Combined signalling and PCM cross-connect and packet engine

MICROFICHE APPENDIX
A microfiche appendix containing a single microfiche with twenty-nine (29) frames is submitted as part of the application and is incorporated by reference herein.
FIELD OF THE INVENTION
The present invention pertains to the field of communication systems and, more particularly, to cross-connect circuitry for use in telecommunication networks.
BACKGROUND
In most modern telecommunication networks, a community of subscribers are connected to a central office switch through a two-way distributed network, which may include one or more transmission facilities, e.g., microwave, optical, electrical, etc., and which may utilize both digital and analog transmission protocols. In particular, between the central office switch and a respective subscriber termination, e.g., a remotely located subscriber line card, telecommunication signals are commonly digitized and multiplexed for transport over relatively high bandwidth shared transmission facilities, e.g., optical fiber, for greater network efficiency.
By way of example, in an exemplary optical communication network, an optical line terminal ("OLT") located at a central office transmits digital telecommunications data signals--i.e., pulse-code-modulated ("PCM") voice or data--between a co-located switch and a number of remotely located optical network units ("ONUs"), wherein the data is transmitted optically between the OLT and the respective ONUs over optical fibers, and electrically between the OLT and the switch. More particularly, the OLT terminates respective "bus side," or "switch side," communication lines (e.g., E1 or T1 lines) on one end and "fiber side" optical fibers on another. In the downstream direction, the OLT cross-connects PCM data contained in byte-size time-division-multiplexed bus side channels into designated fiber side channels, which are then packetized and converted from electrical to optical signals for transmission to respective ONUs. Similarly, in the upstream direction, the OLT converts receives optical signals to electrical signals, disassembling the incoming data packets into fiber side channels, which are then cross-connected into respective assigned switch channels for transmission to the switch.
Cross-connecting the respective switch and fiber channels is conventionally performed with a timeslot interchange switch circuit, wherein up to several thousand channels may be cross-connected in a single interchange system. For example, in a known cross-connect system, byte-size time-division-multiplexed switch channels transported on an eight-bit wide bus are input/output on one side of a timeslot interchange cross-connect system, and byte-size time-division-multiplexed fiber channels transported in serial bit streams are output/input at the other side, wherein the transmission of data channels on both sides of the cross-connect system is repeated at a data frame rate of eight KHz, i.e., with eight bits/channel transmitted every 125 microseconds in each direction, for an overall data rate of 64,000 bits per second (i.e., 64 kbps), per channel.
To perform the cross-connection of the respective PCM data channels, the timeslot interchange circuit must temporarily store the respective bytes of data for each channel in a unique location in an associated cross-connect memory, which is duplicated in both the upstream and downstream directions. For example, in the downstream direction, this has been done by addressing a left port of a dual-port memory with a timeslot counter for the switch channel side, and by addressing a right port of the same dual-port memory with the data from a control store memory containing the bus side channel number to be cross-connected to each fiber timeslot for the fiber channel side. The data buses of the control store memory are connected directly to byte-wide registers that receive and transmit data on the bus side from and to a byte-wide bus, and on the fiber side to parallel/serial and serial/parallel converters connected to opto-electrical circuitry, respectively. The control store memory is addressed by a fiber side timeslot counter. Data for each channel is stored in the cross-connect memory at an address corresponding to its bus side timeslot number. The control store memory is programmed through a second port by a microprocessor, with each address containing the bus side timeslot number of the channel to be cross-connected to the fiber timeslot having the same number as the address.
Higher capacity interfaces, such as Integrated Services Digital Network (ISDN) channels, can be carried over multiple 64 kbps channels internal to the network. To maintain the byte sequencing of these higher rate interfaces, timeslot interchange memories are usually organized into two banks that are alternately read from, and written to, respectively, by the switch and fiber sides of each frame. Thus, for a first frame in the downstream direction, the bus side writes to bank "zero," while the fiber side reads from bank "one." In the next frame, the bus side writes to bank one while the fiber side reads the data in bank zero, i.e., that was just written by the bus side, albeit in a different order. In this manner, all the data contained in a single frame on the bus side is carried in the next frame on the fiber side, and byte sequencing is maintained when multiple internal channels are reassembled into a higher rate interface to a respective ONU.
The cross-connection of signalling information associated with individual switch and fiber PCM data channels is complicated by the variations in multiframe signalling protocol used throughout the world. For example, under the applicable international (i.e., "E1") telecommunication transmission standard, signalling information is carried over a "signalling multiframe" comprising sixteen data frames, wherein each E1 data frame comprises thirty-two byte-size data channels transmitted at the basic telephony rate of 8 KHz (i.e., every 125 usecs), for an overall data rate of 2.048 mega bits per second (mbps). In particular, each E1 data frame contains a framing channel followed by fifteen subscriber PCM channels, a signalling channel, and another fifteen subscriber PCM channels, respectively, wherein each signalling channel carries four bits of signalling information each for two of the respective subscriber PCM channels (i.e., with the first signalling channel of each multiframe unused), such that, over the course of a sixteen frame signalling multiframe, each of the thirty PCM data channels has four signalling bits transmitted to go with sixteen bytes of PCM data.
On the other hand, under the applicable U.S. (i.e., "T1") standard, signalling multiframes comprising twenty-four T1 data frames are used, wherein each T1 data frame comprises twenty-four byte-size subscriber PCM channels transmitted, along with a single framing bit, every 125 usecs for an overall data rate of 1.544 mbps. Unlike the E1 standard, a "bit robbing" methodology is employed to carry signalling information in a T1 frame. In particular, each subscriber PCM channel carries one bit of its own signalling information in its least significant bit location every sixth frame. In this manner, over the course of a T1 signalling multiframe, each of the twenty-four subscriber PCM channels has four signalling bits transmitted to go with the twenty-four bytes of PCM data.
More recently, some signalling cross-connect systems have expanded on the above-described timeslot interchange circuit, by enlarging the cross-connect control store memory to contain sixteen locations for use in communication systems operating under the E1 standard (or twenty-four locations for communication systems operating under the T1 standard) for each fiber side timeslot, so that different fiber and switch timeslots can be cross-connected in each frame of a signalling multiframe. This configuration allows for signalling information for each of the fifteen (or twenty-four) subscriber channels associated with each bus side signalling channel to be cross-connected to the signalling information location associated with any of the fifteen (or twenty-four) subscriber channels in any of the fiber side signalling channels.
However, to ensure that the signalling information is not overwritten by another channel before it is "read out" by the respective other side of the cross-connect circuit, the signalling information associated with each subscriber channel must be stored at a unique address, which requires that the cross-connect memories be fifteen times (for E1) or twenty-four times (for T1) as large as the memory of a "simple" timeslot interchange circuit described above. In addition, the control store memory must also have fifteen (or twenty-four) times as many addresses with five more bits per location, since five is the first integer greater than the base-two logarithm of twenty-four. The larger control store also requires more complex firmware to write to fifteen (or twenty-four) locations for each subscriber channel and other locations for the associated signalling channels.
In addition to cross-connection of subscriber channels, the OLT must also perform the assembly and disassembly of data packets transmitted to, and received from, the respective ONUs. Conventionally, this packet assembly/disassembly ("PAD") function is performed by a "packet engine" that is separate from the cross-connect circuitry, wherein large memory buffers are required to accommodate the different data transmission rates on the respective switch and fiber sides of the OLT. A further control store is also needed to be able to control the PAD function for each fiber timeslot, e.g., wherein each further control store location contains a word that specifies whether the particular timeslot contains an overhead data pattern (e.g., for clock recovery or optical level acquisition), an OLT/ONU communications channel, or a subscriber PCM data channel from the bus side backplane bus.
Thus, there is a need for a system that combines cross-connect and packet engine circuitry, stores signalling information efficiently, and significantly reduces both the complexity and the memory requirements of the respective system hardware and firmware over the known art.
SUMMARY OF THE INVENTION
The present invention provides systems and apparatus that combine cross-connect and packet engine circuitry in a manner which improves efficiency of transient storage of signalling information channels, thereby significantly reducing cross-connect memory requirements, as well as reducing hardware and firmware complexity over the known art.
As deployed in a preferred OLT of an optical telecommunications network, a plurality of combined PCM and signalling channel cross-connect and packet assembly/disassembly ("XCONN/PAD") circuits each receive downstream PCM and signalling channels from a common eight-bit wide backplane bus, which are transmitted in successive 1536 channel frames at the standard telecommunications frame rate of 125 usec, i.e., for a bus side data rate of 12.288 mega bytes per second (MBps). The OLT may be configured to operate under either the E1 or T1 standard. When operating under the E1 standard, each 1536 channel bus side frame supports PCM and signalling channels for up to thirty-two E1 lines. When operating under the T1 standard, each 1536 channel bus side frame supports PCM and signalling channels for up to forty-eight T1 lines. In alternate preferred embodiments, multiple backplane busses may be used to increase system capacity.
In conjunction with a plurality of externally disposed memories, each XCONN/PAD circuit cross-connects the downstream "bus side" PCM and signalling channels into fiber side PCM and signalling channels, which are output in 960 channel frames over a downstream serial interface, i.e., for a fiber side data rate of 7.68 MBps. In particular, each XCONN/PAD circuit is provided with a two-page downstream PCM channel memory that cross-connects downstream PCM channels and a separate downstream signalling channel memory that cross-connects associated downstream signalling channels between the downstream backplane bus and fiber side serial interface.
In accordance with a more specific aspect of the invention, the respective downstream PCM and signalling channel memories are configured to be substantially the same size, wherein each downstream signalling channel is mapped to an address in the downstream signalling memory which corresponds to the address of its associated PCM channel in the downstream PCM memory. The downstream signalling channels are stored in the downstream signalling channel memory over a first signalling multiframe period and then read out into respective fiber side frames during the ensuing signalling multiframe period. Packetizing of downstream data frames for transmission to respective downstream ONUs is thereby easily accomplished because overhead channels are inserted in the proper fiber timeslots and PCM and signalling channels are cross-connected at the appropriate timeslots, thus avoiding the need for a separate control store and large FIFO for packet assembly. In other words, "real time" assembly of the fiber side PCM and signalling channel frames is accomplished.
Similarly, each XCONN/PAD circuit disassembles received upstream packets into successive 960 byte frames of received upstream packets into upstream fiber side PCM and signalling channels, which are transmitted over an upstream serial interface at a data rate of 7.68 MBps. In a reciprocal manner as the downstream data channels, the upstream fiber side PCM and signalling channels are cross-connected into respective bus side channels and multiplexed onto a common eight-bit wide upstream backplane bus at a rate of 12.288 MBps. More particularly, each XCONN/PAD circuit is also provided with a two-page upstream PCM channel memory that cross-connects the upstream PCM channels with a maximum delay of two frames and a separate upstream signalling channel memory that cross-connects associated upstream signalling channels with a maximum delay of two respective signalling multiframes between the respective upstream serial interface and backplane bus.
As with the downstream memories, the respective upstream PCM and signalling channel memories are configured to be substantially the same size, wherein each upstream signalling channel is mapped to an address in the upstream signalling memory which corresponds to the address of its associated PCM channel in the upstream PCM memory. The upstream signalling channels are stored in the upstream signalling channel memory over a first signalling multiframe period and then read out into respective 1536 channel bus side frames during the ensuing signalling multiframe period.
Advantageously, upstream and downstream "expansion" buses can be employed on the bus side, in conjunction with expansion, or duplicate, upstream and downstream memories, to effectively double system capacity. In an exemplary embodiment, the expansion buses are implemented as eight-bit-wide buses, similar to the primary buses. The fiber side cross-connect address control store memory preferably includes an extra bit that can be used to choose which memory is employed and to indicate whether upstream data is to be taken from the primary bus or the expansion bus.
Accordingly, it is a general object of the present invention to provide systems and apparatus that simultaneously perform the functions of PCM and signalling channel cross-connect and packet assembly/disassembly, with minimal memory and firmware requirements.
Other and further objects, features, aspects, and advantages of the present invention will become better understood with the following detailed description of the accompanying drawings.

