This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
1 Basics 91.1 Service Overview GPRS 101.2 Service Overview EGPRS 111.3 Support of GPRS QoS classes 121.3.1 Radio Network Planning Impact 13
1.4 Dual Transfer Mode 141.4.1 Radio Network Planning Impact 15
1.5 (E)GPRS MS Multislot Classes 161.6 (E)GPRS General Architecture 171.7 Alcatel (E)GPRS Architecture 191.8 (E)GPRS Protocol Layers (Transmission Plane) 221.9 Alcatel (E)GPRS BSS Hardware support 231.10 Modulation Technique: 8-PSK only for EGPRS 241.11 8-PSK TRA Power Aspects 251.12 (E)GPRS Radio Blocks Structure 291.13 GPRS Channel Coding 311.14 EGPRS Channel Coding 341.15 Radio Link Adaptation Overview 391.16 Automatic ReQuest for repetition (ARQ) 401.17 Type-I ARQ mechanism 411.18 Type-I ARQ in GPRS 421.19 Type-I ARQ in EGPRS 431.20 (E)GPRS radio physical channel: PDCH Concept 471.21 (E)GPRS Multiframe 481.22 (E)GPRS Logical Channels 491.23 Master/Slave PDCH concept 511.24 Temporary Block Flow 521.25 Resources Sharing 541.26 MS multiplexing co-ordination 581.27 GPRS mobility management (GMM) states for MS 611.28 Radio Resource (RR) operating modes for MS 621.29 Attach procedure 631.30 PDP context activation 651.31 Location management 661.32 Routing Area 671.33 Network Mode of Operation (NMO) 681.34 TBF establishment 691.35 UL TBF establishment on CCCH, 1 phase access 701.36 UL TBF establishment on CCCH, 2 phases access 721.37 DL TBF establishment on CCCH 741.38 System information broadcasting on BCCH 751.39 System information broadcasting on PBCCH 771.40 (E)GPRS Transmission Aspects 801.40 TRX Classes Concept 811.41 Two Abis Links per BTS 84
2 B9 features 852.1 Enhanced Packet Cell Reselection (R4 MSs) 862.1.1 Radio Network Impact 87
2.2 Extended Uplink TBF Mode 882.2 Radio Network Planning Impact 892.3 Enhanced support of E-GPRS (EDGE) in uplink 912.3.1 Radio Network Planning Impact 93
2.4 Counter Improvements for Release B9 942.4.1 Radio Network Planning Impact 98
2.11 RMS_I2 Improvements 1142.11.1 Radio Network Planning Impact 115
3 (E)GPRS Radio Algorithms 1163.1 Cell Reselection Overview 1173.2 Cell reselection: NC0 mode, no PBCCH established 1213.3 Cell reselection: NC0 mode, PBCCH established 1233.4 Cell reselection execution: NC0 in PTM 1303.5 Cell reselection: NC2 mode 1323.6 GPRS redirection 1433.7 GPRS Power Control: Overview 1453.8 GPRS Power Control: Measurements 1463.9 GPRS Power Control: Algorithm 1503.10 Link adaptation: DL GPRS Radio Link Control 1533.11 Link adaptation: UL GPRS Radio Link Control 1573.12 Link adaptation in EGPRS: New metrics 1603.13 Link adaptation: DL EGPRS Radio Link Control 1613.14 EGPRS Link Adaptation Decision 1633.15 TRX ranking/TRX transmission pool set-up 1643.16 TRX capability for PS traffic 1663.17 Radio Resource Allocation: Overview 1673.18 Radio Resource Allocation: PDCH state 1683.19 TRX selection for EGPRS TBFs 1713.20 Radio Resource Allocation: EGPRS TBFs 1763.21 Radio Resource Allocation: TBF Re-allocation 1793.22 Radio Resource Allocation: Min_PDCH 1803.23 Radio Resource Allocation: Fast initial (E)GPRS access 181
4 General (E)GPRS planning principels 1824.1 Throughput Dependency -> Interference (and Level) 1834.2 Packet data throughput 1844.3 Reference performance point 1854.4 Saturation effect 1864.5 Cell area and throughput 1884.6 Throughput <-> C/I 189
5 (E)GPRS Network intoduction 1915.1 GPRS network planning 1925.2 GPRS Greenfield planning 1935.3 GPRS traffic calculation and traffic analysis 1955.4 GPRS traffic calculationand PS traffic 1965.5 GPRS traffic calculation and user profile 1985.6 GPRS traffic calculation and market applications 1995.7 GPRS traffic calculation and user behavior 2005.8 Customer questionnaire 2015.9 Traffic Model (Example) 2035.10 User mapping 2045.11 Multi-Service 205
5.12 QoS per User Application 2065.13 GPRS traffic calculation 2075.14 Exemplary results of the 3 traffic calculation methods 2125.15 GPRS traffic calculation result 217
6 (E)GPRS Network design 2186.1 General 2196.2 Frequency planning 2226.3 Throughput 2246.4 Link budget 2256.5 Interference analysis on BCCH frequencies 2286.6 Interference analysis on TCH frequencies 2296.7 TRX assignment to GPRS service 2306.8 GPRS Analysis 2316.9 LA and RA planning 2356.10 Quality of Service 245
7 Considerabele features to react (E)GPRS target 2487.1 General 2497.1 Optimization campaign on parameters 2507.2 MPDCH 2517.3 Enhanced PDCH Adaptation & Fast pre-emption 2547.4 User multiplexing 2557.5 PDCH Resource Multiplexing 2567.6 Radio (TBF) Resource Reallocation 2577.7 Coding Scheme Adaptation 2597.8 Cell Reselection 2607.8 GPRS Power Control 2627.8 Features on DL TBF establishment and release 2637.8.1 Delayed DL TBF release 2647.8.2 Fast Downlink TBF re-establishment process 2667.8.3 Non-DRX feature 267
8 GPRS introduction into oerational GSM network 2688.1 General 269
9 GSM Network enhancement features & GPRS 2759.1 Frequency Hopping 2769.2 µ-cell 2789.3 Dual Band 2809.4 Concentric cell 283
10 E-GPRS 28410.1 E-GPRS main differences 285
11 GPRS traffic calculation example 28811.1 Customer questionnaire (Example) 28911.2 User and area distribution determination 29111.3 Traffic demand for CS traffic 29211.4 Traffic demand for packet traffic 29311.5 Network capacity calculation 29711.6 Traffic dimensioning 301
� ETSI standardized solution and can be introduced in two ways:
� CS enhancement: Enhanced circuit-switched data or ECSD
� PS enhancement for GPRS � EGPRS
� EGPRS relies on the introduction of 8-PSK (Eight Phase Shift Keying)modulation technique:
� Same qualities in terms of generating interference on an adjacent channel as GMSK � makes possible to integrate EDGE channels into existing frequency plan
� 8-PSK Symbol rate = GMSK Symbol rate, but one symbol represents now 3 bits instead of 1 bit in GMSK � increased data rates
� Four QoS classes (or traffic classes) are defined:
� The conversational class will be very likely dedicated to real-time conversation. Speech and video conferencing tools are some examples of such applications
� The streaming class corresponds to a real-time stream and enforces mainly constraints on jitter. Video streaming or PoC (Push to takover Celullar) are typical applications for the streaming traffic class.
� The interactive class corresponds to mainly to traditional Internet applications like web browsing. Some differentiation can be donebetween two services by using the traffic handling priority attribute.
� The background class is typically corresponding to Best Effort services. Applications that make use of this class might be e-mail downloading, SMS, or even ftp downloading.
� PFC procedure
• Packet Flow Context (PFC) is a concept introduced starting with R99 3GPP release to ensure that the
BSS is involved in the R99 QoS negotiation. The interest of PFC is to differentiate on the radio
interface the conversational and streaming traffics and to reserve resources for these traffics.
Without the PFC, the BSS only knows the R97/98 QoS parameters (correspond to the interactive and
background R99 QoS classes). It enables to perform admission control and QoS based resource
allocation in the BSS.
• R99 QoS is taken into account if the PFC (Packet Flow Context) procedures are supported by the MS,
the BSS and the SGSN. It allows the BSS B9 to handle streaming and interactive traffics and also to
negotiate the QoS parameters.
• R97/98 QoS should be also taken into account (OP12) if PFC is not supported by the MS or the SGSN in
order to handle interactive traffics or some specific applications as PoC (Push over Cellular).
� QoS subfeatures are of great interest in traffic-driven networks (number of sites determined by the traffic to be carried, not by the coverage per site). They will define the actual traffic shape in the cell by allocating, in a selective manner, resources for (CS and) PS calls. Here a traffic capacity gain is expected (higher traffic levels can be handled with feature activated than without).
� Radio interface impact
• a) Support of PFC feature by RLC/MAC :
- PFC_FEATURE_MODE: this 1 bit field is a part of the R99 extensions in the GPRS_Cell_Options. It is
broadcasted on BCCH (SI13) or PBCCH (PSI1, PSI13 and PSI14) and indicates to the MSs if the network supports
the PFC feature.
- The PFC impact on the one phase access: "If the PFC_FEATURE_MODE is set in the system information and if a
PFC exists for the LLC data to be transferred then the PFI shall be transmitted along with the TLLI of the
mobile station in the RLC extended header during contention resolution. The PFI is not used for contention
resolution but is included to indicate to the network which PFC shall initially be associated with the uplink
TBF.„
• b) RLC/MAC/… messages impacts:
- PI bit (PFI indicator) is created, it indicates the presence of the optional PFI field:
› 0 PFI is not present
› 1 PFI is present if TI field indicates presence of TLLI
› The PFI field indicates a PFI coded as it is defined in TS 44.018.
� RLC/MAC messages impacted are:
• Packet Resource Request : PFI field is added
• (EGPRS) Packet DL ACK/NACK: PFI field is added (if a Channel Request Description is also present)
• UL (EGPRS) RLC data blocks : PFI field is added after the TLLI field (see 44.060 § 10.2.2 and 10.3a.2).
� PFI is included in the following SM messages :
• Activate_PDP_Context_Accept,
• Activate_Secundary_PDP_Context_Accept,
• Modify_PDP_Context_Request (sent by the network) and
• Modify_PDP_Context_Accept (in case the request to modify is sent by the MS).
� PFC_FEATURE_MODE is included in the MS_Network_Capability I.E. (which is sent in the Attach_Request and
� This feature allows a dual transfer mode capable MS to use a radio resource for CS traffic and simultaneously one or several radio resources for PS traffic.
� Single slot operation DTM MSs are not supported in Alcatel BSS because the implementation of these MSs is difficult compared to the throughput expected in PS services. Only multislot operation DTM MSs are supported.
� In Alcatel’s implementation, the Gs interface is required to support DTM to ensure CS paging co-ordination. It avoids the BSS to ensure the paging co-ordination.
� While in dual transfer mode, the BSS only allocates full rate PDCH to the MS.
� The dynamic Abis feature allows to simplify the radio resource allocations. It avoids defining new TBF re-allocation triggers.
� Some restrictions towards BSS in deploying DTM exist. They are presented below: � Half rate
� Support of half rate configurations (one single timeslot encompassing one half rate circuit channel + one half rate packet channel) was not considered in the first implementation of DTM.
� Inter-cell handovers� The number of inter-cell handovers should be minimized for DTM calls, as an inter-cell HO leads to the re-allocation of the packet session. Therefore, handover causes having a low priority should be inhibited for the time the MS is operating in DTM.
� Intra-cell handovers� The number of intra-cell handovers should be minimized for DTM calls, as an intra-cell HO leads to the re-allocation of the packet session.
� Hierarchical networks� As (E)GPRS are preferentially offered in macro cells, the BSS shall ensure that at least one PDCH can be used in micro cells to re-direct the MS towards the macro cells. It means that the BSS shall allow a PDCH used by a MS operating in DTM mode to be shared by other (E)GPRS MS.
