Académique Documents
Professionnel Documents
Culture Documents
1
Baseband Modulation
2
Need for Baseband Modulation
• Due to the limited available system bandwidth, only the major portion of
a digital signal spectrum can be transmitted and restored. Even if there is
no loss or noise in the communication system, the received signal will
have distortion due to the limited channel bandwidth.
time
time
BANDLIMITED an
2 0 → -1 S(t)
1 → 1
0 T
4 00 → -3 4-LEVEL
01 → 1
11 → 1
10 → 3 0 T 2T
1/-p(t) 0/0
0/0
The modulator has two states, labeled “+” and “-”. The modulator responds to a source symbol
0 with a zero waveform and to a source symbol 1 with the waveform p(t) or -p(t) depending on
whether its state is “-” or “+” respectively.
Note that the choice of the states is not obvious. A common mistake is to choose the output levels,
0, 1, and -1 as states. But, if you make this choice, it is hard/impossible(?) to capture the memory
of the system.
5
Examples of Baseband Impulse Responses
h (t) |H(f)|
(1) NRZ (non return to zero)
0 T t 0 1/T f
(4) Manchester
e.g. Ethernet on Coaxial
0 T Cable (3-10 Mbps)
0 1 2 3
T T T 6
Examples of Baseband Impulse Responses (continued)
h (t) |H(f)|
0 T 0 1
2T
t
~10 ms
0 f
Twisted Copper Pair with Transformers (DSL)
h(t)
|H(f)|
~DFE
ang H(f)
t
0 f
Radio h(t)
Nulls require a non-linear
|H(f)| ang H(f) equalizer: DFE or MLSE/VA
t
I/Q Equalizer
Gitlin1_8
0 f0 f
Pulse Distortion Due to MultiPath
9
Digital Baseband Modulation Techniques
NRZ
0 Freq.
1 to 1
2T T
0 Freq.
11
Coded Baseband Digital Modulation Techniques
• Coded Systems
• Can be combined with any of the above
• Performance improvement at the cost of increased receiver delay
and complexity
• Allows rate to approach channel capacity
• Types: - block coding
- convolutional or trellis coding
Uncoded System
Error
Probability
0 T
At A:
{an}
∑ anδ(t-nT)
n
T
Timing Margin
r(t) = ∑ anx(t-nT)
At B:
n
r(t) T 2T 3T 4T 5T 6T 7T 8T PAM
At D:
{ân}
x(t)
• Ideal Pulse t
sinπWt
πWt
X(f)
• Minimum Bandwidth Pulse
1/T Symbols/sec
1
2W
f
-W = - 1 0 W= 1
2T 2T
Nyquist (1928) 14
Why Equalization Is Needed
(A)
Amplitude
0 F1
Frequency
Amplitude
(B)
P(t)
Transmitted Pulse
-3T -2T -T 0 T 2T 3T
Time
(C)
P(t) + P(t-T)
-3T -2T -T 0 T 2T 3T
Time
15
INTERSYMBOL INTERFERENCE: ISI
• Cause of IS I:
16
EQUALIZATION: Pulse Shaping
y(t) = 0 at t = nT [n = 1, 2, 3, . . . ] 17
The Eye Pattern [as a measure of distortion]
0 T 2T 3T 4T 5T 6T 7T 8T
r (kT ) = ak x(0) + ∑ ak − m xm
m≠ 0
Margin
for Error
Due To
Noise
Threshold
Level Margin
for Error
Due to
Timing
Sample Instant
Scope Trigger
From Modem
-0.2 0.5
0
π/T π/T
2π
(a) (b)
π
T, 0 ≤ ω ≤ (1 − α )
sin πt / T cos απt/T T
x (t) = X (ω) =
π
T 1 − sin T ω − , π (1 − α) ≤ ω ≤ π (1 + α )
πt / T 1 - 4α 2 t2 / T 2
2 2α T T T
Raised Cosine Pulse Shaping: (a) Frequency Response, (b) Time Response
20
Performance of PAM Systems: No ISI/Multipath
(Gaussian Noise)
1
10-1
10-2
Pe 10-3 L=2 4 8 16
10-4
10-5
10-6
0 5 10 15 20 25 30 35
Ps / P N [SNR], dB
Probability of error Ps for L-level PAM, where P s / PN is the
signal-to-noise power ratio in the Nyquist bandwidth. Note
that doubling the bit rate requires about 6-7 dB of SNR
1 3 Ps
1/ 2
1
Pe = 1 − Pr {η > d} = 2 1 − Q 2
L L L - 1 PN
1 ∞ −t 22
Q(x ) = ∫ e dt
2π x
21
Adaptive Equalization
Input Output
binary binary
data {a(n)} Transmitting y(n) data
Σ
Pulse Receiving Adaptive Decision
Channel
generator filter filter equalizer device
Sampler
Decision
device
Switch Stored
Input Adaptive reference
signal equalizer generator
-
Error
Σ + Desired Response = Replica of transmitted signal
22