CAPÍTULO IV DE LAS AUDITORÍAS
DEL RECURSO DE RECONSIDERACIÓN
• Configuration Unsupported
If UTRAN instructs the UE to use a configuration that it does not support, the UE shall set the IE failure cause to configuration unsupported. This usually happens when UE goes from UMTS coverage area to GSM area.
• Physical channel establishment failure
When a physical dedicated channel establishment is initiated by the UE, the UE shall start a timer T312 and wait for layer 1 to indicate N312 successive "in sync" indications. On receiving N312 successive "in sync" indications, the physical channel is considered established and the timer T312 is stopped and reset.
If the timer T312 expires before the physical channel is established, the UE shall consider this as a "physical channel establishment failure". This can be due to many reasons like node busy, UE in coverage fringe, low timer value. If the failure is high, T312 wait timer can be increased to give UE more time to sync.
The counters used for intra RNC and inter RNC handovers are:
Other Optimization Strategies
1. Load Estimation based on Throughput
This approach is simple and fast. However, it ignores other cell interference issues that may result in reduced throughput, thereby giving false impression of the spare system capacity
2. Load Estimation based on Wideband Received Power
This approach provides a better view of the used resources (cell of interest plus interferers) to provide an indication of the loading conditions. The problem with this approach lies in the accuracy of
the reported measurements and the fact that they have to be averaged over a period of time for increased reliability.
3. Network Audit Strategy
Characterization of Radio Properties Complete neighbor relations overview Complete feeder check
Complete parameter value and consistency check
Counter statistics collection: starting point of performance Setting up routine for monitoring all other activities in the
network (SW upgrades, re-parenting etc) Network Audit Strategy
Radio Network Aspects
• Cell Planning Coverage
• Cell Planning , Interference
• Neighbor Definitions
• Location Area / Routing Area Planning
Careful geographical placement of LA/RA borders to minimize the number of LA/RA updates and Iur handovers
• In low traffic areas
• Perpendicular to main traffic flow Number of Cells in LA/RA
• Load from LA/RA updates
• Paging Load Basic Rule
• 1 to 3 RNC per LA and RA
• No sharing of LA/RA i.e separate LA/RA for GSM / GPRS and for WCDMA even if they cover the same geographical area and the same sites in case of co- siting.
• UL / DL Power Balancing
• Output Power of common control channels
• Code Planning
• Parameter Tuning
O & M Aspects
• Parameter consistency
• Software status
• Detecting Crossed Feeders
• Measurement Principles
Some Optimization examples done on live UMTS network: 1 Optimization of ISHO Thresholds
There is a main parameter in the ISHO procedure that determines whether to leave the 3G network for the 2G network (connected mode) on RSCP or EcNo. Even if these parameters are link (through the RSSI), there are rather independent and it is difficult to guaranty a certain level of RSCP by a level of EcNo and vice versa.
On IDF, we noticed that at RSCP levels of -100dBm the quality of the radio links was rather poor and therefore it was difficult to correctly keep the connection alive till the ISHO operation successes.
There are 3 events involved in the ISHO procedures namely e2d, e2f and e3a.
ed2 à compressed mode activation à UE starts measuring 2G cells based on the 3G2G neighbors list
e2f à compressed mode deactivation à UE stops the measurements e3a à handover from UTRAN command (ISHO) à UE sends the elected 2G cell to perform the HHO
The initial thresholds for these 3 events are respectively -100dBm, -97dBm and -103dBm.
It is important to notice that during the compressed mode the UE is really busy with the monitoring and the power control. Most of the time, it is during this phase that we experience the worst quality of the voice. Hence the settings will tend to reduce this time of CM.
The propose settings are:
The UE goes into CM 2dB early thus it guaranty a better quality of the RL. The main modification is that e3a is higher than e2d which mean that the UE is allowed to go on GSM has soon as the best suitable 2G cell is found.
Remark: The procedure works correctly only and only if the 3G2G neighbor plan is optimized e2f @ -97dBm e3a @ -103dBm C M e2f @ -96dBm e3a @ -97dBm e2d @ -98dBm AFTER BEFORE
2 Optimization of Qrxlevmin for 3G Network Eligibility
Ping-pong type 1 – Green Area :
FDD_Qmin vs QRxLevmin_3G
The S criteria is there to prevent the UE selecting a cell with RSCP below QRxLevmin and EcNo below QQUAL_min but the radio conditions change quickly (especially at low levels). Hence it happens that the UE reselects the 3G layer and goes rapidly out of 3G coverage after the network reselection. Also, when the RSCP is low the measurements of EcNo reported by the UE are less accurate and this ping-pong effect is therefore emphasized.
Ping-pong type 2 – Purple Area :
FDD_Qmin vs QQUALmin + SSearch-RAT
If the hysteresis between the 2 thresholds FDD_Qoffset and QQUALmin + SSearch-RAT is not big enough then it will happen that the UE decides to reselect the 2G layer right after a 3G reselection (and vice versa). This
QRxLevmin_3G is lowered from -111dBm to -115dBm
FDD_Qmin
behavior is due to the fluctuating radio conditions. The actual margin of 4dB is considered wide enough to regulate this effect.