� EGPRS MS is characterized by two multislot classes:� GPRS multislot class
� EGPRS multislot class
� Typically, EGPRS multislot class < GPRS multislot class E.g. the multislot class of the mobile can be 3 RXs + 2 TXs (class 6) in pure GPRS mode and 2 RXs + 1 TX (class 2) in pure EGPRS mode
� Type 1: class 1-12, class 19-29 recognized as class 10
� Type 2: class 13-18, allocation is limited to max. 5+5 timeslots
� MS type
• Type 1 are simplex MSs, i.e., without duplexer: they are not able to transmit and receive at the same time
• Type 2 are duplex MSs, i.e., with duplexer: they are able to transmit and receive at the same time
� Rx
• The maximum number of received time slots that the MS can use per TDMA frame. The receive TS shall be
allocated within window of size Rx, but they do not need to be contiguous. For SIMPLEX MS, no transmitted
TSs shall occur between receive TS within a TDMA frame. This does not take into account the measurement
window (Mx).
� Tx
• The maximum number of transmitted time slots that the MS can use per TDMA frame. The transmitted TS
shall be allocated within the window of size Tx, but they do not need to be contiguous. For SIMPLEX MS, no
received TS shall occur between transmit TS within a TDMA frame.
� SUM
• The maximum number of transmitted and received time slots (without Mx) per TDMA frame.
� The meaning of Ttb, Tra et Trb changes regarding MS types.
• For SIMPLEX MS (type 1):
- Ttb is the minimum time (in time slot) necessary between the Rx and Tx windows.
- Tra is the minimum time between the last Tx window and the first Rx window of the next TDMA in
order to be able to open a measurement window.
- Trb is the same as Tra without opening a measurement window.
• For DUPLEX MS (type 2):
- Ttb is the minimum time necessary between 2 Tx windows belonging to different frames.
- Tra is the minimum time necessary between 2 Rx windows belonging to different frames in order to be
able to open a measurement window.
- Trb is the same as Tra without opening a measurement window.
� (E)GPRS defines a network architecture dedicated to packet service domain, with radio access, which allows service subscriber to send and receive data in an end-to-end packet transfer mode
� (E)GPRS uses the BSS architecture, but defines a fixed network (GPRS backbone) which is different from the NSS, and which links the BSS to PDNs (packet data networks). The BSS is used for both circuit-switched and (E)GPRS services
� The BSS has 2 clients:
� the MSC, for circuit-switched services (A interface)
� the GPRS backbone network, for GPRS (Gb interface)
� one or more 64 kbit/s channels on one or more 2 Mbit/s links
� Gb interface: Layer 1 specified in GSM 08.14The protocol stack defined in the stage 2, GSM 03.60
� GPRS backbone is an IP network and is composed of routers:
� Serving GPRS Support Node (SGSN), at the same hierarchical level as the MSC, which is linked to several BSSs. It keeps track of the individual MS’s location and performs security functions and access control
� Gateway GPRS Support Node (GGSN), which is linked to one or several data networks, provides interworking with external packet-switched networks and is connected with SGSNs via an IP-based GPRS backbone network
� Four different coding schemes, CS-1 to CS-4, are defined for the GPRS Radio Blocks carrying RLC data, and are applied depending from the actual radio conditions
� The first step of the channel coding procedure is to add a BlockCheck Sequence (BCS) for error detection
� For CS-1 to CS-3, the second step consists of pre-coding USF (except for CS-1), adding four tail bits and a half rate convolutional coding for error correction that is punctured to give the desired coding rate
� For CS-4 there is no coding for error correction
� The most protected mode is CS-1 which is therefore always used for GPRS signaling (even for EGPRS)
� Nine different coding schemes are defined: MCS-1 to MCS-9
� First step of the EGPRS coding procedure, is to add a Block Check Sequence (BCS) to each RLC data block, for error detection
� Second step consists of adding six tail bits (TB) and a 1/3 rate convolutional coding for error correction that is punctured to give the desired coding rate
� The Pi (puncturing schemes) for each MCS correspond to differentpuncturing schemes achieving the same coding rate
� Puncturing is a technique of removing bits in predetermined locations of the data block after the block has been channel coded
� MCS-9, MCS-8, MCS-7, MCS-4, MCS-3: are possible P1, P2, and P3
� MCS-6, MCS-5, MCS-2, MCS-1: P1 and P2 are possible
The puncturing process consists of transmitting only some of the coded bits obtained after the rate 1/3
convolutional coding. Depending on the considered puncturing scheme, different coded bits are transmitted.
Therefore, when the receiver receives two versions of the same RLC block sent with two different puncturing
schemes, it obtains additional information leading to an increased decoding probability.
� In the ARQ method, when the receiver detects the presence of errors in a received RLC block, it requests and receives a re-transmission of the same RLC block from the transmitter
� The retransmission can be performed using:
� Type-I ARQ mechanism. This applies for both GPRS and EGPRS mode
� Type-II hybrid ARQ mechanism, also called Incremental Redundancy (IR). This applies only for DL EGPRS mode
� IR is optional for the BTS, but is mandatory for the EGPRS MS (3GPP requirement)
B9!!! ARQ type-II applies for UL and DL EGPRS mode !!!
� In the selective type-I ARQ mechanism, the receiver discards the erroneous blocks, and indicates in the acknowledgement messages the reference of these erroneous blocks for their retransmission. Then, the sending side has to retransmit the erroneous data RLC blocks
MS
Uplink RLC data block B1 / PDTCH (1)
MFS
Packet UplinkAck/Nack /PACCH (3)
Uplink RLC data block B2 / PDTCH (2)
Uplink RLC data block B2 / PDTCH (4)
Uplink RLC data block B3 / PDTCH (5)
The Block 2 has been
unsuccessfully received
MS retransmits the uplink
RLC data block B2
� With the type 1 ARQ mechanism, the decoding of a re-transmitted RLC block does not use the
previously transmitted versions (not correctly received) of this RLC block. The decoding of a RLC data
block is only based on the current transmission.
� The type 1 ARQ mechanism is always used for the GPRS
� GPRS CSs are designed independently from the others with its own basic payload unit size, so the family concept does not exists in GPRS
� Before its transmission over the radio interface, the LLC frame is segmented into payload units according to CS that will be used to transmit the radio block
� In case of erroneous reception, the RLC data block can be retransmitted only with the same CS (segmentation is not possible)
���� If the radio conditions have changed and the coding rate is not appropriate to them, the receiver will never be able to decode the retransmission of the RLC data block. This will lead to the release of the TBF and the establishment of a new one in order to transmit the LLC frame
� In order to avoid this problem, the choice of the CS on the network side has to be made carefully. This often results in an non-optimized use of the radio interface, leading to a reduction of network capacity compared with its theoretical capacity
� MCSs have been designed to offset the GPRS disadvantage
� MCS family concept is applied
� In EGPRS, in case of retransmission request (type-I ARQ) for a RLC data block, the same or a next lower MCS within the same family is used
� The retransmission can be performed with or w/o RLC data segmentation (e.g. from MCS-9 to MCS-6 w/o, and MCS-6 to MCS-3 with segmentation)
� When one RLC data block is retransmitted with a lower MCS, the coding rate is decreased by two, but the redundancy transmitted is increased
���� That increases the capability to decode the radio block !
� Retransmission operates in connection with the link adaptation
� E.g. if the LA mechanism orders the usage of MCS-5 and the first transmission of an erroneous RLC block was with MCS-6, the transmission will be performed with MCS-3. The blocks that are sent for the first time will be transmitted with the last-ordered MCS
� Type-II ARQ (IR) is an efficient combination of 2 techniques:
� Automatic Repeat reQuest : in case of error detection in a received RLC block, a re-transmission of the same RLC data block is requested
� Forward Error Correction : adds redundant information to the user information at the transmitter, the receiver uses the info to correct errors causes by radio disturbances
� In the IR mechanism:
� The information which is sent first results from an initial “puncturing scheme” (PS1) applied to the encoded RLC data block
� If an error is detected by the receiver:
� the received message is stored
� selective retransmission of the RLC data block is requested
� a second “puncturing scheme” (PS2) is applied to the same MCS, by the sender
� the receiver decodes (combines) the resulting message together with the previously received message(s)
� multiple retransmission can be requested until decoding succeeds
� The type 2 ARQ mechanism or incremental redundancy (IR) is an ETSI function, mandatory for the EGPRS MS
receiver (downlink path) and optional for the BTS receiver (uplink path). In B8 release, the IR feature is only
available on the downlink path. It is important to notice that the IR feature is always running in the EDGE MS
receiver (except in case of MS memory shortage). The DL incremental redundancy is not used for the signaling
blocks, the GPRS data blocks and the data blocks in RLC unacknowledged mode. It is only used for the EGPRS data
blocks in RLC acknowledged mode.
� In the type II ARQ mechanism (IR):
• the first emission of a RLC data block is done using a first puncturing scheme (PS1),
• in case of re-transmission of this RLC block, the transmitter uses the same MCS or a MCS of the same family
than the one used for the initial block. On the DL path, depending on the value of the parameter
EN_FULL_IR_DL, re-segmentation of the RLC block may be performed or not,
• at the output of the demodulator, the receiver combines the information of soft bits corresponding to the
first transmission of the block and its different re-transmissions, thus increasing the decoding probability of
the RLC block.
• Remark : according to the 04.60 (RLC/MAC layers) GSM recommendation, the soft-combining inside the MS
receiver is only performed between an :
- MCSx block and MCSx block (that is the same MCS is used for the re-transmission),
- MCS9 block and an MCS6 block (in that case the RLC data blocks carry the same number of payload
units),
- MCS7 block and an MCS5 block (in that case the RLC data blocks carry the same number of payload
units).
� If the "MS OUT OF MEMORY" field is set by the mobile in the EGPRS Packet DL Ack/Nack message, the type I ARQ
shall apply in the MS receiver (ARQ without IR). This occurs when the memory for IR operation runs out in the MS
(that is when the memory of the MS is full due to the storage of the different versions of a RLC block not
� (1) The BSS sends a DL data block using the puncturing scheme P1 and MCS-6. B1 is not successfully decoded by the MS. The MS stores the received block
� (2) The MS requests a selective retransmission of the erroneous block, in the next EGPRS Packet DL Ack/Nack
� (3) The MS retransmits the DL data block using a new puncturing scheme P2 and the same MCS-6.If the block header is correctly decoded, the MS decodes the data making soft combination with the previous transmission
� In the puncturing scheme selection for re-transmission, 2 cases have to be considered:
• if the selected MCS has not changed : if all the different punctured versions of the data block have
been sent, the procedure shall start over and PS1 shall be used, followed by PS2, then by PS3 (if
available for the considered MCS), so that the PS selection is cyclic,
• if the selected MCS has changed : the PS to be used is indicated by the table below.
� B9 release: the IR mechanism is implemented in uplink and downlink
� This mechanism is associated with link adaptation in order to provide superior radio efficiency on the air interface
� IR feature is always running in the EGPRS MS receivers, except when a memory shortage is reported by the MS � the stored packets are discarded and type-I ARQ is set !
� Parameter for IR activation:
� EN_FULL_IR_DL which enable or disable the RLC data segmentation for retransmissions
� EN_FULL_IR_DL = disable; e.g. if MCS-5 is ordered by LA, and the first transmission was with MCS-6 then, the retransmission is performed with MCS-3 (segmentation on the initial RLC data block, ARQ Type-I)
� EN_FULL_IR_DL=enable; even if MCS-5 is ordered, the retransmission is performed with MCS-6 (no segmentation, ARQ Type-II)
� Packet Data Channel (PDCH)� (E)GPRS radio access method = GSM TDMA (8 timeslots per carrier)
� One PDCH represents a physical channel (1 timeslot) dedicated to packet data traffic (GPRS/EDGE), over the radio interface
� PDCH group� The available PDCH’s are grouped into “PDCH groups”
� One PDCH group contains consecutive timeslots (without TS holes)belonging to the same TRX, having the same radio configuration
� possible to have hopping and non hopping PDCH groups in one cell
� maximum number of PDCH groups/cell is equal to 16 (equal to maximum number of TRX / cell)� 16 TRX/cell achieved with help of the B7 feature “cell split over 2 BTS’s”, EVOLIUM™ BTS
� The packet data ‘call’ is a Temporary Block Flow (TBF)� For a data packet transmission, a temporary physical connection (TBF) will be set up as an unidirectional link
� Each TBF is unidirectional: Uplink TBF and Downlink TBF for the same mobile are uncorrelated
� One TBF allocates radio resources on one or more PDCH and comprise a number of RLC/MAC blocks carrying one or more LLC PDUs
� TBF is only temporary and maintained for the duration of the data transfer
� Either the mobile or the network can initiate a TBF
� Temporary Flow Identity (TFI):
• Each TBF is assigned a TFI by the MFS.
