WO2002087186A2

WO2002087186A2 - Channel characterization and rate selection in dsl systems

Info

Publication number: WO2002087186A2
Application number: PCT/US2002/012857
Authority: WO
Inventors: Farrokh Rashid-Farrokhi; Cindy C. Wang; Steven R. Blackwell
Original assignee: Centillium Communications, Inc.
Priority date: 2001-04-23
Filing date: 2002-04-22
Publication date: 2002-10-31
Also published as: WO2002087186A3

Abstract

Techniques for characterizing channel condition that do not require equalizer training are disclosed. Probe signals having different symbol rates are used in the channel characterization. Achievable SNR at various symbol rates (both transmitted and not transmitted) can be predicted. Achievable SNR is calculated based on the received signal PSD when a probe signal is present and the estimated crosstalk/noise PSD during silence periods. An optimal combination of modulation type and symbol rate based on the channel condition can also be identified.

Description

Channel Characterization and Rate Selection in DSL

Systems

Inventors:

Farrokh Rashid-Farrokhi

Cindy C. Wang

Steven R. Blackwell

RELATED APPLICATIONS [0001] This application claims the benefit of U.S. Provisional Application No. 60/285,904, filed April 23, 2001.

FIELD OF THE INVENTION [0002] The invention relates to telecommunications, and more particularly, to fast optimal rate selection techniques for digital subscriber line systems.

BACKGROUND OF THE INVENTION [0003] The Telecommunications Standards Section of the International Telecommunication Union (sometimes designated as ITU-T) provides recommendations to facilitate the standardization of the telecommunications industry. One of these recommendations is referred to as G.991.2. ITU-T Recommendation G.991.2 describes a digital subscriber line (DSL) standard referred to as G.SHDSL (Symmetric High-bit-rate DSL). Similarly, the standards group TlEl .4 of the American National Standards Institute (ANSI) is developing an industry standard for very-high-bit-rate digital subscriber line (VDSL) technology.

[0004] Such standards support an activation process. In general, the activation process establishes a communication link with required transmission parameters between a physically connected and powered transceiver pair (e.g., central office and customer premises transceivers). The activation process can also modify transmission parameters of the link. Either transceiver of a transceiver pair can initiate the activation process. The activation process may be started, for example, by a power-up request, or after a link interruption or deactivation. A successfully completed activation process makes the link ready for steady-state data communication (referred to as data mode herein). [0005] Prior to activation of the link (before entering data mode), a handshake procedure may be established between the newly coupled transceiver pair. One such handshake procedure is defined in ITU-T recommendation 994.1. Generally, such handshaking procedures allow communicating modems to exchange information regarding their respective capabilities and protocols thereby facilitating an efficient communication session.

[0006] During the activation process, training signals are transmitted. The training signal characteristics are generally specified in the standards. However, the symbol rate and duration of the training signals, as well as the implementation of the receiver algorithm, are not defined in the standards, and are therefore left to vendor discretion. Different vendors use different techniques to investigate the possibility of supporting different symbol rates given the current channel conditions (e.g., interference and loop response). Typically, such techniques employ receiver algorithms that rely on training an equalizer included in the receiver, and measuring the signal-to-noise ratio (SNR) at the equalizer output. In such a preactivation mode, the symbol rate would be selected so that a desired bit error rate (e.g., 10^"7) can be maintained during the data mode (once the activation procedure is completed). However, such techniques are associated with a number of problems. [0007] For example, the process of equalizer training is time consuming. For example, in a low symbol rate loop, equalizer training may require up to 10 seconds. Moreover, since the equalizer convergence time may be longer than the maximum training signal length (e.g., G.991.2 specifies a maximum training signal length of 3 seconds), such techniques may not provide an accurate estimation of the achievable SNR in the activation mode. In addition, equalizer training has to be repeated for many training signals, where each training signal has a different symbol rate. This is because the performance of the equalizer at disparate symbol rates may not be highly correlated due to factors such as variations in the channel response and interference characteristics. Also, such techniques only characterize the existing noise profile in the system at the time equalizer training is performed. Any subsequent changes in the noise profile may therefore compromise the integrity of the communication channel. In such a case, it may be necessary to run the equalizer training process again. Otherwise, the communication channel may run inefficiently and data may be lost.

[0008] What is needed, therefore, are techniques for characterizing channel condition that do not require equalizer training.

SUMMARY OF THE INVENTION [0009] Techniques for characterizing channel condition that do not require equalizer training are disclosed. Probe signals having different symbol rates are used in the channel characterization. An optimal combination of modulation type and symbol rate based on the channel condition can also be identified. The features and advantages described in the specification are not all-inclusive and, in particular, many additional features and advantages will be apparent to one of ordinary skill in the art in view of the drawings, specification, and claims. Moreover, it should be noted that the language used in the specification has been principally selected for readability and instructional purposes, and not to limit the scope of the inventive subject matter.

BRIEF DESCRIPTION OF THE DRAWINGS [0010] Figure 1 is a block diagram illustrating functional layers of a generic DSL system.

[0011] Figures 2a and 2b illustrate block diagrams of preactivation transmitter and receiver reference models, respectively, in accordance with embodiments of the present invention.

[0012] Figure 2c is a block diagram of a receiver processing module in accordance with one embodiment of the present invention.

[0013] Figures 2d and 2e are block diagrams of a received signal PSD estimation module in accordance with embodiments of the present invention. [0014] Figure 3a illustrates a typical timing diagram for a preactivation sequence. [0015] Figure 3b is a table illustrating an example timing of the signals that make up the preactivation sequence of Figure 3 a.