BRIEF DESCRIPTION OF THE DRAWINGS
The drawings illustrate both the design and utility of preferred embodiments of the present invention, in which:
FIG. 1 is a simplified block diagram of an exemplary optical communication network employing a preferred OLT for transmitting time-division-multiplexed PCM and signalling channels between a central office switch and a number of remotely located ONUs;
FIG. 2 is a simplified functional block diagram of a preferred fiber interface module ("FIM") circuit board employed in the OLT of FIG. 1;
FIG. 3 is a simplified functional block diagram of the downstream data flow through a preferred cross-connect and packet assembly/disassembly ("XCONN/PAD") circuit employed in the FIM circuit board of FIG. 2;
FIG. 4 is a more detailed functional block diagram of a preferred XCONN/PAD circuit employed in the FIM circuit board of FIG. 2;
FIG. 5 is a block diagram of a preferred bus side cross-connect ("BUSX") module for use in the preferred XCONN/PAD circuit of FIG. 4;
FIG. 6 is a block diagram of a preferred fiber side cross-connect ("FIBX") module for use in the XCONN/PAD circuit of FIG. 4;
FIG. 7 is a block diagram of a preferred fiber front end ("FIBFE") module for use in the XCONN/PAD circuit of FIG. 4;
FIGS. 8A-D are tables describing the mapping of multiple dual port memories into external single port memories that are accessed and controlled by a preferred external memory access module ("XMAM") in the XCONN/PAD circuit of FIG. 4;
FIG. 9 is a table describing internal memories of the XMAM of the XCONN/PAD circuit of FIG. 4;
FIG. 10 is a table showing exemplary timeslot assignments for bus side PCM data and signalling channels in a preferred OLT when operating under the E1 standard; and
FIG. 11 is a table showing exemplary timeslot assignments for bus side PCM data and signalling channels in a preferred OLT when operating under the T1 standard.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
Referring to FIG. 1, an exemplary telecommunications network 20 includes a central office 21 housing a switch 26 and an OLT 22. A plurality of E1 or T1 transmission lines 31 transport time-division-multiplexed PCM and signalling data channels between the switch 26 and the OLT 22. In particular, the transmission lines 31 may be E1 lines if the OLT 22 is operating under the E1 standard (e.g., as set forth in CCITT 802.4), or T1 lines if the OLT 22 is operating under the applicable T1 standards. Within the OLT 22, the E1/T1 lines 31 are terminated on a plurality of respective E1 or T1 module ("ETM") circuits 28. In alternate preferred embodiments, some of the ETMs 28 may be replaced with other types of communication interface circuit boards, such as, e.g., Sonet interface circuit boards.
In the downstream direction, the ETMs 28 reformat the respective T1 or E1 lines into successive frames containing 1536 byte-size subscriber PCM and signalling channels, which are then multiplexed onto an eight-bit wide downstream data bus 33 (i.e., for a "bus side" data rate of 12.288 MBps). In particular, the ETMs 28 reformat the downstream signalling data into separate respective signalling channels associated with a single subscriber channel, wherein a single subscriber signalling channel containing four bits of signalling information is transmitted every signalling multiframe--i.e., every sixteen frames when operating under the E1 standard or every twenty-four frames when operating under the T1 standard.
In a presently preferred embodiment, when operating under the E1 standard, each downstream frame on bus 33 delivers PCM and signalling channels for up to thirty-two E1 lines, wherein the individual subscriber channels for each respective E1 are spread over forty-eight byte-size "timeslots." A table showing exemplary timeslot assignments for the PCM data and signalling channels when operating under the E1 standard is shown in FIG. 10. When operating under the T1 standard, each 1536 byte bus side frame delivers PCM and signalling channels for up to forty-eight T1 lines, wherein the individual subscriber channels for each respective T1 are spread over thirty-two timeslots. A table showing exemplary timeslot assignments for the PCM data and signalling channels when operating under the T1 standard is shown in FIG. 11.
Referring to the timeslot assignment table shown in FIG. 10, when operating under the E1 standard, each downstream frame is multiplexed from the ETMs 28 onto downstream bus 33 in forty-eight successive "rows" 0-47, each row containing thirty-two timeslots corresponding to E1 lines 1-32 ("E1-E32")--i.e., where row 0 contains timeslot nos. 0-31; row 1 contains timeslot nos. 32-63, and so on, through row 47, which contains timeslot nos. 1504-1535. In the exemplary timeslot assignments shown in FIG. 10, rows 0-3, 7, 11, 15, 19, 23, 25, 27, 31, 35, 39, and 43 are inactive; row 4 is used for test purposes; rows 5-6, 8-10, 12-14, 16-18, 20-22, 24, 26, 28-30, 32-34,, 36-38, 40-42, and 44-45 are assigned to subscriber PCM channels 1-15 and 17-31 of each respective E1 line; row 46 is used for the signalling channel corresponding to subscriber PCM channels 1-15; and row 47 is used for the signalling channel corresponding to subscriber PCM channels 17-31. Thus, the data in the signalling channels changes progressively with each frame for the duration of the signalling multiframe, such that, e.g., in frame 0, a signalling multiframe patern is carried in row 46; and in frame 1, signalling data for channel 1 is carried in row 46; etc. It is to be understood that these are merely exemplary timeslot assignments.
Referring to the timeslot assignment table shown in FIG. 11, when operating under the T1 standard, each downstream frame is multiplexed from the ETMs 28 onto downstream bus 33 in thirty-two successive "rows" 0-31, each row containing forty-eight timeslots corresponding to T1 lines 1-48 ("T1-T48")--i.e., where row 0 contains timeslot nos. 0-47; row 1 contains timeslot nos. 48-95, and so on, through row 31, which contains timeslot nos. 1488-1535. In the exemplary timeslot assignments shown in FIG. 11, rows 3, 7, 11, 15, 19, 23, and 27 are inactive; rows 0-2, 4-6, 8-10, 12-14, 16-18, 20-22, 24-26, and 28-30 are used for subscriber PCM channels 0-23 of each respective T1 line; and row 31 is used for the signalling channel corresponding to subscriber PCM channels 1-48. It is to be understood that these are merely exemplary timeslot assignments. As will be appreciated by those of ordinary skill in the art, timeslots labeled "inactive" for purposes of the above example are merely inactive vis a vis E1 (or T1) PCM subscriber information. Other information can readily be transmitted in channels labeled "inactive" for purposes of the above example.
From the ETM boards 28, the respective downstream PCM and (reformatted) signalling channels are multiplexed onto the common eight-bit wide data bus 33 and transmitted to a plurality of fiber interface module ("FIM") circuit boards 34, also located in the OLT 22. Similarly, a second eight-bit wide data bus 35 carries upstream PCM and signalling channels from the respective FIM boards 34 to the ETMs 28. As is described in greater detail herein, each of the FIM circuit boards 34 cross-connects downstream bus side PCM and signalling channels received from downstream data bus 33 into fiber side PCM and signalling channels, respectively. The fiber side channels are assembled into time-division-multiplexed packets for optical transmission over respective downstream fiber cables 23 to a respective plurality of remotely located ONUs 24. At each ONU 24, the received downstream optical data packets are converted to electrical signals, disassembled back into PCM and signalling channels, and forwarded to respective subscriber lines (not shown).
Similarly, upstream time-division-multiplexed data packets are assembled at the individual ONUs 24 and optically transmitted over a plurality of upstream fiber cables 27 to the plurality of FIM boards 34 at the OLT 22. Received upstream optical data packets are converted to electrical signals and disassembled into respective upstream fiber side PCM and signalling channels. The upstream fiber side channels are cross-connected into upstream bus side channels and multiplexed onto the upstream data bus for transmission to the plurality of ETM boards 28.
In alternate preferred embodiments, the respective downstream and upstream data busses 33 and 35 may be duplicated to increase the system capacity. Additional back-up busses (not shown) are also preferably provided in both directions in case of failure.
Referring to FIG. 2, an exemplary preferred FIM circuit board 34 includes a cross-connect and packet assembly/disassembly ("XCONN/PAD") circuit 36, which can be implemented as part of an application specific integrated circuit ("ASIC"), and an opto-electronic converter ("O/E") circuit 38 configured to receive downstream data from, and transmit upstream data to, respectively, the XCONN/PAD circuit 36. In particular, the O/E circuit 38 converts downstream electrical signals transmitted from the XCONN/PAD circuit 36 into information-equivalent optical signals for transmission over a downstream fiber 23 to one or more ONUs 24 (not shown in FIG. 2) associated with that particular FIM board 34, and vice-versa. A presently preferred embodiment of the O/E circuit 38 is described more fully in U.S. patent application Ser. No. 08/608,166, entitled "DC Coupled receiver for Shared Optical System," filed Mar. 20, 1996, which is fully incorporated herein by reference.
The FIM board 34 also includes a processor 42 with a memory (not shown), that controls, configures, and monitors the performance of the XCONN/PAD and O/E circuits 36 and 38, respectively. The FIM processor 42 communicates via a processor control link 68 with the processors of each of the other FIM boards 34, as well as respective processors (not shown) of the ETM circuit boards 28, in the OLT 22. A twenty-seven-bit wide processor address and data bus 29 links the FIM processor 42 to the XCONN/PAD circuit 36. The FIM processor also receives the incoming optical power level 25 from the O/E circuit 38. Control inputs 37 sent from the XCONN/PAD circuit 36 are used to control the O/E circuit 38.
As is described in greater detail below, in conjunction with the FIM processor 42 and four external memories 80, the XCONN/PAD circuit 36 cross-connects the downstream bus side PCM and signalling channels into fiber side PCM and signalling channels, which are packetized in 960 channel downstream data frames over a downstream serial interface 30, i.e., for a fiber side data rate of 7.68 MBps. Similarly, the XCONN/PAD circuit 36 receives 960 channel packets of upstream PCM and signalling channels from the O/E circuit 38 over an upstream serial interface 32 at the data rate of 7.68 MBps, and cross-connects the upstream fiber side channels into respective bus side upstream channels, which are then multiplexed onto the upstream data bus 35 at a data rate of 12.288 MBps.
