Transcript Macrocell/Microcell Architectures in CDMA Systems

Capacity and Coverage in
Two-Tier CDMA Cellular Networks
Shalinee Kishore
Department of Electrical Engineering
Princeton University
Supported by: AT&T Labs Fellowship
Advisors: H. V. Poor, S. Schwartz,
L. J. Greenstein (WINLAB)
November 25, 2002
Two-Tier System: Macrocells and Microcells
Macrocell
Microcell
Macrocells - cells in the traditional cellular system
• Cell radii are 1 to 10 km.
• Base stations are costly, antenna tower heights  30 m.
Microcells - smaller cells embedded within macrocells
• Cell radii are less than 1 km.
• Base stations are compact, low-cost, at heights of ~10 m.
Why Microcells? An Example
Desired Coverage
High Density of Users
Actual Coverage
Due to high-user-density regions, actual performance of
macrocell falls short of desired performance.
Why Microcells? (Cont’d)
Other Reasons: Users can be separated based on
Fast moving users  Macrocell
• mobility
Slow moving users  Microcell
Voice users  Macrocell
• desired data rates
Data users  Microcell
Microcells in Single-Frequency
Code Division Multiple Access (CDMA) Systems
• CDMA is employed in current cellular phones in US and is
standard for third generation systems worldwide.
• CDMA uplink (user-to-base): users assigned random codes.
• Every user’s signal interferes with signals from every other user.
• In single-tier systems (macrocells only), there is in-cell and out-of-cell
interference.
• CDMA downlink (base-to-user): base station uses orthogonal
codes to transmit to all in-cell users.
• In single-tier systems, there is ideally only out-of-cell interference.
• Dispersive wireless channels cause loss-of-orthogonality, leading to
in-cell interference.
• In both the uplink and downlink of two-tier systems, there is
additionally cross-tier interference.
Two Classes of CDMA Microcells
Hotspots:* Small cells
embedded inside a larger
macrocell to provide
coverage in small region
with high user/traffic
density or poor coverage.
Handoff between tiers.
Clusters/Overlay: Small
cells that tesselate and span
almost all of macrocell
coverage area. No handoff
between tiers.
- Single-frequency (near-far problem)
- Dual-frequency (spectral efficiency issues)
* Focus of our research
Previous Work on CDMA Microcells
Hotspots
• Shapira, “Microcell Engineering in CDMA Cellular Networks,” IEEE Transactions on
Vehicular Technology, 1994.
• Gaytan and Rodriguez, “Analysis of Capacity Gain and BER Performance for CDMA
Systems with Desensitized Embedded Microcells,” ICUPC, 1998.
• Wu, et al., “Performance Study for a Microcell Hot Spot Embedded in CDMA Macrocell
Systems,” IEEE Transactions on Vehicular Technology, 1999.
Overlays
• I, et al., “A Microcell/Macrocell Cellular Architecture for Low- and High-Mobility Wireless
Users,” IEEE Journal on Selected Areas in Communications, Vol. 11, Issue 6, Aug. 1993.
• Hamalainen, et al., “Performance of CDMA Based Hierarchical Cell Structure Network,”
IEEE TENCON, 1999.
• Ghaleb, et al. “Tiered Services/Private System Support for CDMA Systems,” VTC, 1999.
Research Goals
Expand understanding of Macrocell/Microcell architectures in
CDMA networks.
• Develop new methods of analysis for evaluating such
systems.
• Evaluate impact of propagation, user distribution,
channel fading, maximum transmit power constraints,
and dispersion on uplink and downlink capacity and
coverage area.
• Devise techniques, tradeoffs, and engineering rules for
performance improvement and system deployment.
Summary of Thesis
• Ideal Conditions:
No variable fading of user signal powers
Uplink: no transmit power constraint
Downlink: no in-cell interference
- Single-Macrocell/Single-Microcell (Two-Cell) System
- Multiple-Macrocell/Multiple-Microcell (Multi-Cell) System
- Other Issues in Two-Cell Systems
1)
2)
3)
4)
Effect of soft-handoff
Effect of voice activity detection
Effect of propagation parameters
Microcells as Data Access Points (DAP’s)
Summary of Thesis (Cont’d)
• Non-Ideal Conditions:
- Uplink Capacity and Coverage
1) Effect of transmit power constraints
2) Effect of received power fading
- Downlink Capacity: No Multiuser Detectors
1) Effect of Channel Dispersion
2) Alternative methods of power control
Two-Cell System:
Uplink and Downlink in Ideal Conditions
Uplink Capacity of Two-Cell System: Problem Statement
Given:
• CDMA system with single macrocell and single microcell
• Matched filter receiver and SINR-based power control
• Probability density of user location over coverage region
• Processing gain (W/R) and desired SINR (G )
• Propagation characteristics, including shadow fading
• Criterion for base station selection (e.g., strongest path gain,
minimum required transmit power)
• Hard-handoff: each user communicates with only one base
Determine:
Uplink user capacity (number of simultaneous voice users)
Feasibility
In order to meet SINR requirements for N M macrocell and
N microcell users,
(K  NM )(K  N )  IMI
where
Cross-Tier
Interference
Terms
W
K
1
RG
TMj
IM  
j  Tj
Tj
I  
j M TMj
(Feasibility)
(single-cell pole capacity)
TMj
j  

