GG 313 Lecture 26 11/29/05 Sampling Theorem Transfer Functions.

Slides:



Advertisements
Similar presentations
| Page Angelo Farina UNIPR | All Rights Reserved | Confidential Digital sound processing Convolution Digital Filters FFT.
Advertisements

All slides © S. J. Luck, except as indicated in the notes sections of individual slides Slides may be used for nonprofit educational purposes if this copyright.
Ch 3 Analysis and Transmission of Signals
Math Review with Matlab:
ECE 8443 – Pattern Recognition EE 3512 – Signals: Continuous and Discrete Objectives: Response to a Sinusoidal Input Frequency Analysis of an RC Circuit.
HILBERT TRANSFORM Fourier, Laplace, and z-transforms change from the time-domain representation of a signal to the frequency-domain representation of the.
Lecture 7: Basis Functions & Fourier Series
GG 313 Lecture 25 Transform Pairs and Convolution 11/22/05.
Math Review with Matlab:
Leo Lam © Signals and Systems EE235. Transformers Leo Lam ©
SIMS-201 Characteristics of Audio Signals Sampling of Audio Signals Introduction to Audio Information.
Properties of continuous Fourier Transforms
Convolution and Fourier Transforms in Digital Signal Analysis
Department of EECS University of California, Berkeley EECS 105 Fall 2003, Lecture 2 Lecture 2: Transfer Functions Prof. Niknejad.
Discrete-Time Convolution Linear Systems and Signals Lecture 8 Spring 2008.
FILTERING GG313 Lecture 27 December 1, A FILTER is a device or function that allows certain material to pass through it while not others. In electronics.
Leo Lam © Signals and Systems EE235. Transformers Leo Lam ©
Leo Lam © Signals and Systems EE235 Lecture 27.
CELLULAR COMMUNICATIONS DSP Intro. Signals: quantization and sampling.
The Nyquist–Shannon Sampling Theorem. Impulse Train  Impulse Train (also known as "Dirac comb") is an infinite series of delta functions with a period.
Content Op-amp Application Introduction Inverting Amplifier
Leo Lam © Signals and Systems EE235. Leo Lam © Fourier Transform Q: What did the Fourier transform of the arbitrary signal say to.
Chapter 25 Nonsinusoidal Waveforms. 2 Waveforms Used in electronics except for sinusoidal Any periodic waveform may be expressed as –Sum of a series of.
Experiment 17 A Differentiator Circuit
Digital Signals and Systems
Filters and the Bode Plot
EE513 Audio Signals and Systems Digital Signal Processing (Systems) Kevin D. Donohue Electrical and Computer Engineering University of Kentucky.
EE421, Fall 1998 Michigan Technological University Timothy J. Schulz 08-Sept, 98EE421, Lecture 11 Digital Signal Processing (DSP) Systems l Digital processing.
Vibrationdata 1 Unit 5 The Fourier Transform. Vibrationdata 2 Courtesy of Professor Alan M. Nathan, University of Illinois at Urbana-Champaign.
DTFT And Fourier Transform
1 Let g(t) be periodic; period = T o. Fundamental frequency = f o = 1/ T o Hz or  o = 2  / T o rad/sec. Harmonics =n f o, n =2,3 4,... Trigonometric.
Chapter 2. Signals Husheng Li The University of Tennessee.
Module 2 SPECTRAL ANALYSIS OF COMMUNICATION SIGNAL.
1 BIEN425 – Lecture 10 By the end of the lecture, you should be able to: –Describe the reason and remedy of DFT leakage –Design and implement FIR filters.
Signals CY2G2/SE2A2 Information Theory and Signals Aims: To discuss further concepts in information theory and to introduce signal theory. Outcomes:
Chapter 5: Fourier Transform.
Vibrationdata 1 Unit 5 The Fourier Transform. Vibrationdata 2 Courtesy of Professor Alan M. Nathan, University of Illinois at Urbana-Champaign.
EE104: Lecture 5 Outline Review of Last Lecture Introduction to Fourier Transforms Fourier Transform from Fourier Series Fourier Transform Pair and Signal.
11/22/20151 Digital Filters. 11/22/20152 Time domain to Frequency Domain and vice versa Filter Characteristics.
Leo Lam © Signals and Systems EE235 Leo Lam.
Vibrationdata 1 Unit 6a The Fourier Transform. Vibrationdata 2 Courtesy of Professor Alan M. Nathan, University of Illinois at Urbana-Champaign.
GG313 Lecture 24 11/17/05 Power Spectrum, Phase Spectrum, and Aliasing.
Lecture 7: First Order Filter and Bode Plot
2D Sampling Goal: Represent a 2D function by a finite set of points.
Chapter 2. Fourier Representation of Signals and Systems
Geology 6600/7600 Signal Analysis 21 Sep 2015 © A.R. Lowry 2015 Last time: The Cross-Power Spectrum relating two random processes x and y is given by:
ELECTRICAL ENGINEERING: PRINCIPLES AND APPLICATIONS, Fourth Edition, by Allan R. Hambley, ©2008 Pearson Education, Inc. Lecture 17 Fourier Analysis, Low.
2D Fourier Transform.
Eeng360 1 Chapter 2 Linear Systems Topics:  Review of Linear Systems Linear Time-Invariant Systems Impulse Response Transfer Functions Distortionless.
Leo Lam © Signals and Systems EE235 Lecture 26.
Learning from the Past, Looking to the Future James R. (Jim) Beaty, PhD - NASA Langley Research Center Vehicle Analysis Branch, Systems Analysis & Concepts.
(4) Filters.
Maj Jeffrey Falkinburg Room 2E46E
Digital Signal Processing Lecture 4 DTFT
Unit 5 The Fourier Transform.
Filters and the Bode Plot
Sampling and Quantization
Computational Data Analysis
Signals and Systems EE235 Lecture 26 Leo Lam ©
EE Audio Signals and Systems
ECET 345 Competitive Success/snaptutorial.com
ECET 345 Education for Service-- snaptutorial.com.
ECET 345 Teaching Effectively-- snaptutorial.com.
Notes Assignments Tutorial problems
HKN ECE 210 Exam 3 Review Session
Coherence spectrum (coherency squared)
LECTURE 18: FOURIER ANALYSIS OF CT SYSTEMS
8.3 Frequency- Dependent Impedance
Signals and Systems Revision Lecture 2
Chapter 3 Sampling.
Presentation transcript:

