Transcript Slide 1

Intro to Sinusoids
• What is a sinusoid?
» Mathematical equation
x(t )  A cos(2ft   )
t
x(t ) : Function of the time variable
A : Amplitude
f : Frequency (# cycles per sec, Hertz)

: Phase
Ira Fulton School of Engineering
SINUSOIDAL SIGNAL
Acos( t  )
• FREQUENCY

– Radians/sec
– Hertz (cycles/sec)
 (2 ) f
• AMPLITUDE
– Magnitude
• PERIOD (in sec)
1 2
T 
f

• PHASE
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering

A
Intro to Sinusoids
• What is a sinusoid?
» Plot
Ira Fulton School of Engineering
Intro to Sinusoids
• Examples
Ira Fulton School of Engineering
Intro to Sinusoids
• Examples
Ira Fulton School of Engineering
Intro to Sinusoids
• Examples
Ira Fulton School of Engineering
Intro to Sinusoids
• How to plot a sinusoid?
» Determine height
» Determine zero crossings
• Max height is +A
» peaks oscillate: +A and -A
» peaks occur when argument of cosine is a
multiple of  , i.e., 0,  ,2 ,3 ,...
Ira Fulton School of Engineering
Intro to Sinusoids
• Zero crossings
» where plot crosses zero
» located at odd multiples of
 / 2,3 / 2,...
 /2
, i.e.,
• Note: cycles repeat (periodic function)
» sufficient to determine peak and zero
crossings in one cycle and then replicate
cycle
» Period = length of cycle = T  1 / f
Ira Fulton School of Engineering
Intro to Sinusoids
• We can also get equation from plot
» find the parameters A, f , and 
Ira Fulton School of Engineering
Intro to Sinusoids
• We can also get equation from plot
» find the parameters A ,
f, and 
Answer:
Ira Fulton School of Engineering
A  15, f  1000Hz,   0
Intro to Sinusoids
• Procedure to get equation from plot
• Find the amplitude A
• Find the period T by measuring the time
distance between positive peaks: f  1 / T
• Find the phase  by measuringt the “time
1
shift” (delay from 0) t 1 of a peak. Note: t 1
positive if peak moved to right, and
negative if peak moved to left.
• Then, compute   2ft1
Ira Fulton School of Engineering
Intro to Sinusoids
• Sinusoids can be interpreted as a
“rotating phasor”
x(t )  cos( 2ft   )



•  (t ) rotating angle
 (t )
» f : speed of rotation (# cycles per
sec)
»  : initial angle start point
Ira Fulton School of Engineering
Generating Signals using Sinusoids
• Main reason why sinusoids are
important: they are basic tools for
constructing other useful and more
complicated signals
• A new signal can be generated by
combining together different sinusoids
x(t )  A0  A1 cos( 2f1t  1 ) 
A2 cos( 2f 2 t   2 )  ...  AN cos( 2f N t   N )
Ira Fulton School of Engineering
Generating Signals using Sinusoids
• Example: Beat Note Waveform
• Beat notes generated by adding two
sinusoids with nearby frequencies
• They can also be generated by playing two
neighboring piano keys
• Mathematically:
x(t )  A cos(2f1t )  A cos(2f 2 t )
where f1  f 0  d ; f 2  f 0  d ; d small.
Ira Fulton School of Engineering
Generating Signals using Sinusoids
• Example: Beat Note Waveform
• Let
f o  200Hz ; d  20Hz  f1  220Hz ; f 2  180Hz
• Matlab Script File: beatnote.m
dur = 2.0;
fs = 8000;
f1 = 220; f2 = 180;
t = 0: (1/fs): dur;
x = cos(2*pi*f1*t)+cos(2*pi*f2*t);
plot(t,x); xlabel(‘Time,t’); ylabel(‘Beat Signal’);
sound(x,fs);
Ira Fulton School of Engineering
Generating Signals using Sinusoids
• Example: Beat Note Waveform
• Let
f o  200Hz ; d  20Hz  f1  220Hz ; f 2  180Hz
• Matlab Function: beat.m
function [x,t] = beat(f0, d, dur);
fs = 8000;
f1 = f0+d; f2 = f0-d;
t = 0: (1/fs): dur;
x = cos(2*pi*f1*t)+cos(2*pi*f2*t);
Ira Fulton School of Engineering
Matlab Functions
• Example: Beat Note Waveform
• Let
f o  200Hz ; d  20Hz  f1  220Hz ; f 2  180Hz
• Matlab Script File using Function beat.m
[x,t] = beat(200,20,1);
plot(t,x); xlabel(‘Time,t’); ylabel(‘Beat Signal’);
sound(x,fs);
Ira Fulton School of Engineering
Generating Signals using Sinusoids
• A new signal can be generated by combining
together different sinusoids
x(t )  A0  A1 cos( 2f1t  1 ) 
A2 cos( 2f 2 t   2 )  ...  AN cos( 2f N t   N )
• Periodic signals obtained when frequencies
are integer multiples of f 0 :
f k  kf0
where f k is called harmonic frequency of f 0
and f 0 is called the fundamental
frequency.
Ira Fulton School of Engineering
Generating Signals using Sinusoids
• A new periodic signal can be generated
by combining together sinusoids having
harmonically related frequencies
x(t )  A0  A1 cos( 2f 0 t  1 ) 
A2 cos( 2 2 f 0 t   2 )  ... 
AN cos( 2Nf 0 t   N )
• Period = length of one cycle = 1 / f 0
Ira Fulton School of Engineering
PERIODIC SIGNALS
• Repeat every T secs
– Definition
– Example:
x (t )  x ( t  T )
x(t )  cos (3t )
2
T ?
T 
– Speech can be “quasi-periodic”
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
2
3
T  3
Generating Signals using Sinusoids
• Example: Square Wave
N
8
x(t )  
cos( 2 (2m  1) f 0 t   / 2)
m  0 ( 2m  1)
• For N =1:
8
8
x(t )  cos( 2f o t   / 2)  cos( 2 3 f o t   / 2)

