UNITII

Baseband Pulse Transmission:

Introduction, Matched filter, Properties of Matched filter,

Matched filter for rectangular pulse, Error rate due to noise,
Inter-symbol Interference (ISI),Nyquist’s criterion for
distortion less baseband binary transmission, ideal Nyquist
channel, Raised cosine filter & its spectrum, Correlative
coding – Duo binary & Modified duo binary signaling
schemes, Partial response signaling, Baseband M-array
PAM transmission, Eye diagrams.
Matched Filter
 Matched filter is a device for the optimal detection of a
digital pulse. It is so named because the impulse response
of the matched filter matches the pulse shape.
 System model without ISI

channel
Design Criterion
 To find h(t) such that the output signal-to-noise ratio SNRO
is maximized.
x(t)  g(t)  w(t) for 0  t T
y(t)  [g(t)  w(t)]* h(t)
 g(t)* h(t)  w(t)* h(t)
 go (t) n(t)

| g o (T )|2
SNRO 
E[n2 (T)]
Analysis of Matched Filter
2


G( f )H ( f ) exp( j2fT )df
  
N0
2 
| H ( f ) |2 df

Cauchy-Schwarz inequality

  
2


  


1 (x)2 (x) dx  
| 1 (x) | dx
2

| 2 (x) |2 dx

with equality holding if, and only if,  1(x)  k  2* (x) for some constant k.
Analysis of Matched Filter


g o (t)   H ( f )G( f )exp( j2ft)df
2

| g o (T ) |   H ( f )G( f )exp( j2fT )df
2


With w(t) being white with PSD N 0 / 2,

N0
S N ( f )  SW ( f ) | H ( f ) | 2
| H ( f ) |2
2
 N0 
 E[n (T )]   S N ( f )df  
2
2
| H( f ) | df
 2 
 Analysis of Matched Filter
 By Cauchy-Schwarz inequality,
2
  

 H ( f )G( f ) exp( j2fT )df   | H( f ) | df  | G( f ) exp( j2fT ) |2 df

2
  

 

 | H( f ) | df  | G( f ) |2 df
2
2 
  
N0 

  | G( f )|2 df

2
| H ( f ) | df N0
2 
This is a constant bound, independent of the choice of h(t).
Hence, the optimal  is achieved by:

H ( f )  k G* ( f ) exp( j2fT )

Matched Filter for Rectangular Pulse
Also h(t)
Matched Filter for Rectangular Pulse
 hopt(t) in this example can be implemented as integrate-and-
dump circuit
 Error Rate due to Noise
 In what follows, we analyze the error rate of polar non-
return-to-zero (NRZ) signaling in a system with optimal
matched filter receiver over AWGN channel.
Intersymbol Interference
 The channel is usually dispersive in nature.
 In this section, we only consider discrete pulse-amplitude
modulation (PAM). Consideration of PDM and PPM will
be similar but out of the scope of this section.


bk {0,1}, ak  2bk 1 and s(t)   a g(t  kT ).
k
k b
Intersymbol Interference
 Notably, in the previous section, we only consider one
interval of input.
s(t)  I g(t)
This is justifiable because of no ISI.
 However, in this section, we have to consider

s(t)   a g(t  kT ).
k 
k b

since ISI is involved.

 We also assume perfect synchronization to simplify the
analysis.
 

 ak g(t  kTb )
Information of ak is carried at [kTb, (k+1)Tb).
s(t)  We sample at iTb = (k+1)Tb to retrieve ak.
k

x(t)  s(t) * h(t)  w(t)   a [g(t  kT ) * h(t)]  w(t)
k 
k b


y(t)  x(t) * c(t)   a [g(t  kT ) * h(t) *c(t)]  w(t) *c(t)
k
k b


y(iTb )   a [g(t  kT ) * h(t) *c(t)]
k 
k b
 w(t)*c(t)
tiTb
tiTb
 
g(t  kTb ) * h(t) *c(t)   G( f )exp{ j2fkTb }H ( f ) C( f )exp{ j2ft}dt

  G( f ) H ( f ) C( f ) exp{ j2f (t  kTb )}dt
 p(t  kTb )

where p(t)   G( f )H ( f )C( f )exp{ j2ft}dt.

 y(t)  a
k 
k p(t  kTb )  n(t), where n(t)  w(t) *c(t)

 y(iTb )  a
k 
k p(iTb  kTb )  n(iTb )
 
y((i 1)Tb )  a
k 
k p((i 1 k )Tb )  n((i 1)Tb )  ai p(Tb )  n((i 1)Tb )

As a result, without ISI and additive noise,


y((i 1)Tb )  a
k 
k p((i  1 k)Tb )  ai p(Tb )

and {ai } can be completely reconstructed by{y((i  1)Tb )}.

