connection. The operation of the forward channel is as described above with reference to FIG. 3 with one addendum. The echo canceller 442 inserted between the hybrid network 152 and the decoder 156 was added to reduce the influence of the inversion channel. The echo canceller 442 scales the output analog signal 444 and subtracts it from the incoming analog signal 448 to produce an analog signal 154. Such techniques and implementations for echo cancellation are well known. The reverse channel can be implemented using a variety of existing modem technologies. For example, International Telecommunication Union, Telecommunication Standardization Sector (ITU-T), "A Duplex Modem Operation at Signaling Rates of up to 14,400 bit / s for Use on the General Switched Telephone Network and Leased Point to Point 2-wire Telephone-Type Circuit "Recommendation V.32bis, Geneva, Switzerland (1991). The data is modulated by a modulator 446 to form an output analog signal 444 that is carried by the telephone system. The modulation technique used is known. For example, a method capable of transmitting up to 14,000 bits per second has been described above. Similarly, a method capable of transmitting up to 28,800 bits per second is disclosed in International Telecommunication Union, Telecommunication Standardization Sector (ITU-T), Recommendation V.34 Geneva, Switzerland (1994), which is incorporated herein by reference.

The output analog signal 444 is located on the subscriber line 122 using the hybrid network 152 as is practically used in all telephone devices. The hybrid network 512 converts between one four-wire interface (two independent unidirectional signals) and the other two-wire interface (one bidirectional signal). The 2-wire signal is simply the sum of the two signals on the 4-wire side. At the central office of the client, the telephone company device converts the analog signal on subscriber line 122 to a digital network connection 136 sampled at 8,000 samples / second using telephone system clock 236. In North America, this conversion is performed to provide 8 bits per sample using nonlinear mapping, known as the 占 law, to improve the signal to noise ratio of a common speech signal. Once converted to the μ law, the client's signal is carried by the digital telephone network 134 until the signal reaches the server's premises equipment. Since the server has a digital connection with the telephone system, the signal is not converted to analog form by the central station of the server. However, there are several interface layers (such as ISDN 'U' or 'S') interposed between the server and the digital network connection 136. However, since the same data provided to the digital network connection 136 appears later in the digital network connection 130, this arbitration hardware can be ignored. Demodulator 440 performs the inverse of modulator 446, such as is performed in an existing modem with a small exception. Demodulators can be fully implemented in digital hardware because their inputs and outputs are all digital, while existing modems perform their roles using analog inputs. Implementations of the demodulator 440 are also known, such as using a modulator 446, and may be implemented in the International Telecommunication Union, Telecommunication Standardization Sector (ITU-T), "A Duplex Modem Operation at Signaling Rates of up to 14,400 bits / ≪ / RTI > Telephone < RTI ID = 0.0 > Telephone-Type Circuit " Recommendation V.32bis, Geneva, Switzerland (1991). Note that since the performance degradation of the signal occurs only on the user's subscriber line, the inverse channel exhibits superior performance over conventional modems. Existing modems handle distortion that occurs in the subscriber line at both ends of the communication path. Alternate implementations of the present invention may use other known methods or techniques that provide an invert channel, or may eliminate distortion altogether. Thus, the description of possible implementations of the inversion channel is provided for illustrative purposes, rather than limiting the scope of features of the invention. The provision of an inversion channel also simplifies the synchronization of the decoder 156 and the encoder 150 and allows the system to be reinitialized if necessary. The performance of the system can be monitored by the decoder 156 in consideration of the error signal in FIG. If the error signal 272 exceeds one third of the mean difference between a predetermined level, and preferably the μ law linear value, the decoder 156 determines whether the system should be reinitialized to the encoder 150 via the inverse channel .

Combination with source encoder

It is possible to extend the functionality of the encoder 150 and decoder 156 shown in Figure 3 to perform an additional inverse transformable transformation on the data stream 100 before being applied to the encoder 150. [ The effect of this conversion can be eliminated by providing an inverse conversion to the output of the decoder 156 before forming the data stream 126. [ Such a transformation is advantageously not limited thereto and provides a predetermined inverse function.

