K Vasudevan Faculty of EE IIT Kanpur ([email protected].

in) 1

EE320A
Solutions for Tutorial 2
Date: 18th Sept. 2020

1. Solution: We start with the Fourier transform pair:

1 1
exp (−|t|) ⇀
↽ . (1)
2 1 + (2πf )2
Time scaling by 2π we get
1 1
exp (−|2πt|) ⇀
↽ . (2)
2 (2π)(1 + f 2 )
Multiplying both sides by 2π and applying duality we get
1
π exp (−|2πf |) ⇀
↽ = g(t) (say). (3)
1 + t2
Now the Hilbert transform of g(t) in the above equation is best obtained
in the frequency domain. Thus
Z ∞
ĝ(t) = −j sgn (f )π exp (−|2πf |) exp (j 2πf t) df
f =−∞
Z ∞
= −j π exp (−2πf (1 − j t)) df
f =0
Z 0
+ jπ exp (2πf (1 + j t)) df
f =−∞
−1 1
= −j π + jπ
−2π(1 − j t) 2π(1 + j t)
t
= . (4)
1 + t2
Thus proved.
2. Solution: We know that
⇀ 2
Rg (τ ) ↽ |G(f )|
dRg (τ ) 2
⇒ ⇀
↽ j 2πf |G(f )| . (5)
dτ
Let
dRg (τ )
= g1 (τ ). (6)
dτ
Let G1 (f ) denote the Fourier transform of g1 (t). Thus
2
G1 (f ) = j 2πf |G(f )| . (7)
K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 2

By Rayleigh’s energy theorem we have

Z ∞ Z ∞
2 2
|g1 (τ )| dτ = |G1 (f )| df. (8)
τ =−∞ f =−∞

Substituting for g1 (t) and G1 (f ) in the above equation we get the required
result.
3. Solution: Since gp (t) is periodic with a period T0 , the autocorrelation of
gp (t) is also periodic with period T0 . We also have
n m
T0 = =
f1 f2
n−m m+n 2n 2m
⇒ T0 = = = = (9)
f1 − f2 f1 + f2 2f1 2f2
where m and n are integers. Thus, frequencies f1 − f2 , f1 + f2 , 2f1 and
2f2 are also periodic with period T0 . Now, the autocorrelation of gp (t) is
equal to (note that gp (t) is real-valued):
T0
1
Z
∆
Rgp (τ ) = gp (t)gp (t − τ ) dt
T0 t=0
A21 A22
= A20 + cos(2πf1 τ ) + cos(2πf2 τ ) (10)
2 2
where we have made use of the fact that sinusoids integrated over one
period is equal to zero. In computing the autocorrelation, we have lost
the phase information.
4. Solution: We know that
2 2
e−πt ⇀
↽ e−πf
1 −πt2 1 −πf 2
⇒ e ⇀
↽ e . (11)
t0 t0
Using time scaling with a = 1/t0 we get
1 −πt2 /t20 2 2
g(t) = e ↽ e−πf t0 = G(f ).
⇀ (12)
t0
Since G(f ) is real-valued, the Fourier transform of the autocorrelation of
g(t) is simply G2 (f ). Thus
2 2
G2 (f ) = e−2πf t0
= G1 (f ) (say). (13)

Once again we use the time scaling property of the Fourier transform
namely

|a|g(at) ⇀
↽ G(f /a) (14)
K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 3

√
with 1/a = 2 in (12) to get
1 2 2
g1 (t) = √ e−πt /(2t0 ) (15)
t0 2
which is the required autocorrelation of the Gaussian pulse.
5. Solution: We know that

x̂(t) ⇀
↽ −j sgn (f )X(f ). (16)

Therefore

X1 (f ) = (3/4)X(f ) + (1/4)sgn (f )X(f )


 X(f ) for f > 0
= (3/4)A for f = 0 (17)
(1/2)X(f ) for f < 0


The plot of X1 (f ) is given in Figure 1.

X1 (f )

A
3A/4
A/2

−W 0 W

Figure 1: X1 (f ).