The text sets p(0) = 1 for simplicity. but this might be a little confused
(See Slide 4-28)! The text is correct when the information of ai is
carried during [(i1)Tb, iTb).

Information of ai is actually carried during [iTb,(i+1)Tb).

So, in order to recover ai, “correlation” (convolution) operation should
start at iTb, and end (i.e., is sampled) at (i+1)Tb.
Hence, y((i+1)Tb) is used to reconstruct ai.
Nyquist’s Criterion for Distortionless Baseband
Binary Transmission
 Is it possible to completely eliminate ISI (in principle) by
selecting a proper g(t) ?
Nyquist’s Criterion for Distortionless
Baseband Binary Transmission
 Let P(f) = G(f)H(f)C(f).
 Sample p(t) with sampling period Tb to produce P(f).
 From Slide 3-4, we get:

1   n
P ( f )   P
Tb n 
f  
Tb 

 Also from Slide 3-4, we have:


P( f )   p(nTb ) exp j2nTb f   1
n
( !
1, if n = 0
Because p(nT b) =
0, if n 6= 0
Nyquist’s Criterion for Distortionless Baseband
Binary Transmission
 This concludes that the condition for zero ISI is:

 n
 P
n  
f   T b
Tb 

 This is named the Nyquist criterion.

 The overall system frequency function P(f) suffers no
ISI for samples taken at interval Tb if it satisfies the
above equation.
 Notably, P(f) represents the overall accumulative effect
of transmit filter, channel response, and receive filter.
 Ideal Nyquist Channel
 The simplest P(f) that satisfies Nyquist criterion is the
rectangular function:

T , | f | W  1
 b 2Tb
P( f )   and P(W )  P(W )  Tb .
1
 0, | f | W 
 2Tb

sin(2Wt)
 p(t)   sinc(2Wt)
2Wt
Infeasibility of Ideal Nyquist Channel
 Rectangular P(f) is infeasible because:
 p(t) extends to negative infinity, which means that each ak
has already been transmitted at t = – ∞!
 A system response being flat from –W to W, and zero
elsewhere is physically unrealizable.
 The error margin is quite small, as a slight (erroneous)
shift in sampling time (such as, iTb+), will cause a very
large ISI.
 Note that p(t) decays to zero at a very slow rate of 1/|t|.
Infeasibility of Ideal Nyquist Channel
 Examination of timing error margin
 Let t be the sampling time difference between
transmitter and receiver.

y(iTb  t)  a
k 
k p((i  k)Tb  t)

 For simplicity, set i = 0.


y(t)  a
k 
k p(t  kTb )

sin[2W (t  kT)]
 a k b

k 2W (t  kTb )

 
sin[2Wt  k ]
y(t)   ak 2Wt  k
(1) k sin[2Wt]
k

a
 
k
k
2Wt  k
sin[2Wt] sin[2Wt]  (1) a k
k

 a0
2Wt

 
k 2Wt  k
k 0


(1)k a k
There exists{ak }such that    for any fixed small t  0.
k 2Wt  k
k 0

Question: How to make p(t) decays faster?

Answer: Make P(f) smoother.
Raised Cosine Spectrum

For a nonnegative function p(t),

  k P( f )
if  t k p(t)dt  , then exists.
 f k





Raised Cosine Spectrum

 We extend the bandwidth of p(t) from W to 2W, and require

that 1
P( f )  P( f  2W )  P( f  2W ) for | f | W .
2W
 So, the price to pay is a larger bandwidth.
 One of the P(f) that satisfies the above condition is the
raised cosine spectrum.
1
, 0 | f | (1  )W
2W
1   (| f | (1  )W )  
P( f )   1  cos  , (1  )W | f | (1  )W
4W  2W 
0, | f | (1  )W
Raised Cosine Spectrum
 The transmission bandwidth of the raised cosine spectrum
is equal to:
BT  2W (1  )
where  is the rolloff factor, which is the excess bandwidth
over the ideal solution.

