0 оценок0% нашли этот документ полезным (0 голосов)
70 просмотров48 страниц
This optimisation guide deals with all aspects related to the QOS FOLLOW-UP and optimisation oI an EDGE network. In the whole document, B8 MR5 is considered, but the majority of the inIormation is valid also Ior lower releases oI the B8 BSS.
This optimisation guide deals with all aspects related to the QOS FOLLOW-UP and optimisation oI an EDGE network. In the whole document, B8 MR5 is considered, but the majority of the inIormation is valid also Ior lower releases oI the B8 BSS.
Авторское право:
Attribution Non-Commercial (BY-NC)
Доступные форматы
Скачайте в формате PDF, TXT или читайте онлайн в Scribd
This optimisation guide deals with all aspects related to the QOS FOLLOW-UP and optimisation oI an EDGE network. In the whole document, B8 MR5 is considered, but the majority of the inIormation is valid also Ior lower releases oI the B8 BSS.
Авторское право:
Attribution Non-Commercial (BY-NC)
Доступные форматы
Скачайте в формате PDF, TXT или читайте онлайн в Scribd
Abstract : this optimisation guide deals with all aspects related to the QoS Iollow-up and optimisation oI an EDGE network Algorithms, Parameters, Network Dimensioning, Field Tests and QoS Iollow-through. In the whole document, B8 MR5 is considered, but the majority oI the inIormation is valid also Ior lower releases oI the B8 BSS.
Key words : EDGE, radio algorithms, radio parameters, B8, QoS Follow-up
Appraisal and approval authorities PCS C. Brechtmann PCS France E. Desblancs DD-MM-YY : Signature : DD-MM-YY : Signature : PCS/NMO F. Jarreau DD-MM-YY : Signature : DD-MM-YY : Signature :
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 2 Table of contents
1 INTRODUCTION.................................................................................... 4 2 ALGORITHMS & PARAMETERS ....................................................... 5 2.1 CHAPTER STRUCTURE ................................................................................................5 2.2 MAIN CONCEPTS........................................................................................................5 2.3 RADIO RESOURCES MANAGEMENT...........................................................................10 2.4 TRANSMISSION RESOURCES MANAGEMENT...............................................................14 2.5 RADIO LINK ADAPTATION .......................................................................................16 2.6 PARAMETERS SUM-UP ..............................................................................................19 3 NETWORK DIMENSIONING............................................................. 24 4 QOS FOLLOW-UP................................................................................ 27 4.1 MAIN KPI`S ............................................................................................................27 4.1.1 TBF Establishment Rate ................................................................................27 4.1.2 TBF Retransmission Ratios ...........................................................................27 4.1.3 TBF Normal Release Rate (UL and DL) ........................................................29 4.1.4 CS/MCS distribution and useful traffic ..........................................................29 4.1.5 Resource Reallocation...................................................................................32 4.1.6 Ater congestion .............................................................................................33 4.1.7 KPIs order of importance.............................................................................34 4.2 QOS FOLLOW-UP DURING A FIELD-TRIAL..................................................................34 4.3 QOS FOLLOW-UP ON A REGULAR NETWORK..............................................................41 5 KNOWN ISSUES................................................................................... 43 5.1 ATER CONGESTION ..................................................................................................43 5.2 EDGE PERFORMANCE DEGRADATION WITH FREQUENCY HOPPING............................45 5.3 APD ISSUES ............................................................................................................45 5.4 FAULT REPORTS THAT MAY IMPACT EDGE PERFORMANCE.......................................46 5.4.1 Bad PING performance in some Nokia mobiles .............................................46 5.4.2 Wrong RLC block coding when retransmitting with MCS3 (with padding).....47 6 FINAL COMMENTS............................................................................. 48
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 3 History :
Edition Date Originator Comments 1 Proposal 1 30/03/2005 P. F. Costa Creation oI the document 1 Proposal 2 18/04/2005 P. F. Costa Second draIt version 1 22/04/2005 P. F. Costa Edition 1 Release
Edition changes : Corretions to the text New document template
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 4 1 Introduction The present document aims at presenting all the basic inIormation and methods needed Ior optimizing an EDGE network. This document will only Iocus on EDGE networks, but it is nevertheless important to have some knowledge on it. A similar document dedicated to GPRS networks is expected to appear at the time oI writing oI the current edition oI the present document, it was still not the case. It is also important to state that although the majority oI the algorithms are presented in this document, its objective is not to introduce Alcatel`s EDGE solution |1| exists Ior this purpose.
Chapter 3 presents all the EDGE related algorithms and parameters. Chapter 4 presents Network Dimensioning aspects. Chapter 5 presents all issues related to Iield-testing. Chapter 6 presents the proceedings Ior QoS Iollow-up both the main KPI`s and the most typical problems (and their workaround) are presented. Chapter 7 gives some Iinal comments.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 5 2 Algorithms & Parameters 2.1 Chapter structure The present chapter will Iollow the same structure than |1|, so please reIer to it Ior more details regarding the diIIerent algorithms considered.
2.2 Main concepts EDGE introduces some new aspects that Iatally distinguish it Irom the 'classical GPRS service: - New coding schemes are introduced EDGE uses MCS1 to MCS9 (MCS meaning Modulation and Coding Scheme). In release B8, MCS1 to MCS9 is available in DL, while MCS1 to MCS4 is available in UL. - New modulation besides the GMSK modulation, which is still used in MCS1 to MCS4, EDGE uses 8-PSK modulation. This modulation has a variable envelope, which impacts the available output power. - New LLC PDU segmentation while in GPRS the LLC PDU is segmented in RLC blocks taking into account the used coding scheme, in EDGE the LLC PDU is segmented in payload blocks (new concept in EDGE) taking into account the used MCS Iamily in EDGE the RLC block can carry one or two payload blocks. - The coding scheme Iamily concept, reIerred just above is also new in EDGE. Besides what was already reIerred, this new concept changes the way retransmission is made. Table 1 presents the diIIerent coding scheme Iamilies available (currently missing in the table below), as well as its theoretical throughput. - Puncturing Schemes new concept in EDGE also used in the retransmission mechanisms.
Coding scheme Coding scheme Family Modulation Theoretical throughput per PDCH (kbit/s) CS1 GMSK 8 CS2 GMSK 12 CS3 GMSK 14.4 CS4 - GMSK 20 MCS1 Family C GMSK 8.8 MCS2 Family B GMSK 11.2 MCS3 Family A; Family A padding GMSK 14.8 / 13.6
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 6 Coding scheme Coding scheme Family Modulation Theoretical throughput per PDCH (kbit/s) MCS4 Family C GMSK 17.6 MCS5 Family B 8-PSK 22.4 MCS6 Family A; Family A padding 8-PSK 29.6 / 27.2 MCS7 Family B 8-PSK 44.8 MCS8 Family A padding 8-PSK 54.4 MCS9 Family A 8-PSK 59.2 To6|o 1: CoJiog Schomos
Retransmission mechanism, or ARQ (Automatic Repeat request), has been greatly modiIied in EDGE.
The Iirst modiIication comes Irom the introduction oI the payload concept. In an EGPRS TBF the RLC block can be retransmitted either by using the same MCS or by using a MCS Irom the same coding scheme Iamily. A side eIIect Irom this is that even iI Link Adaptation (algorithm that changes the coding scheme according to radio conditions) is disabled, we can observe RLC blocks with diIIerent coding schemes (although always with the same Iamily).
Still in the ARQ mechanism, EDGE introduces a new type oI ARQ. Now we have: - Type 1 ARQ the decoding oI a re-transmitted RLC block does not take into account the previously transmitted versions oI this same RLC block. It is always used in GPRS and in uplink EGPRS. - Type 2 hybrid ARQ (used only in EGPRS) also known as Incremental Redundancy (IR), this mechanism works as Iollows: o The Iirst emission oI the RLC block is done with a Iirst puncturing scheme (PS1). o II the RLC block needs to be retransmitted, the same MCS or a MCS Irom the same Iamily will be used. The block may or may not be re-segmented (depending on parameterisation). The transmitter will also select the Puncturing Scheme to use. o The receiver will use the inIormation in all versions oI the received block as to increase its decoding probability.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 7 Note that 3GPP states that: - IR is mandatory in MS`s receiver; - There is no way to activate/de-activate IR in the MS receiver by signalling over the air-interIace IR can only be de-activated in the case oI insuIIicient memory in the mobile; - The soIt combining can be done between MCSx and MCSx blocks (same MCS used), a MCS9 and a MCS6 block (RLC data blocks with the same number oI payload units) or between a MCS7 block and a MCS5 block (RLC data blocks with the same number oI payload units).
Note also that IR is only available in DL in the B8 version oI the Alcatel BSS.
The RLC Window management also changes in EDGE. While in GPRS the RLC window size is Iixed at 64 blocks, in EDGE: - It is variable and determined Ior each TBF depending on the MS multi-slot class; - It varies between 64 and 1024 blocks (512 blocks in B8); - It is increased Iollowing an increase on the number oI PDCH`s, but not decreased in the case oI a reduction.
Hardware and Power Aspects In terms oI HW compability, EDGE is available (i.e. 8-PSK is supported) in all TRE Irom G4 (also called TRA) onward.
In all cases the output power available in GMSK is always greater than in 8-PSK this is due to design issues concerning the Iact that GMSK is a constant envelope modulation, while 8-PSK is not. Table 2 presents the output powers oI the diIIerent EDGE capable TRE`s available.