As will be apparent to those skilled in the art, the difference in the respective bus and fiber side data rates is due to the number of "inactive" (i.e., unassigned) bus side channels at any given FIM card 34, which are eliminated on the fiber side. In particular, the respective channels in each 1536 byte downstream frame transmitted over the downstream bus 33 are allocated across all of the respective plurality of FIM boards 34, depending on the specific configuration of the network 20.
In order to clock the respective downstream and upstream bus side channels, the XCONN/PAD circuit 36 receives a 12.288 MHz input clock signal 74. A loop filter 60 and associated voltage-controlled crystal oscillator (VCXO) 62 are employed by the XCONN/PAD circuit 36 to generate a 61.44 MHz clock signal 66, which is synchronized to the 12.288 MHz input clock signal 74, for clocking the respective 960 byte fiber side data frames over the downstream and upstream serial interfaces 30 and 32 (i.e., on a bit-by-bit basis).
The external memories 80 store downstream and upstream data, which is transferred between the XCONN/PAD circuit 36 and the memories 80 via a two-way data link 39, and an address/control input 41.
For ease in illustrating aspects and features of the present invention, the downstream cross-connect and packet assembly operation of the XCONN/PAD circuit 36 are described below. As will be apparent from the present disclosure to those skilled in the art, the upstream packet disassembly and cross-connect process is performed in essentially the same (albeit reversed) manner.
Referring to FIG. 3, downstream PCM and signalling channels are received by the XCONN/PAD circuit 36 from the eight-bit wide backplane bus 33 in frames containing 1536 channels every 125 usec, wherein the respective PCM channels are routed to a downstream PCM memory 40, and the signalling channels are routed to a downstream signalling memory 52, respectively. In accordance with known telecommunication cross-connect design methodologies, not all of the bus side PCM and signalling channels are "active" (i.e., assigned to respective subscribers) on any given FIM card 34, and respective write enables (not shown) are input to the downstream PCM and signalling memories 40 and 52, so that only the active PCM and signalling channels are stored therein.
More particularly, as each new downstream data frame is received off the bus 33, the active PCM channels are written into a bus side data port of the downstream PCM memory 40. A bus side timeslot counter 44 clocked by the 12.288 MHz clock signal 74 transmits a twelve-bit bus timeslot address corresponding to each received PCM channel to an address input of the downstream PCM memory 40, i.e., wherein the bus timeslot address corresponds to the respective timeslot number of the PCM channel being stored. In this manner, the data contained in the active PCM channels in each downstream frame is stored in the downstream PCM memory 40 according to its respective bus side timeslot number. Eleven bits of the bus timeslot address are used for addressing a page of the downstream PCM memory 40, which is implemented effectively as a dual port, two-page, memory, as discussed in greater detail hereinbelow. The twelfth bit is used to determine whether to address to the first page or the second page of the memory 40. The counter also receives a multiframe synchronization pulse 75 at the beginning of each new signalling multiframe received off bus 33.
The twelve-bit bus side timeslot address is also sent to a bus side control memory 54 and a bus signalling timeslot conversion circuit 56, which converts bus timeslot numbers within each frame to associated PCM timeslot numbers within the frame. The bus side control memory 54 is decoded to provide write enable signals 71, 73 to the downstream PCM memory 40 and the downstream signalling memory 52, respectively. Additionally, the bus side control memory 54 sends to the bus signalling timeslot conversion circuit 56 a four-bit datafield 201 that specifies whether each timeslot has been assigned to a PCM channel, a signalling channel, or is inactive. The bus signalling timeslot conversion circuit 56 also receives a frame number 57, which runs from 0 to 23 for T1 applications, and from 0 to 15 for E1 applications. The bus signalling timeslot conversion circuit 56 builds a table of the timeslot numbers of active PCM channels that have their signalling carried in the current frame, looks up entries in that table during signalling timeslots, and sends a corresponding twelve-bit address signal to a bus side address input of the downstream signalling memory 52.
The cross-connect address control store memory 46 controls the readout of the respective PCM channels stored in the downstream PCM memory 40 through a fiber side data port by transmitting a twelve-bit fiber side address 203 to a fiber side address input 57 of the downstream PCM memory 40. An associated fiber timeslot counter 48 is clocked by a 7.68 MHz clock signal 67, which is derived from the 61.44 MHz clock signal 66 shown in FIG. 2. The fiber timeslot counter 48 also receives the multiframe synch pulse 75 at the start of each signalling multiframe and sends a ten-bit wide address signal to the cross-connect address control store memory 46. This signal (the output of the counter 48) changes state with each clock pulse such that the counter 48 counts from zero to 959 and then resets to zero with the start of the next frame. The ten-bit wide signal output by the fiber timeslot counter 48 is sent to a fiber signalling timeslot conversion circuit 58 that converts fiber timeslot numbers within each frame to associated PCM timeslot numbers within the frame. The fiber signalling timeslot conversion circuit 58 also receives a frame number 59 and sends a twelve-bit-wide address signal to the downstream signalling memory 52, which sends an eight-bit-wide downstream signalling data signal to the PAD circuit 50.
The cross-connect address control store memory 46 sends an eight-bit control word to a packet assembly/disassembly ("PAD") circuit (or "packet engine") 50, which also receives an eight-bit wide downstream data signal from the downstream PCM memory 40. The PAD circuit 50 sends eight-bit-wide downstream data 63 at a rate of 7.68 MHz to a parallel-to-serial conversion circuit ("P/S") 61, which outputs the cross-connected downstream data 30--which has been adjusted to the fiber rate of 61.44 mbps and packetized with various protocol inserted--from the XCONN/PAD circuit 36 to the O/E 38. As one of skill in the art would readily perceive, the above description of downstream data travel can effectively be reversed to convey a similar understanding of upstream data flow.
Referring to FIG. 4, a presently preferred XCONN/PAD circuit 76 is preferably implemented as a single ASIC, including four major modules: a bus side cross-connect ("BUSX") module 86; a fiber side cross-connect ("FIBX") module 88; a fiber front end ("FIBFE"), or packet engine, module 77; and an external memory access module ("XMAM") 90. The following description discusses functional data flow through the XCONN/PAD circuit 76.
The BUSX module 86 interfaces with the FIM processor (not shown in FIG. 4) over an address/data/control ("A/D/C") bus 78 for configuration and control, and receives a clock signal 79. In a preferred embodiment, the clock signal 79 has a frequency of 12.288 MHz and the A/D/C bus 78 has sixteen bits of address, eight bits of data, and three control signal bits, including chip select and read/write, for a total of twenty-seven parallel bits wide. The A/D/C bus 78 allows the FIM processor to control and monitor the various modules of the XCONN/PAD circuit 36. The BUSX module 86 includes a bus timeslot counter 81, a BUSX state machine 94, and a loopback register block 96.
Incoming downstream data 82 to the BUSX module 86 is routed through a parity check/multiplexer 92 and sent through a multiplexer/demultiplexer circuit ("MUX/DEMUX") 98 to the loopback register block 96 and to downstream PCM and signalling data ports on the XMAM 90. Similarly, a MUX/DEMUX 100 passes upstream data 97 to the A/D/C bus 78 and the loopback register block 96 from upstream PCM and signalling data ports on the XMAM 90. The MUX/DEMUXes 132, 134, 108, 114, 110, 112, 130, 98, 100, 102, 104 thereby allow access to respective SRAM memories 83, 84, 85, 93 by the FIM processor for configuration and test. The bus timeslot counter 81 addresses a bus side timeslot type control memory (BTS.sub.-- TYP) 84, and the data read from the BTS.sub.-- TYP memory 84 is sent to the BUSX state machine 94. In addition, the bus timeslot counter 81 is stored as data during PCM timeslots to build the signalling timeslot lookup table in a bus signalling timeslot memory (BSIGTS) 85. The data from the signalling timeslot memory 85 is routed to the XMAM module 90 to address a downstream signalling memory ("DSIG.sub.-- M") 87 and an upstream signalling memory ("USIG.sub.-- M") 138.
The FIBX module 88 interfaces with the A/D/C bus 78 for configuration and control, and includes a fiber timeslot counter 91 and a FIBX state machine 106. In the FIBX module 88, the fiber timeslot counter 91 addresses the fiber control store memory ("FXM") 95 in the XMAM module 90, and the FXM data output of the XMAM module 90 sends control data and cross connect addresses to the FIBX state machine 106, which addresses a fiber signalling timeslot memory ("FSIGSTS") 93 with the signalling timeslot number (i.e., a "SIGNUM" field in the data from the FXM 95). The FSIGTS memory 93 receives data from the fiber timeslot counter 91, and addresses the DSIG.sub.-- M 87 of the XMAM 90. The FIBX state machine 106 addresses upstream and downstream PCM memories 136, 99 and upstream and downstream signalling memories 138, 87 respectively, of the XMAM 90.
In the FIBFE module 77, a phase-frequency detector comprises part of a phase-locked loop 116 that, along with loop filter 60 (shown in FIG. 2), VCO 62 (shown in FIG. 2), and a digital divide by five circuit (not shown), generates a 61.44 MHz clock signal synchronized to the 12.288 MHz downstream bus clock signal 74. The FIBFE module 50 also includes a transmitted-data-byte processing block 118, a transmitted-data serial processing block 120, a FIBFE state machine 122, an electrical/optical (E/O) pulse shape controller block 124, a received-data-byte processing block 126, and a received-data serial processing block 128.