 The text puts BT = W(1+)

 from the baseband viewpoint!


 cos(2Wt) 
p(t)  sinc(2Wt) 22 
 116  2
W t 
1
~ 3 as | t | large
| t|
Raised Cosine Spectrum

 cos(2Wt) 
 p(t)  sinc(2Wt) 2 2  consists of two terms:
 1 16 W t 
2

 The first term ensures the desired zero crossing of p(t).

 The second term provides the necessary tail
convergence rate of p(t).

 The special case of  = 1 is known as the full-cosine rolloff

characteristic.
sinc(4Wt)
p(t) 
116W 2 t 2
Raised Cosine Spectrum
 Useful property of full-cosinespectrum.
1, i  0
 iTb   1
p    , i  1
 2  2
0, i  2
 We have more “zero-crossing” at 3Tb/2, 5Tb/2,7Tb/2,…
in addition to the desired Tb, 2Tb, 3Tb…
 This is useful in synchronization. (Think of when
“synchronized,” the quantity should be small both at 3Tb/2,
5Tb/2, 7Tb/2,… and at Tb, 2Tb,3Tb…)
 However, the price to pay for this excessive synchronization
information is to “double thebandwidth.”
Correlative-Level Coding

 ISI, when generated in an uncontrolled manner, is an

undesirable phenomenon.
 However, ISI may become a friend if it is added to the
transmitted signal in a controlled manner.
 Known fact: A signal of bandwidth W can be
distortionlessly transmitted using its samples with
sampling rate  2W.
 Conversely, in a channel with bandwidth W Hz, the
theoretical maximum signal rate is 2W symbols per
second.
Correlative-Level Coding
A channel with bandwidth W Hz
B B
W W

W W
The maximum signal rate is
2W samples per second.
W W
Correlative-Level Coding
 Why intentionally adding ISI? Answer: To have better
bandwidth efficiency.
 Ideal Nyquist pulse shaping is efficient; it cannot be
realized.
 Raised consine pulse shaping is realizable; it is
bandwidth inefficient.
 By adding ISI to the transmitted symbols in a controlled
manner, we can achieve the Nyqusit rate 2W in a
channel bandwidth of W Hertz.
 Correlative-level coding or Partial-response
signaling
One Example of Correlative-Level Coding
 Duobinary signaling (or class I partial response)

HduoB ( f )

where {bk } i.i.d.

Duobinary Signaling
 Let us ignore the effect of HNyquist(f) first in the block
diagram in the previous slide. We directly obtain:

 H DuoB ( f )  1  exp( j2fTb )

 Note that ck has three levels (–2,0,2).
 The transfer function of the overall system is thus:
Duobinary Signaling

 HNyqusit(f):
 Give that

 As shown in the next slide, the response HI(f) is

realizable.
Duobinary Signaling

 HI(f)
| |
Duobinary Signaling

 hI(t):

Text omits this term

  hNyquist (t)  hNyquist (t  Tb ) by saying “exceptfor
a scalingfactor.”
Duobinary Signaling

 hI(t):
Duobinary Signaling
 Bandwidth efficiency of duobinary signaling
 Example.

  
The transmitted signal  akg(t  kTb)    ak (t  kTb )  * g(t)
k k  


 a  (t  kT )
k 
k b
g(t)

The input to this filter may not be WSS!