TkA GM5K output power 8-F5K output power 8-F5K output power (EDGE+ TkA} 00 Medium power 45 W / 4.5 d8m 15 W / 41.8 d8m 30 W / 44.8 d8m 00 High power 0 W / 47.8 d8m 25 W / 44 d8m 30 W / 44.8 d8m 1800 Medium power 35 W / 45.4 d8m 12 W / 40.8 d8m 30 W / 44.8 d8m 1800 High power 0 W / 47.8 d8m 25 W / 44 d8m 30 W / 44.8 d8m To6|o 2: Powot chotoc|otis|ics ltom |ho Jillotoo| ovoi|o6|o Tk/ (EDGE copo6|o TkEsj.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 8 Because GMSK output power is oIten diIIerent Irom 8-PSK output power, there is a new concept introduced with EDGE the Average Power Decrease (APD).
According to the 3GPP specs, there are some constraints on this parameter when the TRE responsible Ior the EDGE service is also carrying the BCCH. |4| states that: 'Furthermore, 8-PSK modulated timeslots on the BCCH carrier may use a mean power which is at most 4 dB lower than the mean power used Ior GMSK modulated timeslots with the exception oI the timeslot preceding a slot used Ior BCCH/CCCH where at most 2 dB lower mean power is allowed. (3GPP 45.008 5.i.0, section 7.1)
3GPP also predicts some speciIic problems Ior Iast moving mobiles, in |4|: 'In the case that 8-PSK modulation is allowed on the BCCH carrier and Irequency hopping including the BCCH carrier is used, the reception quality in connected mode Ior some Iast moving MS (meaning MS experiencing Doppler Irequencies oI 100 Hz or more) may be degraded. This may be seen as a backwards compatibility problem Ior some existing MS, most likely occurring iI the used APD is larger than 2 dB. (3GPP 45.008 5.i.0, section 7.1)
A very important new concept (Alcatel SpeciIic) introduced with EDGE (and in GPRS CS3/CS4) is the TRX class concept. Depending on the used MCS, Irom one to Iive 16kbit/s channels (GCH) are needed per PDCH between the BTS and the MFS. So, each TRX has a TRX class (Irom one to Iive) that deIines the number oI Abis/Ater GCH per radio TS or PDCH. Note also that 1 EGCH n GCH n 16 kbit/s channels).
A concept that is associated to the notion oI TRX class n is the TRX Transmission pool this is the set oI extra Abis nibbles (Abis nibble (16k) used to carry traIIic when TRX is a class n TRX , n ~ 1).
Regarding the TBF mode and TBF establishment, there exist some diIIerences between GPRS and EDGE. Please reIer to |1| and |2| Ior more details.
Table 3 presents the most important parameters related to these new concepts.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 9 Forometer nome Definition Defoult Volue EM_EGFk5 Enobles (disobles} EDGE EMA8LE MAX_EGFk5_MC5 Moximum MC5 (Modulotion ond Coding 5cheme} used for EGFk5 troffic in the cell MC5 M8_EXTkA_A85_T5 Mumber of oll extro Abis (4k} T5 of oll the pools defined on the 2 possible sectors on which the cell is mopped - EM_FULL_k_DL Disobles (enobles} the re-segmentotion for the downlink T8F in the cell EMA8LE 85_TXFWk_MAX 8T5 moximum ollowed tronsmission power. 0 To6|o 3: So|oc|ioo ol potomo|ots to|o|oJ |o moio coocop|s io EDGE
II about EN_EGPRS there is not much that can be said (setting it to enable, or disable, just enables, respectively disables, EDGE service in the cell), regarding the three other parameters some notes should be added.
MAX_EGPRS_MCS is comparable to MAXGPRSCS in GPRS it indicates the maximum MCS achievable in terms oI parameterisation. But note that the maximum MCS achievable depends on other Iactors like the TRX class oI the TRX where the EDGE call is being made. Usually this parameter is set to 9 hence letting that the TRX class available deIine the maximum MCS achievable.
NB_EXTRA_ABIS_TS is a virtually changeable parameter although it is possible to see it in the OMC terminal, it is not possible to change it. Its value is automatically deIined at the OMC depending on the conIiguration oI the cell. Nevertheless it permits us to have an idea oI the EDGE capabilities oI the cell. It is important to note that a class n TRX will need (n-1)*2 extra Abis TS (e.g., a cell with only one class 4 TRX will have NBEXTRAABISTS 6).
EN_FULL_IR_DL has a somewhat deceitIul name in Iact IR is always enabled on the MS receiver, only re-segmentation is disallowed (ENFULLIRDL enable) or allowed (ENFULLIRDLdisable). By deIault, ENFULLIRDL is set to enable (hence, re- segmentation is disallowed). This recommendation comes Irom the diIIerent experimentations perIormed already by PCS team.
BS_TXPWR_MAX is not an EDGE speciIic parameter but a GSM one. This parameter deIines the BTS maximum allowed transmission power, which itselI is deIined relatively to the BTS maximum RF transmit power in GMSK. The BTS will have a GMSK output power equal to BTSGMSKoutputpower BSTXPWRMAX. This permits to reduce the APD oI a cell, as it permits to change the GMSK output power. Normally, this parameter is not changed, as it aIIects the site coverage.
2.3 Radio Resources Management All the subjects Irom this chapter are described in detail in |2| and |3|. Regarding the TRX conIiguration and priority deIinition, it is basically the same in GPRS or EDGE the only distinction is made between CS and PS traIIic. Contrary to the previous version oI the BSS, B7.2, there is only one parameter Ior setting TRX priority both Ior CS and PS traIIic TRXPREFMARK. Furthermore, only TRX`s with TRXPREFMARK 0 can carry PS traIIic.
The process determining the TRX priority, handled by the BSC, can be described in 4 steps: - DR (Dual Rate) TRE selection the priority is given to non EDGE TRE`s, then to Medium Power EDGE TRE`s and Iinally to High Power EDGE TRE`s. - TRX-TRE mapping PS capable TRX are mapped preIerentially (Irom highest to lowest priority) on the Iollowing TRE (the reverse order is used to set the TRX priority Ior the CS traIIic): o Full Rate, High Power, EDGE capable TRE o Dual Rate, High Power, EDGE capable TRE o Full Rate, Medium Power, EDGE capable TRE o Dual Rate, Medium Power, EDGE capable TRE o Full Rate, non EDGE capable TRE o Dual Rate, non EDGE capable TRE - Ordering oI the TRX Ior the PS traIIic the PS capable TRX are ordered as Iollows (Irom highest to lowest priority): o BCCH TRX iI PSPreIBCCH true o HW TRX capability: EGPRS High Power, then EGPRS Medium Power, then non EGPRS o Full Rate then Dual Rate o Size oI the PDCH group: rank Iirst TRX with maximum number oI consecutive SPDCH per TRX o Finally TRX identity (lowest TRX id Iirst) - Association oI the TRX transmission pools to each TRX the biggest TRX transmission pool is associated to the TRX with the highest PS priority.
For the Resource allocation, there is a set oI principles that are Iollowed when establishing an EGPRS TBF: - EGPRS TBF must be established preIerentially on the highest class TRX;
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 11 - It is preIerred to multiplex several EGPRS TBF`s on a high class TRX than allocating the EGPRS TBF on a low-class TRX; - A new re-allocation trigger (T4 re-allocation) is added Ior UL GPRS TBF that share at least one PDCH on an EDGE TRX with a DL EGPRS TBF although it is now possible that an UL GPRS TBF can share the same PDCH than a DL EGPRS TBF, it is not advisable, as it brings some perIormance degradation.
To Iollow the principles stated above, some rules are needed when establishing the EDGE TBF. The BSS selects only the EGPRS TRX oI highest class (that is which oIIer the highest throughput) as long as the maximum number oI EDGE TBF per PDCH on these TRX is not higher than a threshold (NTBFPDCHMCSiMCSj). There are some exception to this rule: in the case oI Immediate UL EGPRS TBF establishment (only EGPRS TRX are taken into account, but they are not necessarily oI the highest class) and in the case oI a UL (respectively DL) EGPRS TBF establishment when there is a concurrent DL (respectively UL) EGPRS TBF and T2 re-allocation is disabled (the TRX which supports the ongoing DL (respectively UL) EGRPS TBF is the only candidate).
AIter selecting the TRXs where to allocate the EGPRS TBF, some additional rules are Iollowed Ior the timeslot allocation. The Iollowing order is respected: - Candidate allocations with the highest number oI non-busy PDCH in the direction oI the bias; - Candidate allocations with the highest number oI non-Iull PDCH in the direction oI the bias; - Candidate allocations with the highest number oI non-busy PDCH in the direction opposite oI the bias; - Candidate allocations with the highest number oI non-Iull PDCH in the direction opposite oI the bias; - Candidate allocations with the lowest number oI established EGPRS TBF in the direction oI the bias; - Candidate allocations with the lowest number oI established EGPRS TBF in the direction opposite oI the bias; - Candidate allocations on the TRX with the highest priority (reIer to the beginning oI part 3.3); - Candidate allocations with the lowest number oI GPRS TBF in the direction oI the bias; - Candidate allocations which have all their PDCH established (iI all the preIerred candidate allocations require additional PDCH, a request is sent to the BSC); - Candidate allocations which do not require a re-allocation oI the ongoing TBF; - Candidate allocations with the PDCH oI lowest index.
By candidate allocation we mean either a double list oI contiguous PDCH in a TRX (one list per direction), which veriIies the concurrent constraints; or a single list oI contiguous PDCH in a TRX in the direction oI the request.
In the present BSS version, B8, 4 re-allocation triggers exist: T1 to T4. The Iollowing priorities are Iollowed: T1 ~ T4 ~ T3 Incoming TBF establishment request ~ T4
T1 and T2 remain unchanged regarding the GPRS and EGPRS TBF.
