Principles of Digital Data Transmission Chapter 7 Lectured by Dr. Yun Q. Shi Dept. of Electrical & Computer Engr. New Jersey Institute of Technology
[email protected] Text used for the course:
, 4th Edition, Lathi and Ding, Oxford
Introduction • A significant portion of communication in 1990s was in analog. • It has been replaced rapidly by digital communication. • Now most of the communications become digital, with analog communication playing a minor role. • This chapter addresses the framework (several major aspects) of digital data transmission • All we learnt in this course so far is utilized here. • Hence, this chapter is naturally a conclusion of this course.
Dr. Shi
Lathi & Ding-Digital Commun
2
1
Digital Communication Systems: Source • Digital Communication Systems: Figure 7.1 Source encoder, Baseband modulation (Line coding), Digital carrier modulation, Multiplexer, Channel, Regenerative repeater, Receiver (detection).
• Source:
The input
– A sequence of digits – We discuss mainly: The binary case (two symbols) – Later in this chapter: The M-ary case (M symbols) more general case Dr. Shi
Lathi & Ding-Digital Commun
3
Figure 7.1 Fundamental building blocks of digital communication systems
Figure 7.1 Fundamental building blocks of digital communication systems
Dr. Shi
Lathi & Ding-Digital Commun
4
2
Line Coder (Transmission Coder) • The output of a multiplexer is coded into electrical pulses or waveforms for the purpose of transmission over the channel. • This process is called a line coding or transmission coding • Many possible ways of assigning waveforms (pulses) to the digital data. 1
• On-off: Dr. Shi
↔ a 0 ↔
pulse p (t) no pulse
Lathi & Ding-Digital Commun
5
Line Coder … • Polar: 1 ↔ ↔ 0
a a
pulse pulse
p (t) − p (t)
• Bipolar (pseudoternary or alternate mark inversion (AMI))
1 ↔ 0 ↔
a pulse p (t) or - p (t) depending on whether the previous 1 is encoded by -p (t) or p (t) no pulse
In short, pulses representing consecutive 1’s alternate in sign. • All the above three could use half-width pulses. It is possible to select other width. • Figure 7.1 (parts a, b and c). Dr. Shi
Lathi & Ding-Digital Commun
6
3
Line Codes
Dr. Shi
Lathi & Ding-Digital Commun
7
Line Coder … • Full width pulses are often used in some applications. – i.e., the pulse amplitude is held to a constant value throughout the pulse interval (it does not have a chance to go to zero before the next pulse begins). – These schemes are called non return-to-zero (NRZ) in contrast to return-to-zero (RZ) – Figure 7.2 shows • An On-off NRZ signal • A Polar NRZ signal in d and e parts of the figure Dr. Shi
Lathi & Ding-Digital Commun
8
4
Multiplexer • Usually, the capacity of a practical channel >> the data rate of individual sources. • To utilize this capacity effectively, we combine several sources through a digital multiplexer using the process of interleaving. • One way: A channel is time-shared by several messages simultaneously. Dr. Shi
Lathi & Ding-Digital Commun
9
Regenerative Repeater • Used at regularly spaced intervals along a digital transmission line to: 1) detect the incoming digital signal and 2) regenerate new clean pulses for further transmission along the line. • This process periodically eliminates, and thereby combats the accumulation of noise and signal distortion along the transmission path. • The periodic timing information (the clock signal at R b Hz) is required to sample the incoming signal at a repeater. • R b (rate): pulses/sec Dr. Shi
Lathi & Ding-Digital Commun
10
5
Regenerative Repeater… • The clock signal can be extracted from the received signal. – e.g., the polar signal when rectified ⇒ a clock signal at Rb Hz • The on-off signal = A periodic signal at Rb + a polar signal (Figure 7.) • When the periodic signal is applied to a resonant circuit tuned to Rb Hz, the output, a sinusoid of Rb Hz, can be used for timing. Dr. Shi
Lathi & Ding-Digital Commun
11
On-off Signal Decomposition
Dr. Shi
Lathi & Ding-Digital Commun
12
6
Regenerative Repeater… rectified
• The bipolar signal an on-off signal ⇒ the clock signal can also be extracted. • The timing signal (the output of the resonant circuit) is sensitive to the incoming pattern, sometimes. – e.g. in on-off, or 0 bipolar
↔
no pulse
• If there are too many zeros in a sequence, will have problem. – no signal at the input of the resonant circuit for a while – sinusoids output of the resonant circuit starts decaying
• Polar scheme has no such a problem. Dr. Shi
Lathi & Ding-Digital Commun
13
Transparent Line Code • A line code in which the bit pattern does not affect the accuracy of the timing information is said to be a transparent line code. – The polar scheme is transparent. – The on-off and bipolar are not transparent (nontransparent).
Dr. Shi
Lathi & Ding-Digital Commun
14
7
Desired Properties of Line Codes 1.
Transmission bandwidth: as small as possible
2.
Power efficiency: for a specific bandwidth and detection error rate, transmitted power should be as small as possible.
3.
Error detection and correction capability: as strong as possible
4.
Favorable power spectral density: desirable to have zero PSD at (dc), called DC null, because of ac coupling requires this. If at dc, ω = 0 , PSD ≠ 0 , then dc wanders in the pulse stream, which is not desired.
5.
Adequate timing content: should be possible to extract the clock signal (timing information).
6.
