DETAILED DESCRIPTION OF THE INVENTION

[0039] Discrete Multi-Tone (DMT) data transmission schemes have been shown to facilitate high performance data transmission. Among the benefits of DMT architectures is that they have high spectral efficiencies and can adaptively avoid various signal distortion and noise problems. Since they have very high data transmission capabilities, in most applications selection of a DMT data transmission scheme will provide plenty of room for the expansion of service as the demands on the data transmission system increase. Discrete Multi-tone technology has applications in a variety of data transmission environments. For example, the ATIS Asymmetric Digital Subscriber Line North (ADSL) American standard contemplates use of a Discrete Multi-Tone data transmission scheme.

[0040] A detailed description of the protocols for ATIS ADSL North American standard Discrete Multi-Tone (DMT) transmission scheme is described in detail in the above referenced ATIS contribution. The standardized system uses 256 “tones” which are each 4.3125 kHz wide in the forward (downstream) direction. The frequency range of the tones is from zero to 1.104 MHz. The lower 32 tones may also be used for duplexed data transmission in the upstream direction. Improvements in this system which contemplate increasing the transmission bandwidth by as much as an order of magnitude have been proposed in other applications by present invention. In other systems, the number of subchannels and/or the subchannel bandwidth used may be widely varied. However when modulation is done, typical values for the number of available subchannels are integer powers of two, as for example, 128, 256, 512, 1024 or 2048 subchannels.

[0041] As described in the background section of this application, one limitation of discrete multi-tone transmission systems is that in order to support a plurality of drop points serviced by a single line, the upstream signals must be synchronized when they arrive at the central unit. This synchronization problem has limited the attractiveness of Discrete Multi-tone (DMT) data transmission schemes in certain applications such as cable systems and wireless cellular television delivery since these systems use a single line (medium) to service a relatively large number of independent remote units, which would typically be operated by different subscribers.

[0042] Referring initially to FIG. 1, a schematic transmission scheme for a typical multi-user subscriber network will be described. A central unit 10 (which includes a central modem) communicates with a plurality of remote units over a common transmission line 17 which is split into a plurality of feeds 18. Each feed 18 services an associated remote unit which typically includes a remote modem 15 which receives the signals and a remote device 22 which uses the data. A service provider 19 would typically be arranged to provide the data to the central modem for transmission to the remote modems 15 and to handle the data received by the central modem from the remote modems. The service provider 19 can take any suitable form. By way of example, the service provider can take the form of a network server. The network server can take the form of a dedicated computer or a distributed system. A variety of transmission media can be used as the transmission line. By way of example, twisted pair phone lines, coaxial cables, fiber lines and hybrids that incorporate two or more different media all work well. This approach also works well in wireless systems.

[0043] As will be appreciated by those skilled in the art, one requirement of discrete multi-tone data transmission systems such as those contemplated herein is that if two or more units (typically two remote units) are attempting to independently transmit information to a third unit (i.e. the central unit 10), the signals from the remote units must by synchronized or at least some of the signals will be incomprehensible to the central unit 10. The problem with using discrete multi-tone transmissions in such a system is that the length of the feeds 18 will typically vary from remote to remote. Therefore, even if the remotes synchronize with the clock of the central unit 10, their communications back to the central unit 10 will be phase shifted by an amount that is dependent at least in part on the length of the associated feed. In practice, these types of phase shifts can make remotely initiated communications unintelligible to the central modem.

[0044] A representative DMT transmission band is illustrated in FIG. 2. As seen therein, the transmission band includes a multiplicity of sub-channels 23 over which independent carrier signals (referred to as sub-carriers 27) may be transmitted. DMT transmission inherently partitions a transmission medium into a number of sub-channels 23 that each carry data independently. The data on each sub-channel 23 can correspond to a different signal or can be aggregated into higher data rates that represent a single or fewer wider-bandwidth transmissions. These sub-channels 23 are implemented entirely with digital signal processing in DMT, which eliminates the need for analog separation filters and maximizes spectral efficiency. The number of sub-channels used may be widely varied in accordance with the needs of a particular system. However, when modulation is performed using an Inverse Fast Fourier Transform (IFFT), typical values for the number of available sub-channels 23 are integer powers of two, as for example, 128, 256, 512, 1024 or 2048 sub-channels 23. By way of example, in one embodiment that is adapted for use in a cable based subscriber system, 1024 sub-carriers 27 may be used with each carrier confined to a 32 kHz sub-channel 23. This provides approximately 32 MHz of frequency bandwidth in which the remote units can communicate with the central unit 10.