[0016] Figure 3c is a table illustrating a set of scrambler polynomials. [0017] Figure 4 is a flow chart illustrating a method for estimating equalizer performance in accordance with one embodiment of the present invention. [0018] Figure 5 illustrates a method for identifying an optimal combination of modulation and symbol rate based on the channel condition in accordance with one embodiment of the invention.

[0019] Figures 6a and 6b are tables illustrating the required SNR for different modulation levels to maintain a particular bit error ratio.

[0020] Figure 7a is a table illustrating the estimated clock cycles associated with carrying out one embodiment of the present invention.

[0021] Figure 7b is a table illustrating the probe signal length for minimum and maximum symbol rates associated with one embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION [0022] Embodiments of the present invention provide a preactivation mode mechanism for predicting the achievable SNR of a communication link for different possible symbol rates. As such, conventional equalizer training is not required. Rather, the disclosed techniques allow equalizer performance to be estimated based on the predicted achievable

SNR. In general, the link transmitter transmits a number of short duration probing

(training) signal sequences to the link receiver. The rate and timing parameters of the probing signal sequence can be established by the receiver, and provided to the transmitter as part of an initial handshaking procedure. Each transmitted probe signal sequence is associated with a different rate, and data characterizing the link is detected at the receiver during active periods of each probe signal sequence, as well as during silent periods (no probe signal present). The achievable SNR can then be calculated for each probe signal sequence. In addition, a technique that identifies an optimal combination of modulation and symbol rate based on the channel condition is disclosed.

[0023] For purposes of clarity and facilitating understanding, the functional layers and modes of operation of a DSL system are generally discussed herein. A preactivation transmitter-receiver pair that operates in accordance with embodiments of the present invention is then discussed. The discussion further includes techniques for predicting achievable SNR for the channel coupling the transmitter-receiver pair, and for identifying an optimal combination of modulation and symbol rate based on the channel condition.

Functional Layers of a DSL System [0024] Figure 1 is a block diagram illustrating functional layers of a DSL system. In particular, the diagram illustrates the functional blocks and interfaces of generic DSL transceivers that are coupled via a transmission medium. Each transceiver has an application invariant section and an application specific section. The application invariant section includes a physical media dependent (PMD) layer 120 and a physical media-specific transmission convergence (PMS-TC) layer 115, while the application specific section includes a TPS-TC layer 110 and a number of interface (I/F) blocks 105. The functions at the customer side (e.g., 105a, 110a, 115a, and 120a) are carried out by the customer transceiver (sometimes referred to as network termination or xTU-R), while functions at the central office side (e.g., 105b, 110b, 115b, and 120b) are carried out by the central office transceiver (sometimes referred to as line termination or xTU-C).

[0025] The system may be, for example, SHDSL-based or VDSL-based. However, the principles of the present invention can be implemented in other DSL systems having similar functionality (e.g., baseband modulation or single carrier-based communication systems such as second-generation high-bit-rate DSL) as will be apparent in light of this disclosure.

In addition, note that the system illustrated may be configured with additional components, such as one or more signal regenerators coupled between the transceivers along the transmission medium to ensure a strong transmission signal over long distances. The number and location of such regenerators depends on factors such as insertion loss and transmission characteristics of the transmission medium. In addition, system features such as management functions (typically controlled by an operator's network management system), and remote power feeding are not illustrated in Figure 1 , but may be added as necessary.

[0026] The transceivers, transition medium (sometimes referred to as a digital local line), and any regenerators units make up a DSL span. The transition medium may be, for example, a single copper twisted pair (also referred to as a loop). In alternative configurations (e.g., such as a two pair SHDSL-based system), the transition medium may be two copper twisted pairs. In such an embodiment, each transceiver has two separate

PMD layers 120 interfacing to a common PMS-TC layer 115.

Layer Functions and Interfaces [0027] The principal functions of the PMD layer 120 include link startup and line equalization, as well as symbol timing generation and recovery, coding and decoding, modulation and demodulation, and echo cancellation. Embodiments of the present invention discussed herein generally operate in the PMD layer 120. However, this is not intended to be a limitation on the present invention, which can operate in any layer generally coupled to the communication medium or otherwise equivalent to a physical media dependent layer. The PMS-TC layer 115 functions include framing and frame synchronization functions, as well as scrambler and descrambler functions. The PMS-TC layer 115 is connected across a logical interface to the TPS-TC layer 110. The TPS-TC layer 110 is application specific and generally operates to package user data within the transmission frame. Functionality here may include multiplexing, demultiplexing, and timing alignment of multiple user data channels. The TPS-TC layer 110 communicates with one or more interface blocks 105 across a logical interface thereby providing support for channels of user data. Note that the logical interfaces are not intended to imply a physical implementation.

Preactivation Mode of PMD Layer [0028] In general, the PMD layer 120 is associated with various modes of operation, such as a data mode, an activation mode, and a preactivation mode. The data mode operates after activation procedures have been completed, and allows payload to be communicated between the communicatively coupled transceivers. The activation mode operates before the data mode is entered, and generally establishes a communication link with required transmission parameters between the physically connected and powered transceivers. The activation mode can also be used to modify transmission parameters of the communication link. The preactivation mode operates before the activation mode is entered, and generally includes one or more handshake sessions and line probing ("training") sequences.