� Important:
• Since B7, it is possible to establish 32 TBFs per PDCH group (See sub-session 2.2 for ‘PDCH group’
definition).
� TBF
• is a group of blocks dynamically allocated to one MS for one transfer of RLC blocks in one direction
inside one cell.
• A Temporary Block Flow is a temporary, unidirectional physical connection across the Um interface,
between one mobile and the BSS. The TBF is established when data units are to be transmitted across
the Um interface and is released as soon as the transmission is completed.
� Two different resource sharing mechanisms exists:� PDCH multiplexing
� Multislot usage
Allows optimum usage of the available radio resources
� PDCH Multiplexing� PDCH multiplexing refers to the sharing of one PDCH by more than two users
(TBFs)
� It occurs when there are more requests for PDCH resources than available PDCH’s
� A maximum number of UL/DL_TBF can share the same PDCH in UL and DL direction respectively
� MAX_UL_TBF_SPDCH=6; MAX_DL_TBF_SPDCH=10
� When a PDCH is shared between an UL GPRS TBF and a DL EGPRS TBF, then the DL EGPRS shall be limited to GMSK (i.e. MCS-4) � GPRS MS becomes candidate for radio resource reallocation
� Multislot usage� Refers to the case when 1 user can request at once more than 2 PDCH resources for the data transmission
� Up to 5 PDCH on different (but consecutive) timeslots on the same frequency could be allocated to one mobile at the same time (MS multislot capability)
� B8 & B9 release supports 4+2 slots for Type 1 MS and 5+5 for Type 2 MS
� The PDCH blocks will be consecutively transmitted over the PDCH only if there is no user multiplexing
� lets assume that the data for user 1 has a length of 3 blocks (length of TBF 1=3 blocks) and is transmitted over PDCH #2
� as soon as one block of user 1 was entirely transmitted, another user 2can use the same PDCH #2 to transmit the blocks of its own TBF of e.g. length = 4 blocks, followed by the user 3 transmission...
� the blocks of user 1, user 2 and user 3 will not be transmitted in consecutive order:
� as soon as one block of user 1 is transmitted, another block of user 2 can be transmitted, continued with a block of the user 3 over the same PDCH #2
� Multislot usage example:
� User 1 has (1+1) and users 2 & user 3 have (3+1) MS multislot capability
� By the means of the polling mechanism, periodically an UL PACCH block is allocated during DL transfer, e.g. to allow an MS to request the establishment of an UL TBF by including a Channel Request description in a Packet DL Ack/Nack message
� the MS has no USF because it is involved in a DL TBF
� use of the RRBP (Relative Reserved Block Period) field transmitted in downlink
� RRBP values indicates the number of TDMA frames the MS shall wait before transmitting its uplink RLC/MAC block
� a special USF value is used: USF = no emission
� RRBP: Relative Radio Block Period
� Allocation of a PACCH block for the sending of acknowledgements in the UL of blocks received in the
DL:
• The MS has no USF because it is involved in a DL TBF
• Use of the RRBP field transmitted in the downlink (MAC header) in association with the TFI of the DL TBF in the RLC header.
• At the exact occurrence of the RRBP, a special USF value is used for the UL TBF taking place on the
same PDCH: USF=no emission.
� It is a semi-boolean parameter. The RRBP field of a RLC/LAC block is checked each time by the MS
whose TFI is written in the RLC header.
• When S/P is false, no UL PACCH is scheduled.
• When the RRBP field is valid, the value gives the number of blocks to wait before sending its PACCH
block in the UL
� S/P is false means MS has to send an acknowledgement message to the MFS.
� Aim� in order to send and receive GPRS data, the MS must activate the PDP (Packet Data Protocol) address, which it wants to use
� Results
� the MS is known in the corresponding GGSN (the GGSN knows the SGSN where the MS is located) and data transmission with external data network can begin
� As total paging is more frequent with GPRS service together withGSM paging, Routing Area (RA) was defined which may be smaller than Location Area (LA)
� One RA is a subset of one and only one LA
� RAI (RA Identity) identifies several cells
� The MS location in Standby state is known in the SGSN at the RA level
� The MS is paged in its RA when MT traffic (MS in Standby State) arrives at the SGSN
� Several modes of TBF establishment in UL and DL exists:
� In PIM mode� UL TBF on the CCCH or PCCCH (with primary MPDCH activation)
� DL TBF on the CCCH or PCCCH (with primary MPDCH activation)
� In PTM mode� UL TBF establishment during a DL TBF on the uplink PACCH
� DL TBF establishment during a UL TBF on the downlink PACCH
� The TBF establishment is performed through two types of access requests:
� One phase access request
� Two phase access request
� B8/B9: The BSS preferentially establishes an EGPRS TBF to an EGPRS MS provided that an EGPRS Packet Channel request message has been received and that there are EGRPS resources (i.e. radio resources supported by an EGPRS capable TRX) available in the cell, otherwise a GPRS TBF will be established
� Access Burst Type : 8 bit or 11 bit access burst
� EGPRS_PACKET_CHANNEL_REQUEST: EGPRS capable MS shall use EGPRS PACKET CHANNEL REQUEST message for uplink TBF establishment on the (P)RACH (En_EGPRS = True)
� BEP_PERIOD: Bit error probability (BEP) filter averaging periodEGPRS c
ell
� The MS has to get SI13 information on a regular basis:
• each time the SI13 content is updated (PSI field = SI13_CHANGE_MARK set to 1).
• every 30 seconds max (even if the TBF has to be interrupted).
• Through 2 different ways: SI13 on the BCCH or PSI13 in a PACCH block.
• The MS has always the time to switch on PSI13 in NMOIII and/or NMOI with a Master PDCH
because PBCCH blocks are always after a I or X TS within the 52 multi-frame.
� Access Burst Type: it defines the access burst (8 bits or 11 bits) to be used on the PRACH, PTCCH and
the “Packet Control Ack” on a PACCH.
� When the Master Channel is present in the cell, the System Information Type 13 message has different
contents from those described above. It mainly consists of:
• The radio description of the Primary Master Channel (in terms of time slot number, training
sequence code and frequency parameters).
• One GPRS Mobile Allocation (MA), if frequency hopping is used for GPRS. This is the GPRS MA of
the Primary Master Channel, if hopping. If the Primary Master Channel is not hopping, the MA
corresponds to the hopping TRX(s) used for GPRS, if any.
� Three modes of cell reselection have been defined by the 3GPP Standard for GPRS MSs. These
Network Control (NC) modes, known as the NC0, NC1 and NC2, are shortly described below:
• NC0: the GPRS MS performs autonomous cell reselection without sending measurement reports to
the network.
• NC1: the GPRS MS performs autonomous cell reselection. Additionally it sends measurement
reports to the network.
• NC2: the GPRS MS shall not perform autonomous cell reselection. It sends measurement reports
to the network. The network controls the cell reselection.
� To support high data throughputs, Alcatel has developed a solution, which aims at providing the best trade-off between offered radio throughput and impact on the telecom resource consumption
� This solution is based on the concept of multiple classes of TRX, which support more or less data throughput. The higher the packet class, the higher the maximum data throughput, the higher the impact on BSS Telecom resources
� Five TRX classes (1 to 5) have been defined
� The Operator defines per cell the number of TRXs of each class
� The B9 added improvements allow reducing the time dedicated to aCell Reselection in packet mode.
� These sub-features impact traffic model, allowing faster CR to a new cell or less number of CRs performed in a cell, will result in a higher aggregated throughput in the cell.
� Traffic model changes: the feature will modify the number of UL TBF activation+release on PACCH for all TCP/IP based applications and WAP.
� The feature will also modify the average duration of an uplink TBF, and as a consequence increase the number of MS multiplexed in uplink.
� If necessary to reserve a certain bandwidth in uplink for QoS, then the maximum number of MS in UL on the concerned PDCH should be limited. (the current default value is of 5 MS multiplexed in uplink)
� Some parameters are to be handled in order to set up and configure this feature:
� EN_EXTENDED_UL_TBF: Enable the extended TBF mode feature on the uplink.
� T_MAX EXTENDED_UL: Maximum duration of the extended uplink TBF phase. Recommended rule: value between 1s and 2s.
� EN_FAST_USF_UL_EXTENDED: Enable the transmission of USF every 20ms in extended mode, when the extended UL TBF feature is activated.
� EN_RA_CAP_UPDATE: Enable the Radio Acces Capability update on Gb. Recommended rule: should be enabled if EN_EXTENDED_UL_TBF is enabled and RA cap. update is supported by SGSN.
� It is recommended not to activate simultaneously extended UL TBFfeature (flag EN_EXTENDED_UL_TBF) and the DL PDU rerouting feature (flag EN_AUTONOMOUS_REROUTING).
Fast USF UL extended : to keep the link alive in order to be ready as soon as needed. If n MSs in extended,
then USF for 1 MS sent every n x 20ms.
RA CAP Update : the MFS can request the RA capabilities of the MS to the SGSN (based on IMSI)
2.3 Enhanced support of E-GPRS (EDGE) in uplink [cont.]
� In B9 release, “Incremental Redundancy” may be activated for both the DL and UL paths. Thanks to Incremental Redundancy, the link adaptation procedure can be more aggressive: if the chosen MCS is a bit too optimistic, IR increases the probability of data recovery and increases data rates considerably specially in poorer radio conditions for higher MCS’s.
� The link adaptation mechanism in UL is based on measurements (MEAN_BEP, CV_BEP) done by the BTS on the radios blocks receivedfrom the mobile. To take into account MCS-5 to MCS-9, the BSS algorithm for link adaptation needs new link adaptation MEAN_BEP/CV_BEP tables. These tables are the same as the one already used for DL.
� 8-PSK in the UL should be considered in the planning tools for the throughput and coverage estimation (based on interference calculation). It impacts cell range estimates if the link-budget is UL limited.
� The IR gain should also be considered in the throughput estimation :
� “Counter Improvements for Release B9” feature covers four candidate “sub-features” for B9:
1. Support of distributions: It introduces a new concept of counters called distributions to obtain improved statistics on (E)GPRS resource usage.
2. Consolidation of cell indicators at GPU level: It allows an operator to consolidate each indicator defined at cell level per GPU. This operation is very useful to follow possible lacks of GCH or GPU resources in a given GPU.
3. Counters defined at TRX level: It introduces a few counters defined at TRX level to follow the radio and transmission resource usage.
4. New MFS counters: It consists in defining a few new counters to ease the dimensioning and optimisation of (E)GPRS networks.
� The sub-feature consists in allowing the operator to consolidate cell counters P105c/d/e/f/g/h at GPU level.Also, without this consolidation, it is up to the MFS to perform the consolidation, which is in contradiction with the usual principles. Indeed, it is not the role of the MFS to perform computation on counters.
CellNumber of UL TBF establishment failures due to a lack of transmission resources.P105h
CellNumber of DL TBF establishment failures due to a lack of transmission resources.P105g
CellNumber of UL TBF establishment failures due to CPU processing power limitations of the
GPU.
P105f
CellNumber of DL TBF establishment failures due to CPU processing power limitations of the
GPU.
P105e
CellNumber of UL TBF establishment failures due to GPU congestion.P105d
CellNumber of DL TBF establishment failures due to GPU congestion.P105c
� The new counters and distributions should allow us to improve the existing (E)GPRS traffic model (i.e. better accuracy of the model can be achieved) but no impact on radio and other telecom performances is expected.
� The purpose of this feature is to give to the MFS all the radio timeslots that are usable for PS traffic, according to the whole BSS load (CS and PS loads). The MFS needs no more to request radio timeslots to the BSC; instead the MFS is always aware of all theavailable radio timeslots.