[0045] The number of remote units that may be used in any particular system may vary greatly in accordance with the needs of a particular system. By way of example, in one embodiment of the described cable based subscriber system, it may be desirable to permit up to 500 remote units to communicate with a single central unit. In systems that contemplate such a large number of remote units, it may be desirable to allocate the remote units in groups. Of course, the groups need not each contain the same number of units. By way of example, a system that permits up to 500 remote units may divide the remote units into eight groups, with each group permitting up to 90 remote units, with each remote unit group being assigned a designated frequency band. For example, the frequency spectrum may be divided into a plurality of equally sized designated frequency bands. In the particular embodiment described, one-eighth of the 32 MHz, or approximately four megahertz would be assigned to each group. Therefore, each group would have about 4 MHz, and correspondingly, 128 sub-channels 23 to use for transmitting to the central unit 10. Grouping allows the central unit 10 to keep track of the remote units in a manageable manner as they come on and off line.

[0046] The groupings can be made using any number of methods. By way of example, a first group could consist of consecutive sub-channels 0-127, a second group sub-channels 128-255 and so forth. Alternatively, the allocation of sub-channels 23 to the respective groups may be interleaved throughout the spectrum. For example, the first group may be assigned sub-channels 0, 8, 16, 24, 32 . . . ; the second group may have sub-channels 1, 9, 17, 25, 33 . . . ; the third group: 2, 10, 18, 26, 34 . . . ; and so forth. The interleaving of sub-channels 23 assigned to the groups helps to reduce the probability that noise located in one particular area of the frequency spectrum will corrupt a significant portion of the transmissions in a single group. Instead, the spurious noise will affect only a portion of the spectrum for each group. As can be appreciated by those skilled in the art, the frequency bandwidth of the upstream channel, size of the sub-channels 23 and the groupings are not restricted to the numbers in the described embodiment but can be chosen to suit the needs of the particular use of the transmission system.

[0047] One method of addressing the synchronization problems pointed out above contemplates the use of dedicated overhead subchannels 28 and 29 (of FIG. 2) to facilitate synchronization. In this embodiment, upstream overhead subchannel 28 carries synchronization signals from the various remotes to the central modem. Downstream overhead subchannel 29 carries synchronization signals from the central modem to the various remotes. The overhead subchannels 28 and 29 may be located at any suitable frequency position within the transmission band. In many embodiments such as the asymmetric digital subscriber line system discussed above, it may be desirable to locate the overhead subchannels near either the upper or lower frequency edge of the downstream signal so as to minimize their interference with adjoining subchannels. When the system constraints permit, it may be further desirable to separate the overhead subchannels from other subchannels used for data transmission by at least one or two subchannels in order to minimize potential interference caused by the synchronization signals. This is desirable since the synchronization signals will often be unsynchronized with other transmissions. Therefore, they will cause more distortion than other signals due to being out of synch. Accordingly, a small buffer is helpful. Along the same lines, it may also be desirable to use relatively low powered signals as the overhead subcarriers to further minimize interference issues in some cases.

[0048] As will be described in more detail below, in another aspect of the present invention, synchronized quiet times are periodically provided in the upstream communication stream. The synchronized quiet times may be used to handle a variety of overhead type functions such as initialization of new remote units, transmission channel quality checking and handling data transfer requests. Referring next to FIG. 7, a representative frame delimited transmission timing sequence is illustrated that provides a number of synchronized quiet periods that are suitable for handling the overhead functions. In the embodiment shown, the transmissions are broken up into string of transmission frames 32. Each transmission frame includes a transmission interval 33 and a first quiet interval S1. Each transmission interval 33 is further divided into a plurality of symbol periods 35 as shown. A plurality of transmission frames 32 are then grouped together into a super-frame 36. In addition to the transmission frames 32, each super-frame 36 also includes a second quiet time interval 38. In the embodiment described, the second quiet time interval 38 may be used as either an initialization interval (S2) or a retraining interval (S3).

