How does a channel is allocated to GPRS mobile for data transfer ?
ngnguru_com 12-February-2009 07:52:51 PM

Comments


I think you must visit www.ee.iitb.ac.in/~prakshep/IBMA_lit/manual/manual110.html
Posted by HamidAliKhan


Please visit:

www.ee.iitb.ac.in/~prakshep/IBMA_lit/manual/manual110.html

www.gsmcommunication.com
Posted by sagitraz


FIELD OF THE INVENTION

The present invention relates generally to data transmission in a GPRS/EDGE system, and in particular, the present invention relates to set up of an uplink packet data transfer in a GPRS/EDGE system using an indirect carrier sense multiple accesswith directed acknowledgement.

BACKGROUND OF THE INVENTION

Global System for Mobile Communications (GSM) General Packet Radio Service (GPRS) and Enhanced Data for Global Evolution (EDGE) are intended to allow the service subscriber the ability to send and receive data in an end-to-end packet transfermode without utilizing network resources in the circuit-switched mode. GPRS and EDGE permit the efficient use of radio and network resources when data transmission characteristics are i) packet based, ii) intermittent and non-periodic, iii) possiblyfrequent, with small transfers of data, e.g. less than 500 octets, or iv) possibly infrequent, with large transfers of data, e.g. more than several hundred kilobytes. User applications may include Internet browsers, electronic mail and so on.

Efforts are presently underway to further develop the European Telecommunications Standards Institute (ETSI) GPRS and EDGE specifications to support the wireline concept of voice over Internet protocol (VoIP). This effort includes the abilityfor a mobile station to terminate and originate a VoIP call as an endpoint on the Internet. The current definition for GPRS and EDGE supports the concept of both a packet-switched radio environment and a packet-switched network environment, i.e. thepacket abstraction of the Internet is carried through to the air interface in the form of intermittently accessible radio resources based upon the availability of radio resources and the demand for the interchange of user data.

FIG. 1 is a schematic diagram of a complete packet data transfer in a GPRS/EDGE radio environment. As illustrated in FIG. 1, packet switching in the radio environment is achieved using the concept of a packet data transfer 100, referred to as a"temporary block flow" (TBF). The temporary block flow 100, which includes a data transfer setup phase 102, a data transfer phase 104, and a data transfer teardown phase 106, is regarded as the basic unit of data interchange within the GPRS/EDGEenvironment. As a result, temporary block flow 100 may be thought of conceptually as its three components, data transfer setup phase 102, data transfer phase 104, and data transfer teardown phase 106, occurring sequentially in time. It is understoodthat the amount of time for the setup of a temporary block flow for GPRS varies, and is dependent on channel conditions, radio resource availability, network congestion and so on.

Although GPRS and EDGE have been specified with the objective of interchanging packet based user data, the application for most such data interchange is not of a real-time nature. Voice over IP presents several challenges to the GPRS/EDGEdomain, one of which is the availability of data transfer capability in the uplink direction. For example, when the mobile VoIP user speaks into the phone, a temporary block flow is required to be set up in the uplink direction as soon as possible. However, the time required by GPRS and EDGE to set up such an uplink temporary block flow is prohibitive when compared to the generally accepted maximum turnaround delay for voice telephony, which is 125 ms. Furthermore, VoIP telephony would require theaddition of other mechanisms which would enable the radio layers to have knowledge of the type of information they are carrying at any given time.

In particular, the amount of time required for data transfer setup phase 104 has proven to be excessively long, resulting in problems associated with both round-trip turnaround time, and throughput, as a function of the duty-cycle reductionrequired for setting up an acknowledgement at the upper (network) layers, e.g. the transport layer.

Accordingly, what is needed is a method for enabling a mobile station to more rapidly set up an uplink packet data transfer in a GPRS/EDGE environment.
BRIEF DESCRIPTION OF THE DRAWINGS

The features of the present invention which are believed to be novel are set forth with particularity in the appended claims. The invention, together with further objects and advantages thereof, may best be understood by making reference to thefollowing description, taken in conjunction with the accompanying drawings, in the several figures of which like reference numerals identify like elements, and wherein:

FIG. 1 is a schematic diagram of a complete packet data transfer in a radio environment.

FIG. 2 is a schematic diagram of a GPRS system according to the present invention.

FIG. 3 is a schematic diagram of modification of a user data stream as the user data stream passes through specified layers of a GPRS system.

FIG. 4 is a schematic diagram of a multiframe structure for packet data channels.

FIG. 5 is a data flow diagram of stream-oriented data transmitted between a mobile station and a network.

FIG. 6 is a schematic diagram of a dynamic timeslot allocation for medium access control.

FIG. 7 is a schematic diagram of a fixed timeslot allocation for medium access control.

FIG. 8 is a schematic diagram of signaling logic for establishing an uplink packet data transfer.

FIG. 9 is a flowchart of indirect carrier sense multiple access with directed acknowledgement in a mobile station.

FIG. 10 is a flowchart of indirect carrier sense multiple access with directed acknowledgement in a base station

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

The present invention is related for allowing mobile stations to more rapidly set up an uplink packet data transfer in a GPRS/EDGE system using an indirect carrier sense multiple access with directed acknowledgement (ICSMA/DA), whereby the mobilestation would be notified of when a transmit resource is not in use, allowing the mobile station to transmit on this resource only if a downlink transfer is in progress and then acknowledging the mobile station's access to the medium directly.

FIG. 2 is a schematic diagram of a GPRS system according to the present invention. As illustrated in FIG. 2, a GPRS system 200 includes a mobile station 202 sending and receiving packet data from an internet application 204 to a remote internetapplication 206 through a base station system 208. While a single base station system 208 and mobile station 202 is illustrated in FIG. 2, it is understood that GPRS system 200 includes multiple numbers of base station systems and mobile stations. Mobile station 202 includes a GPRS/EDGE subsystem 210 for processing signaling messages received from base station system 208, and signals received from internet application 204 through transport and network layers 212. GPRS/EDGE subsystem 210 includesa medium access control (MAC) layer 211, and adds header overhead for sub-network convergence/divergence protocol (SNDCP), and logical link control (LLC). A protocol control unit 214, includes a medium access control layer 213, and is coupled to orcontained within base station system 208, and interfaces with GPRS/EDGE subsystem 210 of mobile station 202, and with internet application 206 through transport and network layers 216. Internet transport layers 212 and 216 include a transmission controlprotocol (TCP) layer 218 which TCP packetizes stream-oriented user data, and an internet protocol (IP) layer 220 which assigns an address to the packetized data.

FIG. 3 is a schematic diagram of modification of a user data stream as the user data stream passes through specified layers of a GPRS system. As illustrated in FIG. 3, a user data stream of infinite length is modified as the user data streampasses through GPRS system 200. For example, as illustrated in FIGS. 2 and 3, as the user data stream passes through transmission control protocol layer 218, and an RLC layer, the data stream is divided into a TCP packet 222 that includes a payload 224that is 536 octets in length and a transmission control protocol header packet 226 that is twenty octets in length, giving TCP packet 222 a total length of 556 octets. As TCP packet 222 subsequently passes through internet protocol layer 220, anadditional twenty octet internet protocol header 228 is appended to TCP packet 222, forming an IP packet 230 having a total length of 576 octets. An additional four octet SNDCP header 232 is appended to IP packet 230, forming an SNDCP packet 234 havinga total length of 580 octets, and an additional four octet logical link control header 236 is appended to SNDCP packet 234 forming a logical link control packet 238 having a total length of 584 octets. As a result, the user data stream has a totallength of 584 octets as the data stream exits logical link control.

Next, radio link control divides the 584 octet logical link control packet 238 into a certain number of radio link control data blocks, the exact number of which depends upon the channel coding scheme used. For example, in a CS-1 channel codingscheme, the number of radio link control blocks needed is equal to (LLC frame length/RLC payload length)+(LLC frame length MOD RLC payload length), which for the 584 octet logical link control frame is equal to 31 radio link control blocks. In a CS-2channel coding scheme, the number of radio link control blocks needed is equal to (LLC frame length/RLC payload length)+(LLC frame length MOD RLC payload length), which for the 584 octet logical link control frame is equal to 21 radio link controlblocks.

FIG. 4 is a schematic diagram of a multiframe structure for packet data channels. Assuming a perfect schedule of one radio link control block transmitted on each available block period for a single timeslot transfer, raw throughput may becomputed based upon the length of time required to send a certain number of radio link control data blocks. As illustrated in FIG. 5, a packet data control channel is organized as a multiframe 260 having fifty-two frames 262 and twelve data blocks B0B11, in which each data block B0 B11 is distributed over four time division multiple access (TDMA) frames. An "idle" or "search" frame 264 located after every three data blocks, enables the mobile station to perform adjacent cell signal measurements,synchronization and verification of synchronization status on adjacent cells, interference measurements, and so forth. Each data block B0 B11 is made up of four frames, each of which has a frame period f equal to 4.61538 milliseconds, and a block periodb that is equal to 18.4616 milliseconds, while each idle frame 264 has an idle frame period I that is equal to the frame period f, or 4.61538 milliseconds. The total period of the multiframe 260 structure of the packet data channel is equal to 240milliseconds.

The length of time required TR to send a certain number of radio link control data blocks Nb is calculated using the following equation: TR=(Nb×b)+((Nb/3)×f) EQUATION 1 while the raw data throughput Rd iscalculated using the following equation: Rd=(number of payload octets/TR)×8 EQUATION 2

Using Equations 1 and 2, the time required to send all of the radio link control blocks in a logical link control frame in a CS-1 coding scheme (i.e. 31 blocks) is equal to 0.618462 sec. The throughput is the number of payload octets (584)divided by the time required to send them plus their overhead (0.618462) times 8 bits per octet, which is equal to 7000 bits/second. In terms of an overhead analysis of the CS-1 coding scheme, theoretical throughput is approximately equal to 9050bits/sec. The overhead of scheduling, i.e. the fact that there are idle frames that prevent the scheduling of every consecutive block reduces the effective throughput by 4/52 to approximately 8861 bits/second. The overhead of radio link control headers,i.e. three octets per block, reduces the effective throughput by 3/22 to approximately 7652 bits/second. The overhead of the logical link control header, four octets, reduces the effective throughput by 4/584 to approximately 7599 bits/second. Finally,the overhead of the SNDCP header, 4 octets, reduces the effective throughput by 4/580 to approximately, and the overhead of the Internet protocol suite, i.e. TCP and IP headers, reduces the effective throughput by 40/576 to approximately 7000bits/second.

Similarly, the time required to send all of the radio link control blocks in a logical link control frame (i.e. 21 blocks) in a CS-2 coding scheme is equal to 0.42 second, and the throughput is the number of payload octets (584) divided by thetime required to send them plus their overhead (0.42) times 8 bits per octet, which is equal to 10,209 bits/second. Theoretical throughput on channel at CS-2 is approximately equal to 13,400 bits/sec. The overhead of scheduling, i.e. the fact that thereare idle frames that prevent the scheduling of every consecutive block reduces the effective throughput by 4/52 to approximately 12,369 bits/second. The overhead of radio link control headers, i.e. three octets per block, reduces the effectivethroughput by 3/32 to approximately 11,209 bits/second. The overhead of the logical link control header, four octets, reduces the effective throughput by 4/584 to approximately 11,132 bits/second. Finally, the overhead of the SNDCP header, four octets,reduces the effective throughput by 4/580 to approximately 11,055 bits/second, and the overhead of the Internet protocol suite, i.e. TCP and IP headers, reduces the effective throughput by 40/576 to approximately 10,209 bits/second.

FIG. 5 is a data flow diagram of stream-oriented data transmitted between a mobile station and a network. As illustrated in FIGS. 2 and 5, when stream-oriented data 213 is transmitted from remote internet application 206 to mobile station 202during a downlink period 300 for sending data long a downlink, the data is first divided into packets at TCP layer 218, and given an address at IP layer 220 of transport and network layer 216, and sent to protocol control unit 214 of base station system208 as a TCP/IP packet 302.

As illustrated in FIGS. 3 and 5, during downlink period 300, TCP/IP packet 302 includes overhead associated with logical link control packet 238 and SNDCP packet 234, and it is assumed that for every TCP/IP packet 302, there is a correspondinglogical link control packet 238 and SNDCP packet 234 as well. The actions associated with transmitting the information over the air interface begin when a logical link control frame containing the user information in the form of an encapsulatedtransport/network/SNDCP packet enters protocol control unit 214 of base station system 208.

As illustrated in FIGS. 2 and 5, assuming that mobile station 202 is camped on the network in packet idle mode, when appropriate, base station system 208 begins a setup sequence of a downlink setup period 224 by sending a package paging request215 to GPRS/EDGE subsystem 210 of mobile station 202. In response, after receiving a random access burst 217 from GPRS/EDGE subsystem 210, protocol control unit 214 sends an immediate assignment message 219 and a packet downlink message 221, detailingthe parameters of the assignment, e.g. over what channel the transfer would take place, when the transfer would start, and so on. Protocol control unit 214 sends a series of radio link control data blocks 226 to GPRS/EDGE subsystem 210 after receiving apacket control acknowledge message 222 from GPRS/EDGE subsystem 210.

