Professional Documents
Culture Documents
CTFS, Continuous Time Fourier Series, (Continuous Time, Periodic) CTFT, Continuous Time Fourier Transform, (Continuous Time, Nonperiodic)
Time
Differentiation
Transforms of
Periodic Signals
Parseval’s Theorem The product of two energy signals in the time domain corresponds to the convolution of their transforms in the frequency domain.
Integral Definition
of an Impulse
Duality
Total-Area Integral
Using Fourier
Transforms
Integration
CTFT
The Fourier transform of a signal x(t) exists if (1) it is an energy signal and (2) any discontinuities are finite – This requires the trick of using limits
• True for all signals of finite amplitude and duration • As with CTFS, convergence does not imply that the inverse Fourier transform will recover the signal
• Do periodic signals have a Fourier transform? • However, will be equal at all points except for discontinuities
– No, but we can still apply the transform if we allow X(j!) to be expressed in terms of impulse functions
CTFS
Convergence Since the CTFS includes an infinite series, we must consider under what conditions it converges
• An infinite sum is said to converge so long as it is bounded A sufficient condition for convergence (not proven) is
• In other words, the signal has – Finite power – Finite energy over a single period
Dirichlet The Fourier series representation of a periodic signal x(t) converges if all of the following 3. Finite number of distinct maxima and minima in T
Conditions for conditions are met. 4. x(t) is single valued
Convergence These are sufficient, but not necessary, conditions.
1.
2. Finite number of discontinuities in a period T
Terminology The coefficients X[k] are called the spectral coefficients or Fourier series coefficients of x[n] The function argX[k] is called the phase spectrum of x(t)
The function |X[k]| is called the magnitude spectrum of x(t)
Complex • The phase is odd: X*−k+ = −X[k]
Conjugate • The real part is even: Re,X*−k+- = Re,X*k+-
Symmetry This complex-conjugate symmetry ensures that the imaginary components of the sum cancel • The imaginary part is odd: Im,X*−k+- = −Im,X*k+-
each other and x(t) is therefore real-valued.
The complex-conjugate symmetry of the coefficients ensures that for real-valued signals x(t)
• The amplitude is even: |X*−k+| = |X*k+|
Gibb's • For finite Fourier series estimates there is apparent overshoot • The maximum overshoot (error) does not decrease as N → ∞
Phenomenon • As N increases, the edges become sharper and more accurate
Signals
Instantaneous
Signal Power
Signal Energy
Average Signal
Power
Energy For many signals the energy integral does not converge because the signal is not "time limited" A signal cannot be both an energy signal and a power signal
Comments and therefore the energy goes to infinity.
A signal is called an energy signal if E∞ < ∞ Signals with finite energy have zero average power:
A signal is called a power signal if 0 < P ∞ < ∞ Signals of finite duration and amplitude have finite energy:
A signal can be an energy signal, a power signal, or neither type
Discrete-time signals formed by sampling periodic continuous-time signals may have a different period or may even be aperiodic
Two different-looking analytical descriptions of discrete-time functions may, in fact, be identical A time-shifted version of a discrete-time function can produce decimation or undefined
Time scaling a discrete-time function can produce decimation or undefined values, phenomena that do not occur when time scaling continuous-time functions.
Since discrete-time functions are only defined for integer values of n, the values of expressions line g[2.7] or g[3/4] are simply undefined.
If we create a discrete-time sinusoid by sampling a continuous-time sinusoid, the period of the discrete-time sinusoid may not be readily apparent and, the discrete-time sinusoid may not even be periodic.
Other Notes
Where,
And
The requirement on a discrete-time sinusoid that it be periodic is that, for some discrete time n and some integer m, 2F0n = 2m. In
words, F0 must be a rational number (ratio of integers).
Units Discussion: "n" is a really a time index, not time itself. F0 should have units of cycles/sample to make 2F0n dimensionless and 0 should have units of radians/sample.
Unit Impulse Disc Time
Functions
Property
Unit Ramp Rectangle Periodic Impulse
Ramp[n]= rectNw[n]=
Time Compression Time Scaling of the form nKn, where |K|>1 and K is an integer. Time compression for discrete-time functions is similar to time compression for continuous-time functions in that the
function seems to occur faster in time. But in the case of discrete time functions there is another effect called decimation.
Time Expansion If we want to graph g[n/2] for each integer value of n, we must assign a value to g[n/2] by finding the corresponding value in the original function definition. But when n is one, n/2 is
one-half and g[1/2] is not defined. We could simply leave those values undefined or we interpolate.