Transparency Dr. Shi
Lathi & Ding-Digital Commun
15
PSD of Various Line Codes • Consider a general PAM signal, as shown in Figure 7.4 (b): Tb = 1 t = kTb Rb Rb – the k th pulse in the pulse train y (t ) = ak p (t ) • p (t): the basic pulse • P (ω) [P(f)]: Fourier spectrum of p (t) • a k: arbitrary and random
– The on-off, polar, bipolar line codes are all special cases of the general pulse train y(t), a k = 0, +1, -1 Dr. Shi
Lathi & Ding-Digital Commun
16
8
A Random PAM signal
All 3 line codes can be represented this way
New approach to determine PSD of y(t)
Dr. Shi
Lathi & Ding-Digital Commun
17
PSD of Various Line Codes … • How to determine the PSD? – Figure 7.4 – An attractive general approach Rx (τ ), S x (ω ) need to be studied once Then for different p(t), ⇒ different PSD – x(t): an impulse train – xˆ (t ) : a rectangular pulse train ( ε ⋅ hk = a k ) when ε → 0, xˆ (t ) → x(t ) (Figure 7.5) Dr. Shi
Lathi & Ding-Digital Commun
18
9
Dr. Shi
Lathi & Ding-Digital Commun
19
PSD of Various Line Codes … R0 = ak2
⇒ time average of a k2
R1 = ak .ak +1 ,L L, Rn = ak .ak +n
1 RX (τ ) = Tb S X (ω ) =
1 Tb
∞
∑ R δ (τ − nT )
n = −∞
n
b
∞
∑ Rn e− jnωTb =
n = −∞
1 Tb
∞ [ ] R + 2 R cos( n ω T ) ∑ 0 n b n =1
S y (ω ) = P (ω ) S X (ω ) 2
Different line codes Dr. Shi
⇒ different P(ω ) ⇒ different S y (ω ) Lathi & Ding-Digital Commun
20
10
Polar Signaling 1 ↔ p(t) 0 ↔ -p(t)
} ⋅ ak = ±1, equally
R0 = ak2 = 1
R1 = ak .ak +1 = 0 Rn = 0, n ≥ 1
likely, ak2 = 1
Q ak .ak +1 = 1 or − 1 equally likely Similar reasoning
S y (ω ) = P (ω ) .S X (ω ) 2 1 = P(ω ) . Tb 2
Dr. Shi
Lathi & Ding-Digital Commun
21
Polar Signaling… • To be specific, assume p(t) is a rectangular pulse of width T b (a half-width rectangular 2 pulse) t p (t ) = re c t T b 2 T ω Tb P ( ω ) = b s in c 2 4 T ω Tb S y ( ω ) = b s in c 2 4 4 Dr. Shi
Lathi & Ding-Digital Commun
1st zero − cros sin g :
ω⋅Tb 4
= π ⇒ω =
or ⇒ f =
4π Tb
2 = 2Rb Tb 22
11
PSD of polar signal (half-width rectangle)
Dr. Shi
Lathi & Ding-Digital Commun
23
Polar Signaling … • Comment 1:
• • •
•
The essential bandwidth of the signal (main lobe) = 2 Rb For a full-width pulse, ⇒ Rb “Polar Signaling is not bandwidth efficient.” Comment 2: “No error-detection or error-correction capability.” Comment 3: “Non-zero PSD at dc( ω = 0)” ⇒ This will rule out the use of ac coupling in transmission. Comment 4: “Most efficient scheme from the power requirement viewpoint” for a given power, the detection-error probability for a polar scheme is the smallest possible. Comment 5: “Transparent” Dr. Shi
Lathi & Ding-Digital Commun
24
12
Achieving DC Null in PSD by Pulse Shaping Q P (ω ) =
∫ ∴ P (0) = ∫
∞
−∞
∞
−∞
p ( t ) e − j ω t dt p ( t )dt
⇒ If the area under p(t) = 0, P(0) = 0 .
Dr. Shi
Lathi & Ding-Digital Commun
25
Manchester (split-phase, twinned-binary) Signal
Dr. Shi
Lathi & Ding-Digital Commun
26
13
On-off Signaling • 1 ↔ p(t) 0 ↔ no pulse R
0
R
n
S
X
=
1 2 1 = 4
=
(ω ) =
∞
1 1 + 2Tb 4Tb
1 1 + 4Tb 4Tb
Dr. Shi
n ≥ 1
for all
∞
∑
e
∑
e
− jn ω T b
n = −∞ n ≠0
− jn ω T b
n = −∞
Lathi & Ding-Digital Commun
27
On-off Signaling S X (ω ) =
1 2π + 4Tb 4 T b2
∞
∑
n = −∞
δ ω −
2π n Tb
P (ω ) 2π ∞ 2π n S y (ω ) = 1 + δ ω − ∑ 4Tb Tb n = −∞ Tb For a half-width rectangular pulse, 2
S y (ω ) =
Tb 2π ω Tb sinc 2 1 + T 16 4 b
∞
∑
n = −∞
δ ω −
2π n Tb
Note: 1. For some derivation, refer to the next four pages. 2. PSD is shown in Figure 7.8. 3. PSD consists of both a discrete part and a continuous part. Dr. Shi
Lathi & Ding-Digital Commun
28
14
Some Derivation (Lathi’s book, pp. 58-59, Example 2.12) • A unit impulse train: FS
g (t) =
∞
∑
n = −∞
D ne
jn ω b t
δ T (t ) = CombT (t ) = g (t ) 0
= =
0
1 Tb
∞
∑
e
n = −∞
1 2 + Tb Tb
jn ω b t ∞
∑
n = −∞
cos( n ω b t )
1 ∞ ∑ δ ( f − nf 0 ) T 0 n = −∞ 2π ∞ 2π G ( ω ) = FT { g ( t )} = ) δ (ω − n ∑ T 0 n = −∞ T0 G ( f ) = FT { g ( t )} =
≅ ω 0 COMB Dr. Shi
ω0
(ω )
Lathi & Ding-Digital Commun
29
Example 2.12
Dr. Shi
Lathi & Ding-Digital Commun
30
15
Some Derivation… • Fact 1 (Lathi’s book, p.83, Eq. (3.20a, 3.20b): 1↔ δ(f
)
e j 2πf bt ↔ δ ( f − f b )
• Fact 2:
[e
jω bt
]
↔ δ (ω − ωb )
Also,
(t ) ↔ 1 δ (t − nT b ) ↔ δ
Dr. Shi
e
− jn ω T
b
Lathi & Ding-Digital Commun
31
Some Derivation … • Fact 3:
(
)
∞ ∑ δ t − nT b n1=4−∞ 42443
FS =
∞ jnω t 1 2π b ω = ∑ e b T n = −∞ T b 1b4 44442444443
b FT 644744 8 ∞ − jnω t b ∑e n = −∞ Dr. Shi
b FT =
ω
(
∞ ∑ δ ω − nω b b n = −∞
Lathi & Ding-Digital Commun
) 32
16
PSD of An On-off Signal
Dr. Shi
Lathi & Ding-Digital Commun
33
On-off Signaling • This is reasonable since, as shown in Fig. 7.2, an on-off signal can be expressed as the sum of a polar and a periodic component • Comment 1: For a given transmitted power, it is less immune to noise interference. – Q Noise immunity ∝ difference of amplitudes representing binary 0 and 1. – If a pulse of amplitude 1 or –1 has energy E, then a pulse of amplitude 2 has energy (2)2E = 4E.