In T3 re-allocation, a new condition was added Ior EGPRS TBF. The re-allocation is triggered when: - An EDGE TBF is not established on the TRX which has the highest EGPRS capability AND - More than NCANDIDATEFORREALLOC bytes have been exchanged since the beginning oI the TBF or the last re-allocation AND - T3192 is not running
As Ior T4, this re-allocation occurs when an UL GPRS TBF and a DL EDGE TBF share at least one PDCH on an EDGE TRX the UL GPRS TBF (or more exactly the MS) will be candidate Ior the re-allocation.
One oI the new Ieatures Irom BSS version B8 is the Iast initial (E)GPRS access. II the parameter ENFASTINITIALGPRSACCESS true (hence enabling the Ieature) one PDCH is allocated separately and preIerably on an EDGE TRX (to provide Iast initial access Ior GPRS and EGPRS), then on a TRX with the lowest class (to waste not too much transmission resources), then on a TRX with the highest priority (to avoid a conIlict with the CS traIIic) and Iinally on the timeslot with lowest index (in order to minimise the risk oI having to perIorm a T1 re-allocation).
Table 4 presents the most important parameters related to these new concepts (reIer also to part 3.6 Ior more details on these parameters).
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 13 Forometer nome Definition Defoult Volue TkX_FkEF_MAkK Freference Mork ossigned to o given TkX - F5_FkEF_8CCH_TkX ndicotes whether or not the F5 requests sholl be preferentiolly served with FDCH(s} of the 8CCH TkX D5A8LE EM_kE5_kEALLOCATOM Enobles (disobles} the resource reollocotion feoture (individuol triggers considered} 1 (oll triggers octivoted} EM_FA5T_MTAL_GFk5_ACCE55 This flog indicotes whether or not one 5love FDCH for (E}GFk5 troffic usoge will be stoticolly estoblished in the cell D5A8LE To6|o 4: So|oc|ioo ol potomo|ots to|o|oJ |o koJio kosoutco Mooogomoo|
TRX_PREF_MARK was already discussed above. It is not possible to talk about a deIault value Ior it what is certain is that at least one TRX must have this parameter set to 0 to be able to carry PS traIIic.
Alcatel`s deIault value Ior PS_PREF_BCCH_TRX is 'disable. But several networks have opted to chose a diIIerent value both options present advantages and disadvantages. Opting to give priority Ior PS traIIic Ior the BCCH TRX might have the advantage oI using a cleaner Irequency this is true only in the case oI NH/RH or SFH, not in the case oI BBH. Another advantage may arise Irom the Iact that the TRE recovery mechanism always gives priority to the BCCH TRX. Regarding disadvantages oI the above option (and so, advantages Ior the option oI not giving priority to the BCCH TRX), there are mainly two: - On one hand, the biggest PDCH group possible on the BCCH TRX is, in the best case scenario, 7 TS, and depending on SDCCH conIiguration can be even smaller. In the case where there is a TRX with class above any oI the other TRX`s, this means that the number oI TS 'enjoying this extra transmission capability (and extra throughput that goes with it) is smaller. - On the other hand, there are also some 3GPP requirements regarding the 8-PSK APD when EDGE is on the BCCH TRX (see 2.2). There might be the need to decrease the cell coverage (by decreasing the GMSK power output) in order to Iollow these recommendations this can have a inadmissible impact on CS service. (Note that recent Iield experiments seem to indicate that there is no negative impact in recent MSs behaviour when disregarding these recommendations - it is relatively saIe not to Iollow the 3GPP specs in this aspect.)
EN_RES_REALLOCATION can be changed as to enable any oI the diIIerent Iour re- allocation triggers. The deIault value is nonetheless 'enable all triggers option that will permit a better EDGE perIormance.
EN_FAST_INITIAL_GPRS_ACCESS is by deIault disabled. Anyway, many operators have chosen to activate this Ieature. Its major gains are in terms oI end-user perIormance. Because the resources are allocated in advance, some time can be gained when establishing a DL or UL TBF. The disadvantage oI this Ieature is that PDCH and GCH might be allocated Ior nothing, causing some waste oI transmission resources.
2.4 Transmission resources management Regarding transmission resources, two aspects can be considered: Abis resources and Ater resources.
As Iar as Abis resources are concerned, we have already introduced the TRX class n and the TRX transmission Pool concepts beIore.
As Iar as Ater is concerned, two concepts must be deIined at this moment: Ater congestion state oI a DSP (in the MFS GPU), and Ater usage.
Ater congestion is deIined as: - Beginning, iI the number oI available Ater nibbles Ior the DSP is lower than a threshold (NbAterThreshold); - End, iI the number oI available Ater nibbles Ior the DSP is equal or higher than a threshold (NbAterThreshold). See 0 Ior more details on how this threshold is deIined.
When a DSP enters in Ater congestion state, it triggers an algorithm to reduce the number oI GCH. This process consists, Ior one TRX, in reducing the number oI GCH to 2 (iI MaxGPRSCS ~ 2 in the cell) or to 1 (iI MaxGPRSCS 2 in the cell) Ior the PDCH that do not carry any EGPRS TBF. When a TRX has been processed, the algorithm is stopped iI the number oI available Ater nibbles Ior the DSP is higher or equal than NbAterthreshold. See 0 Ior more details on which TRX`s are processed.
When there is no Ater congestion, each TS oI a class n TRX uses n GCH (1 EGCH n GCH) on the Ater interIace. But during the Ater congestion state, things change: - In the case oI an EGPRS TBF establishment on a class n TRX, there is no impact - one EGCH is established with n GCH
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 15 - In the case oI GPRS TBF establishment on a class n TRX, one EGCH is established with 1 (iI MaxGPRSCS 2) or 2 (iI MaxGPRSCS ~ 2) GCH note that this has absolutely no impact on end-user perIormance, as 1 GCH (respectively 2) are enough to have coding scheme CS2 (respectively CS4).
One slight nuance must be pointed out here. Although EGPRS TBF`s are not supposed to be aIIected by this GCH reduction process, it was seen in the Iield that this is not true the GCH reduction does aIIect EGPRS TBF`s.
What might happen is that iI a EGPRS TBF is established during the time when a GPRS TBF is still established, or iI it is established less than TPDCHINACTIVITY seconds aIter the end oI the GPRS TBF, it may use the same PDCH's that were only associated to 2 GCH's. This will limit the maximum coding scheme to MCS5.
Ater usage is deIined as the percentage oI Ater nibbles that are used in a GPU: AterGPUUsage NbusedAternibbles / NbAternibbles * 100.
In case oI high Ater usage, the maximum number oI PDCH allocated to a new TBF (transitions between normal and high Ater usage have no consequence on on-going TBF) is limited to MAXPDCHPERTBFHighAterUsage whereas it is limited to MAXPDCHPERTBF in case oI normal Ater usage. Table 5 presents the most important parameters related to these new concepts (reIer also to part 3.6 Ior more details on these parameters).
Forometer nome Definition Defoult Volue GFk5_MULT5LOT_CLA55_DEFAULT_VALUE Defoult volue of the GFk5 multislot closs ossumed ot T8F estoblishment when the octuol M5 multislot closs is unknown 8 ATEk_U5AGE_THkE5HOLD Threshold (percentoge of used Ater nibbles, in o GFU} obove which the Ater usoge is soid high. 0.7 MAX_FDCH_FEk_T8F Moximum number of FDCH ollocoted to o single (E}GFk5 connection 5 MAX_FDCH_FEk_T8F_HGH_ATEk_U5AGE Moximum number of FDCH ollocoted to o single (E}GFk5 connection, when the Ater usoge is high 5 To6|o 5: So|oc|ioo ol potomo|ots to|o|oJ |o Ttoosmissioo kosoutco Mooogomoo|
GPRS_MULTISLOT_CLASS_DEFAULT_VALUE is normally set at 8, meaning a 41 MS. There may be some waste oI resources when considering that by deIault all mobile are
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 16 class 8 in some cases (especially in case oI Ater congestion) it is advised to set this parameter to a smaller value.
ATER_USAGE_THRESHOLD is usually set to 70. This parameter may be changed to make the PDCH reduction process kick in sooner (by reducing the threshold) or later (by increasing the threshold).
MAX_PDCH_PER_TBF and MAX_PDCH_PER_TBF_HIGH_ATER_USAGE deIine the maximum number oI PDCH that a TBF can use in respectively normal and high Ater usage conditions. Normally both are set to 5. This value is more than enough, as no mobile will need more than 4 TS Ior establishing the TBF. As both parameters are equal, the PDCH reduction process has generally no impact.
2.5 Radio Link Adaptation In EGPRS new radio measurements are used (as deIined in 0) MeanBEP (Mean Bit Error Probability) and CVBEP (CoeIIicient oI Variation oI the Bit Error Probability).
The MS can report 32 diIIerent MeanBEP values (MEANBEP0 to MEANBEP31), and the reported MeanBEP value depends on the used modulation. As Ior CVBEP values, the MS can report 8 diIIerent CVBEP values (CVBEP0 to CVBEP7). The BTS measures Ior each UL burst the BEP and calculates Ior each UL radio block (4 bursts) the MeanBEP and the CVBEP StdBEP / MeanBEP. The MeanBEP and the CVBEP are reported on a radio block basis by the BTS to the MFS.
The EGPRS link adaptation can be disabled thanks to the ENCSADAPTATIONACK (RLC acknowledged mode) and ENCSADAPTATIONNACK (RLC unacknowledged mode) parameters (same parameters than Ior the GPRS link adaptation).
Regarding the Algorithms used both in DL and UL Ior the Link Adaptation, please reIer to |1| Ior details.
