chapter 4 baseband pulse transmission - fju.edu.tw
TRANSCRIPT
Digital Communication輔仁大學 電子工程系所
Chapter 4Baseband Pulse Transmission
4 - 1Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Outline
•Matched filter which is the optimum system for detecting a known signalim AWGN
•Calculation of bit error rate due to channel noise
• Intersymbol Interference (ISI) which arises when channel is dispersive
•Nyquest’s criterion for distortionless baseband data transmission
•Correlative-level codingfor combating ISI
•Digital subscriber lines
•Equalization of a dispersive baseband channel
•Eye Pattern for displaying combined effects of ISI and channel noise
4 - 2Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Introduction
•Transmission of digital data—Baseband channel: digital computer—Bandpass channel: microwave radio, satellite channelmodulation process
•Baseband transmission: low-pass channel BW > data stream BW
•Major sources of bit errors—Noise: optimum detection in presence of noise Matched filter—Intersymbol interference (ISI):> due to non-ideal low-pass channel
> can be reduced using pulse shaping and equalization
AWGN Channel ISI Channel
Shaping Filter
4 - 3Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Matched Filter
•Optimum design of receiver filter for detecting a pulse transmitted over achannel corrupted by additive noise
x t g t w t+= 0 t T
filter input pulse signal white noise
T: an arbitrary observation interval
y t go t n t+=
g t h t w t h tfilter output
go T2
E n2 t ----------------------=
maximizing the peak pulse SNRmeasured at time t = T
useschwarz’s inequality
hopt t kg T t– =
Hopt f kG* f j2fT– exp=
: a time-reversed and delay version of g(t)
Linear receiver
4 - 4Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Properties of Matched Filter
T2---
Tt
A2---
A2---– A2T 8
A2T 4
tT 2 3T 4 2T
go t
T
hopt tg t
•Peak pulse SNR depends on ratio of signal to power spectral density of noise
Go f Hopt fG f k G f2 j2fT– exp= =
go T Go f j2fT exp fd–
k G f2 fd–
kE= = =
Fourier of resulting matched filter output go(t)
matched filter output at time t = T
E g2 tfd–
G f2 fd–
= = signal energy = integral of square magnitude spectrum of a pulse signal
E n2 t k2No
2---------- G f2 fd
–
k2No E2
--------------= = average output noise power
maxkE 2
k2NoE 2--------------------------- 2E
No------= =peak pulse signal-to-noise ratio =
E No signal energy-to-noisespectral density ratio
=
4 - 5Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Matched Filter for Rectangular Pulse
•Correlator configurationhopt t kg T t– =
y t h t x t g T – x t – d0
T
= =
y T g T – x T – d0
T
g x d0
T
= =
T T – =
td0T
x t g t w t+=
g t
y t y T
y T g tx ttd0
T
=
• For a rectangular pulse, matched filter may beimplemented using a circuit “Integrate-and-dump”
Integrate-and-dump circuit
rectangular pulse
Matched filter output
Integrator output
4 - 6Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Output of Matched Filter
•Consider a binary PCM system based on polar non-return-to-zero (NRZ)—symbols 1 and 0 are represented by positive and negative rectangular pulses—Channel noise is modeled as additive white Gaussian noise w(t) of zero
mean and power spectral density No/2
Receiver for baseband transmission of binary-encoded PCM wave using polar NRZ signaling.
x t+A w t, symbol 1+A– w t, symbol 0+
=y x ttd
0
Tb A– 1
Tb----- w ttd
0
Tb+= =
Received signal Matched filter output sample at time t = Tb
kATb=1 Gaussian distributed
4 - 7Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Output Noise Variance
•The output y represents sample value of a random variable Y—Random variable Y is Gaussian distributed with a mean A—Variance of random variable Y is Y
2 n2=
(b) Probability density function of Y when 1
Noise analysis of PCM system.
