Overview

 Cellular design
18-759: Wireless Networks  Frequency Reuse
Lecture 7: Cellular Networks  Capacity and Interference
 Elements of a cellular network
 How does a mobile phone take place?
Dina Papagiannaki & Peter Steenkiste  Paging
 Handoff
Departments of Computer Science and
 Frequency Allocation
Electrical and Computer Engineering  Traffic Engineering
Spring Semester 2009
http://www.cs.cmu.edu/~prs/wireless09/
Peter A. Steenkiste & Dina Papagiannaki 1 Peter A. Steenkiste & Dina Papagiannaki 2

The advent of cellular networks The cellular network design

 Mobile radio telephone system was based on:  Simplest layout

» High power transmitter/receivers » Adjacent antennas not
» Could support about 25 channels equidistant – how do d
you handle users at the
» in a radius of 80 Km edge of the cell? d √2d
 To increase network capacity:
» Multiple low-power transmitters (100W or less)
d
» Small transmission radius -> area split in cells
» Each cell with its own frequencies and base station  Ideal layout d
» Adjacent cells use different frequencies
» The same frequency can be reused at sufficient distance
 Implications on mobile handsets (per L1)

Peter A. Steenkiste & Dina Papagiannaki 3 Peter A. Steenkiste & Dina Papagiannaki 4

The hexagonal pattern Cell site

 A hexagon pattern d
can provide
equidistant access to R
neighboring cell
towers
 d = √3R
 In practice, variations
from ideal due to
topological reasons

Peter A. Steenkiste & Dina Papagiannaki 5 Peter A. Steenkiste & Dina Papagiannaki 6

Page 1
 Frequency reuse Minimum separation?

 Each cell features one base transceiver

 Through power control cover the cell area
while limiting the power leaking to other co-
frequency cells
 Frequency reuse not possible for adjacent
towers!
 The number of frequency bands assigned to a
cell dependent on its traffic

Peter A. Steenkiste & Dina Papagiannaki 7 Peter A. Steenkiste & Dina Papagiannaki 8

Frequency reuse characterization Capacity and Interference

 S = Total # of duplex channels available for use
 D = minimum distance between centers of co-  k = Total # of duplex channels per cell
channel cells
 N = Cluster of cells which collectively use the complete
 R = radius of cell set of available frequencies
 d = distance between centers of adjacent cells
 N = number of cells in a repetitious pattern,
i.e. reuse factor
 If a cluster is replicated M times within the system, the
N = I 2 + J 2 + (I × J), I,J = 0,1,2,3,...
€
total # of duplex channels C can be used as a measure of
capacity
D or
D
= 3N = N
R d
Peter A. Steenkiste & Dina Papagiannaki 9 Peter A. Steenkiste & Dina Papagiannaki 10

€
€ €

Tradeoff How to Increase Capacity?

 If N ↓ ⇒ k ↑ since S is a constant
∴ M ↑ for a fixed geographical area if the same  Adding new channels
cell radius is maintained
 Frequency borrowing
⇒ Capacity increases
 Microcells
» Antennas on top of buildings, even lamp posts
 PRICE PAID: Co-channel interference ↑ » Form micro cells with reduced power
» Good for city streets, roads and inside buildings
 NOTE: To reduce co-channel interference

∴ Capacity ↓ since kN = S = fixed

 There is a trade-off between capacity and interference
reduction

Peter A. Steenkiste & Dina Papagiannaki 11 Peter A. Steenkiste & Dina Papagiannaki 12

Page 2
 Cell splitting Cell sectoring

 Cell size ~ 6.5-13Km, Minimum ~ 1.5Km

 Cell sectoring
 Requires careful power control and possibly » Cell divided into wedge shaped sectors
more frequent handoffs for mobile stations » 3-6 sectors per cell, each with own channel set
 A radius reduction by a factor of F reduces » Subset of cell’s channel, use of directional antennas
the coverage area and increases the required
number of base stations by a factor of F 2

Peter A. Steenkiste & Dina Papagiannaki 13 Peter A. Steenkiste & Dina Papagiannaki 14

Elements of a cellular system Channels

 Base Station (BS): includes antenna, a  Control channels to exchange information

controller, and a number of transceivers for regarding setup and call maintenance.
communicating on the channels assigned to Establishing relationship between mobile and
that cell closet BS.
 Controller is used to handle the call process  Traffic channels carry voice and data
between the mobile unit and the rest of the connections between users
network  5% of channels for control/95% for traffic
 MTSO: mobile telecommunications switching
office, serving multiple BSs. Connects calls
between mobiles and to the PSTN. Assigns
the voice channel, performs handoffs, billing

Peter A. Steenkiste & Dina Papagiannaki 15 Peter A. Steenkiste & Dina Papagiannaki 16

Call progression Call progression

(a) Monitor for strongest signal (b) Request for connection (c) Paging (d) Call accepted

Peter A. Steenkiste & Dina Papagiannaki 17 Peter A. Steenkiste & Dina Papagiannaki 18

Page 3
 Call progression Paging

 Broadcast mechanism to locate a target

mobile unit
 Normally, there is knowledge on a limited
number of cells where the mobile may be
(Location Area in GSM, Routing Area if data
packet sessions)
 GSM: neighbor cells grouped in Location
Area and subscriber only updates when
(e) Ongoing call (f) Handoff moving across. Paging restricted to the
Location Area itself.
» How do we assign cells to LAs?

