Transmission and Telephony

1
Transmission and Telephony

• Multiplexing techniques
• Asynchronous
• SONET
• WDM
• Telephone Switching
• Digital Switches
• Telephone Network
• Traffic Analysis
• Cellular Networks

2
First Exam 10/2

• Exam will cover material in chapters1-4 and
  associated lectures
• Exam will be a mix of problems requiring analysis
  and calculation, and short answer questions
• You will have an hour and 10 minutes for the exam
• You can/should bring a calculator and one page of
  notes (both sides)

3
Multiplexing
(a)
(b)
A
A
A
A
Trunk group
B
B
B
MUX
MUX
B
C
C
C
C

• A single high capacity facility carries several
  logical paths
• Multiplexers combine and break out the signals

Figure 4.1
4
Frequency Division Multiplexing
A
f
W
0
0
W
0
W

• Individual signals relocated to non-overlapping
  bands
• Combined signal fits into channel bandwidth
• Telephone network used multiple levels

Figure 4.2
5
Digital Time Division Multiplexing
1
2
3
4
5
6
7
8
1
2
3
4
1 Frame

• Multiplexed link runs at a multiple of the
  channel rates
• Bits from input signals packed into a Frame
• Individual signals may be bit or byte interleaved
• Problems arise if the input streams are not
  running at the same rate.

6
A T1 Multiplexer
1
1
2
MUX
MUX
2
. . .
. . .
22
24
b
1
2
23
24
. . .
b
24
frame
24

• Multiplexer combines 24 64Kb/s channels.
• 8 bits from each channel and one framing bit in
  each 193 bit frame
• All channels are synchronized
• T1 link transmits 8,000 Frames/second (1.544 Mbs)
• E1 (ITU) standard combines 32 channels (2.048Mbs)

Figure 4.4
7
North American Digital Hierarchy
European Digital Hierarchy
Figure 4.5
8
Asynchronous Multiplexing
RM
R1
R2

• What happens if R1, R2, and RM are not
  synchronized?
• Slips. If R1 and R2 are not equal or RM is not
  exactly twice R1,R2 output has gaps or lost bits
• If synchronization isnt guaranteed, multiplexing
  must accommodate different rates
• Extra, Stuffed bits included so that they can
  be dropped in a slip
• RM gt needed to guarantee nothing is lost
• Stuffed bits are identifiable and ignored

9
Bit Stuffing
One Masterframe
RM
R1
1
0
g
f
F
e
E
d
D
0
0
c
C
J
I
H
G
F
E
D
C
B
A
R2
Control bits
Stuff bits

• Extra bits are added periodically
• one stuff bit for each channel that may contain
  data
• one control bit indicating if the stuff bit is
  used
• Extra bits must appear often enough to
  accommodate variance in rates
• RM must be enough larger than R1 R2 to
  accommodate the extra bits
• Frame positions of multiplexed signals are not
  synchronized.

10
Exercise

• A multiplexer takes 2 1Mb/s links and multiplexes
  onto a 2.2Mb/s link
• If it has a 28 bit master frame carrying 12 bits
  from each input plus 1 control bit and 1 stuff
  bit
• How long does it take to send one frame?
• What are the maximum and minimum rates for the
  inputs that can be supported?

11
Solution

• 28 bits at 2.2Mbits/second 12.7 microseconds
  per masterframe.
• If 12 bits of data are sent, the rate is
  12/282.2Mbits/second .943 Mb/s
• If 13 bits of data are sent the rate is
  1.021Mb/s.

12
Extracting Channels from Asynchronous Signals
MUX
DEMUX
remove tributary
insert tributary

• Impossible to locate and replace an individual
  signal without demultiplexing
• Accessing one signal may require several levels
  of unwrapping
• A significant problem with high capacity optical
  links.