To solve part (b) let

m(t) = (3/4)x(t) + (3/4)j x̂(t) (18)

Then

M (f ) = (3/4)X(f ) + (3/4)sgn (f )X(f )


 (3/2)X(f ) for f > 0
= (3/4)A for f = 0 (19)
0 for f < 0


Then X2 (f ) = M (f − f0 ) and is given by


 (3/2)X(f − f0 ) for f > f0
X2 (f ) = (3/4)A for f = f0 (20)
0 for f < f0


The plot of X2 (f ) is shown in Figure 2.

K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 4

X2 (f )

3A/2

3A/4

f0 f0 + W

Figure 2: X2 (f ).

To solve part (c) let

m(t) = (3/4)x(t) + (1/4)j x̂(t). (21)

Then

M (f ) = (3/4)X(f ) + (1/4)sgn(f ) X(f )


 X(f ) for f > 0
= (3/4)A for f =0 (22)
(1/2)X(f ) for f < 0


Thus X3 (f ) = M (f − W ) and is given by

X3 (f )

A
3A/4
A/2

0 W 2W

Figure 3: X3 (f ).


 X(f − W ) for f > W
X3 (f ) = (3/4)A for f = W (23)
(1/2)X(f − W ) for f < W


The plot of X3 (f ) is given in Figure 3.

6. Solution: For a single stage, the input and output voltages are related as
follows:

v1 (t) = i(t)R + v2 (t). (24)

However,
dv2 (t)
i(t) = C . (25)
dt
K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 5

Therefore (24) can be rewritten as:

dv2 (t)
v1 (t) = RC + v2 (t). (26)
dt
Taking the Fourier transform of both sides we get:

V1 (f ) = (j 2πf RC + 1)V2 (f )
V2 (f ) 1
⇒ = . (27)
V1 (f ) 1 + j 2πf RC

The magnitude response of the overall cascade connection is

 
 Vo (f )  1
 Vi (f )  = (1 + 4π 2 f 2 τ 2 )N/2 (28)
 
0

where τ0 = RC. Note that

 
 Vo (f ) 
lim  =1 (29)
f →0 Vi (f ) 

and for large values of N

 
 Vo (f ) 
 Vi (f )  → 0
  for 4π 2 f 2 τ02 > 1. (30)

Therefore, we are interested in the magnitude response in the vicinity of

f = 0, for large values of N .
We make use of the MacLaurin series expansion of a function f (x) about
x = 0 as follows (assuming that all the derivatives exist):

f (1) (0) f (2) (0) 2

f (x) = f (0) + x+ x + ... (31)
1! 2!
where
dn f (x) 

f (n) (0) = . (32)
dxn x=0

In the present context

1
f (x) = (33)
(1 + 4π 2 τ02 x)N/2

with x = f 2 . We have

f (0) = 1
4π 2 τ02

−N
f (1) (0) =


2 2
2 (1 + 4π τ0 x) N/2+1 
x=0
K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 6

−N
= (4π 2 τ02 )
2

2 
4π 2 τ02
 
N N 
f (2) (0) = +1

2 2 (1 + 4π 2 τ02 x)N/2+2 

x=0
 2
N 2 2
≈ 4π τ0 (34)
2
where we have assumed that for large N
N N
+1≈ . (35)
2 2
Generalizing (34), we can obtain the nth derivative of the MacLaurin series
as:
   
(n) nN N N
f (0) = (−1) + 1 ... +n−1
2 2 2

n 
4π 2 τ02 
×

2
(1 + 4π 2 τ0 x)N/2+n 

x=0
 n
n N 2 2
≈ (−1) 4π τ0 (36)
2
where we have used the fact that for any finite n ≪ N
N N
+n−1≈ . (37)
2 2
Thus f (x) in (33) can be written as:

y y2
f (x) = 1− x + x2 − . . .
1! 2!
= exp (−xy) . (38)

where

y = (N/2)4π 2 τ02 = (N/2)4π 2 R2 C 2 . (39)

Finally, substituting for x we get the desired result which is Gaussian.