(a) Probability density function of random variable Yat matched filter output when 0 is transmitted. is transmitted.
n t 1Tb----- w ttd
0
Tb=
Output noiseE n2 t 1
T b2------ E w ttd
0
Tb w uud
0
Tb
1T b
2------ E w tw u td ud0
Tb0
Tb= =
1T b
2------No
2----- ud
0
Tb= 1
T b2------ Tb
No
2-----
No
2Tb--------
No
2----- 1
Tb----- = = =
No t u– noise variance
2W 1 Tb=equivalent noise BW
4 - 8Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Error Rate Analysis
fY y 0 12Y
2----------------- y A+ 2–
2Y2
----------------------exp=
1No Tb
---------------------- y A+ 2–No Tb
----------------------exp=
conditional probability of errorgiven symbol = 0
p10 P Y symbol 0 fY y 0 yd0
= =0 0=
p101
No Tb---------------------- y A+ 2–
No Tb----------------------exp yd
0
=
1No Tb
---------------------- A2–No Tb--------------exp yd
A
=1
------- z2– exp zdANo Tb
-----------------
=
z y No Tb =
12--- erfc
ANo Tb
-----------------= 12--- erfc
A2Tb
No---------- 1
2--- erfc
Eb
No-----= =
erfc u1
------- z2– exp zdu=
Q x 12--- erfc
x2
-------=
Eb = A2Tb = bit energy
4 - 9Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Probability of Error
• Average probability of symbol error is abinary symmetric channel dependssolely on Eb/No
- Exponential improvement inaverage probability of symbol
- decreases very rapidly as Eb/No isincreased
- small increase in Eb error free
p10 p01=
Pe p0 p10 p1 p01+ p10= =
12--- erfc
Eb
No-----=
p0 p1 1 2= =
average probabilityof symbol error
Eb /No: ratio of transmitted signal
per bit to noise spectral density
erfc u u2– exp2 u
-------------------------PeEb No– exp
2 Eb No --------------------------------
upper bound
Probability of error in a PCM receiver
4 - 10Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
M-ary PAM Transmission
1.5
-1.5
0.5
-0.5T0 2T 3T 4T
00 01 10 11
T=2Tb
•M-ary PAM is faster than thecorresponding binary PAM
- 1 baud = bits/sec.
- 1/T : signaling rate = bauds rate =symbols/sec.
•Given the same average Pe, M-aryPAM needs more transmit power
Log2 M
Tb Log2 M
3AA-A-3A–
Pe1M---- M 2– P n A 2P n A + =
M 1–M
------------P n A = M 1–M
------------erfcANo T
----------------=
x2 t E ak2 2
M---- 2m 1– 2A2
m 1=
M 2
A2 M 2 1–3
-------------- Pav= = = =
PeM 1–M
------------erfcANo T
---------------- M 1–M
------------erfcPav3TM2 1– No
------------------------= =
1 1M----–
erfc3Es
M 2 1– No-------------------------= Es Eb Log2 M=
4 - 11Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Intersymbol Interference
Baseband binary data transmission system
• Intersymbol Interference (ISI)
- Due to dispersive channel
• A key question: A pulse shape of interest
- Discrete Pulse-amplitude-amplitude (PAM)
Transmitted signal (Binary PAM)s t ak g t kTs–
k=
y t ak p t kTs– k n t+=
Received signalp t g t h t c t =
p 01= p(t) is normalized
P f G fH fC f=
scaling factor. amp. changes
y(t) is sampled at time ti = iTb
y ti ak p i k– Tb k –=
n ti +=
ai ak p i k– Tb k –=k i
n ti + +=
through system Pe12-----E erfc
A +No Tb
------------------=
12-----E erfc
A2
N0Tb----------
N0Tb----------+
p 0A= ISI = SNR A2
N0Tb----------=
S/N is highPerformanceis limited byISI rather thannoise
2L com.