� -100 ms gain in the DL or UL TBF establishment duration
� As the maximum number of radio resources is allocated to the MFS, the TBF establishment duration (DL or UL) is reduced compared to the B8 solution (if the MFS requests for additional radio resources to establish the TBF).
� This could lead to an increase in the average TBF throughputs at cell level.
� The B9 “Improved 3G cell reselection” feature allows the operator to declare per 2G cell basis the 3G neighbor cells (the FDD UMTS frequencies and the scrambling codes). Maximum 3 FDD UMTS frequencies may bedeclared per cell basis. When knowing in advance the frequency and the scrambling code of a 3G cell, an MS should require 10 to 20ms tosynchronize on that cell.
� Regarding the current load, the BSS may reject an external hand-over coming from the UTRAN, provided the hand-over has not been triggered by an emergency cause, i.e. provided the hand-over request does not carry a cause type uplink/downlink quality/strength.
� Radio Network Planning Impact
� These sub-features impact traffic model, allowing faster 2G-to-3G cell reselection to a new cell or denying incoming handovers in a lodedcondition. It will result in a higher aggregated throughput in the cell or in less call drops experienced by a source 3G cell.
� These sub-features impact traffic model, allowing faster 2G-to-3G cell reselection to a new cell or denying incoming handovers in a loaded condition. It will result in a higher aggregated throughput in the cell or in less call drops experienced by a source 3G cell.
� This feature provides a solution to share the Ater and Abis nibbles between the radio timeslots of a TRX so that the transmission resources left available by a PDCH can be re-used by other PDCHsas long as those PDCHs belong to the same TRX. Thus allows reducing the waste of transmission bandwidth on the Ater and Abisinterfaces.
� Terminology
• M-EGCH
- The term M-EGCH (Multiplexed-EGCH) is used to refer to a link established between the MFS and
the BTS. An M-EGCH is defined per TRX (instead of an EGCH per radio timeslot in release B8).
• GCH
- A GCH is the 16kb/s channel between the MFS and the BTS. It is composed of an Ater nibble and an
Abis nibble cross-connected together in the BSC. The MFS or the BTS periodically send blocks on a
GCH every 20 ms.
• GCH frame
- In 20 ms period (also called block period), a number of 320 bits of this GCH can be used: this is the
frame.
• Segment
- A segment is formed by a part of an RLC block (after its segmentation on the M-EGCH link) and a
GCH header (different for first segment and subsequent segments). RLC data might be padded or a
segment can be a “no-data segment”.
- Note that in B9 a frame can be constituted of several segments belonging to different RLC blocks as
now all the RLC blocks sent on several PDCHs of a TRX are multiplexed on the same M-EGCH link.
Padding bits are added to the RLC blocks’ segments to fill the frame to 320 bits.
� The M-EGCH Statistical Multiplexing solution allows to share a given number of GCHs at a TRX level, i.e between the radio timeslots of one TRX, so that:
� the transmission resource left available by one TBF mapped on a set of RTS and being idle (eg, in establishment or delayed release phase) is automatically reused by another TBF mapped on the same RTSs or on another set of RTSs (as long as those sets of RTS are on the same TRX ).
� an increase of MCS, i.e. of throughput experienced by one TBF, does not lead to an increase of transmission links need since this increase can be compensated by a decrease of MCS experienced by another TBF.
� The GCH left while the control blocks are transferred can also be re-used by other TBFs (which is not the case in B8); indeed control blocks are encoded with CS1 and do not use an entire 320-bit frame.
The Statistical Multiplexing introduces a new segmentation of the radio blocks on the M-EGCH link: the blocks
of all the PDCHs of the TRX are sent one after the other without padding between them. As in B8 a block for a
PDCH can be spread over several 320-bit frames but after its last segment the block of another PDCH can be
started (if the remaining transmission capacity is sufficient). So a fixed 320-bit frame can have up to 2 or 3
segments of variable size. As in B8, the unused part of a 320bit frame (once all the PDCHs have been
scheduled) is filled with padding and the unused GCHs with a NODATA PDU.
The EGCH layer is highly impacted to support the statistical multiplexing and is renamed “M-EGCH layer” in
B9. This feature only applies to G3 and G4 TRX while the G2 DRFU TRX uses a B7.2 like GCH stack (1 GCH
� This feature enables, on the Abis, to dynamically allocate nibbles among the different TREs used for PS traffic in a given BTS. Compared to B8, it allows a higher average Abis bandwidth per PDCH, the BSC capacity in terms of TRXs is increased, and in some BTS configurations it may avoid to deploy a second Abis link. The extra Abis nibbles are shared at BTS level.
� Radio Network Planning Impact
� Increase of BSC capacity in terms of # of TRXs handled allows higher PS throughputs and could lead to lower PS blocking/drop probabilities.
� Deals with the determination of the number and of the nature of the 16k GCH channels inside each M-EGCH. It is implemented as a transmission resource manager. The transmission resource manageris located at MFS/GPU level. It handles both Abis and Ater resources at GCH level.
� It is in charge of:
� Creating and removing the M-EGCH links
� Selecting, adding, removing, and redistributing GCHs over the M-EGCH links
� Managing transmission resource preemptions
� Managing Abis and/or Ater congestion states,
� Optionally, monitoring M-EGCH links usage, according to the (M)CS of their supported TBFs (UL and DL).
� Abis nibbles sharing rules:
• To ensure that, anytime, each cell of a given BTS would be able to support PS traffic, we should
guarantee a minimal number of Abis nibbles to every cell in the BTS. Consequently, it has been
decided that basic Abis nibbles are only shared at cell level (i.e. among TRXs of the same cell or
sector). This restriction prevents some cells from using the whole Abis nibbles of the BTS as a
given cell cannot use the basic Abis nibbles of another cell. However, Extra (and Bonus) Abis
nibbles are shared at BTS level.
� Ater nibbles sharing rules:
• A given amount of Ater transmission resource is allocated per GPU. Afterwards, this Ater
transmission resource is shared among the four DSPs of the GPU thanks to the GPU on-board Ater
switch.
• Only 64K Ater TS are handled at GPU-level between DSPs. Thus, a 64K Ater TS may be moved
from one DSP to another if, and only if, all its four 16K Ater nibbles are free. This is the unique
restriction to Ater nibbles sharing at GPU-level.
• Furthermore, to prevent the above restriction from disturbing the First GPRS traffic in a cell, an
Ater reserve shall always be available. The Ater reserve consists on one or several free 64K Ater
TSs and is defined per GPU. Every 64K TS of the Ater reserve may be connected to any DSP of the
GPU to fulfil GCH requests:
• to establish the initial GCH in a cell with the Fast Initial GPRS Access feature activated, or;
• to ensure the First GPRS traffic in a cell with no active initial GCH.
• Each time a 64K TS is taken from the Ater reserve, a process is launched to retrieved another 64K
TS to replace it in the Ater reserve. This is done by means of GCH pre-emption on the Best effort
� The goal of the feature is to monitor the usage of each allowed AMR codecs (FR or HR), and to provide statistics information on timing advance.
� This feature allows monitoring the proper operation of AMR and the quality of the radio coverage in a cell. It also gives the possibility to tune the AMR parameters. Indeed, statistics about frame erasure rate in uplink and comparison between codec distribution and RXLEV allow assessing the voice quality, and adapting AMR thresholds to the situation of a given cell.
� RMS_I1 Indicators:
Mnemonic Definition Formula
RMS_AMR_FR_UL_BAD Number of bad speech frames using any AMR FR codec in uplink
RMS44a
RMS_AMR_HR_UL_BAD Number of bad speech frames using any AMR HR codec in uplink
RMS45a
RMS_AMR_FR_UL_RXLEV_UL Number of speech frames using one AMR FR codec in uplink per Rxlev on the uplink path
RMS46a
RMS_AMR_HR_UL_RXLEV_UL Number of speech frames using one AMR HR codec in uplink per Rxlev on the uplink path
RMS48a
RMS_AMR_FR_DL_RXLEV_DL Number of speech frames using one AMR FR codec in downlink per Rxlev on the downlink path
RMS47a
RMS_AMR_HR_DL_RXLEV_DL Number of speech frames using one AMR HR codec in downlink per Rxlev on the downlink path
� Knowing which codecs are the most used, and comparing them with link level in the cell, the operator could assess the voice quality and possibly adapt the AMR parameters (definition of the subset, thresholds and hysteresis).
� These parameters are different for AMR FR and AMR HR, information shall be provided separately for AMR FR and AMR HR.
� The codecs used in UL and in DL can be different; therefore interpretation of results would be easier if results are provided separately for uplink and downlink.
� The aim of this feature is to provide statistics information on timing advance, in order to understand geographical traffic distribution in a cell, to identify resurgences and hot spots.
� The improvement “RMS_I2: Timing advance” is a good indicator about the mobile position relative to a cell.
� Its usage in RMS B7.2 is very limited: only measurement reports done over a TA threshold are available, along with the max measured TA. This information is not detailed enough to understand geographical distribution in a cell, in order to identify resurgences and hot spot.
� RM_I2 Indicators:
Mnemonic Definition Formula
RMS_TPR_TIMING_ADVANCE The distribution of number of measurement reports for which the value of timing advance is in TA band
RMS50a
RMS_TPR_UL_RXLEV_TA_BAND
The average value of RXLEV per TA band in uplink.
RMS51
RMS_TPR_DL_RXLEV_TA_BAND
The average value of RXLEV per TA band in downlink.
RMS52
RMS_TPR_UL_RXQUAL_TA_BAND
The average value of RXQUAL per TA band in uplink.
RMS53
RMS_TPR_DL_RXQUAL_TA_BAND
The average value of RXQUAL per TA band in downlink.
� This RMS improvement described here would provide help to the operator for optimization of his network planning, through identification of these resurgences and hot spots. Detecting hotspots can be very useful in order to re-design that part of the network in a most adapted way to the experienced traffic load.
� In GSM, when an MS in idle mode moves from cell A to cell B, it performs a cell reselection applying the C1 or C2 criteria. In dedicated mode, MS performs a handover
� For (E)GPRS, the MS does in GMM READY state (PTM) cell reselection
� In the old cell an abnormal TBF release takes place
� In the new cell the MS establishes a new resource. (Different to handover in GSM, where the new channel is reserved by the network in advance) RA B
� Three modes of cell reselection have been defined for a MS in GPRS packet transfer mode:
� NC0 mode: (E)GPRS MS performs autonomous cell reselection without sending measurement reports to the network
� NC1 mode: (E)GPRS MS performs autonomous cell reselection. Additionally it sends measurement reports to the network
� NC2 mode: (E)GPRS MS shall not perform autonomous cell reselection. It sends measurement reports to the network. The network controls the cell reselection
� B9 release supports NC0 and NC2 modes
� NETWORK_CONTROL_ORDER parameter defines whether the MS or the BSS controls the cell reselections� NC0 mode: NETWORK_CONTROL_ORDER = 0
3.2 Cell reselection: NC0 mode, no PBCCH established [cont.]
� C2 criterion:
� PENALTY_TIME <> 11111
C2 = C1 + CELL_RESELECT_OFFSET - TEMPORARY OFFSET * * H(PENALTY_TIME - T) � non-serving cells: H(x) = 0 for x < 0; H(x)= 1 for x ≥ 0
� serving cells: H(x) = 0
� T is a timer implemented for each cell in the list of strongest carriers. T shall be started from zero at the time the cell is placed by the MS on the list of strongest carriers
� CELL_RESELECT_OFFSET may be used to give different priorities to different bands when multiband operation is used
� TEMPORARY_OFFSET applies a negative offset to C2 for the duration of PENALTY_TIME after the timer T has started for that cell.