In the FIBFE module 77, downstream data from downstream cross connect memories ("DPCM.sub.-- M and DSIG.sub.-- M") 99, 87 in the XMAM 90 is sent via MUX/DEMUXes 130 and 134 to transmitted byte processing block 118 and then to transmitted serial data processing block 120. Upstream data received from an O/E circuit (not shown) is passed through received serial processing block 128 to received byte processing block 126, and then via MUX/DEMUXes 132 and 134 to upstream cross connect memories ("UPCM.sub.-- M and USIG.sub.-- M") 136, 138 in the XMAM 90. An electro-optical pulse shape block 124 generates control for the O/E circuitry. Data from the FXM 95 is input to the FIBFE state machine 122 to control FIBFE 77 circuitry.
The XMAM 90 coordinates accesses to a number of "virtual" multi-port memories, time-division multiplexing the accesses at a higher rate (30.72 MHz) than the bus clock signal 74 (12.288 MHz) into four single-port SRAM chips 83, which include the downstream PCM memory 99, the downstream signalling memory 87, the upstream PCM memory 136, the upstream signalling memory 138, and the FXM 95 (the fiber side cross-connect control store memory), which the XMAM 90 maps into four external SRAM memories ("XMEM") (not shown) located outside the XCONN/PAD circuit 36. In particular, the downstream and upstream PCM and signalling memories 99, 136, 87 and 138 each contain four pages (two for the expansion bus) of fifteen hundred thirty-six byte addresses, and the FXM 95 contains two pages of 960.times.24 bytes.
In the XMAM 90, the FXM 95 has an address port and a data port, wherein the address port communicates via MUX/DEMUX 110 with the FIBX module 88. The downstream and upstream PCM and signalling memories 99, 136, 87, and 138 each have left address and data ports and right address and data ports, respectively. In particular, the left address and data ports of the downstream PCM and signalling memories 99 and 87 each communicate with the BUSX module 86, respectively, and the right address and data ports each communicate with the FIBX module 88 and FIBFE module 50, respectively.
In a preferred embodiment, the XMAM module 90, which implements virtual XMEM memories in external SRAMs, receives bus and fiber access request, read/write, and address lines, a 12.288 MHz bus side clock signal, a 7.68 MHz fiber side clock signal, the A/D/C bus 78 from the processor, and a 30.72 MHz clock signal. Additionally, bus side data and fiber side data can be read from or written to the XMAM 90. The external memories have an address input that is driven only by the XMAM 90, and data can be read from or written to the external memories only through the XMAM 90.
In a presently preferred embodiment, the SRAMs comprise four single-port SRAMs. Access to the external memories through the XMAM 90 is time-division multiplexed among the various virtual memory ports of the XMAM 90 at a higher rate, i.e., 30.72 MHz. Because the virtual memory ports are all clocked at a rate that is divided down from the same clock, i.e., 61.44 MHz, accesses can be assigned timeslots at a rate of 30.72 MHz. In particular, bus accesses to the upstream and downstream PCM and signalling memories 87, 99, 136, 138 take two-fifths, fiber accesses one-fourth, FXM 95 accesses one-fourth (plus one-fourth to the external memories not accessed by the fiber side PCM and signalling ports), and FIM processor accesses one-tenth, respectively, of the cycles.
The FIM processor may also access the memories of the XMAM 90 in unused bus or fiber timeslots, e.g., during an overhead timeslot. Input data for all accesses to the XMAM 90 is registered in the XMAM 90 until the 30.72 MHz timeslot for the access port occurs, and outputs are available for fiber side data one 7.68 MHz cycle later and for bus side data one and one-half 12.288 MHz cycles later. Bus side data accesses can occur 12.288 million times each second and fiber side accesses can take place 7.68 million times per second. Multiple accesses of the virtual memories 80 are accomplished per bus side or fiber side clock cycle. In this way the XMAM 90, although implemented with single-port memories, functions effectively as a set of multi-port memories.
Referring to FIG. 5, the BUSX module 86 includes a front-end block 140, which performs parity detection and multiplexing functions on the data. A back-end loopback block 142 performs loopbacks, i.e., sending downstream data back upstream for system testing. Bus side registers 144 interface with the FIM processor, and the bus timeslot counter 81 keeps track of the bus timeslot count within each frame. The BUSX module 86 also includes a block 146 that includes two downstream signalling scan memories and a downstream signalling scan change memory to enable downstream signalling scanning, as described in detail below. An operation and maintenance ("OAM") block 148 handles operation and maintenance channels of the system. A BTS.sub.-- TYP controller 150 and a BSIGTS controller 152 control addressing, decoding, and data input/output for the BTS.sub.-- TYP memory 84 and BSIGTS memory 85, respectively.
Referring to FIG. 6, within the FIBX module 88, upstream and downstream fiber PCM blocks 154 and 156 receive, test, and route upstream and downstream data channels, respectively, from the UPCM.sub.-- M 136 and the DPCM.sub.-- M 99. A plurality of fiber side registers 158 are in communication with the FIM processor, and the fiber timeslot counter 91 keeps track of the fiber timeslot count within each frame. A half-FIFO 160, which stores data sequentially until full, but appears as a random access memory to the FIM processor, serves as a buffer to store fiber errors that accumulate too rapidly for the FIM processor to handle, is read from periodically by the FIM processor. Several types of ONU errors can occur, including, for example, missing packets, bad signalling, or parity errors. The half-FIFO 160 stores the error type and the timeslot number of the error. A fiber control store glitchless switching block 162 performs glitchless, or synchronous map, FXM 95 page switching, as discussed in detail below. A single COMM packet is sent and received in each frame from the FIM to the ONUs with the aid of a communication control block 164.
Referring to FIG. 7, within the FIBFE module 77, downstream data received from the downstream PCM memory 99 of the XMAM 90 is sent to a FLOOP register 166, which stores data contained in a downstream test channel identified in the FXM 95 to be sent back upstream for testing purposes. The downstream data is also sent through MUX/DEMUX 130 to a downstream packet assembly block 168. The downstream packet assembly block 168 assembles the data packets and inserts overhead, checksum, and (if under the T1 standard) TT bits, which identify which of three domestic variations of the T1 standard signalling is being used. A single-bit wide downstream data packet parity memory ("DPAR") 170 stores a parity bit for each data packet for transmission in the next frame, as controlled by the downstream packet assembly block 168. The downstream data is sent to a downstream parallel-to-serial converter 172 and then transmitted over the downstream serial interface 30 to the O/E circuit 38 (not shown in FIG. 7). The downstream data is also routed to a full serial loopback circuit 174, which may be selectively activated for sending the fiber side data stream back upstream for testing purposes.
The FIBFE state machine 122 decodes the data transmitted from the FXM 95 of FIG. 4 (i.e., after selecting data from page zero or page one of the FXM 95). The state machine 122 also receives control signals from the O/E pulse shape controller 124, which generates control signals for the O/E 38.
Upstream data is received from the O/E circuit 38 via the upstream serial interface 32 and sampled by a received-data-sampler ("RXDSAMP") 176 with an associated digital delay-locked loop ("DIGDLY") 178, which generates eight approximately equally spaced (in time) phases of the 61.44 MHz transmit clock. The loop is updated once per frame with an isolated training pulse inserted into the clock phases. The samples are synchronized and sent to a digital phase acquisition (i.e., clock recovery) circuit ("DIGCLK") 180, which acquires accurate phase for the received data by selecting the correct phase from among eight equally spaced (in phase) samples of the received data. Presently preferred embodiments of the RXDSAMP 176, DIGDLY 178, and DIGCLK 180 blocks are described more fully in U.S. patent application Ser. No. 08/787,849, entitled "Digital Phase Acquisition with Delay Locked Loop," filed Jan. 22, 1997, which is assigned to the assignee of the present application and which is fully incorporated herein by reference.
The recovered (and synchronized) upstream data channels are fed through an upstream serial-to-parallel converter 182, which communicates with a time-of-flight counter 184, which counts the time between transmission of a COMM packet sent to a given ONU and the received response of the ONU, thereby measuring the time of flight on the respective downstream and upstream fibers to each ONU. The different times-of-flight for each of the ONUs are preferably compensated for by adjusting individual transmission delays in each ONU, so that every COMM packet response comes back to the FIM at the same time in the upstream frame, although only one ONU transmits a COMM packet response in any given frame.
From the upstream serial-to-parallel converter 182, the upstream data is sent to a packet disassembly block 186, which decodes and disassembles the upstream data packets and "freezes" the signalling bytes if a data error or missing checksum is detected--i.e., the previous signalling data is retained and the corrupt signalling data is discarded. A single-bit upstream data packet parity memory ("UPAR") 188 stores data packet parity for verification in the next received frame under the control of the packet disassembly block 186. Additionally, a FIBERR block 190 detects fiber errors.
The recovered upstream PCM channels are routed, via control mux 132, to the respective upstream PCM and signalling memories 136 and 138 of the XMAM module 90. The upstream signalling channels are routed to an upstream signalling register 192, which stores the respective signalling channel bytes until they can be written into the upstream signalling memory 138, as described in greater detail below. In particular, the upstream signalling register 192 transmits the signalling channels to the upstream signalling memory 138 via a MUX/DEMUX 134.
The various internal memories included in the XCONN/PAD circuit 76 described in connection with FIGS. 4-7 are listed by name in FIG. 9, with size and functional description provided for each. Function of the XMAM module 90, which resides external to the XCONN/PAD circuit 76 and maps five "chunks" of virtual memory into four external XMEM memories, is now described in detail according to a presently preferred embodiment.
