.

0
Signals and Systems-23EC02
UNIT-IV
Correlation and Sampling Theorem
B.Tech., III-Sem., ECE

1. Auto Correlation Functions (ACF) and Properties

2. Energy Spectral Density (ESD)

3. Power Spectral Density (ESD)

4. Parseval’s Theorem

5. Cross Correlation Functions (CCF) and Properties

6. Relation between Convolution and Correlation

7. Detection of periodic signals in the presence of noise by correlation

8. Extraction of signal from noise by filtering

9. Sampling Theorem
➢ Graphical Analysis of Sampling Theorem-Impulse Sampling
➢ Analytical Analysis of Sampling Theorem
➢ Sampling Theorem for Band Limited Signal-Band Pass Sampling
➢ Reconstruction of Signal from its Samples
➢ Statement of Sampling Theorem
➢ Effect of under Sampling-Aliasing Effect
➢ Natural and Flat Top Sampling

10.Solved Problems

11.Assignment Questions

12.Quiz Questions

----------------------------------------------------------------------------------------------------------------------------- -----------------
Dr.B.Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page 1/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

1. Auto Correlation Function (ACF) and Properties:

The measure of similarity between a signal x(t) and time delayed form of the same signal, i.e x(t+τ)
is called Auto Correlation Function (ACF). It is represented with Rxx(τ) and it can be computed
from the formula
∞ ∞
𝑅𝑥𝑥(𝜏) = ∫ 𝑥(𝑡)𝑥(𝑡 + 𝜏)𝑑𝑡 = ∫ 𝑥(𝑡)𝑥(𝑡 − 𝜏)𝑑𝑡 ; 𝑖𝑓 𝑥(𝑡) 𝑖𝑠 𝑎𝑝𝑒𝑟𝑖𝑜𝑑𝑖𝑐 𝑠𝑖𝑔𝑛𝑎𝑙
−∞ −∞

1 𝑇 1 𝑇
𝑅𝑥𝑥(𝜏) = lim ∫ 𝑥(𝑡)𝑥(𝑡 + 𝜏)𝑑𝑡 = lim ∫ 𝑥(𝑡)𝑥(𝑡 − 𝜏)𝑑𝑡 ; 𝑖𝑓 𝑥(𝑡) 𝑖𝑠 𝑝𝑒𝑟𝑖𝑜𝑑𝑖𝑐 𝑠𝑖𝑔𝑛𝑎𝑙
𝑇→∞ 2𝑇 −𝑇 𝑇→∞ 2𝑇 −𝑇

Properties:
1. Auto Correlation Function Rxx(τ) is even function in τ.
𝑅𝑥𝑥(−𝜏) = 𝑅𝑥𝑥(𝜏)
2. Total Energy under the signal x(t) can be computed from the Auto Correlation Function
Rxx(τ) by substituting τ=0.
𝐸 = 𝑅𝑥𝑥(0)
3. Maximum value of Auto Correlation Function Rxx(τ) occurs at origin.
𝑅𝑥𝑥(0) ≥ |𝑅𝑥𝑥(𝜏)|
4. For an Energy signal, Auto Correlation Function Rxx(τ) and Energy Spectral Density ψ xx(w)
form Fourier Transformable pair.
𝐹𝑇
→
𝑅𝑥𝑥(𝑡) Ψxx(w)
←
𝐼𝐹𝑇

Where,
∞
Ψxx(w) = 𝐹𝑇[𝑅𝑥𝑥(𝜏)] = ∫ 𝑅𝑥𝑥(𝜏)𝑒 −𝑗𝑤𝜏 𝑑𝜏
−∞

5. For a Power signal, Auto Correlation Function Rxx(τ) and Power Spectral Density Sxx(w)
form Fourier Transformable pairs.
𝐹𝑇
→
𝑅𝑥𝑥(𝑡) Sxx(w)
←
𝐼𝐹𝑇

Where,
∞
Sxx(w) = 𝐹𝑇[𝑅𝑥𝑥(𝜏)] = ∫ 𝑅𝑥𝑥(𝜏)𝑒 −𝑗𝑤𝜏 𝑑𝜏
−∞

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 2/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

2. Energy Spectral Density (ESD):

If x(t) is energy signal, and FT[x(t)] = X(w) with Auto Correlation Function Rxx(τ), then the
frequency domain of Rxx(τ) or the squared magnitude spectrum, i.e. │X(w)│2 is called Energy
Spectral Density (ESD) or Energy Density Spectrum (EDS). It is represented with ψxx(w) and it can
be computed from the formula;

∞
Ψxx(w) = 𝐹𝑇[𝑅𝑥𝑥(𝜏)] = ∫ 𝑅𝑥𝑥(𝜏)𝑒 −𝑗𝑤𝜏 𝑑𝜏 = | 𝑋(𝑤) |2
−∞

Proof:
We know that for an energy signal, Auto Correlation Function Rxx(τ) and Energy Spectral Density
ψxx(w) form Fourier Transformable pair. i.e.,

Ψxx(w) = 𝐹𝑇[𝑅𝑥𝑥(𝜏)]
∞
= ∫ 𝑅𝑥𝑥(𝜏)𝑒 −𝑗𝑤𝜏 𝑑𝜏
−∞
∞ ∞
= ∫ (∫ 𝑥(𝑡)𝑥(𝑡 + 𝜏)𝑑𝑡) 𝑒 −𝑗𝑤𝜏 𝑑𝜏
−∞ −∞
∞ ∞
= ∫ 𝑥(𝑡) (∫ 𝑥(𝜏 + 𝑡) 𝑒 −𝑗𝑤𝜏 𝑑𝜏) 𝑑𝑡
−∞ −∞
∞
= ∫ 𝑥(𝑡)(𝐹𝑇[𝑥(𝜏 + 𝑡)])𝑑𝑡 , 𝑎𝑝𝑝𝑙𝑦 𝑡𝑖𝑚𝑒 𝑠ℎ𝑖𝑓𝑡𝑖𝑛𝑔 𝑝𝑟𝑜𝑝𝑒𝑟𝑡𝑦
−∞
∞
= ∫ 𝑥(𝑡) 𝑒 𝑗𝑤𝑡 𝑋(𝑤)𝑑𝑡
−∞
∞
= 𝑋(𝑤) ∫ 𝑥(𝑡) 𝑒 𝑗𝑤𝑡 𝑑𝑡
−∞

= 𝑋(𝑤)𝑋 ∗ (𝑤)
= | 𝑋(𝑤) |2
= Ψxx(w)

∞
Ψxx(w) = 𝐹𝑇[𝑅𝑥𝑥(𝜏)] = ∫ 𝑅𝑥𝑥(𝜏)𝑒 −𝑗𝑤𝜏 𝑑𝜏 = | 𝑋(𝑤) |2
−∞

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 3/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

