JP2006505211A - Multistage nonlinear echo canceller for digital communication systems with or without frequency division duplexing. - Google Patents

Abstract

In a digital communication system in which a digitally modulated signal is transmitted in a signal path including a bidirectional channel (218), a multistage echo canceller (207) removes a linear echo signal and a nonlinear echo signal of an output signal from the input signal (220). In order to have two or more stages connected in sequence. The result is important for both systems where the output signal frequency band overlaps with the input signal frequency band and for systems where the bands are separated (here only for non-linear echo signals). Each of the two or more stages is characterized by its function, and the sequence is characterized in that the functions of two or more stages are alternately arranged in the order of a linear function and a nonlinear function.

Description

  This application claims priority to US Provisional Application No. 60 / 422,655, filed Oct. 30, 2002. This application also claims priority as a continuation-in-part of the following two pending US patent applications, which are hereby incorporated by reference:

No. 09 / 968,469, filed Oct. 1, 2001, “Peak-to-Average Power Ratio Reduction in a Digitally Modulated Signal”
No. 09 / 968,063, filed on Oct. 1, 2001, “A Multistage Equalizer for Corrections and Nonlinear Distortion in Digital Digitally”
The present invention relates to information transmission by means of digital modulation over a bidirectional channel in the presence of linear and non-linear distortions. More particularly, the present invention relates to the correction of a received digitally modulated signal to the echo or echo effect of the transmitted signal, particularly when the echo signal is subject to both linear and nonlinear distortion.

  Digital modulation refers to placing one or more characteristics of one or more carrier waves using a digital code and placing information on the modified carrier wave formed in this way. In this way, the modulated carrier wave “carry” information. An unmodulated carrier wave may have a frequency of zero, that is, a constant level such as a voltage, or may change with time like a sine wave. Digital modulation modulates one or more of carrier wave amplitude, phase, and frequency. The purpose of digital modulation is to transmit information via one or more modulated signals, such as in a communication channel or a data storage channel.

  For clarity, the “transmission” of a digitally modulated signal is a signal path that includes the channel and any other element at either end of the channel that the signal must pass to enter and exit the channel. It shall refer to the passage of a digitally modulated signal. The term “channel” means a physical medium used for conducting or storing signals. Examples of channels include twisted pair wires, coaxial cables, optical fibers, and space electromagnetic waves. In addition to channels, the signal path includes components or elements coupled to either channel end to provide digital modulation signals to the channels and receive digital modulation signals from the channels.

  In the transmission of digitally modulated signals in systems designed for digital communication or data storage, the signals are often subject to linear and nonlinear distortions. Since the signal deteriorates due to such distortion, in order to extract the information from the signal with certainty, it is necessary to take corrective measures when receiving the signal.

  Linear distortion changes the state of the signal during transmission. In this connection, the channel through which the signal is transmitted disperses the amplitude and phase of the signal component to an unequal level that depends on the frequency of the signal component. As a result, the received signal is impaired, and intersymbol interference may occur. Such a channel is called a “dispersive channel”. A channel whose output signal changes in direct proportion to changes in the input signal or some input signal components may be considered a “linear channel”. However, in such a channel, components of different frequencies pass through the channel at different speeds and are attenuated by different factors. These effects of linear distortion can be improved by equalizing the received signal. The linear equalizer removes or reduces the effect of linear distortion by adjusting the components of the received signal and correcting for component changes caused by transmission through the channel. The linear effect does not create new frequencies that were not part of the undistorted signal.

  Nonlinear distortion occurs when the proportionality or linearity of a signal is distorted to some extent. Such non-linear effects usually do not spread throughout the signal path, but are concentrated in a specific location. Examples of non-linear distortions include: (1) a driver at the input to a channel or intermediate channel repeater that exhibits some degree of non-linearity depending on the signal amplitude or amplitude derivative (slew rate), and (2) some degree of non-linearity ( Corroded contacts in channels with (non-ohmic) characteristics, and (3) transformers in channels that exhibit large non-linearities that may be related to magnetic hysteresis in the magnetic core. In general, non-linear effects generate new frequencies that were not part of the original undistorted signal.

  A single channel may provide reverse transmission capability for two signal paths. In bi-directional transmission over a single shared channel, it is necessary to provide means in the channel to separate the output signal from the input signal at both ends of the channel. It may also be necessary to provide repeater means in the channel that can separate the reverse signal and recombine the reverse signal in the middle of both ends of the channel.

  Since this separation process is not perfect, a portion of the signal transmitted at one channel end may reverberate or “echo”. This echo signal overlaps with the input signal and may be regarded as if it were noise. In some communication systems, the input signal and the output signal share all or part of the same frequency band. In such a case, the echo signal level is often reduced using a linear echo canceller. This is accomplished by modeling the effects of various system components and channels on the echo signal to produce a predicted value of the echo signal. Subtracting this from the input signal greatly reduces the echo signal and improves the signal-to-noise ratio. In other systems, the bands are separated for the two directions. This is a scheme known as frequency division duplexing or FDD. Although the linear portion of the echo signal of one signal remains in the proper band, this is not the case for nonlinear echo signals. The reason is that nonlinear echo signals often contain new frequency components that extend far beyond this band and may overlap the opposite band. Thus, the multistage nonlinear echo canceller that is the subject of this specification can benefit both FDD and non-FDD systems.

  In some cases, the signal to be transmitted may contain a low level out-of-band component. For example, undesirable effects may occur when using a limited sampling rate, or glitches may occur in data during transitions between adjacent data symbols. These components can be removed by carefully filtering the signal before transmission. Further, the effect can be corrected later by linear echo cancellation. If all out-of-band components are removed before transmission, the linear echo canceller makes no sense for the FDD scheme.

  Some solutions to the echo cancellation problem require both nonlinear echo cancellation and linear echo cancellation. For example, US Pat. No. 5,627,885 discusses a method in which a linear echo canceller and a nonlinear echo canceller are provided separately. However, these echo cancellers are connected not in series but in parallel.

  The multi-stage nonlinear echo canceller (MNEC) according to the present invention addresses signal correction problems to eliminate linear echo and nonlinear echo effects, especially when nonlinear distortion of the echo signal is unavoidable at one or more locations in the linear signal path. It provides an effective solution. Because MNEC with adjustable parameters can adapt to a range of different situations, it is not necessary to know in advance the location and characteristics of the non-linear effect (s). MNEC includes multiple stages. Each stage is characterized by a function that models the distortion in the echo signal of the transmitted signal. The stages are connected in a sequence in which the functions of each stage are alternately arranged in the order of a linear function and a nonlinear function. This sequence includes a plurality of stages. One or more linear stages are provided to remove linear distortion to the echo signal in one or more sections of the signal path where there is no significant nonlinear effect. One or more non-linear stages are provided to correct distortion due to local non-linear effects. As long as the nonlinear distortion is local, the accuracy of modeling and correction by a nonlinear stage characterized by a function having a relatively small number of adjustable parameters can be increased.

