JP4216813B2 - Gigabit Ethernet transceiver - Google Patents

Description

  The present invention relates to a system and method for reducing noise present in signals received and processed by devices in a communication system, and to a system and method for reducing such noise in a communication system having high throughput. The present invention also relates to a system and method for reducing power dissipation of devices in a communication system, and to a system and method for reducing such power dissipation in a communication system having high throughput. The invention further relates to an activation protocol for initiating normal transmission between transceivers in a high throughput communication system. As used in the context of this disclosure, “high throughput” includes, but is not limited to, 1 gigabit (GB) / sec.

(Cross-reference of related applications)
No. 09 / 037,328 (filed Mar. 9, 1998, entitled “Apparatus and Method for Reducing Noise in a Communication System”, inventor Oscar E. Agazzi), No. 09 / No. 078,466 (filed May 14, 1998, entitled “Start-up Protocol for High Throughput Communication System”, inventors Oscar E. Azazzi and John L. Creigh, 09 / 078,933 (May 14, 1998) Application, titled “Start-up Protocol for High-Throughput Communication Systems”, Inventor Oscar E. Agazzi), and 09 / 143,476 (filed Aug. 28, 1998, entitled “Devices for reducing power dissipation in communication systems” Method ", inventors Oscar E. Agazzi, John L. Creigh, and Mehdi. Hatamian) is a continuation-in-part application.

  A basic communication system is illustrated in FIG. The system includes a hub and a plurality of computers that are serviced by the hub in a local area network (LAN). Although four computers are shown as examples, different numbers of computers may be included in the system. Each computer is typically separated from the hub by a distance on the order of about 100 meters (100 meters). The computers are also remote from each other. The hub is connected to each computer by a communication line. Each communication line includes an unshielded twisted pair wire or cable. Generally, the wire or cable is formed from copper. Four unshielded twisted pair wires are provided on each communication line between each computer and hub. The system shown in FIG. 1 operates with several categories of unshielded twisted pair cables designated as categories 3, 4, 6, and 7 in the communications industry. Category 3 cables are the lowest quality (and lowest cost), and categories 6 and 7 are the highest quality (and highest cost).

  Associated with each communication system is “throughput”. System throughput is the rate at which the system processes data, usually expressed in bits per second. Most communication systems have a throughput of 10 megabits (Mb) / second or 100 Mb / second. In the rapidly developing area of communication system technology, 1 Gb / s full duplex communication is possible over existing Category 5 unshielded twisted pair cables. This system is generally called "Gigabit Ethernet".

  A portion of normal Gigabit Ethernet is shown in FIG. Gigabit Ethernet provides for the transmission of digital signals between one computer and the hub and the reception of that signal at the other computer and the hub. A similar system is provided for each computer. The system includes a gigabit medium independent interface (GMII) that receives byte-wide data at a specified rate, eg, 125 MHz, and scrambles, codes, and various control functions. This data is transmitted to a physical coding sublayer (PCS). The PCS encodes the bits from GMII into a 5-stage pulse amplitude modulation (PAM) signal. The five symbol levels are -2, -1, 0, +1 and +2. Communication between the computer and the hub uses four unshielded twisted pair wires or cables, each operating at 250 Mb / s, and eight transceivers, one at each end of the unshielded twisted pair. Achieved. Full duplex bidirectional operation stipulates the use of a hybrid circuit at the two ends of each unshielded twisted pair. This hybrid controls access to the communication line, thereby enabling full duplex bidirectional operation between the transceivers at each end of the communication line.

  A common problem associated with communication systems utilizing multiple unshielded twisted pairs and multiple transceivers is that crosstalk and echo noise or impairment signals are introduced into the transmitted signal. Noise is inherent in all such communication systems regardless of system throughput. However, the impact of these fault signals is significant on Gigabit Ethernet. Fault signals include echo, near end crosstalk (NEXT), and far end crosstalk (FEXT) signals. As a result of such a fault signal, the performance of the transceiver, particularly the receiver part, is degraded.

  NEXT is a fault signal that results from capacitive and inductive coupling of signals from the near-end transmitter to the receiver input. The NEXT failure signal encountered by the receiver in transceiver A is shown in FIG. Receiver A is trying to detect a direct signal from transmitter E, but in receiver A, crosstalk signals from transmitters B, C and D appear as noise. Since each receiver in the system encounters the same effect, the signal passing through the receiver experiences degradation due to the NEXT impairment signal. In FIG. 3, only the NEXT impairment signal experienced by receiver A is illustrated for clarity.

  Similarly, due to the bidirectional nature of the communication system, echo impairment signals are generated by each transmitter, even in a receiver included in the same transceiver as the transmitter. The echo impairment signal encountered by the receiver in each transceiver is shown in FIG. The receiver is trying to detect the signal from the transmitter at the opposite end of the communication line, but the crosstalk signal from the transmitter appears as noise in the receiver. Since each receiver in the system encounters the same effect, the signal passing through the receiver experiences signal distortion due to the echo impairment signal.

  Far-end crosstalk (FEXT) is a failure resulting from capacitive coupling of signals from the far-end transmitter to the receiver input. The FEXT failure signal encountered by the receiver in transceiver A is shown in FIG. Receiver A is trying to detect a direct signal from transmitter E, but in receiver A, crosstalk signals from transmitters F, G and H appear as noise. Since each receiver in the system encounters the same effect, the signal passing through the receiver experiences signal distortion due to the FEXT impairment signal. In FIG. 5, only the FEXT failure experienced by receiver A is illustrated for clarity.

  As a result of such noise disturbance signals, the performance of the communication system is degraded. The signal carried by the system is distorted and the system experiences a high signal error rate. Accordingly, the art provides a method and apparatus that compensates for communication system performance degradation caused by noise disturbance signals, and a method for reducing this type of noise in high-throughput systems such as Gigabit Ethernet. And there is a need to provide equipment. Embodiments of the present invention meet these needs.

  Four transceivers at one end of the communication line are illustrated in FIG. The transceiver components are shown as overlapping blocks, with each layer corresponding to one transceiver. The GMII, PCS, and hybrid shown in FIG. 6 correspond to the GMII, PCS, and hybrid shown in FIG. 2, and are considered to be independent of the transceiver. The combination of the transceiver and the hybrid forms one “channel” of the communication system. Accordingly, FIG. 6 illustrates four channels that each operate in a similar manner. The transmitter portion of each transceiver includes a pulse shaping filter and a digital / analog (D / A) converter. The receiver portion of each transceiver includes a digital adaptive equalizer system and detector including an analog to digital (A / D) converter, a first in first out (FIFO) buffer, a feed forward equalizer (FFE). The receiver portion also includes a timing recovery system and a near-end noise reduction system that includes a NEXT cancellation system and an echo canceller. NEXT cancellation systems and echo cancellers typically include a large number of adaptive filters.

  Communication line characteristics, such as length, can affect the ability of the NEXT cancellation system and echo canceller to effectively cancel NEXT and echo noise. Normal cable response measurements and simulations indicate that a “long” echo and NEXT canceller is required to provide a sufficient cancellation level for such interference sources. The term “long” is used to describe a canceller having a large number of taps required by the characteristics of the cable. For example, FIG. 7 shows the echo impulse response of a 100 m cable with 85 ohm and 100 ohm terminated characteristic impedance. The rated characteristic impedance is 100 ohms, but the manufacturing standard allows 15% tolerance. This impedance mismatch can result in reflections at the far end of the cable and secondary pulses with a delay of about 1 microsecond. Due to the long delay, approximately 140 taps (125 taps for 1 microsecond delay plus about 15 additional taps to cancel the secondary pulse) are required to cancel this pulse.

  The echo impulse response often has additional reflections at intermediate delay values. In addition, due to the unequal reflection attenuation of the cable, a continuous change in characteristic impedance can occur along the cable, resulting in a number of small reflections at the midpoint. What this intermediate reflection means is that the echo canceller should not be configured to cancel only the initial and terminal reflections, but should be configured to cover the full range of impulse responses. As a result of changes in cable characteristics, changes in cable impulse response also vary. Furthermore, the response of an individual cable may change as a result of its operating environment. For example, the impulse response of a cable may change due to changes in operating temperature. Therefore, it is difficult to precalculate the location where a tap is required and incorporate this location into the echo and NEXT canceller design. FIG. 8 shows the NEXT impulse response of a 100 m cable. As shown, the NEXT response is also long, requiring many taps in the NEXT canceller that make up the NEXT cancellation system. The combination of NEXT and echo canceller consumes most of the DSP operation in Gigabit Ethernet.