Depending upon the availability of schedulable blocks, packet paging request message 215 may require from 81 to 1721 ms, followed by random access burst 217 from mobile station 102, which typically requires 9.6 ms. Immediate assignment message219 contains a starting time that may range from 37 ms to 3 minutes in the future, but typically ranges from 13 to 25 TDMA frame periods, or 60 115 ms. Additional signaling associated with exchanging packet downlink assignment message 221 and a packetcontrol acknowledgement message 222 are included in downlink setup period 224. It is therefore assumed that downlink setup period 224 may be equal to a starting time, which is in fact what is observable in an actual system. As a result, the timerequired for downlink setup period 224 is a minimum of approximately 849 ms, a maximum of approximately 2643 ms and an average of approximately 1746 ms.

After the starting time has been reached, protocol control unit 218 sends GPRS/EDGE subsystem 210 a temporary block flow containing radio link control data blocks 226. Once GPRS/EDGE subsystem 210 has received all downlink blocks, GPRS/EDGEsubsystem 210 assembles, processes and transmits a resulting single data packet 228 to IP layer 220 of transport and network layers 212, which then sends data packet 228 to TCP layer 218 of transport and network layers 212.

Assuming perfectly available radio resources so that data may be sent on every schedulable downlink block on a single timeslot, the time to transmit all blocks during data transfer period 225 for a 536 octet user data payload is approximatelyequal to 0.618462 seconds for a CS-1 coding scheme, and 0.420 seconds for a CS-2 coding scheme. The downlink temporary block flow terminates after a last radio link control data block is sent if sending radio link control on protocol control unit 214has no more data to be sent and a radio link controller timer T3192 expires before radio link control receives more data to be sent from logical link control, which is the case when a transmission control protocol transmission starts in "congestioncontrol" (slow-start) mode. The temporary block flow is always torn down after the blocks making up the first transmission control protocol packet are transmitted, causing the downlink temporary block flow to incur the overhead of temporary block flowbeing setup again for the subsequent blocks.

TCP layer 218 of transport and network layers 212 performs redundancy checking and makes a determination that data packet 228 has been received properly. IP layer 220 of transport and network layers 212 then includes the packet data in astream-oriented output 230 to internet application 204 and issues a TCP acknowledgement (TCP ACK) message 232 to TCP layer 218 of transport and network layers 216 on the far end of the virtual circuit. TCP ACK message 232 is processed by SNDCP/LLC andRLC layers as before, but in an uplink direction.

Radio link controller of GPRS/EDGE subsystem 210 of remote transport and network layers 216 receives a TCP/IP/SNDCP/LLC packet containing TCP ACK message 232, but cannot begin a setup sequence corresponding to an uplink setup period 234 fortransmission of TCP ACK message 232 until a radio link control timer T3192 of protocol control unit 114 has expired. As a result, a downlink temporary block flow corresponding to downlink period 300 that carries TCP/IP packet 302 that was initiallysent, must be torn down completely before uplink period 234 for setting up TCP ACK message 232 may begin.

For example, upon receiving TCP ACK message 232, GPRS/EDGE subsystem 210 sends a channel request access burst 236 to protocol control unit 214, which responds by sending an immediate assignment message 238. GPRS/EDGE subsystem 210 then sends apacket resource request message 240 to protocol control unit 214 requesting resources for a temporary block flow. Protocol control unit 214 responds with a packet uplink assignment message 242, which is acknowledged by GPRS/EDGE subsystem 210 in apacket control acknowledge message 244. Data blocks 246 containing TCP ACK message 232 and teardown are then transmitted from GPRS/EDGE subsystem 210 to protocol control unit 214 during an acknowledge data transfer period 248. Protocol control unit 214then transmits data blocks 246 to transport and network layers 216 in a TCP acknowledge message 304. As a result, uplink setup period 234 and acknowledge data transfer period 248 form an uplink period 306 that is required for TCP acknowledge message 232to reach transport and network layer 216 in corresponding TCP acknowledge message 304. Once the required TCP ACK message is received by transport and network layers 216, a next TCP/IP data packet message 250 is sent from transport and network layers 216to GPRS/EDGE subsystem 210 via protocol control unit 214.

The period required for the initial setup of uplink setup period 234 is dependent upon components such as the periodic occurrence of a random access channel (RACH), the starting time sent in immediate assignment message 238, and the starting timesent in packet uplink assignment message 242. The periodic occurrence of a random access channel can range from 41 217 TDMA frame periods, assuming a case of 41 frame periods, or 190 ms. The starting time sent in immediate assignment message 238 mayrange from 9 TDMA frame periods to 3 minutes, but is typically from 9 25 TDMA frame periods, or 42 115 ms, while the starting time sent in packet uplink assignment message 242 may range from 9 TDMA frame periods to 3 minutes, but is typically around 20TDMA periods, or 92 ms. As a result, initial setup of uplink setup period 234 is typically a minimum of approximately 320 ms, a maximum of approximately 480 ms, and an average of approximately 320 ms. This exceeds the generally accepted maximum end toend delay of 125 milliseconds.

TCP ACK message 232 has a length of 40 octets, which combined with the overhead of both logical link control header 236 and SNDCP header 232 is equal to 48 octets. Assuming perfectly available radio resources so that data may be sent on everyschedulable downlink block on a single timeslot, the time to transmit all data blocks 246 during acknowledge data transfer period 248 for a 40 octet TCP/IP ACK payload is equal to 60 ms (3 RLC data blocks) for the CS-1 coding scheme, and 37 ms (2 RLCdata blocks) for the CS-2 coding scheme.

Mobile station 202 receives the right to transmit on the uplink by using either a dynamic timeslot allocation medium control access (MAC) mode or a fixed timeslot allocation medium control access mode. FIG. 6 is a schematic diagram of a dynamictimeslot allocation for medium access control. As illustrated in FIG. 6, in dynamic timeslot allocation, a mobile station 300 receives a downlink radio link control/medium access control (RLC/MAC) control block 302 from a base station 304 that includesa special address, referred to as an uplink state flag (USF) 306, along with RLC/MAC data 308. If USF 306 (a 3-bit quantity) is identical to that of a USF assigned to mobile station 300, then mobile station 300 has the right to transmit in the next timedivision multiple access (TDMA) frame. A data block addressed to a second mobile station 310 may contain USF information for mobile station 300.

FIG. 7 is a schematic diagram of a fixed timeslot allocation for medium access control. As illustrated in FIG. 7, in fixed timeslot allocation, a mobile station 312 periodically receives a starting time and a bit-map 314 from a base station 316,representing a base and offset of future timeslots on which the mobile station is to transmit. In this way, mobile station 312 is informed of when temporary block flow starts and receives bitmap 314 representing timeslots on which mobile station 312 isto transmit relative to the starting time, so that mobile station 312 transmits in timeslots assigned by the starting time and allocation bitmap.

The present invention utilizes a USF field for both fixed and dynamic allocation MAC mode, when the mobile station is not engaged in an uplink temporary block flow, although USF value is given a different meaning, as described below. The presentinvention includes a collision avoidance (CA) mechanism that utilizes the already-present USF address in the RLC/MAC control block to enable the rapid creation of an uplink temporary block flow, when there is already a downlink temporary block flow inprogress. Since the USF value is receivable by multiple mobile stations on the radio resource, the USF value assignment serves as an indirect lock on the resource.

According to the present invention, the USF field is recognized during an active downlink temporary block flow as a "channel availability" indicator and a "directed acknowledgement field". FIG. 8 is a schematic diagram of signaling logic forestablishing an uplink packet data transfer according to the present invention. According to the present invention, the mobile station, when receiving blocks comprising a downlink temporary block flow, examines the USF field when it has information totransmit in an uplink temporary block flow. If the USF were a zero value, then the channel would be evaluated as "available", and the mobile station may therefore begin transmitting its new uplink TBF information.

In particular, as illustrated in FIGS. 5 and 8, while a base station 320 and a mobile station 322 are in a downlink temporary block flow setup 324 of downlink setup period 224, which includes the assignment of a mobile station USF address, suchas "110", for example, base station 320, through medium access control layer 213, sends a USF address to mobile station 322 by which mobile station 322 will be identified for the duration of the downlink temporary block flow for data transfer period 225resulting from downlink temporary block flow setup 324, along with a contingent uplink timeslot number on which mobile station 322 may transmit. Once GPRS/EDGE data flows in the downlink direction in data transfer period 225, base station 320, throughmedium access layer 213, indicates uplink channel availability 326 to mobile station 322 by sending the value USF=000. If mobile station 322 has data to transmit on the uplink, mobile station 322 transmits a first uplink radio link control data block328 on the timeslot indicated by base station 320 as a contingent uplink timeslot number.

Base station 320 receives the first uplink data block 328 and knows how to associate a temporary flow identifier (TFI) to a USF value of mobile station 322. Base station 320 acknowledges the USF address of mobile station 322 in the next downlinkradio link control data block 330 by inserting the USF value (that serves to indirectly address a mobile station) of mobile station 322 into the header of next downlink radio link control data block 330. According to the present invention, the insertedUSF value serves as an acknowledgement to the sending mobile station 322 and as a "channel busy" indication to other mobiles desiring to transmit. Mobile station 322, through medium access layer 211, interprets next downlink radio link control datablock 330 with a USF value located in a header at the beginning of radio link control data block 330 as an acknowledgement that the first uplink data block 328 was correctly received by base station 320, and sends a subsequent uplink radio link controlblock 332. This procedure is then continued for the remaining portion 343 of the uplink temporary block flow 248, until the end of uplink temporary block flow 248, which is indicated in the usual manner to base station 320 by the countdown procedure inthe last several radio link control data blocks. In the countdown procedure, mobile station 322, during transmission of the last few data blocks 246, decrements a variable in the header of data blocks 246 to inform base station 320 that the uplink datablock flow is about to end. This knowledge helps base station 320 allocate to another mobile station.

As a result, according to the present invention, mobile station 322, when receiving blocks comprising a downlink temporary block flow, would examine the USF field when mobile station 322 has information to transmit in an uplink temporary blockflow. If the USF is a zero value, then the channel would be evaluated by mobile station 322 as being "available", and mobile station 322 may therefore begin transmitting new uplink temporary block flow information. Base station 320 acknowledges receiptof the uplink data blocks 328, 332, 336, 340 and so forth, by sending a direct acknowledgement in respective downlink radio link control blocks 330, 334, 338, 342 and so forth. Unlike a true "random access" method, the present invention utilizes thepacket data traffic channel (PDTCH), rather than a random access channel, since the synchronization between mobile station 322 and base station 320 is already known. There is therefore no need for the special shortened GSM "access burst" to be used. Asa result, an initial radio link control block data block (and therefore user information) may be sent along with the uplink access procedure, further streamlining uplink access. The generally used GSM uplink access method is shown in FIG. 5 and includesexchange of channel request access burst 236, immediate assignment message 238, packet resource request 240, and packet uplink assignment 242. In this way, the present invention removes the need for this interchange.

FIG. 9 is a flowchart of indirect carrier sense multiple access with directed acknowledgement in a mobile station, according to the present invention. As illustrated in FIGS. 8 and 9, upon completion of downlink temporary block flow setup 324,mobile station 322 determines whether mobile station 322 has data available to transmit on the uplink, Step 342. Once mobile station 322 has data to transmit on the uplink, mobile station 322 then determines whether downlink temporary block flow setup324 has been completed, Step 344. If downlink temporary block flow setup 324 has not been completed, mobile station 322 waits, Step 340, until downlink temporary block flow setup 324 is completed.

If downlink temporary block flow is complete in Step 344, mobile station 322 determines whether downlink radio link control block 326 has been received, Step 346. If downlink radio link control block 326 has not been received, the processreturns to Step 340 so that mobile station 322 waits until downlink radio link control block 326 is received. If downlink radio link control block 326 has been received, mobile station 322 determines whether USF contained within downlink radio linkcontrol block 326 is equal to zero, Step 348, which indicates that base station 320 has indicated to mobile station that an uplink channel is available and is not being utilized by any mobile station. If uplink channel is available, i.e., USF containedwithin downlink radio link control block 326 is equal to zero in Step 348, mobile station 322 sends first uplink radio link control block 328 to base station using uplink channel, Step 354. Once mobile station 322 sends uplink radio link control block328 to base station 320 in Step 354, the process returns to Step 340, and the process continues for the next uplink radio link control block.

If uplink channel is not available, i.e., USF contained within downlink radio link control block 326 is not equal to zero in Step 348, mobile station 322 determines whether USF of downlink radio link control block is equal to identifier of mobilestation 322, which indicates that mobile station 322 can transfer the next data block.