3
Ira Fulton School of Engineering
Generating Signals using Sinusoids
Square Wave generates using only first 3 harmonic
Frequencies:
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
T=0.1
Generating Signals using Sinusoids
• Example: Square Wave
• Matlab Function square.m
function [x,t] = square(f0, N, dur);
fs = 8000;
t = 0:1/fs:dur;
x = zeros(1,length(t));
for m = 0:N
x = x + (8/((2*m+1)*pi))*cos(2*pi*(2*m+1)*f0*t- pi/2);
end
• Generate and plot square waves with
f 0 = 25Hz, dur = 0.12 sec, and N=1,2,3,25
Ira Fulton School of Engineering
Generating Signals using Sinusoids
• Example: Synthetic Vowel
f 0  100 Hz
A1  12,226, 1  1.508;
A4  29,416,  4  1.876; A5  48,836,  5  0.185;
A16  13,621, 16  1.449; A17  4,723, 17  0
• Generated signal approximates the
waveform produced by a man speaking
the sound “ah”.
Ira Fulton School of Engineering
Vowel Waveform (sum of all 5 components)
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
Generating Signals using Sinusoids
• Example: Synthetic Vowel
• Matlab Script File
dur = 1; f = 100;
fs = 8000;
t = 0:1/fs:dur;
x = 12226 * cos(2*pi*2*f*t + 1.508);
x = x + 29416 * cos(2*pi*4*f*t + 1.876);
x = x + 48836 * cos(2*pi*5*f*t - 0.185);
x = x + 13621 * cos(2*pi*16*f*t - 1.449);
x = x + 4723 * cos(2*pi*17*f*t);
plot(t,x); xlabel(‘Time, t’); ylabel(‘Vowel’)
sound(x,fs)
Ira Fulton School of Engineering
TUNING FORK EXAMPLE
•
•
•
•
“A” is at 440 Hertz (Hz)
Waveform is a SINUSOIDAL SIGNAL
Computer plot looks like a sine wave
This should be the mathematical
formula:
A cos(2 (440)t   )
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
TUNING FORK A-440 Waveform
T  8.15  5.85
 2.3 ms
Time (sec)
f  1/ T
 1000/ 2.3
 435 Hz
Ira Fulton School of Engineering
SPEECH EXAMPLE
• More complicated signal (BAT.WAV)
• Waveform x(t) is NOT a Sinusoid
• Theory will tell us
– x(t) is approximately a sum of sinusoids
– FOURIER ANALYSIS
• Break x(t) into its sinusoidal components
– Called the FREQUENCY SPECTRUM
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
Speech Signal: BAT
• Nearly Periodic in Vowel Region
– Period is (Approximately) T = 0.0065 sec
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
DIGITIZE the WAVEFORM
• x[n] is a SAMPLED SINUSOID
– A list of numbers stored in memory
• Sample at 11,025 samples per second
– Called the SAMPLING RATE of the A/D
– Time between samples is
• 1/11025 = 90.7 microsec
• Output via D/A hardware (at Fsamp)
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
STORING DIGITAL SOUND
• x[n] is a SAMPLED SINUSOID
– A list of numbers stored in memory
•
•
•
•
CD rate is 44,100 samples per second
16-bit samples
Stereo uses 2 channels
Number of bytes for 1 minute is
– 2 X (16/8) X 60 X 44100 = 10.584 Mbytes
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
SINES and COSINES
• Always use the COSINE FORM
A cos(t   )
• Sine is a special case:
sin( t )  cos( t  2 )

Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
Sinusoidal Synthesis
• Sinusoids with DIFFERENT Frequencies
– SYNTHESIZE by Adding Sinusoids
N
x (t )   Ak cos( 2 f k t   k )
k 1
• SPECTRUM Representation
– Graphical Form shows DIFFERENT Frequencies
• SPECTROGRAM Tool
– Shows how frequency content varies in function of
time
– MATLAB function is specgram.m
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
SPECTROGRAM EXAMPLE
• Two Constant Frequencies: Beats
cos(2 (660)t ) sin(2 (12)t )
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
Frequency is the vertical axis
Time-Varying Frequency Diagram
A-440
Time is the horizontal axis
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
Sinusoidal Synthesis: Motivation
• Synthesize Complicated Signals
– Musical Notes
• Piano uses 3 strings for many notes
• Chords: play several notes simultaneously
– Human Speech
• Vowels have dominant frequencies
• Application: computer generated speech
– Can all signals be generated this way?
• Sum of sinusoids?
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
Fur Elise WAVEFORM
Beat
Notes
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
SIMPLE TEST SIGNAL
• C-major SCALE: stepped frequencies
– Frequency is constant for each note
IDEAL
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
SPECTROGRAM of C-Scale
Sinusoids ONLY
From SPECGRAM
ANALYSIS PROGRAM
ARTIFACTS at Transitions
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
Spectrogram of LAB SONG
Sinusoids ONLY
Analysis Frame = 40ms
ARTIFACTS at Transitions
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
Time-Varying Frequency
• Frequency can change vs. time
– Continuously, not stepped
• FREQUENCY MODULATION (FM)
x(t )  cos(2 fc t  v(t ))
• CHIRP SIGNALS
– Linear Frequency Modulation (LFM)
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
VOICE
New Signal: Linear FM
• Called Chirp Signals (LFM)
QUADRATIC
– Quadratic phase
x(t )  A cos( t  2 f0 t   )
2
• Freq will change LINEARLY versus time
– Example of Frequency Modulation (FM)
– Define “instantaneous frequency”
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
INSTANTANEOUS FREQ
• Definition
x(t )  A cos( (t ))
d  (t )
 i (t )  dt
Derivative
of the “Angle”
• For Sinusoid:
x(t )  A cos(2 f 0t   )
 (t )  2 f 0t  
 i ( t ) 
d  (t )
dt
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
 2 f 0
Makes sense
INSTANTANEOUS FREQ of the Chirp
• Chirp Signals have Quadratic phase
• Freq will change LINEARLY versus time
x (t )  A cos( t   t   )
2
  (t )   t   t  
2
 i (t ) 
d  (t )
dt
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
 2 t  
CHIRP SPECTROGRAM
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
CHIRP WAVEFORM
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
OTHER CHIRPS
 (t) can be anything:
x(t )  A cos( cos( t )   )
 (t) could be speech or music:
– FM radio broadcast
 i ( t ) 
d  (t )
dt
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
  sin(  t )
SINE-WAVE FREQUENCY MODULATION (FM)
Courtesy of James McClellan et al, Signal Processing First, 2003
Ira Fulton School of Engineering
Music Synthesis
• Musical notes synthesized using a sinewave at
a given frequency.
• Musical scale is divided into 8 octaves; each
octave consists of 12 notes.
• Notes in each octave are related to notes in
previous and next octave:
– The frequency of a note is twice the frequency of
the corresponding note in the previous adjacent
(lower) octave.
• Each octave contains 12 notes (5 black keys
and 7 white) and the ratio between the
frequencies of the notes is constant between
successive notes: fnext note = 21/12 fprevious note
Ira Fulton School of Engineering
Music - Octave
•D4 denotes D note in 4th octave
•A4 or A-440 note (tone at 440 Hz) is usually a reference note
called middle A.
Ira Fulton School of Engineering
Music - Octave
Ira Fulton School of Engineering
Music - Notation
4
• Musical notation shows which notes are to be played
and their relative timing
Ira Fulton School of Engineering
Sinewave Synthesis
• Methods used
– Recursive
– Taylor’s Series
– Look-up Table
Ira Fulton School of Engineering
Recursive Method
sin( A  B)  sin A cos B  cos A sin B
cos(A  B)  cos A cos B  sin A sin B
Let A =  and B =n where  is the angle and n the index
sin(  n )  sin( [1  n])  sin  cosn  cos sin n
cos(  n )  cos( [1  n])  cos cosn  sin  sin n
For  = 1 and n= 1
sin(1[1  1])  sin(1) cos(1)  cos(1) sin(1)
cos(1[1  1])  cos(1) cos(1)  sin(1) sin(1)
Sin(2) and cos(2) can be computed using only four multiplications and 2
addition/subtraction
Sin(1) and cos(1) must first be precomputed and stored in memory
Ira Fulton School of Engineering
Recursive Method – Practical Example
• Generate a sine wave of frequency F = 100 Hz
having a sample frequency Fs equals to 8000
samples/s.
• Calculate the angle increment 
– Total number of samples/period = Fs/F = 80
– The angle increment  = 360/80 = 4.5
• Precompute sin(4.5) and cos(4.5)
• Apply the recursive method
Ira Fulton School of Engineering
Taylor’s Series
Ira Fulton School of Engineering
Look-up Table
• Calculate the values of L evenly spaced points on a
sinusoid and store them in memory
Lookup Values for L = 16
1
• Suppose L = 16 and Fs =
16000 samples/s
• The total number of cycles
obtained per second
is:16000/16 = 1000 Hz
0.8
0.6
Amplitude
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
0
5
10
Sample
Base Address
0
Base Address + 1
sin(2/L)
Base Address + 2
sin(4/L)
.
.
.
Base Address + L – 1
Ira Fulton School of Engineering
.
.
.
sin(2(L-1)/L)
15
– The frequency calculated is
called fundamental frequency
given by:
f fund 
1
1 F
x  S ( Hz)
L
L sam ples T
Look-up Table (Cont’d)
• If the samples are not read sequentially from
the look-up table (e.g., skipping one sample),
the frequency generated will be:
1
1 FS
f  '
x  ' ( Hz)
L sam ples T L
Example:
Suppose Fs = 16 Khz, L = 16, and  =
2:
f sin 

