Transcript RNO Wind
RNO Wind Part III Confidential 1 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Part III - Content Call Setup Time UL Interference PS Utilization Cell Reselection Confidential 2 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Call Setup Time Confidential 3 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Call setup Time – Preamble PRACH • During drive testing can be noted that there are call setup failures where the network does not seem to respond to RRC Connection Requests with RRC Connection Setup –message.These are problems due to the spiky UL noise and due to that the power ramping is not aggressive enough to provide high enough Tx power for the terminal during open loop PC PowerOffsetLastPreamblePRACHmessage L1ACK/AICH PtxAICH Downlink / BS PowerRampStepPRACHpreamble PRACHRequiredReceivedCI Preamble 1 UEtxPowerMaxPRACH …. …. Preamble n RACH Message part Uplink / UE PRACH_preamble_retrans: The maximum number of preambles allowed in one preamble ramping cycle RACH_tx_Max: # of preamble power ramping cycles that can be done before RACH transmission failure is reported, Note: The power ramp-up process will continue until 1) A positive or negative AI is received from the network 2) RACH_tx_MAX value is reached 3) UE reaches UEtxPowerMaxPRACH value Confidential 4 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Call setup Time – Preamble PRACH The parameters affecting to open loop power control are, in brackets are the recommended values: • PRACH_preamble_retrans (7) • RACH_tx_Max (16) • PowerOffsetLastPreamblePRACHmessage (2 dB) • PowerRampStepPRACHpreamble (2dB) The PRACHRequiredReceivedCI (-20dB) allow to calculate the UEpower for the fist preambleas in the following: Ptx = CPICHtransmissionPower-RSCP(CPICH) +RSSI(BS) + PRACHRequiredReceivedCI (-20dB) Example: CPICH = 33dBm (Parameter per Node-B) RSCP = -80dBm (Measured by UE) RSSI = -85 dBm UL_Required_C/I = -25 dB (Parameter per Node-B) UE PRACH First Preamble Power = 33 dBm – (-80 dBm) + (-85 dBm) + (-25 dB) = 8 dBm The parameter PRACHRequiredReceivedCI can be set to -18…-20dB instead of the default -25dB (typically 20dB is enough) Confidential 5 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Call setup Time – Preamble PRACH Typical improvement passing from -25dB to -20dB: PRACH req. C/I = -20dB PRACH req. C/I = -25dB Clear improvement in number of needed RRC Connection Request messages per call. For –20dB 100% of established calls are setup with only 1 RRC Connection Request message 100% 88% 100% % 80% 60% 40% 20% 0% 2% 0% 5% 2 3 0% 6% 0% 1 4 # RRC Connection Request Messages per call setup PRACH req. C/I = -25dB Clear improvement number of sent preambles per RRC Connection Request for –20dB case. For –20dB 50% of cases the needed number of preambles is <=4 where as for –25dB it is ~6.5 PRACH req. C/I = -20dB 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% There should be significant improvement also for call setup delay 1 2 3 4 5 Confidential 6 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI 6 7 8 Call setup Time – Preamble PRACH The average number of acknowledged PRACH preambles during the RRI period can be calculated based on the KPI below M1000C176SUM_RACH_ACK_P REAMBLES M1000C177DENOM_RACH_ACK_P REAMBLES RACH load due to preamble can then be calculated by dividing the above further by the max number preambles can be received during RRI • For example if RRI period is 200ms the are 10 20ms RACH frames and in each 20ms RACH frame there are 15 RACH sub slots within each it is possible to receive and decode max 4 preambles -> therefore in 200ms it is possible to receive 15*4*10=600 preambles M1000C176SUM_RACH_ACK_PREAMBLES /600*100% M1000C177DENOM_RACH_ACK_PREAMBLES Confidential 7 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Call Setup Time – SRB Rate Why 13.6kbit/s? Use of 13.6 kbit/s SRB also in highly loaded networks Decreased setup times (PDP context activation minimum 0.7s lower) Improved Iub efficiency Typical improvement passing from 3.4 to 13.6 7 Nokia RAN1.5 (3.4 kbps) + M11 6 Nokia RAN04 (13.6 kbps) + M12 Nokia RAN target