One point to keep in mind is that this process does not use O&M parameters but internal tables taking into account the RLC mode oI the TBF (acknowledged or unacknowledged), then the ENFULLIRDL parameter value (TBF acknowledged mode only), then the APD (Average Power Decrease) in 8PSK oI the used TRE and Iinally the modulation scheme currently used (GMSK or 8PSK).
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 17 There is also a DL and UL TBF deIence mechanism available. In DL, iI too many consecutive EGPRS Packet downlink Ack/Nack messages are asked by the MFS but are not received, the modulation and coding scheme is changed to MCS1 (whatever the value oI the current coding scheme). In UL, the deIence procedure is used in case too many consecutive uplink radio blocks are not decoded. II a certain number oI conditions is IulIilled, the coding scheme is changed to MCS1.
Table 6 presents the most important parameters related to these new concepts (reIer also to part 3.6 Ior more details on these parameters).
Forometer nome Definition Defoult Volue 8EF_FEkOD Filter constont for EGFk5 chonnel quolity meosurements 10 EM_C5_ADAFTATOM_ACK/MA CK Enobles (disobles} the link odoptotion in kLC ock (respectively unock} mode EMA8LE T8F_DL/UL_MT_MC5 Volue of the DL/UL modulotion ond coding scheme when the link odoptotion olgorithm is disobled or initiol volue of the modulotion ond coding scheme otherwise MC53 T8F_MC5_DL/UL n DL. For o monoslot T8F olone on its FDCH, threshold defining the number of consecutive EGFk5 Focket Downlink Ack/Mock not received obove which the coding scheme of o downlink ocknowledged or unocknowledged T8F is chonged to MC51 (only in downlink}. For o multi-slot T8F or o T8F thot shores its FDCH(s}, the limit is proportionol to the instontoneous bondwidth ollocoted to the T8F. n UL. Threshold defining the moximum number of consecutive times the network receives on involid UL kLC doto block or nothing from the M5 hoving o monoslot EGFk5 T8F before chonging the coding scheme to MC51 DL. 12 UL. 8 E_TX_EFFCEMCY_FEkOD Mumber of received rodio blocks for on EGFk5 T8F ofter which E_TX_EFFCEMCY is computed. 200 TX_EFFCEMCY_ACK/MACK_TH k Threshold below which the T8F is releosed becouse of o bod tronsmission efficiency in ock (respectively nock} mode ock mode. 10% nock mode. 15% To6|o : So|oc|ioo ol potomo|ots to|o|oJ |o koJio Liok /Jop|o|ioo
BEP_PERIOD is a parameter considered in the calculation oI the MeanBEP value. Its deIault value was Iound through the various Iield trials and experimentations done so Iar.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 18 II a bigger value is used, the MeanBEP calculation will be less reactive, iI it is smaller, it will be more re-active. Even considering this, the value oI 10 should be good enough either Ior normal cells or Ior Iast moving traIIic cells. It is also important to know that in previous versions oI the BSS B8 (MR4 or lower), the deIault value was 25, which is coded 10 this might arise some conIusion. The actual deIault is 10, coded 6.
EN_CS_ADAPTATION_ACK/NACK is the parameter responsible Ior enabling (or disabling) the Link Adaptation mechanism. It is important to note here that even iI Link Adaptation is disabled, it is possible that other MCS diIIerent Irom the initial MCS might be used iI too many erroneous packets are exchanged, MCS switches back to MCS1, and it is possible that retransmissions are done in a diIIerent MCS but Irom the same MCS Iamily.
TBF_DL/UL_INIT_MCS deIine in which MCS the TBF should start. The deIault value can be considered as rather conservative, but advantages and disadvantages exist in changing it to a higher (or lower) MCS. Starting with a high MCS means that the throughput will potentially be high since the beginning oI the TBF. Furthermore, even short TBF`s will use a high MCS. In good radio conditions, it is clear that the average throughput will be higher, particularly in the case oI short TBF`s (on long TBF`s the Link Adaptation mechanism would have enough time to kick in, and the initial lower throughput would be diluted on the whole TBF transIer, hence reducing the impact oI low initial MCS). The disadvantages oI starting with a high MCS are equally evident. In the case oI medium/bad radio conditions, MCS will have to be decreased beIore successIully transmitting any blocks. This 'delay will obviously impact the throughput achieved and again, the impact will be bigger on short TBFs. KPIs will also be impacted by these additional Iailures (principally TBF establishment and retransmission rates see 4.1 Ior details on KPIs).
TBF_MCS_DL/UL is considered on the TBF deIence mechanism. For a monoslot TBF alone on its PDCH, TBFMCSDL deIines the number oI consecutive EGPRS Packet Downlink Ack/Nack not received above which the coding scheme oI a downlink acknowledged or unacknowledged TBF is changed to MCS1 (only in downlink). For a multi-slot TBF or a TBF that shares its PDCH(s), the limit is proportional to the instantaneous bandwidth allocated to the TBF. The TBFMCSUL deIines the threshold Ior the number oI missed reception oI RLC blocks (either because they are invalid or because they were not received at all) beIore switching back to MCS1 (coding scheme which oIIers best protection). These parameters are the EDGE counterparts oI the TBFCSDL/UL parameters.
E_TX_EFFICIENCY_PERIOD and TX_EFFICIENCY_ACK/NACK_THR are considered Ior the DL deIence mechanism, together with EGPRSN3105LIMIT.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 19 Contrary to TBFMCSDL/UL, they deIine thresholds Ior abnormally releasing the TBF. II EGPRSN3105LIMITis Iixed at BSS and MFS level, the Iirst two parameters permit changes. They are not usually considered when optimising parameters, though. Note also that ETXEFFICIENCYPERIOD is the EDGE counterpart oI the TXEFFICIENCYPERIOD parameter.
2.6 Parameters sum-up In the previous sections, only some parameters were presented. Table 7 presents all these parameters and other that were not seen beIore.
Mome Type konge OMC-k occess Definition Generol porometers ACCE55_8Uk5T_TYFE Cell 0 or 1 Formot of the occess burst used by (E}GFk5 M5s when sending o FACKET CHAMMEL kEQUE5T, o FACKET COMTkOL ACKMOWLEDGEMEMT or o FTCCH/U messoge Ater_Usoge_Threshold 855 1 to 100 % Threshold (percentoge of used Ater nibbles, in o GFU} obove which the Ater usoge is soid "high" EM_EGFk5 Cell Enoble/Disoble Enobles/Disobles EGFk5 troffic in the cell EM_FA5T_MTAL_GFk5_ACCE55 Cell Enoble/Disoble This flog indicotes whether or not one 5love FDCH for (E}GFk5 troffic usoge will be stoticolly estoblished in the cell MAX_EGFk5_MC5 Cell 1 to Moximum Modulotion ond Coding 5cheme used for EGFk5 troffic in the cell MAX_GFk5_C5 Cell 2 to 4 Moximum coding scheme used for GFk5 troffic in the cell MAX_FDCH_FEk_T8F_High_Ater_ Usoge Cell 1 to 5 Moximum number of FDCH ollocoted to o single (E}GFk5 connection, when the Ater usoge is "high" M8_EXTkA_A85_T5 Cell 0 to Virtuol chongeoble Mumber of oll extro Abis (4k} timeslots of oll the pools defined on the 2 possible sectors on which the cell is mopped M8_T5_MFDCH Cell 0 to 4 Mumber of rodio timeslots reserved for the primory ond secondory moster FDCHs defined in the cell
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 20 Mome Type konge OMC-k occess Definition MC_kEFOkTMG_FEkOD_T Cell 0.48 to 1.44 s keporting period of MC meosurements reported by the M5 while in pocket tronsfer mode F5_FkEF_8CCH_TkX Cell 0 or 1 ndicotes whether or not the F5 requests sholl be preferentiolly served with FDCH(s} of the 8CCH TkX T_DL_EGFk5_Meoskeport Cell 0 to 3000 ms Time period to request for on EGFk5 Focket Downlink Ack/Mock with meosurements AkQ porometers EGFk5_DL_Ack_Foctor MF5 0 to 1 Mo occess f the EGFk5 troffic is corried through Abis terrestriol link ond Ater terrestriol link, this porometer defines the rotio between the number of tronsmitted kLC doto blocks necessory to request for o EGFk5 Focket Downlink Ack Mock in ocknowledged mode, ond the kLC window size EGFk5_DL_ACK_Foctor_5otellite 855 0 to 1 Mo occess f the EGFk5 troffic is corried through Abis sotellite link or Ater sotellite link, this porometer defines the downlink EGFk5 AkQ period foctor. Acknowledgement period over window size rotio EGFk5_UL_Ack_Foctor MF5 0 to 1 Mo occess f the EGFk5 troffic is corried through Abis terrestriol link ond Ater terrestriol link, this porometer defines the rotio between the number of received kLC doto blocks necessory to send o Focket Uplink Ack/Mock in ocknowledged mode, ond the kLC window size EGFk5_UL_Ack_Foctor_5otellite MF5 0 to 1 Mo occess f the EGFk5 troffic is corried through Abis sotellite link or Ater sotellite link, this porometer defines the rotio between the number of received kLC doto blocks necessory to send o Focket Uplink Ack/Mock in ocknowledged mode, ond the kLC window size EM_FULL_k_DL (from 88 Mk5 releose} 855 Enoble/Disoble Enobles/disobles the re- segmentotion for the downlink T8F in the cell EM_k (before 88 Mk5 releose} 855 Enoble/Disoble Enobles/disobles ncrementol redundoncy for the downlink T8F in the cell
