CA2162611C - Method and apparatus enabling synchronous transfer mode and packet mode access for multiple services on a broadband communication network - Google Patents

Abstract

STM traffic, e.g. voice and video telephony (VT), as well as packet mode (e.g.
ATM) traffic, e.g. broadcast digital video, interactive television, and data, are transmitted via a multiple access broadband fiber/coaxial cable network. Customer premises equipment (CPE) at stations, and a bandwidth controller, which may be at a head end or central office, with which all stations communicate, work together to adapt to the changing demands of the traffic mix, and efficiently allocate bandwidth to a variety of bursty and isochronous traffic sources. The bandwidth allocation defines two types of time slots, STM and ATM, and divides each frame into two corresponding STM and ATM regions. The boundary between the regions can be changed dynamically. A contention access signaling channel is provided in the STM region, for call control and set-up requests. Within the STM region, the time slots can be of variable length and be allocated on a per call basis; the length of the time slots is proportional to the bandwidth requirement of STM calls. Within the ATM region, the time slots are of fixed length, each capable of accommodating one ATM cell. Further, the fixed length ATM time slots may be reserved for a particular user for the duration of a call, or may be shared through a contention process. At least one contention ATM
time slot is always made available for signaling messages related to ATM call control and set-up requests. The downstream time frame is structured in a similar manner, but includes an additional MAP (slot allocation mapping) field to transmit to the stations ATM time slot allocation and status information for time slots in the upstream channel.

Description

METHOD AND APPARATUS ENABLING SYNCHRONOUS TRANSFER
MODE AND PACKET MODE ACCESS FOR MULTIPLE SERVICES
ON A BROADBAND COMMUNICATION NETWORK

Field of the Invention This invention relates generally to digital co.. l.. ication via a two-way high bandwidth (e.g. fiber and coaxial cable) co~ unication medium, and, in particular, to an access protocol and arrangement for allowing colnl)clillg stations col-n~tecl to a common head end via a tree and branch coll....ul~ication nclwulh to use u~ c~ll and dow~llc~l ~h~nn~.le to lla~ l~il to and receive from the head end, a variety of types of information, such as voice, video-telephony, data or control information that may be synchronous, asynchronous or sporadic. The invention also relates to wireless networks, which are akin to fiber/coax llclwulks, in that mobile stations do not usually directly listen to each other, but instead depend on a base station for feeclb~cl~

Background of the Invention There is ~ llly a great deal of activity directed toward enh~ncing the fiber-coax networking infrastructures, such as those used by cable television and telephone co~ -ies The thrust is two-fold: ~nh~ncing the dow~ll~ cal,a~ily ofthe lltlwul~ to support new services, and providing for significant u~ am capacity for newi~ a~,ti~services,includingtelephonyanddata~ctwulhing. Here,"u refers to h~nemiesion from a station to a head end (or central of fice), and "d~w~ ~" refers to tr~nemiesion from a head end to the stations.
Many current plans for providing u~sll~ll capability utilize the well known Time Division Multiple Access CIDMA) method of ~eei~ing bandwidth to stations that have invoked a ~eign~ling method to indicate to the head-end that they wish to ll~lslllil. However, e~ieting versions of TDMA do not provide as much flexibility in the use of the available bandwidth as is desired. Also, TDMA allocates peak bandwidth for the entire duration of â call, and thus does not take advantage of the statistical nature of the traffic.

Existing packet data solutions also do not extend well to the broadband cable environment. Carrier Sense Multiple Access/Collision Detection (CSMA/CD), as described, for example in Canier Sense Multiple Access with Collision Detection (CSMA/CD) Access Method and Physical Layer Specifications, published by the 5 Institute of Electrical and Electronics Engineers, ~m~rir~n National Standard ANSItIEEE Std. 802.3-1985, is arranged so that stations listen to the rh~nn~l for a period to clettorminç that it is IDLE before ~ g This type of ~rr~n~m~nt isvery inefficient in the cable el,vil. n".~ , because of the ~ t~n~es involved. The propagation delay, and corresponding dead time of the ch~nn~l~ is much greater in 10 cable networks than in the Local Area Networks that are the standard CSMA/CD
envi~ nent.
As another eY~mrle, the slotted Aloha approach described by L. Roberts in "ALOHA Packet System With and Without Slots and Capture", Col~uler Co.. ll.~ic~tions Review, April 1975, specifies a system arranged so that time is 15 divided into a series of exact slot periods. Any station that has data may !~ ;l at any time. When collisions occur, stations back off and l~,t~ l according to somerandomized algorithm. This approach has also been ~nh~nced with a reservation scheme in which once a station ac~luires a slot, it can keep succes.sive slots by setting a header field; the station may also signal that it has finished t~n~mittin~ by 20 rl~ngi"g that Seld. See, for example, Lam, S. S. Packet BroadcastNetworks - Ap~ llal~ce analysis of the R-ALOHA Protocol, IEEE Transactions on Comput~,.s, Vol. C- 29, No. 7, July 1980, 596-603. Here again, the problem is that in the coax/fiber i~L~ll,lcture under consi~l~tion~ it is difficult to allow stations to directly listen to each other. ALso, the unique tree and branch structure of the cable 25 ll~lw-~lk r~uiles that a new approach be used.
Also very important is the capability of an access protocol to serve the needs of both Synchronous Tran~sfer Mode (SIM) and Asynchronous Transfer Mode (ATM) applications. Many new applications, such as high speed data and MPEG-2 video, are likely to be transported using ATM techniques from a home or office. At 6 ~

the same time, applications such as voice and video telephony are likely to continue being transported using STM, for reasons of delay, error sensitivity, etc.
A technique known as DQDB, described in "IEEE 802.6: Distributed Queue Dual Bus Access Method and Physical Layer Specifications," is applicable for network 5 spanning a metropolitan area. However, DQDB employs a dual-bus structure and requires a well-defined upstream-downstream relationship between adjacent stations. It is possible to design a system to fulfill these architectural requirements for tree and branch networks, but the scheme would be very complicated and the cost would be extremely high.
Recently, a variation of DQDB, called Extended Distributed Queuing Random Access Protocol (XDQRAP), has been proposed by C. Wu and G. Campbell, "Extended DQRAP: A Cable TV Protocol Functioning as a Distributed Switch," Proc. of 1st International Workshop on Community Networking, San Francisco, July 13-14, 1994,pp. 191-198. One of the drawbacks of this proposal is that it does not support both STM
15 and ATM techniques. In its present form, it does not accommodate bursts of varying lengths, as required for STM connections. Further, the physical layer overhead (i.e., guard band and preamble) is very high, due to the requirement of multiple request mini-slots associated with each data slot.
Accordingly, there remains a strong need for an access protocol for broadband 20 tree and branch networks that adapts to the ch~n~ing demands of a mix of STM and ATM applications, and efficiently allocates bandwidth to a variety of bursty andisochronous traffic sources.

:~..7 Summary of the Invention In accordance with the present invention, a system and technique supports STM applications, e.g. voice and video telephony (VT), as well as packet mode (e.g.
ATM) applications, e.g. broadcast digital video, il~ d.;live television, and data, in S the context of multiple access on broadb~nrl fiber/coaxial cable nclw~lks. Further, the present invention adapts to the ch~nging d~m~n-le of a mix of STM and packetmode applications (e.g. ATM), and efficiently allocates bandwidth to a variety of bursty and isochronous traffic sources. Because the current trend is to ~u~oll packet mode applications using the ATM protocol, the following description will 10 generally refer to ATM. However, it is to be understood that other packet mode protocols may use this invention. If the br~lb~n-l llelwol~ and tPrmin~l~ evolve to full end-to-end ATM support for all services, then the present invention will ~u~ll voice, VT, and other services, all in an ATM format.
In the case of a coaxial cable or fiber tree and branch co..,.,.l..,ication 15 ~,lwol,~, the present invention is carried out both in cmtompr premises eqnirm~nt (CPE) at stations, and in a common controller, which may be a head end with which all stations co,l."lu ficate. A medium access control (MAC) processor provided in each of the stations and in the common controller or head end divides the time domain for a given digital bit stream ch~nn~l into a series of successive frames, each 20 having a plurality of time slots. There are two types of time slots, STM and ATM.
The frame is divided into two regions; in accordance with the present invention, the boundary b~ ~n the regions can be cllanged dynamically. Within the STM
region, variable length time slots can be allocated to calls, such as voice and video telephony, ~ ui,il,g di~ll amounts of bandwidth. A co"lelllion access sign~ling 25 ~h~nn~l is also provided in this region, for STM call control and set-up requests.
Within the ATM region, the time slots are of fixed leng~, each capable of accommodating one ATM cell. Further, the fixed length ATM time slots may be reserved for a particular user for the duration of a call, or may be shared through a collle"lion process. At least one cGlllelllion ATM time slot is always made available 30 for sign~ling messages related to ATM call control and set-up requests.

5 ~ ~ ~7~6~ 1 ~
The downstream time frame is structured in a manner similar to that used in the upstream direction, but includes an additional MAP (slot allocation mapping) field.
This field is used to communicate to the stations, ATM time slot allocations in the upstream channel. It also indicates to the stations which of the time slots, in the next 5 upstream frame, are open or available for contention, and which are reserved for designated stations.
Because of the unidirectional architecture of the up and down communication channels, individual stations do not communicate directly with each other, but can receive downstream broadcast messages (i.e., STM messages and ATM MAP
10 messages) indicating the status of each upstream time slot. These status messages are generated in the common controller or head end, and are transmitted in a broadcast downstream channel so that all stations can receive them.
For systems provisioned for STM traffic, one fixed length time slot in the STM
region of upstream frames is always used as a shared, i.e., contention-based, STM
15 si,~n~ling channel. Similarly, at least one time slot is always provided in the ATM
region in which ATM applications can transmit call set-up related sign~ling messages.
When an STM/ATM call arrives at a station, a si~n~ling message is sent from the station to the head end to request call set up and bandwidth allocation. In one embodiment of the invention, STM time slots are allocated closest to the beginning of a 20 frame, and ATM time slots are allocated closest to the end of a frame. This arrangement leaves the unassigned bandwidth in the center of the frame, i.e., between STM and ATM regions.
When an STM call terrnin~tes~ a "hole" is momentarily created in the STM
region. A "hole" represents released bandwidth that is currently lln~igned. This25 is taken account of by "repacking", which refers to reallocation of time slots to existing STM calls so that all time slots for the STM calls are placed adjacent to each other and closest to the beginning of the frame. In one arrangement, repacking is accomplished immediately after termination of an STM call, and the released bandwidth is used to create additional ATM time slots. In another arrangement, the 30 released bandwidth is added to the unassigned region of the frame, and made use of when a new call arrival occurs. In yet another arrangement, the repacking is deferred until it is cletennined that the released bandwidth due to several "holes" is adequate to admit a new call, but the holes are not contiguous to each other. Following arepacking action at the controller, in each of the arrangements, all stations are notified of the new STM and/or ATM time slot allocations by sending messages in S the STM ~ign~ling and MAP portions of the next dowl~l,~ frame.
In the ATM portion of an U~)Sl~dlll frame, a co..~ ion/les~ lion process is used. Throughput and delay pelru. .~n~e in multiple access nt;lwul~ with branch and tree network topologies are significantly improved by interleaving the ATM time slots within a f~ame structure, whereby the propagation delay constraint is ove.co~e, and by a rese,~lion process by which the head end allocates time slots to stations at a rate adequate to meet the delay requirernent specified by the station. Further details relating to how time slots are ~igne~l to dirr~ .el~ types of ATM calls are described below.
ATM calls can be of several types: (a) constant bit rate (ATM/CBR), (b) delay sensitive variable bit rate (ATM/VBR), (c) delay tolerant ATM/VBR, and (d)ATM/colll~lllion. An application such as voice or video telepho~y in ATM fo~n could be ca~gcjl~ d as ATM/CBR; an interactive data application is an example ofdelay se~ilive ATM/VBR; a file l~rcil is an example of delay tolerant AT~/VBR;
and the u~ messages for VOD and video games are examples of ATM/cont~ntion.
When an ATM call arrives at a station, the station tr~n~mit~ a sign~ling message to the head end, using an u~ ~.l ATM time slot that is available for co.,t~ ion, to specify the type of call. The head end applies dirr~,;e,.l processes in the h~n~lling of each type of call.
ATM/CBR calls are allocated ATM time slots by the head end, in accol~ce with their bandwidth l~ui~ ents. Delay sel~ili-~ ATM/VBR calls are allocated ATM time slots by the head end, in accordance with a statistically weighted incr~m~nt~l bandwidth (SWIB) de~e~ tion that takes account of eYi~ting ATM/VBR traffic and the statistical char~cteri~tics of the new call request. With respect to delay tolerant ATM/VBR, the head end can receive, from the station, any 7 21626Il information concerning burst length and delay tolerance. The head end can then assign ATM time slots at a rate that meets those requirements. For example, a station might not be assigned one slot in every frame, but rather may be allowed to transmit "m" ATM cells in every nth frame (m and "n" are integers).
In ATM time slots that are available for contention, the ATM/contention calls are processed in a manner similar to that used in the above cited Edmon, Li, andSriram patent application. Thus, acknowle~lgment or status messages may be L,~l,filled in a broadcast dowl~L~ çl~nnrl from the common controller or head end, and indicate a particular time slot status, which can, for example, be (a) IDLE, (b) BUSY/CONTINUATION, me~ning that a station has a need for contin~ use of a slot, (c) BUSY/FINAL, m~ning that the co~tim~ed use of a slot has come to an end, or (d) COLLISION, m~ning that a collision has occurred with respect to thatslot. The stations respond to the status messages to ~ e when to transmit the next ATM cell from an ATM/co~ ,Lion type of call.
An additional feature of the bandwidth allocation method of this invention is that ATM traffic is ~e.si~ a ...i~ u~leed bandwidth to meet a Quality of Service (QOS) l~uife.llent, which may, for example, pertain to the m;.x;.~
pf.. iLle-l cell delay or loss ratio. Further, additional "extra" bandwidth can be Rc~i~ne~l if spare bandwidth is available on the rh~nnrl~ to as many as ATM calls as 20 possible, to provide better than the specified QOS. In other words, given that some of the ATM traffic is bursty, such traffic may be allowed to use any spare bandwidth, in addition to its ~u~allleed bandwidth. However, the method of this invention allows the "extra" bandwidth to be taken away from the ATM connectionsto which it is ~ign~3, and used instead to admit a new STM or ATM call. Taking 25 away the "extra" ATM bandwidth potentially allows a new call to be ~-lmilterl but does not lower the QOS for the ongoing ATM calls below the required QOS. An example of QOS is the delay requirement specified by a variable bit rate (VBR) ATM call, such as a file transfer. If "extra" bandwidth is ~i~ed to this call, the file transfer may complete sooner than the specified ~ llin~o., and if the "extra"