Handshake sessions provide a mechanism for exchanging capabilities and negotiating the operational parameters for each transceiver connection. Line probe sequences provide a mechanism to identify or otherwise derive characteristics of the transmission medium, such as achievable SNR.

General Overview of Preactivation System [0029] Figures 2a and 2b illustrate block diagrams of preactivation transmitter and receiver reference models, respectively, in accordance with embodiments of the present invention. The preactivation transmitter can be coupled to the preactivation receiver via a communication medium (line) thereby forming a transmitter-receiver pair for a given communication direction (e.g., upstream or downstream). A corresponding receiver- transmitter pair coupled by the transmission medium can provide the opposite communication direction. Thus, a transceiver-transceiver pair of a DSL span can be provided. The transmission medium can be, for example, one or more copper twisted pairs, or optical fibers.

Preactivation Transmitter [0030] Handshake module 205 operates in accordance with an established handshaking procedure (e.g., G.994.1) that effectively initiates and terminates a preactivation sequence.

After the initial handshake procedure is performed, the line probe function of the transmitter is effectively switched in, where scrambler module 210 scrambles the input bit stream making up the probe signals. Mapper module 215 performs bit-to-level mapping, where the scrambled bit stream is mapped into a stream of constellations in accordance with the modulation scheme employed (e.g., pulse amplitude modulation level N). Spectral shaper module 220 filters the transmit signal thereby shaping its spectrum, and upsample module

225 increases (sometimes referred to as interpolation) the transmitter sampling rate to the sampling rate of the digital-to-analog converter (D/A) 230. D/A 230 converts the transmit signal to its analog equivalent, and line driver (LD) 235 drives the analog equivalent onto the line via line coupling 240. In one embodiment, line coupling 240 is a DSL coupling transformer that couples the transmitter circuitry to the line.

Transmit Scrambler Polynomial and Probe Signal Symbol Rate [0031] During the training sequences, the transmit scrambler module 210 operates pursuant to a scrambler polynomial. This scrambler polynomial can be selected by the receiver during a handshake session from a set of allowed scrambler polynomials illustrated in Figure 3c. For example, the polynomial associated with index 000 for the downstream direction might be selected. Such a table or set can be stored, for example, in a non-volatile memory (e.g., EEPROM or flash memory) included in each of the communicating transceivers that house the corresponding transmitter-receiver pair. Thus, each transceiver is aware of the time sequence of the scrambled bit stream. Generally, the transmitter supports all the polynomials listed in the table. Allowing the receiver to select the polynomial generally simplifies the receiver structure and algorithm. Regardless, the scrambler module 210 operates pursuant to a particular scrambler polynomial thereby generating a random sequence which can be used for a probe signal. The transmit scrambler module 210 can be initialized (e.g., to all zeros) as a preliminary step in the handshake session.

[0032] The symbol rate of a particular probe signal generated pursuant to the selected polynomial can be manipulated by the transmitter. For example, the symbol rate of the generated probe signals can be increased or decreased by respectively increasing or decreasing the system clock. Alternatively, the symbol rate of the generated probe signals can be increased by increasing the upsample ratio of upsample module 225. A combination of varying the system clock frequency and increased upsample ratio can also be used to vary the symbol rate of the probe signals.

Preactivation Receiver [0033] The transmitted analog signal is decoupled from line by decoupling 245, which may be, for example, a decoupling transformer. Note that line decoupling 245 may further include hybrid circuitry for performing two-to-four wire conversion. The received analog signal is then amplified by a programmable gain amplifier (PGA) 250 to facilitate processing of the received signal. The amplified signal is then converted back to its digital equivalent by analog-to-digital converter (A/D) 255. Typically, A/D 255 has a high sampling rate. The downsample module 260 reduces the sampling rate (sometimes referred to as decimation) of the digital equivalent signal to the receiver sampling rate (e.g., eight times the transmit symbol rate). The resulting digital signal, x(n), is then subjected to the receiver algorithm (e.g., probe signal analysis and rate selection), which operates in receiver processing module 265.

Preactivation Sequence [0034] Figure 3a illustrates a typical timing diagram for a preactivation sequence that is transmitted by the preactivation transmitter. Both the upstream (e.g., remote probe signals

P_rl, P_r2 and P_r3) and downstream (e.g., central probe signals V_c\ and P_c2) sequences are shown. Note that the time index m of Figure 2a represents the symbol time, and time index t of Figure 3a represents analog time. Figure 3b is a table illustrating an example timing of the signals that make up a preactivation sequence. Note that the specified tolerances given in Figure 3b are relative to the nominal or ideal value, and are not cumulative across the preactivation sequence. For any one communication direction, the transmitted preactivation sequence, including a number of probe signals, is transmitted by the preactivation transmitter and then received by the preactivation receiver. The preactivation receiver algorithm of the receiver processing module 265 can then analyze the probe signals of the received preactivation sequence, and estimate equalizer performance of the receiver. No equalizer training is necessary. The equalizer performance data (e.g., achievable SNRs, channel response magnitudes, and other information relevant to characterizing the channel) can be provided to the equalizer.

[0035] In the embodiment illustrated, the scrambler module 210 input, d(m), is an initialization signal that is represented by logical ones for all m. Assume that the probe signal modulation format employed by the mapper module 215 is uncoded 2-PAM.

However, modulation schemes other than uncoded 2-PAM can be employed to modulate the probe signals. For example, QAM (quadrature amplitude modulation) and CAP (carrierless amplitude phase) modulation techniques can also be used to modulate the training signals.