3. Power Spectral Density (PSD):

If x(t) is power signal with Auto Correlation Function Rxx(τ), then the frequency domain of Rxx(τ)
is called Power Spectral Density (PSD) or Power Density Spectrum (PDS). It is represented with
Sxx(w) and it can be computed from the formula;
∞
Sxx(w) = 𝐹𝑇[𝑅𝑥𝑥(𝜏)] = ∫ 𝑅𝑥𝑥(𝜏)𝑒 −𝑗𝑤𝜏 𝑑𝜏
−∞

4. Parseval’s Theorem:
If x(t) is energy signal and FT[x(t) ] = X(w), then the total energy under the signal x(t) can be
computed from x(t) as well as X(w ) by using the following relation is called Parseval’s theorem.
∞
1 ∞
𝐸 = ∫ | 𝑥(𝑡) |2 𝑑𝑡 = ∫ | 𝑋(𝑤) |2 𝑑𝑤
−∞ 2𝜋 −∞
Proof:
We know that the total Energy under the signal x(t) can be computed from the formula
∞
𝐸 = ∫ | 𝑥(𝑡) |2 𝑑𝑡
−∞
∞
1 ∞
= ∫ 𝑥(𝑡)𝑥 ∗ (𝑡)𝑑𝑡 ; 𝐼𝐹𝑇[𝑋(𝑤)] = 𝑥(𝑡) = ∫ 𝑋(𝑤)𝑒 𝑗𝑤𝑡 𝑑𝑤
−∞ 2𝜋 −∞
∞ ∗
1 ∞
= ∫ 𝑥(𝑡) ( ∫ 𝑋(𝑤)𝑒 𝑗𝑤𝑡 𝑑𝑤 ) 𝑑𝑡
−∞ 2𝜋 −∞
∞
1 ∞
= ∫ 𝑥(𝑡) ( ∫ 𝑋 ∗ (𝑤)𝑒 −𝑗𝑤𝑡 𝑑𝑤 ) 𝑑𝑡
−∞ 2𝜋 −∞
1 ∞ ∗ ∞
= ∫ 𝑋 (𝑤) (∫ 𝑥(𝑡)𝑒 −𝑗𝑤𝑡 𝑑𝑡 ) 𝑑𝑤
2𝜋 −∞ −∞

1 ∞ ∗
= ∫ 𝑋 (𝑤)𝐹𝑇[𝑥(𝑡)]𝑑𝑤
2𝜋 −∞
1 ∞ ∗
= ∫ 𝑋 (𝑤)𝑋(𝑤)𝑑𝑤
2𝜋 −∞
1 ∞
= ∫ | 𝑋(𝑤) |2 𝑑𝑤
2𝜋 −∞

∞
1 ∞
𝐸 = ∫ | 𝑥(𝑡) |2 𝑑𝑡 = ∫ | 𝑋(𝑤) |2 𝑑𝑤
−∞ 2𝜋 −∞

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 4/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

5. Cross Correlation Function (CCF) and Properties:

The measure of similarity between a signal x(t) and time delayed form of another signal y(t), i.e
y(t+τ) is called Cross Correlation Function (CCF). It is represented with Rxy(τ) and it can be
computed from the formula
∞
𝑅𝑥𝑦(𝜏) = ∫ 𝑥(𝑡)𝑦(𝑡 + 𝜏)𝑑𝑡 ; 𝑖𝑓 𝑥(𝑡) 𝑎𝑛𝑑 𝑦(𝑡) 𝑎𝑟𝑒 𝑎𝑝𝑒𝑟𝑖𝑜𝑑𝑖𝑐 𝑠𝑖𝑔𝑛𝑎𝑙
−∞

1 𝑇
𝑅𝑥𝑦(𝜏) = lim ∫ 𝑥(𝑡)𝑦(𝑡 + 𝜏)𝑑𝑡 ; 𝑖𝑓 𝑥(𝑡) 𝑎𝑛𝑑 𝑦(𝑡) 𝑎𝑟𝑒 𝑝𝑒𝑟𝑖𝑜𝑑𝑖𝑐 𝑠𝑖𝑔𝑛𝑎𝑙
𝑇→∞ 2𝑇 −𝑇

Properties:
1. If x(t) and y(t) are orthogonal signals, then the Cross Correlation Function Rxy(τ) is zero.
𝑅𝑥𝑦(𝜏) = 0
or
𝑅𝑦𝑥(𝜏) = 0

2. Relation between Cross Correlation Functions Rxy(τ) and Ryx(τ) is

𝑅𝑥𝑦(−𝜏) = 𝑅𝑦𝑥(𝜏)
or
𝑅𝑥𝑦(𝜏) = 𝑅𝑦𝑥(−𝜏)

3. Maximum value of the Cross Correlation Function Rxy(τ) occurs at the arithmetic mean of
Rxx(0) and Ryy(0).
𝑅𝑥𝑥(0) + 𝑅𝑦𝑦(0)
≥ |𝑅𝑥𝑦(𝜏)|
2

4. Maximum value of the Cross Correlation Function Rxy(τ) occurs at the geometric mean of
Rxx(0) and Ryy(0).
√𝑅𝑥𝑥(0)𝑅𝑦𝑦(0) ≥ |𝑅𝑥𝑦(𝜏)|
It is Schwartz inequality.

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 5/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

6. Relation between Convolution and Correlation:

➢ Convolution and Correlation are two similar and different operations, which is used in almost
all signal processing applications to analyze signals and systems in both the time and
frequency domains.
➢ Convolution operation includes four different operations, namely Folding, Shifting,
Multiplication and Integration in the case of continues time signals/ Summation in the case of
discrete time signals.
➢ Convolution in continuous time domain can be computed from the formula.
∞
𝑥(𝑡) ∗ 𝑦(𝑡) = ∫ 𝑥(𝜏)𝑦(𝑡 − 𝜏)𝑑𝜏
−∞

➢ The measure of similarity between a signal x(t) and time delayed form of the same signal, or
a different signal y(t) is called Correlation, which includes three different operations, namely
Shifting, Multiplication and Integration in the case of continues time signals/ Summation in
the case of discrete time signals.
➢ The measure of similarity between a signal x(t) and time delayed form of the same signal, i.e
x(t+τ) is called Auto Correlation Function (ACF). It is represented with Rxx(τ) and it can be
computed from the formula
∞
𝑅𝑥𝑥(𝜏) = ∫ 𝑥(𝑡)𝑥(𝑡 + 𝜏)𝑑𝑡
−∞

➢ The measure of similarity between a signal x(t) and time delayed form of another signal y(t),
i.e y(t+τ) is called Cross Correlation Function (CCF). It is represented with Rxy(τ) and it can
be computed from the formula
∞
𝑅𝑥𝑦(𝜏) = ∫ 𝑥(𝑡)𝑦(𝑡 + 𝜏)𝑑𝑡
−∞

➢ It is clear from the above analysis that the folding is one extra operation which included in
the convolution operation.
➢ If the second signal y(t) is even, i.e., y(-t)=y(t), then the convolution and correlation
operations are same.