Peter A. Steenkiste & Dina Papagiannaki 19 Peter A. Steenkiste & Dina Papagiannaki 20

Handoff Handoff between 2 cells

Base station A Base station B
 Could be network or client initiated, in which
case the client provided the network with
measurements
 Target performance metrics:
» Cell blocking probability
» Call dropping probability
» Call completion probability
» Probability of unsuccessful handoff
» Handoff blocking probability
» Handoff probability
» Rate of handoff
» Interruption duration
» Handoff delay

Peter A. Steenkiste & Dina Papagiannaki 21 Peter A. Steenkiste & Dina Papagiannaki 22

Handoff implementations Power Control

 GSM/W-CDMA  Received signal at mobile needs to be

» Inter-frequency handovers will measure the target sufficiently above background noise
channel before moving over
 Mobile transmission power minimized to
» Once the channel is confirmed OK, the network will
command the mobile to move and start bi-directional avoid co-channel interference, alleviate health
communication there concerns and save battery power
 CDMA2000/W-CDMA(same)  In SS using CDMA, need to equalize power
» Both channels are used at the same time – soft handover from all mobiles are the BS
 IS-95 (inter-frequency)
» Impossible to measure channel directly while
communicating. Need to use pilot beacons. Almost
always a brief disruption.

Peter A. Steenkiste & Dina Papagiannaki 23 Peter A. Steenkiste & Dina Papagiannaki 24

Page 4
 Open-Loop Power Control Closed-Loop Power Control

 No feedback from the BS (some SS systems)  Signal strength from mobile to BS adjusted
 BS transmits a pilot signal: according to performance metric on the
» Mobile acquired timing and phase reference for
reverse channel
demodulation » Reverse signal power level, received signal-to-noise ratio,
or received bit error rate
 Transmitted power in the reverse channel
assumed to be inversely proportional  BS makes the decision and communicates a
» Assumes forward and reverse link signal strength closely
power adjustment command to the mobile on
correlated a control channel
» Combats near-far problem in CDMA networks  Mobile provides information about received
 Features: signal strength to the BS, and BS adjusts
» Not as accurate as closed loop power accordingly
» Quick adjustment to rapid signal strength fluctuations
http://www.springerlink.com/content/l48275725160w472/fulltext.pdf
Peter A. Steenkiste & Dina Papagiannaki 25 Peter A. Steenkiste & Dina Papagiannaki 26

Fixed Channel Assignment Dynamic Channel Assignment

(FCA) (DCA)

 Each cell is allocated a predetermined set of voice  Voice channels are not assigned or allocated to
channels. different cells permanently. Instead each time a
 Any call attempt within the cell can only be served request is made, the serving BS requests a channel
by the unused channels in that cell from the MTSO.
 If all the channels in that cell are being used the  MTSO allocates a channel to the requested cell
call is blocked → user does not get service following an algorithm that takes into account the
likelihood of future blocking within the cell, the
 A variation of FCA: the cell whose channels are all freq. of use of the candidate channel, the reuse
being used is allowed to borrow channels from the distance of the channel, and other cost functions.
next cell. MTSO supervises this operation.
 MTSO only allocates a channel if it is available and
not being used in the restricted distance for co-
channel interference
Peter A. Steenkiste & Dina Papagiannaki 27 Peter A. Steenkiste & Dina Papagiannaki 28

FCA/DCA comparison Hybrid Channel Assignment

 Advantage of DCA: Likelihood of blocking  In hybrid channel assignment strategies, the

decreases and trunking efficiency increases concepts of fixed and dynamic channel
 Disadvantage of DCA: MTSO should collect real- assignment schemes are combined.
time data on channel occupancy, traffic  In addition to the fixed set of channels
distribution and RSSI of all channels on a assigned to each cell, there are channels in a
continuous basis. global pool for use in case of shortage.
 Overhead in terms of storage and computational
load on the system.

Peter A. Steenkiste & Dina Papagiannaki 29 Peter A. Steenkiste & Dina Papagiannaki 30

Page 5
 Hybrid Channel Assignment Dynamic Load balancing

 When a call requires service from a cell and  Channel borrowing from neighboring cells.
all of its fixed channels are busy, a channel  co-channel interference important
from the global pool is assigned to the call.
 When a channel is borrowed or dynamically
 The ratio of fixed to global (dynamic) assigned, the interference level in the co-
channels is a significant parameter which channel cells within the reuse distance
defines the performance of the system. should be examined.
 Ratio is a function of traffic load and would  If the interference level is going to be too high, the
vary over time according to the offered load corresponding channels of those co-channel cells should
also be locked.
estimations.
 This not only decreases the capacity of the overall
 Superior performance with non-uniform traffic system, but increases the computational load on the
system as well.