  If there is one major source of nonlinear distortion localized anywhere in the echo signal path, the preferred embodiment of the multistage equalizer has three stages: a linear, nonlinear, and linear sequence. . In this case, MNEC acts directly on the copy of the transmitted signal to model the linear distortion that occurs in the portion of the echo signal path between the start of the echo signal path and the location of the nonlinear distortion source. A first linear stage characterized by a function that produces a result is included. For example, this linear distortion may occur in some filtering that is added to the signal to reduce out-of-band components before transmission. The MNEC further includes a second non-linear stage coupled to the first stage. This second stage features a second function that acts on the first result to produce a second result that models the non-linear strain generated by the non-linear strain source. For example, such non-linear distortion may occur in digital / analog converters, line drivers, line coupled transformers, corrosion of connections in the lines, or combinations thereof. A third linear stage coupled to the second stage operates on the second result to model linear distortion in the signal path portion between the location of the nonlinear distortion source and the end of the echo signal path. For example, this linear distortion is characterized by impedance mismatch between the hybrid circuit and the line, a change in the line number on the line, a bridge tap attached to the line, a bad junction on the line May occur in a receiver input filter to reduce out-of-band components, or a combination thereof. The third result is the output signal of MNEC. In a conventional echo canceller, this output signal is subtracted from the received signal in order to cancel all or part of the echo signal. This configuration of the MNEC can be easily generalized even when there are multiple locations where nonlinear distortion occurs in the echo signal path. In such a case, there may be four or more stages of the multistage equalizer. In these cases, both the first and last stages are often linear, except when the nonlinearity is accurately specified at the start and end points of the echo signal path (in this case, The linear step may be omitted).

  The function characterizing the stage may include adjustable parameters. In order to set these parameters to values that minimize a particular error level, an echo canceller controller is coupled to the linear and non-linear stages. In general, function parameter adjustment utilizes an optimization process to minimize the error level.

  The present invention is illustrated in one or more of the above drawings and is disclosed in detail in the following description. Here, the elements that are “connected” are illustrated and described, but this is to order these elements and basically explain how these elements work in concert with each other. . Accordingly, it is within the scope of the present invention to place other elements not shown or described herein in connection between the elements shown and described. The present invention is applied to digital communication via a bidirectional communication channel. This channel may be embodied in any of a plurality of media.

  Unlike conventional echo cancellers, the multi-stage nonlinear echo canceller (MNEC) according to the present invention is used when upstream and downstream data occupy separate frequency bands, ie, in a process known as frequency division duplexing or FDD. be able to. FIG. 1 shows such a separated upstream band (110) and downstream band (120). Through a non-linear distortion process, the upstream signal may generate a series of distortion factored signal elements in a band (130) that may be wider than the upstream band and overlap the downstream band. Similarly, a downstream signal may generate a band formed by a series of signal elements with distortion coefficients that may overlap an upstream band (not shown). These series of distortion factor signal element echo signals interfere with the true signal in that band, reducing the maximum data rate. The rate of data rate reduction is related to how much the effective "noise floor", which is increased due to a series of distortion factor signal elements, exceeds the background level when there is no signal on the line. The MNEC function simulates both the distortion and echo effect processes, creates a model of the interference signal, and subtracts this model from the input signal to remove these unwanted components and increase the maximum data rate. It is. If the upstream and downstream bands overlap (ie, not FDD), the MNEC performs a dual function of replacing the standard linear echo canceller and attempting to remove a series of signal elements with nonlinear distortion coefficients in the echo signal.

  FIG. 2 is a block diagram illustrating one end of a digital communication system, in which input data 202 to be transmitted to a destination is provided to an encoding and modulation circuit 204. (Note that digital information can be processed in either hardware or software. This is true for all components in FIG. 2 except those labeled with reference numbers 210-224, whose signals are in analog form.) The circuit 204 associates the input data 202 with a digital code. This encoded data is divided into a series of symbols. Each symbol represents digital data having a certain number of bits. These symbols are then used to modulate a carrier or carrier set with one or more amplitudes, frequencies, and phases. For every allowed symbol, there are specific settings for these carrier parameters. These parameters remain fixed for a certain period before switching to the parameter representing the next symbol. A digital modulation signal 206 is generated by circuit 204. These signals 206 represent the modulated carrier to be transmitted in digital form.

  A digital modulation signal 206 is provided to a digital / analog converter (DAC) 208 and an MNEC 207. The DAC 208 converts the digital modulation signal into an analog format 210. The signal 210 output from the DAC 208 is coupled to the analog filter 212. The purpose of this analog filter is to reduce the level of the out-of-band component of the signal 210, but as a side effect, it may cause phase distortion and amplitude distortion in the in-band component. This filtered output signal is connected to a hybrid circuit 216 that serves to separate the output signal from the input signal via the bi-directional communication channel 218. The hybrid circuit 216 (not explicitly shown) includes an output side line driver, an input side preamplifier, a line coupled transformer, and other components that may vary depending on the design.

  It is ideal if the hybrid 216 can completely separate the transmission signal and the reception signal on the line. Unfortunately, it is not completely separated in practice. The communication channel 218 echoes or echoes a portion of the output signal due to various actions, so that this portion of the output signal exits the hybrid 216 as if this portion of the output signal was a portion of the received signal 220. There are things to do. For twisted pair telephone lines, these effects include impedance mismatch between hybrid 216 and line 218, line number changes on the line, bridge taps attached to the line, and poor junctions on the line. . In general, the echo signal is also subjected to a part of the DAC 208, the hybrid 216, and a part of non-linear distortion that may occur due to a defective connection or a defective connection in the line 218. As a result, the MNEC according to the present invention is intended to solve the problem that a series of signal elements with nonlinear distortion coefficients are generated in the echo signal. Received signal 220 is connected to analog filter 222. The purpose of this analog filter is to reduce the intensity of the out-of-band component, but as a side effect, it may cause phase distortion and amplitude distortion in the in-band component. The output side of the analog filter 222 is connected to an analog / digital converter (ADC) 226. The received signal that has passed through the ADC is in a digital format. At this point, the output signal of MNEC 207 is subtracted at 230 from the received signal. The function of the MNEC 207 according to the present invention is to provide a series of distortion factors to be subtracted from the input signal to improve the signal-to-noise ratio by eliminating (or partially eliminating) the actual echo signal contained in the input signal. Generating a model of an expected echo signal including signal elements. After subtracting at 230, the signal is connected to an equalizer 234 to correct for channel effects on the received signal. Finally, the signal is demodulated and decoded 238 to produce output data 240.

