Chapter 3

Digital Transmission
Fundamentals
Analog vs. Digital
Digital Representation of Analog Signals
Why Digital Communications?
 Data vs. Signal
z Data: piece of information formatted in human/machine
readable form: voice, music, image, file
z Signal: electric or electromagnetic (EM) representation of
data; transmission media work by conducting energy along
a physical path; thus, to be transmitted, data must be
turned into energy in the form of EM signals
z Transmission : communication of data through propagation
and processing of signals

Coding Decoding
Data Signal Signal Data

Sender Channel or Communication Medium Receiver

Signal Representation
Signal Representation:
typically in 2D space, function of time, space or frequency
• when the horizontal axis is time, graph displays the
value of a signal at one particular point in space as
a function of time
• when the horizontal axis is space, graph displays
the value of a signal at one particular point in time
as a function of space
 Analog vs. Digital
Analog data: representation variable takes on continuous values in
some interval, e.g. voice, temperature, etc.
Digital data : representation variable takes on discrete
(a finite & countable number of) values in a given interval, e.g. text,
digitized images, etc.
Analog signal: continuous in time and can assume an infinite No. of
values in a given range (continuous in time and value)
Discrete (digital) signal: signal that is continuous in time and can
assume only a limited number of values (maintains a constant
level and then changes to another constant level)
Digitization of Analog Signal
z Sample analog signal in time and amplitude
z Find closest approximation
Original signal
Sample value

7Δ/2 Approximation
5Δ/2
3 bits / sample

3Δ/2
Δ/2
−Δ/2
−3Δ/2
−5Δ/2
−7Δ/2

Rs = Bit rate = # bits/sample x # samples/second

Example: Voice and Audio
Telephone voice CD Audio
z Ws = 4 kHz → 8000 z Ws = 22 kHertz → 44000
samples/sec samples/sec
z 8 bits/sample z 16 bits/sample

z Rs=8 x 8000 = 64 kbps z Rs=16 x 44000= 704 kbps

per audio channel
z Cellular phones use z MP3 uses more powerful

more powerful compression algorithms:

compression 50 kbps per audio
algorithms: 8-12 kbps channel
 Sampling Rate and Bandwidth
z A signal that varies faster needs to be sampled
more frequently
z Bandwidth measures how fast a signal varies
10 10 1 0 1 0 11 1 1 0 000
x1(t) x2(t)
... ... ... ...

t t

1 ms 1 ms

z What is the bandwidth of a signal?

z How is bandwidth related to sampling rate?
Periodic Signals
z A periodic signal with period T can be represented
as sum of sinusoids using Fourier Series:
x(t) = a0 + a1cos(2πf0t + φ1) + a2cos(2π2f0t + φ2) + …
+ akcos(2πkf0t + φk) + …

“DC” fundamental
long-term frequency f0=1/T kth harmonic
average first harmonic

•|ak| determines amount of power in kth harmonic

•Amplitude specturm |a0|, |a1|, |a2|, …

 Example Fourier Series
10 10 1 0 1 0 11 1 1 0 000
x1(t) x2(t)
... ... ... ...

t t

T2 =0.25 ms T1 = 1 ms

4 4
x1(t) = 0 + cos(2π4000t) x2(t) = 0 + cos(2π1000t)
π π
4 4
+ cos(2π3(4000)t) + cos(2π3(1000)t)
3π 3π
4 4
+ cos(2π5(4000)t) + … + cos(2π5(1000)t) + …
5π 5π
Only odd harmonics have power
http://www.nst.ing.tu-bs.de/schaukasten/fourier/en_idx.html
Spectra & Bandwidth
Spectrum of x1(t)
z Spectrum of a signal: 1.2

magnitude of amplitudes as 1

0.8

a function of frequency 0.6

0.4

z x1(t) varies faster in time & 0.2

has more high frequency

0

9
12

15

18

21

24

27

30

33

36

39

42
content than x2(t) frequency (kHz)

z Bandwidth Ws is defined as Spectrum of x2(t)

range of frequencies where 1.2

1
a signal has non-negligible 0.8

power, e.g. range of band 0.6

0.4

that contains 99% of total 0.2

signal power
0
0

9
12

15

18

21

24

27

30

33

36

39

42
frequency (kHz)
Bandwidth of General Signals
“speech”

s (noisy ) |p (air stopped) | ee (periodic) | t (stopped) | sh (noisy)

z Not all signals are periodic

X(f)
z E.g. voice signals varies
according to sound
z Vowels are periodic, “s” is
noiselike
z Spectrum of long-term signal
z Averages over many sounds,
many speakers f
z Involves Fourier transform 0 Ws
z Telephone speech: 4 kHz
z CD Audio: 22 kHz
 Chapter 3
Communication
Networks and Services

Digital Representation of Analog Signals

Digital Transmission of Analog
Information
2W samples / sec m bits / sample
Analog Sampling Quantization
source (A/D)

Original x(t) 2W m bits/sec

Bandwidth W

Transmission
or storage
Approximation y(t)

Display Interpolation Pulse

or filter
playout generator

2W samples / sec
Digital Transmission of Analog
Signals (Cont.)
Digitization Procedure consists of two steps:
(1) sampling – obtain signal values at equal intervals (T)
(2) quantization – approximate samples to certain values

Analog signal (continuous-time, Discrete time signal (discrete -time,

continuous-amplitude) continuous-amplitude)

Discrete time signal (discrete -time, Digital signal (discrete -time,

continuous-amplitude) discrete -amplitude)
Pulse Code Modulation (PCM)
 Sampling Theorem
According to the Nyquist theorem, the sampling rate must be at
least 2 times the highest frequency contained in the signal.