  A high degree of power dissipation occurs because of the large number of taps required to achieve satisfactory performance in such a system. This high degree of power dissipation is undesirable in that it makes high-throughput communication systems, especially Gigabit Ethernet, inoperable and unmarketable. Accordingly, the art provides a method and apparatus for reducing power dissipation in communication systems that utilize a large number of taps, and reduces this type of power dissipation in high throughput systems such as Gigabit Ethernet. There is a need to provide a method and apparatus. Embodiments of the present invention meet these needs.

  One of the most important stages of operation of a Gigabit Ethernet transceiver is startup. At this stage, the adaptive filter included in the transceiver converges, the timing recovery subsystem acquires frequency and phase synchronization, the delay difference between the four wire pairs is compensated, and the pair match and polarity are To be acquired. When startup is successfully completed, normal operation of the transceiver can be started.

  In one activation protocol known as “blind start”, the transceiver converges its adaptive filter and timing recovery system simultaneously, while also acquiring timing synchronization. The disadvantage of this activation is that a high level of interaction is required between the various adaptation and acquisition algorithms in the transceiver. This high level interaction reduces the reliability of convergence and synchronization operations that occur during startup.

  Thus, the art provides an activation protocol for use in high-throughput communication systems such as Gigabit Ethernet that uses an optimal sequence of operations and minimizes interaction between various adaptation and acquisition algorithms. There is a need to do. Embodiments of the present invention meet these needs.

Briefly and in general terms, the present invention relates to a system and method for reducing noise and power dissipation in a communication system. The invention also relates to an activation protocol for use in a communication system.
In a first aspect, the present invention relates to a communication system having a predetermined threshold error. The communication system includes a communication line having a plurality of twisted wire pairs, a plurality of transmitters having one transmitter at each end of each twisted wire pair, and each end of each twisted wire pair. Multiple receivers with one receiver in each, each receiver including a direct signal from a transmitter at the opposite end of the twisted wire pair to which the receiver is associated and a plurality of noise signals Receive a signal. The system also includes a plurality of adaptive filters responsive to the combined signal, each adaptive filter having a plurality of taps each having a coefficient, each tap being switchable between an active and inactive state. The system further periodically adjusts the transfer function of at least one adaptive filter by selectively deactivating taps, ensuring that the communication system error does not exceed a threshold error. A control unit is included.

By selectively deactivating at least one adaptive filter tap, the present invention reduces the power consumption of the filter and thus the overall power consumption of the communication system.
In a more detailed aspect, the communication system further includes a plurality of noise reduction systems each comprising at least one adaptive filter. One noise reduction system is associated with each receiver and provides at least one duplicate noise impairment signal. The system also includes a plurality of devices, one associated with each receiver. Each device is responsive to the combined signal received by the receiver and the duplicate noise impairment signal provided by the noise reduction system associated with the receiver and substantially removes at least one noise signal from the combined signal. In another aspect, the noise signal includes multiple far-end crosstalk, one from each transmitter at the opposite end of the communication line, other than the transmitter at the opposite end of the twisted wire pair with which the receiver is associated. FEXT) fault signals are included and the noise reduction system includes a FEXT cancellation system that provides a duplicate FEXT fault signal as one of the duplicate noise fault signals. In yet another aspect, the noise signal includes a plurality of near-end crosstalk (NEXT), one from each transmitter at the same end of the communication line, other than the transmitter at the same end of the twisted wire pair with which the receiver is associated. ) A fault signal is included and the noise reduction system includes a NEXT cancellation system that provides a duplicate NEXT fault signal as one of the duplicate noise fault signals. In yet another aspect, the noise signal includes an echo impairment signal received from the transmitter at the same end of the twisted wire pair with which the receiver is associated, and the noise reduction system includes one of the duplicate noise impairment signals. One is an echo canceller that provides a duplicate echo impairment signal. In another detailed aspect, the controller includes means for setting the state of each tap, means for calculating the current error of the system, and means for comparing the current error with a threshold error. In yet another aspect, means for setting the state of each tap includes means for specifying a tap threshold for each tap, means for comparing the absolute value of the tap coefficient with the tap threshold for each tap, and absolute Means for deactivating taps whose values have a coefficient less than the tap threshold.

  In a second aspect, the present invention is a method of operating a communication system having a communication line having a plurality of twisted wire pairs and a plurality of transceivers at each end of each twisted wire pair. . Each transceiver has a receiver and a transmitter. Each receiver receives a combined signal including a direct signal from a transmitter at the opposite end of the twisted wire pair with which the receiver is associated and a plurality of noise signals. Each transceiver further includes a plurality of adaptive filters responsive to the combined signal. Each adaptive filter has a plurality of taps each having a coefficient. Each tap can be switched between active and inactive states. The method includes the steps of specifying a system threshold and transmitting at least one adaptive filter by selectively deactivating taps while ensuring that the system error does not exceed the threshold error. Adjusting the function periodically.

  In more detail, the method further includes generating at least one duplicate noise impairment signal for each receiver and at least one duplicate to generate an output signal that is substantially free of at least one noise signal. Combining the noise impairment signal with the combined signal. In another aspect, adjusting the transfer function includes setting the state of each tap, calculating the current error of the system, and combining the current error with a threshold error. It is. In another aspect, one transceiver in the communication system serves as a master and the other transceiver serves as a slave. Each transceiver has a noise reduction system, a timing recovery system and at least one equalizer. The method further includes performing a first stage in which the slave timing recovery system and equalizer are trained and the master noise reduction system is trained; the master timing recovery system and equalizer are trained and the slave Performing a second stage in which the noise reduction system is trained and performing a third stage in which the master noise reduction system is trained. In another aspect, one transceiver in the communication system serves as a master and the other transceiver serves as a slave. Each transceiver has a noise reduction system, a timing recovery system and at least one equalizer. The method further includes performing a first stage in which the slave timing recovery system and equalizer are trained and the master noise reduction system is trained, and the master timing recovery system is trained in both frequency and phase. Performing a second stage in which the master equalizer is trained and the slave noise reduction system is trained, and the master noise reduction system is retrained and the master timing recovery system is retrained for phase and slave timing recovery. Performing a third stage in which the system is retrained for both frequency and phase.

  In a third aspect, the present invention provides a communication line having a plurality of twisted wire pairs, a plurality of transmitters having one transmitter at each end of each twisted wire pair, a plurality of receivers, It is a communication system containing. One receiver is at each end of each twisted wire pair. Each receiver has a direct signal from the transmitter at the opposite end of the twisted wire pair with which the receiver is associated and a plurality of far end crosstalk (FEXT) from each remaining transmitter at the opposite end of the communication line. ) Receive a combined signal including a fault signal. The system also includes a plurality of FEXT cancellation systems, one associated with each receiver. Each FEXT cancellation system provides a duplicate FEXT failure signal. The system further includes a plurality of delay devices, one associated with each receiver. Each delay device responds to the combination signal received by its receiver and delays the combination signal. Also included are a plurality of first devices, one associated with each receiver. Each first device is substantially responsive to the output of the delay device associated with the receiver and the duplicate FEXT failure signal provided by the FEXT cancellation system associated with the receiver, and substantially removes the FEXT failure signal from the combined signal. .

  By providing a plurality of FEXT cancellation systems that generate a duplicate FEXT fault signal and a plurality of devices that combine the duplicate FEXT fault signal with the combined signal, the present invention substantially cancels the FEXT fault signal from the combined signal. Therefore, signal degradation due to noise in the communication system is reduced, and transmitted information is more reliably recovered.

In more detail, the FEXT cancellation system includes means for receiving a signal from each receiver at the same end of the communication system other than the receiver with which the FEXT canceller is associated. The FEXT cancellation system also includes means for generating a separate duplicate FEXT fault signal for each received signal and means for combining the individual duplicate FEXT fault signals to generate a duplicate FEXT fault signal. In another aspect, when the direct signal arrives after the FEXT signal, the delay device delays the combined signal by an amount approximately equal to the time delay between the FEXT impairment signal and the direct signal reaching the receiver. In another aspect, when the direct signal arrives after the FEXT impairment signal, the delay device has an amount approximately equal to the time delay between the FEXT impairment signal and the direct signal reaching the receiver;
The combination signal is delayed by the larger amount such that the combination signal is synchronized with the combination signal from the other receiver.

  In a fourth aspect, the present invention relates to a method for reducing noise in a communication system. The system includes a communication line having a plurality of twisted wire pairs. The system also includes a plurality of transmitters with one transmitter at each end of each twisted wire pair. The system further includes a plurality of receivers with one receiver at each end of each twisted wire pair. Each receiver has a direct signal from the transmitter at the opposite end of the twisted wire pair with which the receiver is associated and a plurality of far end crosstalk (FEXT) from each remaining transmitter at the opposite end of the communication line. ) Receive a combined signal including a fault signal. The method includes, for each receiver, generating a duplicate FEXT impairment signal and combining the duplicate FEXT impairment signal with the combined signal to produce an output signal with little FEXT impairment signal.