Seconds 5 4 3 2 1 0 3G-3G CS call setup Confidential 8 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI PS call setup DCH allocation Call setup Time – KPI In RN2.2 the following counters are available to monitor the Call Setup Time RRC Setup Time M1001C221/M1001C222 RAB Setup Time M1001C223 / M1001C224 M1001C235 / M1001C236 for CS for DATA BACKGR In detail we have: M1001C221 - SUM OF RRC SETUP TIMES Sum of RRC setup times. This counter divided by the DENOMINATOR - M1001C222 gives the average RRC setup time. RRC setup time is defined as the time between the RRC: RRC CONNECTION REQUEST message and the RRC: RRC CONNECTION SETUP COMPLETE message. M1001C223/235 - SUM OF RAB SETUP TIMES FOR CS VOICE/FOR DATA BACKGR Sum of RAB setup times. This counter divided by the DENOMINATOR - M1001C224/236 gives the average RAB setup time. RAB setup time is defined as the time between the RANAP: RAB ASSIGNMENT REQUEST and RANAP: RAB ASSIGNMENT RESPONSE messages during RAB establishment. Confidential 9 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Call setup Time – Annex1 RACH/FACH MO-UE Mobile-to-mobile CS call setup on common channels Delay RRC connection request UE RNC 0 RRC connection setup RNC UE 40 RRC connection setup completeUE RNC 100 CM service request UE CS 200 Security mode command RNC UE 100 Security mode complete UE RNC 200 Setup UE CS 300 Call proceeding CS UE 100 Radio bearer setup RNC UE 100 Radio bearer setup complete UE RNC 300 Alerting CS UE MT-UE Cumulative 0 Parallel RB setup for MO40 UE and paging of MT-UE 140 (CS core feature) 340 440 640 940 1040 Paging RNC 1140 RRC connection request UE 1440 RRC connection setup RNC RRC connection setup complete UE Paging response UE Security mode command RNC Security mode complete UE Setup CS Call confirmed UE Radio bearer setup RNC Radio bearer setup complete UE 250 2980 CS UE RNC UE RNC CS UE RNC UE CS UE RNC UE 400 50 40 100 100 100 200 300 100 100 300 250 <3.0 s mobile-to-mobile AMR call setup time Confidential 10 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI 1340 1390 1430 1530 1630 1730 1930 2230 2330 2430 2730 2980 RACH/FACH Average paging delay of 320 ms assumed (640 ms paging cycle) Typical value for CS Call Setup Time Call setup Time – Annex2 Typical value for PS Call Setup Time Common channels used for setup to avoid slow synchronized reconfigurations later Delay RNC UE RNC PC UE RNC PC UE RNC UE 0 40 100 200 100 200 250 150 300 200 Cumulative 0 40 140 340 440 640 890 1040 1340 1540 <1.6 s PS call setup time Confidential 11 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI RACH/FACH RRC + PDP on common channels RRC connection request UE RRC connection setup RNC RRC connection setup complete UE GPRS service request UE Security mode command RNC Security mode complete UE PDP context activation request UE Radio bearer setup RNC Radio bearer setup complete UE PDP context activation accept PC Parallel RB setup and RL/AAL2 setups (or prereserved Radio links) Initial bit rate DCH allocated directly together with SRB UL Interference Confidential 12 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI What’s Interference? Overload Area PrxTarget [dB] + PrxOffset [dB] Prx Target [dB] Marginal Load Area Feasible Load Area LRT UnloadedRT and LNRT UnloadedNRT Unloaded Area Own cell load factor Any working point turned off from the expected load curve can be considered as interference. Interference can be internal or external. Internal interference can be caused by not appropriate dimensioning, planning or commissioning External is usually referred to mobile or other RF sources Confidential 13 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Load vs. Power Class1_Prx/Load 45000 40000 Rel. Amplitude 35000 30000 25000 20000 15000 10000 5000 0 -5000 0 50 100 150 200 250 WBTS ave_lrt_class_1 ave_lnrt_class_1 ave_prxtot_class_1 Typical mismatch among load and Power can be easily found in a live network. Above is reported a qualitative behaviour in class_1 power for some Wind WBTSs that are experiencing a 1<rt_load<2 (rt_load relative value from 0 to 4) and the related nrt_load and Prx_power. The