[0049] The actual periods provided for the transmission interval 33, the quiet time interval S1, the initialization interval S2 and the retraining interval S3 may be widely varied in accordance with the needs of a particular system. Similarly, the number of transmission frames 32 in a super-frame 36 may be widely varied. By way of example, one suitable embodiment for use in the described cable-based subscriber system, contemplates a transmission interval 33 set to a period sufficient to transmit 63 symbols and the S1 time interval 34 set to one symbol in length of time. The initialization interval S2 may be used as an alternative arrangement for synchronizing the remote units. Thus, the length of the second quiet time interval 38 is typically determined by the physical aspects of the communications system, as will be discussed in more detail below. In general, the remote units are required not to broadcast during an S1 or S3 quiet time interval unless given permission by the central unit 10. In some embodiments, the remote units are also required not to broadcast during an S2 quiet time interval unless they are seeking to initiate installation as will be described in more detail below.

[0050] Referring next primarily to FIGS. 2-4, the use of auxiliary overhead subchannels to facilitate synchronization of newly added remotes will be described in more detail. Initially, the remote modem 50 includes a remote synchronization controller 80 that cooperates with a central controller 60 in the central modem unit. As briefly discussed above, in the described embodiment, two auxiliary overhead subchannels are provided to facilitate communications between the controllers. When the remote modem 50 is initialized and desires to come on stream, its remote controller 80 observes downstream signal transmissions that inherently contain the central modem clock information. This is sometimes done by employing pilot signals although other schemes can be employed as well. The remote modem is then “loop-timed”. That is, it phase locks its own clock with the clock of the central modem. The remote controller then sends a synchronization signal to the central unit 30 via overhead subchannel 28. The synchronization signal passes through the transmission media into the receiver portion of central modem unit 30. When the central modem 30 receives a remotely initiated (upstream) synchronization signal while it is currently in communication with other remote units, it compares the frame boundaries of the remotely initiated synchronization signal with the frame boundaries of signals being received from other remote units. Typically, there would be a phase shift between the frame boundaries that is detected by the controller 60. The controller 6O then generates a downstream synchronization signal that is transmitted back to the remote units via overhead subchannel 29.

[0051] In the embodiment described and shown, the controller 80 is responsible for generating the upstream synchronization signal when the remote modem desires to initiate communications with the central modem. The upstream synchronization signal is fed from the controller 80 to the multiplexer/encoder 143 and directed specifically towards upstream overhead subchannel 28. It should be appreciated that since the nature of the synchronization signal is known, it could be introduced to the transmitter at other locations as well or could even be applied directly to the analog interface 148. Typically, the synchronization signals and/or sequence would be the only signals transmitted by the remote until synchronization is complete. The upstream synchronization signal is then transmitted to the central modem via overhead subchannel 28 where it would be received by receiver 70. The receiver's demodulator 76 then feeds the demodulated synchronization signal to the central modem's controller 60. The central controller 60 detects the remotely initiated synchronization signal and compares its frame boundary to the frame boundaries of any signals that are simultaneously being received from other remote units. When the central modem 30 is in communication with other remotes, it is likely that the frame boundaries of the remote requesting access will be phase shifted from the frame boundaries of those that are already in communication with the central modem due to variations in the feed length. In such cases, the central controller 60 initiates a return (downstream) synchronization signal that indicates the phase shift (which takes the form of a time delay) required to align the frame boundaries. The return synchronization signal is then transmitted to the remotes via the second overhead subchannel 29. Like the upstream synchronization signal, the downstream synchronization signal may be introduced to the downstream data stream at the encoder.