Dr. Shi
Lathi & Ding-Digital Commun
34
17
On-off Signaling … • For polar:Q
1 Tb
∴ • For on-off:
digits are transmitted per second polar signal power 1 2E = power = 4 E 2 T b Tb
1 E = E = Tb Tb
Power is now as large as twice of above.
• Comment 2: Not transparent Dr. Shi
Lathi & Ding-Digital Commun
35
Bipolar Signaling (Pseudoternary or Alternate Mark Inverted (AMI)) A. 0 ↔ no pulse 1 ↔ p(t) or –p(t) depending on whether the previous 1 was transmitted by –p(t) or p(t) B. [p(t), 0 , -p(t)]: In reality, it is ternary signaling. C. Merit: a dc null in PSD demerit: not transparent Dr. Shi
Lathi & Ding-Digital Commun
36
18
Bipolar Signaling… 1 N N 1 ( ± 1) 2 + (0) = N →∞ N 2 2 2 k 1 N 3N 1 R (1) = lim ( − 1) + (0) = − N →∞ N 4 4 4 111, 101, 110, 011, 010, 001, 000 ak ak + n R (2)
D. R ( 0 ) = Nlim →∞
R ( 2 ) = lim
N →∞
R n = lim
N →∞
1 N
1 N
∑a
2 k
= lim
1 N N 5N (1) + ( − 1) + (0) = 0 N 8 8 8
∑a
k
a k + n = 0, n > 1
k
Dr. Shi
Lathi & Ding-Digital Commun
37
Bipolar Signaling… E.
S y (ω ) = P(ω ) S X (ω ) 2 P (ω ) ∞ R n e − jnω Tb = ∑ Tb n = −∞ 2
P (ω ) = Tb
2
P (ω ) = Tb P (ω ) = 2 ⋅ Tb Dr. Shi
2
∞ R + 2 R n cos nω Tb ∑ 0 n =1 2
(7.10b)
(7.10c)
1 1 + 2 ⋅ ( − ) cos ω T b 4 2
P (ω ) (1 − co s ω Tb ) = Tb
2
Lathi & Ding-Digital Commun
ω Tb sin 2 2 38
19
Bipolar Sampling … ⇒
As ω=0 (dc) regardless of P(ω) A dc null (desirable for ac coupling)
S y (ω ) = 0
sin 2 (
ω Tb 2
)= 0
at
ω =
⇒ b a n d w id th = R b H z ,
2π 1 ,∴ f = = Rb Tb Tb reg a rd less
For a half-width rectangular pulse,
Dr. Shi
⇒ Zero:
of
P (ω )
t Tb , P (t ) = rect 2 T b 2 ω Tb P (ω ) = s in c 2 4 T b
Lathi & Ding-Digital Commun
S y (ω ) =
39
Tb ωT ωT sinc 2 b sin 2 b 4 4 2
ɷ1Tb=4π 4π ω1 = Tb f1 = 2 Rb
ɷ2Tb=2π 2π ω2 = Tb f 2 = Rb
⇒
1 Rb = Tb essential ⋅ bandwidth
Half that of polar or on-off signaling. Twice that of theoretical minimum bandwidth (channel b/w). Table 3.1:
Dr. Shi
ω t ωτ rect ↔ τ sin c = τ sin c 2 τ 2 τ Lathi & Ding-Digital Commun
40
20
Dr. Shi
Lathi & Ding-Digital Commun
41
Merits of Bipolar Signaling 1.Spectrum has a dc dull. 2.Bandwidth is not excessive. 3.It has single-error-detection capability. Since, a single detection error ⇒ a violation of the alternation pulse rule.
Dr. Shi
Lathi & Ding-Digital Commun
42
21
Demerits of Bipolar Signaling 1. Requires twice as much power as a polar signaling. Distinction between A, -A, 0 vs. Distinction between A/2, -A/2 2. Not transparent Dr. Shi
Lathi & Ding-Digital Commun
43
Pulse Shaping S y (ω ) = P (ω ) S X (ω ) 2
=
P (ω ) Tb
2
( R0 + 2
∞
∑R
n
cos nωTb )
[Slide 19]
n =1
• Different signaling (line coding, SX (ω) ) ⇒ diff.S y (ω ) • Different pulse shaping (P(ω)) ⇒ diff. S y (ω ) An additional factor. • Intersymbol Interference (ISI) (refer to the next slide) – Time-limited pulses (i.e., truncation in time domain) not band-limited in frequency domain – Not time-limited, causes problem band-limited signal (i.e., truncation in frequency domain)
Dr. Shi
Lathi & Ding-Digital Commun
44
22
Intersymbol Interference (ISI) in Detection B. Sklar’s Digital Communications 1988
Dr. Shi
Lathi & Ding-Digital Commun
Back to slide 44
45
Phase Shaping…
Intersymbol Interference (ISI) • Whether we begin with time-limited pulses or band-limited pulses, it appears that ISI cannot be avoided. An inherent problem in the finite bandwidth transmission. • Fortunately, there is a way to get away: be able to detect pulse amplitudes correctly. Dr. Shi
Lathi & Ding-Digital Commun
46
23
Nyquist Criterion for Zero ISI • The first method proposed by Nyquist. 1, t = 0 1 p (t ) = 0, t = ± nTb Tb = Rb
(7.22)
T b : separation between successive transmitted pulses
• Example 6.1 (p.256) p(t ) = sinc(2π Bt ),
B=
1 1 = Rb 2Tb 2
Since 2πBTb=π, Tb is the 1st zero crossing.
• Minimum bandwidth pulse that satisfies Nyquist Criterion Fig. 7.10 (b and c) Dr. Shi
Lathi & Ding-Digital Commun
47
Minimum Bandwidth Pulse (Nyquist Criterion)
Dr. Shi
Lathi & Ding-Digital Commun
48
24
Nyquist Criterion for Zero ISI… 1 p (t ) = sinc (π Rb t ) = 0
P (ω ) =
ω 1 rect Rb 2π R b
Table 3.1:
t=0 t = ± nTb
1 Tb = Rb
bandwidth =
Rb 2
FT
ω Sinc (Wt ) ↔ rect π 2W
W
Dr. Shi
Lathi & Ding-Digital Commun
49
Problems of This Method •
Impractical 1. − ∞ < t < ∞ 2. It decays too slowly at a rate 1/t. ⇒ Any small timing problem (deviation) ⇒ ISI
•
Solution: Find a pulse satisfying Nyquist criterion (7.22), but decays faster than 1/t.