nrt load added to rt can not give sense of the Prx spike Confidential 14 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI NSN Load Areas & Class of Power PrxTarget [dB] + PrxOffset [dB] Overload Area Class4 Marginal Load Area Class3 Feasible Load Area_2 Class2 Prx Target [dB] PrxTarget [dB] - PrxOffset [dB] Feasible Load Area_1 Class1 LRT UnloadedRT and LNRT UnloadedNRT Unloaded Area Class0 Own cell load factor CLASS AREA CLASS 0 Unloaded (Lrt=<UnloadedRT) AND (Lnrt=<UnloadedNRT) CLASS 1 Feasible_Load_Area_1 (PrxTarget -PrxOffset >= PrxTotal ) AND ((Lrt>UnloadedRT) OR (Lnrt>UnloadedNRT)) CLASS 2 Feasible_Load_Area_2 (PrxTarget > PrxTotal > PrxTarget -PrxOffset) AND ((Lrt>=UnloadedRT) OR (Lnrt>= UnloadedNRT)) CLASS 3 Marginal_Load_Area (PrxTarget + PrxOffset > PrxTotal >=PrxTarget) AND ((Lrt>UnloadedRT) OR(Lnrt> UnloadedNRT)) CLASS 4 Overload_Area (PrxTotal >= PrxTarget + PrxOffset) AND ((Lrt>UnloadedRT) OR (Lnrt>UnloadedNRT)) Confidential 15 INCREMENTED IF © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI UL Interferece Detection Method Different approach can be applied to detect UL interference. Mainly we have: - Field measurement - Counters Analysis Using the Counters Analysis approach dedicate counters are available for UL Interfernce detection as MAXPrxNoise and MINPrxNoise (M1000C12 and M1000C13) The UL interference severity can be estimated by analysing: MAXPrxNoise – MINPrxNoise, but these counters are incremented only when cell is unloaded. Here we propose a line for a method that approximately return the WBTS interfered. The method takes the basis from the autotuning algorithm and use the value of Prx returned to detect the interfered cell. The first step is the localization of reference point for each class Then different kind of statistical model can be applied for evaluating the drawn from them Finally a w.w.w concept is used to derive information from space and time recurrence Some help could come from counters that trigger downgrade or release bocause of interference (e.g. M1000C147RB_DOWNGR_DUE_PBS_INTERF M1000C159RB_RELEASE_DUE_PBS_INTERF if PBS is enabled) Confidential 16 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Prx Autotuning The auto-tuning algorithm moves the reference point of the load curve and this means that all the areas can be shifted up and down during the day this means that a certain value of PrxTotal (which is measured by the bts) may trigger different areas during the day. For example the sample 4 triggers in the first case the class 2 while in the second case the class 1, but it’s the same value of power! Main idea is to use this gap to detect interference t1 t0 Prx Target_t0 [dB] Prx Target_t1 [dB] Overload Area Marginal Load Area Feasible Load Area 2 4 4 Feasible Load Area 1 Unloaded Area Time Confidential 17 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Class Power Reference Point It is not an easy task to find the expected value of Prx in each class. Different masking effect are present either for the granularity of the measurement available that are not appropriate for this kind of analysis or for the inherent difficulty in evaluating the real load experienced. Here a shot for class1 considering the stay time in the class is attempted The spike are more accentuated for low permanence and diluited for the high one Permanence in Class1>45min 0.35 Prx Displacement 0.25 0.2 0.15 Permanence in Class1<15min 0.1 0.05 1.4 0 1.2 1 3 5 7 Prx Displacement 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 WBTS An average can be attempted filtering off the spike and the default value Prx Rel. Amplitude Prx Rel. Amplitude 0.3 1 0.8 0.6 0.4 0.2 0 1 74 147 220 293 366 439 512 585 658 731 804 877 950 1023 1096 1169 1242 1315 WBTS Confidential 18 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Power Class Distribution Function Probable Interfered WCEL Probable Interfered Prx_Dist. function WCEL 1 Rel. Amplitude 0.8 c2 0.6 c1 c0 c3 0.4 c4 0.2 0 0 200 400 600 800 1000 1200 1400 1600 1800 WBTS Here a Prx Distribution over the all WCELs is presented. Typical value of the reference point are represented