  In a fifth aspect, the present invention provides a method for reducing power dissipation in a communication system having a plurality of adaptive filters with a plurality of taps each with a coefficient, each tap being switchable between active and inactive states. Accompanied by. The method includes the steps of a) specifying an acceptable error of the system, b) setting a tap threshold for each active tap, and c) for each active tap the absolute value is the active tap. Deactivating taps having a coefficient less than the tap threshold set for d) calculating d system error; and e) comparing the calculated system error with an acceptable system error. F) increasing the tap threshold for each active tap if the calculated system error is less than an acceptable system error; and g) the calculated system error does not exceed the allowable system error. Repeating steps c) to f) until an acceptable system error is approached. Murrell.

  In a sixth aspect, the present invention provides power dissipation in a communication system having at least one adaptive filter with a plurality of taps, each tap being switchable between active and inactive states, each tap having a coefficient. With a method to reduce. The method includes a) calculating an initial system error, b) setting a tap error threshold for each active tap, and c) for each active tap, the absolute value for the active tap. Deactivating taps having a coefficient less than a set tap error threshold; d) calculating a next system error; e) a difference between the next system error and the initial system error is predetermined. If not, increase the tap error threshold for each active tap; and f) Steps c) to e) until the difference between the next system error and the initial system error exceeds a predetermined value. The step of repeating is included.

  In a seventh aspect, the present invention is an activation protocol for use in a communication system having a master transceiver at one end of a twisted wire pair and a slave transceiver at the other end of the twisted wire pair. is there. Each transceiver has a near-end noise reduction system, a far-end noise reduction system, a timing recovery system, and at least one equalizer. The protocol includes the steps of maintaining the master in a half-duplex mode that transmits signals but does not receive any signals during the first phase, and a half-duplex mode that receives signals from the master but does not transmit any signals. Maintaining the slave at a frequency, converging the master near-end noise reduction system, and the frequency and phase of the signal received by the slave so that the frequency and phase are synchronized with the frequency and phase of the signal transmitted by the master. And adjusting the slave equalizer. Also, during the second stage, maintaining the slave in a half-duplex mode that transmits signals but does not receive any signals, and puts the master in half-duplex mode that receives signals from the slaves but does not transmit any signals. Maintaining, freezing the frequency and phase of the slave, converging the slave near-end noise reduction system, and a signal received by the master so that the phase is synchronized with the phase of the signal transmitted by the slave. And adjusting the master equalizer. Also, during the third stage, maintaining the slave in full-duplex mode such that the slave transmits and receives signals, and maintaining the master in full-duplex mode such that the master transmits and receives signals. And reconverging the master near-end noise reduction system.

  In an eighth aspect, the present invention comprises a master transmitter / receiver at one end of a communication line and a slave transmitter / receiver at the other end of the communication line, each transmitter / receiver having a near-end noise reduction system and a far-end noise reduction. An activation protocol for use in a communication system having a system, a timing recovery system, and at least one equalizer. The protocol includes the steps of maintaining the master in a half-duplex mode that transmits signals but does not receive any signals during the first phase, and a half-duplex mode that receives signals from the master but does not transmit any signals. Maintaining the slave at a frequency, converging the master near-end noise reduction system, and the frequency and phase of the signal received by the slave so that the frequency and phase are synchronized with the frequency and phase of the signal transmitted by the master. And adjusting the slave equalizer. The protocol further includes, during the second phase, maintaining the slave in a half-duplex mode that uses a free-running clock to transmit signals but does not receive any signals, and receives signals from the slaves but does not receive any signals. Maintaining the master in half-duplex mode without transmitting signals, converging the slave near-end noise reduction system, and receiving by the master so that the frequency and phase are synchronized with the frequency and phase of the signal transmitted by the slave Adjusting the frequency and phase of the processed signal and converging the master equalizer. The protocol also includes maintaining the slave in full-duplex mode, during which the slave transmits and receives signals, and the master in full-duplex mode, in which the master transmits and receives signals. Maintaining the master near-end noise reduction system, adjusting the phase of the signal received by the master so that the phase is synchronized with the phase of the signal transmitted by the slave, and the frequency And adjusting the frequency and phase of the signal received by the slave so that the phase is synchronized with the frequency and phase of the signal transmitted by the master.

By dividing the activation protocol into three stages, the convergence of the equalizer and the timing recovery system is separated from the convergence of the noise reduction system. Thus, the interaction between various adaptation and acquisition algorithms in the transceiver is reduced and the reliability of convergence and synchronization operations is improved.
These and other aspects and advantages of the present invention will become apparent from the following more detailed description when viewed in conjunction with the accompanying drawings which illustrate preferred embodiments of the invention by way of example.

  The discussion herein is for the purpose of explaining and understanding the present invention and is particularly concerned with Gigabit Ethernet. However, it is understood that the concepts and claims of the present invention apply to other types of communication systems other than Gigabit Ethernet.

Communication System Overview A communication system incorporating features of the present invention is shown generally at 10 in FIG. The system 10 includes a hub 12 and a plurality of computers to be serviced by a hub in a local area network (LAN). Although four computers 14 are shown by way of example, a different number of computers may be used without departing from the scope of the present invention. Each computer 14 is separated from the hub 12 by a distance of about 100 meters (100 m). Computers 14 are also remote from each other.

  The hub 12 is connected to each computer 14 by a communication line 16. The communication line 16 includes a plurality of unshielded twisted pair wires or cables. Generally, the wire or cable is formed from copper. Four unshielded twisted pair wires are provided between each computer in the system 10 and the hub 12. The system shown in FIG. 1 operates with several categories of twisted pair cables designated as categories 3, 4, 5, 6, and 7 in the communications industry. Category 3 cables are the lowest quality (and lowest cost), and categories 6 and 7 are the highest quality (and highest cost). Gigabit Ethernet uses Category 5 cable.

  FIG. 2 illustrates in detail a portion of the communication system of FIG. 1 including one communication line 16 and one computer 14 and a portion of the hub 12. The communication line 16 includes four unshielded twisted pair wires 18 operating at 250 Mb / sec / pair. A transceiver 20 including a transmitter (TX) 22 and a receiver (RX) 24 is disposed at each unshielded end of each twisted pair 18. There is a hybrid 26 between each transceiver 20 and its associated unshielded twisted pair 18. The hybrid 26 is an interface to the communication line 16 that enables full duplex bidirectional operation between the transceivers 20 at each end of the communication line. The hybrid also functions to isolate the transmitter and receiver associated with the transceiver from each other.

  The communication system includes a standard connector called Gigabit Media Independent Interface (GMII) 28. GMII 28 is an 8-bit wide data path in both transmit and receive directions. Clocked at an appropriate frequency, such as 125 MHz, GMII produces a net throughput of bidirectional data at an appropriate rate, such as 250 Mb / sec / pair. GMII provides a symmetric interface in both transmit and receive directions. A physical coding sublayer (PCS) 30 receives and transmits data between the GMII 28 and the transceiver 20. The PCS 30 performs functions such as scrambling and encoding / decoding data before transferring the data to either the transceiver or the GMII. The PCS encodes the bits from GMII into a 5-level pulse amplitude modulation (PAM) signal. The five symbol levels are -2, -1, 0, +1 and +2. The PCS also controls some functions of the transceiver, such as skew control as described below.

Transceiver Circuits Four transceivers 20 are illustrated in detail in FIG. The components of the transceiver 20 are shown as overlapping blocks with each layer corresponding to one transceiver. The GMII 28, PCS 30, and hybrid 26 in FIG. 9 correspond to the GMII, PCS, and hybrid in FIG. 2 and are considered to be separated from the transceiver. The combination of the transceiver 20 and the hybrid 26 forms one “channel” of the communication system. Accordingly, FIG. 9 illustrates four channels each operating in a similar manner.

The transmitter portion of each transceiver 20 includes a pulse shaping filter 32 and a digital / analog (D / A) converter 34. In the preferred embodiment of the present invention, the D / A converter 34 operates at 125 MHz. The pulse shaping filter 32 receives one one-dimensional (1-D) symbol from the PCS. This symbol is called TXDatax symbol 36 when x is 1-4 corresponding to each of the four channels. The TXDatax symbol 36 represents 2-bit data. The PCS generates one 1-D symbol for each channel. Each channel symbol passes through a spectral shaping filter of form 0.75 + 0.25z ⁻¹ with a pulse shaping filter 32, and emissions are limited to within the FCC requirements. This simple filter shapes the spectrum so that the power spectral density of the transmitter output is lower than that of a communication system operating at 100 Mb / s over two pairs of category 5 twisted pair wires. The symbol is then converted to an analog signal by a D / A converter 34 which also serves as a low pass filter. The analog signal gains access to the unshielded twisted pair wire 18 through the hybrid circuit 26.