Also, other types of PAM can be used as well. In addition, probe signal parameters such as the symbol rate, spectral shape, duration, power backoff and other relevant probe signal characteristics are specified during the initial handshake session (e.g., G.994.1). Probing session results are exchanged by a second or terminating handshaking session (e.g.,

G.994.1). Note that the bit time is equivalent to the symbol time is uncoded 2-PAM is used to modulate the training signals.

Receiver Processing Module [0036] Figure 2c is a block diagram of a receiver processing module in accordance with one embodiment of the present invention. The receiver processing module 265 includes a received signal PSD estimation module 270, a crosstalk/noise PSD estimation module 275, an achievable SNR calculation module 285, a channel response calculation module 290, and an optimizer module 295. Probe signals at various symbol rates are received at the input at the PSD estimation section for processing. The achievable SNR at a number of symbol rates (some actually transmitted, others not transmitted) is provided at an output of module 265. Also provided at an output is the overall optimal PAM level and symbol rate based necessary to achieve the maximum data rate for the given communication channel. [0037] The received signal PSD estimation module 270 is adapted to estimate the received signal PSD when a probe signal having a symbol rate is present, the probe signal having a symbol rate. The crosstalk/noise PSD estimation module 275 is adapted to estimate the crosstalk/noise PSD during silence periods. The achievable SNR calculation module 285 is adapted to calculate achievable SNR at the probe signal symbol rate based on the received signal PSD and the crosstalk/noise PSD. The achievable SNR calculation module is further adapted to calculate achievable SNR for symbol rates lower than the probe signal symbol rate (or higher than probe signal symbol rate if appropriate). In this embodiment, the achievable SNR calculation module 285 operates pursuant to a Decision Feedback Equalizer (DFE) bound.

[0038] The channel response calculation module 290 is adapted to calculate channel response magnitude at a frequency range up to the probe signal symbol rate based on the received signal PSD and known transmit signal PSD associated with the probe signal. The achievable SNR for symbol rates lower than the probe signal symbol rate can be calculated based on received signal PSD at such a lower symbol rate, which is calculated based on the channel response magnitude and known transmit signal PSD associated with the lower symbol rate. In some cases (e.g., for symbol rates under 500KHz), the achievable SNR for symbol rates higher than the probe signal symbol rate can be calculated based on received signal PSD at such a higher symbol rate, which is calculated based on the channel response magnitude and known transmit signal PSD associated with the higher symbol rate. The optimizer module is adapted to calculate the optimal PAM levels for a number of symbol rates based on achievable SNR for each symbol rate, and to calculate the optimal symbol rate based on the optimal PAM levels and their associated symbol rates. Thus, the overall optimal PAM level and symbol rate are defined.

[0039] Figures 2d and 2e are block diagrams of a received signal PSD estimation module in accordance with embodiments of the present invention. In Figure 2d, the received signal PSD estimation module 270 includes blocking module 271, windowing module 272, an fast Fourier transform (FFT) module 272, and an averaging module 274. A probe signal is received by the blocking module 271 for processing. The received signal PSD associated with the probe signal is provided at an output of the module 270. [0040] The blocking module 271 is adapted to divide the probe signal into K blocks, each block having M samples. The windowing module 272 is adapted to window (e.g., Bartlett window) the blocks of the probe signal thereby providing a windowed data stream. The FFT module 272 is adapted to receive each block of the windowed data stream thereby providing an FFT of each input block. The averaging module 274, which operates in the frequency domain, is adapted to receive and average the FFTs of the blocks. Thus, an estimate of the received signal PSD associated with the probe signal is provided. [0041] In the alternative embodiment illustrated in Figure 2e, the received signal PSD module 270 includes an averaging module 276 and an FFT module 277. In this embodiment, the received signal PSD estimation module is adapted to receive a probe signal K times. The averaging module 276, which operates in the time domain, is adapted to average the K received probe signals thereby providing an average received probe signal. The FFT module 277 is adapted to receive the average received probe signal. Thus, an estimate of the received signal PSD associated with the averaged probe signal is provided. The receiver processing module 265 and its sub-modules will be discussed in more detail with reference to Figures 4 and 5.

[0042] Note that alternative configurations for the preactivation transmitter and receiver will be apparent in light of this disclosure, and the illustrated configurations are not intended to limit the present invention. For example, components or modules included in the preactivation transmitter or the preactivation receiver can be integrated with other components or modules (e.g., D/A 230 and LD 235 can be included in one component, just as PGA 250 and A/D 255 can be included in one component). Likewise, some illustrated components may not be included in other embodiments (e.g., upsample module 225 may not be required depending on the sampling rates of the transmitter and D/A 230). Components or modules not illustrated, such as analog filters and sample buffers, may also be included in alternative embodiments. In addition, numerous conventional configurations and componentry can be employed to realize line coupling 240, line driver 235, and D/A 230, as well as line decoupling 245, PGA 250, and A/D 255. Likewise, items such as upsample module 225, spectral shaper module 220, mapper module 215 and scrambler module 210, as well as downsample module 260, and receiver processing module 265 can be implemented, for example, as a set of software instructions or routine executing on a digital signal processor (DSP) or other suitable processing environment, or by special purpose silicon (e.g., application specific integrated circuit chip or chip set). [0043] Figure 4 is a flow chart illustrating a method for estimating equalizer performance in accordance with one embodiment of the present invention. Generally, the method operates in a communication system having a transceiver pair communicatively coupled by a transmission medium. Thus, the system defines a first transmitter-receiver pair for a downstream communication channel, and a second transmitter-receiver pair for an upstream communication channel. Note that the method may be employed in either direction, or both directions. Embodiments of this method can be carried out, for example, by the system (or portions thereof) illustrated by Figures 2a and 2b. Alternative configurations and processing environments will be apparent in light of this disclosure, and the present invention is not intended to be limited to any one particular embodiment or configuration.