8 ~ fi ~
bandwidth is reassigned to another call in the middle, the file transfer still completes within the specified ~le~(lline.
So far, our descriptions of bandwidth allocation and management focused only on the upstream traffic. The downstream STM and ATM time slots are assigned by the 5 head end in accordance with the bandwidth required for each call. There are many similarities between the upstream and the downstream traffic characteristics. Therefore, some of the methods specified by this invention are applicable for handling the upstream and downstream traffic. However, some differences do exist in these methods, because there is contention access for some part of the traffic in the upstream direction, 10 while the downstream bandwidth is shared by a combination of TDMA and statistical packet multiplexing only.
Note that in applications in which voice and video telephony traffic are transmitted in ATM form between each of the stations and the head end, all of the traffic carried in the cable distribution network may be ATM traffic. In this event, there 15 are no STM time slots, and the present invention is nevertheless useful in optimi7ing the handling of a multiplicity of types of ATM traffic.
While our invention has thus far been primarily described in the context of a fiber/coax based tr:~n~mi~sion infrastructure, it is to be understood that the invention is also applicable to a wireless communications environment. In the latter situation, the 20 terms "mobile station", "base station" and "wireless medium", would be applicable in place of the terms "station", "head end" and "fiber/coax medium".
In accordance with one aspect of the present invention there is provided a method of allocating tr~n.~mi.~ion bandwidth among multiple stations interconnected with a common controller via a tr~n~mi.~sion medium having a multiple access 25 upstream channel, said method comprising the steps of: dividing time in said upstream channel on said tr~n~mis.~ion medium into a series of successive time frames; dividing each of said time frames into first and second regions separated by a boundary region, each of said first and second regions cont~ining a one or more time slots; said first region consisting of one or more variable length time slots for STM (Synchronous30 Transfer Mode) calls, and said second region consisting of one or more fixed length time slots assigned to (a) reservation and (b) contention oriented ATM

8a (Asynchronous Transfer Mode) calls; and dynamically adjusting the location and size of said boundary region in each of said frames as a function of the bandwidth requirements of said stations, and for dynamically sharing said second region toaccommodate both reservation and contention types of ATM calls.
In accordance with another aspect of the present invention there is provided appal~Lus for allocating tr~n~mi.~.~ion bandwidth among multiple stations interconnected with a common controller via a tr~n~mi~ion medium having a multiple access upstream channel, said app~us comprising: means for dividing time in said upstream channel on said tr:m.~mi~sion medium into a series of successive time frames; means for dividing each of said time frames into first and second regions separated by a boundary region, each of said first and second regions cont:~ining one or more time slots, said first region consisting of one or more variable length time slots for STM (SynchronousTransfer Mode) calls, and said second region consisting of one or more fixed length time slots assigned to (a) reservation and (b) contention oriented ATM (Asynchronous Transfer Mode) calls; and means for dynamically adjusting the location and size of said boundary region in each of said frames as a function of the bandwidth requirements of said stations, and for dynamically sharing said second region to accommodate both reservation and contention types of ATM calls.

Brief Description of the D~
Fig. 1 illustrates the overall arrangement of a broadband network in which a plurality of stations are interconnected with a common controller called a head end, by a tree and branch tr~n.~mi~sion network, and in which the present invention may be used;
Fig. 2 illustrates an example of the upstream/downstream spectrum allocation that may be used on the tr~n~mi.~ion medium interconnecting the stations and head end shown in Fig. 1;

Fig. 3 illustrates the arrangement by which various communications devices in the stations of Fig. 1 are conn~cted to the tr~n~mi~ion network as well as the arrangement of the Mediurn Access Control (MAC) processor in accordance with our inventlon;
Fig. 4 illustrates the arrangement of a MAC processor similar to the one shown in Fig. 3, which is disposed in the head end of Fig. 1, and which directs con~ u lications signals b~ lw~en the stations and the a~r~l;ate service platforms;
Fig. S illusLl~les the arrangement of the data fields in an u~ ~ message contained within an ATM time slot;
Fig. 6 illu~ les the arrangement of the data fields in an dow~ll~ea~
message contained within an ATM time slot;
Fig. 7 illu~ les a typical arrangement of di~ .elll traffic types within an U~ ~ time frame;
Figs. 8 and 9 illustrate further details of typical arrangements of U~ ,alll and dow~l,~ time frames, respectively, and, in particular, show the positions ofheaders and other frame .~ign~ling messages;
Fig. 10 illustrates the details ofthe MAP field shown in Fig. 9;
Fig. 11 illu~ les, collce~tually, the information m~int~in~ in the bandwidth controller located in the head end, for the purposes of bandwidth allocation andmanagement;
Fig. 12, which is similar to Fig. 7, illustrates the de~ tions for various regions within each time frame and indicates certain constraints on these regions;
Fig. 13 is a process di~grslm illu~ Llg how an STM call arrival is processed at the head end;
Fig. 14 is a process diagram illu~ g how diLl. ~e~l types of ATM calls are identified and preprocessed; this diagram should be read in COllj ull~;lion with Figs. 15 through 21;
Fig. 15 is a process diagram illustrating how bandwidth allocation is performed for an ATM/CBR or ATM/VBR call;

lo 2162611 Fig. 16 is a process diagram illustrating how ATM extra bandwidth is reallocated;
Fig. 17 is a process diagram illustrating how bandwidth for a delay sensitive ATM/VBR call is detç~mined;
Fig. 18 is a process diagram illustrating the concept of Statistically Weighted Incr~ment~l Bandwidth (SWIB) for delay sensitive ATM/VBR calls;
Fig. 19 is a process diagram illusll~ling the ~lçeignAtiorls for various fields or regions within a series of time frames over which bandwidth allocation decisions are made;
Fig. 20 is a process diagram illu~ g how bandwidth allocation is p~,.r,~ ed for an ATM/CBR or ATM/VBR call over a bandwidth allocation cycle of the multiple frame periodicity shown in Fig. 19;
Fig. 21 is a process ~liA~m illu~lldling how an ATM/colll~llion call is processed at the head end;
Fig. 22 is process diagram illu~lldlillg how decisions are made regarding the bandwidth cull~lllly Asei~n~ to ATM/conlelllion traffic;
Fig. 23 is a diagram illu~lldting the process followed in the head end when a call t~ A~ this diagram should be read in conjull ;lion with Fig. 24;
Fig. 24 is a diagram illustrating the process followed in the head end when an ATM/co~ lion call is tc- Ill;l~AI~ and Fig. 25 is a diagram illu~ ling the process followed at a station when a tenninAl or device becomes active.