[0052] The nature of the downstream synchronization signal may vary, however, by way of example, the synchronization signal may simply indicate that the remote should advance or retard the frame boundary by one sample. In a somewhat more complicated system, the controller can attempt to calculate the number of samples that the frame boundary must be advanced or retarded and a signal that dictates the number of samples that the frame boundary should be shifted can be sent. Other signal interpretations can be used as well. As will be discussed in more detail below, in many embodiments, the sample rate for upstream communications will be an integer factor of the sample rate of the downstream communications. The described delay is based on the sample rate of the central modem, as opposed to the remote.

[0053] Since a plurality of remotes are all connected to the same transmission line 17, the synchronization signal will be received by all the operating remote modems. The signal is then passed from each remote modem's decoder to their associated controller 80. However, the remote controllers 80 are arranged to ignore synchronization signals on the overhead subchannel unless they are currently trying to initiate communications with the central modem. This can be accomplished in a variety of ways. By way of example, the downstream synchronization signals may include an address directed at a specific remote. Alternatively, the remotes can simply assume that the central modem signal is directed at them if they are currently attempting to initiate communications. The remote controller 80 of the remote unit that is attempting to initiate communications receives and interprets the centrally initiated synchronization signal and instructs the frame synchronizer 147 to implement the requested phase shift timing delay (or advance). A second remotely initiated synchronization signal would then be sent. If the new synchronization signal is not in synch, the same process will be repeated. In one embodiment, the synchronization signal would merely instruct the frame synchronizer to advance or retard by one sample. It is contemplated that in most applications of DMT, such an incremental system will work well to quickly synchronize the remote unit. By way of example, in a system that has a symbol (frame) rate of 8 kHz (and thus a symbol period of 125 μs) which corresponds to 64 Kbps, with each frame having 128 samples plus a prefix, in distribution networks having feed length variations of as much as two miles, it would still take less than approximately ten milliseconds to synchronize using a simple single sample advance/retard approach.

[0054] When a remotely initiated signal is determined to be in synch, then the central controller would send a return synchronization signal over the second overhead subchannel 29 indicating that no further phase shifting is required and that the remote unit may initiate full communications with the central modem incorporating the desired phase shifting. When the remote is synchronized before it is recognized by the central modem, the data tones transmitted just after initialization are used to identify the remote modem. It is expected that the relative phase shifting of frame boundaries is primarily dependent on fixed constraints such as the transmission length through the various feeds. Therefore, once a remote is synchronized, it does not need to be resynchronized unless the connection is terminated or broken.

[0055] It should be appreciated that when the central unit is not in communication with any other remote units at the time it receives a request to initiate communications, the central controller 60 would merely send back a synchronization signal indicating that no phase shifting was required and that full communications may begin. A similar signal would, of course, also be generated in the event that the requesting remote happens to be in synch with the other remote modems when it first attempts to initiate communications. When the remote modern receives such a signal, the same process may be followed with the required phase shift simply being zero.

[0056] Typically, the central controller 60 would also provide information indicative of the subchannels that the remote unit should utilize for its transmissions, etc. As mentioned above, the subchannel allocation can be dynamically changed during use. Although this feature is important to the discrete multi-tone transmission scheme is not particularly germane to the present invention and therefore will not be described only briefly, although it is described in detail in the cited references.

[0057] Synchronization of a remote modem to the central modem requires the acquisition of the central modem's sampling clock and carrier. In one preferred embodiment, these clocks are recovered by inspecting the phase errors for at least two tones. The phase error for these tones can be computed with respect to a fixed known transmitted phase on the tones (i.e. “pilot” tones). Alternatively, they may be determined by assuming decisions on the transmitted phases are correct and computing the offset between the pre- and post-decision phases (i.e., decision-aided phase-error computation). The slope of the phase error plot; as illustrated in FIG. 6, is proportional to the timing-phase error, while the constant part (the y-intercept) of the phase-error plot is the carrier-phase error. The timing (sampling) phase error and the carrier-phase error are determined by phase detector 181 and input to phase-lock loops 182, 184 that synthesize a sampling clock and carrier frequency at the recovered central modem frequencies as illustrated in FIG. 5. The carrier is used to demodulate the downstream signal to baseband and the sampling clock is (after division by divider 189) used to clock the analog-to-digital converter(s) (ADC). If the data tones and the signal tones occupy separated tones, then more then one analog-to-digital converter at slower sampling clocks may be used in place of a single higher-speed ADC clock. In embodiments that include the notch filter 185, voltage controlled oscillators 183, 186 are provided to control the location of the notch.