[0044] The method begins with exchanging 405 information about a forthcoming training signal session. The training signal session includes a number of probe signals (e.g., Figure 3a). Probe signals are generally a sequence of data bits in a particular pattern. In one embodiment, the exchanging of information is performed in accordance with a G.994.1 handshaking procedure (or other suitable handshaking procedure). The exchanged information includes, for example, the transmit power of the probe signals, the transmit scrambler polynomial, the number of probe signals included in the session, the symbol rate of each probe signal, and other relevant probe signal information. In addition, the particular capabilities of the communicating transceivers can be exchanged. Thus, upon completion of the information exchange, each transceiver is informed of the parameters of a forthcoming training signal session.

[0045] The method further includes transmitting 410 a training signal sequence including one or more probe signals. Note that either the receiving node or the transmitting node can initiate the sequence. In one embodiment, the node that initiates the training signal sequence is established during the exchanging 405 of information (e.g., initial handshaking procedure). The method further includes estimating 415 the received signal power spectral density (PSD) when a probe signal is present, and estimating 420 the crosstalk/noise PSD during silence periods. Note that estimating the received signal PSD can be repeated for a number of different probe signals, each probe signal associated with a different symbol rate. Likewise, estimating the crosstalk/noise PSD can also be repeated under varied conditions, such as silence periods of various durations. Note that the received signal PSD includes the transmitted probe signal sequence plus crosstalk and noise. Further note that different noise profiles may be used to predict the system performance under different crosstalk scenarios.

[0046] In one embodiment, the Welch method is used in estimating 415 the received signal PSD. More specifically, a probe signal having a particular symbol rate is received (e.g., at receiver processing module 265), and is divided into K blocks. Each block has M samples. In one embodiment, the probe signal is divided so that each block has 512 samples (e.g., M=512). A Bartlett window w(n) as defined in Equation 1 can be employed to window the received data stream of the probe signal. The windowed data stream is applied to a fast Fourier transform (FFT) as defined in Equation 2. The FFT may be included, for example, in the receiver processing module 265. The resulting output of the FFT is then averaged according to Equation 3 to estimate the received signal PSD. Equations 1, 2, and 3 mathematically demonstrate this embodiment for estimating the received signal PSD:

(Equation 1)

e where N is the width of the window.

[0047] The k* element of the FFT of the i^th block is defined by:

k = 0,- --,M (Equation 2)

1 M-\ where U is a normalizing factor defined by: U = — xv² (n) . Thus, the FFT of the probe signal is evaluated at M points, where k is the index to one of the points. [0048] The received signal PSD, S_rcv , can be estimated 415 by Equation 3 :

S_rev(^) = ^∑J^ ) (Equation s)

The received signal PSD, S_rcy, is estimated during transmission of the corresponding probe signal.

[0049] Alternatively, the estimating 415 of the received signal PSD can be carried out by averaging the received signal over a number of probe signal sequences as shown in

Equations 4 and 5. For example, assuming a probe signal symbol rate of 200 Kbaud, where each probe signal sequence has an arbitrary length of 511 symbols, one hundred such probe signal sequences would be received in 255.5 milliseconds [((511 symbols / probe signals) * (1 second / 200Ksymbols) * (100 probe signals)) = 255.5 milliseconds]. Each of these received probe sequences can be averaged. The FFT of the resulting averaged probe signal can then be taken. One advantage of this alternate approach is the reduction in noise power due to averaging.

(Equation

4)

[0050] The averaged probe signal, s_av {n) may be calculated by:

1 ^κ~i *«V(O = — ∑*^ω(^Λ) (Equation

Λ- 1=0

5) where J ^(,) («) is the i^m transmission of a particular probe signal sequence, n is the time index, and K is the total number of probe signal sequences used for the received signal PSD estimation. In one embodiment, K is equal to 100 to provide a robust average. Thus, each probe signal that is used in the training sequence (e.g., four probe signals each having a different symbol rate) is transmitted K times.

[0051] The crosstalk/noise PSD, N , can be estimated 420 during the silent periods (e.g., when no probe signal is being transmitted) by Equation 6:

M K ^ M

6) where J and U are defined as before. In one embodiment, the estimating functions of 615 and 620 are programmed or otherwise configured in the receiver processing module 265 of the receiver illustrated in Figure 2b.

[0052] The method proceeds with calculating 425 the achievable SΝR for the symbol rate of a probe signal. In one embodiment, the achievable SΝR for the symbol rate can be predicted from a DFE bound as demonstrated by Equation 7:

(Equation 7)

2nk where f_sym denotes the probe signal symbol rate, and f_k = —^- radian second. Note that the ^~M caret symbol ^A denotes an estimated quantity. In one embodiment, the DFE bound function is programmed or otherwise configured in the receiver processing module 265 of the receiver illustrated in Figure 2b.

