1. Performance Evaluation of Femto-based HetNet
TSG-RAN WG1 Meeting #60
San Francisco, USA, February , 2010 R
Performance Evaluation of Femto-based HetNet
Agenda: 8.2.2
Source : Samsung
Document for: Discussion and Decision

2. Introduction Background
HetNet scenario priority was agreed in RAN1#59
Indoor HeNB clusters (as in TR [1])
Outdoor Hotzone cells (as in TR [1]) with configuration #1 and #4
Indoor Hotzone cells
In HeNB scenarios, companies in [2][3][4][5] provided some results related with Macro UE dead-zone problem by evaluating wideband SINR distribution or user throughput.
In [2], SINR distribution of Macro UEs was shown to be significantly degraded as the percentage of indoor Macro UEs increases, due to the increased interference signal level coming from HeNBs.
Moreover, the evaluations in [4] [5] showed that Macro UE deadzone problem can be overcome by performing a power control method of HeNBs [4], frequency allocation, frequency selective scheduling, and beamforming [5].
This contribution evaluates femto-based HetNet system performance
To assess the system performance, we first observe the wideband SINR distribution according to the change of deployment ratio of HeNBs.
By applying fast-fading model, i.e., SCM, from the Macro eNB to Macro UE, the dead-zone problem of Macro UE is investigated by evaluating Macro UE throughput.
To overcome the dead-zone problem, we observe the benefit of simple HeNB silencing method on Macro UE throughput. 2

3. Simulation Scenarios and assumptions
This contribution compares performance for following scenarios:
Scenario 1: Macro eNBs only Scenario 2: Macro eNBs + Home eNBs Scenario 3: Macro eNBs + Home eNBs with ICIC (TDM silencing)
HetNet evaluation with dual-stripe model: downlink
Macro eNB - Macro UE: 4x2 SU-MIMO with SCM urban macro with low angular spread Home eNB - Home UE: 1x1 SISO
(Please see the Appendix for details
of parameters)
Macro system-level simulation parameters
System Bandwidth
10 MHz (50 PRB)
Subframe Length
1 ms
Subband Size
50 PRB (wideband)
Users per Sector 10
Scheduling
Proportional fair with full bandwidth allocation
Transmission Mode
SU-MIMO
Link Abstraction
Exponential Effective SINR Mapping
Modulation and Coding Schemes
QPSK r = 1/8, 1/6, 1/5, 1/4, 1/3, 2/5, 1/2, 3/5, 2/3, 3/4, 5/6 16QAM r = 2/5, 1/2, 3/5, 2/3, 3/4, 5/6
64QAM r = 3/5, 2/3, 3/4, 4/5, 5/6, 6/7, 7/8, 8/9, 9/10
Receiver Type
MMSE
H-ARQ
Chase combining, non-adaptive/asynchronous
Feedback
Wideband RI/Wideband PMI/Wideband CQI 4 bit CQI quantization with 1.6 dB SINR resolution LTE R8 PMI codebook 5 ms period, 6 ms delay
Link Adaptation
Target block error rate: 5%
Outer-loop link adaptation: ACK +0.5/9 dB NACK -0.5 dB
Overhead
3 symbols per subframe for control, LTE R8 CRS
Home eNB cluster parameters
N (number of cells per row ) 10
M (number of clusters per sector) 1
L (number of floors per cluster)
R (deployment ratio )
10/25/50/75/100 %
P (activation ratio)
100 %
Probability of Macro UE being indoors
(%) 3 3

4. Heterogeneous Network Layout
Cell layout snapshot
Placement of Macro eNBs
Two-tier Macro eNBs with 3-sector antennas are deployed
A HeNB cluster with a single floor is placed in a macro sector area, as shown in figures below
Cellular layout of two-tier Macro eNBs and Home eNBs
Blue x : Macro UE (outside of Apt.)
Red x : Macro UE (inside of Apt.)
Red o : Home eNB
Black * : Home UE 4 4

5. Results 1 – Wideband SINR distributions
Macro/Home UE wideband SINR distribution: Scenario 2
Observations
As deployment ratio (DR) increases, the geometry of Home UEs is drastically degraded due to the increase of interference level from HeNBs.
Asymmetric shift is observed with dual-stripe model, which is different from [3].
However, the geometry of Macro UEs does not change significantly because a limited number of dominant HeNBs gives huge amount of interference and the remaining interference is negligible due to far distances and interior/exterior wall losses.
Asymmetry due to interior/exterior wall losses from dual-stripe model 5

6. Results 1 – Wideband SINR distributions (cont’d)
Macro UE + Home UE wideband SINR distribution: Scenario 2
Observations
Size of coverage hole of Macro UE depends on the location of HeNB cluster
Large coverage hole of Macro UE occurs if the HeNB cluster is deployed in the Macro cell edge.
ICIC technique for Macro UE in dead-zone, i.e., victim UE, is necessary to ensure that the UE is able to work (Control channel BLER < -8 dB [4]).
Different size of Macro UE coverage hole
SINR (dB) 6

