Chapter 2 Physical

1
Chapter 2Physical Data-Link Layers

• Professor Rick Han
• University of Colorado at Boulder
• rhan_at_cs.colorado.edu

2
Announcements

• Textbooks Computer Networks, by Peterson and
  Davie
• CU book stores order has arrived
• Read Chapter 1, start Chapter 2, sections
  2.1-2.8, skip Section 2.9 Network Adaptors
• Larger classroom maybe
• Lectures will be online a few days after each
  class
• Follow the Lectures link on class Web site
• Homework 1 will be handed out next class
• Next, Chapter 2

3
The Layered Network Stack
Internet Stack
Application Layer
Transport Layer (TCP/UDP)
Network Layer (IP)
Data Link Layer
Physical Layer
4
Layer 1 The Physical Layer

• Host A encodes the bit into an analog signal.
  Host B decodes the analog signal into a received
  bit.

5
Encoding Mapping Bits to Analog Waveforms

• 1 -gt 5 volts, 0 -gt -5 volts on copper wire
• light intensity on optical fiber

• Pulse Amplitude Modulation (PAM) encodes bits as
  diff. levels. NRZ is special case.
• Manchester encodes bits as diff. transitions
• FSK (Frequency Shift Key) encodes bits as diff.
  frequencies. 1200 bps modems.

6
Sending More Bits Per Second

• Can reduce the size of time slot
• Widely used
• Limitation smearing of the waveforms and noise
  limit the maximum bit rate (Shannons capacity
  theorem)

• Square wave is sum of many different frequency
  components. analogy stereo signal
• High frequencies get attenuated in wire, thereby
  smearing the waveform
• Reducing time slots causes more smear from
  previous slot to interfere with current slot
  InterSymbol Interference (ISI)

7
Sending More Bits Per Second (2)

• Can send more bits per time slot using
  multi-level D/A converter
• 8 discrete levels
• 111 -gt 7 V, , 000 -gt -7 V
• 100110 -gt 1, 5,

Symbols

• 3 bits per symbol

8
Sending More Bits Per Second (3)

• Transmit a sequence of symbols S11, S25, S3,
  
• Baud rate Symbols/sec
• If baud rate is 1 symbol/sec, then bit rate 3
  bits/sec
• Tripled the bit rate!
• Limitations noise

• Datam(t)Sm(t)SqWv(t-mT)
• Where Sm(t) is the level of mth symbol, T
  duration of each symbol, ttime, SqWv is square
  wave

9
Sending More Bits Per Second (4)

• Can send more bits over different frequencies,
  analogy AM/FM
• Modulate Sm(t) up to 3000 Hz, demodulate down
• Datam(t) Sm(t)cos(2p3000t) Sm(t)cos
  (2p6000t) simplified sq wave not
  shown
• Can send more bits over the same frequency!
• Datam(t) Sm(t)cos(2p3000t) Sm(t)sin
  (2p3000t)
• This is called Quadrature Amplitude Modulation
  (QAM).
• If Sm(t), and Sm(t) both have 4 levels, then
  this is called 16-QAM. 9600 bps modems use this.
  Baud?

10
Physical Layer Effects

• Goal maximize the Signal-to-Noise ratio (SNR) to
  minimize the probability of bit error, then pass
  the bit up to Data-Link Layer
• Unreliability
• due toSmearing, Interference, (Wireless
  shadowing, multi-path, doppler, )
• Apply advanced adaptive filtering and digital
  signal processing (DSP) to improve SNR
• Propagation Delay
• Speed of light c 3x108 m/s
• Over copper wires, propagation speed is 2/3 of c
• Satellite links have long prop. delays (120 ms
  one-way)
• Interactivity requires lt400 ms roundtrip

11
Layer 2 The Data Link Layer

• Next Problem How do I send a message from Host A
  to Host B?

