2 Introduction Signals ADCDAC

Download as pdf or txt
Download as pdf or txt
You are on page 1of 36

Dept.

of Electrical and Electronic Engineering


EEE-3203 : Digital Signal Processing
3rd Year (2019) 2nd Semester
Topic: Introduction to Signals and ADC/DAC

Dr. Abul Kalam Azad


Email: [email protected]
Telephone: 01707076731
Office: Room # 217
Discrete-time (D-T) Signal
 A D-T signal is an indexed sequence of real or complex numbers.

 A D-T signal x(n) is a function of an integer-valued variable, n.


 A D-T signal is undefined for non-integer values of n.
 It is convenient to view x(n) as a vector. The sequence values x(0) to x(N - 1)
of x(n) may be considered to be the elements of a column vector

x   x(0), x(1)......., x( N  1) 
T

AKA_DU_EEE3203_DSP 2
Discrete-time (D-T) Signal
 D-T signals are often derived by sampling a continuous-time signal (such as
speech) with an analog- to-digital (A/D) converter.
 Sampling of a continuous-time signal xa(t) at a rate of Fs = l/Ts samples per
second produces the D-T signal x(n), where
x(n)  xa (nTs )
 In general, a D-T signal may be complex-valued. A complex signal may be
expressed in terms of its real and imaginary parts
x(n)  x1 (n)  jx2 (n)  Re{x(n)}  j Im{x(n)}
 In polar form in terms of magnitude and phase x ( n)  x ( n) exp  j arg{x( n)}
1  Im{ x ( n)} 
o Magnitude: x(n)  Re 2 {x(n)}  Im 2 {x( n)} and Phase: arg{x(n)}  tan  
 Re{x(n)} 
 The complex conjugate of x(n)
x (n)  Re{x(n)}  j Im{x(n)}  x(n) exp[  j arg{z (n)}]
*

AKA_DU_EEE3203_DSP 3
Elementary D-T Signals
 Unit sample sequence δ(n): is defined as
1 n  0
 ( n)  
0 n  0
 Unit step sequence u(n): is defined as
1 n  0
u ( n)  
0 n  0
 Ramp sequence r(n): is defined as
n n  0
r ( n)    nu[n]
0 n  0
 Properties of elementary signals:
(i )  [n]   [n] (ii ) x[n] [n]  x[0] [n] (iii ) x[n] [n  k ]  x[k ] [n  k ]
 
(iv)  [n]  u[n]  u[n  1] (v) u[n]    [n  k ]
k 0
(vi ) x[n]   x(k ) [n  k ]
k 

AKA_DU_EEE3203_DSP 4
Presentation of D-T Signals
 Graphical presentation:  Functional presentation:
1 for n  1
2 for n  2

x[n]  
3 for n  3, 4
0 elsewhere

 Tabular presentation  Sequence presentation

x[n] ... 0 0 1 2 3 3 0 0 ...



x[n]  ... 0 0 1 2 3 3 0 0 ..


n ... 1 0 1 2 3 4 5 6 ... or
x[n]  0 1 2 3 3 0 0 ..............

AKA_DU_EEE3203_DSP 5
Deterministic and Random D-T signals
 Deterministic signals
o Values of such D-T signals are completely specified for any given integer n
o Present, past and future values of the signals are known precisely

 Random signals
o Can take random values for any given integer
o Examples: Output of a noise generator, speech signal

AKA_DU_EEE3203_DSP 6
Even and Odd D-T signals
 A D-T signal x(n) is even signal if x( n)  x(n) for all n

 A signal x(n) is add signal if x(n)   x(n) for all n

AKA_DU_EEE3203_DSP 7
Periodic and Non-periodic signals
 x(n) is periodic with period N if x(n  N )  x(n) for all n
o N is a positive integer
o Fundamental Period: Smallest N for which above equations hold.