𝑥(𝑡) ∗ 𝑦(𝑡) = 𝑅𝑥𝑦(𝜏), 𝑖𝑓 𝑦(𝑡) 𝑖𝑠 𝑒𝑣𝑒𝑛

and
𝑥(𝑡) ∗ 𝑦(𝑡) ≠ 𝑅𝑥𝑦(𝜏), 𝑖𝑓 𝑦(𝑡) 𝑖𝑠 𝑛𝑜𝑡 𝑒𝑣𝑒𝑛

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 6/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

7. Detection of Periodic Signals in the presence of Noise by Correlation:

➢ We know that for a distortion less communication system, the output wave shape is exact
replica of input wave shape.
➢ If the noise is added in the communication system, then output may be corrupted or distorted.
Now the task is to analyze the system to detect the signal by using correlation functions.
➢ Let a periodic signal x(t) is transmitted through a communication system and assume the
noise n(t) is added and produces an output of y(t).

x(t) Communication System y(t) = x(t) + n(t)

➢ Detection of a periodic signal x(t) in the presence of noise n(t) can be analyzed by using auto
correlation and cross correlation functions.
➢ We know that the noise n(t) is random in nature, hence the correlation between a periodic
signal x(t) and noise n(t) is zero. i.e. cross correlation between x(t) and n(t) is zero.
1 𝑇
𝑅𝑥𝑛(𝜏) = lim ∫ 𝑥(𝑡)𝑛(𝑡 − 𝜏)𝑑𝑡 = 0
𝑇→∞ 2𝑇 −𝑇

➢ Now evaluate the auto correlation function of response of a system, y(t) = x(t) + n(t).
1 𝑇
𝑅𝑦𝑦(𝜏) = lim ∫ 𝑦(𝑡)𝑦(𝑡 − 𝜏)𝑑𝑡
𝑇→∞ 2𝑇 −𝑇

1 𝑇
= lim ∫ [𝑥(𝑡) + 𝑛(𝑡)][𝑥(𝑡 − 𝜏) + 𝑛(𝑡 − 𝜏)]𝑑𝑡
𝑇→∞ 2𝑇 −𝑇

1 𝑇
= lim ∫ [𝑥(𝑡)𝑥(𝑡 − 𝜏) + 𝑥(𝑡)𝑛(𝑡 − 𝜏) + 𝑛(𝑡)𝑥(𝑡 − 𝜏) + 𝑛(𝑡)𝑛(𝑡 − 𝜏)]𝑑𝑡
𝑇→∞ 2𝑇 −𝑇

= 𝑅𝑥𝑥(𝜏) + 𝑅𝑥𝑛(𝜏) + 𝑅𝑛𝑥(𝜏) + 𝑅𝑛𝑛(𝜏); 𝑤ℎ𝑒𝑟𝑒, 𝑅𝑥𝑛(𝜏) = 𝑅𝑛𝑥(𝜏) = 0

= 𝑅𝑥𝑥(𝜏) + 𝑅𝑛𝑛(𝜏)
➢ The noise n(t) is random in nature, and its autocorrelation Rnn(τ) is negligible for higher
values of ‘τ’. i.e., Rnn(τ) = 0.

𝑅𝑦𝑦(𝜏) = 𝑅𝑥𝑥(𝜏) or y(t) = x(t)

➢ Where Rxx(τ) is periodic because x(t) is periodic and Ryy(τ) is also periodic because the
noise n(t) is random in nature, and its autocorrelation Rnn(τ) is negligible. therefore the
periodic signal x(t) is said to be detected in the presence of noise.

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 7/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

8. Extraction of Signal from Noise by Filtering:

➢ We know that the correlation functions are used to detect a signal in the presence of noise.

x(t) Communication System y(t) = x(t) + n(t)

Rxx(τ) Communication System Ryy(τ)

𝑤ℎ𝑤𝑟𝑒, 𝑅𝑦𝑦(𝜏) = 𝑅𝑥𝑥(𝜏) + 𝑅𝑛𝑛(𝜏)

➢ Similarly, we can extract the signal from noise by filtering. In general, filters are used to
extract required information and attenuate all other unwanted things. In communication, filter
can be defined as a frequency selective device for various frequencies of signals.

➢ Based on frequency response, filters are classified into four types.

✓ Low Pass Filters (LPF)
✓ High Pass Filters (HPF)
✓ Band Pass Filters (BPF)
✓ Band Stop Filters (BSF)
➢ Depends on type of application use the filter and apply FT[Ryy(τ)] = Syy(w) as input to the
filter and take the output, Sxx(w)=FT[ Rxx(τ) ].

Syy(w) Filter Sxx(w)

➢ The ratio between the filter output, Sxx(w) to filter input Syy(w) is called power transfer
function of the filter.
𝑆𝑥𝑥(𝑤)
|𝐻(𝑤)|2 = ⟹ 𝑆𝑥𝑥(𝑤) = |𝐻(𝑤)|2 𝑆𝑦𝑦(𝑤)
𝑆𝑦𝑦(𝑤)
➢ Computation of Sxx(w) by using above filtering process is called spectral analysis, in this
process all unwanted frequency components are attenuated and produces required signal. This
is called extraction of signal from noise by filtering.

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 8/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

9. Sampling Theorem:
The process of converting a given continuous time (analog) signal into a discrete time signal is called
sampling or sampling process or sampling theorem.
(A)Graphical Analysis of Sampling Theorem-Impulse Sampling:
Discrete time signal can be obtained from the continuous time or analog signal g(t) by multiplying
train of impulse signal, 𝑔𝛿 (𝑡) = 𝑔(𝑡)𝛿𝑇𝑠 (𝑡). It is called ideal sampling or impulse sampling.

g(t)

δTs(t)

…….-4Ts -3Ts -2Ts -Ts 0 Ts 2Ts 3Ts 4Ts .….. t

gδ(t
)

…….-4Ts -3Ts -2Ts -Ts 0 Ts 2Ts 3Ts 4Ts .….. t

Where,
g(t) : Given continuous time or analog signal
δTs(t) : Train of impulse with a time period Ts.
gδ(t) : Discrete time or sampled form of given g(t)
Ts : Sampling period or Sampling interval
∞

𝑔𝛿 (𝑡) = 𝑔(𝑡)𝛿𝑇𝑠 (𝑡), 𝛿𝑇𝑠 (𝑡) = ∑ 𝛿(𝑡 − 𝑛𝑇𝑠)

𝑛=−∞
∞

⇒ 𝑔𝛿 (𝑡) = 𝑔(𝑡) ∑ 𝛿(𝑡 − 𝑛𝑇𝑠)

𝑛=−∞

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 9/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(B)Analytical Analysis of Sampling Theorem:

From the graphical analysis of sampling theorem,
∞

𝑔𝛿 (𝑡) = 𝑔(𝑡) ∑ 𝛿(𝑡 − 𝑛𝑇𝑠)