Detailed Desc~ption The type of bro~dbAn~l network in which the present invention may be used is shown in Fig. 1. A pluMlity of stations 107-1 to 107-8 are conn~ctecl to a comrnon head end 109 by a tree and branch llA~ -"i~ion ~tw~lk. The trAn~mi~ion network can include a main coaxial cable 102 having a plurality of taps 104-1 to 104-4, each of which serves a coll.,s~ollding feeder cable 103-1 to 103-4. Each feeder cable in turn serves one or more of the stations 107 via a tap 114 and a drop, such as drop 105. The network in Fig. 1 includes, for each seNing area, a single fiber node 101 which permits a transition between main coaxial cable 102 and an optical fiber 108.
However, it is to be noted that the entire network can consist of coaxial cable in some implementations, and in others, more than one such fiber node can be employed in5 the 11GtWU~h~ depending upon which portions of the nelwu,h are fiber and which are coaxial cable.
By virtue of taps 104, which are often directional, in the tree and branch arrangement illustrated in Fig. 1, individual stations cannot collllllullicate with or "hear" each other, and therefore, conventional access protocols cannot be used 10 efficiently. Rather, each of the stations COll~ U licates in the u~ ~ direction with a common controller located in a head end or central office 109, while the head end can broadcast dowl~llealll (on coax 102 and fiber 108) to all of the stations orll~slnil selectively to certain stations, using well known add~ssillg techniques. In the rest ofthe ~nming description, the terms "common controller" or "head end" are 15 used i.,l:~.tcha~geably, with the lln~ n(1ing that the common controller is located in a head end. ~Also, it may be noted that some impl~ ;0ns of the present invention may involve a central office in place of a head end. The arrangement of Fig. 1 is further assumed to include unidirectional, u~sk~ll and dowl~
amplifiers (not shown) that also prevent all stations from hearing all other stations.
20 A typical nelwulk of the type shown in Fig. 1 may involve distances in a range up to 100 miles from the farthest station to head end 109.
The traffic anticipated in coax/fiber ~twolhs of the type shown in Fig. 1 can be broadly classified as analog traffic, ~yllcl~onous tl~sr~l mode (STM) traffic, and a~ylldl.uno~ ~f~ mode (AI~I) traffic. An eA~Lulple of analog traffic is the 25 con~ tional analog CATV. Examples of STM traffic applications are voice, narrow-band ISDN, and video telephony, which require bursts of di~lent length and are normally multiplexed using Time Division Multiple Access (TDMA) techniques.
The ATM traffic can be further classified as (a) CO1~ t bit rate (ATM/CBR), (b) delay sensiliv~ variable bit rate (ATM/VBR), (c) delay tolerant ATM/VBR, and (d)30 ATM/colll~lllion. An application such as voice or video telephony in ATM form ' 12 2162611 could be categorized as ATM/CBR; an interactive data application is an example of delay sensitive ATM/VBR; a file transfer is an example of delay tolerant ATM/VBR;
and U~ l cignAling and VCR remote control messages for VOD and button push messages for video games are examples of ATM/col,l~"Lion. ATM/co"le"Lion traffic5 is çss~ntiA11y sporadic in nature, concicting of occasional bursts of one or more packets or cells.
Note that sometimes, delay il~c~ iLi~re traffic like that generated by a file transfer, may also bé served using an ATM/CBR connection. Note also that VBR
video can be transported through the network as ATM/VBR traffic, but it l~uiles 10 resynchlo,~lion at the l~iving station before display on a video monitor. Theasynchronous or VBR traffic is bursty, due to vAriAtinnC in activity/inactivity periods and potentially variable bit rates during the activity periods. Sporadic traffic is also bursty, but is char~ct~o-ri7~d by relatively long inactivity periods followed by bursts of one or a few packets.
As illustrated in Fig 1, calls requiring service beyond the area covered by head end 109 are ~ ...;ll~l by the head end to various service providers. These providers include Public Switched Telephone Network (PSTN) 111, ATM network l 12, as well as other similar fAri1iti~s~ such as other wide-area ~c;lw~lks. The CATV
antenna 110 is used for reception of cable TV pro~A.~ g, at head end 109, which may be trn~mitte(l to stations 107 in dowl~lLedm analog broadcast ~ nnel~ 201 onthe fiber/coax tr~n~mi~sion medium.
Fig. 2 illu~ tes an u~ ,~/downstream sp~~ allocation that may be used in the IlA-~",ic~ion l,~ lwul~ illl~C~ nl~ stations 107 to head end 109.
~sllming that a total bandwidth of at least 750MHz is available on all of the COIllpOl~ i of the l~twu,h, a 400 M~ band 201 from 50 MHz to 450 MHz may be used to carry coll~ lional dow~ ~ analog h~llll~ion, such as COll~r~ lllional cable TV progl~-",..ing A 40 MHz band bc;lw~ll S and 45 MHz can be divided into a plurality of U~ ~ll digital l~hAnn~ Ic 202-1 to 202-n, each of which can be 1 MHz to 6 MHz wide, and capable of carrying 1.6 Mb/s to 10 Mb/s digital bandwidth when 30 used with a suitable modulation scheme. Each of these U~ digital channels can be used for carrying control, ~ip;n~ling, voice, video telephony and other messages in the U~ ealll direction. A larger number, typically 50, of do~ll~llealll digital channels 203-1 to 203-m can be included in the 300 MHz band between 450 and 750 MHz. These ~ h~nnel.~ can be used for the downstream component of voice, video S telephony, video on dem~nrl (VOD), and other interactive (ITV) services. Due to the lower noise in the downstream ~h~nn~l~ (as compa.ed to the u~ eal~ h~nn~
with suitable modulation sl~hPm~os, it is possible to carry up to 28 (or even 40) Mb/s in each of the downstream ~h~nn~lc 203-1 to 203- n. The u~ ea~n l~h~nn~l~ 202 are shared by the stations 107 using a multiple access method such as described by the present invention. The do~~ ll rh~nnPI~ 201 and 203 are "broadcast" ~h~nntol~, in that all stations 107 can receive the messages tr~n~miltecl in those ~h~nn~l~. It should be noted that the present invention can operate with any spccllulll allocation and bit rates, including and other than those discussed in the context of Fig. 2.
Before describing the specific arrangement of the stations 107 (Fig. 3) and head end 109 (Fig. 4), and the specific components in ~sll~ and dow~ll~
messages (Figs. 5-10) e~ch~nged belw~n the stations and the head end, it will beinstructive to present an overall view of the operation of these elements in connection with the present invention. In accor~ce with the principles of the present invention, bandwidth controller 335 in MAC processor 330 in stations 107 in Fig. 3, and a similar bandwidth controller 435 in MAC processor 430 in head end 109 in Fig. 4, are ~l~ged to implement a protocol with respect to information tr~n~mi~ion on the i,l~ co~ecting tree and branch network, v~/ a station 107 can insert il~llllation in a particular tirne slot in an u~ ~ rh~nn~l as long as the time slot is either (a) allocated to that station, or (b) a co..l. .~ion slot. A station where an STM
25 call arrival occurs, initially ~ sses the STM ~ign~ling slot (see Fig. 7) via a colllell~ion process. Then, the head end allocates STM time slots for that station in the U~ and dowl~ll~ frames (see Figs. 8 and 9). When an ATM call arrival occurs at a station, the station 1~ its call set-up messages using a contenLion ATM slot (see Fig. 7). For an ATM/CBR or ATM/VBR call, the head end allocates 30 some y,u~lleed and possibly some "extra" ATM slots to the station, depending on its bandwidth requirement (see Fig. l 2). The guaranteed and extra ATM slots are"reserved" for the ATM call until the call ends or the head end intervenes to make some re-~e.eignments. When a station is said to have "reserved" time slots, the term reserved is used loosely, in the sense that the same number of time slots are allocated 5 periodically over a bandwidth allocation cycle; however, it does not neceee~. ;ly mean that specific time slots are allocated in fixed positions within each cycle. A
bandwidth allocation cycle refers to either one frame or multiple cl nee~ ;vt; frames over which the bandwidth allocation periodicity occurs. The "extra" ATM slots refer to extra bandwidth that is mom~nt~rily ~eeigned to an ATM/VBR call, bec~use of the 10 availability of some 1m~eeign~1 or idle bandwidth on the link. Whenever a newSTM or ATM call set-up requires some ~ua~allleed bandwidth, some or all of the extra bandwidth can be taken away from the calls that cullelllly have it.
The ATM/co~ .,1ion traffic, after going through a call atlmieeion process, does not individually get any ~,u~d~ ed bandwidth allocation. Inete~A such a call 15 l~s,,,,l~ its messages using the ATM time slots deei~ted for "contention" by the head end (see Fig. 7). The method used for this contention process is based on that described in the previously cited patent application by Edmon, Li and Sriram (with some variations as described later in this disclosure). The dow~ ~l STM
messages and the ATM MAP, co..L~ g information about 1~ s~ilv~lion and 20 co~ "lion status of ATM time slots, are 1~ ~1 in the STM ~eigr ~lin~ field 902 and ATM MAP field 920, l~~ livt;ly, of a downstream frame (see Fig. 9).
In the co. ,1. . .1 ;on process, any station that tries to l.a~""l in an ATM/co~ ,l,lion time slot reads the status i~ lalion associated with that time slot that is c~ 1 in a broadcast do~ nn~l If a time slot is not L~. ~,cd, a 25 station can insert inforrnation. If a time slot is reserved for another station, the station ~bst~ine from l~ g in that time slot. If the station detects that its o~-vn ."i~;on c A~ficnced a collision in the p~ g frame, then the station m~ting the tr~nemiee~ion executes a backoffand leh;~ process.
Four contention time slot conditions, which are ~liecllssed in more detail 30 below, may be indicated in a downstream status or acknowleclgm~nt message which is part of the MAP 920 (see Figs. 9 and 10). First, a slot may be IDLE, meaning that the slot was unused in the last frame and is available for tr~n~mi~sion by any station.
Second, the slot may be marked as BUSY/CONTINUATION, me~ning that the slot was busy in the last frame and is reserved in the next frame for the same station.
5 Third, the slot may be marked as BUSY/FINAL, me~ning that the slot was busy inthe last frame, but is available in the next frame for any station to contend. Finally, the time slot may be marked as COLLISION, indicating that it was corrupted because the slot did not have meaningful data due to collisions or noise.
Further in accordance with the invention, stations that have tr~n~mitted in a 10 slot may set a co.,~ ion indication in the bandwidth request field (BRF) 509 in Fig. S in an U~)Sll~ln header, to indicate reservation of the same time slot in the next frame. This provides a mech~ni~m for stations to obtain an initial cont~ntion-based access, and then reserve the same ATM co~ ,llion slot over multiple frames to transmit messages that are multiple ATM cells long. However, if n~cess~ head 15 end 109 can exercise overriding control over the station using the c~ lion indication.
Many details of the method of this invention are described via flow charts and diagrams in Figs. 11-25. These details include specific ways of making time slots allocations for STM and ATM calls, dynamic bandwidth allocation across mnltirle 20 traffic classes in the STM and ATM regions, management of the ATM/contentions traffic, bandwidth release and reallocation when calls t~ , etc.
Referring now to Fig. 3, there is illll~tr~ted the arrangement by which various CO~ul~u ~ications devices in stations 107 of Fig. 1 are co...~P~ d to the tr~n~mi~ion network and can, in dccolda~ce with the inventiorl, access tr~n~mi~ion capacity in u~ ealn ~nn~l~ 202 in order to c~ll.l.,l.. ;cate with head end 109. The co~ ~cations devices in a station may include a telephone 301, a television 302, a co~ ul~ 303, and many other devices. Each device is connPcte~l, for example, to a cornmon bus 320 via an associated device specific processo~ 311 -313. Thus, telephone 301 is coupled to bus 320 via processor 311, which, for example, converts PCM voice to analog signals; television 302 is coupled to bus 320 via processor 312, which, for example, converts an MPEG-2 digital TV signal to an analog signal; and computer 303 is coupled to bus 320 via processor 313 which, for example, provides conversion between user and network protocols.
With respect to U~S~ l information, each of the device specific processors 311-313 processes the data ~riginAte l in its associated collllllu~ication device, and presents the data in the proper form for application to a digital l1A'~ . . 350 that enables a connection to cable 102 via a splitter/combiner 360. With respect to do~ lL~ll information, each of the device specific proc~s.sors 311 -313 processes data destin~oA for display/play-out at the associated con~ ullications device (telephone 301, television 302, COlll~U~ 303, etc.). For ~mple, processor 312 connected to television 302 may do MPEG- 2 decoding of video information.
Transfer of data belw~ll bus 320 and a media access control (MAC) processor dç~ignAted generally as 330 (described in more detail below), as well as b~lweell device specific processors 311-313 and bus 320, is performed under the control of a bus controller 322. Bus 320 and bus controller 322 may be arranged to use a standard time slot i~ cllange (TSI) or Ethernet protocol, or any other suitable arrangement that will be a~palelll to a person skilled in the art.
A digital 1~ 350 and a digital recei~ 340 are provided in each station to modulate u~slleam infollllation and demodulate downstream inf ~rrnAtion, r~s~ilively. For example, digital l~i~ . 340 and !.A.. ~ 350 may be ~ldllg~l to convert an analog signal to a digital bit stream (and vice-versa) using any of the well known techniques such as QPSK, QAM, etc. Splitter/combiner 360 o~.~t. s in both directions; it extracts the signal l~;v~d from head end 109 on coax 102, and feeds it to digital lecei~cl 340, and inserts hlrollllation l~ivc;d from digitall1A~ 350 and CATV ~nttonnA 110 into the coax ~ ulll for tr~n~mi~ion to head end 109. It is to be noted here that the functions pc.rolllled by rc~i~,~,. 340, ll~lllill~l 350 and splitter/combiner 360, and potentially MAC processor 330, can be included or embodied in a single unit, sometimes referred to as a "Network Interface Unit".