Fundamental period, N = 4
 Ifx(n  N )  x(n) for all n , x(n) is nonperiodic
  2 2
 Let x ( n)  cos n    N  6 (integer ) Note:  
3 3  N
 x(n) is a periodic signal with fundamental period 6
AKA_DU_EEE3203_DSP 8
Periodic and Non-periodic signals
 
j  n  2 2
 Let x(n)  e 4
  N   8 (integer ) Note:  
4  N
Thus, x(n) is periodic with fundamental period 8

 Periodicity of sum of two periodic sequences


[Let x1[n] and x2[n] are periodic sequences with fundamental periods N1, and N2,
respectively. Under what conditions is the sum x[n] =x1[n] +x2[n] periodic, and what
is the fundamental period of x[n] if it is periodic]
Comments:
Since x1[n] and x2[n] be periodic sequences with fundamental periods N1, and N2
 x1 (n)  x1 (n  N1 )  x1 ( n  mN1 ), m = positive integer
 x2 (n)  x2 (n  N 2 )  x2 ( n  kN 2 ), k = positive integer
 x(n)  x1 (n  mN1 )  x2 (n  kN 2 )
AKA_DU_EEE3203_DSP 9
Periodicity of sum of two periodic sequences
 In order for x(n) to be periodic with period N
x(n  N )  x1 (n  N )  x2 (n  N )  x1 (n  mN1 )  x2 (n  kN 2 )
 Thus, the following relationship must exists
N1 k
N  mN1  kN 2  
N2 m
 Since, k and m are integers, N1/N2 must be a rational number and N will
be the least common multiple of N1 and N2.
   
 Example: x ( n)  cos n  sin n  1  and  2 
3 3 4 4
2 2 N1 6 3
 N1   6 and N 2   8 Now,   (rational number)
1 2 N2 8 4
 Thus x(n) is periodic with fundamental period N = 4N1=3N2 = 24
AKA_DU_EEE3203_DSP 10
Energy and power signals
 The normalized energy E of x(n) is defined as

E 
2
x ( n) Joule
n 
 The average power of x(n) is defined as
N
1
P  lim 
2
x(n) Watt
N  2 N  1
n  N
N
1
o If x(n) is periodic, P  
2
x ( n ) Watt
N  1 n  N
Notes:
 Non periodic signals are energy signals while power signals are periodic
 If E is finite (i.e., 0 <E< ∞), x(n) is called energy signal. Thus, P = 0
 if P is finite (i.e., 0 <P< ∞), x(n) is called power signal. Thus, E = ∞
AKA_DU_EEE3203_DSP 11
Energy and power signals: Example
 x(n) = u(n)
1 N
1 11 N 1
P  lim  u (n)  lim ( N  1)   
2

N  2 N  1 N  2 N  1 2 1 N 2
n  N
Thus, x(n) is a power signal (Its energy is infinite)
 x(n) = (- 0.5)nu(n)
  
n 2 1 4
E  x(n)    0.5     0.25    
2 n

n  n 0 n 0 1  0.25 3
Thus, x(n) is a energy signal (Its power is zero)
 x(n) = x(n) = 
2ej3n x ( n )  2 e j 3n
 2 e j 3n
2
2 4(2 N  1)
N N
1 1
P  lim  x(n)  lim  4
2
2
N  2 N  1 N  2 N  1 2N 1
n  N n  N

Thus, x(n) is power signal AKA_DU_EEE3203_DSP 12


Manipulation of signals: Index manipulation
 Shifting: x(n – k) is the shifted version of x(n) by k samples
o to the right if k is positive. This is referred to as a delay, and
o to the left if k is negative. This is referred to as an advance.

 Folding / Reversal: Involves flipping the signal x(n) with respect to the index n.

AKA_DU_EEE3203_DSP 13
Manipulation of signals: Index manipulation
 Time scaling: Time scaled version of x(n) is defined by y (n)  x(kn)
o where k is an integer
o If k > 1, y(n) is the compressed (down-sampled) version of x(n)
o If 0 < k < 1, y(n) is the expanded (up-sampled) version of x(n)

 Shifting, Folding and Time scaling operations are order-dependent.