4 - 12Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Nyquist’s Criterion
•Nyquist’s Criterion for distortionless baseband Binary transmission
- No ISI for bandlimited channel
Ideal magnitude response Ideal basic pulse shape
P f nRb– n –=
Tb= P f t 1–
p i k– Tb 1 , i k=0 , i k
=
Pf Rb P f nRb– n –=
p 0 1= =
Fourier Transform
sampling time
Rb 1 Tb=
t 1For zero ISI
•Consider rectangular function
P f 12W------
f2W------ =
p t c 2Wt sin=
2W 1 Tb Rb= =
Nyquest rate (Rect.) = Rb= 2W,W: Nyquest BW
Tb
1–Tb
P f nRb– n –=
t
Rb– Rb 2 RbRb 2f t
Tb Tb 1=
4 - 13Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Rectangular Spectrum
• Two difficulities for rectangular function
- undesirable because of the abrupttransmission at the band edges
- p(t) decays as for large slowdecay; due to timing error
W
1 t tISI
timing error
y t a0 c 2Wt sin 2Wt sin
-------------------------------1– kak
2Wt k– ---------------------------
k 0+=
desired symbol Diverge1k---
k 1=
P(f) is discontinuous p(t) ~ 1/t
d/df P(f) ............... p(t) ~ 1/t2
d2/df2 P(f) ............... p(t) ~ 1/t3
dn/df n P(f) ........... p(t) ~ 1/tn+1
A series of sinc pulse corresponding to sequence 1011010
1 2W
WW– fW
W– f
discontinuous
4 - 14Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Raised Cosine Spectrum
Frequency response
Time response
different rollofffactor
•Raised Cosine Spectrum
- overcome the two above practicaldifficulties with ideal Nyquest channel
- Extend the bandwidth from the minimumvalue W = Rb /2 to W ~ 2W
- dn/df n P(f) ........... p(t) ~ 1/tn+1flat portion; rolloff portion
P f
12W------ , 0 f f1
14W------ 1 f W–
2W 2f1–------------------------sin–
, f1 f 2W f– 1
0 , f 2W f– 1
=
1 f1 W–=Rolloff factorexcess bandwidth over W = (W-f1)/W = 1-f1 /W
Transmission bandwidth: BT = W + (Wf1) = (1+)W
0 1
BT
4 - 15Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Characteristics of Raised Cosine
•Rolloff factor vs.Bandwidth
- Transmission bandwidth: BT = W +(W f1) = (1+)W
- =0 ideal Nyquest channel
- =1 BT = 2W ; full-cosine Rollofff
• Insensitive to sampling errors
•Deduce tails of pulse considerably
- =1 tails of p(t) are samllest
- Higher , Higher BT ; smaller ISI dueto timing error
• Time response of P(f)
• For=1
time response
Two interesting properties: useful inextracting a timing signal from the receivedsignal for synchronization
- p(t) = 0.5 at , pulse width of1/2 Amp. = Tb
- zero crossing at ,in addition to ,
price paid: BT (=1 )= 2W =2BT (=0 )
P f14W------ 1 f
2W------cos–
, f1 f 2W
0 , f 2W
=
p t 4Wt sin1 16W2t2–--------------------------=
t Tb 2=
t 3Tb 2= 5Tb 2t Tb= t 2Tb=
p t c 2Wt sin 2Wt cos1 162W 2t2–---------------------------------=
decay as 1 t2ideal Nyquest channel
4 - 16Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
A Example (BW of T1 System)
•T1 carrier; 24 channels; 8 bit PCM; Tb = 0.647msec;
—Minimum transmission bandwidth; Tb=1/2W
BT (= 0) = W =1/2Tb= 1/2(0.647ms)=> 722 kHz
—For full-cosine case =1
BT (= 1) = 2W =1/Tb= 1/(0.647ms=> 1.544 MHz
•For SSB (single-sided band): Bandwidth = 4 k Hz (Voice bandwidth =3.1k Hz); 24 Channels FDM—BT (SSB) = 4 (24)=196kHz << BT(T1)
4 - 17Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Optimum Linear Receiver
•Two channel conditions for baseband data transmission system—Channel noise acting alone matched filter receiver—Intersymbol interference (ISI) acting alone pulse-shaping transmit filter to
realize Nyquest channel
In a real-life situation: channel noise + Intersymbol interference