The four external memories of the XMAM 90 are organized as shown in tabular form in FIGS. 8A-D, which provide names, addresses, sizes, and descriptions of the contents of the respective memories. As can be seen, the fiber cross-connect memory (FXM) is distributed over all of the four external memories, which is necessary because the FXM requires more bandwidth than do the respective downstream and upstream PCM and signalling memories.
In particular, the first external memory ("XMEMA") includes the downstream PCM memory, the downstream signalling memory, and page zero of the low and high bytes of the FXM. The second external memory ("XMEMB") includes the upstream PCM memory, the upstream signalling memory, and page one of the low (bits seven to zero) and high bytes of the FXM. Similarly, the third and fourth external memories ("XMEMC") and ("XMEMD") include additional downstream and upstream PCM and signalling memories for the expansion bus as well as, respectively, pages zero and one of the middle (i.e., bits fifteen to eight) and high bytes of the FXM, which, in the depicted preferred embodiment, is twenty-four bits wide.
The high byte (i.e., bits twenty-three to sixteen) of the FXM is stored such that page zero is duplicated and contained in both the first and third virtual memories XMEMA and XMEMC, and is read from whichever one of the two memories is idle. Page one of the high byte of the FXM is likewise duplicated and contained in both the second and fourth virtual memories XMEMB and XMEMD, and is read from whichever one of the two memories is idle. Because each fiber timeslot may be cross-connected to either a channel on the main bus (resulting in accesses to XMEMA and XMEMB) or a channel on the expansion bus (resulting in accesses to XMEMC and XMEMD), one pair of XMEMs is always idle and available to read the high byte of the FXM.
Operation of the XCONN/PAD circuit 76 of FIG. 4 is now described in detail according to a presently preferred embodiment.
The downstream and upstream PCM memories are each divided into two pages and preferably duplicated, as discussed above in conjunction with FIGS. 8A-D, for expansion busses. While one page of the downstream PCM memory is written by the bus side cross-connect hardware, the other page is read by the fiber side cross-connect hardware. Switching of the pages takes place on frame boundaries. Data is stored at addresses corresponding to the bus timeslots. The downstream and upstream signalling memories are likewise each divided into two pages and duplicated for the expansion busses. While one page of the downstream signalling memory is written by the bus side cross-connect hardware, the other page is read by the fiber side cross-connect hardware. Page switching takes place on signalling multiframe boundaries, with signalling data stored at addresses corresponding to the associated PCM channel bus timeslots. The FIM hardware computes the associated PCM channel bus timeslot from the signalling timeslot and the frame number, as discussed below.
The FIMs can be programmably assigned to any of the backplane bus timeslots. Typically, the operator of the optical fiber communication network sets up a cross-connection between a channel on an E1 (or T1) interface from the switch to a channel in an ONU. Each E1 (or T1) channel is assigned to a fixed backplane timeslot, and the passive optical network that the ONU is connected to is terminated at a particular FIM. Therefore, firmware on that FIM must assign a fiber timeslot to a line card in a respective ONU and cross-connect the assigned fiber timeslot to the backplane timeslot. Each FIM clocks its downstream bus receivers and enables its upstream bus drivers only in timeslots that are assigned to that FIM. Each FIM also ignores downstream bus data and disables its upstream bus drivers in other timeslots.
In addition, to support a subscriber channel with signalling, the FIM must assign one frame (out of the sixteen or twenty-four frames in a signalling multiframe) of a fiber signalling timeslot to be associated with the fiber channel timeslot. The FIM must also cross-connect the correct frame of the fiber signalling timeslot to the correct frame of the backplane signalling timeslot that is associated with the backplane channel timeslot. Thus, for a signalling timeslot, the effective cross-connection changes for each frame in the signalling multiframe. Consequently, a different FIM may be assigned to each frame of a backplane signalling timeslot.
In a preferred embodiment, serial data traveling at 61.44 mbps is transmitted on separate upstream and downstream fibers between the respective FIMs and ONUs. Eight-bit wide fiber side timeslots are assigned semipermanently with the assignments repeated every frame at a frame rate of eight KHz. Thus, each timeslot forms a 64,000 bps channel. The same timeslot assignments are used on both the upstream and the downstream fibers. Data is scrambled on the fibers to reduce the probability of long sequences without transitions, thereby facilitating clock recovery from the data. Other fiber transmission speeds may be supported with modified FIMs.
Because each ONU has a different length of fiber running to it, with different sources of attenuation and different optical transmitters and receivers, some overhead data is needed at the beginning of each upstream ONU transmission to allow the FIM to recover timing and optical level from the received data. To minimize the number of overhead timeslots necessary, each ONU groups its transmissions into a packet so that all timeslots assigned to an ONU are contiguous in time and only one overhead sequence is needed per ONU.
Each ONU is assigned one or more signalling timeslots to carry signalling for associated channels in its packet. Each fiber signalling timeslot carries signalling for up to fifteen E1 (or twenty-four T1) subscriber channels. To facilitate mapping of fiber signalling timeslots to subscriber timeslots, each fiber signalling timeslot preferably precedes the associated subscriber timeslots in the packet. If more than one fiber signalling timeslot is carried in an ONU packet, i.e., to support more than fifteen channels in an ONU, then each fiber signalling timeslot carries signalling information for the subscriber timeslots that follow until the next signalling timeslot or the end of the packet.
For cross-connection of subscriber channels, bus side, or backplane, timeslot types can preferably be stored in a four-bit bus side control store memory referred to as BTS.sub.-- TYP, which can be written or read by FIM firmware. A second bus side control store memory is used to store timeslot types for the expansion busses. Timeslot types that can be identified include INACTIVE (unused timeslot), VOICE (active voice, or PCM, channel in use by the FIM), DATA (active data channel, for example, to support ISDN service), SIG1 (first signalling timeslot), and SIG2 (second signalling timeslot associated with international (i.e., E1) standard voice channel timeslots 17-31). During VOICE or DATA timeslots, data from the downstream backplane bus is written into the main downstream cross-connect memory, using the timeslot number as the address, and data is read from the main upstream cross-connect memory, again with the timeslot number as the address. The data is driven onto the upstream backplane bus. During INACTIVE timeslots, no data is read or written to the downstream PCM and signalling memories, and the upstream bus drivers are disabled.
To cross-connect signalling channels, the timeslot number of the associated PCM channel is first determined given the signalling timeslot number and the frame number in the signalling multiframe. On the backplane side, signalling channels are directly addressed in the respective downstream and upstream signalling cross-connect memories by the timeslot number of the associated PCM channel. On the fiber side, signalling channels are addressed in the signalling memories by the backplane timeslot number to which the associated fiber channel is cross-connected. Thus, the signalling cross-connect memories can be implemented with the same number of addresses as the respective PCM cross-connect memories, instead of having to be sixteen times as large (or, for T1, twenty-four times as large). In addition, a row register includes a plurality of flip-flops (not all of which are used in T1 applications). The row register stores for each of the forty-eight (or thirty-two for T1) signalling timeslots whether the row may contain PCM channels.
If a signalling timeslot is active, signalling data from the downstream bus is written into the downstream signalling memory and signalling data is read from the upstream signalling memory and driven onto the upstream bus, respectively. If a signalling timeslot is not active during the frame for the FIM, no data is written into the downstream signalling memory and the upstream bus drivers are disabled.
In a presently preferred embodiment, backplane signalling is mapped to associated timeslots during signalling cross-connect by the bus signalling timeslot memory (BSIGTS). In particular, the BSIGTS memory and bus side state machine are used to store the timeslots that will be associated with signalling timeslots in the current frame. Given the timeslot number of a backplane signalling timeslot and the frame number in the signalling multiframe, the timeslot number of the associated timeslot can be determined because assignment of backplane timeslots is fixed. For example, in FIG. 10, which illustrates backplane timeslot assignment, the associated channels are all in the same column as the signalling timeslot. In frame number one, for example, the first timeslot (row number five) is associated with the first signalling timeslot (row number forty-six), and the seventeenth timeslot (row number twenty-six) is associated with the second signalling timeslot (row number forty-seven). In frame number two, the second PCM timeslot (row number six) is associated with the first signalling timeslot, and the eighteenth PCM timeslot (row number twenty-eight) is associated with the second signalling timeslot, and so on, up through frame number fifteen in the signalling multiframe. The associations repeat again in the next multiframe.