 Order of operation: Shifting → Reversal → scaling
AKA_DU_EEE3203_DSP 14
Manipulation of signals
 Order of index manipulation
o Let FD denotes folding operation and TD denotes time-delay operation
o Delaying x(n) by k samples of gives: TDk [ x(n)]  x(n  k ) k  0
o Folding x(n) gives: FD[ x(n)]  x(n)
o Now: TDk {FD[ x(n)]}  TDk  x(n)  x(n  k )
o Whereas: FD{TDk [ x(n)]}  FD  x(n  k )  x(n  k )
o So, TDk {FD[ x(n)]}  FD{TDk [ x(n)]}
o Thus, the operation of Folding and Shifting of a signal are not commutative.
 Amplitude manipulation
o Addition: y(n)  x1 (n)  x 2 (n) Amplitude manipulation are
o Multiplication: y(n)  x1 (n) x 2 (n) sample-to-sample basis operations
o Amplitude scaling: y(n)  cx(n)
AKA_DU_EEE3203_DSP 15
Concept of frequency in Continuous & Discrete-time Signals
 Continuous-time sinusoidal signals: A continuous-time sinusoidal signal can
be expressed as
xa (t )  A cos(t   )
Here, subscript, a denotes analog signal, A = amplitude of the sinusoid
Ω = frequency in rad/second = 2πF, F = frequency in cycles/second (Hertz)
θ = phase in radians
 In terms of F (i.e., Ω = 2πF), xa(t) yields
xa (t )  A cos(2 Ft   )

AKA_DU_EEE3203_DSP 16
Concept of frequency in Continuous & Discrete-time Signals
 Discrete-time sinusoidal signals: A discrete-time sinusoidal signal can be
expressed as
x(n)  A cos( n   )
Here, n is a integer variable, called the sample number
A = amplitude of the sinusoid
ω = frequency in radians/sample = 2πf, f = frequency in cycles/sample (Hertz)
θ = phase in radians
 In terms of f (i.e., ω = 2πf), x(n) becomes
x(n)  A cos(2 fn   )

AKA_DU_EEE3203_DSP 17
Identical Discrete-time Signals
 Discrete-time sinusoidal signals: Let us consider the sinusoid
x(n)  cos(0 n   )
 Now, cos  (0  2 ) n     cos(0 n  2 n   )  cos(0 n   )
 In general, the sinusoidal sequences
xk (n)  cos(k n   ), k  0,1, 2...
where
 k  0  2 k  ,     0  

are indistinguishable (i.e., identical)

AKA_DU_EEE3203_DSP 18
Analog-to-Digital conversion
 Most of the real-world signals (e.g., speech, biomedical, seismic, radar etc.)
are analog in nature.
 Analog-to-digital conversion To process analog signals by digital means
o To process analog signals by DSP, It is necessary to convert them into digital form
o This is done by converting them to a sequence of numbers having finite precision
o The procedure is called analog-to-digital (A/D) conversion
 A/D conversion is a three-step process

AKA_DU_EEE3203_DSP 19
Analog-to-Digital conversion
A/D conversion involves three steps
 Sampling:
o This is conversion of Continuous-Time signal into Discrete-Time signal by taking ‘samples’
at discrete-time instant.
o If xa(t) is the input to the sampler, the output is xa(nT) =x(n), where T is the sampling
interval.
 Quantization:
oThis is the conversion of Discrete-Time continuous-valued signal into a Discrete-Time
Discrete-valued (Digital) signal
o The value of each signal sample is presented by a value from a finite set of possible values
o The difference between the unquantized sample x(n) and the quantized output xq(n) is
called the quantization error
 Coding:
o In this process, each discrete value xq(n) is represented by a b-bit binary sequence
AKA_DU_EEE3203_DSP 20
Analog-to-Digital conversion
Three-steps in A/D conversion