[0058] The same sampling clock (after division by divider 189) is used for upstream digital to analog converters. The upstream carrier may be synchronized to the downstream carrier or may not be so synchronized. When it is not synchronized, the central modem's upstream receiver will need to recover the upstream transmission carrier phase, otherwise the central modem's upstream receiver can use a rational phase-locked multiple of the downstream carrier for data recovery. Wideband remote modems would preferably use a sampling clock that is the same as the sampling clock in the central modem. These remote modems will not divide the recovered sampling clock. Narrow band remote modems that receive only a few tones will use a sample clock that is an integer divisor of the recovered sampling clock. Accordingly, narrow band remote modems can be less costly to implement.

[0059] The DMT symbols transmitted upstream from the remote modems must arrive at the central modem at the same time as discussed above, even when they are generated by different remote modems. Therefore, the delay synchronizer 147 inserts an integer number of sample-clocks delay into the upstream transmitted signals. This delay is programmed under control of the downstream synchronization signal as previously discussed. Again, it should be appreciated that the delay is based on the sample rate of the central modem, as opposed to the remote. Specifically, as illustrated in FIG. 5, the sample rate of the remote may be an integer factor of the sample rate of the central. However, the signals must be synchronized at the central modem and therefore, the synchronization adjustments must be made on the basis of the central modem's sample rate.

[0060] In the event that two remotes simultaneously attempt to initiate communications with the central modem, a conflict will occur and the central controller 60 will likely be confused by the upstream sychronization signals. In such a case, its downstream synchronization signal would indicate an improper phase shift and the confirmation synchronization signals would not be properly synchronized. In one embodiment, the central controller 60 could recognize the problem and instruct the remote units to stop and attempt to establish communications at a later point. In another embodiment, the central controller could simply send another downstream synchronization signal indicative of the additional phase shifting that is required. In either event, the remote unit will quickly recognize that a problem exists and assume that a conflict is occurring. In this situation, a suitable conflict resolution scheme can be employed. One simple conflict resolution scheme is simply to have each remote delay for a random amount of time and attempt to reinitiate communications after the random delay. As long as the delay is determined in a manner that the remotes are not likely to consistently follow the same delay pattern, their requests will eventually be separated sufficiently such that each can be brought on line independently. A variety of wait-time distributions may be utilized. By way of example, a Poisson distribution has been found to work well.

[0061] It should be appreciated that the described IFFT modulation scheme works extremely well for systems that are arranged to transmit relatively large chunks of data and therefore require more than a handful of tones. However, in many situations, the remotes may not need to transmit large blocks of data regularly. In such situations, it may be cost effective to utilize a simpler conventional modulation scheme for transmitting information from the remotes to the central unit. In such circumstances, the remote transmitter and the central receiver would both be replaced with the appropriate components. However, there would still be a need to synchronize the remotes as discussed above.

[0062] In operation, the central modem transmits an aggregate DMT signal that uses all (or the usable) tones in a manner such that each remote knows the tones that it is to receive and the number of bits allocated on each of its received tones. The remotes modems, in turn each use only a subset of the available upstream tones. The signals transmitted from the central modem to the remotes may be used to dynamically allocate the tones available to a particular receiver. Alternatively, in a static system, the allocation could be made in the downstream synchronization signal. Dynamic allocation can take place on either another dedicated overhead or control channel or may be multiplexed with other non-control signals. In the described system, the upstream signals are timed so that they arrive at the central modem at substantially the same time. Precise alignment is not necessary; however, the system works best when the boundaries are closely aligned in terms of the sample rate of the central modem.