[0053] The method further includes calculating 430 the channel response at frequency/

(denoted by H{f) ). Note that frequency/denotes a frequency range up to the symbol rate, f_sym . In one embodiment, the channel response H{f) is calculated by the receiver processing module 265 from the received signal PSD, S_rcv (e.g., as estimated in 415) and the known transmit signal PSD, S_Tx (e.g., made known to the receiver via a previous handshake session) as demonstrated by Equation 8:

(Equation

8)

[0054] Once the channel response is calculated, the method proceeds with calculating

435 the achievable SNR for symbol rates other than the probe signal symbol rate (e.g., symbol rates below and possibly above the probe signal symbol rate). In one embodiment, the achievable SNR for different symbols rates can be predicted as demonstrated by

Equation 9:

(Equation 9) [0055] The calculating functions of 430 and 435 can be, for example, be programmed or otherwise configured in the receiver processing module 265 of the receiver illustrated in

Figure 2b. The achievable SΝR for different symbol rates as well as other estimated or otherwise derived information can then be exchanged between the communicating transceivers during a subsequent handshake procedure (e.g., G.994.1) so that both transceivers will have the information.

[0056] Note that the estimated channel response for a transmitted probe signal symbol rate is sufficiently accurate for symbol rates other than the transmitted probe signal symbol rate. Thus, probe signals at lower symbol rates do not have to be transmitted. Likewise, probe signals at higher symbol rates (depending on the spectrum of the actually transmitted probe signal) may not need to be transmitted. Rather, probe signals having different symbol rates that effectively divide the overall symbol rate range of the channel into a number of sub-ranges can be transmitted. Each transmitted probe signal is used to estimate the channel response associated with a particular sub-range. Using the channel response magnitude associated with a particular sub-range of symbol rates, and the known transmit signal PSD {S_Ά ) of probe signals (not actually transmitted) having symbol rates in that sub-range, the received signal PSD at those other symbol rates can be calculated (e.g., Equation 8). The achievable SNR for all those probe signal symbol rates can then be predicted (e.g., Equation

9).

Symbol Rate Sub-Range Scheme [0057] The symbol rate for any one channel may vary over a wide range (e.g., from

15KHz to 1.5MHz). In order to provide a good estimate for the channel response in such a wide bandwidth, several probe signals can be employed. For example, assume that the symbol rate range of a channel is approximately 18KHz to 770KHz. Further assume that four probe signals are actually transmitted, where the symbol rate of probe signal 1 is

66KHz, the symbol rate of probe signal 2 is 300KHz, the symbol rate of probe signal 3 is

535KHz, and the symbol rate of probe signal 4 is 770KHz. In this example, symbol rate sub-range 1 (e.g., 15KHz or lower to 132KHz) is associated with probe signal 1, symbol rate sub-range 2 (e.g., 133KHz to 300KHz) is associated with probe signal 2, symbol rate sub-range 3 (e.g., 301KHz to 535KHz) is associated with probe signal 3, and symbol rate sub-range 4 (e.g., 536KHz to 770KHz) is associated with probe signal 4. Note that symbol rate sub-range 1 includes symbol rates both lower and higher than the symbol rate of probe signal 1. At higher symbol rates (e.g., greater than 500KHz) where attenuation becomes more pronounced, however, it may not be possible to accurately represent symbol rates that are higher than the actually transmitted probe signal. Numerous such sub-range schemes will be apparent in light of this disclosure. Generally, low symbol rate probe signals provide an accurate estimate of the channel response at low frequency. Likewise, high symbol rate probe signals that have wider bandwidth but lower PSD provide a more accurate estimate of the channel response at high frequency.

Multi-level Modulation Optimizer [0058] The G.SHDSL standard defined in Recommendation G.991.2 supports

16TCPAM (16 Trellis Coded Pulse Amplitude Modulation). A system in accordance with an embodiment of the present invention, on the other hand, is able to support up to 1024

PAM and TCP AM. An optimal PAM level can be used to maximize the data rate. The following discussion demonstrates a technique that finds an optimum PAM level based on the condition of the system's communication channel. For purposes of this discussion, note that the required SNR for the iPAM or iTCPAM is denoted by SNR_req>i. The SNR for different PAM levels required to maintain a bit error ratio (BER) of 10^-7 are shown in Figure 6a. Note that the BER of a system depends on various factors such as allowable communication error and the desired efficiency of the communication channel, and thus can vary from system to system. A margin can be added to the required SNR if necessary. For instance, 6.6dB margin is added to the SNR for non-bridge-tap lines, and 7.1dB margin is added to the required SNR for bridge-tap lines. Note for TCP AM based systems, there is a coding gain (e.g., 5.1dB). As such, the required SNR_reπ i for an iTCPAM based system is given by Equation 10:

{SNR,. + 6.6dB - CodingGain non - bridge - tap (Equation 10)

SNR, + 7 ΛdB - CodingGain Bridge -tap

[0059] The required SΝR_ren i for an iPAM based system is given by Equation 11 : f SNR, + 6.6dB non - bridge - tap SNR_{reπ i} = \ ' (Equation 11) ^req'¹ [SNR, + 7. MB Bridge -tap

[0060] Other DSL-based systems configured in accordance with embodiments of the present invention may support other multi-level modulation schemes. For example, a single carrier NDSL system configured in accordance with the present invention may support up to

256 QAM. The SΝR for QAM levels required to maintain a BER of 10^~7 are shown in

Figure 6b. The SΝRs specified in Figure 6b also work for CAP levels (same levels as

QAM) required to maintain a BER of 10"⁷. Equations 10 and 11, as well as their related discussions, equally apply in the context of QAM and CAP modulation schemes. Note that margin added to ensure a robust communication, as well as coding gain, will generally vary from system to system depending on factors such as the desired system performance, available margin, and the particular modulation scheme employed. Further note that embodiments of the present invention can be implemented in the context of non-standard modulation levels, such as 1024 QAM (or higher) in a VDSL system.