FIG. 10 is a flowchart of indirect carrier sense multiple access with directed acknowledgement in a base station, according to the present invention. As illustrated in FIG. 10, during a downlink temporary block flow setup period, base station320 determines whether a channel is allocated to mobile station 322 in an uplink temporary block flow, Step 362. If a channel has already been allocated, the process returns to the start, Step 360. If a channel has not been allocated, base station 320determines whether uplink radio link control block 328 has been received from mobile station 322, Step 364. If uplink radio link control block 328 has not been received, base station 320 sets the USF in downlink radio link control block 326 equal tozero, Step 370, and sends downlink radio link control block 326, Step 372. The process then returns to Step 360 so that base station 320 continues to send an indication of uplink channel availability to mobile station 322 in subsequent downlink radiocontrol blocks until base station 320 receives the initial uplink radio link control data block on the timeslot allocated by base station 320 as a contingent uplink timeslot number.

If base station 320 determines in Step 364 that uplink radio link control block 328 has been received, base station 320 then makes a determination as to whether uplink radio link control block 328 has a USF value equal to a mobile station havinga valid downlink temporary block flow, Step 366. If uplink radio link control block 328 does not have a USF value equal to a mobile station having a valid downlink temporary block flow, the process returns to Step 370, so that base station 320 continuesto send an indication of uplink channel availability to mobile station 322 in subsequent downlink radio control blocks until base station 320 receives the next uplink radio link control data block on the timeslot allocated by base station 320 as acontingent uplink timeslot number. However, if uplink radio link control block 328 does have a USF value equal to a mobile station having a valid downlink temporary block flow in Step 366, base station 320 sets the USF value in the downlink radio linkcontrol data block 330 to the value of USF in mobile station 322. In the example shown in FIG. 8, the USF value in mobile station 322 is "110" as indicated in downlink temporary block flow setup 324. Base station then sends downlink radio link controldata block 330 with USF equal to "110" as a directed acknowledgement of the USF of mobile station 322. The process then returns to Step 360 so that base station 320 waits for receipt of subsequent uplink radio link control data block 322 in Step 364,and the process continues until the end of the associated uplink temporary block flow indicated by the countdown procedure in the last several radio link control data blocks of the uplink data transfer by mobile station 322, or until mobile station 322no longer has data to transmit.

While a particular embodiment of the present invention has been shown and described, modifications may be made. It is therefore intended in the appended claims to cover all such changes and modifications which fall within the true spirit andscope of the invention.

* * * * *
Other References
• Draft ETSI EN 300 911 V6.5.0 (1999-07) Title: Digital Cellular Telecommunications System (Phase 2+) Radio Sybsystem Link Control (GSM 05.08 Version 6.5.0 Release 1997).
• Draft ETSI EN 301 349 V6.4.0 (1999-07) Title: Digital Cellular Telecommunications System (Phase 2+) General Packet Radio Service (GPRS); Mobile Station (MS)-Base Station System (BSS) Interface; Radio Link Control/Medium Access Control (RLC/MAC) Protocol (GSM 04.60 Version 6.4.0 Release 1997).


Title: Mobile terminal and base station in a packet radio services network
Patent ID: US6791944
Issue Date: September 14, 2004

Abstract A mobile terminal for communicating with a base station in a packet radio services network. The terminal has a processor for determining one of a plurality of channels for communication between the mobile terminal and the base station; for digitally coding speech to provide speech information; for assembling speech information into speech packets; and for generating channel allocation requests for a channel in which to send speech packets. A radio transmitter is provided for transmitting the requests and the packets to a base station in the network. A radio receiver receives identities of channels allocated by the base station for the mobile terminal to transmit on. The processor is responsive to each received channel allocation to determine that packets are sent on the allocated channel. In the GPRS since a channel is released when there is no packet to transmit, higher traffic levels can be obtained using the same number of radio channels
We claim:
1. A mobile terminal for communicating with a base station in a packet radio services network, said terminal comprising a processor for determining one of a plurality of channels forcommunication between the mobile terminal and the base station; for digitally coding speech to provide speech information; and for assembling speech information into speech packets; for generating channel allocation requests in which to send speechpackets; a radio transmitter for transmitting the requests and the packets to a base station in the network; and a radio receiver for receiving identities of channels allocated by the base station for the mobile terminal to transmit on, said processorbeing responsive to each received channel allocation to determine that packets are sent on the allocated channel wherein if a request to send is not granted, the processor is arranged to discard speech information until a further request is granted andthe further request is delayed by a predetermined period where the delay is increased if successive requests are not granted and where following a predetermined maximum delay, the delay is reduced.
2. A mobile terminal as claimed in claim 1, wherein the processor is arranged to implement a layered protocol, and wherein each packet is given a network and transport layer header in a subnetwork dependent convergence protocol layer (SNDCP).
3. A mobile terminal in accordance with claim 1 further comprising a voice activity detector, wherein the processor is responsive to detection of voice activity by the voice activity detector, to generate a request for a channel allocation inwhich to send voice packets, and on receipt of a channel identity, to send an address header uncompressed on that channel and subsequently to send packets with compressed headers which do not contain the destination address on the identified channel,until the voice activity detector detects no voice activity.
4. A mobile terminal as claimed in claim 3, wherein the processor is arranged to construct packets of an equal number n of frames, the processor being further arranged to implement a logical link layer protocol (LLC) which adds its own LLCheader information comprising a service access point identifier defining speech service to each packet, and to divide the total LLC plus an SNDCP header into n parts of equal length and to place one header part before each frame in the packet.
5. A mobile terminal as claimed in claim 4, wherein in the physical layer, in each frame, the header and the most important bits speech information are coded using a convolutional code, and a subset of important bits of the speech information are coded using a cyclic redundancy check

BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to packet radio services networks.
2. Description of the Related Art
Standards are being defined for a general packet radio services network (GPRS)
SUMMARY OF THE INVENTION
The invention is based on the recognition that if suitably designed a packet services network could carry speech.
To this end, in accordance with the invention there is provided a mobile terminal for communicating with a base station in a packet radio services network, said terminal including a processor for determining one of a plurality of channels forcommunication between the mobile terminal and the base station; for digitally coding speech to provide speech information; for assembling speech information into speech packets; and for generating channel allocation requests for a channel in which tosend speech packets; a radio transmitter for transmitting the requests and the packets to a base station in the network; and a radio receiver for receiving identities of channels allocated by the base station for the mobile terminal to transmit on, saidprocessor being responsive to each received channel allocation to determine that packets are sent on the allocated channel.
In the GPRS since a channel is released when there is no packet to transmit, higher traffic levels can be obtained using the same number of radio channels.
Preferably, if a request to send is not granted, the processor is arranged to discard speech information until a further request is granted. As speech is highly time sensitive, it is better to discard the information than to send the informationdelayed. The discard produces clipping which, as long as it is not too frequent, is tolerable by the user.
The processor is preferably arranged so that when a request to send is not granted, a further request is delayed by a predetermined period.
The delay is preferably increased if successive requests are not granted.
Following a predetermined maximum delay, the delay is reduced.
The processor is preferably arranged to implement a layered protocol in which each packet is given a header in a subnetwork dependent convergence protocol layer (SNDCP).
Because of the time sensitive nature of speech the header is preferably a RTP/UDP/IP header.
The mobile terminal preferably includes a voice activity detector, and the processor is preferably responsive to detection of voice activity by the voice activity detector, to generate a request for a channel allocation in which to send voicepackets, and on receipt of a channel identity, to send the an address header uncompressed on that channel once and subsequently to send packets with compressed headers which do not contain the destination address on the identified channel, until thevoice activity detector detects no voice activity.
The processor is preferably arranged to construct packets of an equal number n of frames, the processor being further arranged to implement a logical link layer protocol (LLC) which adds its own LLC header information comprising a service accesspoint identifier defining speech service to each packet, and to divide the total LLC plus SNDCP header into n parts of equal length and to place one header part before each frame in the packet. This provides that every frame in the packet has the sameformat and allows a common protection strategy to be applied to each frame. The header information can be given an error correcting code. Speech is more error tolerant, however. More important parts of the speech information can be coded in order toidentify that there is an error, in which case the frame is discarded. Less important parts of the speech information can be left unprotected.
Thus, in the physical layer, in each frame, the header and the most important bits speech information are preferably coded using a convolutional code, and a subset of the important bits of the speech information are coded using a cyclicredundancy check.
The invention also extends to a base station including a radio receiver for receiving requests from mobile stations to send data packets and requests to send speech packets and operable on a plurality of channels to receive data packets andspeech packets; a processor for reserving a predetermined number of said channels for receiving coded speech packets, and for allocating nominating a free one of said predetermined number responsive to a request channel allocation request in which tosend a speech packet; and a transmitter for transmitting the allocated channel to the mobile station.
By dynamically managing the number of channels reserved for speech, optimum service can be given to both speech services and to data services given changing respective demands.
The invention also extends to a base station including a radio receiver for receiving requests from mobile stations to send data packets and requests to send speech packets and operable on a plurality of channels to receive data packets andspeech packets; a processor for nominating channels for a mobile station to send speech packets and for processing packets in a talk spurt comprising a single destination address header followed by a plurality of speech packets not containing adestination address, for transmission over the network.
The invention further extends to a base station including a radio receiver for receiving requests from mobile stations to send data packets and requests to send speech packets and operable on a plurality of channels to receive data packets andspeech packets, a processor for implementing a protocol which recovers network and transport layer headers and logical link layer headers for a packet, from equal parts of each frame in the packet.
The processor may be operative in each frame to correct errors in the header and the most subjectively important bits of the speech information only.
BRIEF DESCRIPTION OF THE DRAWINGS
One embodiment of the invention will now be described, by way of example, with reference to the accompanying drawings, in which:
FIG. 1 is a block diagram of a GPRS mobile terminal and base station embodying the invention;
FIG. 2 shows schematically the operation of RLC/MAC protocol;
FIG. 3 shows network layer protocol layers;
FIG. 4 shows the SNDCP model operation to support voice;
FIG. 5 shows the format of an LLC-PDU;
FIG. 6 shows the organization of TDMA frames in GPRS;
FIG. 7 shows the GPRS TDMA multiframe structure;
FIG. 8 shows the partition of channels in a GPRS carrying speech;
FIG. 9 shows how source coded bits output from the codec are protected and
FIG. 10 shows the operation of each layer in the protocol.
DETAILED DESCRIPTION
Referring to the drawings, a mobile terminal 2 has an antenna 4 coupled to a duplexer 6. The duplexer 6 is coupled to a transmitter 8 and a receiver 10. Signals received by the receiver 10 are fed to a processor 12. Sound waves of speech aretransduced to analog electrical signals by a microphone 14 and the analog signals are converted to digital by the processor which may have one or more central processing units (not shown). An analog to digital converter may be a self contained unit 16. The processor processes the digitized speech which is then coded by a parametric codec algorithm, e.g. EFR, to produce speech frames. A codec may be a self contained unit 18.
A voice activity detector algorithm detects the presence of speech distinguished from silences. The voice activity detector may be a self contained unit 20. When speech is detected, the processor assembles speech information output from thecodec with network and transport layer headers into fixed length packets of two frames and sends a channel allocation request.
As may be seen from the block diagram of FIG. 2, if the channel allocation request is refused, a delay is introduced before a new request is sent and speech frames to occurring during the delay are discarded.
A base station 22 has an antenna 24 feeding a duplexer 26. A radio receiver 28 sends packets received from the mobile terminal 2, to a processor 30. Data for transmission to the mobile terminal 2 is sent to a radio transmitter 32 coupled to theduplexer 26.
Network layer protocols, illustrated in FIG. 3, are intended to be capable of operating over services derived from a wide variety of subnetworks and data links. GPRS was designed from the outset to support several network layer protocolsproviding network transparency for the users of the service. Introduction of new network layer protocols to be transferred over GPRS was to be allowed without any changes to the GPRS network, a function carried out by the subnetwork dependentconvergence protocol (SNDCP). In addition, SNDCP 40 carries out header and data compression, and multiplexing of data coming from different sources to be sent over the LLC layer 42.
IP is used as the network protocol with RTP being used to provide support for the real time streaming by supplying timestamp information and packet sequencing. SNDCP currently only provides for TCP/IP and EP(v4) header compression byimplementing the RFC1144 compression algorithm. However, the SNDCP specifications also allow for additions to the list of supported compression protocols, according to the requirements of new applications and services. The present system employs theRTP/UDP/IP protocols which involve an overhead of 40 octets, corresponding to 320 bits.
Using packets of two frames length, it is necessary to support some form of compression for these transport and network layer headers. Indeed, if the CS-I channel coding scheme were to be used, the combined RTP/UDP/IP headers would occupy theentire information payloads of two radio blocks, leaving no space for any speech information or logic link control (LLC) headers.
A high compression efficiency may be obtained by treating the IP/UDP and RTP headers together rather than separately. Although it is contrary to the ethos of layered architecture, crossing these protocol layer boundaries is appropriate becausethe same function is being applied across all layers.
There are two main properties of the transmitted packets which are used to carry out header compression. The first factor-of-two reduction in data rate comes from the observation that half of the bytes in the headers remain constant over thelife of the connection. An obvious example is the source and destination addresses and ports. The uncompressed header is sent once, during a connection establishment phase. These fields are then deleted from the compressed headers that follow withoutany real loss of information.
The remaining compression comes from differential coding on the changing fields to reduce their size. In particular, for RTP header compression, a big gain in efficiency comes from the observation that although several fields change in everypacket, such as the sequence number and the timestamp, the difference from packet to packet is often constant, and therefore the second-order difference is zero. By making use of these properties, the massive combined RTP/UDP/IP header can be reduced totwo bytes or three bytes, depending upon whether a header checksum is used. As at least part of the end-to-end link includes at least one mobile propagation path, which is by its very nature subject to error, it would be useful to include the headerchecksum in the scheme employed. Although it is not be used for error correction or frame retransmission schemes, it gives an indication that part of the header may be corrupted and to ignore the timing information for that particular packet.
SNDCP also supports data compression by means of the V.42 bis data compression algorithm. However, as the application layer which sits above the SNDCP layer already includes a lossy source coder in the form of a speech codec, there stands littleto be gained by applying data compression by means of entropy coding, as most redundancy in the original information would have been already extracted. In addition, source coding modifies the speech coder bit patterns and makes it difficult to applydifferential channel coding to the speech frame according to the subjective importance of the different bit positions.
FIG. 4 shows the SNDCP model operation to support voice. Analysis of the Voice over GPRS delay budget showed that maximum payload efficiency can be achieved by encapsulating two speech frames into a single network packet. Increasing the numberof speech frames accommodated by a single network packet brings about a proportional increase in the packet buffering delay, thereby increasing the maximum end-to-end delay threshold of 200 ms.
The SNDCP layer 40 therefore accepts the combined RTP/UDP/IP headers 50 and the speech frames through two different service access points. Header compression 52 is carried out, and the resulting header 55 segmented into two sections 54, 56 foraddition to the two speech frames 58, 60 that is encapsulated into that particular packet. This system allows for the two radio link control (RLC) blocks containing the speech information to have exactly the same layout, and therefore use exactly thesame channel coding scheme for both blocks. As the forward error correction is tailored to catering for the properties of a particular speech coder, it is important to ensure that each bit position with a radio block refers to the same bit positionwithin a speech frame for all transmitted blocks. The first received speech frame belonging to a particular network packet is forwarded directly to the lower layer without waiting for the second frame to arrive.
The Logical Link Control layer 42 operates above the RLC 44 and BSSGP 46 layers in the illustrated architecture to provide highly reliable logical links between a mobile terminal and its serving GPRS support node (SGSN). Its main functions aredesigned towards supporting such a reliable link and they include sequence control of LLC frames across a logical link, the detection of transmission, format and operational errors on logical link connection, the notification of unrecoverable errors andflow control.
The operation of the LLC protocol can be better understood by examining the format of an LLC-PDU shown in FIG. 5.
As can be seen, the LLC frame header is divided into two main sections, the Address Field 70 and the Control Field 72. In the Address field is the Service Access Point Identifier (SAPI) 74. This represents a point at which LLC services can beaccessed and provides a means by which the Quality of Service priority can be defined. As ten out of a possible sixteen different identifiers currently remain vacant in the specifications, a new SAPI can be defined for voice services, instructing thelayers above, namely the SNDCP and the BSSGP about the priority required by voice packets over data traffic. The conventional control field contains two sub-fields, represented by N(S) 76 and N(R) 78, whose function it is to determine the position of aparticular LLC frame within a sequence of frames constituting a single network PDU. However, this function is superfluous within the context of the Voice over GPRS system there is no segmenting of network-PDUs, as each N-PDU fits exactly into a singleLLC-PDU. These fields are therefore to be omitted within the context of transporting real-time voice packets, without any loss of functionality. Each LLC-PDU conventionally ends with a 24-bit long footer containing a frame check sequence. This enablesthe LLC layer 42 to ensure that the LLC frame is free of errors (within the capabilities of the CRC check) before passing it on to the network layer at the SGSN for delivery through the backbone network. Should errors be found, it signals forretransmission by means of the RLC layer 44 selective repeat request system. However, as repeat request systems are not used in the present implementation, and as there already exists a cyclic redundancy check at the physical layer, the FCS field 90within the LLC-PDU is also omitted without affecting the functionality of the system when transporting speech services. Indeed, should this field be retained, it would be merely ignored by the receiving process, as even if errors were to be detected,the process would still forward the packet, because as already described, coded speech has an inherent information corruption tolerance.
The system therefore accepts the two segments of the SNDC-PDU containing the two speech frames which belong to the same network packet, and add the new, 8-bit LLC header containing the SAPI for voice services to the first arriving segment so thatthe two frames in the packet have headers of equal length. This is then forwarded to the RLC 44/MAC 45 layer for immediate dispatch over the radio interface. When the peer LLC process at the BSS receives the first radio block containing informationfrom that particular LLC-frame, it identifies that it contains speech information by examining the service access point identifier. This information instructs the process not to look for further header information after the first 8 bits, and not toexpect any frame check sequence. The space freed by removing these fields is used to carry more user payload information.
The RLC/MAC (medium access control) layer provides services for the transfer of upper layer PDUs using a shared medium between multiple mobile stations and the network. This transfer may take place in the form of unacknowledged operation oracknowledged operation, according to the nature of the service required. As its very name implies, the RLC/MAC protocol actually consists of two separate protocols with different functions. The Radio Link Control Layer defines the procedures forsegmentation and reassembly of LLC PDUs into RLC/MAC blocks and in RLC acknowledged mode of operation, for the Backward Error Correction (BEG) procedures enabling the selective retransmission of unsuccessfully delivered RLC/MAC blocks. When operating inthe RLC acknowledged mode, the RLC layer preserves the order of higher layer PDUs provided to it. On the other hand, the MAC (Medium Access Control) function defines procedures that enable multiple mobile stations to share a common transmission mediumwhich may consist of several physical channels. The GPRS MAC protocol allows for a single user to use more than one timeslot concurrently so as to increase throughput. In addition, the same timeslot may be multiplexed between up to eight users, so asto increase the number of users operating on a given set of system resources. In the Voice over GPRS system, there is no provision for multislotting, and for the duration of time for which a channel is occupied by a single user transmitting speechinformation, this channel is not multiplexed with any other users. In addition, no use is made of the retransmission functions of the RLC layer
GPRS shares the same radio interface as the GSM circuit-switched voice system. This means that each physical frequency channel is divided into eight traffic channels by means of time-division multiplexing. Each timeslot within a TDMA frame canbe dynamically allocated to GPRS services or GSM services according to the relative shifts in demands for the two services, with those channels allocated to packet data traffic being referred to as Packet Data Channels (PDCH)
Each time division multiple access (TDMA) frame 100 lasts for 4.615 ms and can accommodate eight PDCHs within its eight timeslots 102. The data to be transmitted by means of the Packet Data Traffic Channels (PDTCH) is segmented into units of 114bits each, which are then encapsulated into radio bursts for insertion into a single TDMA timeslot which lasts for 576.8 us. This means that each RLC/MAC block consisting of 456 bits is segmented and interleaved into four consecutive radio bursts.
GPRS multiplexing differs from that found GSM circuit-switched speech in the way the TDMA frames are organized into multiframes. Whereas GSM supports 26-frame and 51-frame multiframes, in GPRS the TDMA frames are organized into 239.980 ms-longmultiframes 104 consisting of 52 frames. This organization is divided into 12 radio blocks of four frames each, and four idle frames as shown in FIG. 6.
Referring to FIG. 7, it should be noted that all 52 bursts in the picture belong to the same timeslot, and consequently to the same PDCH. In GPRS, several logical channels not found in GSM circuit-switched services are introduced andaccommodated onto the GSM physical channels (PDCH)
Referring to FIG. 8, the Packet Data Traffic Channel, which is used to actually carry the speech information and the Packet Access Grant channel which is used for channel contention are mapped onto the PDCH by what is known as the `Master-Slave`concept. In this system, at least one PDCH (or timeslot), acting as a master, accommodates user data and dedicated signaling, and packet common control channels that carry all necessary control signaling for initiating packet transfer. Such controlsignaling refers only to the access bursts on the packet random access channel (PRACH) 120. All other PDCHs, acting as slaves are used for user data transfer and dedicated signaling only.
For GPRS the master channel is capable of bearing both the PRACH and a PDTCH simultaneously by sharing the physical channel between the two logical channels by means of a time-division multiplexing mechanism. Usually, the bulk of the physicalresources of the Master timeslot are allocated to carrying user data, with one block every t blocks being dedicated to supporting random access attempts, where t is typically an integer with value 3, 4 or 5. In this way, whereas seven slave channels areentirely dedicated to supporting voice and data traffic, the eighth channel, which is usually time slot zero (TSO) is used both as a traffic channel, and as a random access medium for all terminals operating at that particular radio frequency channel.
The present implementation for supporting voice services requires that the master channel be left for control signaling only and not be allowed to accept any PDTCHs. The reason for making such a reservation is the extra delay that sharing themaster channel would impose on the average access time. If, for example, t was set to three, one radio block out of every three available radio blocks in the master channel would be dedicated to supporting the PRACH. This means that a mobile terminalhas to wait approximately 3 RLC blocks, equivalent to 55.3 ms until it is allowed to fire the next random access. This extra delay is clearly desirable for real-time voice services.
The remaining seven slave channels are then divided into those Packet Data Traffic Channels dedicated to voice services 122 and those dedicated to data services 124. The system does not allow a PDTCH to share voice and data services, as thedelay requirements are radically different. By allowing GPRS voice users access to channels dedicated to carrying voice services, better control can be made by the base station to ensure that the required Quality of Service over the radio link in termsof access delay and speech frame loss rate are met.
The operation of the RLC/MAC protocol is summarized in FIG. 2. If a GPRS-terminal is voice-capable and wishes to initiate a conversation over the GPRS network, it initiates a call-setup procedure. In this process, that the Base Station monitorsthe current load in the cell of operation of the user and determine if it can support another voice user. If it can, the base station informs the mobile terminal that it has been admitted to the network and may initiate transmission of voice packets. The mobile terminal then goes into an idle mode where it waits for an indication from the terminal's voice activity detector (VAD) that a speech activity has been detected and a talkspurt has begun. A random access burst is sent over the PRACH 140, andit waits for a reply from the Base Station over the PAGCH 142. If channel resources are available, it indicates to the MS that it has allocated a single channel for the transmission of the talkspurt. In order not to be any more channel inefficient thanthe equivalent GSM circuit-switched voice services, multi-slotting is not enabled for the transmission of voice services over GPRS. This means that a single talkspurt may be transmitted in a single timeslot only, and the base station does not allocatemore than this single PDTCH for this purpose. The mobile terminal then proceeds to transmit all the RLC/MAC blocks belonging to that particular talkspurt 144. On the onset of the following silence period, the mobile station stops transmitting RLCblocks, thereby indicating to the BS that it is releasing the channel 146. This means that channel contention occurs at the beginning of each talkspurt only, and once a terminal has acquired use of a traffic channel, it only relinquishes it when all theLLC frames corresponding to that particular talkspurt have been transmitted. The Base Station then allocates this channel to the pool of channels available for the transmission of voice services. If, however, there are no available PDTCHs allocated forthe support of voice services, the BS informs the MS of the situation 148. The mobile station then enters a random exponential backoff period, at the completion of which it reattempts to access the channel 150.
As real-time speech is time-sensitive information, all the RLC/MAC blocks corresponding to speech frames generated during the backoff, save the most current one are discarded 152. For every failure to access the channel, a counter is incremented154 and used to determine the duration of the backoff period. If however a collision occurs between access bursts in the same timeframe on the PRACH, the Base Station is not able to determine which MS initiated the request, and so is unable to respond. The mobile terminal notices that such a collision has occurred by employing a time set to a value slightly longer than the average response time of the base station. Should this value expire, the mobile station enters the backoff period in the same wayas when channel access has been denied.
Both data terminals as well as voice terminals implement an exponential backoff algorithm, where after an unsuccessful access attempt, a uniform distributed random number w e[0,2.sub.n+1 ] is drawn, where n is the number of access attempts. Fordata terminals, the next random attempt is tried after a waiting time of w*8.5 ms, while for voice terminals, multipliers of 4.615 ms and 8.5 ms respectively are used.
As real-time conversational voice is a delay-critical service it is important that the medium access algorithm is geared towards keeping the access delay of the system to a minimum. A uniform dist
Posted by sajjad6103275



Posted by sajjad6103275



Posted: 15-February-2009 06:01:51 AM By: sajjad6103275


Posted: 15-February-2009 06:02:32 AM By: sajjad6103275

FIELD OF THE INVENTION

The present invention relates generally to data transmission in a GPRS/EDGE system, and in particular, the present invention relates to set up of an uplink packet data transfer in a GPRS/EDGE system using an indirect carrier sense multiple accesswith directed acknowledgement.

