2. OutlineOutline
CDMA overview and inter-cell effects
Network capacity
Sensitivity analysis
– Base station location
– Pilot-signal power
– Transmission power of the mobiles
Numerical results
3. How to match cell design to user
distribution for a given number of base
stations?
– CDMA network capacity calculation
– Reverse signal power and power control
– Pilot-signal power
– Base station location
Problem StatementProblem Statement
4. CDMA Capacity IssuesCDMA Capacity Issues
Depends on inter-cell interference and
intra-cell interference
Complete frequency reuse
Soft Handoff
Power Control
Sectorization
Voice activity detection
Graceful degradation
5. Relative Average Inter-CellRelative Average Inter-Cell
InterferenceInterference
[ ]
)Area(
10E
.deviationstandardandmean
zerohasandshadowing,todue
nattenuatiodecibeltheis
exponent.losspaththeis
102
j
j
j
ζ
ii
s
i
C
n
ω
|ζχ
σ
ζ
m
i
=
= −
= ∫∫
j
j
C
j
i
m
i
ζm
j
ji x,ydAω
/χx,yr
x,yr
I )(
)(
10)(
E 2
10
7. Soft HandoffSoft Handoff
[ ]
[ ]
[ ]
[ ]∫∫
∫∫
∫∫
∫∫
<=
<=
<=
<=
(c)region
1010210
(c)region
1010210
(b)region
1010210
(a)region
1010210
)(101010E
)(101010E
)(101010E
)(101010E
x,yωdAr|rχ
r
r
I
x,yωdAr|rχ
r
r
I
x,yωdAr|rχ
r
r
I
x,yωdAr|rχ
r
r
I
jkk
kjj
ikk
ijj
ζm
j
ζm
ki
ζ
m
i
m
k
ki
ζm
k
ζm
ji
ζ
m
i
m
j
ji
ζm
i
ζm
ki
ζ
m
i
m
k
ki
ζm
i
ζm
ji
ζ
m
i
m
j
ji
8. Inter-Cell Interference FactorInter-Cell Interference Factor
.toequalcellinceinterferen
averagerelativeaproducecellinusers
.celltocellfrom
factorceinterferencell-interuserper
jij
j
ji
κni
jn
ij
κ
10. Network CapacityNetwork Capacity
.,...,1for
,0subject to
capacity)(network,max
1
1
),...,( 1
Mi
cκnn
n
eff
M
j
jiji
M
i
i
nn M
=
≤−+ ∑
∑
=
=
Transmission
power of mobiles
Pilot-signal power
Base station
location
11. Power Compensation FactorPower Compensation Factor
Fine tune the nominal
transmission power of
the mobiles
PCF defined for each
cell
PCF is a design tool
to maximize the
capacity of the entire
network
12. Power Compensation Factor (PCF)Power Compensation Factor (PCF)
Interference is linear in PCF
Find the sensitivity of the network
capacity w.r.t. the PCF
.,...,1for)(
10
E
1
2
10
Miβc
β
κβ
nn
dAω
/χr
rβ
I
ieff
M
j i
jij
ji
j
C i
m
i
ζm
jj
ji
j
j
=≤+
=
∑
∫∫
=
13. Sensitivity w.r.t. pilot-signal powerSensitivity w.r.t. pilot-signal power
Increasing the pilot-signal power of one
cell:
– Increases intra-cell interference and decreases
inter-cell interference in that cell
– Opposite effect takes place in adjacent cells
14. Sensitivity w.r.t. LocationSensitivity w.r.t. Location
Moving a cell away from neighbor A and
closer to neighbor B:
– Inter-cell interference from neighbor A
increases
– Inter-cell interference from neighbor B
decreases
15. Optimization using PCFOptimization using PCF
.,...,1for
,0)(
,1subject to
capacity)(network,max
1
max
1
Mi
βc
β
κβ
nn
ββ
n
ieff
M
j i
jij
ji
M
i
i
β
=
≤−+
≤≤
∑
∑
=
=
16. Optimization using LocationOptimization using Location
.,...,1for
,0
),(
subject to
capacity)(network,max
)(
1
1
Mi
c
β
LCκβ
nn
n
i
eff
M
j i
ijjij
ji
M
i
i
L
=
≤−+ ∑
∑
=
=
17. Optimization using Pilot-signal PowerOptimization using Pilot-signal Power
.,...,1for
,0
),(
subject to
capacity)(network,max
)(
1
1
Mi
c
β
LCκβ
nn
n
i
eff
M
j i
ijjij
ji
M
i
i
T
=
≤−+ ∑
∑
=
=
19. Twenty-seven Cell CDMATwenty-seven Cell CDMA
NetworkNetwork
Uniform user
distribution profile.
Network capacity
equals 559
simultaneous
users.
Uniform placement
is optimal for
uniform user
distribution.
20. Three Hot Spot ClustersThree Hot Spot Clusters
All three hot spots
have a relative user
density of 5 per
grid point.
Network capacity
decreases to 536.
Capacity in cells 4,
15, and 19,
decreases from 18
to 3, 17 to 1, and 17
to 9.
21. Optimization using PCFOptimization using PCF
Network capacity
increases to 555.