Optimizer Algorithm [0061] Figure 5 illustrates a method for identifying an optimal combination of modulation (e.g., iPAM, iTCPAM, iQAM, or iCAP) and symbol rate based on the channel condition in accordance with one embodiment of the invention. This method may be, for example, be programmed or otherwise configured to operate in the receiver processor module 265 of Figure 2b. The method begins with estimating 505 the achievable SNR ( SNR_S ) for a number of symbol rates included in the range of symbol rates for the corresponding channel. In one embodiment, estimating 505 is performed using Equation 7, and the symbol rates are selected based on a sub-range scheme as previously described. In an alternative embodiment, SNR_S can be estimated for all available symbol rates. [0062] The method proceeds with calculating 510 the optimal modulation level (e.g., PAM_S or QAM_S or CAP_S) for each symbol rate based on the corresponding achievable SΝR.

In one embodiment, PAM_S is calculated using Equation 12:

PAM, = arg max SNR_req < SNR_S (Equation 12)

PAM_S is the largest PAM level that requires an SΝR {SNR_reqJ) that is less than the achievable SΝR for each symbol rate ( SNR_S ). Note that the subscript s denotes the corresponding symbol rate, and i denotes a particular PAM level. Further note that Equation 12 and its related discussion equally applies to QAM_S and CAP_S, where i denotes a particular QAM/CAP level (as opposed to a PAM level).

[0063] The method proceeds with calculating 515 the optimal symbol rate, SR, based on the optimal modulation levels calculated in 510 and their associated symbol rates, s. In one embodiment, the optimal symbol rate SR is calculated by Equation 13:

SR = arg max( 2^PAM' s) (Equation 13)

Thus, an overall optimal PAM level and symbol rate are defined. The optimal PAM level

PAM_S corresponding to the optimal symbol rate SR can be designated as PAM_SR for purposes of distinguishing it from the other PAM_S levels. The maximum data rate, DR, can then be calculated by Equation 14:

DR = log₂ PAM_SRf_SR (Equation 14) where f_$R is the optimal symbol rate, SR. The optimal PAM level and symbol rate identified can then be exchanged between the communicating transceivers during a subsequent handshake session (e.g., G.994.1) so that both transceivers will have the information. Again, note that Equations 13 and 14 and their related discussions equally applies to CAP or QAM, where the optimal symbol rate SR can be designated as CAP_SR or

QAM_SR (as opposed to PAM_SR).

Implementation Example [0064] The following demonstrates an example of one embodiment of the present invention. Numerous implementations employing the techniques described herein will be apparent in light of this disclosure, and the present invention is not intended to be limited to any one implementation. For purposes of this example, assume the following parameters:

Samples per block: 512 samples

Receiver Sampling Rate: s_{K rmx} = 3.1 Msample/s

Number of processing cycles per sample: N_R = f_M /S_R = 250 MHz / 3.1 Msample/s = 80

Number of cycles per 512 samples: 40,000

[0065] This example demonstrates how the number of clock cycles for each 512 sample block are calculated, and how the number of cycles for each operation are calculated. The results are summarized in Figure 7a. A large number of samples (e.g., 10,000 samples) can be used for PSD estimation (for both the crosstalk/noise PSD and the received signal PSD).

The probe signal length for minimum and maximum symbol rate is shown in Figure 7b.

[0066] The complexity of each block is as follows:

FFT (for each block): [0067] The estimated number of instructions for each block of 512 samples is 30,000 cycles.

PSD (for each block): [0068] Employing the Welch method, 512 multiplications are performed for windowing, and 512 multiplications are performed to calculate the magnitude square of the

FFT. The scaling can be postponed until after the average magnitude of the FFT is calculated for the entire probe signal (all K blocks). As illustrated by Figure 7a, the number of cycles per 512 sample block in this example is 1024.

SNR Estimation (for each probe signal): [0069] The SNR estimation is done once for each block and for each symbol rate.

There are 512x4 divisions and 512 log operations. Each log operation takes 1200 cycles and each division takes 16 cycles. As such, the total number of cycles is 647 Kcycles.

Thus, each SNR estimation for each probe signal takes about 2.6ms given a 250 MHz clock signal.

[0070] The foregoing description of the embodiments of the invention has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the invention to the precise form disclosed. Many modifications and variations are possible in light of this disclosure. It is intended that the scope of the invention be limited not by this detailed description, but rather by the claims appended hereto.

Claims

What is claimed is: 1. A method for estimating equalizer performance in a DSL communication system, the method comprising: estimating received signal PSD when a probe signal is present, the probe signal having a symbol rate; estimating crosstalk/noise PSD during silence periods; and calculating achievable SNR at the probe signal symbol rate based on the received signal PSD and the crosstalk/noise PSD.

2. The method of claim 1 further comprising the preliminary steps of: exchanging information about a forthcoming training signal session; and transmitting a training signal sequence including one or more probe signals.

3. The method of claim 2 wherein exchanging of information is performed in accordance with a handshaking procedure.

4. The method of claim 1 further comprising: calculating channel response magnitude at a frequency range up to the probe signal symbol rate based on the received signal PSD and known transmit signal PSD associated with the probe signal; and calculating achievable SNR for symbol rates other than the probe signal symbol rate.