𝑛=−∞

Apply Fourier Transform both sides,

∞
𝑋1 (𝑤) ∗ 𝑋2 (𝑤)
⇒ 𝐹𝑇[𝑔𝛿 (𝑡)] = 𝐹𝑇 [𝑔(𝑡) ∑ 𝛿(𝑡 − 𝑛𝑇𝑠)] ; 𝑢𝑠𝑒 𝐹𝑇[𝑥1 (𝑡)𝑥2 (𝑡)] =
2𝜋
𝑛=−∞
∞

⇒ 𝐺𝛿 (𝑓)] = 𝐹𝑇[𝑔(𝑡)] ∗ 𝐹𝑇 [ ∑ 𝛿(𝑡 − 𝑛𝑇𝑠)]

𝑛=−∞
∞
1 𝑛
= 𝐺(𝑓) ∗ ∑ 𝛿 (𝑓 − )
𝑇𝑠 𝑇𝑠
𝑛=−∞
∞
1 𝑛
= ∑ 𝐺(𝑓) ∗ 𝛿 (𝑓 − ) ; 𝑢𝑠𝑒 𝑥(𝑡) ∗ 𝛿(𝑡 − 𝑡0 ) = 𝑥(𝑡 − 𝑡0 )
𝑇𝑠 𝑇𝑠
𝑛=−∞
∞
1 𝑛
⇒ 𝐺𝛿 (𝑓) = ∑ 𝐺 (𝑓 − )
𝑇𝑠 𝑇𝑠
𝑛=−∞

or
∞

⇒ 𝐺𝛿 (𝑓) = 𝑓𝑠 ∑ 𝐺(𝑓 − 𝑛𝑓𝑠 )

𝑛=−∞

or
∞

⇒ 𝐺𝛿 (𝑤) = 𝑤𝑠 ∑ 𝐺(𝑤 − 𝑛𝑤𝑠 )

𝑛=−∞

∞ ∞ ∞
1 𝑛
⇒ 𝐺𝛿 (𝑤) = 𝑤𝑠 ∑ 𝐺(𝑤 − 𝑛𝑤𝑠 ) = 𝑓𝑠 ∑ 𝐺(𝑓 − 𝑛𝑓𝑠 ) = ∑ 𝐺 (𝑓 − ) − − − (2)
𝑇𝑠 𝑇𝑠
𝑛=−∞ 𝑛=−∞ 𝑛=−∞

Where,
fs: Sampling frequency or sampling rate in Hz (fs = 1/Ts)
ws: Sampling frequency in rad/sec

Note: That’s the end of graphical and analytical proof for Band Limited Signal under impulse
sampling technique.

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 10/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(C) Sampling Theorem for Band Limited Signal-Band Pass Sampling:

Now the task is to analyze the spectrum Gδ(f) corresponds to the Band Limited sampled signal gδ(t)
for the following sampling frequencies by assuming the spectrum of g(t), i.e., G(f) is band limited to
2fm. It is called Band Pass Sampling.
✓ fs = 2fm; exact sampling
✓ fs = 3fm (fs > 2fm); over sampling
✓ fs = fm (fs < 2fm); under sampling

G(f)
G(0)

-fm 0 fm f

(a)fs=2fm 𝐺𝛿 (𝑓) = 𝑓𝑠 ∑∞
𝑛=−∞ 𝐺(𝑓 − 𝑛𝑓𝑠 )

-5fm -4fm -3fm -fs=-2fm -fm 0 fm fs=2fm 3fm 4fm 5fm f

(b)fs=3fm; fs>2fm 𝐺𝛿 (𝑓) = 𝑓𝑠 ∑∞

𝑛=−∞ 𝐺(𝑓 − 𝑛𝑓𝑠 )

-4fm fs=-3fm -2fm -fm 0 fm 2fm fs=3fm 4fm f

(c)fs=fm; fs<2fm 𝐺𝛿 (𝑓) = 𝑓𝑠 ∑∞

𝑛=−∞ 𝐺(𝑓 − 𝑛𝑓𝑠 )

-4fm -3fm -2fm -fs=-fm 0 fs=fm 2fm 3fm 4fm f

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 11/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(D) Reconstruction of Signal from its Samples:

It is clear from the above spectral analysis.
➢ If the sampling frequency, fs =2fm, then all the replicas of G(f) involved in the construction
of Gδ(f) and there is no problem in recovering the original signal g(t) from its sampled
version gδ(t). In this case, reconstruction of signal from its samples is possible.
➢ If the sampling frequency, fs =3fm, i.e., fs > 3fm, then all the replicas of G(f) involved in the
construction of Gδ(f) and there is no problem in recovering the original signal g(t) from its
sampled version gδ(t). In this case also, reconstruction of signal from its samples is possible.
➢ If the sampling frequency, fs =fm, i.e., fs < 2fm, then the replicas of Gδ(f) may be overlapped
and the original signal g(t) can’t be recovered exactly from its sampled version of gδ(t). In
this case, reconstruction of signal from its samples is not possible.

(E) Statement of Sampling Theorem:

A band limited signal of finite energy can be completely reconstructed from its samples when the
samples taken at the rate of fs ≥ 2fm samples/sec or it can be completely reconstructed when the
sampling interval Ts ≤ 1/2fm.

Where,
fm : Message Bandwidth
fs : Nyquist Rate (fs = 2fm)
Ts : Nyquist Interval (Ts = 1/fs = 1/2fm)

(F) Effect of under Sampling-Aliasing Effect:

➢ If the sampling frequency (fs) or sampling rate (1/Ts) is equal to or exceeds the Nyquist rate
(2fm), then all the replicas of G(f) involved in the construction of Gδ(f) and there is no
problem in recovering the original signal g(t) from its sampled version gδ(t).
➢ If the sampling frequency (fs) or sampling rate (1/Ts) is less than the Nyquist rate (2fm), then

the replicas of Gδ(f) may be overlapped and the original signal g(t) can’t be recovered exactly
from its sampled version of gδ(t). The loss of information due to this sampling process is
called aliasing effect.
➢ To avoid aliasing, take the sampling frequency fs ≥ 2fm or the sampling period Ts ≤ 1/2fm.