Nyquist: Perfect reconstruction if sampling rate 1/T > 2Ws

(a) x(t) x(nT)

t Sampler t

(b)

x(nT) x(t)

t Interpolation t
filter
 Quantization
‰ PAM signal samples have amplitudes of ‘∞ precision” –direct encoding of
such amplitudes would require ∞ number of bits (digital pulses) per sample

‰ to convert PAM signal to digital signal (that is practical for transmission), each
sample has to be ‘rounded up’ to the nearest of M possible quantization levels

M quantization levels : m = log2(M) bits per level

M↑ ֜ better precision , more bits per sample

M↓ ֜ poor precision , fewer bits per sample
Quantization
3.5Δ
Quantizer maps input
output y(nT) into closest of 2m
2.5Δ
1.5Δ representation values
0.5Δ
−4Δ −3Δ −2Δ −Δ
−0.5Δ 2Δ 3Δ 4Δ
−1.5Δ
Δ
input x(nT) Quantization error:
−2.5Δ “noise” = x(nT) – y(nT)
−3.5Δ

Original signal
Sample value
7Δ/2 Approximation
3 bits / sample

5Δ/2
3Δ/2
Δ/2
-Δ/2
-3Δ/2
-5Δ/2
-7Δ/2
Quantizer Performance
M = 2m levels, Dynamic range( -V, V) ∆ = 2V/M
error = y(nT)-x(nT)=e(nT)
Δ
2
... −2Δ Δ Δ 2Δ 3Δ ... input

V x(nT)
-V −
Δ
2

If the number of levels M is large, then the error is

approximately uniformly distributed between (-∆/2, ∆2)
Average Noise Power = Mean Square Error:
∆
1 ∆2
∫
2
σe =2 x2 dx =
∆ ∆ 12
2
Quantizer Performance
Figure of Merit:
Signal-to-Noise Ratio = Avg signal power / Avg noise power
Let σx2 be the signal power, then
σx2 12σ 2 σx σx
SNR = = x
= 3( )2 M2
Δ /12 4V2/M2
2 = 3( )2 22m
V V
The ratio V/σx ≈ 4

The SNR is usually stated in decibels:

SNR dB = 10 log10 σx2/σe2 = 6m + 10 log10 3σx2/V2
SNR dB = 6m - 7.27 dB for V/σx = 4.
Quantization (Cont.)
Example [ Quantization of PAM Signal ]
Assume an analog signal, as shown below, has to be quantized using at most
8-bits per sample. How many different quantization levels are allowed /
should be used?
 Quantization (Cont.)
Example [ voice signal in telephone system ]
Natural human voice occupies the range of 80 – 4000 [Hz]. Human ear can tolerate
SNR of 40 [dB]. Assume we want to transmit human voice in digitized form.
What bit rate [bps] should be supported by the channel to enable such transmission?

(1) Sampling rate?!

Based on Nyquist Sampling Theorem:
max frequency = 4 [kHz] ֜ sampling rate = 2*4 [kHz] = 8000 [samples/sec]

(2) # of bits per sample?!

Based on SNR formula:
40 [dB] = 6*m - 7.76 ֜ # bits per sample = 8 ֜ # of levels = 28 = 256

data rate = # samples per second * # bits per sample = 64 kbps

 Digital Signals
– sequence of voltage pulses (DC levels) – each pulse represents
a signal element
• binary data are transmitted using only 2 types of signal
elements ( 1 = positive voltage, 0 = negative voltage )
• key digital-signals terms:
ƒ bit interval – time required to send one single bit – unit: [sec]
ƒ bit rate – number of bit intervals per second – unit: [bps]

Most digital signals are aperiodic, 23

so it is not appropriate / correct to talk about their period.
Digital Signals (Cont.)
Digital Signal as a – digital signal, with all its sudden changes,
Composite Analog Signal is actually a composite signal having an
infinite number of frequencies

‰ a digital signal is a composite signal with

an infinite bandwidth

‰ if a medium has a wide bandwidth, a digital

signal can be sent through it

‰ some frequencies will be weakened or blocked;

still, enough frequencies will be passed to
preserve a decent signal shape

‰ what is the minimum required bandwidth

B [Hz] of a band-limited medium if we want
to send n [bps]?

24