13
Synchronous Optical Networking(SONET/SDH)

• Common Optical and Electrical signal formats to
  allow carrier interconnection
• Accommodates both North American (DS1-DS3) and
  ITU (E1-E4) rates
• Synchronous format Easy to extract individual
  signals
• Byte interleaved signal format
• Rates from 51-10,000 MHZ

14
SONET Multiplexing
Packing and Transport
STS-1
Parity Calculation
DS-3
Byte Interleaved Mux
DS-1
STS-N
Packing and Transport
STS-1
DS-2
Scrambler
E/O
OC-N
E1
STS-1
Packing and Transport
E4
STS-1
STS-1
STS-3C

• All inputs multiplexed to STS-1
• Can contain several lower level signals
• higher level signals are split into concatenated
  STS-1
• Multiplexer byte interleaves signal
• All but identification bytes scrambled for balance

15
SONET Hierarchy
16
SONET Add/Drop Multiplexing
ADM
MUX
DEMUX
insert tributary
remove tributary

• Synchronous format allows any signal to be
  extracted or inserted
• Add/Drop multiplexing is a single low cost
  operation at any level of the hierarchy

17
SONET Rings
a
OC-3n
OC-3n
b
3 ADMs
STS-n STS-n
c
OC-3n
physical loop net

• Optical links physically arranged in rings
• ADMs direct logical paths

Figure 4.10
18
ADMs build the logical network
a
(a)
(b)
OC-3n
OC-3n
b
c
OC-3n
3 ADMs connected in physical ring topology
19
Unidirectional SONET Ring
B
A

• SONET multiplexors connected in a ring by a pair
  of fibers
• One fiber normally carries traffic, other spare
  runs in the opposite direction
• Bidirectional connections go around the ring
• If there is a fiber cut, spare fiber completes
  the path.
• Half the capacity is unused.
• Connections can go through many multiplexers.

20
Bidirectional rings
A
B

• 2 fiber both fibers active, and 50 utilized.
• 4 fiber two active, two spare, additional
  sparing options (i.e bypass one bad pair)
• More efficient (shorter connections)
• Preferred in metro area networks

21
Interconnected SONET rings
Figure 4.13
22
SONET Hierarchy
Section Terminating Equipment
Section Terminating Equipment
Line Terminating Equipment
Path Terminating Equipment
Path Terminating Equipment
Line Terminating Equipment
Section
Section
Section
Line
Path

• Path end-to-end
• Line end-to-end for the level of multiplexing
• Segment regenerator to regenerator

23
SONET Layering
(b)

• SONET format is organized in 3 levels of links
• Each level has associated protocols for
  maintenance and control

Figure 4.14
24
STS1 Frame Format
8,000/second
8 bit bytes
Information Payload
Transport Overhead
C1
A1
A2
J1
Section Overhead
F1
B1
E1
B3
D3
D1
D2
C2
9 Rows
H1
H2
H3
G1
Path Overhead
K2
K1
B2
F2
D6
D5
D4
H4
Line Overhead
D7
D8
D9
Z3
D12
D11
D10
Z4
E2
Z2
Z1
Z5
87 Columns
3 Columns

• Overhead bits for section, line, and path
• Basic rate is overheadinformation
• 9?3?8?8000 9?87?8?8000
• 1.728?106 50.112?106 51.840 Mb/s
• Higher rates byte multiplex lower rates

25
Synchronous Payload Envelope
first octet
Pointer
frame k
87 columns
Synchronous Payload Envelope
9 rows
Pointer
last octet
frame k1
first column is path overhead

• Each Path carries a synchronous Payload Envelope
• One SPE can span frames (not synchronized)
• Pointers track where frames start/stop

26
Byte Stuffing
pointer value p
pointer value p
H1
H2
H3
H1
H2
H3
H1
H2
H3
H1
H2
H3
data
empty
4 frames
inverted
inverted
H1
H2
H3
H1
H2
pointer value p1
pointer value p-1
H1
H2
H3
H1
H2
H3
Positive Stuff (Data rate is slower than nominal)
Negative Stuff (Data rate is faster than nominal)
27
Packing lower rates virtual tributaries

• SPE is divided into 7 VT groups of 12?9 bytes per
  frame each.
• 7?123 87 columns of 9 bytes, SPE payload
• 3 columns of overhead, 2 unused

28
Multiplexing higher levels
STS-1
STS-1
STS-1
STS-1
Map
STS-3
STS-1
STS-1
STS-1
STS-1
Map
STS-1
STS-1
STS-1
STS-1
Map
Incoming STS-1 Frames
Synchronized New STS-1 Frames