7. Results 2 – Macro UE throughput
Macro UE w/o Home eNB ICIC: Scenarios 1 and 2
SU-MIMO based on LTE R8 codebook + PF with full bandwidth allocation
Macro UE throughput
Impact of HeNB deployment ratio: 5%, 10%, 20%, 30% and 40%
Outage probability of Macro UE increases with an increase in deployment ratio.
Macro UE throughput
Cell Average
5% User
Median User
Outage Prob.
DR 0% (Only Macro eNBs)
0.0%
DR 5 %
(-19.1%)
(-57.3%)
(-30.0%)
1.5%
DR 10 %
(-27.7%)
(-77.8%)
(-37.2%)
3.2%
DR 20 %　
(-32.9%)
0.0
(-100.0%)
(-41.7%)
10.0%
DR 30 %
(-35.1%)
(-100.0%) 　
(-44.2%) 　
10.1%
DR 40 %
(-35.2%)
(-43.7%)
10.4% 7

8. Results 3 – A simple ICIC technique
Macro UE w/ Home eNB ICIC: Scenarios 1 and 3
Home eNB ICIC
Home eNB serves only one Home UE with closed subscriber group (CSG) policy
If ICIC function is not working with Home eNB, the Macro UE can not be served by both Macro eNB and Home eNBs.
Macro UE that experiences severe interference requests ICIC function to HeNBs
Victim UE in dead-zone requests silencing to the adjacent aggressive HeNBs
Aggressors
Home eNB ICIC
Home eNB
Home UE
Macro UE
Victim UE
Dead-zone
Macro eNB
Home eNB ICIC for victim UE 8

9. Results 3 – A simple ICIC technique (cont’d)
Macro UE w/ Home eNB ICIC: Scenarios 1 and 3
SU-MIMO based on LTE R8 codebook + PF with full bandwidth allocation + Home eNB TDM silencing
Home eNB ICIC triggering threshold: RSRP 10 dB
Home eNB deployment ratio: 5% and 10%
Macro UE throughput
Outage probability of Macro UE decreases if a simple ICIC technique is exploited.
Macro only
No ICIC
HeNB ICIC
Cell Average
(-19.1%)
(-10.3%)
5% User
(-57.3%)
(-37.8%)
Outage Prob.
0.0%
1.5%　
Macro only
No ICIC
HeNB ICIC
Cell Average
(-27.7%)
(-13.2%)
5% User
(-77.8%)
(-40.1%)
Outage Prob.
0.0%
3.2%　　 9

10. Conclusion Concluding remarks References
System level performance of Femto-based HetNet in dense urban scenario [1] was evaluated.
As the deployment ratio of HeNBs increases, the performance of Macro UE is significantly degraded in terms of throughput and outage probability as well.
To overcome dead-zone problem, we need
ICIC technique to protect victim UEs in dead-zone
Remain issues
Fast fading model for HeNB
ICIC technique incorporating with resource partitioning and beamforming
References
[1] 3GPP TR , “Evolved Universal Terrestrial Radio Access; Further advancements for E-UTRA Physical layer aspects”, v 1.5.1
[2] R , “Evaluation and Analysis for Different Topologies in Het-Net”, CATT.
[3] R , “Interference Conditions in Heterogeneous Networks”, Qualcomm.
[4] R , “Downlink CCH Performance Aspects for Co-channel Deployed Macro and HeNBs”, Nokia.
[5] R , “Effect of Cell Association and Frequency Allocation With and Without FSS and BF - Indoor HeNB Cluster Scenario”, Motorola. 10

11. Appendix (Simulation parameters)
Macro eNB parameters
Parameter
Assumption
Cellular Layout
Hexagonal grid, 19 cell sites with wrap-around, 3 sectors per site, frequency reuse factor 1
Carrier Frequency
2 GHz
Inter-site Distance
500 m
System Bandwidth
10 MHz
eNB Transmit Power
46 dBm
Distance-dependent Path Loss
PL (dB) = log10R, R in m
Penetration Loss
20dB
Lognormal Shadowing
Similar to UMTS 30.03, B
Shadowing Standard Deviation
8 dB
Auto-correlation Distance of Shadowing
50 m
Shadowing Correlation
Between Sites
0.5
Between Sectors
1.0
3D Antenna pattern
Horizontal
3dB BW: 70 deg, Back-lobe: 25 dB
Vertical
3dB BW: 10 deg, Back-lobe: 20 dB Downtilting: 15 deg
eNB Antenna Gain after Cable Loss
14 dBi
eNB Antenna Height
32 m
UE Antenna Gain
0 dBi
UE Antenna Height
1.5 m
UE Noise Figure
9 dB
Thermal Noise Density
-174 dBm/Hz
Assumption
Parameter
UE Distribution
UEs dropped with uniform density within the indoors/outdoors macro coverage area, subject to a minimum separation to Macro and Home eNBs
Minimum Distance between UE and eNB
>= 35 m
UE Speeds of Interest
3 km/h
Inter-cell Interference Modelling
Explicit modelling (all cells occupied by UEs)
Cellular Grid
Antenna bore-sight points toward flat side of cell 11 11