Tj
Tj
1
j M 

TMj
Tij = Transmission gain from base i to user j,  = desensitivity
Transmission Gain (Path Gain) Model
  b 2
H  d   , d  b
T    4
H  b   , d  b
  d 
T = Transmission Gain
d = Distance Between User and Base
b = Breakpoint Distance of Median Path Gain
H = Proportionality Constant, Accounts for Antenna Gains and Wavelength
 = Lognormal Shadow Fading
Finding the CDF for one term of IM: Let TMj/Tj = vM
2
max( bM
,min( zmax ,
P [v M  v | R1, ] 

2
bM
v
w
))
 max
w max
f
ZW
min( w max ,max( b2 ,z
( z,w ) dz dw

))
v
w here
  z  D 2  w  h2  hM2
 
1
 f XY 
 
,
2D
4Dg( z,w )  
 
fZW ( z,w )  g( z,w )  

2
2
2
 z  D  w  h  hM



1


,
  f XY 

2
D
4
Dg
(
z
,
w
)



w here
f XY ( x , y )  Density of Users over locations ( x , y )
Exact analysis is doable but extremely complicated.
Simpler Analysis: Mean Approximation
• Since IM and I are sums, they converge fairly tightly to
their means.
• Instead of computing distribution of IM and I  , we
compute their mean values
• E[IM ]  NE[v M ]  Nv M and E[I ]  NME[v  ]  NMv 
• Obtain the following requirement on NM and N:
K (K  N M )
N 
K  NM (1  v Mv  )
Number of Microcell Users
Capacity Contours for Single-Macrocell/Single-Microcell System
Exact Analysis
Simulation
Approximation
Number of Macrocell Users
Multicell System:
Under Ideal Conditions
Multicell Systems: Key Results
• Showed total user capacity is maximum when there are an
equal number of users served by each cell.
• Showed total user capacity is approximately linear in L and
M (number of macrocell bases) for L small. Specifically,
N(L,M )  K  (NTM  K )M  (NT  K )L
NTM and NT can be calculated using two-cell techniques.
Mutlicell Systems: Key Results (Cont’d)
• Derived a simple and reliable approximation for NT:
NT
2K

1  v Mv 
• Similar analysis yields reliable approximation for NTM.
Total Average Number of Users, 95% Feasibility
Single-Macrocell/Multiple-Microcell System
Simulation Results,  s error bar
Linear Approximation
L, Number of Microcells
Total Average Number of Users, 95% Feasibility
9-Macrocell/Multiple-Microcell System
Simulation Results,  s error bar
Linear Approximation
L, Number of Microcells
Other Issues in Ideal Two-Cell Systems:
Soft-Handoff,
Voice Activity Detection,
Propagation Parameter Sensitivity,
and Microcells as DAPs
Other Issues in Two-Cell Systems: Key Results
•
Effect of Soft-Handoff: Both base stations receive each
user’s signal; two signals added
using maximal ratio combining.
- Developed analytical methods to approximate user
capacity under soft-handoff.
- Showed user capacity increases by at most 20% over
hard-handoff.
Other Issues in Two-Cell Systems: Key Results (Cont’d)
• Effect of Voice Activity Factor: Let a be the fraction of time
voice users speak. Under voice activity detection, mean
approximation contour is modified as:
~
ˆ
ˆ
K (K  N M )
~
N 
~
ˆ
K  NM (1  v M v  )
N
NM
~
~
ˆ
w here, K  K  1  a , NM 
, and N  
.
a
a
• Sensitivity to Propagation Parameters: Fairly insensitive
Microcells as Data Access Points
DAP: Base station with limited coverage that provides
high-speed data access to users one-at-a-time.
Email, voice mail,
and fax to the
pedestrian
Downloading a map to
a passing car
Low bit-rate
cellular coverage
High bit-rate
DAP coverage
Examples of DAP’s: Infostations, Dedicated Short-Range Communications
(DSRC), and Intelligent Transportation Systems (ITS)
Problem Statement
Recall: Microcell coverage shrinks as desensitivity ( )
reduces.
Question: What happens when   0 and microcell
coverage area shrinks to that of a DAP?
Determine: Per-user throughput, tu , and total DAP
throughput, t , as functions of .
Normalized Average Throughput
Normalized Average Throughput (E[ t / W ]) Versus 
, Desensitivity
Uplink Capacity and Coverage:
Max Power Constraints
and Variable Power Fading
Maximum Power Constraints: Problem Statement
Given:
• A Single-Macrocell/Single-Microcell System
• User distribution
• Propagation model
• Pmax = Maximum transmit power level for any user
• dmax = Maximum distance over which users are distributed
• hW = Noise power
Determine:
Uplink user capacity as a function of Pmax and dmax
Maximum Power Constraints: Key Results
• Defined P [Outage] as
P [Outage] = (1-P [Feasibility]) + P [Feasibility] ·P [Transmit Power > Pmax].
• Presented uplink user capacity for given level of outage as
a function of a single, dimensionless parameter F, where
F
Pmax
hW
 b

 d max
4

 .

