1. PDCCH LINK ADAPTATION tuning

2. Summary
In the attempt to improve the performance (throughput) some parameters controlling the use of the PDCCH resources were tuned.
The results is indicating that significant benefits are possible
Highest gains can be expected in multi user scenarios
Exact gain will depend on traffic pattern and radio environment
The parameter changes and tests was done lab in a network without commercial traffic.
The following parameters were tuned:
pdcchCfiMode
pdcchLaGinrMargin (system constant)
nrOfTransmissionsSib1
Some of the parameters might not be suitable for a live network deployment

3. Lab Setup & Execution Single cell – L12B ICP4 software – 5MHz Bw @2
Lab Setup & Execution Single cell – L12B ICP4 software – 5MHz
Spirent VR5 Channel Emulator (mainly EPA5 channel used)
Full buffer UDP or FTP traffic
Programmed Attenuation sweep
from ~100dB path loss  UE release (~150dB) (1dB steps, 30sec/step)
UE’s and Logging Tools
Qualcomm 8960 (FFA) & Quanta (Qualcomm chip)
All data shown from UE side (not network side data included)

4. Definitions & Data Sources
Path Loss
As reported by Qualcomm UE (based on RS power)
SINR
As reported by Qualcomm UE (probably based on C-RS?)
UL & DL Scheduling Ratio
Ratio of subframes when the UE DRB data is scheduled to total subframes (ms) in averaging window
BLER
Ratio of successful Redundancy Version 0 (RV0) transmissions to all RV 0 transmissions

5. Characteristics – Default parameters 5MHz Bw – EPA5 – Single UE - UDP DL full buffer traffic
UDP Downlink – Expected peak TP
(35M) not reached
Only RLC status and polling
sent in uplink
Uplink Limited  lowest MCS not reached
No. of PRB reduced in good RF env.
Sched. ratio decreased
with decreasing RF env.

6. Scheduling Ratio & PRBs Detailed View
Sched. ratio reduced in poor RF
especially for subframe 0 & 5
In good RF 25, 22, 19 and 5 PRBs are used
In poor RF only 25 PRBs are used
The PRB reduction in good RF is mainly
in subframe 0 and 5
System Information (SI) is scheduled in subframe 0 and 5 with various repetition

7. PDCCH Control Chanel elements
Control Channel Element 1 2 3 4 5 6 7
1 CCE – 8 spaces
Capacity Consumption
Low PDCCH
GOOD Radio Env.
Low Robustness 
2 CCE – 4 spaces 
Capacity Consumption
High PDCCH
POOR Radio Env.
High Robustness
4 CCE – 2 spaces 
8 CCE – 1 space
Fewer CCE used for each PDCCH allow for more users scheduled in parallel

8. DL scheduler & Link Adaptation
PDCCH Link Adaptation
One PDCCH can be mapped to 1, 2, 4 or 8 CCEs
Selection of number of CCEs is done based on the same GINR estimate used for PDSCH link adaptation
A margin (back-off) is added to PDSCH GINR to compensate for different interference scenarios between the two channels
Degraded RF conditions 
Higher CCE Aggregation level
Lower MCS (for the same number of PRB’s)
If the required number of CCE’s are not available in the control region (1) the UE can not be scheduled on the PDSCH (2) in that subframe
eNodeB UE
Resource allocation (PDCCH)
Data (PDSCH)
DL scheduler & Link Adaptation
1CCE
2CCE
4CCE
8CCE

9. PDCCH Candidates for 5MHz Example
Exact candidate positions depend on x-RNTI and subframe number
Number of candidates increasing with number of PDCCH OFDM symbols (CFI)
Number of candidates is dependent on number of PHICH groups configured (def. 3)
Both Uplink and Downlink Scheduling need the PDCCH resources
Uplink scheduling has priority over downlink scheduling
Common is always using 8 CCE to ensure full cell coverage (Ericsson)
CFI = 1
No Common scheduling possible
No Poor/Medium coverage UE Scheduling possible
CFI = 2
One 8 CCE UE’s or common can be scheduled in the same Subframe
CFI = 3
Two 8 CCE UE’s (or common + one 8 CCE UE)
can be scheduled in the same subframe

10. CCE Aggregation Level vs. SINR and MCS
CCE Use increasing at low SINR
PDSCH Link Adaptation (MCS) follow SINR
PDCCH LA follow PDSCH LA

11. Problem Summary In good radio environment (1 & 2 CCE required)
Reduced Sched. ratio due to
SI scheduling (higher priority)
uplink scheduling (higher priority)
In good radio environment (1 & 2 CCE required)
DRB (user) data can be scheduled on remaining PRB:s when SI is scheduled since there is enough space in the PDCCH
Possible to fit both uplink and downlink scheduling in the PDCCH
In poor radio environment (8 CCE required)
No DRB (user) data can be scheduled when the SI is schedule due to lack of PDCCH resources (only one 8 CCE candidate)
No downlink data can be scheduled when there is uplink data to schedule
Reduced downlink scheduling ratio