A backplane PCM row counter (not shown) in the BSGTS.sub.-- CTL block 152 of FIG. 5 is reset at the start of each frame and increments at each row of bus side timeslots that may contain a PCM channel, as indicated by the associated flip-flops in the row register, which is written once at initialization by the FIM processor. A signalling frame counter is reset at the start of each signalling multiframe and increments at the start of each frame. When the state of the PCM row counter is equal to the signalling multiframe, the timeslot number of each active PCM channel is stored into the first row of the BSIGTS memory. When the state of the signalling row counter minus fifteen is equal to the signalling multiframe number, the timeslot number of each channel is stored into the second row of the BSIGTS memory. In a preferred embodiment, the first row of the BSIGTS memory is in the first thirty-two addresses of a 64.times.16 memory internal to an ASIC, and the second row of the BSIGTS memory is in the second thirty-two addresses of the same memory.
During the first row of signalling timeslots, the first row of the BSIGTS memory is read to retrieve the timeslot numbers of the associated channels. Similarly, during the second row of signalling timeslots, the second row of the BSIGTS memory is read. Signalling data is then written into the downstream signalling memory and read from the upstream signalling memory at the address equal to the timeslot number of the associated channel, which was read from the BSIGTS memory.
The bus timeslot counter addresses the BTS.sub.-- TYP memory and identifies whether a bus timeslot is SIG, VOICE, UNUSED, etc., and whether the payload timeslot is active. The state machine stores the timeslot numbers of the channels that correspond to the signalling in the current frame in the BSIGTS memory. When a signalling timeslot occurs (always after the PCM timeslots), the associated timeslot is retrieved from the BSIGTS memory and the signalling data is read from the upstream signalling memory and written into the downstream signalling memory at the same address as its associated PCM timeslot.
In a preferred embodiment, fiber signalling is mapped to associated PCM timeslots during downstream signalling cross-connect by means of the fiber signalling timeslot memory (FSIGTS) memory. Fiber side signalling timeslots precede the associated fiber timeslots and are identified in the fiber cross-connect address control store, or fiber control store, memory (FXM). For each fiber channel, the FXM is programmed by the FIM processor with control and cross-connect information, to control the cross-connect circuitry.
The downstream signalling cross-connect hardware uses an approach that is similar to the above-described bus side signalling cross-connect approach. Specifically, the FXM entry for each signalling timeslot includes a unique signalling number (SIGNUM) that is used to address the FSIGTS memory. The FXM entry for each PCM timeslot includes the signalling frame number (SIGFR) that carries the channel's signalling in the preceding fiber signalling timeslot, as well as the backplane timeslot number to which the channel is cross-connected. When the SIGFR number from the FXM equals the signalling frame number plus one, the hardware state machine stores the backplane, or bus side, timeslot number of the channel in the FSIGTS memory at an address equal to the SIGNUM field from the FXM entry for the preceding signalling timeslot.
Put another way, the bus timeslot pointers for channels that will have signalling active in the next frame are stored in the FSIGTS memory. In the next frame, the FSIGTS memory is read during the signalling timeslot at the address equal to the SIGNUM field to get the bus timeslot number of the associated PCM channel. Signalling data (preceding the PCM channels) is then read from the downstream signalling memory at an address equal to the bus timeslot number where the associated PCM data was stored.
In sum, in the downstream data path, the fiber signalling timeslot memory (FSIGTS) is used to store the backplane timeslot number of the channel that will be associated with each fiber signalling timeslot in the next frame. When each fiber signalling timeslot occurs, the FSIGTS memory can be read to determine the associated channel backplane timeslot for the current frame. That timeslot number is then used to address the downstream signalling cross-connect memory to read the signalling data, which is then transmitted to the ONUs. In contrast, the upstream signalling cross-connect hardware is simpler, merely saving the signalling data for a few cycles. When a timeslot occurs with signalling active in the current frame, the signalling data is stored in a register while the data in the subsequent PCM timeslots is written to the upstream PCM memory. The saved signalling data is then written into the upstream signalling memory, at the address equal to the bus timeslot number of the associated PCM channel, at the next idle memory cycle, i.e., the next signalling or overhead timeslot.
In a preferred embodiment, the packet engine circuitry can be combined with the cross-connect circuitry because the FXM can identify a number of other timeslots besides merely PCM and signalling, and because the cross-connect circuitry can read and write fiber data and bus data at the appropriate rates to accommodate the differences in clock rates. Thus, OVERHEAD and COMM timeslots are used to control fiber packet assembly and communicate with hardware in the ONUs. Preferably, different timeslot types are identified by different values in a five-bit field in each FXM entry. Other bits in the FXM can then be used to hold information specific to the timeslot type programmed in the given location. In particular, the twelve-bit field to identify the backplane timeslot number to which fiber timeslots are cross-connected is used for several other packet-assembly purposes in other timeslot types. Thus, the same control store (i.e., the FXM) is used to control both the packet engine and the cross-connect circuitry, thereby reducing the memory required.
The FXM entry for OVERHEAD timeslots is also programmed by the FIM processor with the overhead pattern that is to be transmitted on the downstream fiber. The FIM processor uses the same bit field that holds the backplane timeslot numbers for payload timeslots. In this way, any needed overhead pattern can be generated and interspersed with payload timeslots. Data received from the upstream fiber in OVERHEAD timeslots is generally used to recover timing or optical signal level, and discarded by the packet engine.
The FXM entries for COMM timeslots are also programmed by the FIM processor, with pointers to communications buffers (DCOMM.sub.-- M and UCOMM.sub.-- M in XMEMA and XMEMB of FIG. 8A) used to exchange messages between the FIM and respective ONUs for configuration, performance monitoring and maintenance purposes. These pointers also use the same bits that hold the backplane timeslot number for payload timeslots.
An expansion backplane bus can preferably be used to increase the capacity of the optical fiber communication network from 1536 timeslots to 3072 timeslots. Thus, capacity can be doubled while the speed of the busses is maintained at 12.288 MHz. To avoid having to access the cross-connect memories at 24.576 MHz (i.e., at double speed), two sets of cross-connect memories are used: one set for the primary bus, and one set for the expansion bus. And two BTS.sub.-- TYP memories are used, again one for the primary bus and one for the expansion bus. However, since each fiber timeslot is cross-connected to a backplane timeslot on either the primary or the expansion backplane bus, only one FXM is required. Preferably, a bit is added to the FXM to specify which of the two sets of cross-connect memories to cross-connect to each fiber channel. In an especially preferred embodiment, the expansion bus is paired with a redundant expansion bus. Thus, the expansion bus actually consists of a set of two upstream and two downstream busses. However, only one of the busses carries data at any given point in time.
In a presently preferred embodiment, telephone channels are concentrated to more efficiently use network resources. Concentration channels are provided from the switch up to the cross-connect memories. In this concentration configuration, channels are provided on the fiber for each end-user circuit. On the backplane busses, however, channels are provided only for end-user circuits that are in use (e.g., a circuit that is off-hook is in use). Inactive circuits on the fiber are assigned special inactive backplane timeslot numbers to which to perform dummy cross-connects. This allows signalling from inactive channels to be monitored to detect an end-user request to go active (i.e., for a circuit to go off-hook).
Failed channels or channels under test are preferably controlled by specifying the channel as active on either the backplane side or the fiber side, but not both. The desired idle and signalling data can then be written directly by the FIM processor into the primary cross-connect memories and signalling cross-connect memories. This feature can also be used to generate test patterns on inactive channels. There are two alternating pages in the cross-connect memories, and frame and signalling multiframe interrupts are provided. Therefore, the FIM processor can write to the page not currently in use, thereby giving adequate time to allow for interrupt latency.
In an especially preferred embodiment, signalling information is scanned by special-purpose hardware to reduce the FIM processor power needed to support conversions to message signalling, concentration, testing, and signalling translation. Signalling scanning can operate in a transparent mode, whereby signalling information is still transmitted through the cross-connect memories. Alternatively, signalling scanning can operate in a scan-and-insert mode, whereby signalling information is scanned and discarded, allowing the FIM processor to insert signalling into the cross-connect memories.
Signalling scanning state machines receive the current signalling state for each channel, compare the current state to the previous state and, if it is different, store the new state and channel number in a signalling change memory ("DSIGCHG," included in the downstream signalling scanning block 146 of the BUSX module 86, or "USIGCHG") for the FIM processor to read, and store the new state in a comparison memory ("DSCAN," also included in block 146), at the address corresponding to the associated PCM channel, for the next comparison. The volume of data and the required computations are kept to a minimum because only the changes in signalling states are stored in the DSIGCHG memory and read by the FIM processor.