� The following criteria are applied for cell reselection:
� C1: when C1< 0
� C31, C32: when a non-serving cell is evaluated to be better than the serving cell
� C1: the pathloss criterion � Is used as a minimum signal level criterion for cell reselection for GPRS in the same way as for GSM Idle mode criterion
� Same as defined, but with specific GPRS parameters:
� C1 = A - Max(B,0)
A = RLA_P - GPRS_RXLEV_ACCESS_MIN
B = GPRS_MS_TXPWR_MAX_CCH – P
� GPRS specific parameters, are broadcast on PBCCH of the serving cell
3.3 Cell reselection: NC0 mode, PBCCH established [cont.]
� C31: the signal level threshold criterion parameter for hierarchical cell structures (HCS)
� Is used to determine whether prioritized hierarchical GPRS and LSA cell re-selection shall apply
� For cells that fulfill C31criteria (C31>0):
� The best cell is the cell with the highest C32 value, among those cells that have the highest priority class, among those cells that have highest LSA priority
� If no cell fulfils the C31 criterion:
� The best cell is the cell with the highest C32 value, among all the neighbor cells
� C32: cell ranking criterion parameter is used to select cells amongthose with the same priority class
� The signal level threshold criterion parameter C31NC2 is used in hierarchical cellular networks to
determine whether the signal level received from a neighboring cell n is sufficient to redirect the MS
towards cell n based on a non-radio priority criterion. This criterion parameter is used only if there is a
PBCCH established in the serving cell. HCS_THR(n) defines a signal threshold for applying the
prioritized hierarchical GPRS cell reselection criterion. The cell n denotes either the serving cell or a
neighboring cell. Contrary to the C31 criterion implemented in the MS, the Alcatel BSS does not
manage the timer T implemented for each cell to monitor the time a neighboring cell is present in the
list of the strongest carriers. Therefore, the Alcatel BSS always assumes that
GPRS_TEMPORARY_OFFSET(n) = 0. As the GPRS_CELL_RESELECT_HYSTERESIS,
RA_RESELECT_HYSTERESIS, and C31_HYST are used to control the triggering conditions of a cell
reselection, they are not taken into account in the criterion C31NC2 and C32NC2 parameters.
� The cell ranking criterion parameter C32NC2 is used to order the candidate cells on an radio criterion.
This criterion applies only in serving cells where there is a PBCCH established.
GPRS_RESELECTION_OFFSET(n) applies a positive or negative offset which favors or disfavors the
neighboring cell n. The cell n denotes either the serving cell or a neighboring cell. If the parameter
C32_QUAL is set, the determination of C32NC2 is modified so that the neighboring cell n having the
highest AV_DL_RXLEV_NC2 among all the neighboring cells is applied a GPRS_RESELECTION_OFFSET
(only if the offset is positive) and no GPRS_RESELECTION_OFFSET is applied to the other neighboring
cells.
� The MFS shall take care of avoiding ping-pong effects between the old cell and the new cell (i.e.,
circular NC cell reselections). For that purpose, the MFS handles an anti-ping-pong timer and an anti
ping-pong offset, respectively called T_NC_PING_PONG and NC_PING_PONG_OFFSET. While the timer
T_NC_PING_PONG is running the neighboring cells are disfavored by the offset NC_PING_PONG_OFFSET
(expressed in dB) in the cell ranking process.
� The MFS starts the anti-ping-pong timer at the creation of the NC2 context for the MS.
� The MFS stops the anti ping-pong timer at the deletion of the NC2 context.
3.3 Cell reselection: NC0 mode, PBCCH established [cont.]
� C32 is an improvement of C2. It applies an individual offset and hysteresis value to each pair of cells, as well as the same temporary offsets as for C2.
� Additional hysteresis values apply for a cell re-selection that requires cell or routing area update
� With C32, neighbor cells can be favored through the GPRS_RESELECTION_OFFSET(n) broadcast on the PBCCH. This allows favoring neighbor cells e.g. based on their frequency band
� C32 also gives the possibility to temporarily penalize neighbor cells having the same priority as the serving cell (contrary to C31 that penalizes cells of different priorities). The penalty is computed based on the GPRS_TEMPORARY_OFFSET(n) and GPRS_PENALTY_TIME(n) parameters, like for C31
3.3 Cell reselection: NC0 mode, PBCCH established [cont.]
� If cell B is belonging to another Routing Area (RA) than cell A, the MS has to make RA update
� additional hysteresis are applied to avoid unnecessary RA updates:
� CELL_RESELECT_HYSTERESIS hysteresis for cell reselection applied on C1 criterion (no PBCCH), when the new cell is in a different LA or, for a GPRS MS, in a different RA, or when a GPRS MS is in GMM ready state
� RA_RESELECT_HYSTERESIS indicates in both STANDBY and READY state the additional hysteresis which applies on C31 and C32 (with PBCCH) when selecting a cell in a new RA
� C31_HYST: Determines whether an additional cell hysteresis shall be applied to the C31 criterion in same RA, in READY state
� In NC2 mode of operation, the BSS controls the cell reselections of all MS when in packet transfer mode (PTM) or of all MS when in GMM Ready state (depending on the selected NC2 deactivation mode)
� While the NC2 mode is activated for the concerned MS, the MS sends packet measurements reports (PMR) to the BSS
� Aim: NC2 mode is to limit the number of reselections to the strict necessary ones � increased data throughput
� NC2 activation� An MS transit to NC2 mode when it receives a PACKET MEASUREMENT ORDER message from the BSS, at the beginning of a data transfer. It provides mainly the NC_REPORTING_PERIOD_T which is the reportingperiod of NC measurements sent by the MS while in PTM (default = 0.96s)
� Measurement reporting and processing� MS periodically reports its NC2 measurements on PACCH through a PACKET MEASUREMENT REPORT
� The BSS handles the following measurements:
� UL serving cell: RXQUAL for GPRS TBF and mean BEP for EGPRS TBF
� DL serving cell: RXQUAL for GPRS TBF and mean BEP for EGPRS TBF
� DL serving and neighbor cells: RXLEV measurements of BCCH
� NETWORK_CONTROL_ORDER is a cell parameter tunable at the OMC-R.
� The R97 and R98 MSs are differentiated from the other MSs. Indeed, all the MSs shall support the NC2
mode, however since no network manufacturer has implemented the NC2 mode, the R97 and R98 MSs
may not have been sufficiently tested and therefore there is a risk of interoperability with these MSs.
� The “Packet Measurement Order” message is used to activate and de-activate the NC2 mode of
operation for a given MS.
• Activation
- The “Packet Measurement Order (NC2)” message is sent when:
› establishing the first Downlink TBF of the Packet Transfer Mode or when re-establishing
the DL TBF while T3192 is running and there is not any on-going UL TBF.
› no measurement report has already been received for that MS during its on-going packet
transfer(s) (UL and/or DL).
› the MS has not been forced to operate in NC2 mode by a Packet Cell Change Order
message (during an intra-RA cell reselection).
• De-activation
- The “Packet Measurement Order (RESET)” message is sent at the end of the data transfer, in
case of NC2_DEACTIVATION_MODE = “NC2 deactivation at the end of the packet transfer”.
- When the MS goes back to the STANDBY state, in case of NC2_DEACTIVATION_MODE = “NC2
� Cell reselection detection� NC2 reselection are triggered only for EMERGENCY or for POWER BUDGETcauses:
� Cause PT 1: Too low DL received signal level
� Cause PT 2: Detection of a better cell
� Cause PT 3: Too bad DL radio quality
� Cause PT 4: Too bad UL radio quality
� The criteria calculated by the BSS in NC2 mode are very near from those used by the MS in NC0 mode. This ensures that the target cell selected by the MS in NC0 mode or by the BSS in NC2 mode are identical in quite all cases
� C1NC2, C2NC2, C31NC2 and C32NC2 criteria are calculated by the BSS and the parameters defined for cell reselections in NC0 are re-used
� C31NC2: signal level threshold criterion parameter
� Used in hierarchical networks to determine whether the signal level received from a neighbor cell n is sufficient to redirect the MS towards cell n based on a non-radio priority criterion
� Used only if there is a PBCCH established in the serving cell
� C31NC2(n) = AV_RXLEV_NC2(n) – HCS_THR(n)
� HCS_THR(n) defines a signal threshold for applying the prioritized hierarchical GPRS cell reselection criterion
� The cell n denotes either the serving cell or a neighbor cell
� Used to order the candidate cells on a radio criterion
� Applies only in serving cells where there is a PBCCH established
� Cell n is the serving cell:
C32NC2(n) = C1NC2(n)
� Cell n is a neighbor cell:
C32NC2(n) = C1NC2(n) + GPRS_RESELECT_OFFSET(n)
� GPRS_RESELECT_OFFSET(n) applies an positive or negative offset which favors or disfavors the neighbor cell n. Cell n denotes either the serving cell or a neighbor cell
� Candidate cell evaluation� Cell Filtering: this process removes from the list candidates the cells to which a previous NC2 cell reselection failed
� Cell Ranking:
� No PBCCH� The cell are ranked to their C2NC2 value. The best cell candidate is the cell having the highest C2NC2 value
� PBCCH established� The cell are ranked based on the C31NC2 and C32NC2 criteria. Among the cells, the best cell is the cell with the highest C32NC2 value among:
o For cells that fulfill C31NC2criterion (C31NC2>0):
� B8/B9 release GPRS redirection is actually a NC cell reselection that is triggered at the beginning of the PTM in the serving cell even if the radio link is good
� Redirect the MS towards a target cell more appropriate to carry PS traffic
� The operator may wish to favor GPRS traffic in a particular layer/band:
� MULTILAYER NETWORK, it may be more efficient to define GPRS resources in the UPPER LAYER only
� Reduce the number of cell reselections
� Microcells have smaller traffic capacity and is assigned to CS
� MULTIBAND NETWORK, it may be more efficient to favor GPRS traffic in the 900 MHz band, due to its better indoor penetration
� MS GPRS mainly used in indoor environment
� Gain in stability of the GPRS session
� Operator must tune the NC parameters so that a NC cell reselection is systematically triggered at the beginning of a data transfer on receipt of the first Packet Measurement Report
� E.g. NC cell reselection Cause PT1 can be always activated by setting NC_DL_RXLEV_THR = - 47 dBm (Always)
� GPRS power control is only implemented in uplink in open loopconfiguration
� GSM recommendation 05.08
� During open loop power control, the MS adapts its output power in UL per block (i.e. 4 timeslots), based on the measured average signal strength in DL
� Open loop:
� There is no indication by the BTS whether the output power was sufficiently low or high: the same path loss in UL and DL is assumed by the MS
� When accessing the network on the (P)RACH the MS uses the outputpower defined by (GPRS_)MS_TXPWR_MAX_CCH, which is broadcasted on the (P)BCCH
� MS (E)GPRS performs the necessary LEVEL measurements for power control algorithm, either on the BCCH of the serving cell or on the PDCH (carrying the PACCH):
� The choice is made according to PC_MEAS_CHAN parameter, which is broadcasted on the BCCH:
� PC_MEAS_CHAN = 1, measurements on PDCH (default)
� 24 measurements in 480 ms
� PC_MEAS_CHAN = 0, measurements on BCCH
� 12 measurements in 480 ms
� The LEVEL measurements are averaged with recursive filtering algorithms
� The average levels are calculated by the MS in PIM and PTM modes, thus proper average level available at transfer start
� ΓCH : is sent to the MS. This parameter is used for grading the power control to a target received level at the BTS side
� Min:0; Max: 62; Default: 30 dB in GSM 900, 24 dB in GSM 1800
� α : is send to the MS. This parameter can be described as a reactivity factor. The 05.08 GSM recommendation suggest to use α = 1 in order to have an open loop power control
� C : is the DL level average calculated by the MS
� It relies on RXQUAL except between CS3 and CS4 adaptation, wherethe new metric I_LEVEL_TNi (interference level) is also considered� If CS-4 is used, the MS is allowed to report RXQUAL = 7
� AV_RXQUAL_ST (Short Term average), AV_RXQUAL_LT (Long Term average) and AV_SIR (Signal to Interference Ratio) are respectively averaged values at MFS side, of the RXQUAL and I_LEVEL_TNimeasurements received from the MS in Packet DL Ack/Nackmessages
� AV_RXQUAL_ST
� Triggering condition AV_RXQual_ST aim to decrease the coding scheme number as fast as possible when the radio conditions degraded significantly. Reaction would be much slower if it was only based on a long-term average, which could results in a TBF release
3.10 Link adaptation: DL GPRS Radio Link Control [cont.]
O&M threshold
and hysteresis
new CS
current CS
- AV_RXQUAL_ST
- AV_RXQUAL_LT
- AV_SIR
MS MFS
(RXQUAL, I_Level_TNi)
Packet DL Ack/Nack
(RXQUAL, I_Level_TNi)
Packet DL Ack/Nack
Averaging
Link
adaptation
� Interference measurements performed during idle frames of the 52 multiframe (twice during 240ms):
� I_LEVEL_TN 0 = I > C
� I_LEVEL_TN 1 = C - 2dB < I ≤ C
� I_LEVEL_TN 2 = C - 4dB < I ≤ C - 2dB
� I_LEVEL_TN 3 = C - 6dB < I ≤ C - 4dB
...