[0063] Referring next to FIG. 8, an alternative method of initializing a first remote unit during installation that utilizes the described second quiet times S2 in accordance with another aspect of the invention will be described. As discussed above, when a remote unit first comes on line, it must be initialized such that the transmissions from the first remote unit arriving at the central modem are synchronized with the transmissions of any other currently installed remote units. That is, the frame boundaries of upstream DMT communications from the various remote units to the central unit must be substantially synchronized at the central unit for the transmissions to be understood by the central unit. The method described with reference to FIG. 8 is one method of accomplishing such synchronization utilizing the described quiet times.

[0064] Initially, the remote unit to be installed must establish a connection to the transmission network in step 302. The connection enables the remote unit to listen to the downstream transmissions from the central unit 10 and transmit on any unused sub-channel 23 of the upstream channel. In some systems, there may be certain frequency ranges that the system may not use. By way of example, in many cable networks there may be established systems that utilize specific frequency bands. In order to prevent interference and maintain backward compatibility, it is important that the remote unit never transmit in the forbidden frequency range, even during initialization. Of course, certain frequency bands may be forbidden for other reasons as well. Accordingly, in step 303, the central unit will periodically broadcast an identification of frequencies that may never be used. In systems that utilize the concept of remote unit groups as discussed above, the central unit may also periodically broadcast the group number of the group that should be used by the next remote unit to be installed. Alternatively, the group assignment can be handled at a later point.

[0065] The newly connected remote unit listens to the downstream signals for information indicating that certain sub-channels may not be used. The downstream signal also includes the frame timing and quiet period markers required to synchronize the remote unit with the central unit. After the remote unit has synchronized itself with the downstream signal, in step 304 it transmits an initialization signal at the beginning of an S2 quiet period. In one system, this is done by transmitting an initialization signal immediately upon receiving an S2 quiet period marker signal. The initialization signal indicates to the central unit 10 that a remote unit requests to be installed onto the system. The remote unit may determine the onset of an S2 initialization quiet period in any suitable manner. By way of example, a flag may be provided by the central unit 10 in the downstream communications. The remote unit may transmit its initialization signal over all the sub-channels 23, over a group of sub-channels 23 or on a single sub-channel 23 depending on the needs of a particular system. In a preferred embodiment, the downstream signal indicates the group to be used by the next unit to be installed, and the initialization signal is transmitted over all the sub-channels in that group.

[0066] The upstream initialization transmissions from the remote units to the central unit 10 can be accomplished in any modulation scheme suitable for transmitting digital information. By way of example, amplitude, frequency, and quadrature phase shift key (QPSK) modulation schemes can be utilized. For the synchronization signal, differential QPSK (DQPSK) modulation is desired in a preferred embodiment to decrease the possibility of corruption by noise. Additionally, the synchronization can be encoded with a large amount of error correction and redundancy to ensure coherent communications.

[0067] The initialization signal preferably contains information about the remote unit. In a preferred embodiment the initialization signal carries the global address of the remote unit and the maximum transmission data rate requirement of the first remote unit. A global address is similar to addresses used on ethernet or cellular devices. Such addresses are built into the communications device and are distinct from addresses of all other communicating devices. The maximum data rate required by the remote unit is dependent upon the type of device the remote unit is. For example if the remote unit is a television set it would require minimal communications capacity to the central unit 10, possibly only using the upstream signals to send information about movie selections or viewer feedback. On the other hand, if the remote unit is a teleconferencing transceiver then a large amount of bandwidth would be required to transmit video and audio information from the remote unit to the central unit 10. Other pieces of relevant information about the first remote unit can also be sent along with the initialization signal in other embodiments.

[0068] Upon receiving the initialization signal from the first remote unit, the central unit 10 determines in step 306 whether the initialization signal from the first remote unit has collided with another initialization signal from another remote unit trying to connect at the same time. If a collision is detected then the central unit 10 transmits a collision message back to the remotes in step 308. The collision message indicates to the remote units trying to connect to try again. The colliding remote units then each wait a random number of S2 periods before re-sending an initialization signal. The probability of two remote units trying to initialize at the same time is small. By requiring the colliding units to wait random amounts of time that are independent of each other, the probability of repeat collisions is reduced even further.