Then, we should use the time-average autocorrelation function.
Duobinary Signaling

 
X (t)   a  (t  kT )
k b Y (t)   a g(t  kT )
k 
k b
k 
g(t) (to channel)
Duobinary Signaling

2T
Approximately of them
Tb

–T 0 T
Duobinary Signaling


X (t)   a  (t  kT )
k b Y (t)
k
g(t) hDuoB (t) (to channel)

1
 SY ( f )  | G( f ) |2 | H DuoB ( f ) |2
Tb
Duobinary Signaling
H DouB ( f )  2 cos(fTb ) exp( jfTb )

1, 0  t  Tb
Assume g(t)   . | G f |2  Tb2 sinc 2  fTb 
0, otherwise
 sinc2
( fTb ), No Signal ISI
Y
S ( f )

SY (0) cos2 (fTb)sinc2 ( fTb), With SignalISI
 sinc2 ( fTb ), No Signal ISI

 sinc2
(2 fTb ), With SignalISI
Duobinary Signaling

1 .2

0 .8
sinc2 ( fTb ), without SignalISI
0 .6

0 .4 sinc2 (2 fTb ), with Signal ISI

0 .2

0
-2 -1 .5 -1 - 0.5 0 0.5 1 1.5 2
fTb
Duobinary Signaling
 Conclusions
 By adding ISI to the transmitted signal in a controlled (and
reversible) manner, we can reduce the requirement of bandwidth of
the transmitted signal.
 Hence, in the previous example, {ck} can be transmitted in every
Tb/2 seconds!
 Doubling the transmission capacity without introducing
additional requirement in bandwidth!
 Duobinary signaling : “Duo” means “doubling the transmission
capacity of a straight binary system.”
 A larger SNR is required to yield the same error rate because of an
increase in the number of signal levels (from –1, +1 to –2, 0, 2).
Detailed discussion on error rate impact is omitted here!
Duobinary Signaling
 Conclusions (cont.)
 The duobinary signaling is also named class I partial
response.
 Full response: The transmission wave at each time
instance is fully determined by a single information
symbol.
 Partial response: The transmission wave at each time
instance is only partially determined by one
information.
Decision Feedback for
Correlative-Level Coding
 Recovering of {ak} from {ck}

aˆk  ck  aˆk1 HduoB ( f )

 It requires the previous decision to determine the current

symbol.
 So, the system should feedback the previous decision.
 Error, therefore, may propagate!

 How to avoid error propagation? Answer: Precoding.

Precoding of Correlative Coding

Without precoding
{bk {0,1} i.i.d.} ak  2bk 1  ck  ak  ak 1

With precoding
~ ~  a  2 ~ 1 
{bk {0,1}i.i.d.}  bk  bk  bk1 k bk ck  ak  ak 1
ck  a k a k 1 ~
bk
~
bk1 bk ck
~ ~
 (2bk 1)  (2bk1 1) 0 0 0 –2
~ ~ 0 1 1 0
 2bk  2bk 1  2 1 0 1 0
~ ~
bk  bk  bk1 1 1 0 2
Precoding of Correlative Coding

 Final notes
 The precode must not change the “duo- of the
transmission capacity of a straight binary system.”
 Hence, {b! } must have the same distribution as {b }
k k

and hence must be i.i.d.

 Precoding of Correlative Coding
~
 Uniform i.i.d. of {bk}
~ ~ ~
   ~
 It suffices to show Pr bk | bk1 ,bk 2 ,...  Pr bk
~ ~ ~ ~ ~  
~ ~
bk  bk  bk1  Pr bk | bk1,bk2 ,...  Pr bk | bk 1 
~ ~
 Pr(bk  0 | bk1  0)  Pr(bk  0)  1/ 2
 ~ ~
Pr(bk  0 | bk1  1)  Pr(bk  1)  1/ 2
 ~
Pr(bk
~
 1| bk1  0)  Pr(bk  1)  1/ 2
~ ~ ~
  ~
Pr bk | bk 1 ,bk2 ,...  Pr bk
 ~ ~
 1| bk1  1)  Pr(bk  0)  1/ 2
Pr(bk
 For uniformity,
 Pr(b~k  0)  Pr(b ~ ~ ~
k1  0) Pr(bk  0 | bk1  0)
 ~ ~ ~
  Pr(bk1  1)Pr(bk  0 | bk1  1)

 ~ 1 ~ 1
 Pr(bk 1  0)  Pr(bk 1 1)
 2 2
 1
 
2 ~
 ~ ~ ~
 Pr(b k  1)  Pr(b k1  0) Pr(b k  1| bk1  0)
  Pr(b~  1)Pr(b~  1| b~  1)
 k 1 k k 1
 ~ 1 ~ 1
  Pr(bk1  0)  Pr(bk1  1)
2 2