Error correction

Bits may be added to the data stream to provide error correction and / or detection using any known method for error correction. These include, for example, convolutional codes, block codes or other error correction or detection schemes written in detail in the literature. If the same error correction provided in the data stream 126 is also inserted in the signal path from the linear vs. μ law converter 276 to the μ law to linear converter 278 shown in FIG. 10, the desired output signal 286, The quality of the value 284 and the error signal 272 may be improved and the performance of the decoder 156 may be improved.

A subset of the source alfabet

Even though there are 256 possible code words of code available for data transmission, the μ law mapping allows these words to be non-uniformly spaced in the linear domain. Thus, a given pair of codewords may be more easily confused by the decoder 156 due to line noise or other damage. The source encoder may limit its output to a subset of these codewords to improve the accuracy of the decoder 156 through reduction of the total data rate. This may also be used to adapt the decoder 156 to poor line conditions, by decreasing the codeword alpha, if the decoder detects that the codeword can not be separated within a certain error criterion. By reducing the codeword set, an improved error margin is achieved by sacrificing the reduced data rate. Thus, by lowering the data rate, the system can handle the degraded connection.

In the system shown in Figs. 3, 17 and 18, data is transmitted as a sequence of 8-bit PCM codewords used by the digital telephone network 134. In the simplest implementation, all 256 possible PCM code words can be used, and the data can simply take 8 bits at a time and can be used as a PCM word.

However, this approach poses potential problems. First, in the interpretation of the standard μ law for PCM values, there are two code words from the 256 sets that map to the same analog value. Therefore, it is impossible to distinguish these codewords using only the resulting analog signal. Second, the interpretation of the μ law for the PCM codeword corresponds to a non-uniformly spaced analog level. Levels in tightly spaced confessions rules should be avoided as they can be easily confused. Third, the digital telephone network 134 often uses the least significant bits of the PCM codeword for internal signaling to make these bits unreliable. Fourth, when the PCM codewords are converted to their equivalent analog voltage levels, they are transmitted through various filters and subscriber lines, such as the smoothing filter of the CODEC. The result is that certain frequency components, notably DC and high frequency components, are attenuated. If the decoder 156 equates these frequency components, the noise level necessarily increases, causing increased confusion between levels. Fifth, the digital telephone network 134 may transmit code from the same set to a new value, such as in the case of an internal call using the encoding mode of the μ-law and the A-μ law, or when the network attempts to modify the signal level by remapping. You can remap the word.

The system shown here addresses this issue in two ways. First, the decoder 150 uses only a subset of the 256 PCM codewords. In addition, the encoder 150 may select a PCM codeword with reference to a previously transmitted codeword to reduce a predetermined frequency component of the resulting analog signal 154. [

The first step in this process is to identify the frequency and noise characteristics of the transmission channel. Encoder 150 transmits a known training pattern consisting of randomly selected independent PCM codewords. When converted to an analog signal such as an analog signal 154, this results in an analog signal having a substantially flat frequency spectrum. Decoder 156 receives the distorted analog signal 154 and configures a synchronizer and equalizer that reduces this distortion. In this process, the decoder 156 obtains measurements of the noise level and filtering characteristics of the line. These measurements are transmitted back to the encoder 150. This step was introduced under the heading " Preferred Alternate Training Procedure ".

Once the noise level is known, the encoder 150 may select a subset of PCM codewords such that the probability that a given codeword will be confused with the other due to noise is below a predetermined preset threshold probability. The subset is selected by preferentially calculating the required minimum separation between codewords using noise statistics. The codeword set is then selected by selecting the maximum possible set of codeword pairs that are not separated below the minimum preset threshold. The selection of a codeword set is also limited to not use any codewords that may be performance degraded by bitrobbing, which will be further described below.

An example technique for removing a given frequency from an analog signal reconstructed from a PCM codeword is to insert additional codewords that do not contain data at regular intervals. This insertion is used to specify the output spectrum. If, for example, DC is desired to be removed, then with reference to the DC remover 184, as described above, once for each of the N data carrying codewords, having an analog value as close as possible to the negative sum of the values of all previous codewords A code word is inserted.