BACKGROUND OF THE INVENTION

Global System for Mobile Communications (GSM) General Packet Radio Service (GPRS) and Enhanced Data for Global Evolution (EDGE) are intended to allow the service subscriber the ability to send and receive data in an end-to-end packet transfermode without utilizing network resources in the circuit-switched mode. GPRS and EDGE permit the efficient use of radio and network resources when data transmission characteristics are i) packet based, ii) intermittent and non-periodic, iii) possiblyfrequent, with small transfers of data, e.g. less than 500 octets, or iv) possibly infrequent, with large transfers of data, e.g. more than several hundred kilobytes. User applications may include Internet browsers, electronic mail and so on.

Efforts are presently underway to further develop the European Telecommunications Standards Institute (ETSI) GPRS and EDGE specifications to support the wireline concept of voice over Internet protocol (VoIP). This effort includes the abilityfor a mobile station to terminate and originate a VoIP call as an endpoint on the Internet. The current definition for GPRS and EDGE supports the concept of both a packet-switched radio environment and a packet-switched network environment, i.e. thepacket abstraction of the Internet is carried through to the air interface in the form of intermittently accessible radio resources based upon the availability of radio resources and the demand for the interchange of user data.

FIG. 1 is a schematic diagram of a complete packet data transfer in a GPRS/EDGE radio environment. As illustrated in FIG. 1, packet switching in the radio environment is achieved using the concept of a packet data transfer 100, referred to as a"temporary block flow" (TBF). The temporary block flow 100, which includes a data transfer setup phase 102, a data transfer phase 104, and a data transfer teardown phase 106, is regarded as the basic unit of data interchange within the GPRS/EDGEenvironment. As a result, temporary block flow 100 may be thought of conceptually as its three components, data transfer setup phase 102, data transfer phase 104, and data transfer teardown phase 106, occurring sequentially in time. It is understoodthat the amount of time for the setup of a temporary block flow for GPRS varies, and is dependent on channel conditions, radio resource availability, network congestion and so on.

Although GPRS and EDGE have been specified with the objective of interchanging packet based user data, the application for most such data interchange is not of a real-time nature. Voice over IP presents several challenges to the GPRS/EDGEdomain, one of which is the availability of data transfer capability in the uplink direction. For example, when the mobile VoIP user speaks into the phone, a temporary block flow is required to be set up in the uplink direction as soon as possible. However, the time required by GPRS and EDGE to set up such an uplink temporary block flow is prohibitive when compared to the generally accepted maximum turnaround delay for voice telephony, which is 125 ms. Furthermore, VoIP telephony would require theaddition of other mechanisms which would enable the radio layers to have knowledge of the type of information they are carrying at any given time.

In particular, the amount of time required for data transfer setup phase 104 has proven to be excessively long, resulting in problems associated with both round-trip turnaround time, and throughput, as a function of the duty-cycle reductionrequired for setting up an acknowledgement at the upper (network) layers, e.g. the transport layer.

Accordingly, what is needed is a method for enabling a mobile station to more rapidly set up an uplink packet data transfer in a GPRS/EDGE environment.
BRIEF DESCRIPTION OF THE DRAWINGS

The features of the present invention which are believed to be novel are set forth with particularity in the appended claims. The invention, together with further objects and advantages thereof, may best be understood by making reference to thefollowing description, taken in conjunction with the accompanying drawings, in the several figures of which like reference numerals identify like elements, and wherein:

FIG. 1 is a schematic diagram of a complete packet data transfer in a radio environment.

FIG. 2 is a schematic diagram of a GPRS system according to the present invention.

FIG. 3 is a schematic diagram of modification of a user data stream as the user data stream passes through specified layers of a GPRS system.

FIG. 4 is a schematic diagram of a multiframe structure for packet data channels.

FIG. 5 is a data flow diagram of stream-oriented data transmitted between a mobile station and a network.

FIG. 6 is a schematic diagram of a dynamic timeslot allocation for medium access control.

FIG. 7 is a schematic diagram of a fixed timeslot allocation for medium access control.

FIG. 8 is a schematic diagram of signaling logic for establishing an uplink packet data transfer.

FIG. 9 is a flowchart of indirect carrier sense multiple access with directed acknowledgement in a mobile station.

FIG. 10 is a flowchart of indirect carrier sense multiple access with directed acknowledgement in a base station

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

The present invention is related for allowing mobile stations to more rapidly set up an uplink packet data transfer in a GPRS/EDGE system using an indirect carrier sense multiple access with directed acknowledgement (ICSMA/DA), whereby the mobilestation would be notified of when a transmit resource is not in use, allowing the mobile station to transmit on this resource only if a downlink transfer is in progress and then acknowledging the mobile station's access to the medium directly.

FIG. 2 is a schematic diagram of a GPRS system according to the present invention. As illustrated in FIG. 2, a GPRS system 200 includes a mobile station 202 sending and receiving packet data from an internet application 204 to a remote internetapplication 206 through a base station system 208. While a single base station system 208 and mobile station 202 is illustrated in FIG. 2, it is understood that GPRS system 200 includes multiple numbers of base station systems and mobile stations. Mobile station 202 includes a GPRS/EDGE subsystem 210 for processing signaling messages received from base station system 208, and signals received from internet application 204 through transport and network layers 212. GPRS/EDGE subsystem 210 includesa medium access control (MAC) layer 211, and adds header overhead for sub-network convergence/divergence protocol (SNDCP), and logical link control (LLC). A protocol control unit 214, includes a medium access control layer 213, and is coupled to orcontained within base station system 208, and interfaces with GPRS/EDGE subsystem 210 of mobile station 202, and with internet application 206 through transport and network layers 216. Internet transport layers 212 and 216 include a transmission controlprotocol (TCP) layer 218 which TCP packetizes stream-oriented user data, and an internet protocol (IP) layer 220 which assigns an address to the packetized data.

FIG. 3 is a schematic diagram of modification of a user data stream as the user data stream passes through specified layers of a GPRS system. As illustrated in FIG. 3, a user data stream of infinite length is modified as the user data streampasses through GPRS system 200. For example, as illustrated in FIGS. 2 and 3, as the user data stream passes through transmission control protocol layer 218, and an RLC layer, the data stream is divided into a TCP packet 222 that includes a payload 224that is 536 octets in length and a transmission control protocol header packet 226 that is twenty octets in length, giving TCP packet 222 a total length of 556 octets. As TCP packet 222 subsequently passes through internet protocol layer 220, anadditional twenty octet internet protocol header 228 is appended to TCP packet 222, forming an IP packet 230 having a total length of 576 octets. An additional four octet SNDCP header 232 is appended to IP packet 230, forming an SNDCP packet 234 havinga total length of 580 octets, and an additional four octet logical link control header 236 is appended to SNDCP packet 234 forming a logical link control packet 238 having a total length of 584 octets. As a result, the user data stream has a totallength of 584 octets as the data stream exits logical link control.

Next, radio link control divides the 584 octet logical link control packet 238 into a certain number of radio link control data blocks, the exact number of which depends upon the channel coding scheme used. For example, in a CS-1 channel codingscheme, the number of radio link control blocks needed is equal to (LLC frame length/RLC payload length)+(LLC frame length MOD RLC payload length), which for the 584 octet logical link control frame is equal to 31 radio link control blocks. In a CS-2channel coding scheme, the number of radio link control blocks needed is equal to (LLC frame length/RLC payload length)+(LLC frame length MOD RLC payload length), which for the 584 octet logical link control frame is equal to 21 radio link controlblocks.

FIG. 4 is a schematic diagram of a multiframe structure for packet data channels. Assuming a perfect schedule of one radio link control block transmitted on each available block period for a single timeslot transfer, raw throughput may becomputed based upon the length of time required to send a certain number of radio link control data blocks. As illustrated in FIG. 5, a packet data control channel is organized as a multiframe 260 having fifty-two frames 262 and twelve data blocks B0B11, in which each data block B0 B11 is distributed over four time division multiple access (TDMA) frames. An "idle" or "search" frame 264 located after every three data blocks, enables the mobile station to perform adjacent cell signal measurements,synchronization and verification of synchronization status on adjacent cells, interference measurements, and so forth. Each data block B0 B11 is made up of four frames, each of which has a frame period f equal to 4.61538 milliseconds, and a block periodb that is equal to 18.4616 milliseconds, while each idle frame 264 has an idle frame period I that is equal to the frame period f, or 4.61538 milliseconds. The total period of the multiframe 260 structure of the packet data channel is equal to 240milliseconds.

The length of time required TR to send a certain number of radio link control data blocks Nb is calculated using the following equation: TR=(Nb×b)+((Nb/3)×f) EQUATION 1 while the raw data throughput Rd iscalculated using the following equation: Rd=(number of payload octets/TR)×8 EQUATION 2

Using Equations 1 and 2, the time required to send all of the radio link control blocks in a logical link control frame in a CS-1 coding scheme (i.e. 31 blocks) is equal to 0.618462 sec. The throughput is the number of payload octets (584)divided by the time required to send them plus their overhead (0.618462) times 8 bits per octet, which is equal to 7000 bits/second. In terms of an overhead analysis of the CS-1 coding scheme, theoretical throughput is approximately equal to 9050bits/sec. The overhead of scheduling, i.e. the fact that there are idle frames that prevent the scheduling of every consecutive block reduces the effective throughput by 4/52 to approximately 8861 bits/second. The overhead of radio link control headers,i.e. three octets per block, reduces the effective throughput by 3/22 to approximately 7652 bits/second. The overhead of the logical link control header, four octets, reduces the effective throughput by 4/584 to approximately 7599 bits/second. Finally,the overhead of the SNDCP header, 4 octets, reduces the effective throughput by 4/580 to approximately, and the overhead of the Internet protocol suite, i.e. TCP and IP headers, reduces the effective throughput by 40/576 to approximately 7000bits/second.

Similarly, the time required to send all of the radio link control blocks in a logical link control frame (i.e. 21 blocks) in a CS-2 coding scheme is equal to 0.42 second, and the throughput is the number of payload octets (584) divided by thetime required to send them plus their overhead (0.42) times 8 bits per octet, which is equal to 10,209 bits/second. Theoretical throughput on channel at CS-2 is approximately equal to 13,400 bits/sec. The overhead of scheduling, i.e. the fact that thereare idle frames that prevent the scheduling of every consecutive block reduces the effective throughput by 4/52 to approximately 12,369 bits/second. The overhead of radio link control headers, i.e. three octets per block, reduces the effectivethroughput by 3/32 to approximately 11,209 bits/second. The overhead of the logical link control header, four octets, reduces the effective throughput by 4/584 to approximately 11,132 bits/second. Finally, the overhead of the SNDCP header, four octets,reduces the effective throughput by 4/580 to approximately 11,055 bits/second, and the overhead of the Internet protocol suite, i.e. TCP and IP headers, reduces the effective throughput by 40/576 to approximately 10,209 bits/second.

FIG. 5 is a data flow diagram of stream-oriented data transmitted between a mobile station and a network. As illustrated in FIGS. 2 and 5, when stream-oriented data 213 is transmitted from remote internet application 206 to mobile station 202during a downlink period 300 for sending data long a downlink, the data is first divided into packets at TCP layer 218, and given an address at IP layer 220 of transport and network layer 216, and sent to protocol control unit 214 of base station system208 as a TCP/IP packet 302.

As illustrated in FIGS. 3 and 5, during downlink period 300, TCP/IP packet 302 includes overhead associated with logical link control packet 238 and SNDCP packet 234, and it is assumed that for every TCP/IP packet 302, there is a correspondinglogical link control packet 238 and SNDCP packet 234 as well. The actions associated with transmitting the information over the air interface begin when a logical link control frame containing the user information in the form of an encapsulatedtransport/network/SNDCP packet enters protocol control unit 214 of base station system 208.

As illustrated in FIGS. 2 and 5, assuming that mobile station 202 is camped on the network in packet idle mode, when appropriate, base station system 208 begins a setup sequence of a downlink setup period 224 by sending a package paging request215 to GPRS/EDGE subsystem 210 of mobile station 202. In response, after receiving a random access burst 217 from GPRS/EDGE subsystem 210, protocol control unit 214 sends an immediate assignment message 219 and a packet downlink message 221, detailingthe parameters of the assignment, e.g. over what channel the transfer would take place, when the transfer would start, and so on. Protocol control unit 214 sends a series of radio link control data blocks 226 to GPRS/EDGE subsystem 210 after receiving apacket control acknowledge message 222 from GPRS/EDGE subsystem 210.