• Data Link Layer, also called Layer 2, ensures
  that host B can decode a digital message from a
  stream of bits sent by host A
• Examples PPP (Point-to-Point Protocol), HDLC,
  LAPB, LAPD, Frame Relay,

12
The Data Link Layer (cont.)

• A Data Link Layer Protocol implements
• Delimiting/framing of a message
• Fragmenting of a long message
• Retransmission of a lost message

1011000
Host A
Host B
13
Defining a Protocol

• A protocol is an agreement between two parties or
  endpoints as to how information is to be
  transmitted
• A protocol implements this agreement via
• A Header
• How each endpoint responds to control info in the
  header ( external input)

Host A
Host B
14
Framing
Receivers A/D
15
Character/Byte Stuffing

• BiSync and PPP (common over modems)
• Data is divided into 8-bit bytes
• Byte boundaries synchronized btwn sender and
  receiver
• Define a special start-of-packet N-byte flag,
  e.g. let flag be X (one byte flag)
• Stuff X in data replace X with escape character
  E (DLE in textbook) and X, i.e X -gt (E, X)
• Stuff E in data replace E with (E, E)
• At rcvr, first X is start of packet (variants
  exist)

16
Bit Stuffing
01001000001111110000001110001001

• HDLC uses this
• Similar to byte stuffing, except bit stuffing is
  not confined to byte boundaries
• HDLC denotes beginning and end of a packet/frame
  with 01111110 flag
• Since 01111110 may occur anywhere (across byte
  boundaries) in data, stuff it
• At sender, after 5 consecutive ones, insert a 0
• At receiver, 0111110 gt stuffing, so destuff,
  01111110 gt end of frame, 01111111 gt error

17
Byte Counting

• Deduce End-of-Packet with a length field in the
  header, rather than explicit sentinel flag
• 010010000000000100000011100010010
• Start-of-Packet length data
• Length 2 bytes
• Limitation the length field could be corrupted
  by a transmission bit error, due to noise, etc.
• Bit corruption in header and start-of-packet
  usually catastrophic
• Bit corruption can also occur in data
• How do we handle errors caused by bit corruption?

18
Error Detection

• Simple technique Send two copies of data in
  addition to the data, and then do majority logic
  decoding at the receiver

Original data 0100
Corrupt Bit
Send 0100 0100 0100
Receiver 0100 0110
0100 Decode 0100

• Inefficient
• Two errors in the same bit gt cannot detect
  error,
• 0100 0110 0110 -gt 0110 Error
• so each technique has some probability of error

19
Error Detection (2)

• General technique Add redundancy to the data and
  use this redundancy to detect bit errors
• 01001000000000010011010100000011100010010
• Start-of-Packet length err. det.
  data
• The error detection field can be computed over
  the data only, the header only (IP) or both (TCP)
• Can be located in header or at end of packet

• The error detection field is also called a Cyclic
  Redundancy Check (CRC)
• Most link-layer protocols use CRCs for error
  detection, e.g. HDLC, PPP
• A checksum is a special case of a CRC where the
  CRC is computed using binary addition

20
Error Detection (3)

• The Internet checksum
• Divide your data into 16-bit segments
• Add each of the 16-bit segments, using ones
  complement binary addition
• Ones complement of the final sum is your 16-bit
  IP checksum
• Send IP checksum with the IP packet
• An alternative checksum XOR instead of ones
  complement
• Checksums are relatively weak compared to CRCs
  but are easy to compute
• A bit error twice in the same offset position
  within the 16-bit segment will go undetected

21
Probability of Packet Error

• Suppose N bits are sent, and the link is
  independently corrupting each bit with some
  probability of bit error pb. What is the
  probability of packet error?
• Probpacket error Probat least 1 bit is
  corrupt
• 1
  Probevery bit is clean
• 1-(1- pb)N
• If pb 10-6, and N10000 bits, then
• Probpacket error 9.9510-3 1
• Optical links have much lower probabilities of
  bit error 1012
• Wireless links have much higher probabilities of
  bit error 10-3 in fades send small packets

22
Error Correction

• Forward Error Correction (FEC)
• Rather than just detect a bit error, correct a
  bit error
• Add K bits of redundancy to N bits, to form a
  (NK)-bit long packet, or vector

• N dimensions -gt NK dimensions
• 2N patterns or vectors mapped into 2NK
  possibilities
• Spread out these vectors as far away from
  neighbors as possible in (NK)-dimensional space
• N2, K3, NK5
• Receive 01111 - closest
• to 11111, so decode 11
• and correct one bit error