RLF RCA & Improvement Trial
RLF RCA & Improvement Trial
RLF RCA & Improvement Trial
Improvement Trial
(HO Behavior Modification)
By
Amit Kumar Misra Reliance Jio Infocomm Limited
HANDOVER OVERVIEW
• The Intra-LTE Handover feature is based on measurements and coverage triggers evaluated by the user equipment. The user
equipment measurements are reported to the serving RBS which makes the ultimate decision on inter-cell handover.
• In Inter & Intra LTE Technology we have 5 kinds of Events reported for Handover.
EVENT_A1: Serving cell becomes better than absolute threshold.
EVENT_A2: Serving cell becomes worse than absolute threshold.
EVENT_A3:Neighbor Cell becomes amount of better than the serving.
EVENT_A4: Neighbor cell becomes better than absolute threshold.
EVENT_A5: Serving cell becomes worse than absolute threshold1 & Neighbor cell becomes better than absolute threshold2.
• Connected mode strategy is to focus more on mobility management & offer optimal coverage & quality to a user based on existing
radio conditions.
• Current deployed Strategy in Network is observed to be uniformly set for both inter-frequency & Intra Frequency LTE cells.
• The difference in coverage for TDD & FDD is not taken into account in this.
• The mode of Inter-frequency HO is set to be from A3 Event, after triggering of A2. A5 & A4 Events are currently not used as a part
of Inter-Frequency HO.
• The deployed A3 Event behavior can be fine-tuned further based on user behavior & cell radius, to have an optimal cell
performance based on existing radio conditions.
A3 Event Overview:
• This event is used to trigger handover when an offset between the serving cell and a
neighbor cell is me. i.e. Neighbor becomes offset better than serving
• The criteria basically consists of 3 major components –
Hysteresis: Used to add an additional delay to received level of
neighbor cell.
Offset: Used to encounter the deviation in radio signal of serving cell.
CellIndividualOffset (CIO): Relation specific offset set to encounter
varying radio condition & user behavior.
• With Hysteresis & Offset being set at cell level modifying these values can modify
behavior of complete cell.
• To achieve optimal cell performance , CIO can be fine-tuned to modify user behavior
at cell edge for relations showing poor statistics.
• HO can be triggered earlier/ Later by increasing/decreasing the value of CIO
parameter.
• A check was carried out on poor performing sites ( Average RLF Failures >10) to identify the most impacting cause.
• It was found that RLF failures account for more that 80% of drops, followed by ArqMaxRetransmission.
• Hence to identify the underlying cause for these failures, drilldown was carried out to identify the root cause associated
with it.
Add Neighbor
WrongCell Ue attempting at a cell having poor level compared to another existing cell.
• 1) Offset modification
• As described earlier CIO for a particular existing relation can be fine-tuned to shift triggering points & make sure that UE camps on
to a better cell , once the serving cell performance goes down.
Solution Types:
• The CIO can be added as an +ve or –ve Offset based on the scenario.
1)Apply negative CIO for relation AB. This induces additional offset in neighbor cell, hence delaying the HO process. This reduces
the overall Ping-Pong effect by increasing the gap between calculated perceived level of source & neighbor cell.
2)Apply positive CIO to relations A B. This will bring the triggering point a bit early hence allowing the user to be camped on
neighbor cell prior to further degradation in radio condition.
Ex: For scenarios like early HO & ping pong, -ve CIO delays the triggering point, hence reducing unwanted attempts. Whereas in case
of late HO, a +ve offset allows the HO to happen in the right time, by mitigating the unwanted delay.
Offset determination:
Ideally an 4dB offset makes sure that the Ue can avoid ping pong behavior that can result in poor KPI performance. But this may not be the
same for all the cases.
To have an approx. identification of relevant offset. A process can be used for existing bidirectional relations.
A general behavior between neighbor relations is considered to have an approx. identification in this process. It can be safely assumed that
ping pong is a behavior between two relations. Hence as a consequence the impact is supposed to be mutual on both the involved cells.
To use this behavior at a further extent the Average , max & min A3 measurements are taken for both relations to & fro.
It can be seen in above example that that the delta of avg –min & max levels for these relations is almost
nearing to zero, which indicates towards a similar pattern in triggering point.
• Similar observation when done at a low or no ping pong cell provided delta on a higher side.
As a conclusion on both scenarios, it gives an indication that the ping pong effect is basically due to lack of difference in triggering
points.
Now the question is to identify an offset , so that we can mitigate the ping pong effect , without shifting the radius too much.
To start on a safer note the delta found between the two relations suffering from ping pong behavior can be increased to 3
dB using the CIO parameter. (i.e 0 0+3 , 1 1+2).
This can be fine-tuned later based on further observation on arising scenarios.
• Scenario 1: High RLF failures observed in cell A (TDD) due to poor radio conditions. Co- Located FDD Cell found to be
having comparatively better performance.
Solution:
1) Reduce the A2Threshold to trigger the A2 event measurement a bit earlier. ( Optional: Need to be used only if the inter-frequency
relation has few or no IFHO attempts despite being in poor radio condition )
2)Apply positive CIO to relations A(TDD) B(FDD). This will bring the triggering point a bit early hence allowing the user to be
camped on neighbor cell prior to further degradation in radio condition.