  The receiver portion of each transceiver includes a signal detector 41, an A / D converter 42, a FIFO 44, a digital adaptive equalizer system, a timing recovery circuit, and a noise reduction circuit. The digital adaptive equalizer system includes a feed forward equalizer (FFE) 46, two devices 50, 56, a skew adjuster 54 and two detectors 58, 60. The function of these components is described below in connection with the present invention. The noise reduction circuit includes a NEXT cancellation system 38, an echo canceller 38, and a FEXT cancellation system 70.

  The A / D converter 42 provides digital conversion of the signal received from the hybrid 26 at an appropriate frequency equal to the signal baud rate, such as 125 MHz. The A / D converter 42 samples the analog signal with the analog sample clock signal provided by the decision-oriented timing recovery circuit 64. The FIFO 44 receives the digital conversion signal from the A / D converter 42 and stores it on a first-in first-out basis. FIFO 44 forwards the individual signals to FFE 46 by means of a digital sample clock signal 80 provided by timing recovery circuit 64. The FFE 46 receives the digital signal from the FIFO 44 and filters the signal. The FFE 46 is a least mean square (LMS) type adaptive filter that corrects signal distortion by performing channel equalization and preceding intersymbol interference (ISI) cancellation.

  As noted, the signal that is introduced into the A / D converter 42 and then into the FIFO 44 and FFE 46 has several components. Such components include direct signals received directly from the transmitter 22 at the opposite end of the unshielded twisted pair wire 18 with which the receiver 24 is associated. Also included are one or more NEXT, echo, and FEXT fault signals from other transmitters 22 as previously described. A signal including a direct signal and one or more fault signals is called a “combined signal”.

  The FFE 46 forwards the combination signal 48 to the second device 50, which is usually a summing device. In the second device 50, the combined signal 48 is combined with the outputs of the NEXT cancellation system 38 and the echo canceller 40 to produce a signal with little NEXT and echo impairment signals. This signal is called “first soft decision” 52. The signal detector 41 detects a signal from the second device 50 and transfers the signal to the skew adjuster 54. Upon signal detection, the signal detector 41 initiates various system operations, one of which includes transitions between steps of the activation protocol, as described below. Skew adjuster 54 receives first soft decision 52 from second device 50 and outputs a signal called “second soft decision” 66. The skew adjuster 54 performs two functions. First, the length of the unshielded twisted pair 18 is compensated by delaying the first soft decision 52 so that the second soft decisions 66 from all receivers in the system are synchronized. Second, the delay of the first soft decision 52 is adjusted so that the second soft decision 66 arrives at the first device 56 almost simultaneously with the output of the FEXT cancellation system 70. The skew adjuster 54 receives the skew control signal 82 from the PCS 30.

  The skew adjuster 54 forwards the second soft decision 66 to the first device 56, which is typically a summing device. In the first device 56, the second soft decision 66 is combined with the output of the FEXT cancellation system 70 to produce a signal with little FEXT fault signal. This signal is called “third soft decision” 68. The first detector 58 receives the third soft decision 68 from the first device 56. The first detector 58 provides an output signal, or “final decision” 72. The first detector 58 is a slicer that generates a final decision 72 that corresponds to the analog signal level that is closest in magnitude to the level of the third soft decision 68. The first detector 58 is also a sequential detector that operates on the order of the signals across all four channels, such as a symbol-by-symbol detector or Viterbi decoder.

  In one configuration of the transceiver, the first detector 58 is a symbol-by-symbol detector. A group of detectors 58 for each symbol, one for each channel, is shown in FIG. Each first detector 58 includes a slicer 98, an adaptive feedback filter 100 and a summer 102. Summer 102 combines the third soft decision 68 with the output of adaptive feedback filter 100 and provides an output, which is introduced into slicer 98. The output of the slicer 98 is introduced into the adaptive feedback filter 100. The first detector 58 outputs corresponding to a discrete level from the set [−2, −1, 0, 1, 2] that is closest to the difference between the third soft decision 68 and the output of the feedback filter 100. A signal 72 is provided. The adaptive feedback filter 100 corrects the distortion of the third soft decision 68. This filter 100 uses past slicer 98 decisions to estimate subsequent ISI generated by the channel. This ISI is canceled from the third soft decision 68 and a final decision signal 72 is formed.

  In another configuration of the transceiver, the first detector 58 is a combination of a sequential decoder and a decision feedback equalizer (DFE) that uses an architecture commonly known as a multiple DFE architecture (MDFE) sequential detector. Sequential decoder 58 views all signals from all four channels simultaneously and in consecutive samples from each channel over several periods of unit time. The sequential decoder receives at least one signal from each first device 56 as input. The sequential decoder 58 generally responds to the order of the output signals from the first device 56, (1) passes the acceptable order of the signals, and (2) the constraints established by the code standard associated with the system. Discards the unacceptable sequence of signals. An acceptable order is subject to code constraints and an unacceptable order is a violation of code constraints.

  The second detector 60 (FIG. 9) receives the first soft decision 52 from the second device 50. The second detector 60 is a symbol-by-symbol detector similar to the first detector 58 (FIG. 10). This provides an output signal 74 corresponding to a discrete level from the set [−2, −1, 0, 1, 2] that is closest to the difference between the first soft decision 52 and the output of the feedback filter 100. To do. Since the second detector 60 produces an output signal 74 that does not use FEXT cancellation, this determination has a higher error rate than that by the first detector 58 that uses FEXT cancellation. Because of this fact, this decision is called a “provisional decision”. Subsequent ISI present in the input to the second detector 60 will be canceled using the adaptive feedback filter 100 (FIG. 10) housed in the second detector with the provisional decision 74 as input. It is important to note. The coefficients of this adaptive feedback filter 100 are the same as those of the adaptive feedback filter associated with the first detector 58 (FIG. 9).

  The third device 62, which is typically a summing device, receives the first soft decision signal 52 from the second device 50 and the provisional decision signal 74 from the second detector 60. In the third device 62, the first soft decision 52 is combined with the tentative decision signal 74 to generate an error signal 76 that is introduced into the timing recovery circuit 64. The timing recovery circuit 64 receives the tentative decision 74 from the second detector 60 and the error signal 76 from the third device 62. Using these signals as inputs, the timing recovery circuit 64 outputs an analog clock synchronization signal 78 introduced into the A / D converter 42 and a digital clock synchronization signal 80 introduced into the FIFO 44. As previously mentioned, these signals control the rate at which the A / D converter 42 samples the analog input received from the hybrid 26 and the rate at which the FIFO transfers the digital signal to the FFE 46.

  As previously mentioned, symbols transmitted by transmitter 22 (FIG. 2) in the communication system cause NEXT, echo, and FEXT impairments in the received signal for each channel. Since each receiver 24 has access to the data of the other three channels that generate this interference, each of these effects can be nearly counteracted. NEXT cancellation is achieved using three adaptive NEXT cancellation filters 84 as shown in the block diagram of FIG. Each NEXT cancellation system 38 receives three TXDatax symbols 36 from each transmitter at the same end of the communication line 18 as the receiver with which the NEXT cancellation system is associated. Each NEXT cancellation system 38 includes three filters 84, one for each TXDatax symbol 36. These filters 84 model the NEXT noise impulse response from the transmitter and are implemented, for example, as an adaptive transversal filter (ATF) using the LMS algorithm. Filter 84 generates a NEXT fault signal replica for each TXDatax symbol 36. The summing device 86 combines the three individual duplicate NEXT fault signals 92 to produce a NEXT fault signal replica that is included in the combined signal received by the NEXT cancellation system 38 associated receiver. The duplicate NEXT fault signal 88 is introduced into the second device 50 (FIG. 9) where it is combined with the combination signal 48 to produce a first soft decision signal 52 with little NEXT fault signal.

  Echo cancellation is achieved by an adaptive echo cancellation filter 85 as shown in the block diagram of FIG. Each echo canceller 40 receives a TXDatax symbol 36 from a transmitter at the same end as the receiver with which the echo canceller of twisted wire pair 18 is associated. As shown in FIG. 12, each echo canceller 40 includes one filter 85. The filter 85 models an impulse response of echo noise from the transmitter, and is realized as an ATF using an LMS algorithm, for example. This filter produces a replica of the echo impairment signal contained in the combined signal received by the receiver with which the echo canceller 40 is associated. The duplicate echo impairment signal 90 is introduced into the second device 50 (FIG. 9) and combined with the combined signal 48 to produce a first soft decision signal 52 with little echo impairment signal.

