1_importante speech dcr analysis_revc
DESCRIPTION
1_Importante Speech DCR AnalysisTRANSCRIPT
Speech DCR Analysis
Telefonica Spain
Ericsson Internal | 2012-09-14 | Page 2
› In the last months, speech drop rate is constant increasing on Telefonica Spain network
› Next slides contains the updates on the current analysis being done
› Different behavior seen on different regions, so analysis is done in a separate way
Introduction
Ericsson Internal | 2012-09-14 | Page 3
– Degradation of 0.2 points – Degradation on both UMTS 900 and UMTS 2100 (increase start
mainly on W21):
Madrid
Ericsson Internal | 2012-09-14 | Page 4
– For the step on W21, the following RNCs contribute: › ALTAVISTA-01› ALTAVISTA-02› ATOCHA-03› ATOCHA-01 (aunque tb hay un escalón en W12)› SIMANCAS-01 (aunque tb hay un escalón en W17)
– Other RNCs that also have a degradation step:› DELICIAS-01 – W27› LUCRECIA-02 – W28› MANOTERA-01 – W24› MANOTERA-02 – W14
– Ericsson proposes for troubleshooting:› ATOCHA-03 – W21› DELICIAS-01 – W27
Madrid
Ericsson Internal | 2012-09-14 | Page 5
– Degradation of 0.15 points– Degradation mainly on U900 (big step on W25) and U2100 F1
Barcelona
Ericsson Internal | 2012-09-14 | Page 6
› For the step on W25, it can be seen a correlation between RSSI increase and Speech DCR during W25 and W36:
RSSI - BCN
Ericsson Internal | 2012-09-14 | Page 7
› IRAT degradation of both zones is mainly presented on U900 band (failures due to physical channel failure):
IRAT degradation - MAD
Ericsson Internal | 2012-09-14 | Page 8
› IRAT degradation of both zones is mainly presented on U900 band (failures due to physical channel failure):
IRAT degradation - BCN
Ericsson Internal | 2012-09-14 | Page 9
› Degradation seen on (PERU and VICALVAR present congestion on HS):
IUR degradation - mad
Ericsson Internal | 2012-09-14 | Page 10
› Degradation seen on (GUELL and BOTEROS present congestion on HS):
IUR degradation - bcn
Ericsson Internal | 2012-09-14 | Page 11
› Some correlation seen between SHO Success and Speech DCR, worthy to troubleshoot
› Degradation seen on W21 (but also in the beginning of the year have bad values and Speech DCR was not degraded):
SHO - MAD
Ericsson Internal | 2012-09-14 | Page 12
› The next slides contain the troubleshooting done for RNCs with significant degradation on SHO Success Rate.
SHO – MAD troubleshooting
Ericsson Internal | 2012-09-14 | Page 13
› Correlation of SHO failures with ADM failures
RNC-M-ATOCHA-02
Ericsson Internal | 2012-09-14 | Page 14
› Correlation of SHO failures with ADM failures since W35
RNC-M-Albeniz-02
Ericsson Internal | 2012-09-14 | Page 15
› Responsible for big degradation step on SHO since W33 is utrancell OPORTO_FUTBOL_280603701_01, with 0% of success rate. Probably this is a dummy cell that is causing SHO failures and drops.
RNC-M-Albeniz-02
Ericsson Internal | 2012-09-14 | Page 16
› No real correlation between SHO Failures and ADM Failures
RNC-M-Albeniz-03
Ericsson Internal | 2012-09-14 | Page 17
› Main responsible for big degradation step on SHO since W36 are utrancells PHABANA with 0% of success rate.
RNC-M-Albeniz-03
Ericsson Internal | 2012-09-14 | Page 18
› Correlation between SHO Failures and ADM Failures since W35
RNC-M-albeniz-01
Ericsson Internal | 2012-09-14 | Page 19
› Main responsible cells for SHO degradation:
RNC-M-albeniz-01
Ericsson Internal | 2012-09-14 | Page 20
› No correlation between SHO Failures and ADM Failures
RNC-M-atocha-01
Ericsson Internal | 2012-09-14 | Page 21
› Responsible for big degradation step on SHO since W33 are utrancell sOPORTO_FUTBOL_280603701 with 0% of success rate. Probably this is a dummy cell that is causing SHO failures and drops.
RNC-M-atocha-01
Ericsson Internal | 2012-09-14 | Page 22
› No clear correlation between SHO Success and Speech DCR, although at RNC level, there are some significant correlation (see next slide)
SHO - BCN
Ericsson Internal | 2012-09-14 | Page 23
SHO - BCN
Ericsson Internal | 2012-09-14 | Page 24
Ueh exception analysis - mad
Important contribution of MRAB in total number of drops
Ericsson Internal | 2012-09-14 | Page 25
Ueh exception analysis - mad
Although most of the drops are not in SHO and IRAT, there is a significant amount of drops due to these procedures.
Ericsson Internal | 2012-09-14 | Page 26
Ueh exception analysis – rnc-m-atocha-01
Important to tune SHO/RCS Parameters
Ericsson Internal | 2012-09-14 | Page 27
Ueh exception analysis – rnc-m-atocha-03
Important to tune RCS Parameters
Ericsson Internal | 2012-09-14 | Page 28
› The following slides contain proposals in order to tune Radio Connection Supervision timers on:
– Layer 1 - Physical– Layer 2 – RLC– Layer 3 – SHO
Rcs parameters change proposal
Ericsson Internal | 2012-09-14 | Page 29
L1 Timer (UE side)
N313 : number of consecutive “out of sync” after that Ue starts t313
“Out of sync” generation is Ue dependant, generally when the Ue measures a SIR < target SIR.