Parameter modification: Qrxlevmin
It was noticed on the measurements made in connected mode that there are situations where the UE goes back and forth between the 2G and 3G networks. The 3G-2G reselection is necessary to control when the UE is allowed to go (or should go) on the 3G network but the ping-pong is a side effect of this procedure: the UE is in an undetermined state.
Hence, it is important to reduce this effect because:
- It reduces the availability of the UE (especially for video calls) - It decreases the life time of the battery of the UE
- It loads the RNC with signalling for network selection management in IDLE mode.
The proposition is to modify the Qrxlevmin 3G from -111dBm to the minimal value of -115dBm
Remark: T0 is the 17th of February 2005. The period of observation is from the 31st of January 2005 to the 6th of March 2005
Interpretations:
The inter-RAT cell reselection is a good criterion to know the rate of network switching between 3G and 2G.
A low inter-RAT cell reselection rate means that you stay whether on the 3G or the 2G network.
In our case we stay longer on the 3G network because the minimum level of 3G cell eligibility is lowered giving easier access to 3G cells
Conclusions:
Processing amount on the RNC is decreased. Even if this issue is not critical now, it will become an issue with the traffic growth.
Basically, the RNC processes more “Service requests” than “Signalling requests” in proportion.
The ping-pong effect is reduced since the UE stays longer on the 3G network.
Therefore the UE is more available on the 3G network. Remark:
The quality criterion for a 3G network is the EcNo more than the RSCP. With a Qrxlevmin 3G at the minimum level of -115dm, almost all cells are eligible for 3G registration even when the UE’s are indoor. It is then the minimum EcNo requirement (FFD_Qmin) that will determine if a cell can be selected for reselection. Also, as the lost of 3G coverage is detected à -115dBm or -18dB (Qqualmin) it becomes even more important to tune properly the other cells reselection parameters.
3 Optimization of ROFFSET for Enhanced SHO Management
The event e1a’ is intended to release a call when the interferences brought by the communication become too high. The “Improved HO handling” feature helps in maintaining enough radio links after a SHO failure in order to allow further SHO attempts. If we don’t set a limit to this system then the UE will bring interferences from the serving cells to the neighbour cells. The RRC Re- Establishment procedure will be initiated if e1a’ fails (which is equivalent to a call drop for CS with the Qualcomm chipsets that do not support this call re- establishment).
When e1a’ is received the RNC performs a last try: if the procedure fails then all the RL are released.
If there is a procedure e1a in progress when the e1a’ is received then all the RL are released also if the procedure fails
e1a’ is triggered every “reporting interval” ◊ several attempts are allowed since the radio links are not released after SHO failure
If the UE goes too deep inside the service area of cell 2 (with cell 2 as neighbour and cell 1 as serving cell) then if the SHO attempt fails the procedure asks to release the RL to avoid interferences on cell 2
The settings of event e1a’ are sent in the second measurement control message right after the definition of SHO events e1a, e1b and e1c. To distinguish the 2 measurement report types the measurement identity is
SHO area
EcNo
P CPICH 1 Best Active cell
P CPICH 2 Monitored cell Event 1A' Reporting Range Time Event 1A
set to 9 for e1a and to 14 for e1a’. The event e1a’ inherits the settings of
event e1a except for the triggering threshold defined as follow:
Re1a’ = Re1a – Roffset, with Re1a’ ≥ 0
Description of the problem
The parameter Roffset was initially set to 0dB such that e1a and e1a’ have the same triggering threshold.
With this original setting events e1a and e1a’ are sent at the same time resulting in a release of the radio links in case of SHO addition failure. Indeed, as the e1a’ is sent to release the call in case of SHO failure. Therefore, all e1a are pre-empted by e1a’ events.
à This behaviour is suspected to create additional drops.
Implemented modification
The value of Roffset will be set to 3dB thus setting. Indeed, With the new setting, the event e1a’ is delayed by 3dB compared to e1a which is
equivalent to extra time allowed to perform the HO. Eventually, when the
best SC of the neighbour set reaches the level of the best cell of the active set within the range of 1dB (Re1a’) then the call is released (for CS only – for PS there are different procedures).
Expected gains and behaviors
à Reduction of the number of e1a’ and at the same time an increase of the number of e1a
As the e1a’ is delayed we allow more attempts of e1a and therefore this number should increase. But we should also decrease the number of e1a’ since there will be fewer situations triggering this event. We will verify that this delta between these two events does not increase so much that it may overload the RNC in term of signalling.
As the system is more resistant to SHO failures there should be improvements on the call drop rates.
Assessing Success of Recommended Change
The effect of the recommended changes should be analyzed for its success. This can be verified through,
• Drive Test – real time throughput performance
• Statistics - KPI improvements against Baseline