In accordance with the present invention, information origin~ting in any of the cornmunications devices 301-303 connected to bus 320 and destined for the head end 109 of Fig. 1 is applied to cable 102 and to the other components of the tr~n~mi~ion network under the control of MAC processor 330, which implements 5 the MAC layer protocol of the present invention. Likewise, downstream information from head end 109 that is dçstinç~ for the devices 301-303 is received from the tr~n~mi~sion network and cable 102, also under the control of MAC
processor 330. The elements of MAC processor 330 are a frame demIlItirlexer 333 and a coIlG~ollding frame multiplexer 338, both of which operate under the control of a bandwidth controller 335, ATM and STM buffers 331 and 334, ATM and STM
MAC processors 332 and 336, and a program store 337 that coll~ s the program or software code that controls the operation of bandwidth controller 335.
Frame demux 333 reads the appIopliate ATM and STM time slots in downstream frames (see Fig. 9), and sends them to the ATM and STM MAC
processors 332 and 336. The ATM MAC processor 332 sorts received ATM cells destined for station 107 from other traffic, and then sepa-~les payload data and the MAP messages (920 in Fig. 9) sent from head end 109, and sends them to ATM
buffers 331 and bandwidth controller 335, I~spe-;tively. From ATM buffers 331, the payload data is sent to the a~Iopliate device specific processor 311-313 over bus 320, depending upon the ~lestin~ion address for the collllllullications device 301-303 that is coll~i~ed in the message. For the u~ traffic, ATM MAC processor 332 plocesses the il~Illlalion from devices 301-303, and puts it in an ATM format with the applopIiate MAC layer o~ ad (see Fig. 5), and p~selll~ the same to frame mux 338. Frame mux 338 ~ - r(.. ~ the inverse operation relative to frame demux 25 333, writes the ATM and STM mpss~ges into the appropIiate ATM and STM time slots"~ ly, in ulJSll~ll frames, using il~llllalion ~ /ed from the bandwidth controller 335. The control/status messages are described below in com~eclion with Figs. 5-10.
Both frame mux 333 and demux 338 operate in conjunction with bandwidth 30 controller 335, which performs processes that include (a) scheduling tr~n~mi~.sion of ATM cells or STM bursts from buffers 331 and 334 via ATM and STM MAC
processors 332 and 336 to frame mux 338, and from frame mux 338 to digital transmitter 350, and (b) coortlin~ting the STM and ATM MAC processors. ATM
MAC processor 332 performs actions such as (a) generating call set-up requests S when calls are generated by devices 301-303, (b) controlling the co~lenlion access process, (c) following the instructions of the bandwidth controller to schedule ATM
traffic, and accordillgly s~n-ling ATM cells from the ATM buffers 331 to frame mux 338, (d) lCCcivillg and fol~ding data in downstream ATM cells to device specificprocessors 311-313, and (e) generating and l~.h~rl~ing cyclical redlln-l~ncy codes 10 (CRC's) for Ul ~llC~ll and dc wl~L~ca ll ATM traffic, lc~lJe~;(ivcly. STM MACprocessor 336 ~, rQ~ processes similar to ATM MAC processor 332, with some differences as required for the STM traffic by the overall MAC protocol of the present invention. The details of these processes, which, as stated previously, are performed under the control of software programs stored in a program store 337, are 15 described below in connection with Figs. 12 through 25. The actions in bandwidth controller 335 are coordinated with those in bandwidth controller 435 in head end 109 to ~lrOllll the processes described in Figs. 12 through 25. Buffers 331 and 334 are arranged to store ATM and STM payload data on a tel,lpol~ y basis, and to store parameter il~f~ ion used in the bandwidth allocation and management processes 20 that form part of the MAC protocol.
Fig. 4 illu~l,dtes the arrangement of a MAC processor 430, which is similar to MAC processor 330 shown in Fig. 3. MAC processor 430 is disposed in head end 109 of Fig. 1, and directs co~ ...ications signals from st~tion~ 107 to appropliate service platforms 490. As shown in Fig. 4, signals ~ ~l from stations 107 via 25 coaxial cable 102 are applied to splitter/combiner 472, which, like unit 360, Ç~,t~:
the U~l~ signals from coax 102 and ples~ls those signals to digital lccei~ 470.
Splitter/combiner 472 also inserts signals received from digital l . i~ 474 onto the coax 102 for tr~n~mi~ion to stations 107. Digital lccei\ ~l 470 may also provide COLLISION or IDLE status detection, by using (a) power level analysis, and/or (b) 30 digital error detection, and/or (c) other techniques. If these functions are performed in digital receiver 470, the resulting control signals are communicated to the ATM
and STM MAC processors 432 and 436.
Frame demux 433 reads the applol";ate ATM and STM time slots in upstream frames (sw Figs. 7,8) and sends them to ATM and STM MAC processors 432 and 436. ATM MAC processor 432 separates payload data and the ~ign~ling messages sent from stations 107, and sends them to ATM buffers 431 and bandwidthcontroller 435, ~ Aiv~ly. ATM MAC processor 432 also distinguishes b~lw~ell leS~ ~lion and colllclllion ATM cells, and processes them dirr~l~ lltly, in accordance with the methods specified in Figs. 12-25. A call set-up request is processed bynegotiation b~twwll MAC processor 430, service y,~t~;w~y 486, and service platform 490, and thus a connection is established bclwwn a station and a service platform.
After call establishment, the payload data from ATM buffers 431 are sent to the a~)plopliate service platforms 490 via the service g~w~ 486.
For the dow~l~ l traffic, ATM MAC processor 432 processes inform~tion, which ~lL~ins to an ATM call and is receivtid from a service platform 490, and puts it in an ATM format (if not already in that format ) with the appropliate MAC layer overhead (see Fig. 6), and presents the same to frame mux 438. Frame mux 438 p. lr~lms the inverse operation relative to frame demux 433; it wvrites ATM and STM
information into the appl~pliate time slots ~c~ign~d to ATM and STM calls, and the control/status messages into the MAP field in the dowl~ll~ frame (see Fig. 9).
The insertion of ilrrollllation messages into specific time slots of a dow~ll~ll frame occurs in accol~ce with time slot allocations made by bandwidth controller 435, and collllllullicated to the STM and ATM l~luce~ 436 and 432. The control/status lllessages are described below in connection with Figs. 5-10.
Both frame mux 438 and frame demux 433 operate in conjunction with bandwidth controller 435 and ATM & STM MAC processor~ 432 and 436. The tasks accoi"pli~hed by the bandwidth controller 435 are: (a) call a~lmi~sion control and bandwidth allocation on the fiber/coax in u~ll~ll and downstream direc~ion~,(b) scheduling tr~n~mi.~ion of ATM cells or STM bursts from buffers 431 and 434 via ATM and STM MAC processors 432 and 436 to frame mux 438, and from frame mux 438 to digital transmitter 474, and (c) coor~in~ting the STM and the ATM
MAC processors 436 and 432. ATM MAC processor 432 performs actions such as (a) generating ATM call set-up requests when calls are generated by a network (111, 112 in Fig. 1) via service platforms 490, (b) controlling the ATM contention access S process and gen.,.~ g acknowle~1grnent~ for inclusion in downstream MAP 920, (c) following the instructions of bandwidth controller 435 to s~ edl-le ATM traffic, and accordingly sending ATM cells from the ATM buffers 431 to frame mux 438, (d) receiving and folw~udillg data in uy~ll~n ATM cells to service galcw~y 486 via ATM buffers 431, (e) generating and ~ ng cyclic redlmtl~ncy codes (CRC's) for 10 dowlLsL-e . and u~ ea~l ATM traffic"especliv~ly. STM MAC processor 436 ~.rOIllls processes similar to ATM MAC processor 432, with some dirr~ ces as required for the STM traffic by the overall MAC protocol of the present invention.
The details of these processes, which, as stated previously, are performed under the control of software programs stored in a program store 437, are described below in connection with Figs. 12 through 25. Buffers 431 and 434 are arranged to store ATM and STM payload data on a telll~oldl~ basis, and to store parameter illrollllalion used in the bandwidth allocation and management processes that form part of the MAC protocol.
Service g~lcw~ 486 shown in Fig. 4 m~int~in.c information regarding 20 sllbs~ribers, service providers, and routing information for their services platforms, and provides an int~ e l~lw~n MAC processor 430 and service platforms 490.
These service platforms can be located at sep~le sites apart from head end 109, and therefore illu~l~alivcly are illl~ ~led with ~,alc~a~ 486 via a public switched teleco.. ~ications~clwu~k(PSlN) 111 oranATMnetwork 112(seeFig. 1).
25 G.-tcw~ 486 O~lalcs in conjull.,tion with the MAC processor to reco~i7~ the needs of di~ .ll applications ori~in~ting from stations 107, and routes co.. ications signals coll~;.~l~ding to these di~r~,.c.ll applications to appr~l;a~e ones of the service platforms 490. For example, video on ~leTn~nrl (VOD) requests origin~tç~ in a station 107 would be routed through service g~lcw~y 486 to a VOD service platform, while voice calls would be routed through the gateway 486 to a LEC
switch, which acts as a different type of service platform for voice calls.
Referring to Fig. 5, the basic arrangement of the data contained in a time slot is shown. A preamble 501 allows syncl~loni~lion ofthe receiver in head end 109 to the time slot. A source address (SA) 505 allows head end 109 to identify the particular station, co~ llullications device 301-303 or device specific processor 311-313 from which the tr~n~mi~eion ori~in~te l A bandwidth request field (BRF) 509 generally contains information about the bandwidth requested by a call. BRF field 509 may be an actual bandwidth r~ui~e~lent value, or it may specify the type of call based upon which the head end makes a decision rcg~dillg the bandwidth ~ignment for that call. Stations ~ g coll~ntion traffic may set a subfield in BRF field 509, referred to as co,~ ;on bit, to a value "one", to indicate a request for use of the same time slot in the next frame. When the station does not need that colll~.llion time slot in the next frame, it sets the col,l;".li.lion bit to a value "zero".
Although not limited to any particular protocol, the payload 515 in a time slôt can accommodate an A~yllclllollous Transfer Mode (ATM) cell, which is 53 octets long. A Cyclic Redl-n~l~ncy Code (CRC) 517 is used for the pu~pose of error ~~ detection and correction, and covers all of the slot user data in a manner that would be well known to pel~ons skilled in the art. The use of CRC 517 is optional, i.e., not always required in the implementation of this invention.
Finally in Fig. 5, a guard time 500 is included in each time slot, at the begim~il~g and end ofthe time slot, to allow for l~ . turn-on/offtimes, and time di~,iences caused by plopagalion delay as messages are carried by the ll~ ;on medium from station 107 to head end 109. All stations 107 on the coaxial cable 102 initially go through a coarse ranging process with head end 109, so that they have the same view of timing with respect to the time slots that appear on the tr~ncmi~cion medium provided by coaxial cable 102 and fiber 108. This timingsynclll~ni~alion, which is l,. .~lmed in accordance with techniques well known to persons skilled in the art, is then fine tuned with the help of the guard time 500 and preamble 501 within each time slot. The preamble 501 is a preset digital sequence which allows the head end digital receiver 470 to rapidly acquire clock and synchronization information. All of this allows stations 107 to stay synchronized with the slot boundaries on the tr~n~mis~ion medium, while accounting for their 5 distances from each other and head end 109, and the corresponding propagation delays.