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 12/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(G) Natural and Flat Top Sampling:

➢ In the case of ideal or impulse sampling, width of each impulse is negligible. Hence impulse
sampling is practically not possible, that’s why natural and flat top sampling techniques are
used in digital communication.
➢ In the case of both natural and flat top sampling, the train of pulse is used instead of train of
impulse.
➢ Graphical analysis of natural sampling process as shown, where, s(t)=g(t) c(t).

g(t)

c(t)

…….-4Ts -3Ts -2Ts -Ts 0 Ts 2Ts 3Ts 4Ts .….. t

s(t)

T
…….-4Ts -3Ts -2Ts -Ts 0 Ts 2Ts 3Ts 4Ts .….. t

Where,
g(t) : Given continuous time or analog signal
c(t) : Train of pulse with a duration T
s(t) : Natural sampled signal
Ts : Sampling period or sampling interval

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 13/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

➢ Graphical analysis of flat top sampling process as shown, where, s(t)=g(t) c(t).

g(t)

c(t)

…….-4Ts -3Ts -2Ts -Ts 0 Ts 2Ts 3Ts 4Ts .….. t

s(t)

…….-4Ts -3Ts -2Ts -Ts 0 Ts 2Ts 3Ts 4Ts .….. t

Where,
g(t) : Given continuous time or analog signal
c(t) : Train of pulse with a duration T
s(t) : Flat top sampled signal
Ts : Sampling period or sampling interval

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 14/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

10. Solved Problem:

(1) Evaluate (a)Auto Correlation Function Rxx(τ) (b)Energy Spectral Density ψ xx(w) of
aperiodic signal, x(t)=ke-atu(t).

(a) Auto Correlation Function:

∞
𝑅𝑥𝑥(𝜏) = ∫ 𝑥(𝑡)𝑥(𝑡 + 𝜏)𝑑𝑡
−∞
∞
= ∫ 𝑘𝑒 −𝑎𝑡 𝑢(𝑡)𝑘𝑒 −𝑎(𝑡+𝜏) 𝑢(𝑡 + 𝜏)𝑑𝑡
−∞
∞
= 𝑘 2 ∫ 𝑒 −𝑎𝑡 𝑒 −𝑎𝑡 𝑒 −𝑎𝜏 𝑢(𝑡)𝑢(𝑡 + 𝜏)𝑑𝑡
−∞
∞
2 −𝑎𝜏
=𝑘 𝑒 ∫ 𝑒 −2𝑎𝑡 𝑢(𝑡)𝑢(𝑡 + 𝜏)𝑑𝑡 ; 𝑢(𝑡) = 1, 𝑡 > 0 & 𝑢(𝑡 + 𝜏) = 1, 𝑡 > −𝜏
−∞

𝐶𝑎𝑠𝑒 − 1: 𝐼𝑓 𝜏 < 0, 𝑡ℎ𝑒𝑛 𝑢(𝑡)𝑢(𝑡 + 𝜏) = 1 𝑓𝑜𝑟 − 𝜏 < 𝑡 < ∞

∞
2 −𝑎𝜏
𝑅𝑥𝑥(𝜏) = 𝑘 𝑒 ∫ 𝑒 −2𝑎𝑡 𝑑𝑡
−𝜏

2 −𝑎𝜏
𝑒 −2𝑎𝑡 ∞
=𝑘 𝑒 |𝑔
−2𝑎 −𝜏

𝑒 −∞ − 𝑒 2𝑎𝜏
= 𝑘 2 𝑒 −𝑎𝜏
−2𝑎
0 − 𝑒 2𝑎𝜏
= 𝑘 2 𝑒 −𝑎𝜏
−2𝑎
𝑘 2 𝑎𝜏
= 𝑒 ,𝜏 < 0
2𝑎

𝐶𝑎𝑠𝑒 − 1: 𝐼𝑓 𝜏 > 0, 𝑡ℎ𝑒𝑛 𝑢(𝑡)𝑢(𝑡 + 𝜏) = 1 𝑓𝑜𝑟 0 < 𝑡 < ∞

∞
𝑅𝑥𝑥(𝜏) = 𝑘 2 𝑒 −𝑎𝜏 ∫ 𝑒 −2𝑎𝑡 𝑑𝑡
0

𝑒 −2𝑎𝑡 ∞
= 𝑘 2 𝑒 −𝑎𝜏 |𝑔
−2𝑎 0

2 −𝑎𝜏
𝑒 −∞ − 𝑒 0
=𝑘 𝑒
−2𝑎
0−1
= 𝑘 2 𝑒 −𝑎𝜏
−2𝑎
𝑘 2 −𝑎𝜏
= 𝑒 ,𝜏 > 0
2𝑎

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 15/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

𝑘 2 𝑎𝜏
𝑒 ,𝜏 < 0
𝑅𝑥𝑥(𝜏) = 2𝑎
𝑘 2 −𝑎𝜏
{2𝑎 𝑒 , 𝜏 > 0
or
𝑘 2 𝑎𝜏 𝑘 2 −𝑎𝜏
𝑅𝑥𝑥(𝜏) = 𝑒 𝑢(−𝜏) + 𝑒 𝑢(𝜏)
2𝑎 2𝑎
or
𝑘 2 −𝑎|𝜏|
𝑅𝑥𝑥(𝜏) = 𝑒
2𝑎

Rxx(τ)
𝑘2
2𝑎

τ
0

(b) Energy Spectral Density:

Ψxx(w) = 𝐹𝑇[𝑅𝑥𝑥(𝜏)]
𝑘 2 −𝑎|𝜏|
= 𝐹𝑇 [ 𝑒 ]
2𝑎
𝑘2
= 𝐹𝑇[𝑒 −𝑎|𝜏| ]
2𝑎
𝑘 2 2𝑎
=
2𝑎 𝑎2 + 𝑤 2
𝑘2
= 2
𝑎 + 𝑤2

Ψ𝑥𝑥(𝑤)

k2/a2

0 w

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 16/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(2) Evaluate (a)Auto Correlation Function Rxx(τ) (b)Power Spectral Density Sxx(w) of
periodic signal, x(t) = ACos(w0t+θ).

(a) Auto Correlation Function:

1 𝑇
𝑅𝑥𝑥(𝜏) = lim ∫ 𝑥(𝑡)𝑥(𝑡 + 𝜏)𝑑𝑡
𝑇→∞ 2𝑇 −𝑇

1 𝑇
= lim ∫ 𝐴𝐶𝑜𝑠(𝑤0 𝑡 + 𝜃)𝐴𝐶𝑜𝑠(𝑤0 (𝑡 + 𝜏) + 𝜃)𝑑𝑡
𝑇→∞ 2𝑇 −𝑇

𝐴2 1 𝑇
= lim ∫ 2𝐶𝑜𝑠(𝑤0 𝑡 + 𝜃)𝐶𝑜𝑠(𝑤0 𝑡 + 𝑤0 𝜏 + 𝜃)𝑑𝑡
2 𝑇→∞ 2𝑇 −𝑇
𝐴2 1 𝑇
= lim ∫ (𝐶𝑜𝑠(2𝑤0 𝑡 + 𝑤0 𝜏 + 2𝜃) + 𝐶𝑜𝑠(𝑤0 𝜏))𝑑𝑡
2 𝑇→∞ 2𝑇 −𝑇
𝑇 𝑇
𝐴2 1
= lim (∫ 𝐶𝑜𝑠(2𝑤0 𝑡 + 𝑤0 𝜏 + 2𝜃)𝑑𝑡 + ∫ 𝐶𝑜𝑠(𝑤0 𝜏) 𝑑𝑡 )
4 𝑇→∞ 𝑇 −𝑇 −𝑇