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 21 Mome Type konge OMC-k occess Definition kodio meosurements 8EF_FEkOD Cell 1 to 25 Filter constont for EGFk5 chonnel quolity meosurements MC5_AVG_FEkOD MF5 0.02 to 5 s Mo occess Feriod Tegprs ofter which o weight of 0. is opplied to the current UL or DL meon8EF somples in the overoging Modulotion ond Coding 5cheme odoptotion filter M_AVG_ Cell 0 to 15 M_AVG_ volue is broodcost to the (E}GFk5 M5 in the 513. The (E}GFk5 M5 uses it for filtering the interference signol level meosurements thot it sholl perform during the pocket tronsfer mode. These meosurements ore used by the (E}GFk5 M5 for output power control, but olso reported to the network ond used for performing downlink link odoptotion kodio link supervision E_TX_EFFCEMCY_FEkOD 855 1 to 500 Mumber of received rodio blocks for on EGFk5 T8F ofter which E_TX_EFFCEMCY is computed EGFk5_M3105_LMT MF5 1 to 1 Mo occess n o mono-slot DL EGFk5 T8F tronsfer procedure, if the EGFk5 troffic is corried through Abis terrestriol link ond Ater terrestriol link, this porometer defines the moximum number of expected EGFk5 Focket Downlink ACK/MACK or Focket Control Acknowledgement messoges consecutively lost (from the M5} on the rodio interfoce, before releosing obnormolly the DL T8F. For o multi-slot EGFk5 T8F or on EGFk5 T8F which shores its FDCH(s}, the limit is proportionol to the instontoneous bondwidth ollocoted to the T8F
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 22 Mome Type konge OMC-k occess Definition EGFk5_M3105_LMT_5ATELLTE MF5 1 to 1 Mo occess n o mono-slot DL EGFk5 T8F tronsfer procedure, if the EGFk5 troffic is corried through Abis sotellite link or Ater sotellite link, this porometer defines the moximum number of expected EGFk5 Focket Downlink ACK/MACK or Focket Control Acknowledgement messoges consecutively lost (from the M5} on the rodio interfoce, before releosing obnormolly the DL T8F. For o multi-slot EGFk5 T8F or on EGFk5 T8F which shores its FDCH(s}, the limit is proportionol to the instontoneous bondwidth ollocoted to the T8F Link odoptotion for EGFk5 EM_C5_ADAFTATOM_ACK Cell Enoble/Disoble Enobles / disobles the link odoptotion in kLC ocknowledged mode EM_C5_ADAFTATOM_MACK Cell Enoble/Disoble Enobles / disobles the link odoptotion in kLC unocknowledged mode T8F_DL_MT_MC5 Cell 1 to Volue of the downlink modulotion ond coding scheme when the link odoptotion olgorithm is disobled or initiol volue of the modulotion ond coding scheme otherwise T8F_MC5_DL 855 0 to 15 For o monoslot T8F olone on its FDCH, threshold defining the number of consecutive EGFk5 Focket Downlink Ack/Mock not received obove which the coding scheme of o downlink ocknowledged or unocknowledged T8F is chonged to MC51 (only in downlink}. For o multi-slot T8F or o T8F which shores its FDCH(s}, the limit is proportionol to the instontoneous bondwidth ollocoted to the T8F T8F_MC5_FEkOD MF5 1 to 25 Mo occess Mumber of rodio blocks ofter which the MC5 odoptotion decision moy be mode, in EGFk5
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 23 Mome Type konge OMC-k occess Definition T8F_MC5_UL 855 1 to 12 Threshold defining the moximum number of consecutive times the network receives on involid UL kLC doto block or nothing from the M5 hoving o monoslot EGFk5 T8F before chonging the coding scheme to MC51. For o multislot EGFk5 T8F, T8F_MC5_UL_limit = T8F_MC5_UL x n_ollocoted_timeslots T8F_UL_MT_MC5 Cell 1 to 4 Volue of the uplink modulotion ond coding scheme when the link odoptotion olgorithm is disobled or initiol volue of the modulotion ond coding scheme otherwise To6|o 7: EDGE ooJ HSDS potomo|ots (88 to|oosoj
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 24 3 Network Dimensioning The present document does not intend to go deeply in the Network Architecture and Dimensioning. Nonetheless, the system architecture is a very important part oI the whole system description and will be looked at in this chapter in particular some notions will be described. Moreover the architecture topic is very important when considering EDGE as the EDGE introduction has a major impact on the BSS dimensioning due to the static Abis allocation in B8.
For more details concerning System Architecture and Dimensioning, please reIer to 0.
By allowing new modulations and coding schemes, the data traIIic throughput oI a given radio timeslot was increased. This implies that this increase oI throughput is handled on Abis and Ater interIaces too.
Up to now, only a 16 kbit/s nibble was allocated per timeslot in each interIace, clearly insuIIicient Ior the actual possible throughputs.
MFS Telecom allocates dynamically several Ater nibbles (16kbit/s channels) on the Ater interIace, to allow higher than 16 kbit/s throughputs.
On the Abis interIace, several Abis nibbles are also used. But contrary to the Ater interIace, the conIiguration oI these Abis nibbles is done statically. Depending on the requested throughput, a radio timeslot will have associated up to 4 extra Abis nibbles (in addition to the basic one). All radio timeslots must have the same the same throughput capability, and so a TRX may need up to 8 extra Abis timeslots (as seen above, these extra Abis timeslots are called a 'TRX Transmission Pool).
8ocuuso lhoro is u big incrouso in lho numbor o Abis rosourcos noodod, ono singlo Abis link mighl nol bo onough. 88 inlroducos lhon lho possibilily o udding u socond Abis link. This socond Abis link cun only bo usod or puckol lruic - CML, PSL und busic limoslols uro ulwuys muppod lo lho irsl link, whilo lho oxlru limoslols uro splil ovor lho lwo Abis links.
Other important notion is that Extra Abis timeslots are mapped only on a TCU supporting TRE FR (a DR TCU does no support extra Abis TS). It implies that the operator may need to trigger a complete reshuIIling to group all DR TRE and to Iree some TCU Ior FR TRE or extra Abis TS.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 25 Regarding the FR TCU occupancy, we need to use the 'Equivalent Full-Rate (Eq-FR) concept to look at it, as described in Table 8.
Number oI Eq-FR available on a TCU 4 Number oI Eq-FR needed Ior a TRE FR 1 Number oI Eq-FR needed Ior a TRE DR 2 Number oI Eq-FR needed Ior an extra Abis timeslot 0.5 To6|o 8: Fk TCU occupoocy co|cu|o|ioo
As Ior TSU connectivity: o 1 Dual Rate TRX is equivalent to 2 Full Rate TRX o 1 TRX with a Transmission pool type 2 is equivalent to 2 Iull rate TRX o 1 TRX with a Transmission pool type 3 is equivalent to 3 Iull rate TRX o 1 TRX with a Transmission pool type 4 is equivalent to 4 Iull rate TRX o 1 TRX with a Transmission pool type 5 is equivalent to 5 Iull rate TRX
It is possible that Dual Rate is supported on a class N TRX; a 'dual rate class N TRX will consume in the BSC equivalent oI the connectivity consumed by (N1) simple Full Rate TRXs.
To be able to map the cell onto the GPU, the MFS needs to know the 'GPRS weight oI the cells. In B8, this 'GPRS Weight corresponds to the number oI extra Abis timeslots (NBEXTRAABISTS).
Regarding the GPU capability, this depends on the TRX class and iI EDGE is activated or not.
Table 9 shows the GPU capability in the diIIerent possible cases.
TRX Class Coding Schemes Max nb oI PDCH per GPU EDGE not-Activated 1 CS1, 2 240 2 CS1, 2, 3, 4 208
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 27 4 QoS Follow-up 4.1 Main KPI`s 4.1.1 TBF Establishment Rate This indicator shows iI the TBF's are being correctly established.
For DL (in both RNO and NPA): F_roquosls GPPS_DL_T8 F_succoss GPPS_DL_T8 (CPDTESUP} rulo F_succoss_ GPPS_DL_T8 =
For UL (in both RNO and NPA): F_roquosls GPPS_UL_T8 F_succoss GPPS_UL_T8 (CPUTESUP} rulo F_succoss_ GPPS_UL_T8 =
The name oI the indicators above might be deceiving, as both GPRS and EDGE TBF`s are considered. Moreover it is not possible to distinguish GPRS and EDGE TBF with the available counters. 4.1.2 TBF Retransmission Ratios This indicator serves to have an idea iI aIter establishing the TBF's, the RLC blocks are correctly received or not.
The Iollowing RNO indicator (which has no correspondent in NPA) gives this ratio in DL:
ln bolh cusos wo culcululo lho rulio o DL (or UL} PLC dulu bylos oncodod in MCSx (sonl lhrough EGPPS} und rolrunsmillod duo lo unucknowlodgomonl o lho MS (or MFS}, luking us rooronco lho ovorull numbor o DL (or UL} PLC dulu bylos sonl on PDTCH und oncodod in MCSx (sonl lhrough EGPPS} in PLC uck modo.
It is possible in UL to take as reIerence Ior the ratio the total number oI UL RLC data bytes retransmitted both in GPRS and EGPRS (instead oI the overall number oI EGPRS RLC data bytes sent on PDTCH); in that case we`ll have the RNO indicator: GPRSULRLCbytesMCSxretransmissingackratio (QRUDTRRBMO)
Again, there is no NPA counterpart. UnIortunately there isn`t also a correspondent indicator Ior DL.