Observe that, in order to satisfy the condition n ≪ N , the nth term of the
series in (38) must tend to zero for n ≪ N . This can happen only when

xy < 1
2 2 2 2
⇒ (N/2)4π R C f < 1
1
⇒ |f | < √ (40)
πRC 2N
which is in the vicinity of f = 0 for large N .
K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 7

EE320A
Solutions for Tutorial 3
Date: 18th Sept. 2020

g1 (t) g2 (t)
2
2
t t

−3 0 3
−2

−3 −1 0 1 3

Figure 4: Pulses g1 (t) and g2 (t).

1. Solution: We know that

R12 (t) = g1 (t) ⋆ g2∗ (−t) ⇀
↽ G1 (f )G∗2 (f )
(m+n) ⇀ m n
⇒ R12 (t) ↽ (j 2πf ) G1 (f ) (j 2πf ) G∗2 (f )
(m+n) ⇀ m n n
⇒ R12 (t) ↽ (j 2πf ) G1 (f ) (−1) (−j 2πf ) G∗2 (f ). (41)
Let
(n) ⇀ n
h2 (t) = g2 (t) ↽ (j 2πf ) G2 (f )
n
⇒ h∗2 (−t) ⇀
↽ (−j 2πf ) G∗2 (f ). (42)
Similarly
m
h1 (t) = g (m) (t) ⇀
↽ (j 2πf ) G(f ). (43)
Using the fact that multiplication in the frequency-domain is equivalent
to convolution in the time domain we get:
(m+n) n
R12 (t) = (−1) (h1 (t) ⋆ h∗2 (−t))
Z ∞
(m+n) n
⇒ R12 (τ ) = (−1) h1 (t)h∗2 (t − τ ) dt. (44)
t=−∞

Hence proved. For the signal in Figure 4, using m = 1 and n = 0 we get

h1 (t) = 2δ(t + 3) − 2δ(t − 3)
h2 (t) = g2 (t). (45)
Thus
Z ∞
(1)
R12 (τ ) = h1 (t)h2 (t − τ ) dt
t=−∞
= 2g2 (−τ − 3) − 2g2 (3 − τ ). (46)
(1)
R12 (τ ) and R12 (τ ) are plotted in Figure 5.
K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 8

g2 (t)

2
t

0
−2

(1)
R12 (τ )

4
τ

−4

R12 (τ )

−8

−6 −4 −2 0 2 4 6

−3 −1 1 3

Figure 5: Cross correlation R12 (τ ).

K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 9

2. Solution: We start from the basic definition of the cross-correlation func-

tion:
Z ∞
Rg1 g2 (τ ) = g1 (t)g2∗ (t − τ ) dt
t=−∞
= g1 (τ ) ⋆ g2∗ (−τ )
Z ∞
= G1 (f )G∗2 (f ) exp (j 2πf τ ) df. (47)
f =−∞

Here

g1 (t) = g(t) ⇀↽ G(f )

g2 (t) = ĝ(t) ⇀
↽ −j sgn (f )G(f )
⇒ g2∗ (−t) ∗ ⇀ j sgn (f )G∗ (f ).
= ĝ (−t) = ĝ(−t) ↽ (48)

since g(t) and hence ĝ(t) are real-valued. Hence

Z ∞
Rgĝ (τ ) = j sgn (f )G(f )G∗ (f ) exp (j 2πf τ ) df
f =−∞
Z ∞
2
= j sgn (f ) |G(f )| exp (j 2πf τ ) df. (49)
f =−∞

However we know that

⇀ 2
Rg (τ ) ↽ |G(f )|
⇀ 2
⇒ R̂g (τ ) ↽ −j sgn (f ) |G(f )| . (50)

From (49) and (50) it is clear that

Rgĝ (τ ) = −R̂g (τ ). (51)

The second part can be similarly proved.