12. Appendix (Simulation parameters)
Macro system-level simulation parameters
Parameter
Assumption
General
Parameters and assumptions not explicitly stated here according to 3GPP specifications
Duplex Mode
FDD
System Bandwidth
10 MHz (50 PRB)
Network Synchronization
Synchronized
UEs per Sector 10
Handover Margin
1 dB
Subframe Length
1 ms
Subband Size
50 PRB (wideband)
Scheduling
Proportional fair with full bandwidth allocation
Transmission Mode
SU-MIMO
Modulation and Coding Schemes
QPSK r = 1/8, 1/6, 1/5, 1/4, 1/3, 2/5, 1/2, 3/5, 2/3, 3/4, 5/6 16QAM r = 2/5, 1/2, 3/5, 2/3, 3/4, 5/6
64QAM r = 3/5, 2/3, 3/4, 4/5, 5/6, 6/7, 7/8, 8/9, 9/10
Link Abstraction
Exponential Effective SINR Mapping
H-ARQ
Maximum 3 retransmissions
Chase combining, non-adaptive/asynchronous
No error on ACK/NACK
For the retransmission in subframe n,
ACK/NACK is reported subframe n+4 and retransmission is available in subframe n+8
Feedback
Wideband RI/Wideband PMI/Wideband CQI
5 ms period, 6 ms delay (measurement in suframe n is used in subframe n+6)
4 bit CQI quantization with 1.6 dB SINR resolution
No CQI measurement error
No PMI feedback error (perfect and ideal feedback)
Codebook
LTE R8 Codebook
Parameter
Assumption
Link Adaptation
Target block error rate: 5%
Outer-loop link adaptation: ACK +0.5/9 dB NACK -0.5 dB
Receiver Type
MMSE based on DM-RS of serving cell and CRS for interfering cells (assuming identity covariance matrix for the precoding of the interfering cells)
Channel Estimation
Perfect channel estimation on LTE R8 CRS, CSI-RS and DM-RS
PAPR
No constraint on per-antenna power imbalance
Inter-cell Interference Modelling
Dynamic interference with rank adaptation based on actual scheduling in interfering cells
CQI calculated based on MMSE receiver assuming identity covariance matrix for the interferers
Control Channel and Reference Signal Overhead
Overhead of control channel: 3 symbols per subframe
Overhead of LTE R8 CRS: RE
No overhead of DM-RS or CSI-RS
Traffic Model
Full buffer
Fading Model
SCM urban macro with low angular spread for 3GPP case 1
Antenna Configuration
4 x 2 MIMO vertically polarized antennas eNB: 4 lambda spacing, UE: 0.5 lambda spacing
Ideal antenna calibration 12 12

13. Appendix (Simulation parameters)
Home eNB parameters
Parameter
Assumption
HeNB Frequency Channel
Reuse 1 with Macro
System Bandwidth
10 MHz
Min/Max HeNB Transmit Power
0/20 dBm
Distance-dependent Path Loss
Dual-stripe model
Exterior Wall Penetration Loss
20 dB
Lognormal Shadowing
Similar to UMTS 30.03, B
Shadowing Standard Deviation
4 dB
Auto-correlation Distance of Shadowing for HeNB (optional)
3 m
HeNB Antenna Gain
5 dBi
Minimum Distance
between UE to HeNB
Home eNB cluster parameters
N (number of cells per row ) 10
M (number of clusters per sector) 1
L (number of floors per cluster)
R (deployment ratio )
10/25/50/75/100 %
P (activation ratio)
100 %
Probability of Macro UE being indoors
(%)
Sector area = 5002/sqrt(3)/2 m2
HeNB cluster area = 120 X 70 m2
HeNB cluster area/sector area = %
Geometry of dual-stripe model 13

14. Appendix (Simulation parameters)
Dual-stripe model
Cases
Path Loss (dB)
UE to
Macro
eNB
(1) UE is outside
PL (dB) = log10R, R in m
(2) UE is inside an apt
PL (dB) = log10R + Low, R in m
Home
(3) Dual-stripe model: UE is inside the same apt stripe as HeNB
PL (dB) = log10R + 0.7d2D,indoor n ((n+2)/(n+1)-0.46) + q*Liw
R and d2D,indoor are in m
n is the number of penetrated floors
q is the number of walls separating apartments between UE and HeNB
In case of a single-floor apt, the last term is not needed
(4) Dual-stripe model: UE is outside the apt stripe
PL (dB) = max( log10R, log10R) + 0.7d2D,indoor
n ((n+2)/(n+1)-0.46) + q*Liw + Low
(5) Dual-stripe model: UE is inside a different apt stripe
PL(dB) = max( log10R, log10R) + 0.7d2D,indoor
n ((n+2)/(n+1)-0.46) + q*Liw + Low,1 + Low,2
(6) Dual-stripe model or 5x5 Grid Model: UE is within or outside the apartment block
PL(dB) = log10(R/1000), R in m
This is an alternative simplified model based on the LTE-A evaluation methodology which avoids modelling any walls.
(3)
(4)
(5)
Liw is the penetration loss of the wall separating apartments, which is 5 dB.
0.7d2D,indoor takes account of penetration loss due to walls inside an apartment. 14