The configuration, function, and operation of a multi-stage nonlinear echo canceller (MNEC) 237 according to the present invention may be understood with reference to FIGS. 3a-3d, 4, 5a-5g, and 6a-6b. . The MNEC consists of a sequence that includes at least two stages. Each stage has a digital time series x ₁ , x ₂ , x ₃ ,. . . And another digital time series y ₁ , y ₂ , y ₃ ,. . . Is generated. Is that certain elements of the output signal sequence depends on the value of a plurality of elements of the input signal sequence (e.g., y _n is _{x n-2, x n-} 1, x n, x n + 1, and x _{n + 2} ). A stage features each function that may depend on a plurality of configurable parameters as required. The number of configurable parameters need not be the same at each stage. Since the function completely describes the structure and operation of the stage characterized by the function, the function will be described below in place of the description regarding the stage. The configurable parameters of the function characterizing the stage allow the MNEC to adapt to signal paths that include channels with unknown or variable characteristics. Each stage is connected in the following order. That is, the input side to the MNEC becomes the input side to the first stage, the output side of the first stage is connected to the input side of the second stage, the output side of the second stage is connected to the input side of the third stage, This is repeated until the stage, and the output side of the last stage is connected to become the output side of the MNEC. Stages are divided into two types: “linear” and “non-linear”. Which type is the linearity of the change in the output signal time series of the stage (and in all parameter settings) to the changes made to the input signal time series elements of that stage (ie, It depends on whether it changes (in direct proportion). MNEC always follows a sequence that includes at least one linear stage and one non-linear stage. The output signal value in the linear stage usually depends on a large number of input signal values required to correct the channel dispersion. In contrast, non-linear stage output signal elements typically depend only on one or a few input signal elements. Generally, in a sequence of stages, two types of stages are alternately arranged. However, this rule may not be observed in special cases.

  The following guidelines apply when selecting a specific MNEC configuration. An echo signal is generated when the signal follows an “echo path” that is subject to various sources of linear and nonlinear distortion. As can be seen by referring again to FIG. 2, this path is from the input 206 of the DAC to the output 228 of the ADC. This path includes DAC 208, analog filter 212, hybrid 216, channel 218, analog filter 222, and ADC 226. In some cases, other devices may be connected to the channel and interact with the echo signal (eg, a telephone that shares the same twisted pair wire as digital communications). For short channels, the echo signal may interact with a modem or other device connected to the opposite end of channel 218 (not shown).

  It is believed that the echo signal path includes one or more local nonlinear distortion sources. Each nonlinear effect is believed to act at a particular location in the echo signal path according to a one-dimensional (or at least low-dimensional) mechanical variable, such as the amplitude of the signal or its first derivative. The design approach underlying the multi-stage approach assumes the idea that the echo signal path can be viewed as a series of linear intervals separated by these local nonlinear effects. There is no need to know the details and location of these effects in advance. One of these effects occurs near the beginning of the echo signal path. The first step in the echo cancellation process is to use linear steps to reproduce the relevant dynamic variable (s) from the transmitted (unknown) signal. The nonlinear strain is then removed from the dynamic variable (s) by a nonlinear step following the linear step. In many cases, this non-linear stage may be characterized by a power series expansion of a single variable with only a few terms. This non-linear stage output signal is used to represent the echo signal immediately after receiving the non-linear effect. If there is only one significant nonlinear effect in the echo signal path, this output signal may be passed to another linear stage. MNEC uses this linear step to correct for additional linear distortion of the echo signal to the end of the echo signal path. The above is the description of the three-stage configuration of the linear, non-linear, and linear sequences shown in FIG. Alternatively, a sequence can be added in which the non-linear and linear stages alternate in sequence until all non-linearities are corrected. FIG. 3b shows a five-step sequence that can be used to model an echo signal path that includes nonlinear effects occurring at two different locations. Note that usually the beginning and end of MNEC are linear stages. However, if one of the nonlinear distortion locations is at or near one of the echo signal path ends, an exception is made so that no significant linear distortion effects remain between this location and the echo signal path ends. May be.

  Since echo signals may occur at multiple locations in the system, multiple “echo signal paths” may need to be considered in order to properly model their effects. In order to realize this, a plurality of echo cancellers may be arranged in parallel. In this case, all may be sequentially MNEC type, or a part may be a linear echo canceller. In the typical example shown in FIG. 3c, a three-stage MNEC is arranged in parallel with a one-stage linear echo canceller. As can be seen, both echo cancellers receive the same input data, but both output signals are summed. Note that the added echo canceller can only be placed in parallel with some stages of MNEC. For example, linear stages can be placed in parallel with the last three stages of MNEC shown in FIG. 3b. More complex configurations are possible. FIG. 3d shows a particularly useful configuration among them.

  The configuration shown in FIG. 3d is a multistage apparatus that combines the functions of a multistage equalizer and the multistage nonlinear echo canceller. In this device, the two linear stages each have an input side for receiving transmission data input to the channel and an input side for receiving reception data connected from the channel, and at least two input sides and Each output side is provided in a non-linear stage having one output side. Such a device may be useful when both the input signal (received data) and the output signal (transmitted data) echo effects interact with the same source of nonlinear distortion. A typical example is a non-linear characteristic of an incomplete junction in a twisted pair wire. The output signals of the two linear stages that provide the respective signals to the two inputs of the nonlinear stage are not necessarily summed, but rather the role of the individual inputs to some nonlinear function that characterizes the nonlinear stage. Can be fulfilled. As a result, the output signal and the input signal can be combined. If such coupling is not considered important, both the non-linear multistage equalizer and the MNEC can be used without coupling, if desired.

  FIG. 4 shows a preferred embodiment of the linear stage. Here, this linear stage is embodied as a finite impulse response (FIR) filter. One skilled in the art will appreciate that such elements can be characterized or described by the functions in which they are implemented. In this case, this embodiment is implemented by the following function 410.

図４に示した関数４１０を図３ａに示したＭＮＥＣの第１段階として用いる場合、入力信号はデジタル値ｕ₀，ｕ₁，ｕ₂，ｕ₃．．．の時間系列となる。この関数では、連続する各デジタル値が、デジタル値と組み合わされて（乗算されて）積を算出する値を有するパラメータ（この場合は係数α_k）に関連付けられる。指数ｋの範囲には通常、選択された開始値ｋ₀と終了値ｋ₁との間のすべての整数値が含まれる。これらの値は、正数、負数、またはゼロのいずれであってもよい。どの値を用いるかは、特定チャネルの分散性およびその他の特性によって決まる。必要であれば、図に示したように、信号レベルの変化を補正するために定数パラメータＡを含めてもよい。すべての積が合計され、Ａに加算されて、デジタル値ｘ₀，ｘ₁，ｘ₂，ｘ₃．．．の出力信号時間系列の単一要素が生成される。各パラメータとデジタル値の関連付けを時間系列において１つ前に進めることで、出力信号時間系列の次の要素が生成される。この時間系列が、図３ａに示したＭＮＥＣの第１段階により生成される第１結果を構成する。以下に詳細に説明する方法で、エコーキャンセラコントローラにより係数の値が設定・変更される。

When the function 410 shown in FIG. 4 is used as the first stage of the MNEC shown in FIG. 3a, the input signal has digital values u ₀ , u ₁ , u ₂ , u ₃ . . . It becomes the time series of. In this function, each successive digital value is associated with a parameter (in this case a coefficient α _k ) having a value that is combined (multiplied) with the digital value to calculate a product. The range of index k typically includes all integer values between the selected start value k ₀ and end value k ₁ . These values may be positive numbers, negative numbers, or zero. Which value to use depends on the dispersibility and other characteristics of the particular channel. If necessary, a constant parameter A may be included to correct changes in signal level as shown in the figure. All products are summed and added to A to produce digital values x ₀ , x ₁ , x ₂ , x ₃ . . . A single element of the output signal time sequence is generated. The next element of the output signal time series is generated by advancing the association between each parameter and the digital value one time in the time series. This time sequence constitutes the first result generated by the first stage of the MNEC shown in FIG. 3a. The coefficient value is set / changed by the echo canceller controller by the method described in detail below.