It is important to point out that it is not possible to do the distinction between diIIerent MCS when seeing the retransmission all retrasnmissions in MCS are accounted in only one counter per direction, P20e Ior DL and P21e Ior UL (contrary to GPRS where a diIIerence counter accounts the retransmissions in each one oI the diIIerent CS).
It is interesting to recall the correspondent indicator Ior classical GPRS. In GPRS it is possible to calculate the ratio considering as reIerence: - Overall number oI RLC data bytes transmitted in GPRS mode, corresponding to indicators: o GPRSDLRLCbytesPDTCHCSxretransmissingratio (QRDDTRRBCO); o GPRSULRLCbytesCSxretransmissingackratio (QRUDTRRBCO); o No NPA correspondent. - Overall number oI RLC blocks transmitted in GPRS mode, corresponding to indicators: o GPRSDLRLCblockPDTCHCSxretransackratio (QRDDTRRO); o GPRSULRLCblockPDTCHCSxretransackratio (QRUDTRRO); o No NPA correspondent. - Overall number oI RLC data blocks sent on PDTCH in RLC ack/nack mode: o GPRSDLRLCPDTCHCSxretransmissingackrate (QRDDTRRR); o GPRSULRLCCSxPDTCHretransmissingackrate (QRUDTRRR);
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 29 o Correspond to DLRLCBLOCKRETRANSMISIONRATE and ULRLCBLOCKRETRANSMISIONRATE in NPA.
To do the comparison with EGPRS, only the Iirst two indicators should be considered (QRDDTRRBCO and QRUDTRRBCO).
4.1.3 TBF Normal Release Rate (UL and DL)
This indicator is useIul to have an idea iI the TBF's Iinished well.
Both indicators above are in RNO. Their correspondents in NPA are DLTBFNORMALRELEASERATE and ULTBFNORMALRELEASERATE, Ior DL and UL respectively.
Note that like in the establishment indicators, both EGPRS and GPRS TBF`s are taken into account (again, the name is somewhat deceiving). It is also not possible to distinguish between GPRS and EGPRS TBF.
4.1.4 CS/MCS distribution and useful traffic
There are large possibilities to calculate/present the distribution oI CS and/or MCS blocks. But all ratios/rates/useIul throughput/etc. indicators will be based on the number oI useIul blocks received per (M)CS. The indicators in RNO Ior counting the number oI blocks in each type oI MCS (and so, only EDGE TBF`s are considered) are:
GPRSDLuseIulthroughputMCSack (TRPDDMN), corresponding to the number oI useIul DL RLC blocks sent in RLC acknowledge mode and encoded in CS
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 30 (retransmitted blocks and RLC blocks containing LLC Dummy UI Commands are not taken into account).
GPRSULuseIulthroughputCSack (TRPDUMN), corresponding to the number oI useIul UL RLC blocks sent in RLC acknowledge mode on PDTCH and encoded in MCS (blocks not correctly received are not taken into account).
And Ior GPRS: GPRSDLuseIulthroughputCSack (TRPDDCN), corresponding to the number oI useIul DL RLC blocks sent in RLC acknowledge mode and encoded in CS (retransmitted blocks are not taken into account).
GPRSULuseIulthroughputCSack (TRPDUCN), corresponding to the number oI useIul UL RLC blocks sent in RLC acknowledge mode and encoded in CS (blocks not correctly received are not taken into account).
Note that each one oI these indicators correspond directly to one counter: - P55a to P55d, Ior DL RLC blocks encoded in CS1 to CS4; - P55e to P55m, Ior DL RLC blocks encoded in MCS1 to MCS9; - P57a to P57d, Ior UL RLC blocks encoded in CS1 to CS4; - P57e to P57m, Ior UL RLC blocks encoded in MCS1 to MCS9.
The distribution oI (M)CS blocks is only relative to useIul blocks it is not possible to have it relative to established blocks, dropped blocks, etc there are no counters Ior it.
Note also that all the indicators above have no correspondent in NPA. As said in the beginning oI the section, with the amount oI useIul blocks in each (M)CS it is possible to calculate diIIerent rates. For instance, the Iollowing RNO indicator:
GPRSDLuseIulthroughputMCSx8PSKackratio (TRPDDM8PO). This indicator corresponds to the ratio oI DL useIul blocks in EGPRS mode, RLC ack mode and encoded in 8-PSK (MCS5 to MCS9), relatively to the total number oI useIul blocks in EGPRS mode and RLC ack mode (again, no correspondent in NPA).
Besides this indicator, many other are available, and are easily calculated.
It is also possible to calculate the useIul traIIic. The calculation is done by multiplying the number oI blocks encoded in each (M)CS by its payload.
Regarding traIIic, it also possible to know it at LLC level through the Iollowing RNO indicators: - GPRSDLLLCbytesEGPRSackmode (TGBVCDLEAN), which gives the number oI DL LLC bytes transmitted and acknowledged on established DL TBF in EGPRS mode and RLC ack mode. - GPRSDLLLCbytesEGPRSunackmode (TGBVCDLENN), which gives the number oI DL LLC bytes transmitted and acknowledged on established DL TBF in EGPRS mode and RLC unack mode. - GPRSULLLCbytesEGPRSackmode (TGBVCULEAN), which gives the number oI UL LLC bytes transmitted and acknowledged on established DL TBF in EGPRS mode and RLC ack mode. - GPRSULLLCbytesEGPRSunackmode (TGBVCULENN), which gives the number oI UL LLC bytes transmitted and acknowledged on established DL TBF in EGPRS mode and RLC unack mode.
The above RNO correspond to the Iollowing NPA indicators: - DLLLCBYTESEGPRSACKNUMBER - DLLLCBYTESEGPRSNACKNUMBER - ULLLCBYTESEGPRSACKNUMBER - ULLLCBYTESEGPRSNACKNUMBER
Obviously, the GPRS traIIic is also available, through the corresponding RNO indicators: - GPRSDLLLCbytesGPRSackmode (TGBVCDLGAN); - GPRSDLLLCbytesGPRSunackmode (TGBVCDLGNN); - GPRSULLLCbytesGPRSackmode (TGBVCULGAN); - GPRSULLLCbytesGPRSackmode (TGBVCULGNN);
Which in turn correspond to the NPA indicators DL/ULLLCBYTESGPRSACK/NACKNUMBER.
Finally note that each one oI these indicators corresponds directly to one counter only (p43a to p43d Ior DL, p44a to p44d Ior UL, where p43a and p44a correspond to GPRS and ack mode; p43b and p44b to GPRS and unack mode, p43c and p44c to EGPRS and ack mode, and pp43d and p44d to EGPRS and unack mode).
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 32 4.1.5 Resource Reallocation To analyse the resource reallocation, either the number oI requests is evaluated (in RNO): GPRSDLTBFreallocrequest (TRDRRRQN) GPRSULTBFreallocrequest (TRURRRQN) Only in DL there is a corresponding NPA indicator: DLTBFRESREALLOCREQUESTNUMBER.
Or looking at it per allocation trigger: For DL: GPRSDLTBFreallocTrequest (TRDRRTN) 1, 2, 3 or 4 Corresponding to DLTBFRESREALLOCTREQUESTNUMBER in NPA.
For UL: GPRSULTBFreallocTrequest (TRURRTN) 1, 2, 3 or 4 No correspondent indicator in NPA.
The number oI successes is also available: GPRSDLTBFreallocsuccess (TRDRRSUN) GPRSULTBFreallocsuccess (TRURRSUN) Only in DL there is a corresponding NPA indicator: DLTBFRESREALLOCSUCCESSNUMBER.
Or per reallocation trigger: GPRSDLTBFreallocTnsuccess (QRDRSUN) 1, 2, 3 or 4 (DLTBFRESREALLOCTREQUESTNUMBER in NPA).
GPRSULTBFreallocTnsuccess (QRURSUN) 1, 2, 3 or 4 (No correspondent indicator in NPA).
With all the indicators above, it is easy to calculate the success rate:
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 33 roquosl F_roulloc_ GPPS_DL_T8 succoss F_roulloc_ GPPS_DL_T8 (TPDPPSUP} lo succoss_ru F_roulloc_ GPPS_DL_T8 = in DL; roquosl F_roulloc_ GPPS_UL_T8 succoss F_roulloc_ GPPS_UL_T8 (TPUPPSUP} lo succoss_ru F_roulloc_ GPPS_UL_T8 = in UL.
In both cases, there is no correspondent indicator in NPA.
To calculate the success rate per triggering condition, the Iormula used is similar, but considering the requests and successes per triggering cause.
4.1.6 Ater congestion Two types oI congestion can be considered when talking about Ater congestion.
Either the congestion is at Ater Mux level, meaning that the resources in the Ater Mux Ior PS traIIic are not enough, or the congestion is at DSP level, meaning that the DSP`s inside the GPU don`t have enough processing power Ior all PS traIIic available.
SpeciIic counters exist Ior both congestions: - GPRSGPUAtercongtime (QAGALCTT) Ior congestion at Ater Mux level; - GPRSGPUDSPcongtime (QRGPUCDT) Ior congestion at DSP side.
To obtain the Ater congestion rate, just divide the timers above by the GPRSObservationPeriod, which is normally one hour, or 3600 seconds.
There is also a third indicator that can indicate us that problem exist: - GPRSPDCHdegradedtime (ARPDCUSDGT), corresponding to the cumulative time during which a slave PDCH is degraded due to Ater congestion (that is the number oI GCH channels on the Ater interIace is lower than n Ior a PDCH on a class n TRX). Or in terms oI rate: - GPRSPDCHdegradedtimerate (ARPDCUSDGR), which corresponds to the rate between GPRSPDCHdegradedtime (ARPDCUSDGT) and GPRSPDCHtraIIictime (ARPDCUSBUT) (corresponding to the cumulative time during which the slave PDCHs are established and carry at least one UL or DL TBF established in GPRS mode or EGPRS mode).