•
Nyquist: Such a pulse requires a bandwidth k⋅ Dr. Shi
Rb , 2
1≤ k ≤ 2
Lathi & Ding-Digital Commun
50
25
p (t ) ↔ P (ω )
• Let
– The bandwidth of P(ω) is in (Rb /2, Rb) – p(t) satisfies Nyquist criterion Equation (7.22)
• Sampling p(t) every Tb seconds by multiplying p(t) by an impulse train δ T (t ). b
⇒ b ⇒
p (t )δ
p (t) =
T
b
(t) = δ (t)
FT 1 T b
∞
∑
P (ω
− n ω
b
P (ω
− n ω
n = −∞
Equation G (ω ) =
) = 1
( 6 .4 ) 1 T s
∞
∑
n = −∞
s
)
Dr. Shi
Lathi & Ding-Digital Commun
51
Dr. Shi
Lathi & Ding-Digital Commun
52
26
Derivation
• |P(ω)| is odd symmetric in y1-o-y2 system. Dr. Shi
Lathi & Ding-Digital Commun
53
Derivation… ω
b • The bandwidth of P(ω) is 2 + ω x – ω x : the excess bandwidth
–
γ
roll − offfactor
=
=
excess bandwidth theoretica l min imum bandwidth
ωx ω = 2⋅ x ωb ωb
0 ≤ r ≤1
2 R
rR
R
• The bandwidth of P(ω) is BT = 2b + 2b = ( 1 + γ ) 2b • P(ω), thus derived, is called a Vestigial Spectrum. Dr. Shi
Lathi & Ding-Digital Commun
54
27
Realizability • A physically realizable system, h(t), must be causal, i.e., h(t) = 0 , for t < 0. (The n.c. & s.c.) • In the frequency domain, the n.c. & s.c. is known as Paley-Wiener criterion ∞
ln H (ω )
−∞
1+ω 2
∫
dω < ∞
• Note that, for a physically realizable system, H(ω) may be zero at some discrete frequencies. But, it cannot be zero over any finite band • So, ideal filters are clearly unrealizable. Dr. Shi
Lathi & Ding-Digital Commun
55
• From this point of view, the vestigial spectrum P(ω) is unrealizable. • However, since the vestigial roll-off characteristic is gradual, it can be more closely approximated by a practical filter. • One family of spectra that satisfies the Nyquist criterion is ωb 1 / 2 P (ω ) = 0 1
Dr. Shi
) π (ω − 2 1 − sin 2ω x
Lathi & Ding-Digital Commun
ω− ω > ω <
ωb 2
ωb 2
ωb 2
< ωx +ωx −ωx
56
28
Realizability … ωx = 0 ω ωx = b 4
ωx =
ωb 2
(r = 0) ( r = 0 .5 ) ( r = 1)
Shown in the figure on next slide
• Comment 1: Increasing ω x (or r) improves p(t), i.e., more gradual cutoff reduces the oscillatory nature of p(t), p(t) decays more rapidly. Dr. Shi
Lathi & Ding-Digital Commun
57
Pulses satisfying Nyquist Criterion
Dr. Shi
Lathi & Ding-Digital Commun
58
29
Realizability… • Comment 2: As
γ
= 1, i.e., ω x
=
ωb 2
ω ω 1 rect P (ω ) = 1 + cos 2 2 Rb 4πRb
ω ω rect = cos 2 4 Rb 4πRb
• This characteristic is known as: The raised-cosine characteristic The full-cosine roll-off characteristic Dr. Shi
Lathi & Ding-Digital Commun
59
Realizability… p (t ) = Rb
A. B. C.
D.
co s(π R b t ) S in c (π R b t ) 1 − 4 R 2bt 2
Bandwidth is Rb (γ =1) p(0) = Rb p(t) = 0 at all the signaling instants & at points midway between all the signaling instants p(t) decays rapidly as 1/t3 ⇒ relatively insensitive to derivations of Rb , sampling rate, timing jitter and so on. Dr. Shi
Lathi & Ding-Digital Commun
60
30
Realizability… E. Closely realizable F. Can also be used as a duobinary pulse. G. H c (ω ) : channel transfer factor Pi ( k ) : transmitted pulse Pi (ω ) ⋅ H c (ω ) = P (ω )
Received pulse at the detector input should be P(ω) [p(t)]. Dr. Shi
Lathi & Ding-Digital Commun
61
Signaling with Controlled ISI: Partial Response Signals • The second method proposed by Nyquist to overcome ISI. n = 0,1 1 • Duobinary pulse: p(nTb ) = 0 for all other n
Dr. Shi
Lathi & Ding-Digital Commun
62
31
Signaling with Controlled ISI… • Use polar signaling 1 ↔ p (t ) 0 ↔ − p (t )
• The received signal is sampled at t = nTb p(t) = 0 for all n except n= 0,1. • Clearly, such a pulse causes zero ISI with all the pulses except the succeeding pulses. Dr. Shi
Lathi & Ding-Digital Commun
63
Signaling with Controlled ISI… • Consider two such successive pulses located at 0 and Tb. – If both pulses are positive, the sample value at t = Tb ⇒ 2. – If both pulses are negative, ⇒ -2 – If pulses are of opposite polarity, ⇒ 0
• Decision Rule: If the sample value at t = Tb is • Positive ⇒ 1, 1 • Negative ⇒ 0, 0 • Zero ⇒ 0, 1 or 1, 0
Dr. Shi
Lathi & Ding-Digital Commun
64
32
Signaling with Controlled ISI… • Table 7.1
Transmit ed Seq. Sample of x(Tb) Detected seq.