AKA_DU_EEE3203_DSP 21
Sampling of analog signals
 Uniform sampling of an analog signal xa(t) is described by
xa (t )  xa (nT )  x(n),    n  
o x(n) is the discrete-time signal obtained by sampling xa(t) every T second, T = sampling period
1
o Sampling frequency (Hz) or sampling rate (samples/second): Fs 
n T
o For periodic sampling (T is constant): t  nT 
F s

 Let, C-T sinusoidal signal: xa (t )  A cos(2 Ft   )


1
 If xa (t ) is sampled periodically at a rate of Fs 
T
2 nF F
xa (nT )  x(n)  A cos(2 FnT   )  A cos(  )  f 
Fs Fs
‘f’ of a D-T signal can only be used to determined F if Fs is known
AKA_DU_EEE3203_DSP 22
Sampling of analog signals
 Consider an analog signal xa (t )  3cos100 t . What is the frequency of the signal
xa(t)? If the signal is sampled at Fs = 200 samples/second, determine the
discrete-time signal obtained after sampling. x (t )  A cos(2 Ft   )
a
Solution: Frequency of xa(t) is F = 50 Hz.
 n 100 
D-T signal obtained after sampling: x(n)  xa (nT )  xa    3cos n  cos n
 Fs  200 2
 Consider two signals both of which are sampled at a sampling rate of 40 Hz
x1 (t )  cos 20 t and x2 (t )  cos100 t
10  50 5
o The D-T signals are: x1 (n)  cos 2 n  cos n and x2 (n)  cos 2 n  cos n
40 2 40 2
5  n  
o Note that 2x ( n )  cos n  cos  2 n    cos n  x1 (n)
2  2  2
oThus x1(n) and x2(n) are identical, i.e., D/A gives ambiguous results [x1(t) or x2(t)]
o Thus, x2(n) is an alias of x1(n). In general all sinusoids cos2π(F1+Fsk)t, k=1, 2,…yields identical
values. Consequently, they are all aliases of F1.
AKA_DU_EEE3203_DSP 23
Sampling of analog signals
 n 
 Consider a D-T sequence x(n)  cos   . Find two different C-T signals that
 8 
would produce the same sequence when sampled at a rate of 10 kHz.
Solution: The frequency of the D-T signal is
  1
   2 f   f  samples/second
8 8 16
Thus, the frequency of the C-T time signal is F
1 f 
F  f  Fs   10  10  625 Hz
3
Fs
16
Now, any sinusoids with a frequency F+kFs, k=0, 1,2,3...will produce the same
sequence x(n) when sampled with a sampling frequency Fs.
Thus, such C-T signals are xk (t )  cos 2  F  kFs  t . For examples,
x0 (t )  cos 2  F  0  Fs  t  cos(1250 t )
x1 (t )  cos 2  F  1 Fs  t  cos 2 (625  10000)t  cos(21250 t )

AKA_DU_EEE3203_DSP 24
Sampling theorem
 If
o The highest frequency component of an analog signal xa (t ) is Fmax  B , and
o The signal is sampled at Fs  2 Fmax  2 B
 Then
o The signal xa (t ) can be exactly recovered from its sample values using the
interpolation function
sin 2 Bt
g (t ) 
2 Bt
 And  n 

n
o The signal xa (t ) can be expressed as: xa (t )   xa  F  g  t  F 
n   s  s 

 n 
Where xa    xa (nT )  x(n) are the samples of xa (t )
 Fs 

AKA_DU_EEE3203_DSP 25
Nyquist Rate (FN = 2Fmax = 2B)

 If the sampling of an analog signal is performed at Fs  2 Fmax  2 B , then the


sampling rate is called Nyquist rate, FN
When
 The sampling of xa (t ) is performed at Nyquist rate (i.e., FN  2 B ):
Then
 The reconstruction formula becomes
 n 
sin 2 B  t   sin 2 Bt

 n   2B  g (t ) 
xa (t )   xa   2 Bt
n   2 B  2 B  t  n 
 
 2B  
 n  n
xa (t )   xa   g  t  
n   Fs   Fs 

Thus, Nyquist rate FN is the minimum sampling rate required to avoid aliasing
AKA_DU_EEE3203_DSP 26
What happens when (Fs < FN)
 Aliasing occurs due to under-sampling
 Aliasing: Aliasing is an effect that causes different signals to become
indistinguishable (or alias of one another) when sampled using sampling rate
less than Nyquist rate FN.

