Signals and spectra
Signals and spectra in electrical system
• Signals in electrical systems
– Voltage and current waveform
– Defined frequency content
• EM field/wave is defined in terms of frequency,
amplitude and space
• Design of EMC circuits
– requires knowledge of signals in electrical systems, and
their frequency content as high frequency signals are often
present in a system
Recap of signals in electrical systems
• Signals – voltage or current waveforms
Energy
Power
Signal
classification Deterministic
Non-
deterministic
Periodic
Non-periodic
Signal classification
• Energy signals – possess finite energy over infinite
time interval i.e., (t1:-; t2: +)
• Power signals – possess finite average power
• Deterministic signals – precise time behavior is
known; can be expressed mathematically. E.g.
sinusoid, clock
• Non-deterministic signals – random nature; only
signal statistics is known. E.g. digital data signals
• Periodic signals – repetitive time domain signals
– g(t) = g(t nT); t: time, T: period, n=1,2,3,…, f=1/T
– Periodic signals are power signals
– E.g. sinusoids, digital clock
g(t) = A cos(2ft+); A: amplitude, : phase
• A-periodic/non-periodic signals – do not have
periodic behavior
– Non-periodic signals are energy signals
Signals and spectra in electrical system
• Signal types encountered in electrical systems
– narrowband continuous wave (E.g. sinusoid)
– repetitive broadband signals (E.g. digital clock)
– single event broadband signals (E.g. spark discharges,
switching)
• Relationship between rise time and wave shape, and
their spectra is essential for EMC circuit design
Signal representation
• Series expansions and basis functions
– Linear combination of simple signals also known as basis
functions
• Periodic signal, g(t) with period T can be expressed
as,
g(t)
where cn :expansion coefficients; n(t): periodic orthogonal
basis function with period, T satisfying the condition,
• Expansion coefficients are as follows:
• Thus, system response to a complex periodic signal
can be determined by finding the linear superposition
of the responses for simple inputs
• Choice of basis function is important
• E.g. digital clock with period T is composed of
several sinusoids oscillating at some integer multiple
of f
Fourier series
• Most common representation of periodic signals
• Signal must satisfy the following conditions for the
series to converge
1. Function/signal, f(t) must be absolutely integrable over a
period
2. f(t) is finite for all t in a period, T
3. f(t) has a finite number of maxima and minima over T
4. f(t) has a finite number of discontinuities in a period, T
Trignometric Fourier series
• Periodic signal f(t) can be represented as
where
• The terms an and bn are found using orothogonal
properties of the basis functions
Complex exponential Fourier series
• More common and useful representation
• Basis function, with
orthogonality property,
• Fourier series expansion:
with expansion coefficients,
; c-n=cn*
• Relationship with trignometric Fourier Series coeff
An=Re(2*Cn)
Bn= Im(-2*Cn)
Justification for using
• Let Sn(t) be a Fourier series representation with 2n+1
terms,
• The number n is selected such that error,
is minimized in mean square terms, i.e.,
Fourier series properties
• Linearity
If waveform is a linear sum of more functions,
Then f(t) can be written as linear combination of the sum of the Fourier
series.
• Time shifting
If a waveform is shifted in time by , the Fourier series representation of the
signal is obtained by scaling the series expansion coefficients by exp(jn0)
• Unit impulse function is defined as,
or
• Thus,
• Let us consider a periodic train of impulses,
• Fourier coefficient of an impulse,
Differentiation property
• Consider the signal,
• Taking derivation w.r.t time, we get
• Taking ith derivative w.r.t time we get,
• The Fourier coeff of the ith derivative is related to
Fourier coeff of f(t) by,
Assignment
• Calculate the Fourier series coeff. of rectangular
pulse train:
Unit impulse representation of pulse train
• Cn of pulse train can be computed in terms of Cn1 i.e.
derivative of f(x),
• Using, we have
Trapezoidal waveforms
• Finite rise and fall times, pulse width,
• Cn of x(t) can be represented in terms of its
derivatives Cn(2)
If then,
Upon simplifying,
Using property,
Spectral content and EMI sources
Envelope of trapezoidal waveform
Envelope:
Slope of 0dB/decade for,
Second term is 20 dB/decade till f=1/pitau_r
Spectra of a-periodic signals
• Fourier transform,
where
• Frequency content of short pulse created due to ESD, an
EM pulse, glitch in IC, or lightning event can be
examined using Fourier transform, random data signals
• Satisfies linearity, superposition, differentiation, time
shifting, impulse function property
Eg. 1. Short rectangular pulse
Pulse width governs the spectrum
• Fourier transform,
E.g. 2. Gaussian Pulse
Let the short pulse be represented by, x(t)=e-at^2then X(f) is given by,
a=4
• Larger a ==> broader spectrum
Pulse spread governs the spectrum
E.g. 3. Random data signals
• Frequency content is represented by power spectral
density,
Rx(t): auto correlation function,
overbar indicates statistical average of the signal
• Let x(t) represent the random signal that transitions
between 0 and X0. Then, and m(t) is
random variable that varies between +1 and −1 with
equal prob. in bit interval: nT to (n+1)T, T: bit rate
• PSD of random data signal,
• Average power, Note:
T: bit rate governs the spectral
content of random data signals
like the rise time in periodic
signals