• Section and line overheads terminated
• Incoming SPE mapped to a synchronized STS-1
• Signals byte interleaved

Figure 4.17
29
A Typical Circuit Switch

• Inputs/Outputs come through Interfaces
• Switching Fabric provides interconnection
• Control determines how to operate the switching
  fabric

Figure 4.21
30
Space Division Switching
M
Crosspoint
N

• Connects two signals separate in space (e.g. two
  wires)
• Crosspoints each capable of connecting two
  signals together.
• N inputs and M outputs uses N X M crosspoints
• Fully (Strictly) Non-blocking
• Folded (same inputs/outputs) switches use N2/2
  Crosspoints

31
A Multi-stage switch
2(N/n)nk k (N/n)2 crosspoints
kxn
nxk
N/n x N/n
1
1
1
kxn
nxk
N inputs
2
2
N outputs
N/n x N/n
kxn
nxk
2
3
3
? ? ?
? ? ?
? ? ?
kxn
nxk
N/n
N/n
N/n x N/n
k

• Switch has multiple stages of crossbar switches

32
How Many Crosspoints are needed for a 3 stage
non-blocking switch?
k
...
n-1 (busy)
n
n
n-1 (busy)
So, k ? 2(n-1)1
...
Nx 2N(2n-1) (2n-1)(N/n)2
Nx is minimum for n(N/2)½
Result of Charles Clos, often called a Clos
Network
33
How Much does it Matter?
CLOS networks save a lot, but big non-blocking
networks are still costly
34
Excercise

• Design a CLOS network for a 200 by 200 switch

35
Solution

• N 200, best n is square root of N/2 10.
• k 2n-1 19.
• The switch will have
• N/n (20) input stages of n by k (10 by 19).
• k (19) center stages of N/n by N/n (20 by 20),
• and N/n (20) output stages of k by n
• Total crosspoints are
• First stage 20x10x19 3800
• Center stage 19x20x29 7600
• Third stage 20x19x103800
• Total 15200, vs 40,000 for crossbar

36
How to reduce the number of crosspoints

• The Clos network has k (N/n)?(N/n) crosspoint
  switches. We could convert each to a clos
  network internally and save more
• By building networks that could block, the number
  of crosspoints can be much lower
• The average telephone is not busy 100 of the
  time so the chance that all endpoints are in use
  is very small.
• Some blocking is acceptable if chance of blocking
  is much less than chance of busy, dont answer,
  or broken equipment

37
Computing Blocking Probability (Lee)
If p is the probability that a link is busy and
q1-p is the probability that its available
Bpn
n links
n links
B1-qn
38
Blocking probabilities of a 3-stage network
1
p'
p'
2
p
p
p'
...
p'
k
p is the probability an incoming or outgoing link
is busy
p' is the probability a center stage link is busy
p ? n/k
B 1-(1- p ? n/k)2 k
39
Exercise network design

• Design a network to connect 2,048 endpoints that
  are 20 busy with blocking probability lt .001
  (.1)

40
Solution
assume n(N/2)½ (Not necessarily optimum) 32
B 1-(1- p ? n/k)2 k
.001 1-(1- .2 ? 32/k)2 k Solve for k
or, search for k, given small number of values
k 21 gives .00065, k20 is too large
Crosspoints 2Nk k(N/n)2 172032
41
Time Slot Interchange
TSI
Control Memory
Slot Counter
Buffer Memory
42
Time Slot Interchangers

• b (usually 8) bits switched in one frame
• Frames repeat f (usually 8,000) times/second
• Buffer memory is n locations of b bits wide
• Buffer memory accessed twice for each channel
• Tm lt 1/(2?f?n) Memory cycle time limits size
  and rate
• Number of input and output timeslots may be
  different
• Control memory is n locations of log2(n) bits
• In general each signal is delayed 1-2 frame times

43
Time Multiplexed Switches
Control Memory
N
N

• A different set of crosspoints are closed in each
  timeslot according to the control memory