𝐴2 1
= lim (0 + 2𝑇𝐶𝑜𝑠(𝑤0 𝜏))
4 𝑇→∞ 𝑇
𝐴2
= lim (𝐶𝑜𝑠(𝑤0 𝜏))
2 𝑇→∞
𝐴2
= 𝐶𝑜𝑠(𝑤0 𝜏)
2
Rxx(τ)

A2/2

τ
0

-A2/2

(b) Power Spectral Density:

𝑆𝑥𝑥(𝑤) = 𝐹𝑇[𝑅𝑥𝑥(𝜏)]
𝐴2
= 𝐹𝑇 [ 𝐶𝑜𝑠(𝑤0 𝜏)]
2
𝐴2
= 𝐹𝑇[𝐶𝑜𝑠(𝑤0 𝜏)]
2
𝜋𝐴2
= (𝛿(𝑤 + 𝑤0 ) + 𝛿(𝑤 − 𝑤0 ))
2

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 17/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(3) Evaluate (a)Auto Correlation Function Rxx(τ) (b)Power Spectral Density Sxx(w) of
periodic signal, x(t) = ASin(w0t+θ).

(a) Auto Correlation Function:

1 𝑇
𝑅𝑥𝑥(𝜏) = lim ∫ 𝑥(𝑡)𝑥(𝑡 + 𝜏)𝑑𝑡
𝑇→∞ 2𝑇 −𝑇

1 𝑇
= lim ∫ 𝐴𝑆𝑖𝑛(𝑤0 𝑡 + 𝜃)𝐴𝑆𝑖𝑛(𝑤0 (𝑡 + 𝜏) + 𝜃)𝑑𝑡
𝑇→∞ 2𝑇 −𝑇

𝐴2 1 𝑇
= lim ∫ 2𝑆𝑖𝑛(𝑤0 𝑡 + 𝜃)𝑆𝑖𝑛(𝑤0 𝑡 + 𝑤0 𝜏 + 𝜃)𝑑𝑡
2 𝑇→∞ 2𝑇 −𝑇
𝐴2 1 𝑇
= lim ∫ (𝐶𝑜𝑠(𝑤0 𝜏) − 𝐶𝑜𝑠(2𝑤0 𝑡 + 𝑤0 𝜏 + 2𝜃)+)𝑑𝑡
2 𝑇→∞ 2𝑇 −𝑇
𝑇 𝑇
𝐴2 1
= lim (∫ 𝐶𝑜𝑠(𝑤0 𝜏) 𝑑𝑡 − ∫ 𝐶𝑜𝑠(2𝑤0 𝑡 + 𝑤0 𝜏 + 2𝜃)𝑑𝑡 )
4 𝑇→∞ 𝑇 −𝑇 −𝑇

𝐴2 1
= lim (2𝑇𝐶𝑜𝑠(𝑤0 𝜏) − 0)
4 𝑇→∞ 𝑇
𝐴2
= lim (𝐶𝑜𝑠(𝑤0 𝜏))
2 𝑇→∞
𝐴2
= 𝐶𝑜𝑠(𝑤0 𝜏)
2
Rxx(τ)

A2/2

τ
0

-A2/2

(b) Power Spectral Density:

𝑆𝑥𝑥(𝑤) = 𝐹𝑇[𝑅𝑥𝑥(𝜏)]
𝐴2
= 𝐹𝑇 [ 𝐶𝑜𝑠(𝑤0 𝜏)]
2
𝐴2
= 𝐹𝑇[𝐶𝑜𝑠(𝑤0 𝜏)]
2
𝜋𝐴2
= (𝛿(𝑤 + 𝑤0 ) + 𝛿(𝑤 − 𝑤0 ))
2

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 18/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(4) Verify the Parseval’s theorem for a signal 𝒙(𝒕) = 𝟑𝒆−𝟐𝒕 𝒖(𝒕)
3
𝑊𝑒 𝑘𝑛𝑜𝑤 𝑡ℎ𝑎𝑡, 𝐹𝑇[𝑥(𝑡)] = 𝑋(𝑤) = 𝐹𝑇[3𝑒 −2𝑡 𝑢(𝑡)] =
2 + 𝑗𝑤
∞
1 ∞
𝑃𝑎𝑟𝑠𝑒𝑣𝑎𝑙𝑙 ′ 𝑠 𝑇ℎ𝑒𝑜𝑟𝑒𝑚, 𝐸 = ∫ | 𝑥(𝑡) |2 𝑑𝑡 = ∫ | 𝑋(𝑤) |2 𝑑𝑤
−∞ 2𝜋 −∞

∞
𝐿𝐻𝑆 = ∫ | 3𝑒 −2𝑡 𝑢(𝑡)|2 𝑑𝑡
−∞
∞
= ∫ 9𝑒 −4𝑡 𝑑𝑡
0

𝑒 −4𝑡 ∞
=9 |𝑔
−4 0
9
=
4

1 ∞ 3 2
𝑅𝐻𝑆 = ∫ | | 𝑑𝑤
2𝜋 −∞ 2 + 𝑗𝑤
1 ∞ 9
= ∫ 𝑑𝑤
2𝜋 −∞ 4 + 𝑤 2
9 ∞ 1
= ∫ 2 𝑑𝑤
2𝜋 −∞ 2 + 𝑤 2
9 1 −1
𝑤 ∞
= tan | 𝑔
2𝜋 2 2 −∞
9
= (tan−1 ∞ − tan−1(−∞))
4𝜋
9 𝜋 𝜋
= ( + )
4𝜋 2 2
9
= 𝜋
4𝜋
9
=
4

LHS = RHS

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 19/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(5) Evaluate the output of a low pass filter, when the input message signal m(t) = Cos(2πfmt) with
fm = 50Hz is sampled at sampling frequency fs = 40Hz is transmitted through low pass filter with
cutoff frequency fc = 20Hz.