N, Total Number of Users, 5% Outage
Capacity in System with Max Power Constraints
P
F   max
 hW
 b 


d
 max 
4
Variable Power Fading: Background
• Thus far: considered infinitely-dispersive uplink channel 
user signal has constant output power after RAKE processing.
• Actual channels have finite number of paths with variation
about mean path power  user signal has variable fading.
• Can model fading with modified transmission gain:
Tij’ = kTij, k is a unit-mean random variable.
• Examine performance for four scenarios:
• Rural Area (RA) environment
• Typical Urban (TU) environment
• Hilly Terrain (HT) environment
• Uniform multipath channel
Uniform Multipath Channel
Channel Delay Profile
power
Height of each path is mean square
value of a Rayleigh random-variable.
delay
Lp
Number of Paths
• Diversity Factor (DF) measures the amount of multipath
diversity in channel. Computable for any delay profile.
• Uniform channel has DF = Lp.
• Non-uniform channels with Lp paths have DF < Lp. For example,
DFRA = 1.6, DFHT = 3.3, and DFTU = 4.0.
Variable Power Fading: Problem Statement
Given:
•
•
•
•
•
Single-macrocell/single-microcell system
Propagation model with variable fading
Pmax = Maximum transmit power level
dmax = Maximum distance over which users are distributed
hW = Noise power
Determine:
Uplink user capacity so that P[Outage] does not exceed g.
• for the three standard environments, i.e., RA, TU, and HT,
as functions of F.
• for any environment when F >> F* (unlimited terminal
power).
Variable Power Fading: Key Results
• Uplink capacity constant for RA, HT, and TU environments when
F < 0.1 and decreases sharply in F when F < 0.1.
• Capacity reduces by as much as 15% for the RA environment.
• When F >> F*, user capacity in uniform multipath channel can
be approximated as:
2K
N
, for Lp > 1.
Lp
1
v Mv 
Lp  1
• Showed uplink capacity is the same for channels with the same
DF.
DF
Non-Uniform
Delay Profile
Replace Lp in
with DF
Napprox
Obtaining NT for RA, HT, and TU Channels via the Uniform Channel
RA
HT
TU
Uplink Capacity
using Simulation
33
36
37
Uplink Capacity
using Approximation
(via Uniform Channel)
32.5
35.86
37.1429
Downlink Capacity:
Channel Dispersion
and Effect of Alternate Power Control
Downlink Capacity: Background
• CDMA downlink: Base stations transmit orthogonal
signals to users.
• Channel dispersion causes loss of orthogonality at user
terminals.
• Orthogonality factor, , captures loss-of-orthogonality of
user signals in a channel.   [0,1], where  = 0 when no
dispersion in channel and  = 1 when infinite dispersion.
•  can be computed from channel delay profile.
• Thus far: assumed  = 1 (infinite dispersion) but ideal
multiuser detectors removed all in-cell interference.
Downlink Capacity: Problem Statement
Given:
•
•
•
•
•
•
•
Single-macrocell/single-microcell system
Channel delay profile, i.e., orthogonality factor, .
Conventional receivers at user terminals
Base station k transmits total power PTk, k  { M, }
Macrocell user i assigned fraction xi of PTM
Microcell user j assigned fraction yj of PT
Downlink power control scheme for allocating xi and yj
Determine:
Downlink user capacity, number of simultaneous voice
users
Downlink Capacity: Key Results
• Recast uplink capacity, NT, as a function of .
• Capacity of any channel (  ) approximated using
capacity of uniform channel.
• For two of three power control strategies studied (uniform
and slow), overall capacity dominated by uplink for all .
• Under fast power control, user capacity can be
approximated (by relating  to u) as:
N
2K
1 
1
v Mv 
2
,
for   0.
• Fast power control leads to downlink capacity that is
higher than uplink.
Conclusion
• Analytical methods developed for estimating attainable
uplink user capacity in two-tier CDMA systems.
• Analysis done in progression from single-macrocell/singlemicrocell, to single-macrocell/multiple-microcells, to
multiple-macrocells/multiple-microcells.
• Results general with respect to system and propagation
parameters and accurate, as confirmed via simulation.
• Analysis extended to DAP, showing how microcells can
be modified to support high speed data.
• Computed effect of soft-handoff and voice activity
detection on uplink user capacity.
• Quantified effect of maximum power constraints on
coverage area and capacity.
• Used the uniform multipath channel to approximate the
uplink user capacity and downlink user capacity under fast
power control for finitely-dispersive channels.
• Demonstrated the importance of fast downlink power
control in two-tier CDMA systems.