Peter A. Steenkiste & Dina Papagiannaki 31 Peter A. Steenkiste & Dina Papagiannaki 32

Dynamic Load Balancing

with/without (directional) locking Traffic Engineering

 If the cell has L subscribers..

 … and can support N simultaneous users.
 If L<=N, nonblocking system
 If L>N, blocking system
 If blocking:
» What is the probability of a call being blocked?
» What N do I need to upper bound this probability?
» If blocked calls are queued, what is the average delay?
» What capacity is needed to achieve a certain average
delay?

Peter A. Steenkiste & Dina Papagiannaki 33 Peter A. Steenkiste & Dina Papagiannaki 34

Trunking theory terminology Thunking theory terminology

 Set-up Time: The time required to allocated a  Load: Traffic intensity across the entire
trunked radio channel to a requesting user. trunked radio system, measured in Erlangs.
 Blocked Call (Lost Call): Call which cannot be  Grade of Service (GOS): A measure of
completed at time of request, due to congestion specified as the probability of a
congestion. call being blocked (for Erlang B), or the
 Holding Time: Average duration of a typical probability of a call being delayed beyond a
call. Denoted by h (in seconds). certain amount of time (for Erlang C).
 Traffic Intensity: Measure of channel time  Request Rate: The average number of call
utilization, which is the average channel requests per unit time. Denoted by λ calls per
occupancy measured in Erlangs. second.

Peter A. Steenkiste & Dina Papagiannaki 35 Peter A. Steenkiste & Dina Papagiannaki 36

Page 6
 Trunking theory Simple example

 Traffic intensity:  A cell has a capacity of 10 channels

A = λh  In 1 hour it received 97 calls lasting 294
(average number of calls received during the minutes in total
average holding time)  The rate of calls per min = 97/60
 If channel capacity is N – system can be seen  The average holding time = 294/97
as a multiserver queuing system  A = (97/60) x (294/97) = 4.9 Erlangs
λh = ρN
ρ is server utilization, fraction of time server  Mean number of calls in progress is 4.9
is busy
 Mean number of channels engaged is 4.9
 A also average number of channels required

Peter A. Steenkiste & Dina Papagiannaki 37 Peter A. Steenkiste & Dina Papagiannaki 38

Infinite source LLC –

Cellular network design Grade of Service

 Sized to sustain the average demand in the AN

busy hour (not peak demand!)
 Based on carried traffic and not offered!  Erlang B formula P = N N! x
A
 Model depends on: ∑ x!
» How are blocked calls handled? x= 0
– Could be put in a queue (lost calls delayed) A = offered traffic, N = #servers, P = blocking
– Rejected or dropped probability
1. user may hang up and try again after some random
time interval – lost calls cleared (LCC)
2. User repeatedly attempts – lost calls held (LCH)
€
» Number of traffic sources
– Finite or infinite?
– Infinite source assumption reasonable when sources
at least 5 to 10 times the capacity of the system
Peter A. Steenkiste & Dina Papagiannaki 39 Peter A. Steenkiste & Dina Papagiannaki 40

Example – Erlang B Example – Erlang B

N N

A larger capacity system is more efficient than a smaller-

capacity one for a given grade of service
Peter A. Steenkiste & Dina Papagiannaki 41 Peter A. Steenkiste & Dina Papagiannaki 42

Page 7
 Example – Erlang B What’s Next?

N  Cellular Protocol Standards

 Mobile IP
 Wireless Application Protocol (WAP)
30%

10%

A larger capacity system is more susceptible to an

increase in traffic
Peter A. Steenkiste & Dina Papagiannaki 43 Peter A. Steenkiste & Dina Papagiannaki 44

Not Used FCA vs. DCA

FCA DCA
 Performs better under heavy  Performs better under light/
traffic moderate traffic
 Low flexibility in channel  Flexible allocation of channels
assignment
 Not always maximum channel
 Maximum channel reusability
reusability
 Low computational effort
 High computational effort
 Low call set up delay
 Low implementation  Moderate to high call set up delay
complexity  Moderate to high implementation
 Low signaling load complexity
 Sensitive to time and spatial  Moderate to high signaling load
changes  Insensitive to time and spatial
 High forced call termination changes
probability
 Low to moderate forced call
 Suitable for large cell termination probability
environment
 Suitable for micro-cell
Peter A. Steenkiste & Dina Papagiannaki 45 Peter A. Steenkiste & Dina Papagiannaki
environment 46

Handoff Strategies
 First Generation Analog Systems (AMPS):
 Signal strength measured by BS and supervised by the MTSO
 Each BS constantly monitors the signal strengths of all of its reverse
voice channels to determine relative location of each MU w.r.t. BS.
 Locator Receiver at each BS measures RSS of users in neighboring
cells
 Second Generation Systems (TDMA  IS-54/ GSM/
PHP):
 Mobile Assisted Handoff (MAHO)
 Every MS measures the received power from surrounding BS and
continually reports these to the serving BS
 When the power received from BS of other cells > power received
from the serving BS, then handoff initiated
 advantage: Much faster than AMPS handoffs

Peter A. Steenkiste & Dina Papagiannaki 47

Page 8