12. Solution Summary Optimized Parameter Setting
Increase max number of PDCCH symbols to 3
pdcchCfiMode: 0  5
0: (Static by bandwidth – 2 for 5Mhz), 5: (Adaptive CFI, max 3)
Reduce SIB Scheduling
Sib 1 is scheduled every 80ms and repeated 4 times, i.e. every 20ms (default)
nrOfTransmissionsSib1: 4 1 (i.e. 1 repetition = every 80ms)
Reduce PDCCH Link Adaptation Margin (relative PDSCH LA)
Delaying the CCE increase to a lower SINR will make the PDCCH less robust
Analysis needed to find suitable tradeoff

13. Tuning of PDCCH LA Margin
Low Margin  UE will fail to decode the PDCCH
UE will consider itself not scheduled  Reduced Scheduling Ratio as seen from the UE side
High Margin  UE will use unnecessarily many CCE’s
Increasing the probability of lack of PDCCH resources  Reduced Scheduling Ratio
In case of a single UE with DL UDP traffic this will only happen when SI is scheduled*
Strategy:
Reduce margin until a distinct decrease in Scheduling Ratio is seen*
Initially test in large steps, reduce later
Margin steps: 10dB, 6dB, 3dB, 0dB -3dB & -5dB

14. DL Scheduling Ratio vs. MCS 10- -5dB Margins - EVA70 Evaluation
Reference: 10dB
(default)
6dB – no degradation
0dB – no obvious
degradation
3dB – no degradation
-3dB –obvious
degradation
-5dB –severe
degradation

15. analysis EVA70 Evaluation
Based on the EPA70 channel tests it was found that the Scheduling ratio was reduced with a PDCCH LA margin of between 3dB and -3dB
A focused test in the range of 4dB to -3dB margin was made
Step size: 0.5dB
Channel model was changed to EPA5
Results are shown as
Downlink Scheduling Ratio vs. MCS
Distribution of PDCCH Aggregation level vs. MCS

16. DL Scheduling Ratio & CCEs vs. MCS 4- 3dB Margins – EPA5
Reference 4dB Margin
3.5dB Margin
No degradation
3dB Margin
No degradation

17. DL Scheduling Ratio & CCEs vs. MCS 3- 2dB Margins – EPA5
No degradation
PDCCH decoding failures
2.5dB Margin
PDCCH decoding failures
2dB Margin

18. DL Scheduling Ratio & CCEs vs. MCS 1.5- 0dB Margins – EPA5
PDCCH decoding failures
1.5dB Margin
PDCCH decoding failures
1dB Margin
PDCCH decoding failures
1dB Margin

19. LA Margin Selection
Based on the results the PDCCH LA margin was selected as 3dB
Slightly lower margin might be possible without major degradation
Considering the limited test scenarios and channel models evaluated the chosen setting was considered reasonably safe for the non commercial network
3dB Margin
No degradation relative to
the reference case

20. Evaluation of performance improvement multi ue scenario
2 UE’s in the same radio environment (EPA5 channel)
UE1: Full buffer TCP Downlink Traffic
UE2: Full buffer TCP Uplink Traffic
In this scenario the UE’s are competing on the PDCCH space rather than PDSCH
TCP Traffic is generating additional traffic in the reverse direction
TCP ACK’s
Programmed Attenuation Sweep of both UE’s together (splitter)
From ~100dB path loss  UE release (~150dB) (1dB steps, 30sec/step)
3 Cases Evaluated:
Default Parameter Set
Flexible CFI with a max of 3 symbols
Fully Optimized Parameters
PDCCH LA margin – 3dB
eNodeB
UE1
Data (PDSCH)
Data (PUSCH)
UE2

21. UE1: TCP Downlink UE Downlink & uplink Throughput
Increased UL Thp. Is an effect of
increased rate of TCP ACK’s
Improved Thp. (~ 2MBps) from
Improved scheduling ratio
Increased sched. Ratio at low
SINR due to uplink PRB reduction

22. UE1: TCP Downlink UE Downlink Scheduling ratio per subframe number
Default Parameters
Flexible CFI (max 3)
Optimized Parameters
Dowlikn
Side by side with flexible cfi and adamantium (need new labeling)

23. UE2: TCP Uplink UE Downlink & uplink Throughput
Increased DL Thp. Is an effect of
increased rate of TCP ACK’s
Improved Thp. (~ 1MBps) from
Improved scheduling ratio
Flexible CFI enough to give maximum improvement.
UL scheduling mainly competing with other UL scheduling's and the SI scheduling since the UL is prioritized
Main improvement from Flexible CFI

24. Field performance

25. Filed performance optimized parameter setting
The optimized parameter setting was tested in the field
~1 Hour drive route in suburban-rural area
Both unloaded and 100% OCNG downlink load was tested
Single UE FTP download
Limited file size causing idle periods during drive test
EPA5 Lab result curve included as reference

26. Field: Throughput performance Suburban-rural area
The radio environment in field appears a bit less challenging for the UE than EPA5 in the Lab

27. Field: Scheduling ratio Suburban-rural area
In general lower scheduling ratio during field test relative to lab
Handovers, quick variations in radio environment in connection to the TCP protocol etc. could account for this lowered scheduling ratio
Samples with very low scheduling ratio is due to idle times between filed downloads

28. THNAK YOU