  FEXT cancellation is achieved by three adaptive FEXT cancellation filters 87 as shown in the block diagram of FIG. Each FEXT cancellation system 70 receives three provisional decision symbols 74, one from each receiver at the same end as the receiver with which the communication line FEXT cancellation system is associated. Each FEXT cancellation system 70 includes three filters 87, one for each provisional decision symbol 74. These filters 87 model the impulse response of the FEXT noise from the transmitter, and are realized as an ATF using the LMS algorithm, for example. Filter 87 generates a copy of FEXT fault signal 96 for each individual provisional decision symbol 74. The summing device 108 combines the three individual duplicate FEXT fault signals 96 to produce a replica of the FEXT fault signal contained in the combined signal 48 received by the receiver with which the FEXT cancellation system is associated. The duplicate FEXT fault signal 94 is introduced into the first device 56 (FIG. 9) where it is combined with the second combination signal 66 to produce a third soft decision signal 68 with little FEXT fault signal. Since the decision used to cancel FEXT is statistically independent of the final decision 72 made by the receiver where FEXT is canceled, the high error rate of provisional decision 74 does not degrade the performance of FEXT cancellation system 70. It is important to note that.

  The symbols provided by the first detector 58 are decoded and descrambled by the receiving section of the PCS 30 before being introduced into GMII. Changes in the twisting of the wire pairs can cause delays through four channels up to 50 nanoseconds. As a result, symbols may not be synchronized across the four channels. As previously mentioned, if the first detector is a sequential detector, the PCS also determines the relative skew of the four streams of 1-D symbols and adjusts the skew before reaching the first detector 58. Since the symbol delay is adjusted by the unit 54, the sequential decoder can operate on properly configured 4-dimensional (4-D) symbols. Furthermore, because the cable plant may introduce a wire exchange or a pair exchange between the four unshielded twisted pairs, the PCS 30 also determines and corrects these conditions.

Noise Reduction As mentioned previously, FEXT is a failure that results from capacitive coupling of signals from the far-end transmitter to the receiver input, as shown in FIG. The crosstalk signals from transmitters F, G and H appear as noise to receiver A trying to detect the signal from transmitter E. A similar situation applies to all other receivers with respect to the signal from a suitable transmitter located at the opposite end of the line.

  The FEXT noise experienced by receiver A and generated from transmitter F depends on the characteristics of the cable, with an impulse response that models the coupling characteristics of the unshielded twisted pair used by transmitter F and receiver A, It can be modeled as the convergence of the data symbols transmitted by F. A commonly measured FEXT impulse response 104 is shown in FIGS. Similar descriptions can be made for all other possible receiver and transmitter combinations. Thus, there are a total of 12 FEXT impulse responses that describe FEXT noise signals from transmitters E, F, G, and H to receivers A, B, C, and D. These twelve impulse responses are not identical, but each have a general shape similar to that shown in FIGS.

  Although FEXT is also an obstacle for many communication systems other than Gigabit Ethernet, in those systems the receivers may not be physically co-located and / or for receivers experiencing FEXT. Because other receivers operate at a rate that is not synchronized with the data rate, a given receiver typically does not have access to symbols detected by other receivers. Aspects of the invention take advantage of the fact that in a Gigabit Ethernet transceiver, decisions corresponding to all four channels are available to the four receivers and the decisions are synchronized.

  In operation, there may be a delay associated with transmitting a signal over the communication line. Signal synchronization within the system is very important for effective noise cancellation. It is important that the replica noise impairment signal arrives at the total device almost simultaneously with the combined signal and / or the soft decision signal. With respect to the FEXT failure signal, since the failure occurs at the transmitter at the opposite end of the receiver, the time between when the second soft decision signal 66 reaches the first device 56 and when the duplicate FEXT failure signal 94 arrives. The delay is likely to occur. Some channels have a group delay of the FEXT signal 104 that is less than the desired group delay of the signal 106, as illustrated in FIG. In this case, the provisional decision 74 provided by receivers B, C, and D of FIG. 5 arrives too late at receiver A's FEXT cancellation system 70 and is therefore not effective in negating the FEXT fault.

  To compensate for this delay, the present invention utilizes a skew adjuster 54, which, as previously mentioned, is the time between the direct signal reaching the receiver and the FEXT impairment signal associated with that receiver. The first soft decision signal 52 is delayed by a time approximately equal to or greater than the delay. If the output is delayed by an amount greater than the time delay, the situation illustrated in FIG. 16 results, but adaptive feedback filter 87 (FIG. 13) in FEXT cancellation system 70 delays duplicate FEXT fault signal 94. So that it reaches the first device 56 (FIG. 9) almost simultaneously with the second soft decision signal 66.

  The third soft decision 68 that results from the FEXT cancellation allows the first detector 58 to make a more reliable final decision 72 with a much lower error rate. Computer simulations show that the normal improvement that can be achieved by the present invention described in this application is about when the signal-to-noise ratio at the input of the first detector 58 before FEXT cancellation is about 25 dB. 2 to 3 dB. This corresponds to a reduction in symbol error rate of 1000 times or more.

  If there is no delay associated with the transmission of the signal over the communication line, both the FEXT fault signal 104 and the direct signal 106 reach the receiver almost simultaneously as shown in FIG. In this situation, the skew adjuster 54 is set to zero. Alternatively, an embodiment of the present invention is used that has only one detector 110 and one summing device 112 as shown in FIG. In this configuration, the summing device 112 receives the duplicate NEXT, echo and FEXT fault signals 88, 90, 94 and the combined signal 48 and generates a first soft decision 52 with little fault signal. This first soft decision 52 is introduced into the detector 110 and the third device 62. Detector 110 may include a single symbol-by-symbol detector or both a symbol-by-symbol detector and a sequential detector. In the case of a symbol-by-symbol detector, the final decision 72 and the second output 114 of the detector 110 are identical. In both the symbol-by-symbol detector and the sequential detector, the final output is provided by the sequential detector and introduced into the PCS 30. The second output 114 is provided by a symbol-by-symbol detector and is introduced into the timing recovery circuit 64 and the third device 62 and used in determining the error signal 76.

Power Dissipation Reduction As previously mentioned, the NEXT cancellation system, echo canceller and FEXT cancellation system use ATF to effectively cancel noise from the combined signal. An example of ATF used is shown in FIG. The ATF 120 includes a plurality of taps 122 each including a multiplexer 124 and an adder 126. Associated with each tap 122 is a coefficient C _n , where n is 0 to x−1 when x is the number of taps in the ATF. The circuitry associated with each tap includes a 1-bit storage device (not shown) that allows tap activation and deactivation. The value of the coefficient C _n is adjusted by the LMS algorithm as previously mentioned. A register 128 is inserted between the taps 122. Such a register 128 provides data to the tap 122 at time intervals that are aligned with the clock signal.

  As shown in FIGS. 7 and 8, the echo and NEXT impulse responses show that not all taps 122 in NEXT and echo cancellers 38, 40 contribute significantly to the performance of the communication system. . Aspects of the present invention determine which taps 122 do not contribute significantly to reducing the mean square error (MSE) of the system and deactivate those taps, thereby removing them from the filtering computation and reducing system power dissipation. Is greatly reduced. Furthermore, as the impulse response of FIGS. 7 and 8 shows, it is difficult to avoid the need to construct a long-span NEXT and echo cancellers 38,40. Since the individual cable responses may differ from those shown in FIGS. 7 and 8, more or less taps 122 may be required than required for the cables of FIGS. As previously mentioned, it is difficult to determine a priori which tap 122 is required for each cable.

  In accordance with aspects of the present invention, the NEXT, echo and FEXT cancellers 38, 40, 70 are comprised of an ATF 120 that utilizes a sufficient number of taps 122 to provide sufficient cancellation for the worst case expected impulse response. To do. This may require 140 taps 122 as in the example of FIG. 7, and may require more taps for longer cables. To reduce power dissipation, tap 122 is examined after convergence, and taps that are found to contribute less to system performance are deactivated. When the tap 122 is deactivated, it is removed from NEXT, echo and FEXT replication calculations and adaptations, and the contribution to overall system power dissipation is almost eliminated.

When the system is initially converged, all taps 122 are active and NEXT, echo and FEXT cancellers 38, 40, 70 are converged along their entire length. After convergence, tap 122 is examined using the tap scanning algorithm shown in FIG. 19 to determine which taps can be deactivated. In step S1, an acceptable level of error S _ea is specified for the system. In step S2, a tap coefficient threshold T _th is set for each active tap. Each individual tap may have a unique tap threshold T _th , but in the preferred embodiment of the present invention, the tap coefficients of all taps are approximately equal. Since the initial value of the tap coefficient threshold T _th is sufficiently low, only a few taps 122 are deactivated and the performance of the system is not significantly affected. In a preferred embodiment of the invention, the tap coefficient threshold T _th is initially set equal to the tap coefficient C _n having the smallest absolute value. A reasonable value can also be determined by simulation. This initial value is not important as long as it is low enough to avoid significant degradation of system performance when the tap scanning procedure is first applied.