The second stage of MNEC in FIG. 3a (assumed to be a non-linear stage here) receives the first result as an input signal time sequence and outputs signal time sequences y ₀ , y ₁ , y ₂ , y ₃ . . . As a second result. With reference to FIGS. 5a-5g, embodiments of multiple non-linear stages may be understood. In FIG. 5a, the non-linear stage is implemented by a basic power series function 514. This is the preferred embodiment in most cases, especially for amplitude-based nonlinearities. This is a one-dimensional (1-D) stage. Here, one dimension means that the output signal y _n depends only on the single output signal x _n . Due to the low dimensions, this stage is simple compared to the linear stage (FIG. 4), where there are often dozens or hundreds of dimensions or numbers of “taps”. The function 516 shown in FIG. 5b, when the exact expression f (x) of the nonlinear function is known at or near the channel, is embodied in this case by f (x) that also characterizes the nonlinear stage, Non-linear stages can be characterized. FIG. 5c shows a difference-based 1-D nonlinear function 518 that can characterize the nonlinear stage. This nonlinear stage embodiment may be used as an alternative to function 514 to correct for slew rates associated with nonlinearity. Depending on the number of parameters, this non-linear stage may be implemented according to the general 1-D function 520 shown in FIG. 5d. This is true for other types of series expansions that are convenient for a particular application, splines or other methods that approximate the shape of the nonlinear function directly. In the non-linear stage embodiment shown in FIG. 5e, the function 522 that characterizes this stage is implemented as a non-linear 2-D power series that cannot be reduced to 1-D as it depends on amplitude and slew rate simultaneously. May be. This is often an embodiment where the function depends on two consecutive input signal values. The non-linear stage embodiment of FIG. 5 f features a D-dimensional general function 524. As D increases, this embodiment becomes less desirable. FIG. 5g shows an embodiment of the double input side non-linear stage as shown in FIG. 3d. The non-linear function that characterizes this stage may be implemented as a series to utilize one element from each of the two input signal sequences. Up / down coupling can be corrected when there are cross terms in the series that include various powers of both x and y. It is clear that this can be generalized to any D-dimensional nonlinear function that depends on the A element from the first input signal and the B element from the second input signal (where A + B = D). . Since A and B must be at least 1, D must be at least 2.

  Steps may be added as needed, as shown in FIG. 3b. In this case, generally, linear stages and non-linear stages are alternately arranged.

By configuring the MNEC so that the linear stage and the non-linear stage are alternately arranged, the number of adjustable parameters required for modeling the echo signal can be reduced. For example, each of the two linear stages is characterized by a function having 50 input variables (50 values from k ₀ to k ₁ ), and the nonlinear stage is characterized by a function using the power series of one variable up to the fifth power. Let's consider an MNEC configuration consisting of three stages. If any constant term is included in each linear stage, this configuration requires a total of 106 adjustable parameters. When these steps are combined to form an MNEC, each output signal value depends on 99 input signal values. For this reason, in order to obtain the same result as in the single stage, the nonlinear echo canceller needs a nonlinear function for processing 99 variables. A suitable non-linear function can be created using all 5th power forms possible with 99 variables, in which case 91,962,520 adjustable parameters are required. This number is undesirable because it significantly exceeds the number of tunable parameters 106 required in the above example.

  The linear stage may be implemented in the frequency domain by placing a Fourier transform such as the fast Fourier transform (FFT) 610 shown in FIG. 6a before the linear stage. In this case, the linear stage is characterized by a function 620 of the form shown in FIG. This function is simpler than the function 410 of FIG. 4 because there is no addition. Any linear stage can be implemented in this way. In that case, an FFT is placed before each linear stage and an inverse FFT is placed after each linear stage. If the MNEC wants to have an input signal and / or output signal in the frequency domain, the first FFT and / or the last inverse FFT can be removed. This is useful when a Fourier transform is used in the demodulation process. This applies to DMT (Discrete Multitone) modulation, which is often used in ADSL (Asymmetric Digital Subscriber Line) communication systems. The number of coefficients required in the frequency domain (two for each channel of the DMT signal) may be undesirably high for optimization. In such a case, the number of coefficients may be reduced by assuming that these series of parameters are on one or more smooth curves. These curves can be characterized by sparse marker points used for optimization. A complete parameter set can be obtained from these marker points using cubic splines or other interpolation methods.

  FIG. 7a is an example showing what can be achieved using the MNEC of the present invention. The system of FIG. 2 was partially implemented in hardware and software, respectively. A standard DMT (Discrete Multitone) signal modulated with known data was transmitted over a 12,000 foot actual twisted pair and the echo signal was received for processing at the same (transmission) end. Possible sources of nonlinear distortion are a line driver (part of a hybrid) and a line coupling transformer (but not limited to these). This graph shows the residual noise level corresponding to the numbers of the DMT channels at intervals of about 4 kHz in a uniform scale. The upper curve 710 shows the best result that could be achieved with linear echo cancellation. The middle curve 720 shows the results improved by the MNEC of the present invention using a configuration as complex as the linear echo canceller result 710. Noise is reduced to less than 1/2 (almost 8 dB). The bottom curve 730 shows the background level when no signal is transmitted. The upward protrusion near channel 14 is due to an unidentified noise source that is part of the background.

  FIG. 7b shows the experimental results when there is a large out-of-band echo signal due to non-linear effects when using separated upstream and downstream bands. The transmission signal (upstream band) is limited to channel 6 to channel 31 (not shown). All the channels shown in the figure are included in the downlink band. Without non-linearity, there will be no upstream echo signal on these channels. However, it can be seen from the curve 740 that is not affected by linear echo cancellation at all that there is a large echo signal. If this echo signal is not corrected, a corresponding error will occur in the recovery of the input data contained in these downstream channels. Applying the MNEC according to the present invention gives a result like curve 750 which shows a considerable improvement. Thereby, the data rate which a track | line can respond | corresponds becomes high. The lower curve 760 shows the background level when there is no transmission signal. It has been found that the source of the nonlinear echo signal in this particular example is a contact failure in the line.

  In FIG. 8, the MNEC or its copy is connected in a different configuration to optimize the parameters. A test data set is generated and transmitted over the line. The echo signals of this signal are collected and stored. An echo canceller controller 824 is used to identify optimal parameters. The echo canceller controller provides a series of experimental parameters to each stage of the MNEC 806 (808, 810, and 812). Next, test data 802 is processed by the MNEC and subtracted 814 from the captured echo data 816. The result is processed by the controller 824 and used in the parameter value refinement process.

  FIG. 9 shows a series of initial configuration steps for operating the MNEC according to the present invention. Since the optimization technique is used, the process of FIG. 9 can also be called an optimization process. This figure is a flowchart specifically showing the processing executed by the echo canceller controller in accordance with the error measurement value. In the description in this flowchart, specifically, it is assumed that the configuration of each stage of the MNEC characterized by the functions shown in FIGS. 4 and 5a to 5f is standard. Such configurations include general purpose or dedicated digital processors, programmable logic, application specific integrated circuits (ASICs), digital signal processors (DSPs), and software routines executed by any equivalent means (provided that However, the present invention may be implemented in any of various forms. Usually, these functions are implemented in storage cell arrays connected in sequence. These cells can be accessed from the tap to obtain the values stored in these cells and combine them (typically multiplication) with the coefficients. The coefficient values are stored in a location accessible for the purpose of setting and changing these values.