Referring to Fig. 6, there is shown the arrangement of the data fields irl a dc wllsll~ll message co~ ed within an ATM time slot 908 of a dow~l~ frame (see Fig. 9). Field 603 co~ s the destination address for the payload in the time slot, and field 604 co~,~ins the payload. Finally, field 60S contains a cyclic redlm-1~ncy code, used for error detection (and correction) purposes. The use ofCRC 605 is optional, i.e., not always required in irnplement~tion of this invention.
Fig. 7 shows the time slot allocations in an u~sll~ frame-700 at a typical instant. This figure also illustrates the di~l~ types of STM and ATM services that are supported by the access method ofthis invention. In the STM region 702, STM
~ign~ling iS col~t~ined in time slot 706, which is shared by colltel~lion access by all stations 107 for STM call si n~ling. Once bandwidth is allocated by head end 109, an STM call makes use of a periodic time slot allocation in each frame. For example, in Fig. 7, a video telephony (VT) call, requiring 384 Kbps bandwidth, uses a relatively long time slot 708, while three separate voice calls, each ~ g 64 kb/s, use somewhat narrower time slots 711, 712, and 713.
Also in Fig. 7, the ATM region 704 of the frame 700 is partitioned into several ATM time slots, each capable of ~l~commodating one ATM cell. These cellsare labeled as "R" (Reservation) or "C" (Contention), col,es~nding to the mode in which they are being used in this frame and as dictated by the head end bandwidth controller 435. An ATM/CBR data call ~ ing a guaranteed fixed bandwidth uses a reservation time slot 718. Such time slot is made available to the call in a periodic manner to provide a fixed bandwidth for the duration of the ATM/CBR call. An ATM/VBR delay sensitive call also uses a reservation time slot 720. However, this time slot may not be reserved for the entire duration of the call. Tn~te~-l, it may be reserved only for a burst of data (i.e., group of ATM cells) that is ~;ullclllly being generated by the ATM/VBR source. Time slot allocations for this burst is requested in an upstream ATM/contention time slot such as si n~linE time slot 722, and theallocation acknowledgment is received in the next dow~ ll ATM MAP 920 (see S Fig. 9). When this burst is ll~lllilled completely, the periodic allocation of the time slot 720 is t~rrnin~t~ for this ATM/VBR call. When the same ATM/VBR
source produces another burst of data at a later time, the source would then seek allocation of a new lesel~alion ATM time slot (periodic over one or more frames).
An ATM/VBR delay insensitive call, such as a file 1. nl .~r. .., also uses a lesel ~alion 10 time slot 714. Due to its delay in~on~itive nature, a file ll~r~. call may be ~igne~
only one ATM time slot in every n th frame where "n" may be large.
In Fig. 7, interactive TV (ITV) and ATM ~i n~ling traffic use co..l~.";on time slots 716 and 722. ATM si~n~lin~ traffic co~ onds to ATM call set-up related messages. Examples of ITV are U~ elilll remote control signals from video on 15 clem~n~l (VOD) or button push signals in a video game. Although not shown in Fig.
7, if more than one station transmit in a co,lltlllion time slot simlllt~neously, a collision would result, and then the stations would apply a coll~"lion resolution scheme.
Now referring to Figs. 8 and 9, it may be noted that in the u~ ,~ direction, 20 stations 107 transmit in a "burst" tr~n~mic~ion mode, and in the d~wl~
direction, head end 109 ll~lllil~ in a "continuous" tr~n~mi~ion mode. The stations go through a ranging process and l,~lllil or "burst" based on timing i~fo. ~ I;on ~ d from the head end. Each station uses guard times and preambles (see Fig. 5) to avoid hll~r~ ce with ~dj~nt time slots in which another 25 station may ll~LsllliL However, in the dowl~ll~ll direction, the tr~n~mi~ion ~~h~nn~.l is controlled and used only by head end 109. T~ ef~ " there is no need for guardtimes or preambles, and ~e tr~n~mi.c~ion takes place in continuous way.
In Fig. 8, there is shown an Ul~llealll frame 700, similar to that of Fig. 7, showing overheads associated with STM and ATM time slots. Associated with each 30 ATM and STM time slot is a burst header 812, which includes guard time 500 and prearnble 501 shown in Fig. 5. Burst header 812iS typically of the sarne size, whether it is an STM burst of a ~i~n~ling message 802, or a DS0 804, or an NxDS0806 call, or a burst corresponding to an ATM cell tr~n~mi~sion 808. The source address 505 and BRF 509 of Fig. S constitute the MAC message 814 in an ATM
5 burst (Fig. 8). The payload in an ATM burst is an ATM cell, which consist3 of an ATM header 816 and an ATM payload 818. The boundary 801 bctwccll the STM
and ATM regions is movable or dynamic, depending on arrivals and de~allu,~3 of STM and ATM calls. Certain rules for the movement of this boundary are ~.Ypl~in~below, in the context af Figs. 12 through 25.
In Fig. 9, there is shown a dow~LI~ frame. The STM region (to the left of boundary 901) con~i~t~ of time slots corresponding to STM ~ign~ling 902, DS0 904 (e.g. voice) and NxDS0 906 (e.g. video telephony), and the ATM region (to the right of boundary 901) consist3 of one or more ATM cells 908 which include headers 916. MAP field 920 carries control and status information from the head end 109 to 15 the stations 107. It indicates the status of con~cl-lion time slots in the previous frame and also informs stations of the bandwidth (or time slot) allocations in the next uysllca~l~ frarne.
The boundary bclwccn the STM and ATM regions of the frame shown in Fig.
9 is dynamic in a manner similar to that described above for the u~sll~ frame in20 Fig. 8. The locations of the boundaries in the u~L~ ll and do-wlLsll~Lu frames are each dete .,~in~cl indepçnrl~ntly by the bandwidth allocation decisions made in the head end 109.
Now l~,Ç~lillg to Fig. 10, there is shown the col~lr.~L~i of a MAP field 920. Ifthere are "m" ATM slots in the ATM region of a frame, there would be "m"
25 se~nent~ 1012 in the MAP field, each of which co..~ollds to one ATM time slot in ~e dow.~ll~. frame. As mentioned earlier in this disclosure, the ATM slots are located as close to the end of a frame as possible. Accordingly, the m th ATM time slot is closest to the end of a frame, the (m-l) th time slot is the second closest to the end of a frame, and so on. The number "m" may change from one frame to the next but the length of the MAP field provides that information to the stations. The fields within a segment 1012 of the MAP are also shown in Fig. 10.
The SID field 1014 in the i th segment 1012 provides the station ID of the station which either had a s~lccesefill tr~nemi~sion (via contention access) in the 5 previous col,G~onding tirne slot, or to which the time slot in the next frarne is allocated by head end 109 via a reservation process. The Reservation/Contention Indicator (RCI) field 1016 indicates whether the slot is reserved ("R") or open for co,llcl,lion ("C"). The Upsllc~l Slot Occllp~ncy Status Indicator (USOSI) field 1018 has di~ lcl,l mP~nin~e depending on the value of the RCI field 1016 in the 10 same segmPnt 1012. The head end can change the value of the RCI field from C in one frame to R in the next frarne or vice-versa. In the case of a co"lc~llion slot (RCI
= C), the USOSI 1018 in~ ~tes one of the following conditions: (a) IDLE, (b) COLLISION (or equivalently noise), (c) BUSY/RESERVED, (d) BUSY/FINAL.
The definitions of these conditions are similar to those stated in the earlier cited 15 ,ef~ ce (Edmon et al.). When the RCI value is R, i.e., the time slot is ,~ se.~cd, then the USOSI field may be left unused in some irnplernPnt~tions. However, in another irnpk~ c ..~ .~;on ofthis invention, the USOSI 1018 may be used to inform the station of any additional p&l~cl~ eeori~ted with bandwidth allocation and control. For example, the head end may indicate to the station about the actual bandwidth allocated, i.e., the periodicity and nurnber of ATM cells that will beallocated in the frames that will follow.
In the above description, it was stated that the ATM MAP 920 inform~tion is of variable length depending on the number of the ATM time slots in an u~u frame. In another re~li7~tion of this invention, a fixed length MAP 920 (in Figs. 9 and 10) can be used. In such a re~li7~tion, MAP 920 can always have a fixed number of se~mPnte col,~s~ding to the largest anticipated value of "m" (see Fig.10). For cases when one or more ofthe segmpnte 1012 are not meant to contain valid MAP inform~tion (i.e., when "m" is smaller than the its largest anticipated value), then a special value in the SID field will indicate that the said ~egmPnt 1012 is invalid and to be ignored (i.e., there is no ATM time slot corresponding to that segment). A fixed length MAP designed and used in this manner may be desirable in some systems that use the present invention.
In Fig. 11, there is shown the schematic diagram of the view m~int~ined within bandwidth controller 435 at head end 109 for the purpose of bandwidth S allocation and call a lmi~ion decisions. Bandwidth controller 435 (a) receivesu~ ~ull messages, such as messages contained in fields 706 and 722 of Fig. 7, and in field 814 of Fig. 8, which include ATM and STM call setup ~ ling messages, bandwidth requirements for each call, and buffer fill values at each station 107, and (b) generates dow~ll~ sign~lin~ messages such as those cont~ined in STM
~i~n~ling field 902 and ATM MAP 920 of Fig. 9. Based on the inforrnation in the u~ c~l messages, bandwidth controller 435 lI.~ status related to di~el~,nl types of calls. The STM and ATM/CBR calls require dir~ nt amounts of fixed bandwidth for the duration of a call and, accordingly, the bandwidth allocations of these calls is ,~ ed in register 1112. In registers 1101-1 through 1101-n, buffer 15 fill status of the active ATM/VBR delay sensitive calls is m~int~in~ Similarly, in registers 1102-1 through 1102-m, buffer fill status of ATM/VBR delay tolerant calls is Ill~ ;nç~ The number and status of stations using ATM c ntPntion time slots is m~int~ined in register 1114. The registers shown in Fig. 11 may be part of the bandwidth controller, or they may reside in the ATM and STM buffers 431, 434 (see 20 Fig. 4).
Based on this overall view of di~l~ lll types of calls m~int~in~d in the di~ l registers, bandwidth controller 435 makes time slot allocation decisions for the next fsame or the next several frames. These time slot allocation decisions are ~.~.. ~icated in a dow.~ nnf.l to each ofthe stations via downstream sigr1~lin~ 902 and MAP 920 fields in a dowllsl.~l~ frame (see Fig. 9). The p,ocesses used in bandwidth controller 435 for time slot allocation decisions are described in the flo-w~ hall~ of Figs. 13 through 25.
In Fig. 12, there are shown the various regions and boundaries coll~s~ollding to STM and ATM connections within a frame. In this figure, the different symbolsrepresent the bandwidth allocation parameters. The parameter C lC~leSe;llls the overall bandwidth in a channel, and is proportional to the duration of a frame. The pararneters BS and BA represent the bandwidth allocations for the STM and ATM
regions, ~ e.;lively. The parameter U represents the nn~ ned bandwidth. The parameters QS and QA represent the m;1xi~n~ . limits on Bs and BA~ respectively.5 The parameters GA and XA repleselll the ~ual~ulleed and extra bandwidth allocations for the ATM calls. All these parameters are illl.~ ed in Fig. 12 relative to a frame, l~use the bandwidth allocations are pfol)ollional to the collc;~ ding lengths ofthe regions within a frame (in terms of time). The process flow ~ ms of Figs. 13-25 make use of the notations of Fig. 12 to describe the actions performed at each 10 step. It is to be understood that the following co~