In step S3, the absolute value of the tap coefficients C _n for each active tap is compared with the tap coefficient threshold T _th. If the tap coefficient C _n is less than the tap coefficient threshold T _th , the tap 122 is deactivated at step S4. This procedure is repeated for each tap 122 in the filter 120. Preferably, the decision to deactivate tap 122 is made in a sequential manner starting from the input of filter 120. In step S5, the system error S _ec is calculated. This error is calculated by first calculating the MSE of each active tap 122 by multiplying the absolute value of the tap coefficient C _{n by} the average energy signal. The filter 120 error associated with tap 122 is determined by summing the individual tap errors. The system error is then determined by summing the individual filter errors.

In step S6, the calculated system error S _ec is compared with a specified allowable system error S _ea . If the calculated system error S _ec is less than the allowable system error S _ea , the tap threshold T _{th for} each active tap is increased by a small amount in step S7 and steps S3 to S6 are repeated. As a result, some additional taps 122 are deactivated, but the increase in tap threshold T _th is small, so the number of deactivated taps is usually not very large. Steps S3 to S6 are repeated until the calculated system error S _ec approaches the allowable system error S _ea without exceeding the allowable system error. If the calculated system error S _ec is greater than the allowable system error S _ea , the tap scanning algorithm stops.

As an alternative to deciding whether to deactivate a tap based on the MSE of the communication system, a decision may be made based on the MSE of the individual filters. In this embodiment of the invention, an acceptable level of error F _{ea for the} filter is specified. As explained previously, the individual taps are deactivated and the calculation filter error F _ec is calculated. If the deactivation of the tap does not cause the calculated filter error F _ec to exceed the allowable filter error F _ea , the tap remains inactive.

  In yet another embodiment of the present invention, the contribution of each deactivated tap to the filter and thus to the MSE of the system is calculated. If it is determined that the tap's contribution to the MSE is an acceptable amount, the tap remains deactivated. This method is generally preferred during initial startup of a system where the overall system MSE is large because the filter coefficients in the system are in a non-convergent state. When the system's filter coefficients initially converge, the decision to deactivate the tap is generally made based on the tap's contribution to the system's MSE as previously described in connection with FIG.

  The net result of the tap scanning algorithm is that in a normal channel of the communication system, a large number of taps 122 are deactivated and power dissipation is reduced by a large percentage. As an example, a computer simulation of a tap scanning algorithm with an echo response operating on the channel shown in FIG. 7 is presented in FIG. This figure shows the MSE to signal ratio of both master and slave as a function of time during the initial convergence of the system. At time t = 360,000 baud, the tap scanning algorithm is started so that the MSE to signal ratio increases to a predetermined target of 24 dB. FIG. 21 shows the tap of the echo canceller after convergence at the threshold of 24 dB, and FIG. 22 shows the tap of the threshold of 26 dB. Deactivated taps are shown as 0. In FIG. 21, the total number of active taps of the echo canceller is 22. Similarly, the number of active taps 122 in the three NEXT cancellers (not shown) forming the NEXT cancellation system is 6, 2 and 0, respectively. In FIG. 22, the total number of active taps of the echo canceller is 47. Similarly, the number of active taps 122 in three NEXT cancellers (not shown) is 6, 2, and 0, respectively.

  For the 24 dB threshold, only 30 of the 440 initial active taps remain active after the tap scanning algorithm is applied, maintaining a 5 dB margin for the required bit error rate. As can be seen from FIGS. 21 and 22, taps 122 that remain active occur at sparse locations, and the tap location is highly dependent on the individual cable response, so NEXT and echo cancellers are being designed. It would be difficult to statistically assign these taps.

In another embodiment of the invention, the tap scanning algorithm monitors changes in MSE during the tap scanning algorithm. The algorithm is applied until the change in the MSE, not the MSE itself, exceeds a predetermined value, for example 1 dB. If the MSE to signal ratio before the scan is applied is 25 dB, the final MSE to signal ratio is 24 dB and a number of taps are deactivated. This embodiment of the tap scanning algorithm is shown in FIG. In step S10, an initial system error S _ei is calculated. In step S11, an allowable system error difference _De is specified. In step S12, a tap coefficient threshold T _th is set for each active tap. As with other tap scanning algorithms, the value of the tap threshold T _th is sufficiently low so that only a few taps 122 are deactivated and the performance of the system is not significantly affected.

In step S13, the absolute value of the tap coefficients C _n for each active tap is compared with the tap coefficient threshold T _th. If the tap coefficient C _n is less than the tap coefficient threshold T _th , the tap 122 is deactivated in step S14. This process is repeated for each tap 122 in the filter 120. Preferably, the decision to deactivate tap 122 is made in a sequential manner starting from the input of filter 120. In step S15, the next error S _es of the system is calculated.

In step S16, the next error S _es of the system is compared with the initial system error S _ei . If the difference between the next system error S _es and the initial system error S _ei is smaller than a predetermined value, the tap coefficient threshold T _th is increased by a small amount in step 17 and steps S13 to S16 are repeated. As a result, some additional taps 122 are deactivated. Steps S13 to S16 are repeated until the difference between the next system error S _es and the initial system error S _ei exceeds a predetermined value.

In some cases, the NEXT, echo, and FEXT impulse responses may change during normal operation, for example, as a result of temperature changes. Thus, preferably in a sequential manner, the previously deactivated tap 122 is periodically activated to _recheck whether the absolute value of the tap coefficient C _n is less than or equal to the tap threshold T _th. desirable. If the tap coefficient C _n is greater than the tap threshold T _th , the tap 122 remains active, otherwise it is deactivated. Similarly, the active tap 122 may be below the tap threshold T _th , in which case the tap is deactivated. All of this can be accomplished by periodic reapplication of the sequential tap scanning algorithm during normal operation.

  In an alternative embodiment of the present invention, a selected number, such as ten taps 122 located at the input end of the filter 120, are not subject to deactivation. Usually, these first few taps 122 have a large skew rate, which means that the numerical value can vary greatly if the sampling phase changes. This sampling phase changes dynamically as a result of jitter and may make some of the previously deactivated taps 122 significant. By keeping several taps at the input of the filter active, the possibility of degradation in the presence of jitter is avoided.

  When the total number of taps is very large, e.g. 440, power dissipation may be significant during the initial initial convergence transient before the tap scanning algorithm has the potential to deactivate the taps. Even if the average power dissipation of the system is greatly reduced, the peak power dissipation is not reduced. The preferred embodiment of the present invention compensates for this by stepwise converging NEXT, echo and FEXT cancellers. For example, 20 tap blocks are converged at a time, and then a tap scanning algorithm is applied to those taps on a block-by-block basis. For example, if the initial block of 20 taps is not large enough to provide a low MSE, and the MSE during initial convergence is large, the deactivated tap 122 as a measure of whether to terminate the algorithm. It is better to monitor the sum of the square values of the coefficients.

Activation Protocol One of the most important steps in the operation of the communication system is the activation of the transceiver. At this stage, the adaptive filters included in the FFE 46 (FIG. 9), echo canceller 40, NEXT cancellation system 38, FEXT cancellation system 70, timing recovery system 64 and detector 58 of the receiver portion of each transceiver are included. Converge. During convergence, the actual output of the adaptive filter is compared with the expected output of the filter to determine the error. The error is reduced to nearly zero by adjusting the coefficients of the algorithm that defines the filter transfer function. Similarly, the timing recovery system is converged by adjusting the frequency and phase of the phase locked loop and local oscillator included in the timing recovery system so that the signal to noise ratio of the channel is optimized. In addition, the delay difference between the four wire pairs is compensated to obtain pair match and polarity. Successful start-up ensures that the transceiver can start normal operation.

  In accordance with aspects of the present invention, each transceiver channel operates in a loop time manner, as shown in FIG. The two end transceivers 20 of each twisted wire pair 18 have two different roles as far as synchronization is concerned. One transceiver, referred to as master 130, transmits data using an independent clock GTX_CLK provided through GMII interface 28 (FIG. 9). This clock signal is fixed in both frequency and phase and is provided to the master transceiver 130 of each of the four transceiver channels of the communication system. In practice, the transmit clock used by master 130 may be a filtered version of GTX_CLK obtained using a very narrow bandwidth phase-locked loop to reduce jitter. The other end transceiver 20 of the twisted wire pair 18, called slave 132, uses the timing recovery system 64 (FIG. 9) located in the receiver 24 to receive the frequency and phase of its receive and transmit clocks. Both are synchronized with the signal received from the master 130. The slave 132 transmission clock always maintains a constant phase relationship with the slave reception clock. The reception clock of the master 130 is synchronized in phase with the signal received from the slave transmitter 22 instead of the frequency. That is, after the initial acquisition period, the master 130 receive clock follows the master transmit clock with a phase difference determined by the round trip delay of the loop. This phase relationship may change dynamically as the master 130 receive clock needs to track jitter present in the signal received from the slave 132.