Sampler Low Pass Filter

m(t) mδ(t) m0(t)
with with
Sampling frequency Cutoff frequency
M(f) Mδ(f) M0(f)
fs=40Hz fc=20Hz

Input message signal

𝑚(𝑡) = 𝐶𝑜𝑠(2𝜋𝑓𝑚 𝑡); 𝐹𝑇[𝐶𝑜𝑠(𝑤0 𝑡)] = 𝜋(𝛿(𝑤 + 𝑤0 ) + 𝛿(𝑤 − 𝑤0 ))
⇒ 𝑀(𝑓) = 𝛿(𝑓 + 𝑓𝑚) + 𝛿(𝑓 − 𝑓𝑚)
= 𝛿(𝑓 + 50) + 𝛿(𝑓 − 50)

Sampled form of M(f)

∞

𝑀𝛿 (𝑓) = 𝑓𝑠 ∑ 𝑀(𝑓 − 𝑛𝑓𝑠)

𝑛=−∞
∞

= 40 ∑ 𝑀(𝑓 − 40𝑛)
𝑛=−∞

= ⋯ + 40𝑀(𝑓 + 40) + 40𝑀(𝑓) + 40𝑀(𝑓 − 40) + ⋯

= ⋯ + 40𝛿(𝑓 + 90) + 40𝛿(𝑓 − 10) + 40𝛿(𝑓 + 50) + 40𝛿(𝑓 − 50)
+40𝛿(𝑓 + 10) + 40𝛿(𝑓 − 90) + ⋯
= ⋯ + 40[𝛿(𝑓 + 90) + 𝛿(𝑓 − 90)] + 40[𝛿(𝑓 + 50) + 40𝛿(𝑓 − 50)]
+40[𝛿(𝑓 + 10) + 𝛿(𝑓 − 10)] + ⋯

Output of a LPF with cutoff frequency fc=20Hz is

𝑀0 (𝑓) = 40[𝛿(𝑓 + 10) + 𝛿(𝑓 − 10)]
⇒ 𝑚0 (𝑡) = 40𝐶𝑜𝑠(2𝜋 × 10𝑡)
= 40𝐶𝑜𝑠(20𝜋𝑡)

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 20/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(6) Evaluate the output of a low pass filter, when the input message signal m(t) = Cos(2πfmt)
with fm = 50Hz and it is sampled at a sampling frequency fs = 40Hz is transmitted through a
low pass filter with cutoff frequency fc = 40Hz.

Sampler Low Pass Filter

m(t) mδ(t) m0(t)
with with
Sampling frequency Cutoff frequency
M(f) Mδ(f) M0(f)
fs=40Hz fc=40Hz

Input message signal

𝑚(𝑡) = 𝐶𝑜𝑠(2𝜋𝑓𝑚 𝑡)
⇒ 𝑀(𝑓) = 𝛿(𝑓 + 𝑓𝑚) + 𝛿(𝑓 − 𝑓𝑚)
= 𝛿(𝑓 + 50) + 𝛿(𝑓 − 50)

Sampled form of M(f)

∞

𝑀𝛿 (𝑓) = 𝑓𝑠 ∑ 𝑀(𝑓 − 𝑛𝑓𝑠)

𝑛=−∞
∞

= 40 ∑ 𝑀(𝑓 − 40𝑛)
𝑛=−∞

=. . . +40𝑀(𝑓 + 80) + 40𝑀(𝑓 + 40) + 40𝑀(𝑓) + 40𝑀(𝑓 − 40) + 40𝑀(𝑓 − 80)+. ..

=. . . +40𝛿(𝑓 − 30) + 40𝛿(𝑓 − 10) + 40𝛿(𝑓 + 50) + 40𝛿(𝑓 − 50) + 40𝛿(𝑓 + 10)
+40𝛿(𝑓 + 30)
= ⋯ + 40[𝛿(𝑓 + 50) + 𝛿(𝑓 − 50)] + 40[𝛿(𝑓 + 30) + 40𝛿(𝑓 − 30)]
+40[𝛿(𝑓 + 10) + 𝛿(𝑓 − 10)]+. ..

Output of a LPF with cutoff frequency fc = 40Hz is

𝑀0 (𝑓) = 40[𝛿(𝑓 + 30) + 𝛿(𝑓 − 30)] + 40[𝛿(𝑓 + 10) + 𝛿(𝑓 − 10)]
⇒ 𝑚0 (𝑡) = 40𝐶𝑜𝑠(2𝜋 × 30𝑡) + 40𝐶𝑜𝑠(2𝜋 × 10𝑡)
= 40𝐶𝑜𝑠(60𝜋𝑡) + 40𝐶𝑜𝑠(20𝜋𝑡)

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 21/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

11. Assignment Questions

1. Evaluate the autocorrelation function RXX() corresponds to the power spectral density
157 + 12𝑤 2
𝑆𝑋𝑋 (𝑤) =
(16 + 𝑤 2 )(9 + 𝑤 2 )

2. Evaluate the cross-correlation function, RXY() between the stochastic process X(t) and Y(t),
which has the cross power spectral density
1 + 𝑗𝑤, | 𝑤| ≤ 1
𝑆𝑋𝑌 (𝑤) = {
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒

3. A stochastic process X(t) has an autocorrelation function, 𝑅𝑋𝑋 (𝜏) = 3𝑒 −2|𝜏|.

(i)Evaluate the power spectral density, SXX(w).
(ii)Draw the power spectrum.

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 22/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

12. Quiz Questions

(16) The autocorrelation function of a periodic signal x(t) is
𝟏 𝑻/𝟐
(A) 𝑹𝒙𝒙 (𝝉) = 𝑳𝒕 ∫ 𝒙(𝒕)𝒙(𝒕 + 𝝉)𝒅𝒕
𝑻→∞ 𝑻 −𝑻/𝟐
𝟏 𝑻/𝟐
(B) 𝑹𝒙𝒙 (𝝉) = 𝑳𝒕 ∫ 𝒙(𝝉)𝒙(𝒕 + 𝝉)𝒅𝒕
𝑻→∞ 𝑻 −𝑻/𝟐
𝟏 𝑻/𝟐
(C) 𝑹𝒙𝒙 (𝝉) = 𝑳𝒕 𝑻 ∫−𝑻/𝟐 𝒙(𝒕)𝒙(−𝒕 + 𝝉)𝒅𝒕
𝑻→∞
1 𝑇/2
(D) 𝑅𝑥𝑥 (𝜏) = 𝐿𝑡 ∫ 𝑥(𝜏)𝑥(−𝑡 + 𝜏)𝑑𝑡
𝑇→∞ 𝑇 −𝑇/2

(19) The measure of similarity between a signal x(t) and delayed form of another signal y(t) is
(A) Convolution
(B) Auto Correlation
(C) Cross Correlation
(D) Energy Spectral Density

(20) Autocorrelation function of aperiodic signal x(t) is

∞
(A) 𝑹𝒙𝒙 (𝝉) = ∫−∞ 𝒙(𝒕)𝒙(𝒕 − 𝝉)𝒅𝒕
∞
(B) 𝑹𝒙𝒙 (𝝉) = ∫−∞ 𝒙(𝒕)𝒙(−𝒕 + 𝝉)𝒅𝒕
∞
(C) 𝑹𝒙𝒙 (𝝉) = ∫−∞ 𝒙(𝝉)𝒙(𝒕 + 𝝉)𝒅𝒕
∞
(D) 𝑅𝑥𝑥 (𝜏) = ∫−∞ 𝑥(𝑡 − 𝜏)𝑥(𝑡 + 𝜏)𝑑𝑡