Depending upon the availability of schedulable blocks, packet paging request message 215 may require from 81 to 1721 ms, followed by random access burst 217 from mobile station 102, which typically requires 9.6 ms. Immediate assignment message219 contains a starting time that may range from 37 ms to 3 minutes in the future, but typically ranges from 13 to 25 TDMA frame periods, or 60 115 ms. Additional signaling associated with exchanging packet downlink assignment message 221 and a packetcontrol acknowledgement message 222 are included in downlink setup period 224. It is therefore assumed that downlink setup period 224 may be equal to a starting time, which is in fact what is observable in an actual system. As a result, the timerequired for downlink setup period 224 is a minimum of approximately 849 ms, a maximum of approximately 2643 ms and an average of approximately 1746 ms.

After the starting time has been reached, protocol control unit 218 sends GPRS/EDGE subsystem 210 a temporary block flow containing radio link control data blocks 226. Once GPRS/EDGE subsystem 210 has received all downlink blocks, GPRS/EDGEsubsystem 210 assembles, processes and transmits a resulting single data packet 228 to IP layer 220 of transport and network layers 212, which then sends data packet 228 to TCP layer 218 of transport and network layers 212.

Assuming perfectly available radio resources so that data may be sent on every schedulable downlink block on a single timeslot, the time to transmit all blocks during data transfer period 225 for a 536 octet user data payload is approximatelyequal to 0.618462 seconds for a CS-1 coding scheme, and 0.420 seconds for a CS-2 coding scheme. The downlink temporary block flow terminates after a last radio link control data block is sent if sending radio link control on protocol control unit 214has no more data to be sent and a radio link controller timer T3192 expires before radio link control receives more data to be sent from logical link control, which is the case when a transmission control protocol transmission starts in "congestioncontrol" (slow-start) mode. The temporary block flow is always torn down after the blocks making up the first transmission control protocol packet are transmitted, causing the downlink temporary block flow to incur the overhead of temporary block flowbeing setup again for the subsequent blocks.

TCP layer 218 of transport and network layers 212 performs redundancy checking and makes a determination that data packet 228 has been received properly. IP layer 220 of transport and network layers 212 then includes the packet data in astream-oriented output 230 to internet application 204 and issues a TCP acknowledgement (TCP ACK) message 232 to TCP layer 218 of transport and network layers 216 on the far end of the virtual circuit. TCP ACK message 232 is processed by SNDCP/LLC andRLC layers as before, but in an uplink direction.

Radio link controller of GPRS/EDGE subsystem 210 of remote transport and network layers 216 receives a TCP/IP/SNDCP/LLC packet containing TCP ACK message 232, but cannot begin a setup sequence corresponding to an uplink setup period 234 fortransmission of TCP ACK message 232 until a radio link control timer T3192 of protocol control unit 114 has expired. As a result, a downlink temporary block flow corresponding to downlink period 300 that carries TCP/IP packet 302 that was initiallysent, must be torn down completely before uplink period 234 for setting up TCP ACK message 232 may begin.

For example, upon receiving TCP ACK message 232, GPRS/EDGE subsystem 210 sends a channel request access burst 236 to protocol control unit 214, which responds by sending an immediate assignment message 238. GPRS/EDGE subsystem 210 then sends apacket resource request message 240 to protocol control unit 214 requesting resources for a temporary block flow. Protocol control unit 214 responds with a packet uplink assignment message 242, which is acknowledged by GPRS/EDGE subsystem 210 in apacket control acknowledge message 244. Data blocks 246 containing TCP ACK message 232 and teardown are then transmitted from GPRS/EDGE subsystem 210 to protocol control unit 214 during an acknowledge data transfer period 248. Protocol control unit 214then transmits data blocks 246 to transport and network layers 216 in a TCP acknowledge message 304. As a result, uplink setup period 234 and acknowledge data transfer period 248 form an uplink period 306 that is required for TCP acknowledge message 232to reach transport and network layer 216 in corresponding TCP acknowledge message 304. Once the required TCP ACK message is received by transport and network layers 216, a next TCP/IP data packet message 250 is sent from transport and network layers 216to GPRS/EDGE subsystem 210 via protocol control unit 214.

The period required for the initial setup of uplink setup period 234 is dependent upon components such as the periodic occurrence of a random access channel (RACH), the starting time sent in immediate assignment message 238, and the starting timesent in packet uplink assignment message 242. The periodic occurrence of a random access channel can range from 41 217 TDMA frame periods, assuming a case of 41 frame periods, or 190 ms. The starting time sent in immediate assignment message 238 mayrange from 9 TDMA frame periods to 3 minutes, but is typically from 9 25 TDMA frame periods, or 42 115 ms, while the starting time sent in packet uplink assignment message 242 may range from 9 TDMA frame periods to 3 minutes, but is typically around 20TDMA periods, or 92 ms. As a result, initial setup of uplink setup period 234 is typically a minimum of approximately 320 ms, a maximum of approximately 480 ms, and an average of approximately 320 ms. This exceeds the generally accepted maximum end toend delay of 125 milliseconds.

TCP ACK message 232 has a length of 40 octets, which combined with the overhead of both logical link control header 236 and SNDCP header 232 is equal to 48 octets. Assuming perfectly available radio resources so that data may be sent on everyschedulable downlink block on a single timeslot, the time to transmit all data blocks 246 during acknowledge data transfer period 248 for a 40 octet TCP/IP ACK payload is equal to 60 ms (3 RLC data blocks) for the CS-1 coding scheme, and 37 ms (2 RLCdata blocks) for the CS-2 coding scheme.

Mobile station 202 receives the right to transmit on the uplink by using either a dynamic timeslot allocation medium control access (MAC) mode or a fixed timeslot allocation medium control access mode. FIG. 6 is a schematic diagram of a dynamictimeslot allocation for medium access control. As illustrated in FIG. 6, in dynamic timeslot allocation, a mobile station 300 receives a downlink radio link control/medium access control (RLC/MAC) control block 302 from a base station 304 that includesa special address, referred to as an uplink state flag (USF) 306, along with RLC/MAC data 308. If USF 306 (a 3-bit quantity) is identical to that of a USF assigned to mobile station 300, then mobile station 300 has the right to transmit in the next timedivision multiple access (TDMA) frame. A data block addressed to a second mobile station 310 may contain USF information for mobile station 300.

FIG. 7 is a schematic diagram of a fixed timeslot allocation for medium access control. As illustrated in FIG. 7, in fixed timeslot allocation, a mobile station 312 periodically receives a starting time and a bit-map 314 from a base station 316,representing a base and offset of future timeslots on which the mobile station is to transmit. In this way, mobile station 312 is informed of when temporary block flow starts and receives bitmap 314 representing timeslots on which mobile station 312 isto transmit relative to the starting time, so that mobile station 312 transmits in timeslots assigned by the starting time and allocation bitmap.

The present invention utilizes a USF field for both fixed and dynamic allocation MAC mode, when the mobile station is not engaged in an uplink temporary block flow, although USF value is given a different meaning, as described below. The presentinvention includes a collision avoidance (CA) mechanism that utilizes the already-present USF address in the RLC/MAC control block to enable the rapid creation of an uplink temporary block flow, when there is already a downlink temporary block flow inprogress. Since the USF value is receivable by multiple mobile stations on the radio resource, the USF value assignment serves as an indirect lock on the resource.

According to the present invention, the USF field is recognized during an active downlink temporary block flow as a "channel availability" indicator and a "directed acknowledgement field". FIG. 8 is a schematic diagram of signaling logic forestablishing an uplink packet data transfer according to the present invention. According to the present invention, the mobile station, when receiving blocks comprising a downlink temporary block flow, examines the USF field when it has information totransmit in an uplink temporary block flow. If the USF were a zero value, then the channel would be evaluated as "available", and the mobile station may therefore begin transmitting its new uplink TBF information.

In particular, as illustrated in FIGS. 5 and 8, while a base station 320 and a mobile station 322 are in a downlink temporary block flow setup 324 of downlink setup period 224, which includes the assignment of a mobile station USF address, suchas "110", for example, base station 320, through medium access control layer 213, sends a USF address to mobile station 322 by which mobile station 322 will be identified for the duration of the downlink temporary block flow for data transfer period 225resulting from downlink temporary block flow setup 324, along with a contingent uplink timeslot number on which mobile station 322 may transmit. Once GPRS/EDGE data flows in the downlink direction in data transfer period 225, base station 320, throughmedium access layer 213, indicates uplink channel availability 326 to mobile station 322 by sending the value USF=000. If mobile station 322 has data to transmit on the uplink, mobile station 322 transmits a first uplink radio link control data block328 on the timeslot indicated by base station 320 as a contingent uplink timeslot number.

Base station 320 receives the first uplink data block 328 and knows how to associate a temporary flow identifier (TFI) to a USF value of mobile station 322. Base station 320 acknowledges the USF address of mobile station 322 in the next downlinkradio link control data block 330 by inserting the USF value (that serves to indirectly address a mobile station) of mobile station 322 into the header of next downlink radio link control data block 330. According to the present invention, the insertedUSF value serves as an acknowledgement to the sending mobile station 322 and as a "channel busy" indication to other mobiles desiring to transmit. Mobile station 322, through medium access layer 211, interprets next downlink radio link control datablock 330 with a USF value located in a header at the beginning of radio link control data block 330 as an acknowledgement that the first uplink data block 328 was correctly received by base station 320, and sends a subsequent uplink radio link controlblock 332. This procedure is then continued for the remaining portion 343 of the uplink temporary block flow 248, until the end of uplink temporary block flow 248, which is indicated in the usual manner to base station 320 by the countdown procedure inthe last several radio link control data blocks. In the countdown procedure, mobile station 322, during transmission of the last few data blocks 246, decrements a variable in the header of data blocks 246 to inform base station 320 that the uplink datablock flow is about to end. This knowledge helps base station 320 allocate to another mobile station.

As a result, according to the present invention, mobile station 322, when receiving blocks comprising a downlink temporary block flow, would examine the USF field when mobile station 322 has information to transmit in an uplink temporary blockflow. If the USF is a zero value, then the channel would be evaluated by mobile station 322 as being "available", and mobile station 322 may therefore begin transmitting new uplink temporary block flow information. Base station 320 acknowledges receiptof the uplink data blocks 328, 332, 336, 340 and so forth, by sending a direct acknowledgement in respective downlink radio link control blocks 330, 334, 338, 342 and so forth. Unlike a true "random access" method, the present invention utilizes thepacket data traffic channel (PDTCH), rather than a random access channel, since the synchronization between mobile station 322 and base station 320 is already known. There is therefore no need for the special shortened GSM "access burst" to be used. Asa result, an initial radio link control block data block (and therefore user information) may be sent along with the uplink access procedure, further streamlining uplink access. The generally used GSM uplink access method is shown in FIG. 5 and includesexchange of channel request access burst 236, immediate assignment message 238, packet resource request 240, and packet uplink assignment 242. In this way, the present invention removes the need for this interchange.

FIG. 9 is a flowchart of indirect carrier sense multiple access with directed acknowledgement in a mobile station, according to the present invention. As illustrated in FIGS. 8 and 9, upon completion of downlink temporary block flow setup 324,mobile station 322 determines whether mobile station 322 has data available to transmit on the uplink, Step 342. Once mobile station 322 has data to transmit on the uplink, mobile station 322 then determines whether downlink temporary block flow setup324 has been completed, Step 344. If downlink temporary block flow setup 324 has not been completed, mobile station 322 waits, Step 340, until downlink temporary block flow setup 324 is completed.

If downlink temporary block flow is complete in Step 344, mobile station 322 determines whether downlink radio link control block 326 has been received, Step 346. If downlink radio link control block 326 has not been received, the processreturns to Step 340 so that mobile station 322 waits until downlink radio link control block 326 is received. If downlink radio link control block 326 has been received, mobile station 322 determines whether USF contained within downlink radio linkcontrol block 326 is equal to zero, Step 348, which indicates that base station 320 has indicated to mobile station that an uplink channel is available and is not being utilized by any mobile station. If uplink channel is available, i.e., USF containedwithin downlink radio link control block 326 is equal to zero in Step 348, mobile station 322 sends first uplink radio link control block 328 to base station using uplink channel, Step 354. Once mobile station 322 sends uplink radio link control block328 to base station 320 in Step 354, the process returns to Step 340, and the process continues for the next uplink radio link control block.

If uplink channel is not available, i.e., USF contained within downlink radio link control block 326 is not equal to zero in Step 348, mobile station 322 determines whether USF of downlink radio link control block is equal to identifier of mobilestation 322, which indicates that mobile station 322 can transfer the next data block.