 Two C-T signals in the lower figures are alias to each other.
AKA_DU_EEE3203_DSP 27
Why Fs > FN is practically used?
 Consider the signal xa (t )  10sin 300 t , sampled at sampling rate of FN = 300
o Note that the frequency F of xa(t) is 150 Hz, i.e., Fs = FN
o Sampling of xa(t) gives x(n), i.e.,
 n   300 
x(n)  xa (nT )  xa    10sin  n   10sin( n)  0
 Fs   300 
o Here the samples of the signal are taken at its zero-crossing point which misses
the signal completely.
 What to do?
o The signal can be offset in phase by some amount (θ). Thus,
10sin( n   )  10(sin  n cos  cos  n sin  )  10sin  cos  n  (1) n10sin 
o If θ ≠ 0 or π, the samples of the signal taken at FN are not all zero.
¤ However, it is not possible to obtain the correct amplitude of the signal when θ is unknown
o An simple remedy to avoid the troublesome is to sample the analog signal at a rate higher
that FN.

AKA_DU_EEE3203_DSP 28
Sampling of signals: Math Problems
 Example-1:Consider the analog signal xa (t )  3cos100 t
(a) Determine the minimum sampling rate to avoid aliasing.
xa (t )  A cos(2 Ft   ) , F = 50 Hz, so, FN  2  F  100 Hz
(b) If the signal is sampled at Fs = 200 Hz, determine the discrete-time signal obtained after
sampling  n 100 
x(n)  xa (nT ) xa    3cos n  cos n
 Fs  200 2
(c) If the signal is sampled at Fs = 75 Hz, Find the discrete-time signal obtained after sampling
 n  100 4  2  2
x(n)  xa (nT )  xa    3cos n  3cos n  cos  2   n  3cos n
 Fs  75 3  3  3
(d) Determine the frequency and the signal of the sinusoid that yields samples identical to those
obtained in part (c)
For part (c): Fs = 75 Hz, f = 1/3, F of the sinusoid = Fs ×f = 25 Hz, Sinusoid, ya (t )  3cos 50 t
(e) Comments on the results obtained in part (c) and (d)
Comment: For the sampling rate Fs = 75 Hz, F =50 Hz is an alias of F = 25 Hz

AKA_DU_EEE3203_DSP 29
Sampling of signals: Math Problems
Example-2:Consider the analog signal xa (t )  3cos 2000 t  5sin 6000 t  10cos12000 t
(a) What is the Nyquist rate for this signal?
F1  1000 Hz , F2  3000 Hz , F3  6000 Hz , Fmax  6000 Hz , FN  2  Fmax  12000 Hz
(b) If the signal is sampled at Fs = 500 Hz, determine the D-T signal obtained after sampling.
1 3 6 1  2  1
x(n)  3cos 2   n  5sin 2   n  10cos 2   n  3cos 2   n  5sin 2 1   n  10cos 2 1   n
5 5 5 5  5  5
1  2 1 1 2 1
 3cos 2   n  5sin 2    n  10cos 2   n  3cos 2   n  5sin 2   n  10cos 2   n
5  5 5 5 5 5
1 2
 13cos 2   n  5sin 2   n
5 5
(c) What will be the C-T signal ya(t) constructed from x(n) using ideal interpolation
Solution: Since only the frequency components at 1 kHz and 2 kHz are present in the sampled
signal, the analog signal that can be recovered is
ya (t )  13cos 2000 t  5sin 4000 t
AKA_DU_EEE3203_DSP 30
Quantization of continuous-valued signals
 Quantization:
oThis is the conversion of Discrete-Time continuous-valued signal into a Discrete-Time
Discrete-valued (Digital) signal
o The value of each signal sample x(n) is presented by a quantized value xq(n) from a finite
set of possible values