Capacity in cells 4,
15, and 19,
increases from 3 to
12, 1 to 9, and 9 to
14.
Smallest cell-
capacity is 9.
22. Optimization using Pilot-signal PowerOptimization using Pilot-signal Power
Network capacity
increases to 546.
Capacity in cells 4,
15, and 19,
increases from 3 to
11, 1 to 9, and 9 to
16.
Smallest cell-
capacity is 9.
23. Optimization using LocationOptimization using Location
Network capacity
increases to 549.
Capacity in cells 4,
15, and 19,
increases from 3 to
14, 1 to 8, and 9 to
17.
Smallest cell-
capacity is 8.
24. Combined OptimizationCombined Optimization
Network capacity
increases to 565.
Capacity in cells 4,
15, and 19,
increases from 3 to
16, 1 to 13, and 9 to
16.
Smallest cell-
capacity is 13.
25.
26.
27. Combined Optimization (m.c.)Combined Optimization (m.c.)
.,...,1for
,
,0)(
),(
,1subject to
capacity)(network,max
min
1
max
1
Mi
nn
βc
β
LCκβ
nn
ββ
n
i
ieff
M
j i
ijjij
ji
M
i
i
T,L,β
=
≥
≤−+
≤≤
∑
∑
=
=
28.
29.
30. Combined Optimization (m.c.)Combined Optimization (m.c.)
Network capacity
increases to 564.
Capacity in cells 4,
15, and 19,
increases from 3 to
17, 1 to 17, and 9 to
17.
Smallest cell-
capacity is 17.
31.
32.
33.
34. ConclusionsConclusions
Solved cell design problem: given a user
distribution, found the optimal location
and pilot-signal power of the base stations
and the reverse power of the mobiles to
maximize network capacity.
Uniform network layout is optimal for
uniform user distribution.
Combined optimization increases network
capacity significantly for non-uniform user
distribution.
Editor's Notes
We are interested in solving the cell design problem. Given a user distribution profile what are the best locations of the base stations, the pilot-signal power and the transmission power of the mobiles to maximize capacity.
I will start with an overview of CDMA. I will describe what I mean by capacity and find the sensitivity of capacity to base station location, pilot signal power and transmission power of the mobiles. I will use the sensitivity analysis to maximize capacity. I will introduce mobility to differentiate between new calls and handoff calls. I will describe a call admission control algorithm that will guarantee the blocking probabilities for an arbitrary traffic distribution. Finally I will analyze the network performance by calculating the maximum throughput the network can achieve.
CDMA is interference limited. It’s capacity depends on the inter-cell interference and the intra-cell interference. All mobiles use the same frequency not only in the same cell but in all cells. So no frequency planning is necessary. When users move from one cell to another they switch channels. This is called handoff. In CDMA a user will communicate with both base stations near the boundary and will not relinquish its old channel until the new one is well established. Power control is a necessity in CDMA to solve the near-far problem. The signal of users close to a base station will drown out the signals of users that are far from the base station. So power control ensures that all users’ signals are received at the base station with the same level. Sectorization gives a gain of a factor of almost 3. During silent periods in two-way conversations the transmission rate is reduced so voice activity detection give us a gain of 2.6. Finally CDMA does not have a hard limit on the number of users that the system can admit. The more users the more interference so we get a graceful degradation in quality.
Consider cells i and j. This user in cell j is at distance rj from its own base station and ri from base station i. The inter-cell interference from cell j to cell i is given by this equation. The numerator is the gain adjustment through power control. The denominator is the propagation loss to base station i. This equation is for hard handoff.
READ SLIDE
The interference on cell i is now given by these equations. To calculate Iji we divide the cells into grids and calculate the integrals numerically. This could be a computationally intensive task. But it depends on the user distribution profile.
Therefore we define the per user interference factor kappa ji. For n_j users in cell j, the inter-cell interference to i is n_j kappa ji. So we only calculate the kappa ji once.
The signal to interference density ratio required for a given BER give these inequalities. From each we get this inequality for cell i. This has to hold for i equal 1 to M. These set of inequalities have to be satisfied by the ni. This feasible set forms the capacity region. The capacity region is the set of simultaneous users that we can have that satisfy the Eb over Io constraint.
Starting with the transmission power parameter we introduce the notion of power compensation factor.
Beta J is the power compensation factor for cell j. Relative average inter-cell interference is linear in the Beta j so adjusting the beta j of one cell does not require the recalculation of the inter-cell interference factors. So the capacity region for the network is now given by these inequalities.
Increases the cell coverage increases the number of users in that cell. Find the effect of increasing pilot-signal power of one cell on the capacity of the network
Find the effect of moving one cell on the capacity of the network
So now we have found all the derivatives… we want to use them to maximize capacity.
Use the derivatives in steepest descent algorithm
Next I will introduce a call admission control algorithm that will guarantee blocking probability for arbitrary traffic distribution.
Next I will introduce a call admission control algorithm that will guarantee blocking probability for arbitrary traffic distribution.
Uniform network optimal for uniform user dist. PCF best of three parameters. Increasing pilot signal power is better. Imposing a minimum constraints does not penalize total network capacity when used with our combined optimization but cases a severe penalty for a uniform network topology.