[0069] After the central unit 10 receives a valid initialization signal from the first remote unit, the central unit 10 transmits a synchronization signal 310 back to the remote unit. In one embodiment, the synchronization signal includes the global address of the first remote unit, a nodal address assigned to the first remote address, delay correction information, and information about the allocation of the sub-channels 23 in the upstream channel. Either the global address and the nodal address can serve as a unique remote unit identifier, albeit with differing degrees of transmission efficiency. The global address allows the first remote unit to identify that the synchronization signal is intended for it. The nodal address is assigned to the first remote unit in order to facilitate efficient future communications. The global address can be quite long (as for example 48 bits) to allow for an adequate number of global addresses for all the communicating devices that are likely to be manufactured. The nodal address is a shorter address since only a limited number of remote units will be communicating with any single central unit 10. When a multi-grouped system is used, the nodal address also contains group identifier information, e.g. information about the group to which the first remote unit is assigned. In the embodiment described above which contemplates a total of eight groups, that part of the address would be three bits to identify which of the eight groups the first remote unit is in. The remainder of the bits can uniquely identify the node, e.g. the specific remote unit, within its group.

[0070] It should be appreciated by those skilled in the art that the part of the nodal address that specifies the group, i.e. the group identifier information, may be omitted altogether when a remote unit needs to uniquely identify itself to the central unit. This is because the central unit may, by inspecting the frequency band of the unique identifier message, determine the group from which the remote unit's message is sent. In this manner, a remote unit needs to send only the bit pattern in the nodal address that identifies itself in the group, i.e. the unique intra-group identifier information, in order to uniquely identify itself to the central unit. This received intra-group identifier bit pattern, in combination with the ascertained group identifier information, provides the central unit with the complete nodal address of the requesting remote unit. In the preferred embodiment which has 128 sub-channels per group, the unique remote identifier information may be as short as 7 bits in the upstream direction.

[0071] The delay correction information tells the first remote unit how much the frames being broadcast from the first remote unit must be delayed in order to synchronize them with signals from the other connected remote units. The delay correction is determined from the amount of delay that the central unit detects between the time it transmits a quiet period (S2) marker and its reception of the initialization signal. By way of example, if the maximum delay in the channel is T_RT(Max), e.g. maximum round-trip delay, and the delay associated with a given remote unit is T_RT(i), the delay correction for that remote unit is T_RT(Max)−TRT(i). The round-trip delay for a remote unit is defined as the time taken for a signal to travel from the central unit to that remote unit, and an immediate response to be returned to the central unit, including any minimal, incidental delay attributable to processing. Using this information the first remote unit can adjust its transmissions and become synchronized with the other connected remote units, such that the frames of the remote units arrive at the central unit 10 at the same time. The first remote unit may also learn which sub-channels 23 are currently in use by the other connected remote units. In another embodiment, information about sub-channel 23 characteristics are regularly transmitted to all the remote units through the downstream channel. In such systems, channel usage information would not be required to be sent along with the synchronization signal.

[0072] One advantage of transmitting the initialization signals over a broad portion of the available spectrum is that delays may vary to some extend depending upon the frequency at which the signal is transmitted. Therefore, when the initialization signals are transmitted over a variety of the sub-channels 23 the required phase shift can be calculated based on an average of the individual delays.

[0073] The length of the S2 time interval, as discussed earlier, is dependent upon the physical nature of the communications network. In a preferred embodiment the S2 time interval need only be longer than the duration of the initialization signal plus the difference between the maximum and minimum round-trip delays for the network. By way of example, in a typical system employing a fiber optic trunk as the transmission line 17 and coaxial cables as the feeds 18, the fiber trunk is common to all paths between central and remote units, and the difference between the maximum and minimum round-trip delays for the network depends only on the cable part of the network. Using the length of 2 miles for the coaxial line and given its propagation time of approximately 7.5 microseconds per mile, the maximum and minimum round-trip delays would be approximately 32 and 2 microseconds. In a preferred embodiment a symbol is approximately 30 microseconds long, and an initialization signal would comprise two symbols, so that, by way of example, an S2 time interval of 4 symbols would be appropriate.