Dr. Shi
1 1 0 1 1 0 0 0 1 0 1 1 1 1 2 0 0 2 0 -2 -2 0 0 0 2 2 1 1 0 1 1 0 0 0 1 0 1 1 1
Lathi & Ding-Digital Commun
65
Signaling with Controlled ISI: Duobinary Pulses • If the pulse bandwidth is restricted to be it can be shown that the p(t) must be: p (t) =
Rb 2
sin( π R b t ) π R b t (1 − R b t )
ω 2 P (ω ) = cos Rb 2 Rb
rect
ω 2π R b
− e
j
ω 2 Rb
ωb
• Note: The raised-cosine pulse with ω = 2 , ( r = 1), satisfies the condition (7.3b) to be signaling with controlled ISI – refer to Figure 7.13 (slides 57). x
Dr. Shi
Lathi & Ding-Digital Commun
66
33
Signaling with Controlled ISI Duobinary Pulses…
Dr. Shi
Lathi & Ding-Digital Commun
67
Use of Differential Coding • For the controlled ISI method, 0-valued sample ⇒ 0 to 1 or 1 to 0 transition. • An error may be propagated. • Differential coding helps. • In differential coding, “1” is transmitted by a pulse identical to that used for the previous bit. “0” is transmitted by a pulse negative to that used for the previous bit. • Useful in systems that have no sense of absolute polarity. Fig 7.17 Dr. Shi
Lathi & Ding-Digital Commun
68
34
Differential Code
Dr. Shi
Lathi & Ding-Digital Commun
69
Scrambling • Purpose: A scrambler tends to make the data more random by – Removing long strings of 1’s or 0’s – Removing periodic data strings. – Used for preventing unauthorized access to the data.
• Example. (Structure) Scrambler: S ⊕ D 3T ⊕ D 5T = T D nT : ⊕:
T delayed by n units Modulo 2 Sum
1⊕1 = 0 0⊕0=0 Dr. Shi
→ Incomplete version
1⊕ 0 = 1 0 ⊕1 = 1
Lathi & Ding-Digital Commun
70
35
Dr. Shi
Lathi & Ding-Digital Commun
71
Scrambling… • Descrambler:
Incomplete version
R = T ⊕ (D
3
⊕ D 5 )T
= [1 ⊕ ( D 3 ⊕ D 5 )] T = (1 ⊕ F ) T ∆
F = D
Dr. Shi
3
⊕ D5
Lathi & Ding-Digital Commun
72
36
Example 7.2 • The data stream 101010100000111 is fed to the scrambler in Fig. 7.19a. Find the scrambler output T, assuming the initial content of the registers to be zero. – From Fig. 7.19a we observer that initially T = S, and the sequence S enters the register and is returned as through the feedback path. This ( D 3 ⊕ D 5 ) S = FS new sequence FS again enters the register and is returned as F 2 S , and so on. Hence, (7.41) T = S ⊕ FS ⊕ F 2 S ⊕ F 3 S ⊕ ... = (1 ⊕ F ⊕ F 2 ⊕ F 3 ⊕ ...) S
Complete version
Dr. Shi
Lathi & Ding-Digital Commun
73
Example 7.2 … • Recognizing that We have
(
)(
F = D3 ⊕ D5
)
F 2 = D 3 ⊕ D 5 D 3 ⊕ D 5 = D 6 ⊕ D10 ⊕ D 8 ⊕ D8
• Because modulo-2 addition of any sequence with itself is zero, D 8 ⊕ D 8 = 0 and F 2 = D 6 ⊕ D10
• Similarly,
(
)(
)
F 3 = D 6 ⊕ D10 D 3 ⊕ D 5 = D 9 ⊕ D11 ⊕ D13 ⊕ D15 Dr. Shi
Lathi & Ding-Digital Commun
74
37
Example 7.2 … …… and so on …… Hence, T = (1⊕ D3 ⊕ D5 ⊕ D6 ⊕ D9 ⊕ D10 ⊕ D11 ⊕ D12 ⊕ D13 ⊕ D15 ⊕...)S
Because DnS is simply the sequence S delayed by n bits, various terms in the preceding equation correspond to the following sequences:
Dr. Shi
Lathi & Ding-Digital Commun
75
Example 7.2 … S = 1010101000 00111 D 3 S = 0001010101 00000111 D 5 S = 0000010101 0100000111 D 6 S = 0000001010 1010000011 1 D 9 S = 0000000001 0101010000 0111 D 10 S = 0000000000 1010101000 00111 D 11 S = 0000000000 0101010100 000111 D 12 S = 0000000000 0010101010 0000111 D 13 S = 0000000000 0001010101 00000111 D 15 S = 0000000000 000001010101000 00111 T = 1011100011 01001 Dr. Shi
Lathi & Ding-Digital Commun
76
38
Example 7.2 … • Note that the input sequence contains the periodic sequence 10101010…, as well as a long string of 0’s. • The scrambler output effectively removes the periodic component as well as the long strings of 0’s. • The input sequence has 15 digits. The scrambler output up to the 15th digit only is shown, because all the output digits beyond 15 depend on the input digits beyond 15, which are not given. • We can verify that the descrambler output is indeed S when this sequence T is applied at its input. • Homework: Learn to be able to determine where no terms needs to be considered. Dr. Shi
Lathi & Ding-Digital Commun
77
Regenerative Repeater •
Three functions: 1. Reshaping incoming pulses using an equalizer. 2. Extracting timing information Required to sample incoming pulses at optimum instants. 3. Making decision based on the pulse samples
Dr. Shi
Lathi & Ding-Digital Commun
78
39
Regenerative Repeater…
Dr. Shi
Lathi & Ding-Digital Commun
79
Preamplifier and Equalizer • A pulse train is attenuated and distorted by transmission medium, say, dispersion caused by an attenuation of high-frequency components. • Restoration of high frequency components ⇒ increase of channel noise • Fortunately, digital signals are more robust. – Considerable pulse dispersion can be tolerated
• Main concern: Pulse dispersion ⇒ ISI ⇒ increase error probability in detection Dr. Shi
Lathi & Ding-Digital Commun
80
40
Zero-Forcing Equalizer • Detection decision is based solely on sample values. ⇒ No need to eliminate ISI for all t. All that is needed is to eliminate or minimize ISI at their respective sampling instants only. • This could be done by using the transversal-filter equalizer, which forces the equalizer output pulse to have zero values at the sampling (decisionmaking ) instant. (refer to the diagram in next slide) Dr. Shi
Lathi & Ding-Digital Commun
81
Dr. Shi
Lathi & Ding-Digital Commun
82
41
Zero-Forcing Equalizer … c0 =1 ck = 0 ∀k ≠ 0
• Let
tap setting
⇒ P0 (t ) = Pr (t − NTb ) P0 (t ) = Pr (t )
If ignore delay.