Fixed Slave Transmission Clock The sequence of events in the activation protocol of the present invention is shown in FIG. The protocol consists of three stages 134, 136, 138, during which the receiver is trained, eg after the adaptive filter is converged, timing synchronization is acquired, etc. Is started. In the first stage 134, the master initiates transmission to the slave using a transmission clock signal that is fixed in both frequency and phase. The master trains the near-end noise reduction system by converging the adaptive filter included in the echo canceller and the NEXT cancellation system (E). At the same time, the slave trains the equalizer and the far-end noise reduction system by converging the adaptive filters included in the DFE, FFE and FEXT cancellation systems (D). While training the equalizer and far-end noise reduction system, the slave simultaneously acquires both frequency and phase timing synchronization (T). The slave also ensures a differential delay between the four twisted wire pairs at this time, identifies the four pairs, and corrects the polarity of the pairs.

  In one embodiment of the protocol, the transition from the first stage 134 to the second stage 136 on both the master and slave occurs after a fixed period of time. However, in the preferred embodiment, when the slave detects that its receiver has converged the adaptive filters included in the DFE, FFE and FEXT cancellation systems (D) and has acquired timing synchronization (T), the first Transition from stage 134 to a second stage 136. As previously mentioned, the master receiver includes a signal detector 41 (FIG. 9) that detects the energy in the line coming from the slave. When the master detects this energy from the slave, it transitions from the first stage 134 to the second stage 136. Therefore, the slave has the initiative when transitioning from the first stage 134 to the second stage 136, and the master follows when detecting a signal from the slave.

  The convergence of the echo canceller and NEXT cancellation system during the first stage 134 at the master is done with the goal of enabling the master signal detector to detect the signal from the slave. In the absence of proper echo and NEXT cancellation, the signal detector may be triggered by echo and NEXT noise present in the receiver. After the transition is made, the master discards the echo canceller and NEXT cancellation system coefficients that resulted from the convergence in the first stage 134. This is done by resetting the adaptive filter in the echo canceller and NEXT cancellation system. Since the correct sampling phase of the four receivers of the master is obtained in the third stage 138, the coefficients of the echo canceller and NEXT cancellation system obtained in the first stage 134 are the final values reacquired in the third stage 138. It is important to note that can be different.

  In a second stage 136, the slave trains the near-end noise reduction system by converging the adaptive filter included in the echo canceller and the NEXT cancellation system (E). At the same time, the master trains the equalizer and far-end noise reduction system by converging adaptive filters included in the DFE, FFE and FEXT cancellation systems (D). While training the equalizer and far-end noise reduction system, the master simultaneously acquires phase-only timing synchronization (P). At this time, the master also compensates for the differential delay between the four twisted wire pairs, identifies the four pairs, and corrects the polarity of the pairs. In the second stage, the slave saves the timing recovery state variable obtained in the first stage 134 and freezes the frequency and phase. Doing this ensures that when the master resumes transmission at the start of the third stage 138, the slave samples the signal coming from the master to the slave in the correct phase. The slave also freezes the DFE, FFE and FEXT cancellation system coefficients obtained in the first stage 134.

  Similar to the transition from the first stage 134 to the second stage 136, the transition from the second stage 136 to the third stage 138 occurs after a fixed predetermined period of time. The duration of the first, second and third stages 134, 136, 138 is fixed, but the duration is not necessarily equal for all stages. However, in a preferred embodiment, when the master detects that the receiver has converged the adaptive filters included in the DFE, FFE and FEXT cancellation systems (D) and has acquired timing synchronization (P), the second stage Transition from 136 to a third stage 138. Similar to the master, the slave receiver includes a signal detector 41 (FIG. 9) that detects the energy in the line coming from the master. When the slave detects this energy from the master, it transitions from the second stage 136 to the third stage 138. Therefore, the master has the initiative when transitioning from the second stage 136 to the third stage, and the slave follows when detecting a signal from the master.

  In the third stage 138, the slave freezes the coefficients of the echo canceller and the NEXT cancellation system and maintains steady state conditions, while the operating characteristics of the slave are not adjusted. Similarly, the master freezes the DFE, FFE and FEXT cancellation system coefficients and their clock signal phases. The master also maintains a near-end noise reduction system by reconverging the echo canceller and NEXT cancellation system during the third stage 138 (E). Note that in the third stage 138, the slave resumes transmission using the clock recovered from the signal transmitted from the master, so that the master already knows the correct frequency at which the receiver operates. is important. The “relative sampling phase” of the four receivers, ie the difference between the sampling phase of the three receivers and one receiver that is optionally used as a reference, is also known because it was acquired in the second stage 136. Yes. However, the “total sampling phase” of the receiver, ie the sampling phase of the receiver arbitrarily selected as a reference, is not known and must be obtained in the third stage 138. When both the master and slave complete the training operation, they exchange a message indicating that they are ready to send valid data. In a fourth stage 140, all coefficients of the adaptive filter that have been frozen before are unfrozen and ready for data transmission.

Self-running slave transmission clock The sequence of events in the activation protocol of the present invention is shown in FIG. The protocol consists of three stages 144, 146, 148 during which the receiver is trained, eg, the adaptive filter is converged, timing synchronization is acquired, etc., then the normal operation in the fourth stage 150 Is started. The activation protocol is initiated by the master when the master starts transmitting a signal to the slave using a transmission clock signal that is fixed in both frequency and phase. In the first stage 144, the master trains the near-end noise reduction system by converging the adaptive filter included in the echo canceller and NEXT cancellation system (E). As previously mentioned, the slave receiver includes a signal detector 41 (FIG. 9) that detects the energy in the line coming from the master. Upon detecting the signal from the master, the slave trains the equalizer and far-end noise reduction system by converging adaptive filters included in the DFE, FFE and FEXT cancellation systems (D). While training the equalizer and far-end noise reduction system, the slave simultaneously acquires both frequency and phase timing synchronization (T). The slave also compensates for the differential delay between the four twisted wire pairs at this time, identifies the four pairs, and corrects the polarity of the pairs.

  In one embodiment of the protocol, the transition from the first stage 144 to the second stage 146 on both the master and slave occurs after a fixed period of time. However, in the preferred embodiment, when the slave detects that its receiver has converged the adaptive filters included in the DFE, FFE and FEXT cancellation systems (D) and has acquired timing synchronization (T), the first Transition from stage 144 to a second stage 146. The master receiver also includes a signal detector 41 (FIG. 9) that detects the energy in the line coming from the slave. When the master detects this energy from the slave, it transitions from the first stage 144 to the second stage 146. Therefore, the slave has the initiative in transitioning from the first stage 144 to the second stage 146, and the master follows when it detects a signal from the slave.

  The convergence of the echo canceller and NEXT cancellation system during the first stage 144 at the master is done with the goal of allowing the master signal detector to detect the signal from the slave. In the absence of proper echo and NEXT cancellation, the signal detector may be triggered by echo and NEXT noise present in the receiver. After the transition has taken place, the master discards the echo canceller and NEXT cancellation system coefficients that resulted from the convergence in the first stage 144. This is done by resetting the adaptive filter in the echo canceller and NEXT cancellation system. Since the correct sampling phase of the four receivers of the master is obtained in the third stage 148, the echo canceller and NEXT cancellation system coefficients obtained in the first stage 144 are the final values reacquired in the third stage 148. It is important to note that can be different.

  In the second stage 146, the slave trains the near-end noise reduction system by converging the adaptive filter included in the echo canceller and NEXT cancellation system (E). The slave also freezes the DFE, FFE and FEXT cancellation system coefficients obtained in the first stage 144. At the same time, the master trains the equalizer and far-end noise reduction system by converging the adaptive filters included in the DFE, FFE and FEXT cancellation systems (D). While training the equalizer and far-end noise reduction system, the master simultaneously acquires both frequency and phase timing synchronization (T). At this time, the master also compensates for the differential delay between the four twisted wire pairs, identifies the four pairs, and corrects the polarity of the pairs. The slave may deviate from any stable clock having a frequency offset below a predetermined limit, for example 200 ppm, when compared to the master transmit clock. In the second stage 146, the slave transceiver transmits using a free-running clock. As a result, the master acquires both frequency and phase synchronization in the second stage 146. Phase synchronization is a normal function that the master performs with any form of activation protocol, but frequency synchronization is not a normal master function and is a timing recovery state variable when transitioning from the first stage 144 to the second stage 146. This is only done in the second stage of activation 146, which is intended to perform an appropriate operation with a slave transceiver that does not save.