Properties
Differencing The operation on a discrete-tie signal that is analogous to the derivative. The first forward difference of the discrete-time function g[n] is g[n+1]-g[n]. The first backward difference of a
discrete time function is g[n]-g[n-1], which is the first forward difference of g[n-1].
Accumulation The discrete-time counterpart of integration is accumulation (or summation). The accumulation function is described by:
Periodic Functions The fundamental frequency is F0 = 1/N0 in cycles or 0=2/N0 in radians
Signal Energy Signal
Power
One common use for difference equations is to approximate differential equations. The properties of discrete-time systems are basically the same as they are for continuous-time systems.
P
A discrete-time system is stable if all its eigenvalues are less than one in magnitude.
Just as for continuous-time systems, any time a discrete-time system can exhibit an unbounded response to a bounded excitation of any kind, it is classified as BIBO unstable system. So the stability of feedback
systems depends on the nature of the feedback.
Other
Notes
If we multiply the excitation of this system by any constant, the response is multiplied by the same constant, so this system is homogeneous. If we delay the excitation of this system by any time n0, we delay
the response by that same time. Therefore, this system is also time invariant. If we add any two signals to form the excitation of the system, the response is the sum of the responses that would have occurred
by applying the two signals separately. Therefore, this system is an LTI discrete-time system. This system also has a bounded response for any bounded excitation. it is also stable.
Discrete-time LTI systems are usually described by linear, constant-coefficient difference equations. The eigenfunctions of these equations are functions of the form z^n where z is a complex constant.
Chapter 7 - Time-Domain Analysis of Discrete-Time Systems
Every LTI system is completely characterized by its impulse response. The response of an LTI system to an arbitrary signal can be found by convolving the signal with the system impulse response.
SOIP
The impulse response of a cascade connection of LTI systems is the convolution of the individual impulse responses
The impulse response of a parallel connection of LTI systems is the sum of the individual impulse responses A LTI system is BIBO stable if its impulse response is absolutely summable.
LTI systems can be represented by block diagrams and this type of representation is useful both in synthesizing systems and in understanding their dynamic behavior.
Convolution: The excitation for any discrete-time system is made up of a sequence of impulses with different strengths, occurring at different times. Therefore, invoking linearity and time-invariance, the
response of a LTI system will be the sum of all the individual responses to the individual impulses.
Convolution The value of the response y at any discrete time n can be found by summing all the products of the excitation x at discrete times m
with the impulse response h at discrete time n-m for m ranging from negative to positive infinity. For an LTI system, the impulse
Sum response of the system is a complete description of how it responds to any signal.
Convolution
P
S
The sinusoids used in the Fourier series to represent a signal are all orthogonal to each other. The complex trigonometric forms of the Fourier are related through Euler's identity.
The DTFS of a signal is a finite summation because of the nature of discrete time. If a stable LTI system is excited by a periodic signal, the response is also a periodic signal with the same fund. period.
The fundamental concept of the discrete-time Fourier series is to find a way of expressing any arbitrary signal as a linear combination of complex sinusoids. Then we can take advantage of linearity and time-
invariance and find the response to each complex sinusoid one at a time and then add all those response.
The discrete-time Fourier series (DTFS) expresses arbitrary signals as linear combinations of complex sinusoids so we can use superposition to find the response of any LTI system to any arbitrary signal simply by
summing the responses to the individual complex sinusoids.
In discrete time, exact representation is always achieved with a finite number of sinusoids.
The k=Nf complex sinusoid and the k=0 complex sinusoid are identical functions (because n must be an integer) as are the k=Nf+1 and the k=1 complex sinusoids. This proves that when we add any integer
multiple (including negative integers) of the representation time Nf to the harmonic number of any particular discrete-time complex sinusoid, we get an identical discrete-time complex sinusoid. Therefore, any
range of consecutive harmonic numbers k exactly Nf in length is a complete set of complex sinusoids.