Please take note that this last indicator is based on counter P38d. This counter is not reliable beIore B8 MR5. So consider this indicator only iI BSS sub-release is MR5 ed1 onwards.
4.1.7 KPI`s order of importance The Iirst 3 indicators - establishment success, retransmission and normal release rate - are the most important Ior QoS Iollow-up.
On a second approach, it's important to have a metric oI the traIIic exchanged. In terms oI importance, next Iollows the (M)CS distribution and last the resource reallocation.
OI course other indicators (which are not considered as Key indicators) are available and used Ior a deeper technical analysis oI the QoS.
4.2 QoS Iollow-up during a Iield-trial
A Iield trial has usually two main objectives: - Validate the gains brought by EDGE on the Iield; - Help the customer to deIine the best EDGE introduction strategy.
For detailed guidelines on how to perIorm the Iield tests, 0 should be checked.
The QoS Iollow-up during a Iield trial should be a combination oI Iield test results analysis and KPI analysis.
Field tests are normally perIormed with the help oI a series oI possible test tools. Normally the Iollowing tools are used/recommended: - DEUTRIP, Ior generating PS traIIic (i.e. end user PS applications like FTP or WAP). DEUTRIP is a PCS internal tool; - Agilent NITRO, Ior logging all air-interIace related measurements (messages, radio measurements, etc); - DMS Ior post-processing oI air-interIace traces (e.g. Irom Nitro). DMS is a PCS internal tool.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 35 DEUTRIP (which stands Ior Data End User TRaIIic generator & Indicators Provider) is a collection oI modules that automatically generate a diIIerent set oI applications deIined by the user. At the same time, it provides application layers QoS indicators. For more details on DEUTRIP, please reIer to 0 and 0.
Agilent E6474 (usually known as NITRO) is an air-interIace measurements tool. This tool logs all air-inteIace measurements and messages, displaying them in a convenient graphical way. These measurements can come Irom one mobile or more, and/or Irom an external digital scanner. More details regarding Agilent E6474 optimization platIorm can be Iound in Agilent`s website (www.agilent.com).
DMS (which stands Ior Drive Measurement Statistic) is a set oI small tools Ior drive tests statistics. It uses text exports Irom Agilent Nitro (although, other tools can be considered, provided text export is available), and provides three types oI statistics: - GSM Network Statistics; - GSM Call Statistics; - GPRS/EDGE Network Statistics.
For inIormation concerning Agilent NITRO and DMS (Iormerly known as 'GPRSMeasEval), please reIer to 0 and 0.
Obviously, there are alternatives to the tools above cited. For generating PS traIIic, the operator may use operator internal tools, or even no tools at all (it can have someone generating the traIIic manually Alcatel dissuades this option). Make sure oI asking the operator Ior tool documentation when operator internal tools are used.
Ericsson TEMS, Nemo TOM, among others, are alternatives to the Agilent tool. DMS can be used Ior post-processing oI the traces logged with these other tools.
As Iar as post-processing is concerned, there exists some commercial tools alternatives to DMS: Actix Analyser, Nemo Analyse, TEMS Deskcat, etc. As they are commercial tools, their prize is obviously higher than DMS.
Regarding KPI Iollow-up during a pilot, Figure 1 presents quickly the procedure recommended.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 36 8egin dentify cells Field kesults ok End Mo Yes Verify EDGE KFs per cell & per oreo Verify Generol KFs per cell & per oreo All KFs ok All KFs ok Check Optim. Tobles Check counter volidity Check test volidity nform TD Yes Yes Mo Mo
Figuto 1: kocommooJoJ ptocoJuto lot KPl lo||ow-up Jutiog o pi|o| |tio|
The Optimisation tables reIerred in Figure 1 show a quick analysis oI what might be the cause oI degraded KPI`s or bad Iield test results. They can be Iound in Table 10.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 37 Indicator Threshold Investigation Path TBF Establishment Success Rate 95 in DL; 95 in DL; 95 in both directions. VeriIy the distribution oI TBF Establishment Failures per cause: - Failures due to Gb this might indicate problems at Gb link side (no BVC available, which would aIIect the cell this implies that the cell`s operational state is 'disabled); - Failures due to BSS this might indicate a system problem (e.g.: Iaulty board at MFS side). VeriIy also iI CS traIIic is being aIIected due to BSS causes (e.g. TCH assign Iailures due to BSS); - Failures due to congestion: these can also be divided in other subcauses: o Radio congestion this might be due to badly Iunctioning resources (like a TRE not working, with unavailable TS or even 0 available TS), Irom too much CS traIIic, or even Irom too much PS traIIic. o Ater congestion, DSP congestion and CPU congestion either due to lack oI Atermux resources or to not enough DSP and/or CPU processing power. See 5.1 Ior details. Failures due to radio might indicate radio problems (e.g. bad Irequency plan) TBF Retransmission Ratio 5 A high retransmission rate might indicate numerous problems: - The initial MCS could be set too high (deIault value, MCS3, shouldn`t Iace this problem). To veriIy this possibility, check iI retransmissions are being made on the MCS corresponding to the initial MCS; - Radio conditions on the cell are bad, and so are causing many blocks to be not decoded; - Other causes.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 38 Indicator Threshold Investigation Path TBF Normal Release Rate 95 note that: -in DL, it is expected a lower normal release rate in UL most oI the traIIic is signalling, and so is short and coded in MCS1
-the global normal release rate will be closer to the UL value there are many more UL TBFs due to signalling VeriIy how TBF`s are being abnormally released. The Iollowing causes are possible: - RLC blocks are being lost during the GPRS connection - RLC blocks are being retransmitted too many times check the TBF Retransmission Ratio KPI; - Connection drop is high this might be due to 3 other sub-causes: o System drop indicating a system problem. VeriIy other KPI (GSM included); o Gb drop indicating problem on Gb interIace; Radio drop this might be due either to real radio drops or to acceptable releases (due to cell reselection; suspend/resume procedure; PDCH pre- emption). MCS Distribution - No clear threshold exists Ior this KPI, because it depends a lot Irom the conIiguration oI the cell and the end user traIIic (short (e.g. WAP) or long (e.g. FTP) transIers). Nevertheless some very useIul indications might be taken Irom it: - II MCS2 ratio is very high, this might indicate that extra GCH might be badly conIigured. In this case, veriIy iI any particular cell is having maximum MCS2, as this might indicate a Iaulty secondary Abis link; - II MCS5 ratio is very high, this might indicate that there is some Ater congestion see Ater congestion KPI analysis. Resource Reallocation - In B8 and with EDGE, it is natural that some re- allocation triggers will occur very Irequently (T2 and T4, Ior instance). Other triggers should not occur so Irequently: - a high Irequency oI occurrence oI T1 might indicate that the cell is sub-dimensioned to accommodate all CS and PS traIIic; - a high Irequency oI occurrence oI T3 might indicate that there is too much PS traIIic which in itselI is not a bad thing, but in this case congestion indicators should be checked.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 39 Indicator Threshold Investigation Path Ater Congestion 0.2 Ior Ater Mux congestion is acceptable provided it occurs only occasionally For more details concerning Ater congestion issues, please check 5.1. To6|o 10: Cp|imiso|ioo |o6|os - tocommooJoJ iovos|igo|ioo po|hs togotJiog |ho moio KPls
Finally, Table 11 and Table 12 present the FTP results expected during the Iield trials, based on the results obtained so Iar.
Some notes are due, though: - The presented intervals are in Iact the intervals Iound in the results Irom various Iield trials and pilots; - DiIIerent results considering same TRX class, same Irequency hopping scheme were averaged. All values were considered independently iI resegmentation was enabled or not, and iI PSPREFBCCH was set to true or not. - For DL results, only tests with Iilesize oI 1 MB, 2 MB or 3 MB were considered. In UL Iilesizes oI 200kB and 500 kB were considered. - TraIIic conditions in all cases were such not to aIIect EDGE tests. - DiIIerent MS types were used: Nokia 6220 (21 in EDGE), Nokia 6230 (41 or 32 in EDGE) and Philips DragonIly (41 or 32 in EDGE) hence the option to present results in per RTS throughput. - The objective oI these tables is to give a quick idea oI what kind oI results to expect. Further details can be Iound in the EGPRSQoSdatabase (a intranet version is currently being prepared by PCS); - Note that in UL, as 8PSK is not available in B8 release, it is indiIIerent to have TRX class 2 or a superior TRX class; - Both RxLEv and MeanBEP measurements usually deIine radio conditions. UnIortunately this was not the case in some pilots. Also, there were some cases where the radio conditions were perhaps badly deIined (normal radio conditions that are in Iact poor, Ior instance). - The deIinition oI radio conditions can be Iound in Table 13.
As seen beIore, other services can be tested during a Iield trial. The results depend too much on the deIinition oI the test.
WEB test results will depend a lot on the used web page structure a simple text page with Iew big images will obviously give a better throughput than a page with many small elements.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 41 WAP tests are seldom made due to two main reasons: on one hand, the results depend a lot on the page structure and size, and rarely it is possible to have a Iixed WAP page (normally, commercial pages are used); on the other hand, a WAP session uses many short connections, so the improvements oI EDGE are less visible.
MMS tests would normally show improvements in EDGE. Again, they are very dependent on the message considered.
Video streaming tests are also seldom made. There is still not a clear way oI perIorming these tests, so they are normally avoided.
Last but not least, PING tests are regularly made to prove non-regression in EDGE. Notice has to be made to the Iact that some mobiles may be buggy, and so present bad results. See Iault report 3BK A25FBR 153113 Ior more details.