A timestamp from a multiframe counter is recorded in the DSIGCHG memory with the first signalling change recorded in a signalling multiframe. Preferably, the timestamp is written into the USIGCHG memory before the first signalling change in each signalling multiframe. The timestamp is preferably fourteen bits long and counts absolute multiframes continuously from reset until wraparound. If no signalling changes in a signalling multiframe, no timestamp is recorded, thereby reducing the data that the FIM processor must process. The FIM processor uses the timestamp to determine which multiframe each of the signalling changes occurred in. The FIM processor likewise determines the new signalling state for each channel when the signalling changes. Thus, no signalling information is lost, even if the FIM processor falls behind in processing signalling change information for more than a full signalling multiframe.
If the DSIGCHG memory fills, the signalling scanning state machines cannot store any more signalling state changes. To avoid losing a state change, the signalling scanning state machines do not store the new signalling states in the comparison memory (DSCAN) for comparison to the states in the next multiframes. Instead, the previous signalling states are left in the comparison memory. In this way, when the FIM processor empties some of the DSIGCHG memory and changes can again be recorded, the signalling scanning state machines will compare against the most recent signalling state of which the processor was notified. Thus, no signalling states will be lost unless multiple signalling transitions occurred while the signalling change memory was full.
To allow signalling to be intercepted before it is written to the cross-connect memories, the downstream signalling scanning circuitry scans backplane signalling data. The upstream signalling scanning circuitry scans fiber signalling data before it is written to the cross-connect memories.
Preferably, upstream fiber side signalling timeslots are scanned for changes if scanning is enabled on the timeslot. Scanning is enabled on a given timeslot if a particular bit within the FXM is set and the timeslot has an active cross-connect. Scanning may be disabled by the FIM processor. The new signalling state is compared to the prior signalling state from an upstream signalling scan memory ("USCAN"). If the two are different, the timeslot number and new signalling state are stored in the USIGCHG memory. The new signalling state is then stored in the USCAN memory at the timeslot corresponding to the channel associated with the signalling.
Upstream scanning is performed on the fiber side before signalling data is written into the upstream signalling cross-connect memory, thereby allowing insertion of interpreted signalling data by the FIM processor. The preferred upstream scanning circuitry supports a scan-and-insert mode only, whereby signalling data is not written into the upstream signalling cross-connect memory. To achieve firmware-transparent upstream signalling scanning, the FIM processor writes each new signalling state into the upstream signalling cross-connect memory when the new signalling state is received from the USIGCHG memory.
The FIM processor can be maskably interrupted when the USIGCHG memory is nonempty at the end of a signalling multiframe, or immediately when the USIGCHG memory is half full. The nonempty interrupt occurs once per multiframe. The half-full interrupt allows firmware to read the USIGCHG memory more frequently so as to prevent saturation under high-traffic peaks. When the USIGCHG memory is full, new signalling states are not stored in the USCAN memory. Thus, when the USIGCHG memory is emptied, signalling transitions are not lost (with the exception, as discussed above, of multiple changes on the same channel before the USIGCHG memory is emptied).
The upstream signalling scanning circuitry could detect false changes, and even possibly miss a signalling state change, when fiber timeslots are moved. Only channels that move timeslots can be affected. A preferred method of compensating is to read the new signalling state for all channels that moved timeslots after completion of the FXM page switch. USIGCHG memory entries on those channels are then ignored for the duration of one multiframe. An alternative preferred method is to write the current signalling state into the new timeslot for each channel that will shift timeslots.
Downstream scanning is performed on timeslots on the primary bus with scanning enabled. Preferably, downstream scanning is not performed on the expansion bus. As in the upstream direction, signalling changes and a fourteen-bit timestamp are written into the DSIGCHG memory with the aid of downstream signalling scanning memories ("DSCAN1, " "DSCAN2," included in block 146 of FIG. 5). Similar interrupts are provided, as for the upstream scanning hardware.
Preferably, one of three available downstream signalling scanning modes may be selected for each timeslot in the BTS.sub.-- TYP bus side control store memory: transparent scan, scan-and-insert, or no scanning. Transparent scanning checks for signalling changes and notifies the FIM processor, and signalling data is also written into the downstream signalling cross-connect memory and transmitted to the ONUs. Scan-and-insert mode likewise notifies the FIM processor of signalling changes, but signalling data is not written into the downstream signalling cross-connect memory. Instead, the FIM processor writes interpreted signalling data into the downstream signalling cross-connect memory after each signalling change.
In a preferred embodiment, a single parity bit is used to detect errors in fiber payload packets. This bit precedes the data packet as the last bit in a multipurpose header byte used to establish timing, signal an interrupt request, and identify the byte boundary of the packet. Putting the parity bit in the preceding multipurpose overhead byte, rather than in a byte at the end of the packet, minimizes the number of overhead timeslots needed per fiber payload packet.
Because the parity bit precedes the packet, the parity bit is actually the parity computed on the data in the packet sent in the previous frame. The packet engine has a single bit of memory (DPAR, designated block 170 in FIG. 7) for each ONU in which the packet engine stores the parity computed on each data packet until the next frame. At the beginning of each data packet, the parity computed on the data packet in the previous frame is retrieved. The FXM contains a unique ONUID number for each ONU. The ONUID number is used to address the parity bit memory (DPAR). (A second parity bit memory (UPAR, designated block 188 in FIG. 7) is similarly used for upstream data.)
Preferably, the XCONN/PAD circuit can prevent glitches or loss of data when the addition or deletion of channels (e.g., to increase the size of a packet sandwiched between several other packets) causes the fiber timeslots of other channels to be reassigned. First, when fiber timeslots must be reassigned, the change must be synchronized with the ONUs. Second, the data stored in the FSIGTS memory and parity bit memories (DPAR, UPAR) must still be addressed correctly with the new fiber timeslot assignment.
To synchronize fiber timeslot changes with the ONUs, downstream fiber data is first transmitted with the new fiber timeslot assignment. Next, the ONUs are commanded to switch to the new fiber timeslot assignment in the frame in which the ONUs start receiving the new assignment. The FIM switches the upstream part of the packet engine, or PAD, circuit (i.e., the packet disassembly portion) to use the new fiber timeslot assignment in the frame in which the FIM starts receiving upstream data from the ONUs with the new timeslot assignment. This switch can occur up to several frames after the FIM started transmitting the new assignment on the downstream fiber.
The XCONN/PAD circuit is controlled by a single fiber control store memory (the FXM) with two pages: an active (i.e., online) page and an offline page. Each FXM page contains a location corresponding to each fiber side timeslot. To synchronize a fiber timeslot switch with the ONUs, the FIM processor programs the offline page of the FXM with the new timeslot assignment. The FIM processor then sends communications messages to the ONUs to set up the new timeslot assignment in their offline timeslot maps. Finally, the FIM processor commands the FIM hardware to perform a synchronous map switch. The prior offline page of the FXM is used first by the downstream (i.e., packet assembly) portion of the PAD circuit. At the same time, the prior online page of the FXM is still used by the upstream (i.e., packet disassembly) portion of the PAD circuit. Several frames later the upstream portion of the PAD circuit stops using the prior online bank of the FXM. The switch is then complete, and the prior offline page of the FXM has become the current online page of the FXM.
Preferably, the FIM synchronizes downstream FXM control store switches to two frames before the start of a signalling multiframe. The FIM preferably has a programmable delay of between zero and seven frames from a time two frames before the start of the signalling multiframe to the time of the downstream FXM switch. A similar programmable delay is advantageously provided from the time two frames before the start of the signalling multiframe to the time that the COMM packet containing the map switch command for the ONUs is transmitted to the ONUs. Preferably, another similar programmable delay is provided from the time two frames before the start of the signalling multiframe to the time of the upstream FXM switch.
In an especially preferred embodiment, the North American standard for ISDN over T1 is supported. This requires each D timeslot to contain two bits from each of four sixteen-kilobit-per-second D channels. Downstream, two-bit D channels from the T1 interface are collected over four frames in the ETM module and sent once every four frames to the FIM as a full byte for each channel. Four channels share one timeslot in a four-frame multiframe. The data from each timeslot is stored in both pages of the signalling cross-connect memory and both pages of the PCM cross-connect memory, addressed by two bits of frame number and the timeslot number. Downstream ISDN D-channel data is read from the cross-connect memories at a D-channel address that is contained in the two associated B-channel entries in the control store memory (FXM). As will be apparent from the present disclosure, upstream operation is substantially similar.
In order to provide a still more further detailed description of a presently preferred embodiment of the above-described XCONN/PAD circuit, the following verilog design code is provided: ##SPC1##
While preferred embodiments have been shown and described, it will be apparent to one of ordinary skill in the art that numerous alterations may be made without departing from the spirit or scope of the invention.
Therefore, the invention is not to be limited except in accordance with the following claims.