DTFS Formulas If we let n0=0 the DTFS
becomes the DFT:
DTFS Formulas
So the DFT and the DTFS are very similar, differing only by a scale constant N F if the choice of the first n in the DTFS summation is n0=0. ;
Summarizing, if a signal x[n] has a fundamental period N0, and NF is an integer multiple of N0, its DTFS harmonic function is:
where 0≤n<NF
In the very important special case in which we represent a periodic function by its DTFS
using its fundamental period N0 as the representation time, the forward and inverse
DTFS operations become:
If we use an integer multiple m of the fundamental period as the
representation time, the formulas become:
Time Time Linearit Frequency Conjugation For these
shifting reversal y shifting properties
Time Let z[n]=x[an], a>0. If a is not an integer, then some values of z[n] will be undefined and a
Scaling DTFS cannot be found for it. If as is an integer, the z[n] is a decimated version of x[n], and
Properties
some values of x[n] do not appear in z[n]. The result to the right says the at the harmonic
function of z is the same as the harmonic function for x except divided by m.
Change of Using two periods instead of one does not add any information because the signal is exactly
Period the same in each period. We would in general get extra harmonic information, but here all
of the extra harmonics have zero amplitude.
For x[n] even and real-valued, X[k] is even and real-valued For x[n] odd and real-valued, X[k] is odd and purely imaginary
Parseval's theorem The average signal power of the signal is equal to the sum of the average signal powers in ;
its DTFS harmonics
Chapter 11 - Discrete Time Fourier Transform
The DTFS is a special case of the DTFT The DTFT is always periodic with period one in the F domain or period 2 in the domain
O
P
S
For periodic signals, there are simple conversions between a Fourier transform and a Fourier Series.
Defin
ed
Multiplication- Accumulation
Convolution Definition of a
Duality periodic Impulse
Parseval's Theorem
Concept:
Remark #1 X(θ) is computed from samples of x(t) (i.e. from x*n+) but X(θ) is a continuous function
Remark #2 ;
Remark #3 X(θ) is periodic with period 2, 0 to 2, but in practice it's
calculated from - to +.
"What should fs be? " per "Nyquist Criteria" to prevent overlap distortion, let Fx be the highest frequency component of X(θ), then fs>Fx
DFT presumes x(t) is periodic, DTFT works on "aperiodic" signals, DFT gives sample values of DTFT only over certain time duration. This is some times called "windowing".
Mcnames Lower frequencies are those that are near 0, high frequencies are those near ±, intermediate frequencies are those in between. The highest frequency, radians per sample is equal to 0.5 cycles per sample.
Miscellaneous
Series
Eqs.
Chapter 16 - The z transform
Some signals that do not have a DTFT do have a z transform Every z transform has an associated region of convergence in the z plane
SOI
An inverse z transform can be found by the direct inversion integral, synthetic division or partial-fraction expansion. Use of the direct inversion integral is rare and synthetic division does not provide a closed-
P
Exam 1
N samples of a signal x(t) are taken over one period to calculate a set of 10 Fourier coefficients using a DFT. How The spacing between Fourier Coefficients is is the sample frequency and N is the number
1
would the DFT results change if the N samples were zero padded with 3N additional samples? of samples. Zero padding with 3N samples makes the spacing
Four samples are taken over one period of a signal. The sample values are:
2 ; ; ;
x[0]=1; x[1]=0; x[2]=0; x[3]=0. Plot the first 5 Fourier Coefficients
3 A periodic signal g(t) is frequency limited. What effect does frequency limiting have on the CTFS? The CTFS of g(t) will be finite.
In the following LTI system ;
4
If
Given a) ; Thus,
5 a) Find all of the Fourier Series Coeffs for x(t).
b) The average power is
b) What is the average power in x(t)?
Briefly explain how the time compression of x(t) effects X(ω). The Fourier transform property is
6
Illustrate the effect by plotting the Fourier transform of sin(t) and the Fourier transform of sin(5t). Thus,
Consider the pulse train. The CTFS will produce a pulse train with near infinite slopes but you get overshoot and ringing at the corners. That overshoot and ringing
8 Briefly describe the Gibbs Phenomenon.
is the Gibbs phenomenon. Even taking an infinite CTFS will not eliminate it.
One of your classmates says "If the input to a LTI system Agree. The input is of the form: ;
can be represented with a CTFS, then the output can also But each term is an eigenfunction for a LTI system. Hence the same complex exponential appears on the system output. By superposition, a linear
9
be represented by a CTFS". Do you agree or disagree? combination of eigenfunctions on the input of a LTI system creates a linear combination of eigenfunctions on the output of the system. The output of the
Justify your answer. system is thus of the form: which is a CTFS.
For each of the following signals, indicate if they are power signals, energy signals or neither power or energy signals.
a) Therefore it is an energy signal.
10
b) ; Therefore it is neither