individuating areas where interference can be detected. The different shape of the curve of the Feasible_Load_Area_2 and the Marginal_Load_Area_2 respect to the Class_0, Class_1 and Class_4 seems due to the different behaviour of the algorithm. The step visible in C2 and C3 could be due to the strict margin in term of Power Budget to react to the load increase. The overshoot of the C0 curve over the C1 is due to to the different triggering condition that for C0 is load based instead of Power Level driven. Finally C1 having a greater budget maintain a smoother shape. Confidential 19 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI F_time W.W.W. Approach stable interference for a adjacent cluster of cell + Commissioning / Fixed Ext. Source Dimensioning - + F_space Adj missing Mobile Ext. Source periodical spot interference - A single interfernce event can not raise any relevant bother. A statistical analysis is needed. The Who? When? Where? approach is used to derive information and troubleshoot the probable interferer source. The space-time diagram has to be intended as a recurrence indicator for the interference event. In the left side of the F_space axis are reported occurences not adjoined in space. Same concept for F_time. Confidential 20 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Class 0 Class0 can act as the third dimension of the WWW Approach diagram. Considering Class0 as the unlaoded class in the sense that the unloaded limit for RT and NRT (1% and 2% respectively) is not exceeded the interference detection in this class can have two advantages: a) b) More interference sentivity because of low load Easier discrimination between internal and external interference The first point is assured by the triggering condition and can be strenghtened superimposing a second condition over the load. Imposing the LoadRT = 0 and LoadNRT = 0 we have more reliable result for interference This condition triggered mainly during the nigh-time returns the possibility to have an easier troubleshooting Confidential 21 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI PS Utilization Confidential 22 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Traffic Mix KPI E The KPI provides an indication of the percentage of CS voice, CS data, PS data RAB establishment attempts relative to the total number of RAB establishment attempts The KPI is meaningful for cluster/cell level and on day/hour basis. Same KPI can be obtained using RAB ACC COMP These KPI are intended to provide a high level indication of the traffic profile loading the network: • CS_VOICE Traffic Mix • CS_CONV • CS_STREA 16% • PS_CONV R• A BPS_STREA _ ST P _ A T T _ C S_ C O N V R A B _ ST P _ A T T _ C S_ ST R E A R A B _ ST P _ A Voice T T _ P S_ C O N V R A B _ ST P _ A T T _ P S_ ST R E A R A B _ Data Conv 51% • PS_INTER PS Inter 32% PS Backg • PS_BACKG 1% Example for CS_VOICE: RAB _ STP _ CS _ VOICE RAB _ STP _ CS _ VOICE RAB _ STP _ CS _ CONV RAB _ STP _ CS _ STREA RAB _ STP _ PS _ INTER RAB _ STP _ PS _ BACKG To take into consideration that PS might cause many attempts in each call another option is to consider the duration counters! Confidential 23 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Traffic Mix KPI For each traffic class there are counters for RAB Holding time (incremented when the RAB is released only on the cell that was the reference when the RAB is released) AVG _ RAB _ HLD _ TM _ PS _ INTER / 100( s) DENOM _ RAB _ HLD _ TM _ PS _ INTER For each Traffic Class If a distribution on cell level is required the RAB_HOLD_TIME_IN_REF_CELL can be used For NRT traffic classes (inter and backg) there are also counters for DCH Holding time (incremented when the RAB is released only on the cell that was the reference when the RAB is released) AVG _ DCH _ HLD _ TM _ PS _ INTER / 100( s) DENOM _ DCH _ HLD _ TM _ PS _ INTER DCH Holding Time [s] RAB Holding Time [s] 20 20 40 60 80 100 120 Confidential 24 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Only for NRT Traffic Class 40 60 80 100 120 140 160 140 180 200 More 180 160 200 More CELL _ DCH UE CELL _ FACH CELL_DCH state