• Fig 7.22 (b) indicates a problem : a1, a-1, a2, a-2, …are not negligible due to dispersion. • Now, want to force a1= a-1= a2= a-2 =…= 0 • Consider ck’s assume other values (other tap setting). Dr. Shi
Lathi & Ding-Digital Commun
83
Zero-Forcing Equalizer … • Consider ck’s assume other values (other tap setting). N
p0 ( t ) =
∑c
n pr ( t
− nTb )
n =− N
N
p0 ( kTb ) =
∑c
n pr [(k
− n )Tb ],
k = 0, ± 1, ± 2 ,...
n =− N
• For simplicity of notation: p0 ( k ) =
∞
∑c
n =−∞
n
pr ( k − n)
• Nyquist Criterion:
Dr. Shi
p0 ( k ) = 0
∀k ≠ 0
p0 ( k ) = 1
for k = 0 Lathi & Ding-Digital Commun
84
42
Zero-Forcing Equalizer … •
⇒ A set of infinitely many simultaneous equations:
2N+1: cn’s Impossible to solve this set of equations. • If, however, we specify the values of p0(k) only at 2N+1 points: k =0 1 p0 ( k ) = 0 k = ±1,±2,...,± N
then a unique solution exists. Dr. Shi
Lathi & Ding-Digital Commun
85
Zero-Forcing Equalizer … • Meaning: Zero ISI at the sampling instants of N preceding and N succeeding pulses. • Since pulse amplitude decays rapidly, ISI beyond the Nth pulse is not significant for N>2 in general.
Dr. Shi
Lathi & Ding-Digital Commun
86
43
Zero-Forcing Equalizer … • Page 7.37
Think about when n = 2, 1, 0, -1, -2; an example in next slide Dr. Shi
Lathi & Ding-Digital Commun
87
Zero-Forcing Equalizer … • Page 7.37
Dr. Shi
Lathi & Ding-Digital Commun
88
44
Eye Diagram • A convenient way to study ISI on an oscilloscope • Figure
Dr. Shi
Lathi & Ding-Digital Commun
89
Timing Extraction • • •
The received signal needs to be sampled at precise instants. Timing is necessary. Three ways for synchronization: 1.
2. 3.
Dr. Shi
Derivation from a primary or a secondary standard (a master timing source exists, both transmitter and receiver follow the master). Transmitting a separate synchronizing signal (pilot clock) Self synchronization (timing information is extracted from the received signal itself). Lathi & Ding-Digital Commun
90
45
Timing Extraction … • Way 1:
Suitable for large volumes of data high speed comm. Systems. High cost. • Way 2: Part of channel capacity is used to transmit timing information. Suitable for a large available capacity. • Way 3: Very efficient. • Examples: – On-off signaling (decomposition: Fig. 7.2) – Bipolar signaling rectification on-off signaling Dr. Shi
Lathi & Ding-Digital Commun
91
Timing Jitter • Small random deviation of the incoming pulses from their ideal locations. • Always present, even in the most sophisticated communication systems. • Jitter reduction is necessary about every 200 miles in a long digital link to keep the maximum jitter within reasonable limits. • Figure 7.23 Dr. Shi
Lathi & Ding-Digital Commun
92
46
Figure 7.23
Dr. Shi
Lathi & Ding-Digital Commun
93
Detection –Error Probability • Received signal = desired + AWGN Example: Polar signaling and transmission (Fig. 7.24, the next slide) Ap : Peak signal value • Polar: Instead of ± Ap , received signal = ± Ap + n • Because of symmetry, the detection threshold = 0. – If sample value > 0, ⇒ “1” – If sample value < 0, ⇒ “0” – Error in detection may take place
Dr. Shi
Lathi & Ding-Digital Commun
94
47
Figure 7.24
Dr. Shi
Lathi & Ding-Digital Commun
95
Error Probability for Polar Signal • With respect to Figure 7.24 P (∈ / 0 ) = probability that n > Ap P (∈ / 1) = probability that n < -Ap
p (n) =
n: AWGN ⇒ P(∈ 0) = =
Dr. Shi
1 2π
1
∞ Ap
∫σ
n
e
1
σ n 2π −
1 2 ( x) 2
∫
∞ Ap
2π σ n e
1 n − 2 σn
e
1 n − 2 σ n
2
2
dn
x=
n
σn
Ap dx = Q σn Lathi & Ding-Digital Commun
96
48
Error Probability for Polar Signal… 1 Q( y ) = 2π
where • Similarly,
∫
∞
y
e
−
x2 2
dx
Ap P(∈ 1) = Q σn
− Ap to − ∞ Symmetric
• Summary:
P (∈) = P (∈,0) + P (∈,1)
Here, assume P(0)= P(1)=1/2.