 
1
 2 Q.E.D.
Example : Duobinary Coding with Precoding
 Table 4.1 in text

{b~k}       
{ }
bk        
{ak}    – –  – –
{ck}    –   –
{bˆ}       
k

~ ~
 b  ~
{bk {0,1}i.i.d.} bk k bk1  a k
 2bk 1  ck  ak  ak 1
Modified Duobinary Signaling
 The PSD of the signal is nonzero at the origin.
 This is considered to be an undesirable feature in some
applications, since many communication channels cannot
transmit a DC component.
 Solution: Class IV partial response or modified duobinary
technique. 1 .2

0 .8

0 .6

0 .4

0 .2

0
-2 -1 .5 -1 -0 .5 0 0 .5 1 1 .5 2
Modified Duobinary Signaling
~ ~ ~
{bk {0,1}i.i.d.}  bk  bk  bk2  a k  2bk 1  ck  ak  ak 2

H MDuoB ( f )

 H MDuoB ( f )  1  exp( j4fTb )

Modified Duobinary Signaling

 H MDuoB ( f )  1  exp( j4fTb )

 [exp( j2fTb )  exp( j2fTb )]exp( j2fTb )
 2 j sin(2fTb ) exp( j2fTb )

1, 0  t  Tb
Assume g(t)   . | G f |2  Tb2 sinc 2  fTb 
0, otherwise
SY ( f ) / Tb  sinc2 (2 fTb ), Duobinary (See Slide 4-59)

 Y
S ( f ) /(4T b )  sin 2
(2fTb )sinc 2
( fTb), Modified Duobinary
Modified Duobinary Signaling

1 .2
sinc2 (2 fTb ), Duobinary
1

0 .8 sin 2 (2fTb )sinc 2 ( fTb ),

0 .6 Modified duobinary

0 .4

0 .2

0 -1 - 0 .5 0 0.5 1 1.5 2
-2 -1 .5
fTb
Modified Duobinary Signaling

 Precoding is added to eliminate error propagation in

decision system.
ck  a k a k 2 ~
bk
~
bk2 bk ck
~ ~
 (2bk 1)  (2bk2 1) 0 0 0 
~ ~ 0 1 1 –2
 2bk  2bk2 1 0 1 2
~ ~
bk  bk  bk2 1 1 0 0

~ ~ ~
 b   a  2 1 
{bk {0,1}i.i.d.} bk k bk2 k bk ck  ak  ak 2
Generalized Form
of Correlative Level
Coding (CLC) or
Partial Response
Signaling

H CLC ( f )  w0  w1 z 1  !  w N 1 z N 1,
where z  exp( j2fTb ).
 2
III
1 .5

1
IV I
0 .5 V II

0
-1 .5 -1 -0 .5 0 0 .5 1 1 .5
Baseband M-ary PAM Transmission

Gray code
Any dibit differs from
an adjacent dibit in a
single bit position.
Baseband M-ary PAM Transmission
 For M-ary PAM transmission, there are M possible symbols
with symbol duration T.
 1/T is referred to as the signaling rate or symbol rate or
symbols per second or baud. Baud = the number of times a
 Some equivalences signal changes state persecond

 Each symbol can be equivalently identified with log2M

bits.
 The baud rate 1/T can be equivalently transformed to
bps as:
T  Tb log2 (M )
Baseband M-ary PAM Transmission
 Equivalences
 Virtually fix the symbol error, i.e., fix the level distance
(to be 2). For example, (, –) for M = 2, and (,
, –, –) for M = 4. Then, the transmitted power per
unit time for M-ary PAM transmission becomes:

2
1
E[S ]  M [(M  1)]2
 [(M  3)]2
 !  (M  3)2
 (M  1)2

T Tb log2 (M )

 (M 2  1)   1  (M 2  1)

3Tblog (M)
2  Tb  3log2 (M )
Baseband M-ary PAM Transmission

E[S 2 ]  1  (M 2  1)
 
T  Tb  3log2 (M )

For fixed Rb = 1/Tb (bps) and level distance = 2, the

transmitted power of an M-ary PAM transmission signal is
increased by a factor M2/log2M.