In FIG. 9, the MNEC according to the present invention is initialized at step 940. Initialization means that the values of all parameters of each function are set to predetermined initial values. As an option for setting, the parameter can be initialized to a value that allows the input signal to be converted to an output signal with little or no change. For example, when initializing the linear stage of FIG. 4, the parameter α _k can be set to 1 and all other parameters including the constant A can be set to 0. Also, all parameters can be set to 0 when initializing the non-linear stage of FIG. 5a. As another option, the best value of the parameter obtained in the previously attempted optimization can be used as the initial value. Next, a known signal, usually containing a pseudo-random sequence of data transmitted and received and stored over the channel 942, is processed by the MNEC. Using the error measurement obtained at 944, usually the least square error value obtained by subtraction (820 in FIG. 8) and a specific optimization routine (eg, a modified Powell routine), a new value set of function parameters is obtained at step 946. Selected. At 948, the routine loops repeatedly through step 942, step 944, and step 946. At each loop, the stored data set is reprocessed with a new set of parameters throughout each stage until the error measurement converges to a minimum value. The parameters are then fixed and can be used to model the echo signal to obtain actual (unknown) data. A plurality of methods for defining an error value at 944 are conceivable. For example, if the residual signal after subtraction (820 in FIG. 8) is processed by Fourier transform, the magnitude of the error remaining in the specific frequency channel of interest can be examined.

  In MNEC configurations with changes as shown in FIGS. 3c and 3d, additional steps need to be made to optimize the parameters. In the case of the parallel linear echo canceller shown in FIG. 3c, there are at least two approaches.

  In the first method, all parameters are treated equally and all parameters are optimized simultaneously using the above method. In the second approach, the parameters of the linear echo canceller (stage 1B in FIG. 3c) are first set using standard methods well known in the art, and the remaining stages (stages 1A, 2A, and 3A in FIG. 3c). Keep off. Then all stages can be turned on. Next, with the parameters of the linear echo canceller fixed, the remaining parameters are optimized using the above method. For the equalizer / echo canceller multi-stage combination (MCOM) shown in FIG. 3d, the optimization process is a bit complicated. In this case, it is necessary to collect input data having known signals that are simultaneously transmitted and received via the line. This received data set and known transmission data is then sent to the appropriate input of MCOM and processed by MCOM. The output signal can then be subtracted from the expected result (a known signal transmitted from the opposite end of the line). The difference is used to generate an error measure that allows the optimization routine to iteratively find the best selected value for all parameters of MCOM.

  There are conceivable ways to effectively improve this optimization process. By deploying a dual MNEC (or MCOM), a new parameter set can be optimized while continuing to use the current parameter set without interrupting the actual data flow during the optimization process. When the number of parameters related to optimization is large, the progress of optimization may be slowed. Therefore, this method can avoid a situation where data cannot be processed for a long time. If there is currently no satisfactory parameter set, the preferred method is to quickly calculate a low optimization but functional parameter set that can be used while calculating a parameter set with a high level of optimization. . Another refinement is to divide the optimization process into two (or more) loops. This is particularly effective when the final stage is a linear stage. In this case, sufficient optimization can be performed only for the final stage parameters, while other parameters can remain fixed. Since the final stage is linear, this can be done very quickly in a fast inner loop. Other parameters are adjusted by a slow outer loop using a different optimization routine. While all other parameters pass through the slow loop, the final stage parameters are again fully optimized by the fast loop. Instead of a fast loop, the optimal parameters for the final stage can be directly calculated using a technique such as singular value decomposition.

  In a non-linear function such as can be used in this MNEC, if a set of coefficients that results in minimal error is found, a “local minimum” that is not as small as the true “global minimum” may be generated. There are well-known techniques for resolving such ambiguity. It may be necessary to start the optimization process from various initial conditions or use a somewhat slower optimization routine to find a global minimum. Useful techniques include (but are not limited to) modified Powell methods, conjugate gradient methods, multidimensional dithering, and simulated annealing methods. These and other means and methods for optimization and minimization are described, for example, in “Numeric Recipes: The Art of Scientific Computing with Listings in C”, William Press, et al. ., University University Press, 1989.

The frequency band of a system with frequency division duplexing or FDD is shown. Note that the upstream and downstream bands are separated, but the upstream non-linear distortion band may overlap the downstream band greatly. 1 is a block diagram showing elements at one end of a digital communication system incorporating a multi-stage nonlinear echo canceller according to the present invention. FIG. 2 shows two embodiments of the multi-stage nonlinear echo canceller (MNEC) of the present invention. 2 shows two embodiments of the multi-stage nonlinear echo canceller (MNEC) of the present invention. An embodiment in which two echo cancellers are combined in parallel is shown. The method of combining MNEC with a multi-stage nonlinear equalizer is shown. 1 illustrates one embodiment of the linear stage of the multi-stage nonlinear echo canceller of the present invention. Each embodiment of the non-linear stage of a multi-stage non-linear echo canceller is shown. Each embodiment of the non-linear stage of a multi-stage non-linear echo canceller is shown. Each embodiment of the non-linear stage of a multi-stage non-linear echo canceller is shown. Each embodiment of the non-linear stage of a multi-stage non-linear echo canceller is shown. Each embodiment of the non-linear stage of a multi-stage non-linear echo canceller is shown. Each embodiment of the non-linear stage of a multi-stage non-linear echo canceller is shown. Fig. 4 shows an embodiment of a double-input-side nonlinear stage that can be used in the combination of the multi-stage equalizer and echo canceller shown in Fig. 3d. Each of the linear stage Fourier transform and discrete frequency domain embodiments is shown. Each of the linear stage Fourier transform and discrete frequency domain embodiments is shown. For each of the in-band echo signal and the out-of-band echo signal, it is shown that when the multistage nonlinear echo canceller according to the present invention is used, the residual echo signal of the signal element with a series of nonlinear distortion coefficients of the transmission signal is improved. . For each of the in-band echo signal and the out-of-band echo signal, it is shown that when the multistage nonlinear echo canceller according to the present invention is used, the residual echo signal of the signal element with a series of nonlinear distortion coefficients of the transmission signal is improved. . A method for optimizing parameters for controlling the performance of a multistage nonlinear echo canceller using a controller is shown. 4 is a general flow diagram illustrating an optimization procedure according to the present invention for initializing and setting parameter values of a function characterizing a stage of a multi-stage nonlinear echo canceller.