� I_LEVEL_TN 14 = C - 28dB < I ≤ C - 26dB
� I_LEVEL_TN 15 = I ≤ C - 28dB
� MFS uses the I_LEVEL_TNi received to calculate the AV_SIR value
� In case of DL GPRS TBF with PDCH allocated on BCCH TRX and no frequency hopping on the BCCH TRX, the MS does not report any interference levels � usage of BLER (Block Erasure Rate) instead of interference levels
� Drawback of putting GPRS on BCCH freq : no measure of interference levels
� Two new metrics are introduced in EGPRS, Mean_BEP (mean Bit error Probability) and CV_BEP (Coefficient of Variation of BEP), to offset the fact that RXQUAL, does not provide an accurate estimation of the bit error rate of the radio channel
� BEP measured on burst basis, is a reflection of the current C/I, time dispersion of the signal and the velocity of the terminal
� The variation of BEP value over several bursts also provides additional information regarding velocity and frequency hopping
� The mechanism is more efficient than in GPRS, since measurements are taken on
every burst and not only during the idle frames
3 (E)GPRS Radio Algorithms
3.12 Link adaptation in EGPRS: New metrics
∑=
=4
1iiburstblock BEP
4
1MEAN_BEP
∑
∑ ∑
=
= =
−
=4
1iiburst
24
1k
4
1iiburstkburst
block
BEP4
1
BEP4
1BEP
3
1
CV_BEP
�For more details about MEAN_BEP and CV_BEP averages performed by the MS, refer to 3GPP 05.08.
�Raw measurements on a radio block basis
• For EGPRS (that is during an EGPRS DL TBF), the MS shall calculate the following values, for each radio block (1
radio block = 4 bursts) addressed to it (the DL TBF TFI contained in the radio block must be decoded) :
• Mean Bit Error Probability (BEP) of a radio block:
• Coefficient of variation of the Bit Error Probability of a radio block:
• In the above equations, the BEP is measured on a burst basis by the MS before channel decoding.
� Averaging of the raw measurements on a TS basis
•The raw measurements made by the MS on a radio block basis are averaged by the MS per TS (TN in the below
equations) and per modulation type (GMSK (MCS1 to MCS4), 8-PSK (MCS5 to MCS9)) as follows:
•
•
• with (Rn gives the reliability of the averaged quality parameters)
� In the above equations :
• n is the iteration index, incremented for each DL radio block,
• e is a forgetting factor and is calculated according to the BEP_PERIOD cell parameter (new in B8, OMC-R
� To offer high throughput to EGPRS MSs :� EGPRS TBFs are preferentially allocated on high class TRXs
� Multiplexing, on the same PDCH, a DL EGPRS TBF with an UL GPRS TBF has to be avoided, since in this case, the DL EGPRS is limited to GMSK (i.e. MCS4) � new PDCH state: “EGPRS”
� To fairly share throughput between EGPRS TBFs:
� A higher number of EGPRS TBFs has to be piled up on high class TRXsthan on low class TRXs. This ratio has to take into account the maximum throughput which can be offered by each class of TRX� specific TRX selection for EGPRS TBFs
� To optimize GPRS throughput (i.e. high class TRX usage), as long as it does not conflict with EGPRS traffic
� A new reallocation trigger (T4) is created in order to reallocate an UL GPRS TBF which is multiplexed with a DL EGPRS TBF
� All the following PDCH states are related to establish TBFs:
� Allocated : The PDCH is a slave PDCH, which has been indicated as usable for PS traffic by the BSC
� Active : An allocated PDCH is active if it supports at least one radio resource allocated for a TBF or for a RT PFC
� Full : � For GPRS TBF:
The number of established TBFs (GPRS + EGPRS TBFs) is equal to MAX_UL/DL_TBF_SPDCH.
� For EGPRS TBF:
The number of established EGPRS TBFs is equal to MAX_UL/DL_TBF_SPDCH.
� EGPRS : SPDCH used in the DL direction by a 8-PSK capable EGPRS TBF. This state is meaningful only for non-EGPRS capable MSs and only in the UL direction.
!!! New states in B9 !!!
Full : for GPRS TBF : GPRS + EGPRS ts are counted, because some EGPRS TBF on GPRS PDCH are using GMSK
� Specific conditions are defined for TRX selection in case of allocation or reallocation for EGPRS capable MS
� To allocate EGPRS TBFs preferentially on TRX which allows a high throughput
� Principle:
� As long as the TRXs with the highest throughput do not support a maximum number of EGPRS TBFs, the other EGPRS capable TRXs are not taken into account by the algorithm
� N_TRX_EGPRS : number of TRXs on which EGPRS MSs are served in EGPRS mode
� MAX_TBF_PDCH_Current(TRXi) : maximum number of EGPRS TBFs per PDCH, currently allocated in TRXi
� N_TBF_PDCH_MCSi_MCSj
� It defines for each EGPRS TRX capability (MCSi) in the cell the number of EGPRS TBFs per PDCH beyond which it becomes more interesting to serve upcoming EGPRS MSs on TRXs with a lower EGPRS capability (MCSj).
� Max_PDCH_Throughput_MCSi / Max_PDCH_Throughput_MCSjwith Max_PDCH_Throughput_MCSx is the maximum theoretical throughput that can be achieved at RLC/MAC per PDCH using MCSx encoding
3.20 Radio Resource Allocation: EGPRS TBFs [cont.]
E) The candidate timeslot allocations which have the lowest numberof established EGPRS TBFs in the direction of the bias are preferred � It is preferred to multiplex an EGPRS TBF with a GPRS TBF, rather than with another EGPRS TBF
F) The candidate timeslot allocations which have the lowest numberof established EGPRS TBFs in the direction opposite to the bias are preferred
G) The candidate timeslot allocations which are on a TRX with highest priority are preferred
H) The candidate timeslot allocations which have the lowest numberof established GPRS TBFs in the direction of the bias are preferred � H has a lowest priority than G, in order to avoid to establish EGPRS TBFs on low class TRXs, because of GPRS TBFs
3.20 Radio Resource Allocation: EGPRS TBFs [cont.]
I) The candidate timeslot allocations which have all their PDCHsestablished are preferred. If all the preferred best candidate timeslot allocations require additional PDCHs, then a request is sent to the BSC and the algorithm is stopped
J) If the MS has already one or 2 TBFs established, preference is given to the candidate timeslot allocation which does not require a T2 reallocation of the on-going TBFs
K) The candidate timeslot allocation with the PDCHs that have the lowest index is preferred
�maintain a TBF alive despite a pre-emption on a PACCH of a TBF
� or if MEGCH becomes too low to provide MAX MCS of the TBF [B9]
� T2: re-allocation of an on-going TBF when establishing a concurrent TBF� in order to provide a better throughput
� T3: re-allocation to offer a better throughput to an on-going TBFs� In order to provide a higher throughput, if it is possible, to any TBF in the cell.
� T4: re-allocation condition to move
� UL GPRS TBF sharing one PDCH with a DL EGPRS TBF
� � PDCHs which do not carry a DL EGPRS TBF
B9 : Same types as in B8, but extended possibilities
T2 : It is the case in the following scenarios:
- establishment of a downlink TBF, concurrent to an existing uplink TBF, which is
allocated in such a way that the maximum number of timeslots supported in the
direction of the bias cannot be offered to the MS.
- similar situation in case of uplink TBF establishment concurrent to a downlink TBF;
� ETSI -> Simulation of coding scheme performance under different environment and fading conditions
� typical urban environment with mobile speed of 3 km/h (TU3)
� typical urban environment with mobile speed of 50 km/h (TU50)
� typical hilly terrain with mobile speed of 100 km/h (HT100)
� typical rural area with mobile speed of 250 km/h (RA250)
� The impact of Level and interference has been studied in order to find the minimum required Level and C/I ratio for the reference error performance, defined by a block error rate Block Error Rate (BLER) of 10%, the reference performance point
� Why is this important?
� Saturation effect
For data, most users are static (TU3)
Japanese/Korean behaviour : they use data while in subways and trains. Appearing in France due to tv
� Less retransmission has to be performed (less data blocks are erroneous)� since saturation occurs. e.g. for CS-1 starting with 7.2 kbit/s at a C/I ratio of 9dB
� With an increasing C/I ratio the data throughput increases only little up to its maximum value of 8kbit/s (saturation point)
�Data throughput increases
� Due to this saturation effect, a further increase of the C/I ratio does not have large impact on the data throughput of a single coding scheme: possibly a switch to a higher CS may occur (C/I ~ 7 dB for CS-1 to CS-2)
� Reference Performance Point : A tradeoff between the maximisation of the network throughput and excessive C/I constraints.
5.6 GPRS traffic calculation and market applications
� Market applications
� Different services are possible for packet data use e.g. new designed services or services known from the fixed network
� Market applications and user profiles are related to each other, thus some applications are assigned to one user profile only
� Each service is characterized by its occurrence: action time per month and the related bit rate per action.
� In some applications, the data exchange traffic is oriented to downlink, in some others to uplink. Generally the downlink traffic is preponderant in asymmetrical applications such as: web browsing, information downloading, audio downloading etc.
� This shall be taken into account for the dimensioning process: so the dimensioning will be downlink oriented.
Difference between prepaid and postpaid ?
Daily services : weather forecast, news
Hourly : road traffic, market shares
Uplink bias applications : MMS, ftp upload. Create problem for dimesionning ? No, because MMS are
uploaded and then downloaded. They create equal traffic in both ways.
Current Ms use 2ts in uplink, class 11 and 12 are coming (up to 4 TS in uplink, but still simplex.
� GOAL : to categorize the quality of the three calculation methods
� User mapping
� One certain resource can be shared simultaneously by different users. Behavior in GPRS -> Packet switched service for different users on one timeslot.
User 1
User 2
User 3
User
Timeslot 1
TS 2
TRX
TS 3 TS 4 TS 5 TS 6 TS 7 TS 8
In dimensionning, never take maximum usage as an average value!
User mapping should be quite low, in order to allow a high throughput � but requires higher capacity
� gives the smallest number of needed PS TS among the traffic calculation methods
� It calculates for the whole data volume, sum of all users data, the number of PDCH TS needed to transfer this data volume, regardless of data transfer peaks
� This method is not taking into account parallel data transfer, which is the benefit of packet transfer (GPRS).
� So no service attempt queuing and no service multiplexing is taken into account by this method.
� A calculation method to get in the first step of GPRS planning an idea of minimum needed PDCH TS.
� gives for a required service attempt probability (Quantile) and the queue delay time of it (e.g. 2 s delay can be set if no resource is available at service attempt), the number of needed resources (TS).
� The result of Erlang C will give the biggest number of needed PDCH TSamong the presented packet traffic calculations.
� The reason is that a constant data flow is considered which is not the case for different applications like WAP
� For all different services the PDCH TS with Erlang C has to be calculated and summarized. Afterwards the sum of PDCH TS for the different services leads to an over dimensioning.
� This method can be used to give very fast a planning result on how many PDCH as maximum can be expected.