Number	Name	Date
4763317	Lehman et al.	Aug 1988
4782479	Rozema	Nov 1988
4967405	Upp et al.	Oct 1990
5165092	Rickenbach et al.	Nov 1992
5241534	Omuro et al.	Aug 1993
5282195	Hood et al.	Jan 1994
5375121	Nishino et al.	Dec 1994
5390175	Hiller et al.	Feb 1995
5412498	Arstein et al.	May 1995
5452306	Turudic et al.	Sep 1995

Combined signalling and PCM cross-connect and packet engine

Information

Patent Number

Date Filed

Date Issued

Inventors

Original Assignees

Examiners

Agents

CPC

US Classifications

Field of Search

US

International Classifications

Abstract

Description

Claims

US Referenced Citations (10)

Non-Patent Literature Citations (3)

Entry
Hornbach, B.H. et al:"5ESS-2000 Switch: The Next Generation Switching System" AT&T Technical Journal, vol. 72, No. 5, Sep. 1, 1993, pp. 4-12, XP000425021, New York (US) see whole document.
Yukio Nakano et al.: "Signal Processing for SDH Digital Cross-Connect System" Proceedings of the International Conference on Communications (ICC), vol. 2, May 23-26, 1993, pp. 900-904, XP000371210, Geneva (CH), see whole document.
Horning K: "Automatic Crossconnect Equipment," Hasler Mitteillungen, vol. 48, No. 1, 1989, pp. 26-32, XP000096191, Bern (CH), see paragraph 4.3.