• Scenario 2: High RLF failures observed in cell A (FDD) due to poor radio conditions. Co- Located TDD Cell found to be having comparatively
better performance, with existence of TDD in neighbor Cells
Solution:
Since FDD over here has a better footprint compared to TDD over here. It is quite unlikely to have an co-located TDD cell to have a better coverage. Hence, it is advisable to
have a relationship with TDD of neighbor cells.
1) Reduce the A2Threshold to trigger the A2 event measurement a bit earlier. ( Optional: Need to be used only if the inter-frequency relation has few or no IFHO attempts
despite being in poor radio condition )
2)Apply positive CIO to relations A(FDD) B(TDD). This will bring the triggering point a bit early hence allowing the user to be camped on neighbor cell prior to further
degradation in radio condition.
The additional 3rd step is used to make sure that the HO is limited to 1 st & 2nd Tier neighbors. Hence reducing the chances of failures that can arise from call dragging in an
overshooting cell ( eg: PCI collision/confusion with a far existing neighbor )
Please find attached Tilt modification guide to have further insight on modification techniques & procedures.
Antenna Tilt
Modification Guide
• A group of cells having high RLF were considered for root cause analysis based on earlier explained algorithm.
• It can be seen that after having a drilldown the scenario’s associated with above mentioned cases was mostly based on Ping Pong behavior &
poor coverage.
• So necessary analysis was done based on identified scenario to find an appropriate solution.
• Consistent call drops observed in this ENB for cell 0,2 (TDD) & 3,5 (FDD), mostly due to RLF failures & MaxArqTransmission.
• On further drilldown under High RLF observed in this cell has RLF majorly due to consistent ping pong behavior.
• Cell Availability was found to be rather consistent with no major impact.
• For affected cells of ENB 463 further drilldown was carried out with A3 measurements to identify the behavior associated with Pin Pong.
• It can be seen that the delta for Average , Min & Max is quite correlated on both sides for these relations.
• Hence as a solution for our existing scenario , a positive CIO is recommended to be added on these relations to avoid the ping pong behavior.
• Certain relations were found to have missing measurements from target to source. Hence for these it is recommend to add a default positive CIO of 3 dB.
Confidential Reliance Jio Infocomm Limited 19
ENB – 776 Analysis
• Consistent call drops observed in this ENB for cell 2 (TDD) & 5 (FDD), mostly due to RLF failures & MaxArqTransmission.
• On further drilldown under High RLF observed in this cell has RLF majorly due to consistent ping pong behavior.
• Cell Availability was found to be rather consistent with no major impact.
KL-KL-REM-01-A,776,5=>1431,5 -101.304 -117.585 -76.4151 4146 4121 -99.9697 -117.76 -75.7867 6342 6306 -1.33442 0.175094 -0.62843 -2196 -2185Add CIO of 3dB
KL-KL-REM-01-A,776,2=>776,0 -109.063 -124.889 -89.5556 1151 1144 -107.705 -123.811 -88.7838 224 222 -1.35802 -1.07808 -0.77177 927 922Add CIO of 2dB
KL-KL-REM-01-A,776,2=>1431,2 -104.063 -121.479 -83.8125 698 678 -101.822 -119.746 -82.2687 809 769 -2.24098 -1.7329 -1.54384 -111 -91Add CIO of 2dB
CIO already
KL-KL-REM-01-A,776,5=>463,5 -99.3034 -116.92 -77.88 493 483 -99.0837 -117.352 -76.4444 2004 1990 -0.2197 0.431852 -1.43556 -1511 -1507 added for 463
CIO already
KL-KL-REM-01-A,776,5=>463,3 -100.751 -119.633 -79.5306 363 360 -100.93 -120.078 -78.8824 804 795 0.179192 0.445778 -0.64826 -441 -435 added for 463
• For affected cells of ENB 776 further drilldown was carried out with A3 measurements to identify the behavior associated with Pin Pong.
• It can be seen that the delta for Average , Min & Max is quite correlated on both sides for these relations.
• Hence as a solution for our existing scenario , a positive CIO is recommended to be added on these relations to avoid the ping pong behavior.
• Certain relations were found to have missing measurements from target to source. Hence for these it is recommend to add a default positive CIO of 3 dB.
• For relations existing with ENB-463 no modification are done as CIO is also applied in ENB – 463.
• Consistent call drops observed in this cell mostly due to RLF failures & MaxArqTransmission.
• On further drilldown under High RLF observed in this cell has RLF majorly due to poor coverage , consistently.
• Cell Availability was found to be rather consistent with no major impact.
• Also the drops were found to be limited to 1 band , with the other colocated band functioning with adequate performance.
• Further investigation was carried out to identify the root cause behind poor coverage.
• It was found that this site is a standalone site with no proper neighbors within feasible
distance.
• TA Samples also indicated the same , which was quite in sync with the TA found in CSL
for RLF. Most of the failures are within 5-6 Km from cell & from band 3, which
indicates the user to be far away from site & is being served by FDD ( FDD having
better coverage compared to TDD).
• As a solution, since it is a standalone site modification of HO Strategy & physical
settings is not feasible.
We can modify Fractional Path Loss (α) compensation component from 80% to 100% to
enable the Ue to radiate at high Tx power a bit early compared to current existing
scenario. This will safeguard some calls from failing too early.