44
Time Multiplexed Switches

• Switch connects a different configuration of
  inputs to outputs during each of n frames.
• Control memory has n locations of crosspoint
  configurations
• One configuration is Nlog2(N) bits assuming point
  to point connections only
• Memory and crosspoints accessed once each frame.
• Inputs and Outputs are synchronized (gate delays
  only)
• Size limited by crosspoints and device speed

45
Time-Space-Time Switches
N/n time switches k by n each
N/n time switches c by k each
k timeslots
N/n inputs n timeslots each
TSI
TSI
N by k
N by k
N/n by N/n
TSI
TSI
...
...
TSI
TSI

• Like a 3 stage space switch connecting N inputs
  to N outputs
• First stage is N/n n by k time switches
• Second stage is like k N/n by N/n switches
• Third stage is N/n k by n time switches
• Non-blocking if k2n-1
• Blocking in smaller switches can be computed
  using Lee formulas

46
Space-Time-Space Switches
k time switches
n timeslots like n switches
n timeslots like n switches
N/n inputs n timeslots each
TSI
N by k
N by k
N/n by k
k by N/n
TSI
...
TSI

• A 3 stage switch like the multi-stage space
  switch for N inputs and N outputs
• First space stage acts like n N/n by k switches.
• Time stage is k n by n switches
• Last stage is n k by N/n switches
• Same formulas for blocking apply

47
The Telephone Network

• A Call requires connections from several switches
• Switches decide how calls are routed to the
  destination
• Path between switches is almost always digital
• The switches communicate via signaling
• In band (Tones) or Out of band (Messages)

48
Signaling Message flow
Source
Signal
Go Ahead
Signal
Message
Release
Signal
Destination

• Messages cascade through switches
• Typical call procedure uses 4-5 messages
• Complicating issues
• Routing for 800 and other special numbers
• Rerouting around busy switches and trunks
• Managing network resources

49
Local Distribution System
Figure 4.33
50
Echo in telephone networks
Transmit pair
Original signal
Received signal
Hybrid transformer
Echoed signal
Receive pair

• Digital signals use separate transmit and receive
  paths (4 wire)
• Imperfect hybrid transformers cause echo
• Echo is returned to the speaker after traveling
  the length of the connection and back.
• Long connections must control echo to avoid
  speech quality problems

51
Digital Cross Connect
Digital cross-connect System
Channel-switched traffic (digital leased lines)
Local analog
Tie lines
Foreign exchange
Local digital
Local Switch
Digital trunks
Circuit-switched traffic

• Cross connect electronically rewires local
  connections
• Multiple levels of cross connects (DS0, DS1, DS3,
  )
• Cross connects and Multiplexers form a layer
  below the circuit switches.

52
Example Topologies
ADM
ADM
ADM
ADM
Physical SONET Topology using ADMs and DCCs
ADM
ADM
DCC
Logical Topology Switches see this topology
Figure 4.36
53
Signaling between switches
Office A
Office B
Trunks
Switch
Switch
Modem
Modem
Processor
Processor
Signaling

• Signaling links carry messages between switches
• Keeps user data and signaling separate (prevents
  fraud)
• Makes it easier to send more information
• Direct links between all pairs of switches are
  impractical
• Requires a Signaling Network

54
The SS7 Signaling Network
STP
STP
STP
STP
SSP
SSP
Signaling Network
Transport Network
SSP Service switching point (signal to
message) STP Signal transfer point (message
transfer) SCP Service control point (processing)

• Two kinds of traffic
• Trunk Signaling (ISUP/TUP)
• Control signaling (SCCP/TCAP)
• STPs implement a packet network to route messages
• Routing based on point codes (like IP addresses)
• Routing based on phone numbers (like domain names)

55
Layering of SS7 Protocols
Figure 4.42
56
Intelligent Network
Standard State Model
Intelligent Peripheral
Trigger
Switch
Switch
Switch (SSP)

• Intelligent Network is an architecture for
  programming services without programming switches
• Switch (SSP) implements a standard state model
  for a call
• At any transition, the switch can be told to
  trigger a message to an SCP
• The SCP can respond with an action and a new
  state
• One SCP can serve many switches
• Intelligent Peripheral provides user interaction
  functions (menus, speech recognition,
  announcements)