T313: After n313 consecutive “out of service”, the UE starts T313 and doesn’t stop it doesn’t
Measure an “in Sync” that is SiR> SIR Target
Timer for Radio Link Failure in DL = n313*10 (ms)+ t313 . After this timer expiry the call is
Dropped
Ericsson Internal | 2012-09-14 | Page 30
L1 Timer (proposed value)• Actual Telefónica Settings:
N313=100
T313=3
DL Radio Link Failure = 3 (sec) + 100*10 (ms) = 4 seconds.
This means that after 4 seconds Ue exits the DCH state and stays in Idle; there is no possibility to recover the call for the RNC.
• For Radio Link Failure in uplink Teflefónica has the DchRcsLost=5 (sec) and RLfailureT=1 (sec) noutSyncInd=10
That means DchRcsLost+ RLfailureT + nOutSyncInd*0.01= 6,1 seconds.
So uplink and downlink RLF timers are unbalanced
It is recommended to align them to a value that is enough robust. Proposed values:
RadioLinkFailureT = 20 (2 sec)
T313 = 5 sec
N313 = 200
So UL and DL will be balanced to 7 seconds for total L1 supervision
Ericsson Internal | 2012-09-14 | Page 31
L2 TimersUE RNCActive set update (1 or 2 RLC PDUs)
RLC
RLC
Signaling (rrc) messages associated with speech (and other RABs) are made
of RLC PDUs. If they are not received by the peer entity (don’t get an “ack”
by RNC / UE), may due to bad radio radio condition, they are retransmitted;
The total timer of retransmission is:
Timerpoll*(MaxDat-1)+TimerRST*(MaxRST-1)
Consider that if this total time expiries, the UE goes into Idle , while on the RNC
Side, it drops the call (maximum number of RLC retransmission reached)
Ericsson Internal | 2012-09-14 | Page 33
L2 Timers• Current Telefónica settings for SRB:
Timerpoll (both sides RNC and Ue)=160
MaxDat(both sides RNC and Ue)= 40
rncMaxRST = 1
UeMaxRST= 0 (means 1)
rncTimerRST=550
UeTimerRST= 10 (means 550)
Current in Telefónica Total L2 timer (both UL or DL)= Timerpoll*(MaxDat-1)+TimerRST*(MaxRST-1)= 160*(39)+ (1-1)(550) = 6,24sec < 7,1 sec (L1 proposed supervision timers)
Even if L1 = 6,1s, L2 should be greater enough to make L2 supervision algorithm more robust
In Ericsson opinion L2 supervision timers need to be made more robust (longer timeouts) to handle temporary bad radio, and it is suggested to set them enough higher than L1 timers and less than L3 timers
Ericsson Internal | 2012-09-14 | Page 34
L2 Timers : proposed values- On RNC side there is no limit on maxDat, so use that to increase the timeout
- On UE side maxDat is limited, set it to the max value (15) and then increase timerPoll to get the timeout length wanted.
UeRC=2 / UeRcRb=2,3,4 (UeRc=speech and DCH 2,3,4 associated with speech)
UeTimerPoll = RncTimerPoll= proposed = 240
In this case the RLC supervision timing will last
In RNC 240*39 = 9,36 seconds
In Ue 240* 39 = 9,36 seconds
The proposal to change them not only on pure speech (uerc=2, ueRcRb=2,3,4), but also for Speech + Ps 0/0, and to all active multiRab associated with Speech.
Ericsson Internal | 2012-09-14 | Page 35
L3 Timers•L3 Timers start when a rrc procedure takes place (i.e. Radio Bearer / Radio Bearer Setup complete - active set / active set update complete)
•The more classical example is L3 soft handover procedure
Active set update
UERNC
Iu relase request to the CN (internal System release triggered)
tRrcActiveSetUpdate=5 seconds
Rrc connection relase “unspecified”
After sending an rrc active set update, if the RNC doesn’t receive an active set update complete by the UE within timer: tRrcActiveSetUpdate the call is dropped , even if the Ue should recover the L1 or the L2 levels. So L3 timers (in particular the soft handover) should expiry only after the L1 and L2 timers have expiried (Ue / RNC has had all time considered useful to recover The connection)
5 sec
Ericsson Internal | 2012-09-14 | Page 36
L3 Timers proposed value
•Actual value of tRrcActiveSetUpdate in Telefónica = 6 seconds
•Proposed value= 10 seconds.
Ericsson Internal | 2012-09-14 | Page 37
› The following checks were performed, but no correlation with speech dropped call was found:
– Downtime– # Cells integrated– Total number of RAB Attempts– Correlation between IRAT Failures and high interference on 2G– Prop Delay checks:
› Check done on Madrid downtown and no significant impact found, next step to check on regions with big penetration of U900
Other Checks performed
Ericsson Internal | 2012-09-14 | Page 38
– Cell level analysis– Prop Delay impact on DCR on U900– GPEH analysis focused on Terminal Type Impact– Choose one/two RNCs and implement Parameters change Proposal– Check SHO Degradations on BCN RNCs– Study SHO Parameters and check possible parameter changes– Optimizaton Tools trial:
› IRAT Neigbours› Intra-Frequency Neigbours› R99 power paremeters
Next steps