3. Solution: We need to prove the following:
Z ∞ Z ∞ 
1 ∗
ℜ {g1+ (t)} ℜ {g2+ (t)} dt = ℜ g1+ (t)g2+ (t) dt (52)
t=−∞ 2 t=−∞

and
∞ ∞
1
Z Z
2 2
g (t) dt = |g̃(t)| dt. (53)
t=−∞ 2 t=−∞

From the basic definition of the pre-envelope we have:

g1+ (t) = g1 (t) + j ĝ1 (t)

g2+ (t) = g2 (t) + j ĝ2 (t) (54)
K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 10

where g1 (t) and g2 (t) are real-valued signals. Hence ĝ1 (t) and ĝ2 (t) are
also real-valued. Next, we note that the real-part of the integral is equal
to the integral of the real-part. Hence the right-hand-side of (52) becomes:
1 ∞
Z
(g1 (t)g2 (t) + ĝ1 (t)ĝ2 (t)) dt. (55)
2 t=−∞
Next, we note that
Z ∞ Z ∞
g1 (t)g2 (t) dt = g1 (t) ⋆ g2 (−t)|t=0 = G1 (f )G∗2 (f ) df. (56)
t=−∞ f =−∞

where “⋆” denotes convolution. Similarly

Z ∞ Z ∞
ĝ1 (t)ĝ2 (t) dt = ĝ1 (t) ⋆ ĝ2 (−t)|t=0 = (−j sgn (f )) G1 (f )
t=−∞ f =−∞
(j sgn (f )) G∗2 (f ) df
Z ∞
= G1 (f )G∗2 (f ) df
f =−∞
Z ∞
= g1 (t)g2 (t) dt (57)
t=−∞

where we have used the fact that

sgn2 (f ) = 1 for f 6= 0. (58)

Now, the left-hand-side of (52) is nothing but

Z ∞
g1 (t)g2 (t) dt. (59)
t=−∞

Thus (52) is proved. Now, let us see what happens when g2 (t) is replaced
by g2 (−t). Let

g3 (t) = g2 (−t)
⇒ g3+ (t) = g3 (t) + j ĝ3 (t)
= g2 (−t) + j ĝ2 (−t)
∗
⇒ g3+ (t) = g3 (t) − j ĝ3 (t)
= g2 (−t) − j ĝ2 (−t). (60)

Clearly (52) is still valid with g2+ (t) replaced by g3+ (t) as given in (60),
that is
Z ∞ Z ∞ 
1 ∗
ℜ {g1+ (t)} ℜ {g3+ (t)} dt = ℜ g1+ (t)g3+ (t) dt . (61)
t=−∞ 2 t=−∞

To prove part (b) we substitute

g1 (t) = g2 (t) = g(t) (62)

K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 11

in (52) and note that

g+ (t) = g(t) + j ĝ(t)

= g̃(t) exp (j 2πfc t) (63)

where g̃(t) is the complex envelope of g(t). Thus (53) follows immediately.
4. Solution: Let H(f ) denote the Fourier transform of h(t). We have
Z ∞
H(f ) = h(t)e−j 2πf t dt
t=−∞
Z ∞ 
−j 2πf t
 
⇒ |H(f )| =   h(t)e dt
t=−∞
Z ∞
h(t)e−j 2πf t dt
 
≤
t=−∞
Z ∞
⇒ |H(f )| ≤ |h(t)| dt (64)
t=−∞

where we have used the Schwarz’s inequality. Since h(t) is stable

Z ∞
|h(t)| dt < ∞
t=−∞
⇒ |H(f )| < ∞. (65)

Let

A = max |H(f )|. (66)

The energy of the output signal y(t) is

Z ∞ Z ∞
|y(t)|2 dt = |Y (f )|2 df (67)
t=−∞ f =−∞

where we have used the Rayleigh’s energy theorem. Using the fact that

Y (f ) = H(f )X(f ) (68)

the energy of the output signal is

Z ∞ Z ∞
|H(f )|2 |X(f )|2 df ≤ A2 |X(f )|2 df
f =−∞ f =−∞
Z ∞ Z ∞
⇒ |y(t)|2 dt ≤ A 2
|x(t)|2 dt. (69)
t=−∞ f =−∞

Since the input signal has finite energy, so does the output signal.
K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 12