InactivityTimerUL(DL)DCH CELL_FAC H state RLC buffer payload (transport channel traffic volume) From Cell_DCH to Cell_FACH After the inactivity timer expires the RRC radio bearer reconfiguration–procedure is performed. RRC sends an RRC: RADIO BEARER RECONFIGURATION message to the UE. UE acknowledges by sending the RRC: RADIO BEARER RECONFIGURATION COMPLETE – message to the RRC signaling entity of the RNC which starts L2 reconfiguration (as well as PS is informed about the cell state change). Radio link and AAL2 resources are then released and UE is changed to CELL_FACH state. In case the UE is having RT RB which has become inactive and at the same time it is having inactive NRT RB then RADIO BEARER RELEASE procedure is used (instead of RADIO BEARER RECONFIGURATION). Confidential 25 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI CELL _ FACH UE CELL _ DCH CELL_D CH state CELL_FA CH state RLC buffer payload (transport channel traffic volume) From Cell_FACH to Cell_DCH TrafVolThresholdDL(UL) High TrafVolThresholdDL(UL)L ow (WCEL) In uplink direction the need for the capacity is detected by the MAC of UE. UE requests dedicated capacity by sending an RRC: MEASUREMENT REPORT message on RACH to the RRC signaling entity of RNC After the procedure, data transmission on DCH can begin and UE is in CELL_DCH state. In downlink direction the capacity need is detected by the UE MAC entity of RNC. PS requests the RRC signaling entity of RNC to start transport channel reconfiguration –procedure The RRC signaling entity sends an RRC: TRANSPORT CHANNEL RECONFIGURATION message to the UE on FACH, which is acknowledged with an RRC: TRANSPORT CHANNEL RECONFIGURATION COMPLETE After the procedure, data transmission on DCH can begin and UE is in CELL_DCH state. Confidential 26 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Cell-DCH/Cell-FACH KPIs DCH Time % 10% Percentage of time in cell dch: 20% 30% AVG _ DCH _ HLD _ TM _ PS _ INTER 100% AVG _ RAB _ HLD _ TM _ PS _ INTER 40% 50% 60% 70% 80% 90% Similar KPI giving the ratio between FACH and DCH can be constructed starting from 100% More M1006C90 SUM OF UE OPERATING TIME IN CELL_FACH M1006C87 SUM OF UE OPERATING TIME IN CELL_DCH Dividing per the number of UE is possible to have average time for user: CELL_FACH CELL_DCH CELL_FACH M1006C90 SUM OF UE OPERATING TIME IN CELL_FACH/M1006C92 NUM OF UE MEASURED IN CELL_FACH Uplink DCH M1006C87 SUM OF UE OPERATING TIME IN CELL_DCH / M1006C89 NUM OF UE MEASURED IN CELL_DCH Downlink DCH The number of transition can be monitored as well: M1006C45 CELL DCH STATE TO CELL FACH NRT RB data transfer active NRT RB inactivity timer running Confidential 27 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI M1006C46 CELL FACH STATE TO CELL DCH Measuring the RACH/FACH Channel The RACH channel average throughput for both data and signaling can be measured by the following KPI M1000C60AVE_RACH_THROUGHPUT /1000kbps M1000C61RACH_DENOM_3 The FACH Total throughput means all the user related data (FACH-u) and signalling (FACH-c) for a SCCPCH including PCH can be measured by the follwing KPI M1000C66AVE_FACH_U_TOT_TPUT_SCCP_PCH bit/s M1000C67FACH_USER_TOT_TPUT_DENOM_0 Load KPI are available as well using the following counters M1000C64 AVE SCCPCH INC PCH LOAD M1000C65 SCCPCH LOAD DENOM 0 When the throughput approach the maximum allowed or the load the 100% for the actual configuration a parameter tuning to avoid the starvation in CCH or an expansion of RACH and FACH channel is required. The decision outcomes from different input: DCH resources available Marketing Strategy Confidential 28 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Cell Reselection Confidential 29 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Cell Reselection 2G -> 3G Start measurement GSM MS starts WCDMA measurements if : RLA_C< F(Qsearch_I) for 0<Qsearch_I<=7 or RLA_C> F(Qsearch_I) for 7<Qsearch_I<=15 If, for suitable UMTS cell & for a period of 5 s: CPICH RSCP > RLA_C + FDD_Qoffset and CPICH Ec/No FDD_Qmin Confidential 30 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI WCDMA cell reselection 2G -> 3G Measurement