Explanation of symbols

110: Up band 120: Down band 130: Band 202: Input data 204: Coding and modulation circuit, circuit 206: Digital modulation signal, signal, input signal 208: Digital / analog converter, DAC
207: MNEC, multistage nonlinear echo canceller (MNEC), multistage echo canceller 210: analog format, signal 212: analog filter 216: hybrid circuit, hybrid 218: bidirectional communication channel, communication channel, line, channel, bidirectional channel 220: Received signal, input signal 222: analog filter 226: analog / digital converter (ADC), ADC
228: output side 230: subtraction 234: equalizer 238: demodulation and decoding 240: output data 410: function 514: power series function, function 516: function 518: 1-D nonlinear function 520: 1-D function 522: Function 524: Function 610: Fast Fourier transform (FFT)
620: function 710: curve, linear echo canceller result 720: curve 730: curve 740: curve 750: curve 760: curve 802: test data 806: MNEC
808: Step 810: Step 812: Step 814: Subtraction 816: Echo data 820: Subtraction 824: Echo canceller controller, 940: Step 942: Step 944: Error level, Step 946: Step 948:

Claims (46)

In a digital transmission system, a combination for canceling an echo signal of a digital modulation signal to be transmitted through the dispersive medium from a digital modulation signal received from the dispersive medium,
A digital modulation signal source;
A digital to analog converter (DAC) having an input coupled to receive a digitally modulated signal;
An input coupled to receive the digitally modulated analog signal generated by the DAC and provide the digitally modulated analog signal for transmission in a dispersive channel; and a digitally modulated analog signal received from the dispersive channel A hybrid having an output for providing;
An analog to digital converter (ADC) coupled to receive the digitally modulated signal received from the dispersive channel and having an output side;
A multi-stage echo canceller having an input side coupled to the transmission digital modulation signal source and an output side;
Here, the multistage echo canceller includes a first function that generates a first result that models a linear distortion in an echo signal of the transmission digital modulation signal, and the echo of the transmission digital modulation signal. At least one second stage characterized by a second function that generates a second result modeling the nonlinear distortion in the signal from the first result;
A combination comprising a subtracter comprising a first input side connected to the output side of the multistage echo canceller, a second input side connected to the output side of the ADC, and an output side. The multistage echo canceller is
It has three or more stages,
Each stage has an input side and an output side, and in the sequence of these sequential stages, the input side of any stage following another stage is connected to the output side of the other stage. And
Each of the steps includes an input signal digital time series x ₁ , x ₂ , x ₃ ,. . . Corresponding to the output signal digital time series y ₁ , y ₂ , y ₃ ,. . . Characterized in that each element value of the output signal time series depends on the value of one or more elements of the input signal series,
The combination according to claim 1, wherein the respective functions of the stages in the sequence are alternately arranged in the order of a linear function and a nonlinear function. A third stage in the sequence is followed by one or more stages, and the respective functions of the one or more stages are alternating (non-linear, linear, non-linear, ...). 2. The combination according to 2. The linear function includes the following functions:

The combination according to claim 2, wherein A is a constant, α _k is a parameter of the function, and the parameter has a settable value. One or more of the nonlinear functions are a one-dimensional power series function, an inverse function of a known nonlinear function, an inverse function of a known piecewise linear function, a one-dimensional differential power function of order P, and a k parameter. The combination according to claim 2, which is a function in a group including a dependent one-dimensional function, a two-dimensional power series of order P, and a D-dimensional function depending on the k parameter. A digital to analog converter (DAC) having an input coupled to receive a digitally modulated signal;
A filter coupled to the DAC to filter the digitally modulated analog signal generated by the DAC;
The digitally modulated analog signal generated by the DAC and filtered by the filter is coupled to the filter to provide a dispersive channel for transmission and to receive the digitally modulated analog signal from the dispersive channel Hybrid and
An analog to digital converter (ADC) coupled to receive the digitally modulated analog signal received from the dispersive channel and having an output for providing a converted digitally modulated signal;
A multi-stage echo canceller having an input side coupled to a digital modulation signal source and an output side;
Here, the multistage echo canceller includes a first function characterized by a first function that generates a first result modeling a first linear distortion in the echo signal of the digitally modulated signal, and in series with the first stage. A second stage coupled to generate a second result from the first result that models non-linear distortion in the echo signal of the digitally modulated signal; and in the echo signal of the digitally modulated signal; And at least one third stage characterized by a third function that generates a third result modeling the second linear distortion from the second result,
In order to subtract the modeled echo signal of the transmitted digital modulated signal, corrected for linear and nonlinear distortion, from the transformed received digital modulated signal, the output side of the ADC and the multistage echo canceller A transmission / reception device for a digital modulation transmission system, comprising a subtractor coupled to the output side. The multistage echo canceller is
Comprising three or more stages connected in sequence to form a sequence;
Each stage has an input side and an output side, and the input side of any stage following the other stage in the sequence is connected to the output side of the other stage;
Each of the steps includes an input signal digital time series x ₁ , x ₂ , x ₃ ,. . . Corresponding to the output signal digital time series y ₁ , y ₂ , y ₃ ,. . . Characterized in that each element value of the output signal time series depends on the value of one or more elements of the input signal series,
The transmission / reception apparatus according to claim 6, wherein the functions of the steps in the sequence are alternately arranged in the order of a linear function and a nonlinear function. A third stage in the sequence is followed by one or more stages, and the respective functions of the one or more stages are alternating (non-linear, linear, non-linear, ...). 8. The transmission / reception device according to 7. The linear function includes the following functions:

The transmission / reception apparatus according to claim 6, wherein A is a constant, α _k is a parameter of the function, and the parameter has a settable value. One or more of the nonlinear functions are a one-dimensional power series function, an inverse function of a known nonlinear function, an inverse function of a known piecewise linear function, a one-dimensional differential power function of order P, and a k parameter. The transmission / reception apparatus according to claim 6, wherein the transmission / reception apparatus is a function in a group including a dependent one-dimensional function, a two-dimensional power series of order P, and a D-dimensional function depending on a k parameter. A method for modeling and canceling an echo signal of a digital modulation signal to be transmitted through the dispersive medium from a digital modulation signal received from the dispersive medium in a digital transmission system, comprising:
Providing a first digitally modulated signal;
Converting the first digital modulated signal to a first digital modulated analog signal;
Providing the first digitally modulated analog signal to a hybrid for transmission over a dispersive channel;
Providing from a hybrid a second digitally modulated analog signal received from the dispersive channel having an echo signal of the first digitally modulated analog signal;
Converting the second digitally modulated analog signal to a second digitally modulated signal;
Modeling a distorted echo signal of the first digital modulation signal;
Correcting the model of the distorted echo signal of the first digital modulation signal with respect to linear distortion using a first function and then at least using the echo function of the first digital modulation signal with respect to nonlinear distortion using a second function Generating a corrected echo signal of the first digital modulation signal by correcting the model of
Subtracting the corrected echo signal from the second digital modulation signal. Said step of generating a corrected echo signal comprises:
Correcting the model of the distorted echo signal of the first digitally modulated signal by a function comprising a sequence;
Each function has an input signal and an output signal, and the input signal of any function following another function in the sequence receives the output signal of the other stage;
Each of the functions is an output signal digital time series y ₁ , y ₂ , y ₃ ,. . . , Input signal digital time series x ₁ , x ₂ , x ₃ ,. . . And the value of any element of the output signal time series depends on the value of one or more elements of the input signal series,
The method according to claim 11, wherein the functions are alternately arranged in the order of a linear function and a nonlinear function in the sequence. The linear function includes the following functions:

13. The method of claim 12, wherein A is a constant, [alpha] _k is a parameter of the function, and the parameter has a configurable value. One or more of the nonlinear functions are a one-dimensional power series function, an inverse function of a known nonlinear function, an inverse function of a known piecewise linear function, a one-dimensional differential power function of order P, and a k parameter. 13. The method of claim 12, wherein the method is a function in a group comprising a dependent one-dimensional function, a two-dimensional power series of order P, and a D-dimensional function depending on the k parameter. Converting the first digital modulation signal to an analog format;
Filtering the first digital modulated signal in the analog format;
Coupling the filtered analog form of the first digital modulated signal to one end of a dispersive channel for transmission;
Receiving a second digitally modulated analog signal from the one end of the dispersive channel;
Modeling a distorted echo signal of the first digital modulation signal;
Generating a first result using a first function to correct a first linear distortion in the model of the distorted echo signal of the first digital modulation signal; and using a second function, the first digital Generating a second result from the first result to correct non-linear distortion in the model of the distorted echo signal of the modulation signal, and using at least one third function, the distortion of the first digital modulation signal. Generating a third result from the second result for correcting the second linear distortion in the model of the echo signal, and generating the corrected echo signal from the second digital modulation signal. Subtracting the echo signal, a transmitter / receiver echo canceling method. Said step of generating a corrected echo signal comprises:
Correcting the model of the distorted echo signal of the first digital modulation signal in a sequence of functions,
Each function has an input signal and an output signal, and the input signal of any function that follows another function in the sequence receives the output signal of the another function;
Each of the functions is an output signal digital time series y ₁ , y ₂ , y ₃ ,. . . , Input signal digital time series x ₁ , x ₂ , x ₃ ,. . . And the value of any element of the output signal time series depends on the value of one or more elements of the input signal series,
The method according to claim 15, wherein the functions are alternately arranged in the order of a linear function and a nonlinear function in the sequence. 17. The method of claim 16, wherein a third function is followed by one or more functions in the sequence, wherein the one or more functions are alternating (non-linear, linear, non-linear, ...). . The linear function includes the following functions:

The method according to claim 16, wherein A is a constant, α _k is a parameter of the function, and the parameter has a configurable value. One or more of the nonlinear functions are a one-dimensional power series function, an inverse function of a known nonlinear function, an inverse function of a known piecewise linear function, a one-dimensional differential power function of order P, and a k parameter. The method according to claim 16, wherein the function is in a group comprising a dependent one-dimensional function, a two-dimensional power series of order P, and a D-dimensional function depending on the k parameter. A digital to analog converter (DAC) having an input coupled to receive a digitally modulated signal;
A filter coupled to the DAC to filter the digitally modulated analog signal generated by the DAC;
A hybrid coupled to the filter to provide the digitally modulated analog signal generated by the DAC and filtered by the filter to a DSL channel for transmission and to receive the digitally modulated analog signal from the DSL channel; ,
An analog to digital converter (ADC) coupled to receive the digitally modulated analog signal received from the DSL channel and having an output for providing a converted digitally modulated signal;
A multi-stage echo canceller having an input side coupled to a digital modulation signal source and an output side;
Here, the multistage echo canceller has a first function characterized by a first function that generates a first result modeling a first linear distortion in an echo signal of the digital modulation signal, and in series with the first stage. A second stage coupled to generate a second result from the first result that models non-linear distortion in the echo signal of the digitally modulated signal; and in the echo signal of the digitally modulated signal; And at least one third stage characterized by a third function that generates a third result modeling the second linear distortion from the second result,
In order to subtract the modeled echo signal of the transmitted digital modulated signal, corrected for linear and nonlinear distortion, from the transformed received digital modulated signal, the output side of the ADC and the multistage echo canceller A transmitter / receiver for a digital subscriber line (DSL) system, comprising a subtractor coupled to the output. The multistage echo canceller is
Comprising three or more stages connected in sequence to form a sequence;
Each stage has an input side and an output side, and the input side of any stage following the other stage in the sequence is connected to the output side of the other stage;
Each of the steps includes an input signal digital time series x ₁ , x ₂ , x ₃ ,. . . Corresponding to the output signal digital time series y ₁ , y ₂ , y ₃ ,. . . Characterized in that each element value of the output signal time series depends on the value of one or more elements of the input signal series,
21. The transmission / reception device according to claim 20, wherein the respective functions of the stage in the sequence are alternately arranged in the order of a linear function and a nonlinear function. A third stage in the sequence is followed by one or more stages, and the respective functions of the one or more stages are alternating (non-linear, linear, non-linear, ...). 22. The transmission / reception device according to 21. The linear function includes the following functions:

The transmission / reception apparatus according to claim 21, wherein A is a constant, α _k is a parameter of the function, and the parameter has a settable value. One or more of the nonlinear functions are a one-dimensional power series function, an inverse function of a known nonlinear function, an inverse function of a known piecewise linear function, a one-dimensional differential power function of order P, and a k parameter. The transmission / reception apparatus according to claim 21, wherein the transmission / reception apparatus is a function in a group including a dependent one-dimensional function, a two-dimensional power series of order P, and a D-dimensional function depending on a k parameter. Converting the first digital modulation signal to an analog format;
Filtering the first digital modulated signal in the analog format;
Coupling the filtered analog form of the first digital modulated signal to one end of a DSL channel for transmission;
Receiving a second digitally modulated analog signal from the DSL channel;
Converting the second digitally modulated analog signal to a second digitally modulated signal;
Modeling a distorted echo signal of the first digital modulation signal;
Generating a first result using a first function to correct a first linear distortion in the model of the distorted echo signal of the first digital modulation signal; and using a second function, the first digital Generating a second result from the first result to correct non-linear distortion in the model of the distorted echo signal of the modulation signal, and using at least one third function, the distortion of the first digital modulation signal. Generating a corrected echo signal by generating from the second result a third result for correcting a second linear distortion in the model of the echo signal;
Subtracting the corrected echo signal from the second digital modulation signal, a transmitter / receiver echo cancellation method in a digital subscriber line (DSL) system. Said step of generating a corrected echo signal comprises:
Correcting the model of the distorted echo signal of the first digital modulation signal by a sequential function comprising a sequence;
Each function has an input signal and an output signal, and the input signal of any function following another function in the sequence receives the output signal of the other stage;
Each of the functions is an output signal digital time series y ₁ , y ₂ , y ₃ ,. . . , Input signal digital time series x ₁ , x ₂ , x ₃ ,. . . And the value of any element of the output signal time series depends on the value of one or more elements of the input signal series,
26. The method of claim 25, wherein in the sequence the functions alternate in the order of linear and non-linear functions. 27. The method of claim 26, wherein a third function is followed by one or more functions in the sequence, and the one or more functions are alternating (non-linear, linear, non-linear,...). . The linear function includes the following functions:

27. The method of claim 26, wherein A is a constant, [alpha] _k is a parameter of the function, and the parameter has a configurable value. One or more of the nonlinear functions are a one-dimensional power series function, an inverse function of a known nonlinear function, an inverse function of a known piecewise linear function, a one-dimensional differential power function of order P, and a k parameter. 27. The method of claim 26, wherein the function is in a group comprising a dependent one-dimensional function, a two-dimensional power series of order P, and a D-dimensional function depending on the k parameter. In a digital transmission system, a combination for canceling an echo signal of a digital modulation signal to be transmitted through the dispersive medium from a digital modulation signal received from the dispersive medium,
A digital modulation signal source;
A digital to analog converter (DAC) having an input coupled to receive a digitally modulated signal;
An input coupled to receive the digitally modulated analog signal generated by the DAC and provide the digitally modulated analog signal for transmission in a dispersive channel; and a digitally modulated analog signal received from the dispersive channel A hybrid having an output for providing;
An analog to digital converter (ADC) coupled to receive the digitally modulated signal received from the dispersive channel and having an input;
A multi-stage echo canceller having an input side coupled to the transmission digital modulation signal source and an output side;
Here, the multistage echo canceller includes a first function that generates a first result modeling a distortion in the echo signal of the transmission digital modulation signal, and the echo signal of the transmission digital modulation signal. A plurality of stages in the sequence including at least one second stage characterized by a second function that generates a second result modeling the distortion in the first result,
The sequence includes at least one linear stage and one non-linear stage;
A combination comprising a subtracter comprising a first input side connected to the output side of the multistage echo canceller, a second input side connected to the output side of the ADC, and an output side. The multistage echo canceller is
It has three or more stages,
Each of the stages has an input side and an output side, and the input side of any stage following the other stage in the sequential stages in the sequence is connected to the output side of the other stage. And
Each of the steps includes an input signal digital time series x ₁ , x ₂ , x ₃ ,. . . Corresponding to the output signal digital time series y ₁ , y ₂ , y ₃ ,. . . 31. The combination of claim 30, characterized by each function that generates a value, wherein the value of any element of the output signal time series depends on the value of one or more elements of the input signal series. A third stage in the sequence is followed by one or more stages, and the respective functions of the one or more stages are alternating (non-linear, linear, non-linear, ...). The combination according to 31. The linear function includes the following functions:

32. The combination of claim 31, wherein A is a constant, [alpha] _k is a parameter of the function, and the parameter has a configurable value. One or more of the nonlinear functions are a one-dimensional power series function, an inverse function of a known nonlinear function, an inverse function of a known piecewise linear function, a one-dimensional differential power function of order P, and a k parameter. 32. The combination of claim 31, wherein the combination is a function in a group that includes a dependent one-dimensional function, a two-dimensional power series of order P, and a D-dimensional function that depends on a k parameter. 32. The combination of claim 31, wherein one or more of the linear functions are implemented in the frequency domain. A linear echo canceller having an input side connected in the same way as the input side of one said stage and an output side;
A first input side connected to the output side of the linear echo canceller and a second input side connected to the output side of one stage of the plurality of stages to another stage of the plurality of stages. 31. The combination of claim 30, further comprising an adder that provides an input signal or an output signal that is combined as the output signal of the multistage echo canceller. The plurality of stages further includes a last stage;
The input side of the linear echo canceller is connected similarly to the input side of the first stage,
The output side of the linear echo canceller is connected to the first input side of the adder, and the output side of the last stage is connected to the second input side of the adder. 36. The combination according to 36. The multi-stage echo canceller includes three or more stages forming a sequence, and the linear echo canceller includes a single stage;
Each stage has an input side and an output side,
Each of the steps includes an input signal digital time series x ₁ , x ₂ , x ₃ ,. . . Corresponding to the output signal digital time series y ₁ , y ₂ , y ₃ ,. . . 37. A combination according to claim 36, characterized by each function generating, wherein the value of any element of the output signal time series depends on the value of one or more elements of the input signal series. The linear function includes the following functions:

39. The combination of claim 38, wherein A is a constant, [alpha] _k is a parameter of the function, and the parameter has a configurable value. One or more of the nonlinear functions are a one-dimensional power series function, an inverse function of a known nonlinear function, an inverse function of a known piecewise linear function, a one-dimensional differential power function of order P, and a k parameter. 40. The combination of claim 38, wherein the combination is a function in a group that includes a dependent one-dimensional function, a two-dimensional power series of order P, and a D-dimensional function that depends on a k parameter. 40. The combination of claim 38, wherein one or more of the linear functions are implemented in the frequency domain. In the digital transmission system, the echo signal of the digital modulation signal to be transmitted through the dispersive medium is erased from the digital modulation signal received from the dispersive medium, and the effect of nonlinear coupling between the transmission signal and the received signal Is a device for equalizing the received signal,
A digital modulation signal source;
A digital to analog converter (DAC) having an input coupled to receive a digitally modulated signal;
An input coupled to receive the digitally modulated analog signal generated by the DAC and provide the digitally modulated analog signal for transmission in a dispersive channel; and a digitally modulated analog signal received from the dispersive channel A hybrid having an output for providing;
An analog to digital converter (ADC) coupled to receive the digitally modulated signal received from the dispersive channel and having an output;
A multi-stage combination having a first input side coupled to the transmission digital modulation signal source, a second input side coupled to the output side of the ADC, and an output side;
The multi-stage combination has a first sequence including zero or more stages having an input side and an output side connected to the first input side, and an input side and an output side connected to the second input side A second sequence including zero or more stages, two input sides and one output side, wherein the first input side of the two input sides is connected to the output side of the first sequence. The second input side includes an engagement stage connected to the output side of the second sequence and zero or more stages having an output side and an input side, wherein the input side is the coupling A third sequence connected to the output side of the stage,
The first sequence generates a first result that models distortion in the transmission signal, and the second sequence generates a second result that corrects distortion in the received signal, and the combining step Is to generate a third result that corrects the coupling between the transmitted signal and the received signal and eliminates the echo signal, and the third sequence generates a fourth result that corrects the distortion in the received signal. Yes,
The apparatus wherein the multi-stage combination includes at least one non-linear stage. Each of the first sequence, the second sequence, and the final sequence includes a single linear stage, each linear stage has an input side and an output side, and each linear stage is an input signal Digital time series x ₁ , x ₂ , x ₃ ,. . . Corresponding to the output signal digital time series y ₁ , y ₂ , y ₃ ,. . . Wherein the value of an arbitrary element of the output signal time series depends on the value of one or more elements of the input signal series, and the combining step comprises: The sequences x ₁ , x ₂ , x ₃ ,. . . And y ₁ , y ₂ , y ₃ ,. . . Corresponding to the function of generating the output signal digital time series z ₁ , z ₂ , z ₃ , K, and the value of an arbitrary element of the output signal time series generated by the combining stage is the input signal 43. The apparatus of claim 42, wherein the apparatus is dependent on the value of one or more elements of each series. The linear function includes the following functions:

44. The apparatus of claim 43, wherein A is a constant, [alpha] _k is a parameter of the function, and the parameter has a configurable value. The combination stage is characterized by a two-dimensional power series function that depends on one element from each input side, a two-dimensional function that depends on a k parameter and one element from the two input sides, and a k parameter. 44. A non-linear function from a group comprising an A element from the first input side and a B element from the second input side, where A + B = D, and a D-dimensional function depending on The device described. 44. The apparatus of claim 43, wherein one or more of the linear functions are implemented in the frequency domain.

Images