(23) Cross correlation function between x(t) & y(t) is

∞
(A) 𝑹𝒙𝒚 (𝝉) = ∫−∞ 𝒙(𝒕)𝒚(𝒕 − 𝝉)𝒅𝒕
∞
(B) 𝑹𝒙𝒚 (𝝉) = ∫−∞ 𝒙(𝒕)𝒚(−𝒕 + 𝝉)𝒅𝒕
∞
(C) 𝑹𝒙𝒚 (𝝉) = ∫−∞ 𝒙(𝝉)𝒚(𝒕 + 𝝉)𝒅𝒕
∞
(D) 𝑅𝑥𝑦 (𝜏) = ∫−∞ 𝑥(𝑡 − 𝜏)𝑦(𝑡 + 𝜏)𝑑𝑡

(24) If x(t) and y(t) are orthogonal, then the cross-correlation function is
(A) 0
(B) 1
(C) -1
(D) Infinity

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 23/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(25) Choose the correct property of autocorrelation function

(A) 𝑹𝒙𝒙 (−𝝉) = 𝑹𝒙𝒙 (𝝉)
(B) 𝑹𝒙𝒙 (𝟎) = 𝑬𝒏𝒆𝒓𝒈𝒚 𝒐𝒇𝑺𝒊𝒈𝒏𝒂𝒍
(C) 𝑹𝒙 (𝟎) ≥ |𝑹𝒙 (𝝉)|
(D) All the above

(26) The Auto correlation function and Power spectral density form
(A) Laplace Transform Pair
(B) Z-Transform Pair
(C) Fourier Transform Pair
(D) All the above

(27) Which are Fourier Transformable pairs?

(A) ESD & ACF for an Energy signal
(B) PSD & ACF for Power signal
(C) Impulse and Constant signal
(D) All the above

(28) Auto correlation function of 𝒙(𝒕) = 𝟐𝑪𝒐𝒔(𝟐𝒕)𝒖(𝒕) is

(A) 𝑹𝒙𝒙 (𝝉) = 𝟐𝑺𝒊𝒏(𝟐𝝉)
(B) 𝑹𝒙𝒙 (𝝉) = 𝟐𝑪𝒐𝒔(𝟐𝝉)
(C) 𝑹𝒙𝒙 (𝝉) = 𝟒𝑺𝒊𝒏(𝟐𝝉)
(D) 𝑹𝒙𝒙 (𝝉) = 𝟒𝑪𝒐𝒔(𝟐𝝉)

(29) Auto correlation function of 𝒙(𝒕) = 𝟐𝑺𝒊𝒏(𝟐𝒕)𝒖(𝒕) is

(A) 𝑹𝒙𝒙 (𝝉) = 𝟐𝑺𝒊𝒏(𝟐𝝉)
(B) 𝑹𝒙𝒙 (𝝉) = 𝟐𝑪𝒐𝒔(𝟐𝝉)
(C) 𝑹𝒙𝒙 (𝝉) = 𝟒𝑺𝒊𝒏(𝟐𝝉)
(D) 𝑹𝒙𝒙 (𝝉) = 𝟒𝑪𝒐𝒔(𝟐𝝉)

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 24/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(25) The process of converting a continuous-time signal into a discrete-time signal is called
(A) Sampling
(B) Quantization
(C) Both (A) & (B)
(D) None of the above

(26) The time interval between any two adjacent samples of sampled signal is
(A) Nyquist Rate
(B) Nyquist Interval
(C) Sampling period
(D) Both (B) and (C)

(27) What is the minimum sampling frequency required for the sampling process with no
distortion, given analog signal 𝒙(𝒕) = 𝑪𝒐𝒔(𝟏𝟎𝟎𝝅𝒕) + 𝟔𝑺𝒊𝒏(𝟏𝟖𝟎𝝅𝒕) is
(A) 100 Hz
(B) 90 Hz
(C) 180 Hz
(D) 360 Hz

(30) A band limited signal with highest frequency content of 1000 Hz is undergoing sampling at
uniform intervals. For recovery of the signal in an unambiguous way, the sampling frequency
should be necessarily greater than
(A) 2000 Hz
(B) 1500 Hz
(C) 1000 Hz
(D) 500

(31) A band-limited signal is sampled at the Nyquist rate. The signal can be recovered by
passing the samples through
(A) High Pass Filter
(B) Low Pass Filter
(C) Band Pass Filter
(D) Band Stop Filter

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 25/26
 Signals and Systems-23EC02 UNIT-IV Correlation and Sampling Theorem
------------------------------------------------------------------------------------------------------------------------

(32) The process of converting an analog signal into a digital signal involves
(A) Sampling
(B) Quantization
(C) Both (A) & (B)
(D) None of the above

(33) Consider the signal m(t) with spectrum M(f) which is sampled with an ideal impulse train
corresponding to sampling frequency fs. The spectrum of resulting sampled signal is
∞

(𝐴)𝐹𝑇[𝑚𝛿 (𝑡)] = 𝑤𝑠 ∑ 𝛿(𝑤 − 𝑛𝑤𝑠 )

𝑛=−∞
∞

(𝐵)𝐹𝑇[𝑚𝛿 (𝑡)] = 𝑓𝑠 ∑ 𝛿(𝑓 − 𝑛𝑓𝑠 )

𝑛=−∞
∞
1 𝑛
(𝐶)𝐹𝑇[𝑚𝛿 (𝑡)] = ∑ 𝛿 (𝑓 − )
𝑇𝑠 𝑇𝑠
𝑛=−∞

(𝑫)𝑨𝒍𝒍 the above are correct

(49) What is the minimum sampling frequency fs required for no distortion if the maximum
message frequency is fm.
fs = 2fm

(50) Consider the signal m(t) with spectrum M(f) which is sampled with an ideal impulse train
corresponding to sampling frequency fs. The spectrum of resulting sampled signal is
∞ ∞ ∞
1 𝑛
𝐹𝑇[𝑚𝛿 (𝑡)] = 𝑤𝑠 ∑ 𝑀(𝑤 − 𝑛𝑤𝑠 ) = 𝑓𝑠 ∑ 𝑀(𝑓 − 𝑛𝑓𝑠 ) = ∑ 𝑀 (𝑓 − )
𝑇𝑠 𝑇𝑠
𝑛=−∞ 𝑛=−∞ 𝑛=−∞

(51) What is the minimum sampling frequency fs required for no distortion if the message
signal m(t)=Cos(2fmt) with fm=50Hz.

fs = 2fm = 2x50 = 100Hz

(52) What is the minimum sampling frequency fs required for no distortion if the message
signal m(t)= Cos(3fmt) + Cos(2fmt)+ Cos(fmt).
m(t)= Cos(2(3fm/2)t) + Cos(2(fm)t)+ Cos(2(fm/2)t).
Minimum sampling frequency fs=2 x Maximum message frequency
fs = 2( 3fm/2 ) = 3fm

----------------------------------------------------------------------------------------------------------------------------- --------------------
Dr. B. Ramesh Reddy, Professor of ECE & Vice Principal, LBRCE, Mylavaram, NTR Dist., AP. Page. 26/26