2 16000
FS 
 2000 Hz
L
16
Ira Fulton School of Engineering
where:
L
L

'
and  is the
lookup table
increment index.
The maximum value 
can reach is bounded by
the Nyquist rate:
= L/2 => fsin= Fs/2;
Look-up Table (Cont’d)
• Frequency synthesized for different values of 
(L = 16)
 Frequency Generated (Hz)
What if  not integer?
Ira Fulton School of Engineering
1
1000
2
2000
3
3000
4
4000
5
5000
6
6000
7
7000
8
8000
Maximum Frequency
Look-up Table – Non Integer 
• Two Methods
– Round – down
– Interpolation
Ira Fulton School of Engineering
Look-up Table – Round Down Method
1.5
“+”: Look-up Table
“o”: Desired Output
1
L=8
0.5
 = 2.5
0
-0.5
-1
-1.5
0
1
2
Ira Fulton School of Engineering
3
4
5
6
7
8
High Distortion!!
Look-up Table – Interpolation
Linear interpolation method used
y = mx + b
where:
m = line slope given by (Base_address + l +1) – (Base_address + l )
Base_address is the beginning address of the lookup table.
b = (Base_address + l ).
x = fractional part of the pointer with 0<x<1
1.5
y = linear approximated output sample.
(2,1)
Note that () denotes indirect addressing.
Interpolation
0.5.(0.7071-1)+1 =
0.8536
1
(3,0.7071)
0.5
0
-0.5
-1
-1.5
Ira Fulton School of Engineering
0
1
2
3
4
5
6
7
8
Methods - Comparison
Look-up Table
Taylor’s Series
Recursive Method
Memory Space
Yes
No
No
Speed
Fast
Slow
Slow
Complexity
Low
High
Moderate
Accuracy
Moderate
High
High
Ira Fulton School of Engineering
Amplitude Modulation of Tones - ADSR envelope
• Smoothing may be needed since the musical note will
not reach its full intensity instantaneously
• Attack – Decay – Sustain- Release (ADSR) envelope
- Attack time: time during which the musical note
reaches its peak
- Decay time: time required for the tone’s intensity to
partially die away
- Sustain time: time where the tone’s intensity remains
unchanged
- Release time: time for final attenuation
Ira Fulton School of Engineering
ADSR - Examples
Ira Fulton School of Engineering
ADSR Generation
- One segment at a time (A – D – S – R)
- Approximated as rising/decaying exponentials
Ira Fulton School of Engineering
ADSR Generation – cont’d
- If Fs = 16000 samples/s and assuming a max
note duration of 1 second, we need to store 16K
in memory for the ADSR envelope values, which
consumes a lot of memory
- An alternative to the lookup ADSR is to try to
compute the values “on the fly” using the
equation:
y[n]  g.x(n)  (1^  g ).y[n  1]
where g is the rise/decay rate and x(n) is the desired
target value; initially x(-1)= 0.
Ira Fulton School of Engineering
ADSR Generation – cont’d
- The equation presented previously is a firstorder IIR filter given by the following difference
equation:
y[n]  g.x[n]  (1  g ).y[n  1]
x[n]
g
y[n]

z-1
1-g
Ira Fulton School of Engineering
ADSR Generation – cont’d
Why???
IIR Filter
Solve for the equation and show that the
output have an exponential shape
Ira Fulton School of Engineering