= P (∈ 0) P(0) + P (∈ 1) P(1) =
A 1 [P(∈ 0) + P(∈ 1)] = Q p 2 σn
• Q(x):
Complementary error function: erfc (x) 1 Q ( x) ≅ x 2π
Dr. Shi
x2
0 .7 − 1 − 2 e 2 , x > 2 x Lathi & Ding-Digital Commun
97
Error Probability for Polar Signal… • If
Ap
σn
= k , P(∈ 0) = P(∈ 1) = P(∈) = Q(k )
k Q(k)=P(ε)
1 0.1587
e.g. 3.16 ×10−5 ⇒ one of wrongly
Dr. Shi
2
3
4
5
6
0.0227 0.00135 3.16 ×10−5 2.87 ×10 −7 9 . 9 × 10 − 9
3.16 × 105
pulses will be possibly detected
Lathi & Ding-Digital Commun
98
49
Error Probability for On-off Signals • In this case, we have to distinguish between Ap and 0. 1 Ap • Threshold: 2 • Error Probabilities: A 1 P(∈ 0) = prob. of
P(∈ 1) = prob. of P(∈) =
n>
p Ap = Q σ 2 n Ap −1 n< Ap = Q 2 2σ n
A 1 [P(∈ 0) + P(∈ 1)] = Q p 2 2σ n
2
Also assume 0 and 1 equally probable
P(∈) > P(ε ) (in the case of polar signaling) Dr. Shi
Lathi & Ding-Digital Commun
99
Dr. Shi
Lathi & Ding-Digital Commun
100
50
Error Probability for Bipolar Signals 1 ↔ Ap or − Ap 0 ↔ 0v
⇒
A
A
p p If the detected sample value: ∈ − 2 , 2 ↔ 0
Ap P(∈ 0) = prob. n > 2 Ap A + prob. n < − p = prob. n > 2 2
o.w. ↔ 1
Ap A = 2Q p = 2 prob. n > 2 2σ n Dr. Shi
Lathi & Ding-Digital Commun
101
Error Prob. for Bipolar Signals… Ap when P (∈ 1) = prob . n < − positive pulse used 2 Ap when negative or prob . n > pulse used 2 Ap = Q 2σ n 1 P (∈) = P (∈ 0 ) P ( 0 ) + P (∈ 1) P (1) = [P (∈ 0 ) + P (∈ 1) ] 2 Ap > P (ε ) = 1 . 5Q for on − off signaling 2σ n Dr. Shi
Lathi & Ding-Digital Commun
102
51
Detection Error Probability… Ap Q σn A on-off is in middle, Q p 2σ n A bipolar is the worst, 1.5Q p 2σ n
• Summary, polar is the best,
Assumption: “0” and “1” Equally likely
• Another factor: P(∈) decreases exponentially with the signal power. Dr. Shi
Lathi & Ding-Digital Commun
103
Comparison among three line codes • To obtain:
P(∈) = 0.286 × 10−6 Ap
• We need:
σn Ap
σn Ap
σn
=5
for polar case
= 10
Q Q(5) = 0.286 × 10−6
Ap Q P (∈) = Q 2 σ n for bipolar case Q P (∈) = 1.5Q Ap 2σ n −6 = 0.286 × 10
for on-off case
= 10 . 16
c Dr. Shi
Lathi & Ding-Digital Commun
Ap 2σ n
= 5.08 104
52
Comparison among three line codes… Ap
• For the same error rate, required SNR σ n , Polar < On-off < Bipolar
• For the same
A SNR σ p n
(from previous example)
, the caused error rate
Polar < On-off < Bipolar
(from next example)
• Polar is most efficient in terms of SNR vs. error rate • Between, on-off and bipolar: Only 16% improvement of on-off over bipolar • Performance of bipolar case ≈ performance of on-off case in terms of error rate. Dr. Shi
Lathi & Ding-Digital Commun
105
Example a) Polar binary pulses are received with peak amplitude Ap = 1 mV. The channel noise rms amplitude is 192.3 µV. Threshold detection is used, and 1 and 0 are equally likely. b) Find the error probability for (i) the polar case, and the on-off case (ii) the bipolar case if pulses of the same shape as in part (a) are used, but their amplitudes are adjusted so that the transmitted power is the same as in part (a) Dr. Shi
Lathi & Ding-Digital Commun
106
53
Example … a) For the polar case
Ap
σn
=
10 −3 = 5 .2 192.3(10 −6 )
From Table 8.2 (4th ed), we find
P (∈) = Q ( 5 . 2 ) = 0 . 9964 × 10 − 7
b) Because half the bits are transmitted by nopulse, there are, on the average, only half as many pulses in the on-off case (compared to the polar). Now, doubling the pulse energy is accomplished by multiplying the pulse by 2 (in order to keep the power dissipation same) Dr. Shi
Lathi & Ding-Digital Commun
107
Example… Thus, for on-off Ap is 2 times the Ap in the polar case.
•Therefore, from Equation (7.53) A P(∈) = Q p = Q (3.68) = 1.66 × 10 − 4 2σ n
•As seen earlier, for a given power, the Ap for both the on-off and the bipolar cases are identical. Hence, from equation (7.54) Ap P (∈) = 1.5Q 2σ n Dr. Shi
= 1.749 × 10 −4
Lathi & Ding-Digital Commun
108
54
M-ary Communication • Digital communications use only a finite number of symbols • Information transmitted by each symbol increases with M.
IM = log2 M bits
IM
: information transmitted by an M-ary symbol
• Transmitted power increases as M2 , i.e., to increase the rate of communication by a factor of log2 M , the power required increases as M2. (see an example below) Dr. Shi
Lathi & Ding-Digital Commun
109
M-ary Communication… • Most of the terrestrial digital telephone network: Binary • The subscriber loop portion of the integrated services digital network (ISDN) uses the quarternary code 2BIQ shown in Figure. 7.28
Dr. Shi
Lathi & Ding-Digital Commun
110
55
Dr. Shi
Lathi & Ding-Digital Commun
111
Pulse Shaping in Multi-amplitude Case • Nyquist Criterion can be used for M-ary case Controlled ISI • Figure 7.28: One possible M-ary Scheme • Another Scheme: Use M orthogonal pulses: ϕ1 (t ), ϕ 2 (t ),..., ϕ M (t ) • Definition: Tb c i = j ϕ i ( t )ϕ j ( t ) = 0 0 i ≠ j
∫
Dr. Shi
Lathi & Ding-Digital Commun
112
56
Pulse Shaping in Multi-amplitude Case • The figure in next slide: One example in M orthogonal pulses: 2π ⋅ k ⋅ t , 0 < t < Tb k = 1, 2,..., M sin Tb ϕk (t ) =
0, otherwise
• In the set, pulse frequency: k
1 1 2 M : , ,..., Tb Tb Tb Tb
M times that of the binary scheme
(see an example below)
Dr. Shi
Lathi & Ding-Digital Commun
113
Dr. Shi
Lathi & Ding-Digital Commun
114
57
Pulse Shaping in Multi-amplitude Case • In general, it can be shown that the bandwidth of an orthogonal M-ary scheme is M times that of the binary scheme – In an M-ary orthogonal scheme, the rate of communication is increased by a factor of log 2 M at the cost of an increase in transmission bandwidth by a factor of M. Dr. Shi
Lathi & Ding-Digital Commun
115
Digital Carrier Systems • So far: Baseband digital systems – Signals are transmitted directly without any shift in frequency. – Suitable for transmission over wires, cables, optical fibers.