•Objective: design a linear optimum receiver for linear dispersive andnoisy channel—Zero-forcing equalizer: ISI is forced to zero at all instants;> Optimal if channel is zero noise; simple because effect of channel noise is ignored
> Performance is degraded due to noise enhancement
—A refined approach mean-square error criterion (MSE)> A balanced solution to reduce both channel noise and ISI
> computational complexity
> performs as well as and often better than zero-forcing equalizer
4 - 18Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
MMSE Receiver (I)
w t
channel Receiverfilter
c t
x t y tImpulse response of receiver filter = c(t)Channel output = x(t)
x t ak q t kTb– k w t+=
q t g t h t=Noise
TX filter Channel
y t c x t – d– c t x t= =
Receiver filter output
Resulting output y(t) at t = iTb
y iTb i= ni+
i ak c q iTb kTb –– d–
k=
ni c w iTb – d–=
Error signal
ei y iTb ai– i ni+ ai–= =
transmitted symbol
J 12---- E ei
2 =mean-square-error
J 12---- E i
2 E ni2 E ai
2 + + E ini E niai E iai ––+=0 0
: stationary,i E i independent of instant time t = iTb
Rq 1 2 Rq 1 2– q kTb 1– q kTb 2– k= =
E i2 Rq 1 2 c 1 c 2 1 2dd––
=
E ni2 c 1 c 2 RW 2 1– 1 2dd
––
=
No
2----- c 1 c 2 2 1– 1 2dd
––
=
E ai2 1=
E iai E ak ak c q iTb kTb –– d–
k=
c q – d–=
4 - 19Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
MMSE Receiver (II)
J 12---- E i
2 E ni2 E ai
2 + + E ini E niai E iai ––+=
12--- 1
2--- Rq t –
No
2-----t – +
c tc t dd––
c tq t– td
––+=
c tJ 0=
Rq t – No
2----- t – +
c d– q t– =
Fourier transform
Sq fNo
2-----+ C f Qf=
Sq f 1Tb----- Q f k
Tb-----+
2
k=
C f Qf
Sq fNo
2-----+
--------------------------=Optimum Linear receiver
power spectral density of sequence
g t G f
gt G f–
Gf g t–
q kTb
A matched filter q(-t)
A transversal (tapped-delay-line) equalizer
Heq f 1
Sq fNo
2-----+
--------------------------=
4 - 20Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Optimum Linear Receiver
Optimum linear receiver consisting of the cascade connection of matched filter and transversal equalizer.
• Practical considerations
- Channel is usually time varying in a real-life telecommunications environment
• Full transmission capacity of telephonechannel adaptive receiver
- Adaptive implementation of matchedfilter and equalizer
• Two factors contribute to pulse distortion ina public switched telephone network
- Difference in transmission of individuallinks that may be switched together
- Difference in number of links in aconnection
• Telephone channel is random
C(f)ck k N–=N
4 - 21Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Optimum Receiver with Finite-tap Equalizer
x iT ak p iT kT– k iT + xi= =
y iT yi cjxi j–j N–=
N
= =
Equalizer output
Equalizer input
cuJ 0=
J E yi ak– 2 =
cj RX i j–
j N–=
N
RX i– RX i= =
p t q t q t– =
t w t q t– =
autocorrelation of x
matched filter
Jmin c2 1 cj p i–
j N–=
N
– c2 1 go– = =
C fP fPf No
a2
------ Pf+ Pf=
as j
pi q t q t– t It==
as j cj pi j– p i j– –No
c2
------ pj i–+j N–=
N
p i–=Jmin c
2 1 go– c2T 1 P f+ 1– fd
1 2T–
1 2T
= =
c2 T2------- 1 P w+ 1– wd
T–
T
=
C f Heq f 1P f No a
2+----------------------------------= =
P f 1T--- Q f k
T---+
2
k –=
=
P w 1T--- Q f 2k
T----------+
2
k –=
=
w t
channel matchedfilter
q t–
x t
x iT Equalizer
y iT
4 - 22Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Adaptive Equalization
Block diagram of adaptive equalizer
LMS Algorithm
Two operating modes of an adaptive equalizer
• Two operatingmodes of anadaptive equalizer
- Training mode(Position 1)
- Tracking mode(Position 2)