� TRX calculation for CS and PS with application of reuse of CS TS for PDCH (PS) when dynamic/smooth PDCH adaptation and /or fast preemptionfeature is activated
� With the input from GPRS traffic calculation the GPRS Design process can start:
� Basis: The knowledge of the amount of timeslots makes it possible to go to the next step of GPRS network design process
� The user throughput demand is then related to a daily traffic occurrence (user capacity) and in combination with the CS traffic demand, the needed equipment amount is calculated:
� Number of timeslots which may be reserved for GPRS in normal and high load state of the BSC
� Number of timeslots which have to be reserved exclusively for GPRS
� The result of traffic analysis gives the standard BTS configuration for the different traffic areas. The traffic areas are most commonly linked to a specific morpho class
� Next steps:
1. GPRS Field strength prediction is done as for the GSM network planning [A9155]
2. � Inputs for mutual interference calculation [A9155]
3. � Inputs for a GSM/GPRS frequency planning [A9155/AFP module]
� Some differences compared to the well-known power budget is the handling of some losses and margins:
� body loss, for PS: 2 dB, due to the fact, that for the most PS applications the MS is not close to the body , but on an other, from the propagation point of view unfavorable position (e.g. on the table)
� interference margin: minimum 3 dB (urban and dense urban area up to 5 dB, depending of the frequency re-use), due to the high dependency of the PS service on C/I
� (lognormal) fading margin can be added to increase the coverage probability from 50% up to 95%; e.g. assuming standard deviation sigma = 7 dB =>fading margin:1.65 sigma ~11 dB
Diversity Gain: 3.00 dB 0.00 dB 3.00 dB 0.00 dB 3.00 dB 0.00 dBInterference Margin 3.00 dB 3.00 dB 3.00 dB 3.00 dB 3.00 dB 3.00 dBFading Margin 0.0 dB 0.0 dB 0.0 dB 0.0 dB 0.0 dB 0.0 dBIsotr. Rec. Power: -126.00
� Some general considerations apply independently from the BSS software release:
� GPRS/EDGE shall be mapped on the TRX(s) with the best radio quality (lowest interference probability); this can be any TRX in the cell.
� Identification of less interfered frequencies and their ranking
� Assigning the preference for PS traffic handling to the best ranked frequencies (e.g High Power TRX, Full rate capable TRX) with the help of the parameters:
TRX_PREF_MARK; PS_PREF_BCCH_TRX, TRX Classes,
� Since B7 up to 16 TRX per cell are available for GPRS service. So a differentiation of GSM and GPRS TS allocation priority on the TRX must be fixed during planning. The allocation priority for GPRS shall be set according to GPRS QoS needs.
How to map TRE with TRX ?
PS capable TRXs have to be preferentially mapped (from the best choice to the worst) on:
- FR, HP, EGPRS capable TREs
- DR, HP, EGPRS capable TREs
- FR, MP, EGPRS capable TREs
- DR, MP, EGPRS capable TREs
- FR, non-EGPRS capable TREs
- DR, non-EGPRS capable TREs
(When PS_Pref_BCCH_TRX = TRUE, the TRX supporting the BCCH is mapped on the “best” TRE)
� The system keeps always the highest coding scheme (and due to this, the highest achievable throughput), until the C/I proportions lead to change to a lower coding scheme
� By driving through the CS4 area from the centerto the border, a stepwise degradationof the throughput depending from theC/I ratio is visible
�the MS in idle mode will be paged in all cells belonging to the LA where the MS is assigned. The signalling effort for paging is thus focused to a certain area, the LA.
� GPRS: the SGSN pages the MS in STANDBY state, in case of a downlink TBF (comparable to a CS MT call).
� Paging GSM+ paging GPRS additional signalling effort will be produced in the network
� ETSI introduced Routing Areas (RA), which are smaller than LA.
�The signalling effort for paging is now more focused to a smaller area. Since not all cells of a LA are involved in the paging process, the signalling load for the cells is reduced
� the fast identification of the RA membership of the serving cell and neighbour cells (what cell belongs to what RA)
� As a consequence, the assignment of the cells belonging to RA has to be done carefully, to avoid additional signaling load on the cell (additional to the signaling for the CS traffic too)
� For the dimensioning of the number of RA in a LA, of the number of cells belonging to a RA and the number of RA_Codes per LA, the following steps are proposed, function of different network growths
� Step 1: Network with low GPRS/E-GPRS traffic
� RA as big as LA =>1 RA_Code (same for each cell) per LA, 1 RA_colour(same for each cell) per LA
� The first introduction step is based on the assumption, that in the beginning not much PS services is expected. The expense of this implementation is low
� For each RA in a LA one unique RA_Code is assigned
� A balanced number of cells per RA needs to be acquired, however for identified hot spots an unbalanced assignment is possible (smaller RA for hot spots)
� This step represents a reasonable split of the LA into RA if packet data traffic rises
� It can also be carried out right from the start to be prepared for the traffic growth
� According to GSM, QoS indicators for the Air interface available for GPRS
� Indicators based on counters, computed by the MFS, transferred to the OMC-R
� Note: To obtain the QoS for GPRS, it is not sufficient to study only the GPRS indicators. There is always an influence of GSM service on GPRS service, e.g. TCH congestion in GSM could be influenced by high CS traffic or the additional high packed data traffic.
7 Considerable features to reach (E)GPRS QoS target
7.2 MPDCH [cont.]
� Planning Recommendation on MPDCH
� Till the penetration rate of GPRS MS, which support master channel feature, is unclear the MPDCH should be not enabled
� So it is guaranteed that all GPRS mobiles in the network can access for GPRS service. MS, which do not support MPDCH, cannot access the GPRS service if MPDCH is enabled. Note: MPDCH can be enabled in network mode of operation: NMO I and NMO III.
7 Considerable features to reach (E)GPRS QoS target
7.2 MPDCH [cont.]
� Traffic dependent recommendation (with respect to condition for MPDCH):
� Low GPRS traffic
� If GPRS traffic is low no Primary Master Channel needs to be activated
� High GPRS traffic
� Static Primary Master channel� If the available TS are not scarce
� Operator wants the GPRS MS to perform autonomous cell re-selection based on C31 and C32 criterion
� Dynamic Primary Master Channel� If the CS signaling channels CCCH getting overloaded due to high GPRS traffic and signaling in addition to CS signaling
7 Considerable features to reach (E)GPRS QoS target
7.4 User multiplexing
� The strategy of the TBF resource sharing is to use the PDCH resources in a most effective way, that means not to ‘waste’ a PDCH just with one user and therefore to limit the available PS capacity. On the other hand, the more users (different TBFs) share a PDCH, the less effective the data flow and the longer the download or upload time is
� Trade-off between radio resource capacity sharing and optimum data throughput
� Since GSM speech service users are still to be preferred, it is recommended to set N_MAX_UL/DL_TBF_PDCH ≠ 1 (e.g.=2)
� E.g. if N_MAX_DL_TBF_PDCH and CS-2 is used, the DL bit rate per MS will be 6.0 kbit/s (=12/2) per used timeslot for this MS
� If operators goal is to maximize the PS throughput then N_MAX_UL/DL_TBF_PDCH = 1 (default value) is recommended
7 Considerable features to reach (E)GPRS QoS target
7.5 PDCH Resource Multiplexing
� Multislot access is the allocation of more than one PDCH to one MS (multislot access). However to prevent one multislot MS to use too many PDCHs each time it wants to transmit data (detriment of other users), following parameter is used:
� MAX_PDCH_PER_TBF : Maximum number of PDCHs, which can be allocated to a single TBF (or MS)
� Range: [1..5], default value: 5 (today’s MS capabilities)
� Radio Network Planning Impacts
� A few multi slot mobiles can occupy all resources with the default value of MAX_PDCH_PER_TBF. Thus the parameter has to be set, depending from the expected load and in combination with N_TBF_PER_S/MPDCH to reflect operator’s strategy on GPRS QoS.
7 Considerable features to reach (E)GPRS QoS target
7.6 Radio (TBF) Resource Reallocation
� With the feature TBF reallocation, the radio resources allocated to a TBF can be changed during TBF lifetime, which increases successful and efficient TS allocation (according to multislot capability) during ongoing data transfer for PS case.
� Radio Network Planning Impacts
� EN_RES_REALLOCATION is enabling / disabling the Radio Resource reallocation feature per trigger and per BSS
� All event triggers for TBF resource reallocation shall be considered:
� Trigger T1
� Trigger T2
� Trigger T3
� Trigger T4 (new in B8 for EGPRS purposes)
!!! B9 : this feature is always activated !!!Not changeable !
� Trigger T1 (target maintain a TBF alive when its PACCH is fast preempted):
• Reallocate all impacted TBFs using the pre-empted PDCHs instead of releasing them using the
Packet TBF Release procedure
� Trigger T2 (target attempt offering more PDCHs to an MS upon concurrent TBF establishment):
• get rid of the concurrence constraints imposed by the multislot class of the MS and an existing TBF
to offer the best throughput, the initial TBF can be “moved” to other PDCHs
� Trigger T3 (target periodically attempt offering more PDCHs to an MS which has a TBF in the
direction of the bias with less PDCHs than it can support according to its multislot class):
• take benefit of PDCH resource usage variations in a cell to reallocate the resources granted to a
Mobile Station, in case those resources were not using the full multislot class capabilities of the MS
• to offer the best throughput in the direction of the bias and even adapt to bias changes in the
course of a packet transfer
• Parameters for trigger T3 :
- T_CANDIDATE_TBF_REALLOC: Timer value controlling the time duration between successive
resource reallocation attempts for candidate MSs with the trigger T3
7 Considerable features to reach (E)GPRS QoS target
7.6 Radio (TBF) Resource Reallocation [cont.]
� Advantages
� The advantage of the feature TBF resource reallocation is to serve a better PDCH allocation to a TBF (throughput can be optimized), according to the available radio, transmission, DSP and CPU resources, during establishment and lifetime of TBF
� Drawback
� The allocation process is based on the number of PDCHs that the TBF can be mapped on a new resource and not on the throughput the TBF will get on these PDCHs
Consequence: in certain cases, available PDCHs will not be used for TBF reallocation, whilst using them would have improved the TBF throughput
7 Considerable features to reach (E)GPRS QoS target
7.8 GPRS Power Control
� Compatibility of GSM and GPRS UL Power control
� For GPRS rollouts it is recommended to disable the GPRS UL PC bysetting: α=0 and ΓTNX=0
� The reasons why GPRS UL PC shall be disabled:
� MS controlled open loop PC is not working reliably (MS software implementation)
� Field tests show a better throughput performance since the acknowledge message is sent in UL with full power
� Remark: It is possible to deactivate GPRS UL power control (GCH=0 and a=0) and to let GSM UL power control activated (EN_MS_PC=enabled, default), different power control parameters for GSM and GPRS
� Increase UL GPRS throughput
� If TMA (Tower Mounted Amplifier) is used and UL GPRS PC is disabled on a site than better throughput in UL is expected
7 Considerable features to reach (E)GPRS QoS target
7.8 Features on DL TBF establishment and release
� 3 different features are presented which preemptively delay the TBF release to speed up the setup of subsequent TBF
� Delayed DL TBF release
� Fast DL TBF re-establishment
� Non DRX Mode
� Their success depends on the users download behavior e.g. how often pages are changed and the content of the downloaded http looks like. For Web browsing and WAP applications where the PS traffic is bursty, the gain of the features to delay TBF release will be very high
� The 3 features are complementary and can be activated independently from each other. Delays to start download of new LLC PDU depending on feature
7 Considerable features to reach (E)GPRS QoS target
7.8.1 Delayed DL TBF release
� This feature should be enabled if there is no lack of resources to achieve higher user application throughput
� Main beneficiaries will be the applications consecutive pings, WAP and HTTP (clustered web page). The round trip time (RTT) can be shortened by the availability of an already opened TBF. This, in turn, is affected by the TBF hold time and the time between pings
� So in fact less signaling is needed for e.g. download of successive WAP pages or HTTP links because there is no need to establish a new TBF during T_DELAYED_DL_TBF_REL time
� T_DELAYED_DL_TBF_REL should be in between of 1.5s up to 2s depending on available resources in the cell. The higher the TS capacity in a cell is the higher the value of T_DELAYED_DL_TBF_REL can be tuned
DL TBF
T3192
Non-DRX mode
DRX_TIMER_MAX
DL TBF establishmentvia PCH or PPCH of MS paging group
DL TBF
T3192
Non-DRX mode
DRX_TIMER_MAX
DL TBF establishmentvia PCH or PPCH of MS paging group
T_NETWORK_RESPONSE_TIME
p.82
During the delayed release of the DL TBF, the BSS periodically sends to the MS a DL RLC data block (with a
polling request) containing a Dummy UI Command which is a LLC PDU whose checksum is deliberately
wrong. This LLC PDU is hence discarded by the LLC layer of the MS.