  Similar to the transition from the first stage 144 to the second stage 146, the transition from the second stage 146 to the third stage 148 occurs after a fixed predetermined period of time. The duration of the first, second and third stages 144, 146, 148 is fixed, but the duration is not necessarily equal for all stages. However, in the preferred embodiment, when the master detects that the receiver has converged the adaptive filters included in the DFE, FFE and FEXT cancellation systems (D) and has acquired timing synchronization (T), the second stage Transition from 146 to the third stage 148. The master starts sending signals to the slave. When the slave detects this signal from the master, it transitions from the second stage 146 to the third stage 148. Therefore, the master has the initiative in making the transition from the second stage 144 to the third stage 148, and the slave follows when detecting a signal from the master.

  In the third stage 148, the slave freezes the coefficients of the echo canceller and NEXT cancellation system and maintains the near-end noise reduction system in steady state conditions. Since the timing recovery state variable has not been saved on transition to the second stage 146, the slave also reacquires both frequency and phase timing synchronization at the third stage 148 (T). In a third stage 148, the master freezes the DFE and FEXT cancellation system coefficients and the frequency of the clock signal. The master also retrains the near-end noise reduction system by reconverging the echo canceller and NEXT cancellation system (E). The master also reacquires phase-only timing synchronization (P). Note that in the third stage 148, the slave resumes transmission using a clock recovered from the signal transmitted by the master, so that the master already knows the correct frequency at which the receiver operates. is important. The “relative sampling phase” of the four receivers, ie the difference between the sampling phase of the three receivers and one receiver that is optionally used as a reference, is also known because it was acquired in the second stage 146. Yes. However, the “total sampling phase” of the receiver, ie, the sampling phase of the receiver arbitrarily selected as a reference, is not known and must be obtained in the third stage 148. When both the master and slave complete the training operation, they exchange a message indicating that they are ready to send valid data. In the fourth stage 150, all the coefficients of the previously frozen adaptive filter are unfrozen and ready for data transmission.

  Although the invention has been disclosed and illustrated in connection with specific embodiments, the principles encompassed therein can be used in numerous other embodiments that will be apparent to those of ordinary skill in the art. It is. Accordingly, the invention is limited only as indicated by the scope of the appended claims.

1 is a schematic configuration diagram of a communication system that provides a plurality of computers connected to a hub by a communication line to form a local area network (LAN). FIG. 1 is a schematic block diagram of a communication system that provides a Gigabit Media Independent Interface (GMII), a Physical Coding Sublayer (PCS), and a plurality of unshielded twisted pair wires each having a transceiver at each end. FIG. 3 is a schematic configuration diagram of a part of the communication system of FIG. 2 showing a NEXT failure signal received by a receiver A from adjacent transmitters B, C and D. It is a schematic block diagram of a part of the communication system of FIG. FIG. 3 is a schematic configuration diagram of a part of the communication system of FIG. 2, showing FEXT failure signals received by receiver A from transmitters F, G, and H on the opposite side. 1 is a schematic configuration diagram of a communication system including a plurality of transceivers each having a NEXT cancellation system, an echo canceller, a feedforward equalizer, a digital adaptive filter system including one detector, and a timing recovery circuit. An impulse response of an echo signal passing through a 100 m communication line is shown. The impulse response of the NEXT signal passing through the 100 m communication line is shown. Schematic configuration of a communication system including a plurality of transceivers each having a NEXT cancellation system, an echo canceller and a FEXT cancellation system, a digital adaptive filter system including a plurality of detectors and skew adjusters, and a timing recovery circuit FIG. FIG. 10 is a schematic block diagram of the symbol-by-symbol detector of FIG. 9 that each includes a plurality of slicers, feedback filters and adders and receives a soft decision as input. FIG. 10 is a schematic configuration diagram of the NEXT cancellation system of FIG. 9 that includes a plurality of adaptive transversal filters (ATF) and adders, each receiving a transmission signal from an adjacent transmitter as an input. FIG. 10 is a schematic configuration diagram of the echo canceller of FIG. 9 each including an ATF and receiving a transmission signal from the same transmitter as an input. FIG. 10 is a schematic structural diagram of the FEXT cancellation system of FIG. 9 that receives a transmission signal from an opposite transmitter as an input, each including a plurality of ATFs and an adder. Fig. 4 shows a direct impulse response reaching the receiver after a FEXT impulse response. A direct impulse response and a FEXT impulse response reaching the receiver almost simultaneously are shown. Fig. 4 shows a direct impulse response reaching the receiver before the FEXT impulse response. 1 is a schematic configuration diagram of a communication system including a plurality of transceivers each having a NEXT cancellation system, an echo canceller and a FEXT cancellation system, a digital adaptive filter system including one detector, and a timing recovery circuit. FIG. 14 is a schematic diagram of an ATF including a cascade of taps present in the NEXT cancellation system of FIG. 11, the echo canceller of FIG. 12, and the FEXT cancellation system of FIG. 3 is a flow diagram illustrating one embodiment of a method for reducing power dissipation. Fig. 2 shows mean square error (MSE) versus signal ratio as a function of time during initial convergence of a communication system. Fig. 5 illustrates an echo canceller tap that is active after convergence in a communication system having an error threshold of 24 dB. Fig. 5 shows an echo canceller tap that is active after convergence in a communication system with an error threshold of 26dB. 5 is a flow diagram illustrating another embodiment of a method for reducing power dissipation. It is a schematic block diagram which shows the master-slave relationship between the transmitter / receiver of each transmitter / receiver channel of FIG. FIG. 6 is a timing diagram illustrating the steps of one embodiment of an activation protocol. FIG. 6 is a timing diagram illustrating steps of another embodiment of an activation protocol.

Claims (2)

An activation protocol for use in a communication system having a master transceiver connected to one end of a twisted wire pair and a slave transceiver connected to the opposite end of the twisted wire pair. ,
Each of the master transceiver and the slave transceiver includes a near-end noise reduction system , a timing recovery system that adjusts frequency and phase so that a signal-to-noise ratio is optimized , and at least one equalizer. ,
The activation protocol is:
In the first stage,
The master transceiver transmits a signal to the slave transceiver using a transmission clock signal with both fixed frequency and phase;
The master transceiver converges the near-end noise reduction system;
The slave transceiver converges the equalizer; and
Adjusting the frequency and phase of the slave transceiver to synchronize with the frequency and phase of the signal transmitted from the master transceiver using the timing recovery system;
In the second stage after each step of the first stage is completed,
The slave transceiver starts transmitting a signal to the master transceiver at the frequency and phase adjusted in the first stage, and fixes the frequency and phase ;
The slave transceiver converges the near-end noise reduction system;
The master transceiver converging the equalizer; and
Adjusting the phase so that the master transceiver is synchronized with the phase of the signal transmitted from the slave transceiver;
In the third stage after each step of the second stage is completed,
The master transceiver restarts signal transmission at the phase adjusted in the second stage and the frequency adjusted in the first stage;
The master transceiver includes re-converging the near-end noise reduction system. An activation protocol for use in a communication system having a master transceiver connected to one end of a communication line and a slave transceiver connected to the opposite end of the communication line,
Wherein each of the master transceiver and said slave transceiver has near-end noise reduction systems, timing recovery system, and at least one equalizer,
The protocol is
In the first stage,
The master transceiver transmits a signal to the slave transceiver using a transmission clock signal with both fixed frequency and phase;
The master transceiver converges the near-end noise reduction system;
The slave transceiver converges the equalizer; and
Adjusting the frequency and phase of the slave transceiver to synchronize with the frequency and phase of the signal transmitted from the master transceiver using the timing recovery system;
In the second stage after the completion of each step of the first stage ,
A step before Symbol slave transceiver, and initiating a transmission of the signal using a free-running clock, the slave transceiver, for converging the near-end noise reduction system,
Adjusting the frequency and phase so that the master transceiver is synchronized with the frequency and phase of the signal transmitted from the slave transceiver;
The master transceiver converging the equalizer; and
In the third stage after each step of the second stage is completed,
Adjusting the frequency and phase so that the slave transceiver is synchronized with the frequency and phase of the signal transmitted from the master transceiver ;
The master transceiver reconverges the near-end noise reduction system;
And a step of adjusting the phase so that the master transceiver is synchronized with the phase of a signal transmitted from the slave transceiver .

Images