Random Processes Spectral Characteristics
Random Processes Spectral Characteristics
Random Processes Spectral
Characteristics
Dr. Ali Hussein Muqaibel
Dr. Ali Muqaibel 1
Introduction
• You should be familiar with F.T. by integration or
by using tables.
↔
For deterministic signals we can sketch the magnitude
and phase of
• We do not need full pdf to find power. or are
enough.
Dr. Ali Muqaibel 2
1
4/7/2013
Power Density Spectrum & Properties
• The power density spectrum (power spectral density) PSD.
• Parsaval’s Theorem:
• lim lim
→ →
• lim
→
•
• The factor is there because we are using .
• Power is related to the time average of the second moment
, for w.s.s.
Dr. Ali Muqaibel 3
Example
• Find the power in the following random process
cos , & 0,
• cos cos 2 2
• lim
→
• Alternatively,
∗
• →
• , =
• Try it yourself!
Dr. Ali Muqaibel 4
2
4/7/2013
Relationship between Power Spectrum
and Autocorrelation Function
↔
,
,
For W.S.S. , , we ge t the Wiener‐Khintchine relations
2
2 2
Explain what happens through RC filters
2
Dr. Ali Muqaibel 5
Example II:
1
•
0
• 0 & , 0, . Find the PSD.
•
1 1
Dr. Ali Muqaibel 6
3
4/7/2013
Properties of the PSD
• Note: some books use for power spectral density. Others use …etc
1. 0
2. , for real .
3. is real
4.
5.
– recall taking the derivative in frequency domain is equivalent to multiplying by
6. PSD and time average of the auto‐correlation form a Fourier
transform pair
,
Experiment with Matlab
,
For W.S.S. ,
Dr. Ali Muqaibel 7
Practice 1
• Determine which of the following functions can and
cannot be valid power density spectrums. For those
that are not, explain why.
Dr. Ali Muqaibel 8
4
4/7/2013
Practice 2
A random process has the autocorrelation function
exp
Where , are positive constants
(a) Find and sketch the power spectrum of X(t)
(b) Compute the average power in the lowpass part
of the power spectrum.
(c)Repeat for bandpass case.
In each case assume ≫ .
Dr. Ali Muqaibel 9
Bandwidth of Power Density Spectrum
for baseband processes
• Low pass or Baseband: most of the spectral
components are around 0, and the
magnitude decreases at higher frequencies.
• The PSD is similar to the PDF for being real and
positive but the area need not be 1.
• The rms bandwidth , , is defined like the
standard deviation of the normalized PSD to be:
Dr. Ali Muqaibel 10
5
4/7/2013
Bandwidth of Power Density Spectrum
for Bandpass Processes
• Bandpass process: the significant components cluster
near some frequency and . (assume real
processes)
• Mean frequency:
=
• rms bandwidth of bandpass (real) processes:
• Why there is a factor of 4?
Dr. Ali Muqaibel 11
White Noise
2 2
• Unrealizable because of the infinity average power:
1
∞
2
• Thermal noise generated be electrons in any electrical conductor has a PSD that is constant
up to a very high frequency and then decreases.
• A resistor at temperature Kelvin produces noise across the open circuited terminal having a
power spectrum
1
• where 7.64 10 , 290 .
• Remains above 0.9 for very high frequencies around 1000GHz .
Dr. Ali Muqaibel 12
6
4/7/2013
Bandlimited White Noise
sin
Dr. Ali Muqaibel 13
Bandpass Bandlimited White Noise
sin
2 cos
2 2
0
2
Dr. Ali Muqaibel 14
7
4/7/2013
Practice
A wide‐sense stationary noise process has
an autocorrelation
Find and sketch the PSD
Answer
Using the FT def.
Dr. Ali Muqaibel 15