Generally, the embedded codeword can be selected to specify the required spectrum. For example, the codeword may be transmitted via a digital filter at encoder 150, and the embedded codeword is selected to minimize the output energy of the filter. If the digital filter is selected to pass the desired component that is removed, such as a low frequency or a frequency near 4 kHz, this procedure can minimize the above components.

Since the data carrying codeword is selected from a subset of the 256 possible codewords, the data can not generally be located simply within 8 bits of the codeword at a time. Instead, a group of bits from the data is used to select the codeword sequence. For example, if the subset consists of only three codewords, a group of six input bits (with 2 ⁶ or 64 likelihoods) is used to form the four codewords (with the possibility of 3 ⁴ = 81) You can select the value of the group. In this case, there is a certain amount of waste for the available information content, but this decreases as the length of the group increases.

56,000 bits per second Using the Phone System

For a given PCM transmission scheme used by the telephone system, the least significant bit of each 8-bit codeword is used for internal synchronization. This inserts zero bits per 8 bits so that the encoding process described with reference to Figure 5 places the embedded bit at the location of the least significant bit of each encoding value applied to the digital network connection 132, ). ≪ / RTI > These inserted zeros are removed from the decoder 156 by the previous processing data stream 126. [ In this way, the use of the low order bits of the telephone system does not impair the transmitted data, and the maximum data rate is reduced to 56,000 bits / second.

For long distance telephone transmissions, including transmission over the digital telephone network 134, certain calls are made on the line using in-band signaling. In this case, the digital telephone network 134 may use (or infringe) the least significant bit (" LSB ") of every sixth PCM codeword for ring instructions and other signals. This technique is generally known as " roving bit signaling " or simply " bit roving ". If a dial-up is used to carry data, this means that only 7 bits are used during a bit-routed frame. One way to address this problem is to never use the LSB because the transmitter does not have any control over bit roving so that the maximum data rate (7 bits / codeword x 8000 codewords / second) is 56 kbps .

For example, the system shown in FIG. 3 improves upon this by instructing the decoder 156 to identify the bit-routed frame, and then instructing the encoder 150 to avoid only the LSB in the frame in which bit roving occurs. Thus, a single hop using bit roving has a lower bit rate; For example, to 62.7kbps instead of 56kbps. In the case of multiple hops over an asynchronous link, bit roving can occur at different phases, resulting in further reduction in bit rate.

Detection of bit robbed frames may be performed during initial training of the system. Encoder 150 transmits a known pattern of 8 bit PCM codewords over a transmission channel and decoder 156 stores samples from the resulting analog waveform received by it. Decoder 156 attempts to resynchronize and equalize such a signal to minimize the difference between its output and a known pattern, assuming no bit roving is generated. Decoder 156 then measures the average equalized value in each of the six phases. That is, the errors in the first, seventh, thirteenth, and ninth samples are average, and the errors in the second, eighth, fourteenth, etc. samples provide six average error measurements. The decoder 156 then determines whether 1) the bit has not been routed, 2) the bit has been roped by replacing the LSB with 0, 3) the bit has been roped by replacing the LSB with 1, 4) ≪ / RTI > for each phase. A selection is made by determining whether any bit lobing scheme (1, 2, 3 or 4) minimizes the difference between the pattern equalized and the signal known after the bit lobing is performed.

When the bit robbings generated in each frame are determined, the equalization processing is re-executed because the first equalization is performed in a state in which bit robbing is not recognized. This second processing provides a better equalization signal. The bit roving decision may then be made using a second equalized signal in the same manner as described above.

If bitrobing is known for each of the six phases, then the decoder 156 preferably sends this information to the encoder 150. [ For continuous data transmission, encoder 150 and decoder 156 avoids the use of bits that can be degraded by bit roving.