BS<QS
BA = GA + XA ' QA
BA+BS+U=C

are always satisfied in the flow diagrams of Figs. 13-25, although they may not be shown explicitly in each diagram.
Fig. 13 shows the process associated with STM call ~timi~ion and bandwidth allocation, which is followed in the bandwidth controller 435 and STM MAC
20 processor 436 (see Fig. 4) in head end 109. Following the arrival of an STM call at step 1301, head end 109 le~ives a signal about the type of call and det~rmin.o.s its l~uh~ d bandwidth, Ri, at step 1302. A decision is made at step 1303 to dt;l~ e if the call can be admitted using the ~ llly nn~sign~d bandwidth. If the result is yes, then a d~t~ on is made at step 1304 to check whether or not the limit on 25 total STM bandwidth is still satisfied after ~ mitting the call. If the result at step 1304 is positive, then in step 1310, an STM time slot is allocated to the call, and a colle;,pollding STM message is l~ ;lled to the station which ~rigin~t~l or which istoreceive,thecall. If theresultatstep 1303 isnegative,thenatstep 1305 a d~ 1ion is made to check if the call can be ~lmitted by adding the current extra 30 ATM bandwidth to the lln~igned bandwidth. If the result at step 1305 is positive, 28 21 62Sl I
-then the constraint on the maximum STM bandwidth is checked in step 1306, and ifthe result is still positive, then a decision is made to admit the STM call and the process proceeds to step 1307. If the result at step 1306 or 1305 or 1304 is negative, then the call is rejected.
At step 1307, the process proceeds in the following manner: (a) first, just enough ATM time slots from the XA region of the frame are taken away to create the necess~ ~ bandwidth for the STM call (i.e., making use of U and as much of XA asneeded), (b) the value of XA is updated, (c) the values of the STM, ATM and lm~si~n~l regions (i.e., Bs, BA~ and U) are updated, (d) the process of Fig. 16 is applied to reallocate the modified ATM extra bandwidth to eYi~tin~ ATM/VBR calls.
The process proceeds from step 1307 to step 1310, where an STM time slot is allocated to the call, and STM messages and ATM MAP are ge~ led to inform all stations of the new time slot allocations.
Fig. 14 shows the process associated with ATM call ~ sion and bandwidth allocation which is followed in the bandwidth controller 435 and ATM
MAC processor 432 (see Fig. 4) in head end 109. Following the arrival of an ATM
call at step 1401, head end 109 lcceiv~s a signal about the type of call at step 1402.
Based on the ~ign~ling i~ alion received in step 1402, the head end identifies the type of ATM call in step 1403. There are four possible outcomes associated with step 1403. If the type of call is identified at step 1403 as ATM/CBR, then the process proceeds to step 1404, where the bandwidth required, Ri, to support the call is ~l~t~ ....;,.~A based on the ci~n~ling h~follllalion leceiv~d at step 1402. No extra bandwidth is r~uil~d for an ATM/CBR call, and th~ ,rw~, the extra bandwidth allocation pal~ Rx, for the call is set to zero. If the type of call is identified at step 1403 as ATM/VBR delay 3e~ilive, then in step 1405, a bandwidth pal~el~r known as statistically weighted i~cl~ nt~l bandwidth (SWIB) is ~ - ",i~,~l and Ri is set equal to the SWIB value. No extra bandwidth is assigned for an ATM/VBR
delay se~ ive call, and therefore the extra bandwidth allocation parameter, Rx, for the call is set to zero (also in step 1405). An example of a process to determine SWIB is described in the context of Fig. 17, and the concept of SWIB is explained below in the description of Fig. 18.
If the type of call is identified at step 1403 as ATM/VBR delay tolerant, then the process proceeds to step 1406. At step 1406, the bandwidth controller 435 5 obtains delay tolerance limits for the call in consideration, and det~rminee the guaranteed bandwidth Ri required to support the call. The gu~a~lleed bandwidth Ri is d~ ed so that it is just adequate to meet the upper bound of the delay requirement specified by the ATM/VBR delay tolerant call. Also at step 1406, thebandwidth controller 435 determin~s extra bandwidth Rx that can be ~esigned to the 10 connection to meet the specified lower bound of the delay ~ lent. The process proceeds from steps 1404, 1405, or 1406 to step 1408, where the process of Fig. 15 is used to d~t~ e call admission and bandwidth allocation based on bandwidth request p~alllct~ (Ri, Rx) as detc....i..~d at steps 1404, 1405, or 1406. If the type of call is identified at step 1403 as ATM/contention, then the process proceeds to step 1407, where the process of Fig. 21 is used to de(~.. ;.. c call ?flmiesion.
After (ielr~ lion of the bandwidth allocation parameters (Ri, Rx) in Fig. 14, the process in Fig. 15 det~nin~s whether or not the call can be admitted. In step 1501, the pararneters Ri and Rx and the Im~e.cign~d bandwidth U are ~
into nearest multiples of ~, which l~l~sell~ a quanta of bandwidth col,t~ ollding to 20 allocation of one ATM time slot per bandwidth allocation cycle. The bandwidthallocation cycle may be one frame long (as in Fig. 12), or it may be "n" frames long (as ~liecllesed later in Fig. 19). The 4ll~ ~ value of U is l~r~sellled by U' in step 1501. At step 1502, it is det~.. i.. ~d whether U' is adequate to assign Ri to the call.
If the result in step 1502 is posili~" then the process proceeds to step 1503. At step 2S 1503, it is dct~ ed whether the ,~ g lln~eei~n~d bandwidth (U' - Ri) is adequate to ~sign extra bandwidth Rx to the call. If the result in step 1503 is positive, then a decision is made to allocate Rx to the call, and the process proceeds to step 1507. If the result in step 1503 is negative, then in step 1504 it is det~.rrnin~d whether (U' - Ri) is larger than or equal to ~. If the result of step 1504 is negative, 30 then Rx is set to ~ro in step 1505, and the decision is made not to allocate any extra bandwidth to the call. However, if the result in step 1504 is positive, then thedecision in step 1506 is to allocate extra bandwidth equal to (U' - Ri), because that is the most l~n~igned bandwidth that is ~ ly available.
From steps 1505 or 1506, the process proceeds to step 1507. At step 1507 5 the values Of GA~ XA and U are l-pd~tPd in accordance with the ~tlmi~ion of the ATM call7 and the process proceeds to step 1512. The call is ~-imitted and STM
messages and ATM MAP are generated in step 1512, to inform all stations of the new time slot allocations.
If the result in step 1502 is negative, it means that the lm~ d bandwidth 10 by itself is inadequate to allocate the gu~lleed bandwidth Ri. A cl~ ;on is made in step 1508 as to whether or not the un~ nPA bandwidth U' and the extra bandwidth XA together are adequate to assign bandwidth Ri to the call. If the result in step 1508 is negative, then the call is rejected in step 1509. If the result in step 1508 is positive, then a decision is made to admit the call. In this case, a portion of 15 the bandwidth needed to admit the call comes from the extra ATM bandwidth region XA. Then, in step 1510, the parameters GA~ U and XA are updated. Now, due to thechange in XA, the ATM extra bandwidth reallocation process of Fig. 16 is used instep 1511. Then the process proceeds to step 1512, which was previously described.
In the process of call a-lmi~ion and bandwidth allocation in the ATM region, 20 wLcu~c~l a change occurs in the value of XA, the extra bandwidth allocations for eYi~ing ATM/VBR col~tions have to be revised. In the process shown in Fig. 16, in step (A), a suitable crit~rion is used to del~ ...;..f reallocation of XA over ~xi~ting ATM/VBR calls. An example of this ~rit~rion is to ~ lionately reallocate the extra bandwidth in aGcol~cG with the extra bandwidth requests ori~in~lly made by25 each of the ~oYi~tin~ ATM/VBR calls. In step (B) of the process in Fig. 16, new values of Rx are cG~ cd for cach of the arrc tcd co~ ions by the decision made in step (A). Then in step (C), new ATM MAP information is generated to notify all stations of the ATM time slot reallocations.
The process of Fig. 17 is applied in step 1405 of the process in Fig. 14. It 30 shows how the Statistically Weighted Incremental Bandwidth (SWIB) is determined.

First, in step 1702, the ATM/VBR call is classified into one of classes 1, 2, ..., J.
Assume the outcome of step 1702 is that the call is of class j for some value of j.
Before we proceed further in the flowchart of Fig. 17, we turn to Fig. 18 for a moment to describe the concept of SWIB.
Fig. 18 shows SWIB values required to admit a new call of a particular class, given that there are already "n" such calls Arlmitted in the system. The SWIB values are shown by the vertical bars, and the values of "n" are shown on the ho~
axis. When "n" is small, such as 1, 2 or 3, the value of SWIB 1801 is equal to the source peak rate 1804. When the number of cA~ls multiplexed together is small, statistical averaging across multiple ATM/VBR calls does not yield any St~ti~tic~l Multiplexing Gain (SMG), and therefore the call can only be A-lmi1te~1 with a SWIB
value 1801 equal to the peak rate bandwidth 1804. When the number of calls in progress has an intermediate value, such as 12, some SMG is possible, and th~lc;fore the call may be Atimi~te(l with a bandwidth allocation 1802 that is smaller than the peak rate 1804, but larger than the average rate 1805. When the number of calls in progress has a high value, such as 27, significantly high SMG is possible, and therefore the call may be ~imitted with a bandwidth allocation 1803 that is muchsmaller than the peak rate 1804 and fairly close to the average rate 1805. The bandwidth d~ ed from Fig. 18 is called SWIB, bec~u~e it is statistically weighted by the number of calls in progress, and r~,esell~ the incrPn ~ntAl bandwidth that is required to admit a new call while m-oeting pclr ~. lllA~
le.luh~ ~ents specified for the class of calls in consideration. The ~ r~....Ance r~quire~lents may be specified in terms of ATM cell delay and loss ratio. Higher cell delay and loss ratios may occur when the congestion condition of the system is 25 severe. It may also be noted that the SWIB value co~ ,ol~ding to "n" calls inFig. 18 also r~y~3{~l~ the decrease in bandwidth allocation for the class of calls as a whole, when one cAIl terminAtes out of (n+l) calls that are ~ llly ~imitte~
Now l~l.. ;~-g to Fig. 17, step 1704, bandwidth controller 435 reads from ATM buffers 431 (see Fig. 4), a set of values, n = (n1, n2, ..., nJ), represçnting the 30 number of ATM/VBR calls of type 1, 2, ..., J, respectively, that are ~ lelllly admitted. Also in step 1704, bandwidth controller 435 looks up a traffic table stored in ATM buffers 431, to deterrnine SWIB required to admit the new ATM/VBR call of type j. The traffic table referred to here for SWIB ~let~tTnin~tion is a generalization of the concept of SWIB illustrated in Fig. 18, but for a multiplicity S (i.e., one or more) classes of ATM/VBR calls that are statistically multiplexed together. Next, in step 1706, R; is set equal to the SWIB value det~-rrnined in step 1704 and Rx is set to zero. Then in step 170~, the process returns to Fig. 14, step 1405.
In Fig. 19, there is shown "n" col1~ec~ e frames which leplesenl a bandwidth allocation cycle 1901. Fig. 19 is similar to Fig. 12, which representç~l the case when n =1. When the bandwidth allocation cycle 1901 spans over multiple frames, the lengths of the ATM regions (BAl~ BA2, --, BAn) and the ~m~igned regions (Ul, U2, ..., Un) can vary from one frame to another within the cycle of "n"
frames. The ~ ed ATM regions (GA1~ GA2~ .., GAn) and the extra ATM
15 regions (XA1, XA2, ..., XAn3 also can vary from one frame to another within the cycle of "n" frames. However, the length of the STM region, Bs does not vary from frame to frame, because the STM time slots are allocated synchronously with a periodicity of one frame. The reason for the periodicity of one frame for STM time slots is due to the tighter delay constraint for STM calls. The reason why the ATM traffic 20 allows for a bandwidth allocation cycle of multiple frames is due to delay tolerance allowed for certain types of ATM traffic. An ATM/VBR file lld~rel may be completed, for example, with the allocation of one ATM time slot every fd~h frame, and its delay r~uile~ent can still be met. If a second ATM call is allocated an ATM
time slot once in every second frame, ~en the combined bandwidth allocation cycle 25 length for the previously mentioned file llansfer and ~e second ATM call would be ten (i.e., n = 10 in Fig. 19). Now a quanta of ATM bandwidth, ~, that was earlier referred to in the context of the bandwidth allocation process of Fig. 15, represents allocation of one ATM time slot in "n" frames. With this definition of ~, the process of Fig. 15 can be suitably modified to reflect a bandwidth allocation cycle of "n"
30 frames (n > 1).

Fig. 20 shows a modified bandwidth allocation process for ATM calls without repeating all the details shown in Fig 15. Those details can be worked out easily by a person skilled in the art, based on Fig. 15, to elaborate the process of Fig.
20.
In Fig. 20, there is shown a process flowchart for call ~tlmi~ion and allocation of ~ ed ATM time slots and extra ATM time slots, when the bandwidth allocation cycle spans over "n" frames, as shown in Fig. 19 and described above. If bandwidth allocation is done over a cycle of "n" frames (n > 1), then the process of Fig. 14 at step 1408 would proceed to Fig. 20 rather than Fig. 15. In step 2001 of Fig. 20, the bandwidth controller uses an o~tilni~tion (e.g., integer pro~ ,..i"g) process to determinç the numbers of guaranteed ATM time slots (gl, g2~ ~--, gn) and the numbers of extra ATM time slots (xl, x2, ..., xn) over "n" frarnes (periodically) to allocate bandwidth (Ri, Rx) required for the call. If the o~ on process in step 2001 ~l~tr....i~ s that there is no set of values for (g1, g2, ~--, gn) that 15 can be ~csi n~d to meet the ~ c.llcnt of y,ll~A~ bandwidth allocation Ri, then the call is rejected in step 2003. Otherwise, a decision is made to admit the call with the values of (gl, g2, ~--, gn) and (xl, x2, ..., xn) as det~rmin~A by the ~t;...;~l;on process of step 2001. The process proceeds to step 2005 and updates the values of GAi, XAi, i=1,2,... n. Then, in step 2010, the call is a(1mitte~1~ and STM messages and 20 ATM MAP are ge~ ted to inform all stations of the new time slot alloc~tion.c.Before describing Figs. 21 and 22, it would be useful to make a few general observations about bandwidth allocation for ATM/co..lr~.l;on traffic. It should be noted that every ATM time slot is norm~lly treated as open for co-~ l;on~ unlessn~l to an ~Yi~ting ATM/CBR or ATMtVBR call. Depending on the ongoing 25 ATM/coll~..lion traffic, a certain number of ATM time slots, KG~ are g.,~. Ai ~t~ for the ATMt~--~ 1;on traffic. All the le.ll~ g nn~signç~l ATM time slots remain "open" for collk;lllion, and such time slots may be thought of as extra ATM
bandwidth available for colllelllion traffic. The "open" co..l~.l;on slots can be ~csi~n~1 upon request to a new ATM/CBR or ATM/VBR call. An open time slot 30 can also be assigned as a guaranteed time slot for the ATM/contention traffic by the 21 62Sl I