57
Example State Models (ITU)

58
Characteristics of IN

• Benefits
• New services without changing switches
• Service and data in one place (the SCP)
• More efficient networking than service nodes
• Problems
• Getting SSP models and triggers deployed
• Messaging overhead and SS7 cost
• Varying SCP implementations
• Example IN Services
• 800 calling, call screening, call centers,
  Virtual Private Network, Card Services, Personal
  number (follow me)

59
Concentration in networks
Many Lines
Fewer Trunks

• Network has fewer internal links than needed to
  support all the customers active at once
• Attempts to call may be blocked
• Blocking can be analyzed with standard models
• Model is always an approximation of real behavior

Figure 4.43
60
Trunk Usage
N(t)
all trunks busy
t
1
2
trunk
3
4
5
6
7
Figure 4.44
61
Erlang B Analysis

• If traffic follows a Poisson process
• In a small interval ?, probability of new request
  is ??
• probability of no new requests is 1-??
• And call holding time is a random variable X
• Offered loada?EX
• measured in Erlangs, represents average number of
  simultaneous calls.
• Utilization given by the Erlang B Formula

a offered load c number of trunks
62
Blocking (Erlang B)
trunks
Blocking Probability
10
9
8
7
1
6
2
3
4
5
offered load
Utilization ?(1-Pb)EX/c (1-Pb)a/c
Figure 4.45
63
Trunk Utilization
64
Routing and Efficiency
Trunk group
(b)
Tandem Switch 2
Tandem Switch 1
F
C
E
D
B
A
10 Erlangs between each pair
90 Erlangs

• 10 erlangs require 18 trunks
• 9 times 18 or 162 trunks total

• 90 Erlangs require 106 trunks
• 106 trunks total
• Higher blocking if wrong

How many trunks are required between each switch
and the Tandem? What is their utilization?
Figure 4.46
65
Typical Network Routing
Tandem Switch
Alternate Route
High Usage Route

• High Usage route designed to carry most traffic
• Traffic overflows to alternate route through
  tandem switch
• Total blocking is less than on either route.
• Traffic to alternate route is not Poisson
• Route sees arrivals only when high usage route is
  full
• Cant use the Erlang B Formula to predict blocking

Figure 4.47
66
A more complex network
Tandem Switch 2
Tandem Switch 1
Alternate Routes for B-E, C-F
Switch D
Switch A
Switch E
Switch B
High Usage Route B-E
Switch C
Switch F
High Usage Route C-F

• Tandem switches provide some primary and some
  alternate routes.
• Primary routes must have lower blocking
• Implies Tandem switches know about high usage
  routes in deciding which calls to accept

Figure 4.48
67
Dynamic Non-Hierarchical Routing
NCP

• All switches have direct trunks between them
• Switches track the utilization of those trunks
  and forward information to Network Control Points
  (NCPs)
• NCPs compute all two link paths between switches
  and forward those tables to the switches
• Calls are routed on direct path if available,
  then try two link routes
• Rather than blocking a call can be cranked back
  to try a new route

68
Network Management
Ideal Network
1
Load Carried
Goal for Managed Network
Real, unmanaged network
0
1
Load Offered

• Ideally a network would handle every call offered
  until it was 100 busy
• Real networks arent ideal
• Network does work on calls before deciding it
  cant complete them
• The wasted effort on ineffective calls reduces
  capacity
• The proportion of ineffective calls rises with
  rising load
• Network management is needed to improve
  performance

69
Examples of wasted effort

• Routing calls through local and toll switches
  when the end destination is busy
• Collecting telephone number and billing
  information when no trunks are available to take
  the call.
• Retransmitting/receiving signaling messages
  because the original message was dropped due to
  overload
• Processing calls the user gives up on because of
  delay or quality problems
• Routing through multiple switches because a
  direct route wasnt available.
• Searching large sets of calls in progress

70
Some Network Controls

• Split trunk groups
• 2-way groups (can be used by switch at either
  end) are more efficient, but prone to collisions)
• Simplified routing
• Alternate routes complete more calls, but they
  take more effort and use more trunks
• Priority for terminating calls
• When resources are scarce, block new calls to
  save resources to allow incoming calls to be
  completed

71
More Network Controls
1
2
3
1
2
3

• Call gapping
• Limit the rate of calls passed on.
• Others can be delayed or blocked
• Effective in spreading out traffic bursts
• Code blocking
• Blocks all or a of calls made to a particular
  destination that is overloaded
• Theory is that most of these calls will block or
  reach a busy number, and extra traffic may block
  calls that succeed.
• Often used in disaster situations

201-xxxxxx
415-xxxxxx
72
Can load control increase load?