5. Solution: To prove
Trms Wrms ≥ 1/(4π) (70)
we need to use the Schwarz’s inequality which states that:
Z ∞ 2 Z ∞
∗ ∗
(g1 (t)g2 (t) + g1 (t)g2 (t)) dt ≤ 4 |g1 (t)|2 dt
t=−∞ t=−∞
Z ∞
× |g2 (t)|2 dt.
t=−∞
(71)
For the given problem we substitute
g1 (t) =
tg(t)
dg(t)
g2 (t) = . (72)
dt
Thus, the right-hand-side of the Schwarz’s inequality in (71) becomes
 dg(t) 2
Z ∞ Z ∞  
2 2
4 t |g(t)| dt  dt  dt.
  (73)
t=−∞ t=−∞
Let
dg(t)
g3 (t) = ⇀
↽ j 2πf G(f ) = G3 (f ). (74)
dt
Then, by Rayleigh’s energy theorem
Z ∞ Z ∞
|g3 (t)|2 dt = |G3 (f )|2 df
t=−∞ f =−∞
 dg(t) 2
Z ∞   Z ∞
2 2
⇒  dt  dt
  = 4π f 2 |G(f )| df. (75)
t=−∞ f =−∞

Thus, the right-hand-side of the Schwarz’s inequality in (71) becomes:

Z ∞ Z ∞
2 2
16π 2 t2 |g(t)| dt f 2 |G(f )| df. (76)
t=−∞ f =−∞

Let us now consider the square-root of the left-hand-side of the Schwarz’s

inequality in (71). We have:
Z ∞   ∗  Z ∞ 
dg(t) dg(t) dg(t)
t g ∗ (t) + g(t) dt = t g ∗ (t)
t=−∞ dt dt t=−∞ dt

dg (t)
∗
+ g(t) dt
dt
Z ∞
d
= t (g(t)g ∗ (t)) dt
t=−∞ dt
Z ∞
d  2

= t |g(t)| dt.
t=−∞ dt
(77)
K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 13

In the above equation we have used the fact (this can also be proved using
Fourier transforms) that:
 ∗
dg(t) dg ∗ (t)
= . (78)
dt dt

Let
2
|g(t)| ⇀
g4 (t) =
↽ G4 (f )
dg4 (t)
⇒ g5 (t) = ⇀
↽ j 2πf G4 (f ) = G5 (f )
dt
j dG5 (f )
⇒ tg5 (t) ↽ ⇀
2π df
Z ∞ 
j dG5 (f ) 
⇒ tg5 (t) dt =
t=−∞ 2π df f =0
Z ∞ 
j d 
⇒ tg5 (t) dt = (j 2πf G4 (f ))
t=−∞ 2π df f =0
∞
d
Z
2
⇒ t
|g(t)| dt = −G4 (0)
t=−∞ dt
Z ∞ Z ∞
d 2 2
⇒ t |g(t)| dt = − |g(t)| dt (79)
t=−∞ dt t=−∞

Using Rayleigh’s energy theorem, the left-hand-side of the Schwarz’s in-

equality in (71) becomes:
Z ∞ Z ∞
2 2
|g(t)| dt |G(f )| df (80)
t=−∞ f =−∞

Taking the square-root of (76) and (80) we get the desired result in (70).
To prove

Teq Weq ≥ 0.5 (81)

we invoke the Schwarz’s inequality which states that

Z ∞  Z ∞
 

 g(t) dt  ≤
 |g(t)| dt. (82)
t=−∞ t=−∞

Squaring both sides we obtain

Z ∞ 2 Z ∞ 2
 

 g(t) dt ≤ |g(t)| dt
t=−∞ t=−∞
Z ∞ 2
2
⇒ |G(0)| ≤ |g(t)| dt . (83)
t=−∞
K Vasudevan Faculty of EE IIT Kanpur ([email protected]) 14

Using the above inequality and the Rayleigh’s energy theorem which states
that
Z ∞ Z ∞
|g(t)|2 dt = |G(f )|2 df (84)
t=−∞ f =−∞

we get the desired result

Teq Weq ≥ 0.5. (85)