• Baseband signals cannot be transmitted over a radio link or satellites. – Since it needs impractically large size for antennas – Modulation (Shifting signal spectrum to higher frequencies is needed) Dr. Shi
Lathi & Ding-Digital Commun
116
58
Digital Carrier Systems… • A spectrum shift to higher frequencies is also required when transmitting several messages simultaneously by sharing the large bandwidth of the transmission medium, • FDM: Frequency-division Multiplexing. (FDMA: Frequency-division Multiplexing Access) Dr. Shi
Lathi & Ding-Digital Commun
117
Several Types of Modulation • Amplitude-Shift Keying (ASK), also known as onoff keying (OOK) m(t ) cos(ω c t ) : m(t ) : on-off baseband signal (modulating signal)
cos(ω c t )
Dr. Shi
: carrier
Lathi & Ding-Digital Commun
118
59
Modulation Types… • Phase-Shift Keying (PSK) m(t ) : polar Signal 1 ↔ p(t) cos(ωct),0 ↔ − p(t) cos(ωct) = p(t) cos(π −ωct) = p(t) cos(ωct −π )
Dr. Shi
Lathi & Ding-Digital Commun
119
Modulation Types… • Frequency-Shift Keying (FSK) – Frequency is modulation by the base band signal.
1 ↔ ω c1 ,0 ↔ ω c0 – Information bit resides in the carrier frequency.
Dr. Shi
Lathi & Ding-Digital Commun
120
60
PSD of ASK, PSK & FSK
Dr. Shi
Lathi & Ding-Digital Commun
121
FSK • FSK signal may be viewed as a sum of two interleaved ASK signals, one with a modulating frequency ω c , the other, ω c 0
1
– Spectrum of FSK = Sum of the two ASK – Bandwidth of FSK is higher than that of ASK or PSK
Dr. Shi
Lathi & Ding-Digital Commun
122
61
Demodulation A. Review of Analog Demodulation • Baseband signal and bandpass signal The end of the 2nd part of the slides of Ch. 2 and Ch. 3
• Double-sideband Suppressed Carrier (DSB-SC) Modulation m(t) ↔ M(ω)
message signal
1 cos(ωct) ↔ [δ (ω +ωc ) +δ (ω −ωc )] carrier 2 1 m(t)cos(ωct) ↔ [ M(ω +ωc ) + M(ω −ωc )] modulated signal 2 Dr. Shi
Lathi & Ding-Digital Commun
123
Figure 4.1 (a), (b), (c) DSB-SC Modulation
124
62
DSB-SC Demodulation • Demodulation: Synchronization Detection (Coherent Detection) • Figure 4.1 (e) and (d)
Using a carrier of exactly the same frequency & pulse
Dr. Shi
125
DSB-SC Demodulation …
[m(t ) cos(ωct )]cos(ωct ) = e(t ) = 1 [m(t ) + m(t ) cos(2ωct )] 2
1 1 M (ω ) + [M (ω + 2ωc ) + M (ω − 2ωc )] 2 4 1 Aftera LPF ⇒ M (ω ) 2
E (ω ) =
1 m(t ) 2
Dr. Shi
126
63
Figure 4.1 (d), (e) DSB-SC demodulation
127
AM Signal • Motivation: – Requirements of a carrier in frequency and phase synchronism with carrier at the transmitter is too sophisticated and quite costly – Sending the carrier ⇒ a. eases the situation b. requires more power in transmission c. suitable for broadcasting Dr. Shi
128
64
Amplitude Modulation (AM) ϕ AM (t ) = A cos(ω c t ) + m(t ) cos(ω c t ) = [A + m(t )]cos(ω c t ) ϕ AM (t ) ⇔
1 [M (ω + ω c ) + M (ω − ω c )] 2
+ π ⋅ A[δ (ω + ω c ) + δ (ω − ω c )]
Dr. Shi
129
• AM Signal and its Envelope
130
65
• Demodulation of AM signals – Rectifier Detection:
Figure 4.11 Output:
– Envelope Detector:
1
π
m(t )
Figure 4.12 Output: A + m(t )
Dr. Shi
131
132
66
133
B. Back to Digital Communication System: Demodulation •
Demodulation of Digital-Modulated Signals is similar to Demodulation of Analog-Modulated Signals
•
Demodulation of ASK Synchronous (Coherent) Detection Non-coherent Detection: Say, envelope detection
Dr. Shi
Lathi & Ding-Digital Commun
134
67
Demodulation of PSK • Cannot be demodulated non-coherently (envelope detection). – Since, envelope is the same for 0 and 1
• Can be demodulated coherently. • PSK may be demodulated non-coherently if use: Differential coherent PSK (DPSK)
Dr. Shi
Lathi & Ding-Digital Commun
135
Differential Encoding • Differential coding: – “1” encoded by the same pulse used to encode the previous data bit no transition – “0” encoded by the negative of pulse used to encode the previous data bit transition
• Figure 7.30 (a) • Modulated signal consists of pulses ± A cos(ω ct ) with a possible sign ambiguity Dr. Shi
Lathi & Ding-Digital Commun
136
68
Differential Encoding • Figure
Dr. Shi
Lathi & Ding-Digital Commun
137
Differential Encoding… • Tb : one-bit interval • If the received pulse is identical to the previous pulse 1, A2 Y (t ) = A2 cos 2 (ω c t ) = [1 + cos(2ω ct )] 2
z (t ) =
2
A 2
• If 0 is transmitted
Y (t ) = − A2 cos 2 (ω c t ) = − z (t ) = −
Dr. Shi
A2 [1 + cos(2ω ct )] 2
A2 2 Lathi & Ding-Digital Commun
138
69
Demodulation of FSK • FSK can be viewed as two interleaved ASK signals with carrier frequency ω c and ω c , respectively. • Hence, FSK can be detected by non-coherent detection (envelope) or coherent detection technique 0
Dr. Shi
Lathi & Ding-Digital Commun
1
139
Demodulation of FSK
Dr. Shi
Lathi & Ding-Digital Commun
140
70
Demodulation of FSK • In (a), non-coherent detection (envelope), ωc
– H0(ω)
0
Turned to – H1(ω)
respectively ωc
1
• Comparison: above or bottom Whose output is large 0 or 1
Dr. Shi
Lathi & Ding-Digital Commun
141
71