5. The method of claim 4 wherein calculating achievable SNR for symbol rates other than the probe signal symbol rate includes: calculating received signal PSD at a symbol rate lower than the probe signal symbol rate based on the channel response magnitude and known transmit signal PSD associated with the lower symbol rate; and calculating the achievable SNR at the lower symbol rate based on the received signal PSD at the lower symbol rate.

6. The method of claim 4 wherein calculating achievable SNR for symbol rates other than the probe signal symbol rate includes: calculating received signal PSD at a symbol rate higher than the probe signal symbol rate based on the channel response magnitude and known transmit signal PSD associated with the higher symbol rate; and calculating the achievable SNR at the higher symbol rate based on the received signal PSD at the higher symbol rate.

7. The method of claim 1 wherein the method is repeated for a number of probe signals having different symbol rates.

8. The method of claim 7 wherein each probe signal represents a sub-range of symbol rates included in an overall symbol rate range, and the probe signal symbol rate is included in the represented sub-range.

9. The method of claim 1 wherein the method is implemented by a receiver processing module.

10. The method of claim 1 wherein estimating the received signal PSD includes: dividing the probe signal into K blocks, each block having M samples; windowing the probe signal thereby providing a windowed data stream; applying each block of the windowed data stream to a fast Fourier transform thereby providing an FFT of each block; and averaging the FFTs of the blocks to estimate the received signal PSD.

11. The method of claim 1 wherein estimating the received signal PSD includes: receiving the probe signal K times; averaging the K received probe signals thereby providing an average received probe signal; and applying the average received probe signal to a fast Fourier transform thereby providing an estimate of the received signal PSD.

12. The method of claim 1 wherein calculating the achievable SNR at the probe signal symbol is performed using a Decision Feedback Equalizer bound.

13. The method of claim 1 further comprising: exchanging information including the achievable SNR at the probe signal symbol rate in a subsequent handshaking procedure.

14. A method for identifying an optimal combination of modulation and symbol rate based on the channel condition, the method comprising: estimating the achievable SNR for a number of symbol rates included in the range of symbol rates for the corresponding channel; calculating the optimal modulation level for each symbol rate based on the corresponding achievable SNR; and calculating the optimal symbol rate based on the optimal modulation levels and their associated symbol rates thereby defining an overall optimal modulation level and symbol rate.

15. The method of claim 14 wherein the method is implemented by a receiver processing module.

16. The method of claim 14 wherein the symbol rates at which the achievable SNR is estimated are selected based on a symbol rate sub-range scheme.

17. The method of claim 14 further comprising: calculating the maximum data rate based on the overall optimal modulation level and symbol rate.

18. The method of claim 14 wherein the overall optimal modulation level and symbol rate are exchanged during a subsequent handshake session.

19. The method of claim 14 wherein the modulation is one of pulse amplitude modulation, quadrature amplitude modulation, or camerless amplitude phase modulation.

20. A device for estimating equalizer performance in a DSL transceiver, the device comprising: a received signal PSD estimation module adapted to estimate received signal PSD when a probe signal is present, the probe signal having a symbol rate; a crosstalk/noise PSD estimation module adapted to estimate crosstalk/noise PSD during silence periods; and an achievable SNR calculation module adapted to calculate achievable SNR at the probe signal symbol rate based on the received signal PSD and the crosstalk/noise PSD.

21. The device of claim 20 further comprising: a channel response calculation module adapted to calculate channel response magnitude at a frequency range up to the probe signal symbol rate based on the received signal PSD and known transmit signal PSD associated with the probe signal.

22. The device of claim 20 wherein the achievable SNR calculation module is further adapted to calculate achievable SNR for symbol rates other than the probe signal symbol rate.

23. The device of claim 22 wherein the achievable SNR for symbol rates other than the probe signal symbol rate is calculated based on received signal PSD at a lower symbol rate, which is calculated based on the channel response magnitude and known transmit signal PSD associated with the lower symbol rate.

24. The device of claim 20 wherein the device receives a number of probe signals having different symbol rates.

25. The device of claim 24 wherein each probe signal represents a sub-range of symbol rates included in an overall symbol rate range, and the probe signal symbol rate is included in the represented sub-range.

26. The device of claim 20 wherein the device is implemented by one or more application specific integrated circuits, or software instructions executing on one or more processors.

27. The device of claim 20 wherein the received signal PSD estimation module further includes: a blocking module adapted to divide the probe signal into K blocks, each block having M samples; a windowing module adapted to window the blocks of the probe signal thereby providing a windowed data stream; a fast Fourier transform module adapted to receive each block of the windowed data stream thereby providing an FFT of each block; and an averaging module adapted to receive and average the FFTs of each block thereby providing an estimate the received signal PSD.

28. The device of claim 20 wherein the received signal PSD estimation module is adapted to receive the probe signal K times, and further includes: an averaging module adapted to average the K received probe signals thereby providing an average received probe signal; and a fast Fourier transform module adapted to receive the average received probe signal thereby providing an estimate of the received signal PSD.

29. The device of claim 20 wherein the achievable SNR calculation module operates pursuant to a Decision Feedback Equalizer bound.

30. The device of claim 19 further comprising: an optimizer module adapted to calculate optimal modulation levels for a number of symbol rates based on achievable SNR for each symbol rate, and to calculate an optimal symbol rate based on the optimal modulation levels and their associated symbol rates thereby defining an overall optimal modulation level and symbol rate.

31. The device of claim 20 wherein the device is adapted to effect a modulation scheme based on one of pulse amplitude modulation, quadrature amplitude modulation, or camerless amplitude phase modulation.