Depending on operator´s 2G – 3G interworking strategy parameter Q_search_I should planned accordingly. In the best case, 3G cell measurements are possible when RLA_C level < –74 dBm In the best case, 3G cell measurements are restricted to the condition: RLA_C level > –78 dBm GSM GSM 3G 3G 3G GSM Configuration 1 RLA_C< F(Qsearch_I) ( 0<Qsearch_I<=6 ) Confidential 31 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Configuration 2 RLA_C> F(Qsearch_I) ( 7<Qsearch_I<=15 ) Configuration 3 RLA_C< (always). (Qsearch_I=7) 2G -> 3G Cell Re-selection Parameters Qsearch_I and Qsearch_P define the threshold for non-GPRS/GPRS (respectively) capable UEs to measure 3G neighbour cells when a running average of the received downlink signal level (RLA_C) of the serving cell below (0-7) or above (8-15) the threshold Value 0 1 … 6 7 8 9 10 … 14 15 dBm -98 -94 … -74 Always -78 -74 -70 … -54 Never If RLA_C > -70 UE starts 3G measurements UE always measures 3G cells If RLA_C < -94 UE starts 3G measurements FDD_Qoffset and FDD_GPRS_Offset the non-GPRS/GPRS (respectively) capable UEs add this offset to the RLA_C of the GSM cells. After that the UE compares the measured RSCP values of 3G cells with signal levels of the GSM cells Value 0 1 2 3 … 8 … 14 15 dBm Always -28 -24 -20 … 0 … 24 28 Always select irrespective of RSCP value Reselect in case RSCP > GSM RXLev (RLA_C) +28dB FDD_Qmin, defines minimum Ec/No threshold that a 3G cell must exceed, in order the UE makes a cell reselection from 2G to 3G. Confidential 32 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Cell Re-selection Example-Weaker WCDMA Non GPRS case RSCP/ RLA_C Ec/No Cell re-selection to WCDMA RLA_C Serving GSM Cell Qsearch_I=0 (-98 dBm) FDD_Qoffset =6 (-8 dB) Measurements starts (serving cell) Neighbour WCDMA Cell FDD_Qmin=0 (-20 dB) RSCP Ec/N0 Minimum Quality Requirement for WCDMA t 5 sec. Confidential 33 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Cell Re-selection Example-Weaker WCDMA GPRS case RSCP/ RLA_C Ec/No RLA_P Cell re-selection to WCDMA FDD_GPRS_Qoffset =10 (8 dB) Serving GSM Cell (Best) Qsearch_P=0 (-98 dBm) RSCP Measurements starts (serving cell) FDD_Qmin =-20 dB Ec/N0 Neighbour WCDMA Cell Minimum Quality Requirement for WCDMA t 5 sec. Confidential 34 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Cell Reselection 3G -> 2G Whilst camping in a 3G cell the UE performs intra-frequency, inter-frequency, and inter-system measurements based on the measured CPICH EcNo. Serving cell parameters Sintrasearch, Sintersearch and SsearchRAT are compared with Squal (CPICH Ec/No – Qqualmin) in S-criteria for cell re-selection 1 - None (Squal > Sintrasearch ) 2 - WCDMA intra-frequency (Sintersearch < Squal Sintrasearch) 3 - WCDMA intra- and inter- frequency, no inter-RAT cells (SsearchRAT < Squal Sintersearch) 4 - WCDMA intra- and inter-frequency and inter-RAT cells (Squal SsearchRAT ) Sintrasearch Sintersearch 4 3 2 1 WCDMA CELL Confidential 35 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI SsearchRAT Cell Reselection 3G -> 2G CPICH EcNo UE starts GSM measurements if CPICH Ec/No =< qQualMin + sSearchRAT SintraSearch First ranking of all the cells based on CPICH RSCP (WCDMA) and RSSI (GSM) SinterSearch Rs = CPICH RSCP + Qhyst1 Rn= Rxlev(n) - Qoffset1 Serving WCDMA cell calculation, with hysteresis parameter Neighbour WCDMA or GSM cell calculation with offset parameter SsearchRAT qQualMin No Yes Rn (GSM) > Rs (WCDMA) And Rxlev (GSM) >QrxlevMin Second ranking only for WCDMA cells based on CPICH Ec/No Rs = CPICH Ec/No + Qhyst2 Rn=CPICH_Ec/No(n)-Qoffset2 Confidential 36 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Cell re-selection to GSM Cell re-selection to WCDMA cell of highest R value Cell Reselection 3G -> 2G UE ranks the serving cell and the measured neighboring cells to find out if reselection should be made • All the measured suitable cells (S-criteria) are included in the ranking. • Criteria for a suitable cell (S-criteria) is defined as – WCDMA intra-frequency neighbour cell: CPICH Ec/No > AdjsQqualmin and CPICH RSCP > AdjsQrexlevmin – WCDMA inter-frequency cell: CPICH