GG 313 Lecture 26 11/29/05 Sampling Theorem Transfer Functions

Homework review: Data are 128 points, 20 samples/sec. What is the time vector that matches the data? Starting at t=0, we want a new point every 1/20=.05 s time=[0:deltat:6.35]; % The last point is at 127*.05 sec. How about the frequency vector? Our sampling frequency is 20 Hz, so we expect the frequency scale to be symmetric about the middle and for it to go from 0 to the sampling frequency in 128 steps: freq=[0:1/(N*deltat):1/deltat*(1-1/N)];

Now generate and plot the spectrum: Y=fft(y); p=Y.*conj(Y); plot(freq,p) Note the symmetry about the Nyquist (10 Hz) and peaks at 1 and 4 Hz. We actually only need to plot the 1st half.

Now we apply a simple filter to the data by convolution: h=[ ]; out=conv(h,y); Now calculate the power spectrum of “out” and compare it to the input to the filter “y”. OUT=fft(out(2:129)); OUTP=OUT.*conj(OUT); plot(freq,OUTP,'r') But, let’s only plot from zero to the Nyquist: plot(freq(1:65), OUTP(1:65),'r')

The red (filtered) spectrum has lost energy at the low frequencies (attenuated) and gained (amplified) at the high frequencies relative to the input signal. The peak at 1 Hz has been effectively removed.

Now try the other filter: hlp=[ ]; outlp=conv(hlp,y); OUTLP=fft(outlp(2:129)); OUTLPP=OUTLP.*conj(OUTLP); plot(freq(1:65),OUTLPP,'k') In this case, both peaks are present, but the high frequencies have been attenuated relative to the input signal.

Let’s plot the same data in dB (10*log10(p)): blue: input red: [ ] black: [ ] Note that the peaks that look like they had been completely wiped out are actually there - just much smaller. For example the peal at 1 Hz is reduce by ~20 dB by the “black” filter - that’s about a factor of 10 in amplitude.

This is the comb function made up of an infinite string of delta functions separated by T. It’s Fourier Transform is also an infinite string of delta function separated by frequency 1/T: The sampling theorem The sampling theorem says that if all the energy in a continuous signal is below the Nyquist frequency, then all the information in that signal is captured in the sampling process. We can show this by using our last transform pair: (5.117)

Consider what happens to a continuous analog signal in the time domain if we MULTIPLY it (NOT convolve) with a comb function. What we get is a SAMPLE of that function every T. Where T is the sample period.

Now we need to look at the Fourier Transform of each of these functions: The greens are positive frequencies and oranges negative. As long as the signal is BAND LIMITED to frequencies below the Nyquist (1/2T), ALL of the information is preserved after sampling.

No let’s see what happens if the signal is NOT band limited: In this case, the signal contains high frequency energy above the Nyquist. This energy shows up in the 0-Nyquist frequency range as a peak of negative frequency energy. The arrow shows where the peak originated.

Transfer Functions Analysis of time series data often involves determination of how the equipment and processing steps affect the signal as it passes through various steps. For example, a particular sound source will have a particular spectrum, and it will be detected by hydrophones that change the spectrum of the signal in a particular way. These functions must be determined and removed if the characteristics of the “earth filter” are to be determined. There are many ways to determine these TRANSFER FUNCTIONS. For example, we can make a simple filter from a resistor and a capacitor:

This circuit has a well-known transfer function: Notice that this function is already in the frequency domain, and we can plot it using Matlab:

The slope is -6 dB/octave, and the corner frequency (3dB down point) is 1/(2πRC), and the phase response is  =tan -1 (1/(  RC)

One way to define the transfer function of a system that is not well known is to use an impulse for an input signal. Recall that an impulse convolved with any function yields the function itself. So the output of a system when an impulse is the input is called the IMPULSE RESPONSE. The Fourier transform of the impulse response yields the transfer function of the system. Output/Input=Transfer function=FFT(Impulse response) The impulse response of a Linear Time-Invariant (LTI) system yields all the information about transfer properties that the system performs.

Consider the following function: out(i)=out(i-1)+in(i), where out(i) is the output at time (i) and in(i) is the input. USE MATLAB to: 1)get the impulse response of this function 2)find its response to a sine wave in(i)=sin((ii-1)*8*pi*(f/(n*dt))); f=8; Notice several things: The impulse response of the function is a step (=0 if ii =0) What kind of function is this? How does the output change as f changes for the sin wave input?

Recall the integral of a sine wave: This says that the output of an integrator should decrease in amplitude by a factor of 2 with each doubling in frequency. That is, an integrator should modify the spectrum by -6dB/octave. Thus, to integrate a signal, all you need to do it to move it to the frequency domain, subtract 6 dB/octave (with 0 dB at  =1) and return to the time domain. Similarly, to differentiate a signal, ADD 6dB/octave in the frequency domain.

Consider a “perfect” filter in the frequency domain, such a filter passes all energy below a particular frequency and none above. This is a “boxcar”. This is one of our transform pairs, and we know the function in the time domain will be a sinc function: y=sin(x)/x, y=1| x=0.