processes in Figs. 21 and 22, as described below. When the number of ATM/contention slots is incremented or decremented by the processes in Figs. 21 or 22, it is the value of KG that gets incremented or decremented.
Fig. 21 shows the process that is followed in bandwidth controller 435 when 5 a request is received to set up an ATM/contcl.lion call. Examples of ATM/contention calls include VOD and video games (U~ lll messages). In step 2101, head end 109 leccivcs signals specifying the application/session type. In step 2102, head end 109 first detk....ill~s the increase in the average load due to the call.
Then, based on the traffic table for colllcl,lion-oriented calls ...~ ed in ATM
buffers 431, head end 109 d~k .. ;.l~s whether or not the call can be ~-imitted while keeping the probability of collisions and c~ccted delay within specified limits. If the result in step 2102 is positive, then the call is ~mi~ted in step 2105. However, if the decision in step 2102 is negative, then, in step 2103, bandwidth controller 435 detP~nine~ whether or not spare bandwidth is available within the ATM region to 15 allocate additional bandwidth to the contention traffic. If the result of step 2103 is egalivc, then the call is rejected in step 2106; othervvise, the process proceeds to step 2104. At step 2104, one more CO..lf ..l;~ n time slot is added to the ATM region, and the value of KG,1~1e3 I~ g the available bandwidth for colllelllion ATM
traffic, is updated, i.e., incr~m~nte~l by "1". Then the process returns to step 2102.
20 The cycling through steps 2102, 2103, and 2104 stops either when the call is ~timit~A in step 2105, with the addition of extra bandwidth, or the call is rejected in step 2106, due to lack of spare bandwidth to add to the ATM region.
Fig. 22 shows the flow diagram ofthe process imple-..- ..led in bandwidth controller 435 to i~ nc.ll or decrement the number of gu~llced ATM/ccs..~ io~
25 slots, KG~ depending on a measule.llent of the frequency of collisious in theATM/colllclllion slots. The collision me~ulwllents are taken over Tc time intervals, and stored in ATM buffers 431. Mc and Nc denote the total number of ATM time slots allocated for colllclltion traffic and the total nurnber of collisions measured, rc~llectivcly, in interval TC. KU and KL represent the upper and lower bounds on30 KG. fu and fL rcplesellt pre-specified upper and lower collision thresholds. In step 2201, the process waits and proceeds to the next step only when the next collision-measurement clock tick occurs. In step 2202, the values of Mc, Nc, Ku, KL. fu, and fL are read in from ATM buffers 431, and the collision frequency, y, is computed.
Then the process proceeds to step 2204, where it is determined whether the measured 5 collision frequency, y, exceeded fu. If the result in step 2204 is negative, then the process proceeds to step 2206, where it is det~rrnined whether y is below fL. If the result of step 2206 is also negative, then the number of contention slots remains m-~h~n~ed, as shown in step 2208, and the process returns to step 2201. If the result in step 2204 is positive, then in step 2210 it is rhe- ~e(l whether any spare bandwidth 10 is available to allocate additional bandwidth to the co,llt;"lion traffic. If the result of step 2210 is negative, then the process proceeds to step 2208. If, on the other hand, the result of step 2210 is positive, then, in step 2211, it is checked whether KG is equal to or greater than Ku. If the result in step 2211 is negative, then KG is incremented by one, and thus one more contention slot is ~u~leed to 15 ATM/co"lel,lion traffic. The process proceeds to step 2214, where the relevant ATM
bandwidth allocation pa.~" t~ are updated. From step 2214, the process returns to step 2201. If the result in step 2211 is positive, then ~e process proceeds to step 2208 and ~ s the same number of co"l~"lion time slots in step 2208, and thenreturns to step 2201. If the result in step 2206 is positivej then in step 2216, the 20 number of slots, KG, gu~lleed to ATM/contention traffic is co,~al~,d to check if it is equal to or lower than a pre-specified lower bound KL. If the result in step 2216 is positive, then the process pl~ceeds to step 2208 and m~int~in~ the same number of co~ "lion time slots, and then returns to step 2201. If the result in step 2216 is negative, then KG is decr~m~nted by one in step 2218. From step 2218, the process 25 proceeds to step 2214, where the ATM bandwidth allocation parameters are updated, and then the process returns to step 2201.
In Fig. 23, there is shown the process flo~ch~l for actions taken in the bandwidth controller 435 when a call t~ ion occurs. The process in Fig. 23 accomplishes the task of releasing bandwidth due to a call de~ e and updating the 30 values of all the affected parameters associated with ATM and STM bandwidth ... .

regions. In step 2301, the head end 109 receives a call termin~tion signal from a station 107 or a service platform 490. In step 2302, bandwidth controller 435 identifies the type of call that tçrmin~t--, l If the tvpe of call was either STM or ATM/CBR, then in step 2303, the variable B is set to Ri, where Ri was the bandwidth allocated to the call, and the process proceeds to step 2308. In step 2308, the STM bandwidth is decr~m~nt~l by the arnount of the released bandwidth B, thelm~cif~ned bandwidth U is incrPm~onte~l by the same amount, then the ~m~i nç~
bandwidth is ~ d and col,vt~ d into extra ATM time slot allocations XA.
Note that the calculation of the pararneter TEMP in step 2308 includes con~ ration of the CO1L7lI~i11t (BA ~ QA) as specified in Fig. 12. From step 2308, the process proceeds to step 2309. If the call was identified in step 2302 as ATM/VBR delay sel ~jilive, then the process proceeds to step 2304. In step 2304, the bandwidthcontroller looks up traffic tables to detPrminç the statistically weighted increm~nt~l bandwidth (SWIB) released by the call. The method for dt;l~ il)g SWIB at call arrival or d~lulc was described above in connection with Fig. 18. Also in step 2304, the value of B is set to the SWIB value, and Rx is set to zero, and the process proceeds to step 2307. If the call was identified in step 2302 as ATMIVBR delay tolerant, then in step 2305 the value of B is set to (Ri + Rx), where R; and Rx are the gu~l~ed and extra bandwidth allocations released by the call, and the process proceeds to step 2307. In step 2307, the values of the bandwidth par~mçters associated with the ATM region are u~d~led, the released bandwidth is added to the extra ATM region XA, and the process proceeds to step 2309. If the call was identified in step 2302 as ATM/co..~..l;on, then the process of Fig. 24 is applied in step 2306, and the process pl~c~ds to step 2309. In step 2309, the process of Fig.
25 16 is applied for reallocation ofthe extra ATM bandwidth XA, and then, in step 2310, STM messages and ATM MAP are generated to inform all stations of new time slot allocations. It may be noted that at step 2306, XA somçtimÇs may not change (especially in the process of Fig. 24), in which case steps 2309 and 2310would not be necessary.

2l 6~

Fig. 24 shows the process followed in bandwidth controller 435 when ATM/contention call t~, .,.i..~lion occurs. In step 2401, the average bandwidth released by the call is determined, and the same is added in step 2402 to the current cumulative average released bandwidth due to dep~lures of co~ "tion-type calls.
5 Then, in step 2403, the value of the current cllml-ls~tive average (co"~u~d in step 2402) is checked to see if it eY~ a specified threshold value. Since the averagebandwidth released by individual ATM/co~ lllion calls is typically very small, adecision to release bandwidth is made typically after several such calls have departed in succession. If the result in step 2403 is lleg~livt;, then the same number of10 ATM/contention slots are mziint~ined in the ATM region, as indicated in step 2405.
If the result of step 2403 is positive, then in step 2404, the number of ATM time slots, KG, ~;ullclllly allocated to co~ lion traffic is reduced by one, and the current value of cllm~ tive average released bandwidth is set to zero. Also, in step 2304, the value of extra ATM bandwidth XA is incr~m~nte~ by ~, which lepl~,se~ the 15 incremental bandwidth associated with allocation of one ATM time slot per bandwidth allocation cycle. Now the process returns to Fig. 23, step 2309.
In Fig. 25, there is shown the flow diagram of the process impl~.mente~l in station 107 in its bandwidth controller 335, and followed whenever an application device activity occurs. The type of activity is identified in step 2501. If the activity 20 is identified as an arrival of a new call, then the process proceeds to step 2502, where the type of call is id~ntified. From step 2502, the process proceeds to step 2504. If the type of activity identified in step 2501 is an arrival of a burst of ATM cells from an ~Yi~tin~ ATM/VBR delay tolerant call, then also the process proceeds to step 2504. In step 2504, the bandwidth controller 335 llego~les with head end 109 to set 25 up a comle.,tion with the bandwidth l~ ~ir~d to support the call or the burst of ATM
cells. In step 2504, the ..~;.~;...l.." and m~xi~ delay tolerance for a burst of ATM
cells may be specified, if applicable, in the negotiation with head end 109. Theprocess proceeds from step 2504 to step 2506. In step 2506, bandwidth controller335 in station 107 receives from head end 109 the STM messages and ATM MAP, and transmits in accordance wvith head end allocations. In step 2506, station 107 also 2l62~ll follows a pre-specified error correction/recovery technique in case of errors in STM
messages or in the ATM MAP. The use of a particular error correction/recovery technique is not required, and any of several ~ elllly available techniques may be used. The process proceeds from step 2506 to step 2510. When a call t~rrnin~tes or tr~ncmi~ion of a burst of ATM cells is completed, station 107 notifies head end 109 of such an event. In step 2501, if the activity is identified as the g~n~tion of an ATM/cor~t~ntion message, then bandwidth controller 335 in station 107 tr~n~mit~ the ATM/contention message using the cont~.ntion technique as described in detail in the above-cited FAmon, Li and Sriram patent application. The ATM/cc~ ..lion 10 technique used in the present invention differs in some ways from the technique in the above-cited Edmon, Li and Sriram patent application. In particular, the idea of "super slots" is not used. Also, as described in earlier parts of this disclosure, head end 109 exercises control over allocation or deallocation of ATM/cont~ntion timeslots in several ways that si~ifir~ntly differ from that in the above-cited F~lmon, Li 15 and Srirarn patent application. For example, see processes described in Figs. 21 and 22.
Having thus described the present invention in detail, it is now to be noted that the present invention provides significant capacily advantages over known MAC protocols. For STM traffic, it provides at least the same capacity as the 20 conventional TDMA protocol. However, for ATM traffic classes, that will be supported in bro~lb~n~l tree and branch networks (based on fiber/coax or wireless medium), the MAC protocol of the present invention provides significantly highercapacities. The present invention provides superior capacity due, at least in part, to the way that several significant technical issues are addressed. Specifically, with 25 respect to prop~g~tion delay, the present invention is in~n~itive to propagation delay for ~ t~nr~s up to 150 km. A frarne size of 2 ms is ~efell~d for this reason, and is adequate to cope with round trip acknowle~lgm~nt delay for the 150 km distance.
The operation of the invention is generally independent of the frarne length. If the distances are smaller in a specific network, a smaller frame length can be selected.

Further, it is not desirable for stations to be idie while waiting for a fairly long time relative to the packet tr~n~mi~cion time, due to round trip propagation delay, to receive the head end acknowledgments. This problem is avoided by interleaving traffic associated with different stations into multiple time slots within a 5 frame, and ~ lring that for a particular station, the ~uccessive time slots used by that station are spaced apart (in time) from each other by at least the amount of theeYpected round trip propagation delay. Accordingly, by the technique of this invention, the frame is composed of many interleaved ATM time slots; the actual nurnber of time slots depends on the bit rate. Since a frame consists of interleaved 10 time slots, one of the stations can 1....~...;l in one time slot of the frame, and wait for a predetlormined time (e.g., 2 ms) to receive an acknowledgment In the me~ntime the channel does not experience any dead time, be~use other stations can ll~lllil in the other time slots of the frame. The status acknowle l~m~o-nt~ are also interleaved over ~ltern~te frames, since messages such as BUSY/CONTINUATION, 15 BUSY/FINAL, COLLISION and IDLE ~.l~i~ing to ATM time slots in one u~ ~ frame are sent in the MAP field in the next d~w~ll~ frame.
The present invention also provides head end or central controller aLl~ lion, which is particularly advantageous in m~n~gin~ ATM/CBR and ATM/VBR traffic.
The head end can receive, from a station ori~in~tin~ such traffic, any information 20 conc~rnin~ bandwidth l~uhGlllent, burst length and delay tolerance. It can then assign slots at a rate that meets the bandwidth and delay requirements. This is the "lesel./~lion" collll)ol~ent ofthe MAC protocol ofthis invention. By this mech~ni~m, a station with delay se~ilive ATM/CBR traffic may be ~si~ one or more time slots in every frame, to match the rate of gell~,.alion of ATM cells at the station with 25 the rate at which time slots are ~i~ed to that station. On the other hand, a station with delay tolerant ATM/VBR traffic might not be ~ n~d one slot in every frame;
instead, it may, for example, get permi~ion (from the controller) to transmit in one slot every fifth frame, if that satisfies the delay requirement as specified by the station.