In an alternative method, the encoder 150 transmits a code word of known pattern after the clock synchronization and equalization steps described above are completed. Decoder 156 computes statistics taking into account the received signal level, such as the level represented by compensation signal 274 shown in FIG. 10 for each 256 PCM code word transmitted for each of the six phases. The calculated statistics, including the average signal level and change, are used to select a subset of code words that provide a predetermined error probability. The selected subset is sent to the encoder 150 and used in subsequent transmissions. In addition, the μ law versus linear and linear vs. μ law converters 192, 276, 278 use the average level computed above instead of the predetermined level of converters 192, 276, 278. Thus, this method provides an adaptive transducer that can implicitly handle bit roving or remapping that occurs in the digital telephone network 314 without preceding information and unambiguous analysis.

Data compression

The source encoder provides compression of the lossless data stream 100 using any of a variety of known techniques to those skilled in the art. But are not limited to, Lempel-Ziv compression, run-length encoding, and Huffman encoding. The inverse of the selected compression transform is also known and may be applied to the data stream 126. [

Use of another phone system

The above method can be used with a telephone system that uses a nonlinear convolutional operation that is different from the < RTI ID = 0.0 > u < / RTI > For example, most worlds use similar decoding known as A-law. Aspects of the present invention can also be applied to such systems by replacing the μ law versus linear and linear vs. μ law transducers using the equivalent of the A-law. Such equivalents may also be implemented using a 256-element look-up table. In this case, the table is filled with known A-law mappings. Such modifications are obvious to those skilled in the art.

Combination with existing modem

The features of the present invention may also be used with existing modems. In the conventional system shown in FIG. 1, the modem 104 may be modified to include the functions of the encoder 150 described above. In addition, modem 124 may be modified to include the functionality of decoder 156. When a call is connected between the modified modem 104 and the modem 124, both operate as if it were a general connection between the unmodified modems. After initialization is complete, the modem 104 sends an acknowledgment request to the modem 124 using a known acknowledgment protocol, such as a protocol standardized by the International Telecommunication Union (ITU). If the modem 124 includes the realization of the decoder 156, it responds positively to the request. Otherwise, the request is rejected and normal modem communication is used. When an affirmative response is received, the modem 124 and the modem 104 start from the sequence that has been switched and initialized to operate as shown in FIG. In this way, the combined modem / decoder is capable of compatibility with existing modems and, where possible, advantageously uses the features of the present invention to provide increased efficiency.

Combination with database server

The feature of the present invention can be used with a central server to provide data communication of any type (information, audio, video, etc.) between the central side and the plurality of users shown in Fig. The server 450 provides the server data to the server interface 454, which consists of an encoder array, such as the encoder 150 described herein, and possibly a demodulator array, such as demodulator 440. [ The server interface 454 is connected to the digital telephone network 134 via a server connection 456, such as an ISDN PRI interface. Each subscriber to the service includes a client interface 460 comprised of a decoder 156 and optionally an echo canceller 442 and demodulator 446 similar to those shown in FIG. The client interface 460 processes the client connection 458 to provide a client data stream 462. Overall, this configuration allows multiple users to independently communicate with a central server or server. Such configurations include but are not limited to audio or music decomposition, online services, networking services, video or television distribution, voice, information distribution, credit card validation, banking, computer access for interaction, remote management of the present invention, May be usefully utilized for certain types of data services including media. Other implementations and configurations of the invention are also applicable to these and other applications.

High-speed facsimile transmission

The features of the present invention shown in Fig. 19 can be used for high speed facsimile transmission. Transmission FAX 470 scans the image and converts it into a transmission data stream 472 in a known manner. The transport data stream 472 is transformed into a receive data stream 476 via a distribution system 474, for example as shown in FIG. The incoming fax 478 converts the data stream into an image to print or otherwise display. The distribution system 474 may be implemented as shown in Figure 17 using a data stream 126 that is replaced by a received data stream 476 and a data stream 100 that is replaced by a transmit data stream 472 . In addition, data stream 128 and data stream 126 may be stored in a computer-readable medium such as the International Telecommunications Union, Telecommunication Standard (ITU-T), Recommunication V.17, " A 2-Wire Modem for Facsimile Applications With Rates Up to 14,4000 b / s " (1991), Geneva, Switzerland, for receiving the protocol between incoming fax 478 and transfer fax 470. In this manner, the facsimile from the sending FAX 470 is advantageously sent to the receiving fax 478 with a higher jumper than was possible using conventional transmission schemes.