• Look at a switch which when loaded decides to
  return busy to ¾ of the users picking up the
  phone to dial. If processing a call takes 10 ms
  of processor time and returning busy takes 3ms,
  this will save time and reduce load, right?

• What if all users redial until they get through?

• ¾ of the attempts get busy, so on average the
  switch will spend 19ms, 9 giving busy and 10
  completing the call for each call OOPS
• Its important to pick what to control!

73
Mobile Telephony Components

• Databases
• Home Location Register
• Visiting Location Register
• Authentication Center

Mobile Phone

• Base Station
• Controller
• Multiplexing

• Mobile Switching Center (MSC)
• Call Processing
• Features
• Switching
• Transcoding

• Cell Site(BTS)
• Radio gear
• Antennas

74
Mobile Telephony ConceptsCellular Structure

• Wireless Service area split into smaller areas
  known as cells served by antennas
• Division is somewhat arbitrary but based on
  density of phones and obstacles.
• Cells can be large (100s of square km) or very
  small (less than .01 square km)
• Phone communicates through the antenna in a cell
• Frequencies can be re-used in other cells (more
  users served)
• Signal power can be lower (longer battery life)
• Phone location must be tracked and call
  continuity maintained when the phone moves.
• Phone may show up in any cell.
• Phone may move while a call is in progress.

75
Cellular Structure
Omnidirectional
Microcells
3-sector

• Frequencies can be reused every 7 cells
• Cells can be omnidirectional or split in 3
  sectors
• 7 Microcells can replace one full sized cell in
  areas with high density
• Antennas are much lower gain than with fixed
  links
• SNR is low, so codes that tolerate low SNR are
  used.

76
Frequency Reuse

• Closest interfering mobile is about 4 times as
  far from the base as farthest local one.
• Interfering signal will be 20log10(4) 12dB down
• Power control significantly reduces the average
  interference.

R
4R


How many bits/Hz can I get through SNR of 12dB?
77
Mobile Telephony Basics -- Handoff

• If a call is in progress while a phone moves
• Signal received will vary as phone moves
• Adjacent cells monitor signal strength and
  determine when the call needs to move
• Handoff
• Phone is allocated new frequency (and possibly
  timeslots and codes) in a new cell.
• Call is switched to the new antenna and phone is
  told to use new frequency.
• Hard handoff connection is switched, with a
  momentary loss of continuity (TDMA, GSM)
• Soft Handoff New and old cells both monitor the
  phone and assemble signal from combination, no
  loss of continuity (CDMA)

78
Handoff

• Signal strength is monitored in adjacent cells
• When it is determined another cell is closer
• Handset is allocated a new channel
• Adjacent cell antenna is allocated
• Connection switched to new cell
• Complications
• What if there is no available channel?
• What if the adjacent cell is connected to another
  switch (MSC)?
• What if problems cause the mobile to get lost?

79
Mobile Telephony Basics Registration and Roaming

• Phone registers location periodically so that it
  can be located for incoming calls
• Local system records location information, known
  as the Visiting Location Register (VLR)
• VLR usually embedded in the MSC
• If the phone is not in its home service area, the
  local system contacts the home service area which
  maintains phones location and services (HLR)
• HLR can be part of an MSC or stand alone
• HLR/VLR communication via SS7 Messaging

80
Um
Abis
A
mobile station
base transceiver station
base station controller
MSC
Figure 4.53
81
Satellite Networks
satellite motion

• Systems use low earth orbit satelites
• Shorter distance than geostationary
• Less loss (and power needed)
• But the satellites move

• Moving Coverage areas
• Like Cellular
• BTS moves, phone doesnt