¤ Dynamic Range: If x max and xmin are maximum and minimum value of x(n), then
xmax  xmin  x fs is called the dynamic range or full-scale value of A/D conversion
¤ Quantization level (L): The number of values allowed in the digital signal
¤ Quantization step /Resolution (∆): Distance between two successive quantization
levels x  xmin xmax  xmin Here, b is the number of bits
  max 
L 1 2b used in coding process
¤ Quantization error eq (n) : Difference between the quantized value xq(n) and the
actual value x(n)
eq (n)  xq (n)  x(n) 
sampled signal
  eq (n) 
quantized signal
2 2
AKA_DU_EEE3203_DSP 31
Signal-to-Quantization noise Ratio (SQNR)
 When sampling rate Fs satisfies the sampling theorem, quantization is the only
error in A/D conversion process.
 Thus, quantization error can be evaluated by quantizing the analog signal xa(t)
instead of the D-T signal x(n) = xa(nT).
o signal xa(t) is almost linear between
the quantization levels.
o Quantization error: eq (n)  xq (n)  x(n)
o Let  denotes the time that xa(t)
stays within the quantization levels.
o The mean-square

error power Pq is

1 1
Pq    
2 2
e (t ) dt e (t )dt
2  0
q q


o since q  2  t,    t  
e ( t ) 
 
 
1  2 2 2 2
2

Pq     t dt  3  t dt 
 0  2  4 0 12
AKA_DU_EEE3203_DSP 32
Signal-to-Quantization noise Ratio (SQNR)
o If the quantization has b bit accuracy
and quantization covers the entire
range of 2A, the step   2 A . Hence
2b

2 2
A /3 2
A
Pq   2b 
12 2 3  22b
o If xa(t) is a sinusoidal signal, its
average power is
Tp
1 A2
  A cos 0t  dt 
2
Px 
Tp 0 2
o The quality of the output of the A/D
converter is usually measured by SQNR
2
P A  SQNR increases approximately 6 dB for
3
SQNR  x  2 2   22b every bit added to the world-length.
Pq A 2
3 2 2b
 SQNR increases approximately 6 dB for
 SQNR(dB)  10log10 ( SQNR) each doubling of the quantization level.
 1.76  6.02b
AKA_DU_EEE3203_DSP 33
Analog-to-Digital conversion
 When sampling rate Fs satisfies the sampling theorem, quantization is the only
error in A/D conversion process.

 Solution: (a) emax = Δ/2 = ± 0.01 % = ± 0.0001 → Δ = 0.002


xmax  xmin 10  (10)
   L  10000
L 1 L
(b)L  2b  b  ?

AKA_DU_EEE3203_DSP 34
Coding of quantized signals
 The coding process in A/D converter assigns a unique binary number to each
quantized level
 For L level conversion, at least L (L = 2b) different binary numbers are needed,
where b is the number of bits in the worldlength.
 Thus, b = log2L

AKA_DU_EEE3203_DSP 35
Digital-to-Analog (D/A) conversion
 D/A converter is used to convert a digital signal into an analog signal.
 The task of a D/A converter is to interpolate between samples
o The sampling theorem specifies the optimum interpolation for a bandlimited signal.
o However, this type of interpolation is too complicated
o From a practical viewpoint, the simplest D/A converter is the zero-order hold (ZOH) shown
in Fig.(a).

(a) (b)
o ZOH simply holds constant value of one sample until the next one is received
o Additional improvement can be obtained by using linear interpolation as shown in Fig. (b)
AKA_DU_EEE3203_DSP 36

You might also like