ISDN / Digital Phone Relay

The features of the present invention may also be used with some applications that may utilize ISDN or digital technology. It provides a functional equivalent to the ISDN for transmission to the second part, which has only an analog connection from the digitally connected part to the telephone network. This may be performed directly using a system such as that shown in Fig. 17, or may be performed using the mediation relay shown in Fig. The digital subscriber 480 enables a digital call to the analog subscriber 490 and the analog subscriber does not have direct digital access to the digital telephone network and instead has an analog subscriber connection 488. A full digital connection is opened between the digital subscriber 480 and the relay server 484 using a digital connection 482 such as ISDN, Switched-56, T1, or the like. Relay server 484 communicates with analog subscriber 490 along relay connection 486 using any useful means such as a conventional modem or a system such as that shown in Fig. Using appropriate flow control schemes known to those skilled in the art, it becomes clear to the digital subscriber that the digital connection is only open to analog subscribers. Such connections may be used for certain digital communications such as voice, data, digital FAX, video, audio, and the like. It should also be noted that the relay server 484 may be implemented in the actual digital telephone network 134 to provide a digital connection to the analog subscriber.

It will be understood by those skilled in the art that the present invention is shown and described with reference to preferred embodiments and that various changes and modifications may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. For example, an equivalent training request can be achieved using the inversion channel of FIG. The reversal channel of FIG. 17 may provide another equivalent configuration for controlling the flow of information from decoder 156 to encoder 150. However, in such a configuration, the present invention still provides data transmission between the data provider and the user. In addition, compensation of the telephone line may be achieved by other equivalent arrangements known to those skilled in the art; Equivalent training procedures can be used, different equalization methods can be used, and the system can be applied to the equivalents of a central office without departing from the scope of the present invention. Accordingly, one of ordinary skill in the art will appreciate that all such equivalent circuits and modifications are within the scope of the claims.

Claims (36)