[0074] In certain embodiments, it may be desirable to repeat steps 304-310 to validate the information received and/or ensure that the remote is properly synchronized.

[0075] After synchronization has been accomplished, the first remote unit responds by sending a set of synchronized wide band training signals over all the sub-channels 23 during the next available S2 or S3 time interval in step 312. The specifics of the training step will be described in more detail below with reference to FIG. 9. In some embodiments, the central unit 10 will direct the first remote unit to use a specified S3 time interval (e.g., wait for the third S3). Upon receipt of the training signals, the central unit 10 determines the capacities of the various sub-channels 23 to handle transmission between the first remote unit and the central unit 10 (step 314). The central unit 10 preferably has a prior knowledge of the contents of the training signals. This allows the central unit 10 to learn the optimal equalization of the sub-channels 23 and also the maximum bit rates a sub-carrier 27 can carry on the subchannels 23 between the first remote unit and the central unit 10. The central unit 10 saves the channel characteristics of the subchannels 23 with respect to the first remote unit 316. In a preferred embodiment the central unit 10 saves the information in a bits/carrier matrix that contains an indication of the number of bits that each of the sub-channels 23 can carry from each of the remote units. Such a matrix allows the central unit 10 to keep track of the capacity of each of the various sub-channels 23 and is available when allocating bandwidth to the remote units. This also facilitates the dynamic allocation of sub-channels based upon the current characteristics of the transmission environment.

[0076] Referring next to FIG. 9, a method of periodically checking the capacity of the various sub-channels from a selected remote unit to the central unit will be described. As will be appreciated by those skilled in the art, the capacity of the transmission line at various frequencies may vary somewhat over time. Therefore, it is desirable to periodically update the central unit's information concerning the characteristics of the sub-channels 23 with respect to each of the remote units it services. In the embodiment described, such updating is done during the S3 quiet periods. In the embodiment shown, the S3 quiet periods are of the same length as the S2 quiet periods. It should be appreciated that a single transmission line checking process may be used both for the initial training and the periodic checking.

[0077] In the described embodiment, the central unit 10 initiates a retraining event in step 330 by transmitting a retraining command to a first remote unit (remote unit x) that is in current communication with the central unit 10. The first remote unit waits for the next available S3 retraining quiet time interval to transmit a set of training signals over the available sub-channels 23. (Step 332). In an alternative embodiment, the central unit 10 may assign a specific S3 quiet interval to use for transmitting the training signals, instead of the next available S3 time interval. The set of training signals will typically be limited to the sub-channels allocated to the group and will typically be further limited to some subset of the total available group sub-channels to provide a cost effective design. Therefore, the number of training signals that are actually used may be widely varied in accordance with the needs of a particular system. As in the initialization process, the central unit 10 analyzes the signals it receives and updates the bit/carrier rates in the channel characteristics matrix that correspond to the associated remote unit. (Step 334). The central unit 10 then determines whether a change in the sub-channel allocation is necessary for the remote unit. That is, it may determine whether additional or fewer sub-channels 23 should be allocated to the first remote unit in order to meet the first remote unit's throughput and error probability requirements. If a change is necessary, then the central unit 10 re-allocates sub-channels 23 to the first remote unit in step 338.

[0078] If it is determined that no correction is required in step 336 or after any necessary changes have been made in step 338, the central unit 10 checks to see if there have been any requests made by any other remote units for an immediate retraining in step 340. If it is determined in step 340 that there are no immediate retraining requests, the central unit 10 checks to see if the retraining of the first remote unit was a result of a immediate retraining request by checking if there is a valid old address (oldx) in step 347. If there is no valid old address then the central unit 10 increments the counter (x) in step 349 and returns to step 330 where it broadcasts a retrain signal to the next remote unit. On the other hand, if it is determined in step 340 that there was a valid old address, the central unit 10 will adjust the counter such that it reads one more than the old address, which corresponds to the address of the remote unit that would have been next at the time an immediate retrain request was received. (Step 350). That is, x=oldx+1.

[0079] If an immediate retrain request was detected in step 340, then the central unit 10 saves the address of the first remote unit as an old address (oldx) in step 342