To see all results Irom previous experimentations, Ior all services, please reIer to the EGPRSQoSdatabase.
4.3 QoS Iollow-up on a regular network
QoS Iollow-up on a regular network is somewhat diIIerent Irom the Iollow-up done during a Iield trial. Namely: - There is not a speciIic (set oI) cell(s) to look at the traIIic is not speciIically generated on a speciIic cell, but comes Irom end-user usage, and appears scattered all over the network; - There is no control on what the end-users will do the traIIic mix is not controllable, as well as there is no guarantee regarding which mobiles are being used; - While the QoS Iollow-up during a Iield trial is done only during the duration oI the trial itselI (considering some extra week(s) Ior reIerence, oI course), the QoS on a regular network must be checked regularly.
QoS Iollow-up in an EDGE network is not very diIIerent Irom the Iollow-up done in a GSM/GPRS network. The main diIIerences are at KPI level in an EDGE network, the KPI`s to be observed are diIIerent (see 4.1 Ior deIinitions). To be Iair there are Iew EDGE speciIic indicators.
Figure 2 presents quickly the recommended procedure. The optimisation tables reIerred therein correspond to Table 10.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 42 8egin Verify Generol KFs Generol KFs ok G5M/GFk5 Follow-up process Verify EDGE KFs EDGE KFs ok Verify bod KF(s} per Metwork Element dentify Foulty Metwork Element Air-itf meos. possible Do oir meosurements Meos. ok Follow problem during some more time Froblem persists Check Optimizotion Tobles nform TD End Mo Yes Yes Mo Mo Yes Yes Mo Yes Mo
Figuto 2: kocommooJoJ ptocoJuto lot KPl lo||ow-up io o togu|ot oo|wotk
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 43 5 Known Issues 5.1 Ater congestion
As seen in 2.4, during the Ater congestion state the GCH Reduction Mechanism is active.
This GCH Reduction Mechanism only aIIects the establishment oI GPRS TBF`s. But experience on the Iield showed that EGPRS TBF`s are also aIIected!
II an EGPRS TBF is established while a GPRS TBF is still active, or is established less than TPDCHINACTIVITY seconds aIter the end oI the GPRS TBF, it will be aIIected. In both cases the EGPRS TBF will be established on the PDCH already established Ior the GPRS TBF, hence only having 1 or 2 GCH`s associated. Please note that iI EGPRS and GPRS TBF`s are sharing the same TBF`s, it will be the GPRS TBF that will be re- allocated.
This use oI 'limited PDCH`s will have the immediate impact oI limiting the maximum MCS to MCS5, whatever the class considered (obviously only TRX class 3 or above are aIIected). Note also that iI MAXGPRSCS CS2, the maximum will be logically limited to MCS2.
What Iield tests have shown, and it is important to retain, is that even iI there is a correct Ater dimensioning (enough Ater Mux links), Ater congestion at DSP level is possible!
DSP processing is limited and easily the maximum can be reached. Furthermore there are some situations where there is a waste oI resources.
The most Irequent situation oI waste oI resources is linked to the GPRS Resource Anticipation mechanism.
When establishing an UL TBF one-phase access, the MFS may trigger the GPRS Resource anticipation (depending on parameterisation, oI course) meaning that some TS will be reserved Ior PDCH`s in DL anticipating some upcoming traIIic in DL. The mobile class is not known at this time, and so the MFS will consider the GPRSMULTISLOTDEFAULTVALUE as mobile class.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 44 Note that the MFS is not sure that there will be any DL traIIic, and so considering a class n TRX, and considering the deIault value oI GPRSMULTISLOTDEFAULTVALUE, that is 8 (meaning 41 mobile), there could be a potential waste oI 4*(n-1) GCH.
The way to detect the Ater congestion is described in 4.1.6.
Actually two workarounds are known. Either more GPU`s or Atermux links are added, or parameters are changed.
II Ater congestion is at Ater Mux level, there is no way to escape the need to add more Ater Mux capacity (i.e. new link or dedicated link or increased granularity). II Ater congestion is at DSP level, adding GPU`s might solve the problem, but it is neither a practical solution nor an inexpensive one.
In the case oI Ater congestion at DSP level, parameterisation changing is preIerred. The Iollowing parameters can be changed: - GPRSMULTISLOTDEFAULTVALUE to 1(meaning 11 MS) to avoid wasting resources; - TPDCHINACTIVITY to 3s and TPDCHINACTIVITYLAST to 20s to liberate resources sooner. Note that in the case oI congestion at Ater Mux level, the parameterisation changes will not solve (or improve) the problem.
The above parameterisation has theoretically some drawbacks. II during the DL anticipation process the traIIic to be exchanged is indeed data traIIic, and the mobile is oI an higher class (thus needing more than 1 DL TS), it is probable that new resources need to be allocated, hence decreasing the speed gain brought by the DL anticipation mechanism. On the other hand, decreasing the TPDCHINACTIVITY|LAST| might also lead to an increase oI the average time to establish TBF`s, as the delayed DL TBF release mechanism will be slightly less eIIective.
Tests have been made to study the impact oI these parameter changes, and results have shown that impact is negligible.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 45 5.2 EDGE perIormance degradation with Frequency Hopping
In all pilot and Iield trials made so Iar, EDGE perIormance is worse when Frequency Hopping is used.
As a consequence oI these results, a test campaign was launched to quantiIy and veriIy the degradation oI hopping in EDGE perIormance.
First tests made in optimum laboratory conditions (no interIerence, no Iading, etc.) showed that Frequency Hopping and EDGE work well together.
The tests made with the Nokia 6220 showed that in Lab tests, increasing the number oI Ireq. in hopping law makes meanBEP and throughput decrease. But relation between number oI Irequencies and EDGE perIormance is not clear in tests made later! In general, it was seen that Irequency hopping does impact negatively EDGE perIormance.
Tests made with the Nokia 6230 showed that there was negligible perIormance degradation with Irequency hopping.
No clear conclusion can be taken Irom the results so Iar.
Further tests will be perIormed, considering diIIerent mobiles (including the Philips DrangonIly), but considering always the same conditions (radio conditions, number oI TS, etc.).
5.3 APD issues
As seen in 2.2, 3GPP specs give some constraints regarding APD values when EDGE is carried in the BCCH-TRX.
To avoid having a too great APD, it might be needed to decrease the GMSK output power (through BSTXPWRMAX parameter), hence decreasing the cell coverage this might be inadmissible to the operator. Other alternative to respect the 3GPP specs might be to declare TS7 as SDCCH, avoiding the tighter constraints in this TS (2 dB on TS7 vs. 4 dB in all the other TS).
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 46 But again, this option might be impracticable when Cell Broadcast Ieature is activated (CB must use a SDCCH channel on TS0 to TS3 oI the BCCH-TRX).
As can be easily calculated Irom Table 2, the APD will easily be bigger than 4 dB. So to respect the 3GPP recommendations might be an issue.
AIter some trials where the recommendation was disregarded it was seen that no negative impact was Iound. Also, the speciIications Irom 3GPP are a little vague regarding iI they are really mandatory or only recommended.
The reason behind the 3GPP recommendation is to avoid an issue (a MS does not reselect/is not handed over to the best neighbour cell) that occurs only in older mobiles. In theory, the MS is time-synchronised to the best 6 neighbors (in order to be able to read SI3), and so the MS should make RxLev measurements on TS0 oI those neighbour cells. The RSSI in other TS should not be considered.
As the Iield tests didn`t show any negative impact, which just conIirms that only older mobiles are prone to have problems, we can say that: It is possible to use normally the BCCH TRX Ior EDGE without any re-conIiguration or output power modiIication.
5.4 Fault Reports that may impact EDGE perIormance
5.4.1 Bad PING performance in some Nokia mobiles During the EDGE tests made with Orange France (ZE B8), it was detected that the ping perIormance Ior small data ping sizes and large intervals in the Nokia 6230 was degraded. A Iault report 3BK A25FBR 153113 was open.
AIter some talks with Alcatel, Nokia has admitted that the problem is inside its mobile SW. IOT tests made on 19/Jan showed that ping perIormance has improved with new Iirmware version (5.28). The Fault Report linked to this problem was closed during w505.
There are no more details on what was really the problem behind the ping perIormance degradation, but problem is solved with the new Iirmware version.
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 47 When doing PING tests with the Nokia 6230, take care oI the Iirmware version oI the mobile older versions will surely give bad results.
5.4.2 Wrong RLC block coding when retransmitting with MCS3 (with padding) It was seen in the platIorm that whenever there was retransmission oI some block initially coded in MCS8 and re-segmentation was enabled, some problems could occur. When the retransmission is done using MCS3 (with padding) Ior the coding oI the retransmitted blocks, these blocks are incorrectly coded, and so the mobile can not decode them, which impacts the perIormance.
This problem has originated the Fault Report 3BKA20FBR154745, which is actually closed. This problem was solved through Change Request 3BKA20CBR154875, available Irom MR6 Iorward.
The recommendation Irom TD is in the mean time to disable re-segmentation (which is in line with the deIault value Ior ENFULLIR).
Alcatel File Reference Date Edition Page EDGE Optimisation Guide ed1.doc 04/05/05 2.0 48 6 Final Comments
The present document aimed at giving a general view on how to Iollow-up the Quality oI Service and how to optimise an EDGE network.
All assumptions made take into account the knowledge obtained in the diIIerent EDGE networks and in the pilots made so Iar.
Obviously, there is a risk that the document becomes outdated. II this occurs, a new edition oI the document will be prepared so please check that you have the latest version.