2162$11 Although the previous description mentioned several approaches for detçrmining when "repacking" (as defined previously) should occur, there are several specific details and arrangements which may be considered and used for this function, as set forth below. Accordingly, in a specific impl~mt nt~tion of this5 invention, the cost benefit trade-offs for the subscriber population must be taken into consideration in dele. .,.il~ g which rep~t~ing ~ ;ve is pl~relled.
A first approach is called "Quick ~pacl~ing and Creation of Additional ATM
Slotsn. With this approach, ~xi~ting ATM traffic makes use of the bandwidth released by an STM call. This can be arranged as follows. After the depallule of an 10 STM call, the lt ...~;"i~-g STM col~l~ections can be repatl~d within the next frame or two. The released (i.e., lln~igntd) bandwidth can be then used to create as manyATM slots ~ possible. The newly created ATM slots are added to the existing ATM slots, and thereby the ATM region is ~ n~lçd into the lm~sign~d region (see Fig. 12). The STM traffic can reclaim the some ofthe rele~ed bandwidth from the 15 ATM potion of the frame on a dt m~n-l basis, i.e., when new STM calls requires it.
A second approach is called "Quick ~e~fleL il~ and Adding Bandwidth to a Common Spare Pool". This approacll is based upon the fact that, when an STM calldeparts and releases some bandwidth, it is not known apriori as to whether the next col~nection request will be STM or ATM. Th~ rol~, the released bandwidth is 20 added to a spare pool. This can be arranged by (1) ~ L ;l~g the active STM calls, and (2) adding the bandwidth thus rele~ed to the lm~ d region of the frame bc;lw~en the STM and ATM regions. This lln~ci~ or idle region of the frame con~ es the spare pool of bandwidth, and new STM/ATM time slots may be created on a ~l. m~n-l basis by decisions made in the controller. When the controller 25 decides to admit a new STM call, it does so by adding an STM slot and ~ ntlin~ the STM portion ofthe frame into the lln~ign~1 region. When the controller decides to create one or more new ATM slots, those slots are created by ~Yten~ling the ATM
portion of the frame into the lln~s~igned region. In this way, the spare bandwidth always resides between the STM and ATM regions in a frame.

A third approach is called "STM Need Driven Repacking". With this approach, "holes" in the STM portion of the frarne are left untouched, i.e. no repacking is done, until it becomes necessal y to accommodate a new STM or ATM
call. This situation arises when the "holes" are created in a non-contiguous way, 5 and when there is enough aggregate bandwidth to admit a new call, but there is no single time slot or span of several time slots (in the STM portion) that is large enough. When such a situation is identified, the controller does a repacking at that time, so that the new call can be ~lmi~
A fourth approach is similar to the third approach, except that the repacking 10 is ~ r l~ed at periodic or fixed intervals, rather than being triggered by the arrival of a new STM or ATM call.
No matter which rep~ ing ~1L~ I;ve is used, following a rep~ in~ action at the controller, all stations are notified of the new STM and/or ATM time slotallocations, by s~n-1ing messages in the STM ~i n~ling and MAP portions of the 15 next d~wlL~ frame.
Persons skilled in the art will recognize that various modifications may be made to the present invention. Accordingly, the invention should be limited only by the following claims. For example, another modification can achieve higher bandwidth efficiency by considering multiple u~7Ll~n and dow- sl-~ ch~nn~l~
20 collectively as a "group", for bandwidth allocation and call ~lmi~sion decisions. As cll~se~l earlier in the context of Fig. 2, the u~r.7l~e~ll bandwidth is usually O~ ~ in the form of multiple ~h~nn~l~ 202-1 through 202-n. Similarly, the d(lw~l~ bandwidth (see Fig. 2) is also usually o~ ~ in the form of multiple rh~nn~l~ 203-1 ~lhrough 203-m. The time within each of these U1~SL1C~ and 25 dow,~l,~ll rh~nnel~ is divided into a series of time frames, such as those shown in Figs. 7, 8, and 9. In Figs. 12 through 25, the processes for bandwidth allocation for STM and ATM calls were described, con~idering a series of frames within one u~sL~ h~nn~!, and a coll~ onding series of frames in one associated downstream channel. However, it should be noted that the bandwidth allocation and 30 dynamic adjustments of STM regions (Bs), ATM regions (BA, GA~ XA), and the boundary regions (U) (see Fig. 12) can be perforrned on more than one channel, or - even "globally" on all channels, by considering multiple upstream channels 202-1 through 202-n and multiple downstream channels 203-1 through 203-m as a "group". In such an arrangement, the frame shown in Fig. 12 can be interpreted in a generalized manner, in which the STM region BS would represent the current curnulative sum of the individual STM regions in frames in a group of several of the "n" U~ alll ch~nn~l~ Similarly, the ATM regions (BA~ GA~ XA) and the boundary region U would le~les~nt the current cllm~ tive sum of the individual corresponding regions in frames in the same group of several of the "n" u~slr~ ~.h~nnel~. BY
~ some additional status infonn~tion, in the STM and ATM buffers in the head end 109 (see Fig 4), the processes of Figs. 13 through 25 can be implem~nte~
for bandwidth allocation across multiple ~h~nn~l~ Some suitable modifications tothe fl~w~h~l~ of Figs. 13 through 25 may be made, in accordance with the principles on which this invention is based. These modifications can vary, based on specific implementations, but would nevertheless be used for efficient use oftheaggregate bandwidth across multiple ch~nnel~, in accor~ance with the principles of this invention. An example of the stated modification is as follows. When a second call arrives from a station 107, which already has a first call in progress, the head end 109 may determine that bandwidth in not available for the new call on the same upstream ~.h~nnel, say channel 202-1, that the station 107 is ~;ull~nlly using. In that case, the head end 109 may decide to assign bandwidth for the new call on a di~~ Uy~ alll channel 202-n, where bandwidth is ~ nlly available, and then inform the station 107 to switch to the stated different eh~nnel 202-n for the new call set-up ~ well as for col.l;..l.~l;on ofthe previously ongoing first call. This 25 functionality may be accomplished by means of a modem feature called frequency agility, which is eu~ lly available. Then head end bandwidth controller 435 makes adjustments, con~ elllly and in a coordinated manner, to location and sizes of STM/ATM regions and boundary (or lm~igned) regions in the frames of the tvvo channels 202-1 and 202-n. If multiple modems are used in station 107, then the head 30 end bandwidth controller 435 may assign bandwidth to the second call in a different ..

channel, such as channel 202-n, while allowing the existing first call at station 107 to remain in channel 202-1.
Also, while the present invention covers what are generally referred to as (a) OSI Level 2 functions, including functions related to multiple access over a shared S medium at the MA C layer, and (b) OSI Level 3 functions, including functions related to call admission and bandwidth allocation, it is to be lm~erstood that certain other OSI Level 3 functions, such as bandwidth mollilolhlg and policing, and flow and congestion control, can also be provided using presently available techniques, and these other functions would be hll~ led with those provided by this invention.
Also, as previously stated, while the present invention has been described in the context of a fiber/coax based tr~n~mi~sion i~ cture, it is to be understood that the invention is also applicable to a wireless coll~ lications ellviroll~llent. In the latter situation, the terms "mobile station", "base station" and "wh~lcss medium", would be applicable in place of the terms "station", "head end" and 15 "fiber/coax medium".

Claims (10)

1. A method of allocating transmission bandwidth among multiple stations interconnected with a common controller via a transmission medium having a multiple access upstream channel, said method comprising the steps of:
dividing time in said upstream channel on said transmission medium into a series of successive time frames;
dividing each of said time frames into first and second regions separated by a boundary region, each of said first and second regions containing a one or more time slots;
said first region consisting of one or more variable length time slots for STM
(Synchronous Transfer Mode) calls, and said second region consisting of one or more fixed length time slots assigned to (a) reservation and (b) contention oriented ATM
(Asynchronous Transfer Mode) calls; and dynamically adjusting the location and size of said boundary region in each of said frames as a function of the bandwidth requirements of said stations, and for dynamically sharing said second region to accommodate both reservation and contention types of ATM calls.

2. The method defined in claim 1 wherein said dynamic sharing increases the number of time slots assigned to contention oriented ATM calls as a function of the number of collisions that occur when multiple stations contend for time slots.

3. The method defined in claim 1 wherein said dynamic sharing adjusts the number of time slots assigned to contention oriented ATM calls as a function of arrivals and departures of contention oriented calls.

4. A method of allocating transmission bandwidth among multiple stations handling ATM (Asynchronous Transfer Mode) calls, some of said ATM calls generating ATM/contention traffic and other of said calls generating ATM trafficrequiring reserved time slots, said stations being interconnected with a common controller via a transmission medium having a multiple access upstream channel, said method comprising the steps of:
dividing time in said upstream channel on said transmission medium into a series of successive time frames;
dividing each of said time frames into a plurality of time slots; and assigning some of the time slots to said ATM/contention traffic and assigning others of said time slots to said ATM traffic requiring reserved time slots, said assignment being performed as a function of the frequency with which collisions occur due to simultaneous transmission in time slots assigned to said ATM/contention traffic.

5. A method of allocating transmission bandwidth among multiple stations handling ATM (Asynchronous Transfer Mode) calls, a first type of said ATM calls generating ATM/contention traffic, a second type of said calls generating constant bit rate ATM traffic requiring reserved time slots, and a third type of said calls generating variable bit rate ATM traffic requiring reserved time slots, said stations beinginterconnected with a common controller via a transmission medium having a multiple access upstream channel, said method comprising the steps of:
dividing time in said upstream channel on said transmission medium into a series of successive time frames;
dividing each of said time frames into a plurality of time slots; and assigning first ones of said time slots to said first type calls;
assigning second ones of said time slots to said second type calls;
assigning third ones of said time slots to said third type calls;
when any of said assigned first, second or third time slots are vacated because corresponding ones of said first, second or third type calls are terminated, reassigning said vacated time slots to (a) new or (b) ongoing calls, said new or ongoing calls being of (a) the same or (b) a different type, said reassignment being made in accordance with the bandwidth requirement of said (a) new or (b) ongoing calls.

6. Apparatus for allocating transmission bandwidth among multiple stations interconnected with a common controller via a transmission medium having a multiple access upstream channel, said apparatus comprising:
means for dividing time in said upstream channel on said transmission medium into a series of successive time frames;
means for dividing each of said time frames into first and second regions separated by a boundary region, each of said first and second regions containing a one or more time slots, said first region consisting of one or more variable length time slots for STM (Synchronous Transfer Mode) calls, and said second region consisting of one or more fixed length time slots assigned to (a) reservation and (b) contention oriented ATM (Asynchronous Transfer Mode) calls; and means for dynamically adjusting the location and size of said boundary region in each of said frames as a function of the bandwidth requirements of said stations, and for dynamically sharing said second region to accommodate both reservation and contention types of ATM calls.

7. The apparatus defined in claim 6 wherein said dynamic sharing increases the number of time slots assigned to contention oriented ATM calls as a function of the number of collisions that occur when multiple stations contend for time slots.

8. The apparatus defined in claim 6 wherein said dynamic sharing adjusts the number of time slots assigned to contention oriented ATM calls as a function of arrivals and departures of contention oriented calls.

9. Apparatus for allocating transmission bandwidth among multiple stations handling ATM (Asynchronous Transfer Mode) calls, some of said ATM calls generating ATM/contention traffic and other of said calls generating ATM trafficrequiring reserved time slots, said stations being interconnected with a common controller via a transmission medium having a multiple access upstream channel, said apparatus comprising means for (a) dividing time in said upstream channel on said transmission medium into a series of successive time frames and (b) dividing each of said time frames into a plurality of time slots; and means for assigning some of the time slots to said ATM/contention traffic and assigning others of said time slots to said ATM traffic requiring reserved time slots, said assignment being performed as a function of the frequency with which collisions occur due to simultaneous transmission in time slots assigned to said ATM/contention traffic.

10. Apparatus for allocating transmission bandwidth among multiple stations handling ATM (Asynchronous Transfer Mode) calls, a first type of said ATM calls generating ATM/contention traffic, a second type of said calls generating constant bit rate ATM traffic requiring reserved time slots, and a third type of said calls generating variable bit rate ATM traffic requiring reserved time slots, said stations beinginterconnected with a common controller via a transmission medium having a multiple access upstream channel, said apparatus comprising:
means for (a) dividing time in said upstream channel on said transmission medium into a series of successive time frames and (b) dividing each of said time frames into a plurality of time slots;
means for (a) assigning first ones of said time slots to said first type calls; (b) assigning second ones of said time slots to said second type calls; and (c) assigning third ones of said time slots to said third type calls;
means operative when any of said assigned first, second or third time slots are vacated because corresponding ones of said first, second or third type calls areterminated, for reassigning said vacated time slots to (a) new or (b) ongoing calls, said new or ongoing calls being of (a) the same or (b) a different type, said reassignment being made in accordance with the bandwidth requirement said (a) new or (b) ongoing calls.