A method of training a communication system having an encoder and a decoder, the encoder digitally connecting to a digital portion of a telephone network and the digital portion of the telephone network being connected to a decoder by an analog loop, Repeating the N codewords M times from the encoder to the decoder, wherein M and N are integers known to the decoder and the code words correspond to PCM codes used by the digital portion of the telephone network, Sampling a reception sequence of an analog voltage level corresponding to the transmission codeword in the decoder; Storing values associated with each sample in the encoder or decoder; Analyzing the stored value in the encoder or decoder; And And adjusting the parameters of the decoder in accordance with the analysis at the decoder. 2. The method of claim 1, wherein the N codewords are randomly selected from a subset of the two PCM codewords, the two subset of PCM codewords being < RTI ID = 0.0 & ). ≪ / RTI > 3. The method of claim 2, wherein each of the two PCM codewords is selected N / 2 times. 2. The method of claim 1, wherein the receiving sequence is sampled twice the clock rate of the digital portion of the telephone network. 2. The method of claim 1, wherein the receiving sequence is sampled at a rate of 16,000 samples / second. 2. The method of claim 1, wherein analyzing the stored value comprises determining a rate at which M iterations occur. 7. The method of claim 6, wherein the analyzing comprises: determining a sample period associated with the rate at which the M iteration occurs; Comparing the determined period to N / 8000 seconds; And And adjusting the clock of the sampling rate of the decoder according to the comparison. 8. The method of claim 7, further comprising resampling the stored value using a clock of the adjusted sampling rate. 9. The method of claim 8, further comprising: repeating the steps of claim 7 for the resampled value. 2. The method of claim 1, wherein analyzing the stored value comprises measuring a variance of M iterations to measure a noise estimate. 2. The method of claim 1, wherein analyzing the stored value comprises calculating an average received signal level from M iterations. 2. The method of claim 1, wherein the parameter is associated with an equalizer in the decoder. 13. The method of claim 12, wherein the parameter is adjusted according to at least one of a predetermined sequence stored in the decoder, a measurement of an average received signal level, and a noise estimate. 13. The method of claim 12, wherein analyzing the stored values comprises calculating a mean square error estimate. 15. The method of claim 14, wherein the parameter is adjusted to achieve an equalizer having a minimum mean square error. 16. The method of claim 15, further comprising transmitting the parameter back to the encoder. 17. The method of claim 16, further comprising: selecting a codeword set for data transmission based on the parameter at the encoder. 2. The method of claim 1, further comprising updating the training method based on measuring a deviation between the compensation signal and the most recent point codeword. 19. The method of claim 18, further comprising continuously updating the clock and parameters of the sampling rate associated with the equalizer in accordance with the measurement of the deviation. 20. The method of claim 19, further comprising retraining when the measurement of the deviation exceeds a predetermined threshold. An encoder is digitally connected to a digital portion of the telephone network, the digital portion of the telephone network being connected to the decoder by an analog loop, the analog loop converting the digital information from the network to an analog value for transmission to the decoder, A method for detecting the presence of routed bit signaling in a digital portion of a telephone network, Transmitting a predetermined pattern of PCM code words from the encoder to a digital portion of the telephone network; Receiving the PCM codeword of the predetermined pattern in the form of an analog voltage value at the decoder; Determining an error measurement in the decoder in response to the received analog value; And And determining in the encoder or decoder whether there is a roving bit signaling based on the error measurement. 22. The method of claim 21, wherein determining the error measurement comprises comparing the received analog voltage to a predetermined reference voltage. 22. The method of claim 21, wherein the predetermined pattern is divided into M frames, each of the M frames having N time slots. 24. The method of claim 23, wherein determining the error measurement comprises forming N average error measurements corresponding to the N time slots. 25. The method of claim 24, wherein each of the N average error measurements is made by averaging all Nth error measurements. 24. The method of claim 23, wherein N = 6. 22. The method of claim 21, wherein determining the routed bit signaling comprises: Determining whether the size of the error determination is periodically iterative; And And determining whether the periodically repeated error measurement results from roving bit signaling when the error measurement is periodically repeated. A method for selecting a set of PCM codewords in a communication system having an encoder and a decoder, wherein the encoder is digitally connected to a digital portion of the telephone network and the digital portion of the telephone network is connected to the decoder by an analog loop, Transmitting a PCM codeword sequence from the encoder; Receiving the sequence as an analog voltage sequence at the decoder; Determining a noise level in the decoder in response to the analog of the received sequence; Selecting a minimum separation value in the encoder or decoder in response to the noise level determination; And Selecting a PCM codeword set in the encoder or decoder such that no two PCM codewords in the set are separated by less than the selected minimum spacing value. The method of claim 28, wherein selecting the set further comprises selecting a second set of PCM codewords, wherein the PCM codewords from the second set are used during bit- How to. 29. The method of claim 28, wherein determining the noise level further comprises measuring a change in the analog voltage of the received sequence. An encoder is digitally connected to a digital part of a telephone network, the digital part of the telephone network being connected to a decoder by an analog loop, in a communication system with an encoder and a decoder, in a method for transferring information from the encoder to the decoder , Transmitting a sequence of PCM codewords from the encoder; Receiving the sequence as a sequence of analog voltages at the decoder; Determining a noise level at the decoder in response to the analog voltage of the received sequence; Selecting a minimum interval value in the encoder or decoder in response to the noise level determination; Selecting a PCM codeword set in the encoder or decoder such that no two PCM codewords in the set are separated by less than the selected minimum spacing value; And Encoding a sequence of information bits from the selected set of code words into a PCM codeword sequence to convey information from the encoder to the decoder. 32. The method of claim 31, wherein selecting the set further comprises selecting a second set of PCM codewords, wherein the PCM codewords from the second set are used during bit- How to. 32. The method of claim 31, wherein the encoding step is performed by a convolutional encoder. 32. The method of claim 31, wherein the encoding step is performed by a block encoder Lt; / RTI > 32. The method of claim 31, wherein the encoding step is performed by a trellis encoder. 32. The method of claim 31, wherein determining the noise level further comprises measuring a change in the analog voltage of the received sequence.