Sending periodically Dummy UI Commands enables the mobile station to request an UL TBF establishment in
a PACKET DL Ack/Nack message if it has data to send, and prevents defense RLC timers from expiring in the
mobile station.
If new DL LLC PDUs are received for that MS, the DL LLC PDUs can be sent immediately on the DL TBF. If
the BSS does not receive any DL LLC PDU during the inactivity period, it releases the DL TBF through the
= 800 ms + 700 ms (defaults) = 1500 ms if delayed DL TBF Release is enabled by parameter EN_DELAYED_DL_TBF_REL
� Advantage
� no delay to start DL data transfer for new DL LLC PDUs
� less signaling
� throughput improved for reason: long RTT. RTT can be shortened by the availability of an already opened TBF. This, in turn, is affect by the TBF hold time, and the time between pings.
� Drawback
� waste of resources, TBF is kept open during delayed downlink time, available USF values are limited
7 Considerable features to reach (E)GPRS QoS target
7.8.2 Fast Downlink TBF re-establishment process
� Fast Downlink TBF re-establishment process
� After reception of the final block by the MS and after the sending of the last PACKET DL ACK/NACK message, the MS still listens on the PACCH during T3192 sec
� BSS re-establishes a DL TBF on the PACCH of the previous DL TBF (i.e. to send a PACKET DL ASSIGNMENT message on the PACCH)
� fast DL TBF re-establishment without impacting the (P)CCCH resources; i.e. a new TBF is established but with the parameters of the old TBF (TFI, TAI)
� Following aspects are considered if GPRS is introduced into a mature GSM network without network design changes
� Different to the approach of GPRS Greenfield planning
� If the operator foresees design changes due to GPRS QoS requirements than traffic analysis and GPRS network design tasks has to be done before the GPRS introduction step
� Actual status of the GSM network
� GSM QoS and Interference
� All GSM network enhancement features and GSM network problems, mainly GSM QoS and interference, shall be fixed before GPRS is implemented
� New network design/frequency planning
� If a new network design and frequency planning is developed to improve GSM QoS and interference, then the implementation of this design should be done before GPRS is implemented
� TRX assignment to GPRS service and the PDCH planning can be done.
� If resources are not enough for GSM and GPRS
� Additional TRX & frequencies must be allocated to the sites with not enoughtraffic capacity.
� A new frequency planning should be done when a not negligible amount of new frequencies have to be added to a planning area to fulfill (GSM+GPRS) capacity requirements.
� Introduction of GPRS and related features/settings
� The prerequisites for a GPRS analysis are following tasks
� Field strength prediction
� Interference analysis
� If new sites after GPRS analysis are required to fulfill operators GPRS requirements, a new frequency planning with a certain frequency band range planning has to be done.
� Routing area, CAE data
� The routing area (RA) planning is a must for GPRS introduction into GSM network, see chapter 7 for details on RA planning and CAE data generation.
� GPRS QoS increasing tasks to be done are depending on dimensions of QoS requirements. What kind of tasks and references can be done to increase GPRS QoS ? As seen earlier :
� Dedicated TRX for GPRS in a cell can be done if TRX number in the cell is ≥ 2
� Introduction of GPRS Master channels (MPDCH to separate GPRS and GSM signaling
� Open question: Penetration rate of GPRS MS which can decode MPDCH
� The parameters for PDCH dynamic allocation (and TBF resource management) depends on GPRS QoS requirements :
�Weaker GPRS requirements � more TS for GSM can be reserved with a low value of MAX_PDCH
� The dependency between FH usage and Coding scheme distribution and the consequences on CS1-CS3 and CS 4
� Generally Frequency Hopping (FH) leads to Interference averaging. Thus calls having good quality will get worse, bad calls will get better. This is valid for GSM, similar it is valid for GPRS.
� CS1 is used in bad conditions, thus it will be improved if FH isintroduced.
� CS4 is used in very good conditions, which are more seldom in a hopping network. Thus CS4 will perform less good and will be used more seldom.
� The overall gain of CS1 - CS3 will depend on the C/I situation before and after FH.
� CS adaptation parameters can be tuned more optimistic in respect to throughput and Coding Scheme if FH is used:
� The main advantage of a µ-cell environment may be a better frequency re-use possibility, thus better C/I value and higher throughput can be expected (especially for E-GPRS with higher C/I requirements than GPRS s). Following two steps is proposed for GPRS implementation
� 1. Step: GPRS traffic is low => introduction of GPRS for macro and µ-cell together
� Disadvantages in both layers :
� Emergency capacity on macro cell layer reduced
� Higher blocking probability on µ-cell layer for CS traffic
� Solution:
� Reduction of the maximum GPRS capacity of the µ-cell to 30-50% by parameters setting
� Tuning of the GPRS user access handling (TBF and PDCH share)
− Hardware : TRX upgrade, µ-cell and macro cell densification, site design
− Parameter: GPRS capacity and user access handling tuning
� Basis: OMC-R Load measurements and GPRS customer behavior (location)
� Assumption: 80% of packet data traffic is static, 20% is dynamic (driving)
� The strategy is also valid for a different assumption, but this assumption is more probable.
!!! Activate Outgoing Redirection from MICRO !!!EN_OUTGOING_GPRS_REDIR(Umbrella) = DisableEN_OUTGOING_GPRS_REDIR(Micro) = EnableNC_DL_RXLEV_THR(Micro) = -47 dBmNC2_DEACTIVATION_MODE(Umbrella) = at the expiry of the GMM Ready timer
� Concentric cells which are disturbing other cells:
�the inner zone is smaller than the outer zone and keeps the disturbing carriers
� concentric cells which are disturbed by other cells:
�the inner zone and outer zone carriers have the same output powers; nevertheless, the size of the inner zone is dimensioned by proper parameter setting
� the same recommendation holds: the TRX for PS traffic must be configured in the outer zone of the concentric cell
� According the questionnaire, the GPRS user distribution will be calculated
� Due to the different network capacity in urban and rural area and the different ratio of business and private users in the area, the GPRS and speech subscriber are split to urban and rural area
Total GSM subscriber: 1 MioBusiness Private
GPRS share 7% total 5% (50000 subs.) 2% (20000 subs.)Urban area 70% (35000) 50% (10000)Rural area 30% (15000) 50% (10000)
Total CS subscriber: 1 Mio Urban area 80% (800000 subs.) Rural area 20% (200000 subs)
� In our example, the service hours for PS traffic are in total 6 hours, but from 14 to17 o’clock business and private subscriber will make data traffic at the same time
� Thus the busy hours for data traffic are these 3 hours
� It is also visible, that during that time, also for speech traffic a busy time occurs
� Busy hour: GPRS traffic dimensioning will be 14 to17
� Allocating TS to GPRS traffic reduces the capacity within the circuit switched design
� For the busy hour, the BSC is in high load situation, i.e the maximum of PDCHs is equal to MAX_PDCH_HIGH_LOAD (resource control)
� The following table gives the CS capacities based on a blockingprobability of 2% (in Erlang), according to the amount of allocated timeslots for GPRS in BSC high load situation
� One TS is sufficient for PS traffic during the busy hour.
� No CS service degradation during busy hour.
� The reservation of 1 TS for PS traffic represents no service degradation for CS traffic, since the remaining network capacity is still sufficient to handle the CS traffic.
� To guarantee a permanent PS service independent form the load situation, the parameter Min_PDCH_GROUP was set to 1 (I.e. 1 TS/ cell is permanently reserved for PS service and not available for CS traffic),however Min_PDCH_GROUP = 0 is recommended (load reduction on Atermuxinterface)
� Further iterations would be necessary (increase of MAX_PDCH_HIGH_LOAD) if the PS traffic demand could not be handled with MAX_PDCH_HIGH_LOAD = 1 timeslot
� Further, if the CS traffic demand could not be handled with the remaining timeslots some measures are necessary e.g.:
� add a TRX to the considered serving cell
� shrink the cell size of the serving cell (e.g. introduce downtilt) and increase the cell size of a neighbouring cell which offers sufficient capacity to handle the traffic demand surplus of the serving cell
� reduce interference (network changes) to get higher average throughput
� In the Alcatel GPRS implementation step 1, the number of TRX'swhich can be allocated to GPRS is maximum NTRXGPRS =1.
� In our worst case consideration, this TRX comes to its limit when the packet throughput demand is higher than the throughput capacity and cannot be satisfied even if the number of allocated TS for PS reaches Max_PDCH_Group
Switch to notes view!ALMAP: ALcatel MAnagement Platform APN: Access Point Name AS: Alpha Server (Compaq) BG: Border Gateway BSC: Base Station Controller BSS: Base Station Subsystem BSCGP: BSC-GPRS Protocol BSSGP: BSS-GPRS Protocol BVCI: BSSGP Virtual Connection Identifier CCBS: Customer Care and Billing Center CCU: Channel Codec Unit CDR: Call Detail Record CG: Charging Gateway CS: Circuit Switching DHCP: Dynamic Host Configuration Protocol DL: Down Link DLCI= Data Link Connection Identifier DNS: Domain Name System EDGE: Enhanced Data rates for GSM Evolution FUMO : Frame Unit Module FR: Frame Relay GPRS: General Packet Radio Service GGSN: Gateway GSN GMM: GPRS Mobility Management GR: GPRS Register GSL: GPRS Signaling Link GSM: Global System for Mobile communication GSN: GPRS Support Node GSS: GPRS Sub-System GTP: GPRS Tunneling Protocol HLR: Home Location Register HSCSD: High Speed Circuit-Switching Data IMSI: International Mobile Subscriber Identity IP: Internet Protocol ISDN : Integrated Service Digital Network ISP: Internet Service Provider LAN: Local Area Network LLC: Logical Link Control MAC: Medium Access Control MFS: Multi-Bsc Fast packet Server MNRG: Mobile Not Reachable for Gprs MS: Mobile Station MSC: Mobile Switching Center MT: Mobile Terminal NE: Network Element NMC: Network Management Center NNM: Network Node Manager NRPA : Network Requested PDP Context Activation NSAPI: Network Service Access Point Identifier NSC: Network Service Control layer NSEI: Network Service Entity Identifier NSS: Network Sub-System NS-VC: Network Service- Virtual Circuit NTP: Network Time Protocol DB : On Demand Bandwidth OMC: Operation & Maintenance Center OS: Operation System PAGCH: Packet- Access Grant Channel PCCCH: Packet- Common Control Channel
PCO: Protocol PCU: Packet Control Unit PDCH: Packet Data CHannel PDN: Packet Data Network PDP: Packet Data Protocol (IP or X25) PDU: Protocol Data Unit PPCH: Packet- Paging CHannel PRACH: Packet- Random Access CHannel PS: Packet Switching P-TMSI: Packet- Temporary Mobile Subscriber Identity PVC: Permanent Virtual Circuit P-VLR: Packet- Visitors Location Register QoS: Quality of Service RA: Routing Area RIP : Routing Information Protocol RLC: Radio Link Control RADIUS: Remote Authentication Dial In Use Service RRDTUF : Roaming Restriction Data Towards Unknown Foreign PLMN RRM: Radio Resource Management RSZ : Regional Subscription Zone SGSN: Serving GSN SM: Session Management | Short Message SMS: Short Message Service SMS-C: SMS-Center SNDCP: Sub Network-Dependent Convergence Protocol SNMP: Simple Network Management Protocol SNS: Sub-Network Service layer TBF: Temporary Block Flow TC: Trans Coder TCH: Traffic CHannel TCP: Transmission Control Protocol TDMA: Time-Division Multiplexing Access TFI: Temporary block Flow Identifier TID: Tunnel IDentity TLLI: Temporary Logical Link Identity TMN: Telecommunication Management Protocol TS: Time Slot UDP: User Datagram protocol UL: Up Link UMTS: Universal Mobile Transmission System WAP: Wireless Application Protocol WAN: Wide Area Network