FIG. 10 is a flowchart of indirect carrier sense multiple access with directed acknowledgement in a base station, according to the present invention. As illustrated in FIG. 10, during a downlink temporary block flow setup period, base station320 determines whether a channel is allocated to mobile station 322 in an uplink temporary block flow, Step 362. If a channel has already been allocated, the process returns to the start, Step 360. If a channel has not been allocated, base station 320determines whether uplink radio link control block 328 has been received from mobile station 322, Step 364. If uplink radio link control block 328 has not been received, base station 320 sets the USF in downlink radio link control block 326 equal tozero, Step 370, and sends downlink radio link control block 326, Step 372. The process then returns to Step 360 so that base station 320 continues to send an indication of uplink channel availability to mobile station 322 in subsequent downlink radiocontrol blocks until base station 320 receives the initial uplink radio link control data block on the timeslot allocated by base station 320 as a contingent uplink timeslot number.

If base station 320 determines in Step 364 that uplink radio link control block 328 has been received, base station 320 then makes a determination as to whether uplink radio link control block 328 has a USF value equal to a mobile station havinga valid downlink temporary block flow, Step 366. If uplink radio link control block 328 does not have a USF value equal to a mobile station having a valid downlink temporary block flow, the process returns to Step 370, so that base station 320 continuesto send an indication of uplink channel availability to mobile station 322 in subsequent downlink radio control blocks until base station 320 receives the next uplink radio link control data block on the timeslot allocated by base station 320 as acontingent uplink timeslot number. However, if uplink radio link control block 328 does have a USF value equal to a mobile station having a valid downlink temporary block flow in Step 366, base station 320 sets the USF value in the downlink radio linkcontrol data block 330 to the value of USF in mobile station 322. In the example shown in FIG. 8, the USF value in mobile station 322 is "110" as indicated in downlink temporary block flow setup 324. Base station then sends downlink radio link controldata block 330 with USF equal to "110" as a directed acknowledgement of the USF of mobile station 322. The process then returns to Step 360 so that base station 320 waits for receipt of subsequent uplink radio link control data block 322 in Step 364,and the process continues until the end of the associated uplink temporary block flow indicated by the countdown procedure in the last several radio link control data blocks of the uplink data transfer by mobile station 322, or until mobile station 322no longer has data to transmit.

While a particular embodiment of the present invention has been shown and described, modifications may be made. It is therefore intended in the appended claims to cover all such changes and modifications which fall within the true spirit andscope of the invention.

* * * * *
Other References
• Draft ETSI EN 300 911 V6.5.0 (1999-07) Title: Digital Cellular Telecommunications System (Phase 2+) Radio Sybsystem Link Control (GSM 05.08 Version 6.5.0 Release 1997).
• Draft ETSI EN 301 349 V6.4.0 (1999-07) Title: Digital Cellular Telecommunications System (Phase 2+) General Packet Radio Service (GPRS); Mobile Station (MS)-Base Station System (BSS) Interface; Radio Link Control/Medium Access Control (RLC/MAC) Protocol (GSM 04.60 Version 6.4.0 Release 1997).


Title: Mobile terminal and base station in a packet radio services network
Patent ID: US6791944
Issue Date: September 14, 2004

Abstract A mobile terminal for communicating with a base station in a packet radio services network. The terminal has a processor for determining one of a plurality of channels for communication between the mobile terminal and the base station; for digitally coding speech to provide speech information; for assembling speech information into speech packets; and for generating channel allocation requests for a channel in which to send speech packets. A radio transmitter is provided for transmitting the requests and the packets to a base station in the network. A radio receiver receives identities of channels allocated by the base station for the mobile terminal to transmit on. The processor is responsive to each received channel allocation to determine that packets are sent on the allocated channel. In the GPRS since a channel is released when there is no packet to transmit, higher traffic levels can be obtained using the same number of radio channels
We claim:
1. A mobile terminal for communicating with a base station in a packet radio services network, said terminal comprising a processor for determining one of a plurality of channels forcommunication between the mobile terminal and the base station; for digitally coding speech to provide speech information; and for assembling speech information into speech packets; for generating channel allocation requests in which to send speechpackets; a radio transmitter for transmitting the requests and the packets to a base station in the network; and a radio receiver for receiving identities of channels allocated by the base station for the mobile terminal to transmit on, said processorbeing responsive to each received channel allocation to determine that packets are sent on the allocated channel wherein if a request to send is not granted, the processor is arranged to discard speech information until a further request is granted andthe further request is delayed by a predetermined period where the delay is increased if successive requests are not granted and where following a predetermined maximum delay, the delay is reduced.
2. A mobile terminal as claimed in claim 1, wherein the processor is arranged to implement a layered protocol, and wherein each packet is given a network and transport layer header in a subnetwork dependent convergence protocol layer (SNDCP).
3. A mobile terminal in accordance with claim 1 further comprising a voice activity detector, wherein the processor is responsive to detection of voice activity by the voice activity detector, to generate a request for a channel allocation inwhich to send voice packets, and on receipt of a channel identity, to send an address header uncompressed on that channel and subsequently to send packets with compressed headers which do not contain the destination address on the identified channel,until the voice activity detector detects no voice activity.
4. A mobile terminal as claimed in claim 3, wherein the processor is arranged to construct packets of an equal number n of frames, the processor being further arranged to implement a logical link layer protocol (LLC) which adds its own LLCheader information comprising a service access point identifier defining speech service to each packet, and to divide the total LLC plus an SNDCP header into n parts of equal length and to place one header part before each frame in the packet.
5. A mobile terminal as claimed in claim 4, wherein in the physical layer, in each frame, the header and the most important bits speech information are coded using a convolutional code, and a subset of important bits of the speech information are coded using a cyclic redundancy check

BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to packet radio services networks.
2. Description of the Related Art
Standards are being defined for a general packet radio services network (GPRS)
SUMMARY OF THE INVENTION
The invention is based on the recognition that if suitably designed a packet services network could carry speech.
To this end, in accordance with the invention there is provided a mobile terminal for communicating with a base station in a packet radio services network, said terminal including a processor for determining one of a plurality of channels forcommunication between the mobile terminal and the base station; for digitally coding speech to provide speech information; for assembling speech information into speech packets; and for generating channel allocation requests for a channel in which tosend speech packets; a radio transmitter for transmitting the requests and the packets to a base station in the network; and a radio receiver for receiving identities of channels allocated by the base station for the mobile terminal to transmit on, saidprocessor being responsive to each received channel allocation to determine that packets are sent on the allocated channel.
In the GPRS since a channel is released when there is no packet to transmit, higher traffic levels can be obtained using the same number of radio channels.
Preferably, if a request to send is not granted, the processor is arranged to discard speech information until a further request is granted. As speech is highly time sensitive, it is better to discard the information than to send the informationdelayed. The discard produces clipping which, as long as it is not too frequent, is tolerable by the user.
The processor is preferably arranged so that when a request to send is not granted, a further request is delayed by a predetermined period.
The delay is preferably increased if successive requests are not granted.
Following a predetermined maximum delay, the delay is reduced.
The processor is preferably arranged to implement a layered protocol in which each packet is given a header in a subnetwork dependent convergence protocol layer (SNDCP).
Because of the time sensitive nature of speech the header is preferably a RTP/UDP/IP header.
The mobile terminal preferably includes a voice activity detector, and the processor is preferably responsive to detection of voice activity by the voice activity detector, to generate a request for a channel allocation in which to send voicepackets, and on receipt of a channel identity, to send the an address header uncompressed on that channel once and subsequently to send packets with compressed headers which do not contain the destination address on the identified channel, until thevoice activity detector detects no voice activity.
The processor is preferably arranged to construct packets of an equal number n of frames, the processor being further arranged to implement a logical link layer protocol (LLC) which adds its own LLC header information comprising a service accesspoint identifier defining speech service to each packet, and to divide the total LLC plus SNDCP header into n parts of equal length and to place one header part before each frame in the packet. This provides that every frame in the packet has the sameformat and allows a common protection strategy to be applied to each frame. The header information can be given an error correcting code. Speech is more error tolerant, however. More important parts of the speech information can be coded in order toidentify that there is an error, in which case the frame is discarded. Less important parts of the speech information can be left unprotected.
Thus, in the physical layer, in each frame, the header and the most important bits speech information are preferably coded using a convolutional code, and a subset of the important bits of the speech information are coded using a cyclicredundancy check.
The invention also extends to a base station including a radio receiver for receiving requests from mobile stations to send data packets and requests to send speech packets and operable on a plurality of channels to receive data packets andspeech packets; a processor for reserving a predetermined number of said channels for receiving coded speech packets, and for allocating nominating a free one of said predetermined number responsive to a request channel allocation request in which tosend a speech packet; and a transmitter for transmitting the allocated channel to the mobile station.
By dynamically managing the number of channels reserved for speech, optimum service can be given to both speech services and to data services given changing respective demands.
The invention also extends to a base station including a radio receiver for receiving requests from mobile stations to send data packets and requests to send speech packets and operable on a plurality of channels to receive data packets andspeech packets; a processor for nominating channels for a mobile station to send speech packets and for processing packets in a talk spurt comprising a single destination address header followed by a plurality of speech packets not containing adestination address, for transmission over the network.
The invention further extends to a base station including a radio receiver for receiving requests from mobile stations to send data packets and requests to send speech packets and operable on a plurality of channels to receive data packets andspeech packets, a processor for implementing a protocol which recovers network and transport layer headers and logical link layer headers for a packet, from equal parts of each frame in the packet.
The processor may be operative in each frame to correct errors in the header and the most subjectively important bits of the speech information only.
BRIEF DESCRIPTION OF THE DRAWINGS
One embodiment of the invention will now be described, by way of example, with reference to the accompanying drawings, in which:
FIG. 1 is a block diagram of a GPRS mobile terminal and base station embodying the invention;
FIG. 2 shows schematically the operation of RLC/MAC protocol;
FIG. 3 shows network layer protocol layers;
FIG. 4 shows the SNDCP model operation to support voice;
FIG. 5 shows the format of an LLC-PDU;
FIG. 6 shows the organization of TDMA frames in GPRS;
FIG. 7 shows the GPRS TDMA multiframe structure;
FIG. 8 shows the partition of channels in a GPRS carrying speech;
FIG. 9 shows how source coded bits output from the codec are protected and
FIG. 10 shows the operation of each layer in the protocol.
DETAILED DESCRIPTION
Referring to the drawings, a mobile terminal 2 has an antenna 4 coupled to a duplexer 6. The duplexer 6 is coupled to a transmitter 8 and a receiver 10. Signals received by the receiver 10 are fed to a processor 12. Sound waves of speech aretransduced to analog electrical signals by a microphone 14 and the analog signals are converted to digital by the processor which may have one or more central processing units (not shown). An analog to digital converter may be a self contained unit 16. The processor processes the digitized speech which is then coded by a parametric codec algorithm, e.g. EFR, to produce speech frames. A codec may be a self contained unit 18.
A voice activity detector algorithm detects the presence of speech distinguished from silences. The voice activity detector may be a self contained unit 20. When speech is detected, the processor assembles speech information output from thecodec with network and transport layer headers into fixed length packets of two frames and sends a channel allocation request.
As may be seen from the block diagram of FIG. 2, if the channel allocation request is refused, a delay is introduced before a new request is sent and speech frames to occurring during the delay are discarded.
A base station 22 has an antenna 24 feeding a duplexer 26. A radio receiver 28 sends packets received from the mobile terminal 2, to a processor 30. Data for transmission to the mobile terminal 2 is sent to a radio transmitter 32 coupled to theduplexer 26.
Network layer protocols, illustrated in FIG. 3, are intended to be capable of operating over services derived from a wide variety of subnetworks and data links. GPRS was designed from the outset to support several network layer protocolsproviding network transparency for the users of the service. Introduction of new network layer protocols to be transferred over GPRS was to be allowed without any changes to the GPRS network, a function carried out by the subnetwork dependentconvergence protocol (SNDCP). In addition, SNDCP 40 carries out header and data compression, and multiplexing of data coming from different sources to be sent over the LLC layer 42.
IP is used as the network protocol with RTP being used to provide support for the real time streaming by supplying timestamp information and packet sequencing. SNDCP currently only provides for TCP/IP and EP(v4) header compression byimplementing the RFC1144 compression algorithm. However, the SNDCP specifications also allow for additions to the list of supported compression protocols, according to the requirements of new applications and services. The present system employs theRTP/UDP/IP protocols which involve an overhead of 40 octets, corresponding to 320 bits.
Using packets of two frames length, it is necessary to support some form of compression for these transport and network layer headers. Indeed, if the CS-I channel coding scheme were to be used, the combined RTP/UDP/IP headers would occupy theentire information payloads of two radio blocks, leaving no space for any speech information or logic link control (LLC) headers.
A high compression efficiency may be obtained by treating the IP/UDP and RTP headers together rather than separately. Although it is contrary to the ethos of layered architecture, crossing these protocol layer boundaries is appropriate becausethe same function is being applied across all layers.
There are two main properties of the transmitted packets which are used to carry out header compression. The first factor-of-two reduction in data rate comes from the observation that half of the bytes in the headers remain constant over thelife of the connection. An obvious example is the source and destination addresses and ports. The uncompressed header is sent once, during a connection establishment phase. These fields are then deleted from the compressed headers that follow withoutany real loss of information.
The remaining compression comes from differential coding on the changing fields to reduce their size. In particular, for RTP header compression, a big gain in efficiency comes from the observation that although several fields change in everypacket, such as the sequence number and the timestamp, the difference from packet to packet is often constant, and therefore the second-order difference is zero. By making use of these properties, the massive combined RTP/UDP/IP header can be reduced totwo bytes or three bytes, depending upon whether a header checksum is used. As at least part of the end-to-end link includes at least one mobile propagation path, which is by its very nature subject to error, it would be useful to include the headerchecksum in the scheme employed. Although it is not be used for error correction or frame retransmission schemes, it gives an indication that part of the header may be corrupted and to ignore the timing information for that particular packet.
SNDCP also supports data compression by means of the V.42 bis data compression algorithm. However, as the application layer which sits above the SNDCP layer already includes a lossy source coder in the form of a speech codec, there stands littleto be gained by applying data compression by means of entropy coding, as most redundancy in the original information would have been already extracted. In addition, source coding modifies the speech coder bit patterns and makes it difficult to applydifferential channel coding to the speech frame according to the subjective importance of the different bit positions.
FIG. 4 shows the SNDCP model operation to support voice. Analysis of the Voice over GPRS delay budget showed that maximum payload efficiency can be achieved by encapsulating two speech frames into a single network packet. Increasing the numberof speech frames accommodated by a single network packet brings about a proportional increase in the packet buffering delay, thereby increasing the maximum end-to-end delay threshold of 200 ms.
The SNDCP layer 40 therefore accepts the combined RTP/UDP/IP headers 50 and the speech frames through two different service access points. Header compression 52 is carried out, and the resulting header 55 segmented into two sections 54, 56 foraddition to the two speech frames 58, 60 that is encapsulated into that particular packet. This system allows for the two radio link control (RLC) blocks containing the speech information to have exactly the same layout, and therefore use exactly thesame channel coding scheme for both blocks. As the forward error correction is tailored to catering for the properties of a particular speech coder, it is important to ensure that each bit position with a radio block refers to the same bit positionwithin a speech frame for all transmitted blocks. The first received speech frame belonging to a particular network packet is forwarded directly to the lower layer without waiting for the second frame to arrive.
The Logical Link Control layer 42 operates above the RLC 44 and BSSGP 46 layers in the illustrated architecture to provide highly reliable logical links between a mobile terminal and its serving GPRS support node (SGSN). Its main functions aredesigned towards supporting such a reliable link and they include sequence control of LLC frames across a logical link, the detection of transmission, format and operational errors on logical link connection, the notification of unrecoverable errors andflow control.
The operation of the LLC protocol can be better understood by examining the format of an LLC-PDU shown in FIG. 5.
As can be seen, the LLC frame header is divided into two main sections, the Address Field 70 and the Control Field 72. In the Address field is the Service Access Point Identifier (SAPI) 74. This represents a point at which LLC services can beaccessed and provides a means by which the Quality of Service priority can be defined. As ten out of a possible sixteen different identifiers currently remain vacant in the specifications, a new SAPI can be defined for voice services, instructing thelayers above, namely the SNDCP and the BSSGP about the priority required by voice packets over data traffic. The conventional control field contains two sub-fields, represented by N(S) 76 and N(R) 78, whose function it is to determine the position of aparticular LLC frame within a sequence of frames constituting a single network PDU. However, this function is superfluous within the context of the Voice over GPRS system there is no segmenting of network-PDUs, as each N-PDU fits exactly into a singleLLC-PDU. These fields are therefore to be omitted within the context of transporting real-time voice packets, without any loss of functionality. Each LLC-PDU conventionally ends with a 24-bit long footer containing a frame check sequence. This enablesthe LLC layer 42 to ensure that the LLC frame is free of errors (within the capabilities of the CRC check) before passing it on to the network layer at the SGSN for delivery through the backbone network. Should errors be found, it signals forretransmission by means of the RLC layer 44 selective repeat request system. However, as repeat request systems are not used in the present implementation, and as there already exists a cyclic redundancy check at the physical layer, the FCS field 90within the LLC-PDU is also omitted without affecting the functionality of the system when transporting speech services. Indeed, should this field be retained, it would be merely ignored by the receiving process, as even if errors were to be detected,the process would still forward the packet, because as already described, coded speech has an inherent information corruption tolerance.
The system therefore accepts the two segments of the SNDC-PDU containing the two speech frames which belong to the same network packet, and add the new, 8-bit LLC header containing the SAPI for voice services to the first arriving segment so thatthe two frames in the packet have headers of equal length. This is then forwarded to the RLC 44/MAC 45 layer for immediate dispatch over the radio interface. When the peer LLC process at the BSS receives the first radio block containing informationfrom that particular LLC-frame, it identifies that it contains speech information by examining the service access point identifier. This information instructs the process not to look for further header information after the first 8 bits, and not toexpect any frame check sequence. The space freed by removing these fields is used to carry more user payload information.
The RLC/MAC (medium access control) layer provides services for the transfer of upper layer PDUs using a shared medium between multiple mobile stations and the network. This transfer may take place in the form of unacknowledged operation oracknowledged operation, according to the nature of the service required. As its very name implies, the RLC/MAC protocol actually consists of two separate protocols with different functions. The Radio Link Control Layer defines the procedures forsegmentation and reassembly of LLC PDUs into RLC/MAC blocks and in RLC acknowledged mode of operation, for the Backward Error Correction (BEG) procedures enabling the selective retransmission of unsuccessfully delivered RLC/MAC blocks. When operating inthe RLC acknowledged mode, the RLC layer preserves the order of higher layer PDUs provided to it. On the other hand, the MAC (Medium Access Control) function defines procedures that enable multiple mobile stations to share a common transmission mediumwhich may consist of several physical channels. The GPRS MAC protocol allows for a single user to use more than one timeslot concurrently so as to increase throughput. In addition, the same timeslot may be multiplexed between up to eight users, so asto increase the number of users operating on a given set of system resources. In the Voice over GPRS system, there is no provision for multislotting, and for the duration of time for which a channel is occupied by a single user transmitting speechinformation, this channel is not multiplexed with any other users. In addition, no use is made of the retransmission functions of the RLC layer
GPRS shares the same radio interface as the GSM circuit-switched voice system. This means that each physical frequency channel is divided into eight traffic channels by means of time-division multiplexing. Each timeslot within a TDMA frame canbe dynamically allocated to GPRS services or GSM services according to the relative shifts in demands for the two services, with those channels allocated to packet data traffic being referred to as Packet Data Channels (PDCH)
Each time division multiple access (TDMA) frame 100 lasts for 4.615 ms and can accommodate eight PDCHs within its eight timeslots 102. The data to be transmitted by means of the Packet Data Traffic Channels (PDTCH) is segmented into units of 114bits each, which are then encapsulated into radio bursts for insertion into a single TDMA timeslot which lasts for 576.8 us. This means that each RLC/MAC block consisting of 456 bits is segmented and interleaved into four consecutive radio bursts.
GPRS multiplexing differs from that found GSM circuit-switched speech in the way the TDMA frames are organized into multiframes. Whereas GSM supports 26-frame and 51-frame multiframes, in GPRS the TDMA frames are organized into 239.980 ms-longmultiframes 104 consisting of 52 frames. This organization is divided into 12 radio blocks of four frames each, and four idle frames as shown in FIG. 6.
Referring to FIG. 7, it should be noted that all 52 bursts in the picture belong to the same timeslot, and consequently to the same PDCH. In GPRS, several logical channels not found in GSM circuit-switched services are introduced andaccommodated onto the GSM physical channels (PDCH)
Referring to FIG. 8, the Packet Data Traffic Channel, which is used to actually carry the speech information and the Packet Access Grant channel which is used for channel contention are mapped onto the PDCH by what is known as the `Master-Slave`concept. In this system, at least one PDCH (or timeslot), acting as a master, accommodates user data and dedicated signaling, and packet common control channels that carry all necessary control signaling for initiating packet transfer. Such controlsignaling refers only to the access bursts on the packet random access channel (PRACH) 120. All other PDCHs, acting as slaves are used for user data transfer and dedicated signaling only.
For GPRS the master channel is capable of bearing both the PRACH and a PDTCH simultaneously by sharing the physical channel between the two logical channels by means of a time-division multiplexing mechanism. Usually, the bulk of the physicalresources of the Master timeslot are allocated to carrying user data, with one block every t blocks being dedicated to supporting random access attempts, where t is typically an integer with value 3, 4 or 5. In this way, whereas seven slave channels areentirely dedicated to supporting voice and data traffic, the eighth channel, which is usually time slot zero (TSO) is used both as a traffic channel, and as a random access medium for all terminals operating at that particular radio frequency channel.
The present implementation for supporting voice services requires that the master channel be left for control signaling only and not be allowed to accept any PDTCHs. The reason for making such a reservation is the extra delay that sharing themaster channel would impose on the average access time. If, for example, t was set to three, one radio block out of every three available radio blocks in the master channel would be dedicated to supporting the PRACH. This means that a mobile terminalhas to wait approximately 3 RLC blocks, equivalent to 55.3 ms until it is allowed to fire the next random access. This extra delay is clearly desirable for real-time voice services.
The remaining seven slave channels are then divided into those Packet Data Traffic Channels dedicated to voice services 122 and those dedicated to data services 124. The system does not allow a PDTCH to share voice and data services, as thedelay requirements are radically different. By allowing GPRS voice users access to channels dedicated to carrying voice services, better control can be made by the base station to ensure that the required Quality of Service over the radio link in termsof access delay and speech frame loss rate are met.
The operation of the RLC/MAC protocol is summarized in FIG. 2. If a GPRS-terminal is voice-capable and wishes to initiate a conversation over the GPRS network, it initiates a call-setup procedure. In this process, that the Base Station monitorsthe current load in the cell of operation of the user and determine if it can support another voice user. If it can, the base station informs the mobile terminal that it has been admitted to the network and may initiate transmission of voice packets. The mobile terminal then goes into an idle mode where it waits for an indication from the terminal's voice activity detector (VAD) that a speech activity has been detected and a talkspurt has begun. A random access burst is sent over the PRACH 140, andit waits for a reply from the Base Station over the PAGCH 142. If channel resources are available, it indicates to the MS that it has allocated a single channel for the transmission of the talkspurt. In order not to be any more channel inefficient thanthe equivalent GSM circuit-switched voice services, multi-slotting is not enabled for the transmission of voice services over GPRS. This means that a single talkspurt may be transmitted in a single timeslot only, and the base station does not allocatemore than this single PDTCH for this purpose. The mobile terminal then proceeds to transmit all the RLC/MAC blocks belonging to that particular talkspurt 144. On the onset of the following silence period, the mobile station stops transmitting RLCblocks, thereby indicating to the BS that it is releasing the channel 146. This means that channel contention occurs at the beginning of each talkspurt only, and once a terminal has acquired use of a traffic channel, it only relinquishes it when all theLLC frames corresponding to that particular talkspurt have been transmitted. The Base Station then allocates this channel to the pool of channels available for the transmission of voice services. If, however, there are no available PDTCHs allocated forthe support of voice services, the BS informs the MS of the situation 148. The mobile station then enters a random exponential backoff period, at the completion of which it reattempts to access the channel 150.
As real-time speech is time-sensitive information, all the RLC/MAC blocks corresponding to speech frames generated during the backoff, save the most current one are discarded 152. For every failure to access the channel, a counter is incremented154 and used to determine the duration of the backoff period. If however a collision occurs between access bursts in the same timeframe on the PRACH, the Base Station is not able to determine which MS initiated the request, and so is unable to respond. The mobile terminal notices that such a collision has occurred by employing a time set to a value slightly longer than the average response time of the base station. Should this value expire, the mobile station enters the backoff period in the same wayas when channel access has been denied.
Both data terminals as well as voice terminals implement an exponential backoff algorithm, where after an unsuccessful access attempt, a uniform distributed random number w e[0,2.sub.n+1 ] is drawn, where n is the number of access attempts. Fordata terminals, the next random attempt is tried after a waiting time of w*8.5 ms, while for voice terminals, multipliers of 4.615 ms and 8.5 ms respectively are used.
As real-time conversational voice is a delay-critical service it is important that the medium access algorithm is geared towards keeping the access delay of the system to a minimum. A uniform dist

Posted: 16-February-2009 12:30:25 PM By: sagitraz

Please visit:

www.ee.iitb.ac.in/~prakshep/IBMA_lit/manual/manual110.html

www.gsmcommunication.com

Posted: 20-February-2009 06:27:43 AM By: HamidAliKhan

I think you must visit www.ee.iitb.ac.in/~prakshep/IBMA_lit/manual/manual110.html