Ec/No > AdjiQqualmin and CPICH RSCP > AdjiQrexlevmin – GSM cell: Rxlev > Qrxlevmin Ranking is done using Criteria R, and the UE reselects to the cell with highest R-criteria. R-criteria is defined as: • For serving cell: Rs = Qmeas,s + Qhysts • For neighboring cell Rn = Qmeas,n – Qoffsetts,n Qmeas is CPICH Ec/No for WCDMA cell and RxLev for GSM cell Confidential 37 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI How to avoid ping pong ? When phone is camped on 3G, GSM measurements can start when CPICH Ec/Io of serving cell is below Ssearch_RAT + QqualMin. When phone is camped on GSM, cell reselection to 3G is possible if CPICH Ec/Io of the candidate is above FDD_Qmin. Therefore, to avoid ping pongs between 3G and GSM the following condition should be met: FDD_Qmin >= QqualMin + Ssearch_RAT CPICH Ec/Io FDD_Qmin >= -12 dB QqualMin +Ssearch_RAT Ssearch_RAT=4 dB QqualMin=-18 dB Camping on 3G Measure GSM Camping on 3G t Confidential 38 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI How to avoid ping pong ? Parameters for cell reselections • • Qqualmin = -18dB Ssearch_RAT =2dB -> the 3G->2G cell reselection starts when Ec/No hits -16dB FDDQmin(GPRSFDDQmin) = -14dB (6) and QsearchP/QsearchI = always The cell reselection paramters 3G -> 2G and 2G -> 3G provide only 2dB hysteresis which is not enough and should be noticed from the RNC statistics as high amount of INTR_RAT_CELL_RE_SEL_ATTS from all the RRC Connection Setup Attempts • • Recommendation is to adjust the FDDQmin from -14dB to -10dB (or even up to -8dB) to provide 6 to 8 dB hysteresis between 3G to 2G cell reselection and 2G to 3G cell reselection Another parameter to tune is Qrxlevmin On top of Treselection the above parameters will slow down further the 2G to 3G and 3G to 2G cell reselections Confidential 39 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Treselection How long the reselection conditions must be fulfilled before reselection is triggered? Treselection Impacts all cell reselections : Inter RAT, intra frequency and inter frequency The UE reselects the new cell, if the cell reselection criteria (R-criteria, see next slide) are fulfilled during a time interval Treselection As this parameter impacts on all the cell reselections too long Treselection timer might cause problems in high mobility areas but too short timer causes too fast cell reselections and eventually causes also cell reselection ping pong Recommended value 1s should work in every conditions i.e. enough averaging to make sure that correct cell is selected However careful testing is needed to check the performance of different areas • (Dense) Urban area, slow moving UEs with occasional need for fast and accurate (to correct cell) reselections e.g. outdoor to indoor scenarios or city highways – in some cases cell by cell parameter tuning is performed to find most optimal value between 0s and 2s but typically 1s is optimal value when workload is considered as well • Highways, fast moving UEs must reselect correct cell – typically 1s works the best (however occasionally also 0s might be needed in fast speed outdoor to indoor cell reselections e.g. tunnels) • Rural areas, slow or fast moving UEs need very often reselect between different RATs and make proper cell reselections even when the coverage is poor – typically 1s works the best • Location Area Borders, usually the coverage is fairly poor – typically 1s works the best but sometimes to reduce location area reselection ping pong 1s is used when going from LA1 to LA2 and 2s from LA2 to LA1 Confidential 40 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI Cell Reselection KPIs RRC connection request amount for inter RAT cell reselection ratio to all RRC Connection request causes • When hysteresis is increased this KPI should decrease M1001C42INT R_RAT _C ELL_RE_SEL_AT T S M1001C0RRC_CONN_ST P_AT T RRC connection request amount for registrations ratio to all RRC Connection request causes • When hysteresis is increased this KPI should decrease M1001C46REGISTRATI ON_ATTS M1001C0RRC_CONN_STP_ATT Confidential 41 © Nokia Siemens Networks RNO / Wind 18/01/2008 - NMI