4 - 23Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Decision-Feedback Equalization
Block diagram of decision-Feedback equalizer
yn hk xn k–k=
h0 xn hk xn k–k 0 hk xn k–
k 0+ +=
feedback feedforward
• Avoid noise enhancement• nonlinear and therefore more difficult to analyze• Error propagation: not persist indefinitely; occur in burst- L= number of feedback taps. L consecutive correct decisions
decision errors will be flushed out finite duration
- P(next decision error) < 1/2
- Error rate is increased by a factor ~ 2L ; L consecutive correctdecisions
4 - 24Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Distorted Pulse
•System parameters—QPSK modulation—Raised-cosine pulse shaping; Rolloff factor = 0.35—Two-ray mobile radio channel model—Infinite-length equalizer
-20 -15 -10 -5 0 5 10 15 20
Sampling time ( x T )
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Am
plitu
de
-20 -15 -10 -5 0 5 10 15 20
Sampling time ( x T )
-0.5
-0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5in-phase channel Quadrature channel
SNR=10dB
4 - 25Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Equalized Pulse
-20 -15 -10 -5 0 5 10 15 20
Sampling time ( x T )
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Am
plitu
de
-20 -15 -10 -5 0 5 10 15 20
Sampling time ( x T )
-0.5
-0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
-20 -15 -10 -5 0 5 10 15 20
Sampling time ( x T )
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Am
plitu
de
-20 -15 -10 -5 0 5 10 15 20
Sampling time ( x T )
-0.5
-0.4
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
0.5
Linear
DFE
in-phase Quadrature
in-phase Quadrature
SNR=10dB
SNR=10dB
4 - 26Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Eye Patterns (Diagram)
• Eye Patterns: An experimental tool for the channel (noise+ISI)
- Evaluate combined effect of these impairments (channel noise and intersymbolinterference) on overall system performance
- It resembles human eye for binary wave
- Interior region of eye pattern Eye opening
• An eye pattern provides a great deal of useful information about the performance
- Width of eye opening time intervalover which received signal can besampled without error from ISI; preferredtime eye is open the widest
- Sensitivity of system to timing errors rate of closure of eye as sampling time isvaried
- Height of eye opening noise margin ofthe system
4 - 27Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Eye Patterns (Channel Noise)
Noiseless
SNR = 20dB
SNR = 10dB
• Eye diagram of system under idealized conditions
- A quaternary (M=4) baseband PAM
- Source symbols are randomly generated
- no channel noise; no bandwidth limitation
- Raised-cosine pulse-shaping (W=0.5, = 0.5 Hz)
- M-1 = 3 openings; T Tblog2M 2Tb= =
• SNR=20dB; effectof noise is hardlydiscernibleeyeis open
• SNR=10dB;openings of eyediagram arebarely visible
4 - 28Chapter 4: Baseband pulse modulation Dr. Sheng-Chou Lin
Digital Communication輔仁大學 電子工程系所
Eye Patterns (Bandwidth limitation)
•Eye diagram of system under a bandwidth-limitedcondition and a noiseless channel
- Channel is modeled by a low-pass butterworthfilter, squared magnitude response is
, N : order of the filter
•N = 25 and fo = 0.975Hz; BT = W(1+) = 0.75 Hz
- Channel Bandwidth > BT
- A decrease in the size of eye openings
- There is a blurred region at time = 1 sec
•N = 25 and fo = 0.5Hz
- The channel bandwidth is reduced; ChannelBandwidth < BT
- Further reduce the extent to which eyes are open
H f2 11 f fo 2N+------------------------------=