Mitra-Ch9 (1) CH 5 Adsp

T-61.
246 Digital Signal Processing and

Filtering
28.9.2004
Introduction
Analysis of Finite
Finite wordlength effects are caused by:
Word-Length Effects
Rounding / truncation of multiplication

results
Quantization of the filter coefficients
Quantization of the input signal

Dynamic range constraints of the
implementation
2004 Olli Simula
T-61.246 / Mitra: Chapter 9
Analysis of Finite Wordlength Effects
Ideally, the system parameters along with the

signal variables have infinite precision taking
any value between and
In practice, they can take only discrete values
within a specified range since the registers
of the digital machine where they are stored
are of finite length
The discretization process results in
nonlinear difference equations characterizing
the discrete-time systems
These nonlinear equations, in principle, are

almost impossible to analyze and deal with
exactly
However, if the quantization amounts are
small compared to the values of signal
variables and filter parameters, a simpler
approximate theory based on a statistical
model can be applied
2004 Olli Simula
2004 Olli Simula
Copyright 2001, S. K. Mitra
Analysis of Noise Properties and

Dynamic Range Constraints
2004 Olli Simula
e1[n]
Using the statistical model, it is possible to

derive the effects of discretization and
develop results that can be verified
experimentally
Sources of errors (1) Filter coefficient quantization
(2) A/D conversion
(3) Quantization of arithmetic operations
(4) Limit cycles
G1* ( z )
e j [n]
x[n]
G *j ( z )
F1* ( z ) H (z )
y[n]
v1*[n]
Fi * ( z )
vi*[n]
5
2004 Olli Simula
Mitra: Chapter 9 / Copyright 2002

Olli Simula
T-61.246 Digital Signal Processing and

Filtering
28.9.2004
Example: First Order IIR Filter
The Quantization Process and Errors
y[n] = y[n 1] + x[n]
H ( z) =
1
z
=
1 z 1 z
Quantization of input x[n]:
1
1 ' z 1
x'[n] = x[n] + e[n]
Rounding/truncation of v[n]:
v '[n] = v[n] + e [n]
Quantization of coefficients :
Fractional numbers
(sign bit + fractional
part)
The quantization
process model
H ' ( z) =
Error : = Q( x) x
Output y[n] with finite wordlength: y '[n] = y[ n] + [n]

2004 Olli Simula
Rounding /
Truncation:
2004 Olli Simula
The Quantization Errors
Rounding
1
2
(2
r 12 (2 b 2 )
Twos complement
truncation
2 b 2 t 0
Quantization Error
2 b 2 t 0 for x > 0
for x < 0
9
Quantization of Floating-Point Numbers

Only mantissa is quantized; the relative error is relevant!
x = 2 E M , Q( x) = 2 E Q( M )
2004 Olli Simula
Sign-magnitude and
ones complement
truncation
0 t 2 b 2
2004 Olli Simula
Error : e =
Q ( x) x Q ( M ) x
=
x
x
11
2004 Olli Simula
10
Analysis of Coefficient
Quantization Effects
The transfer function H ( z ) of the digital
filter implemented with quantized
coefficients is different from the desired
transfer function H(z)
Main effect of coefficient quantization is to
move the poles and zeros to different
locations from the original desired locations
2004 Olli Simula
12

Olli Simula

Filtering
28.9.2004
The actual frequency response H (e j ) is

thus different from the desired frequency
response H (e j )
In some cases, the poles may move outside
the unit circle causing the implemented
digital filter to become unstable even though
the original transfer function H(z) is stable
2004 Olli Simula
Direct form realizations are more sensitive

to coefficient quantization than cascade or
parallel forms
The sensitivity increases with increasing filter
order
Usually second order blocks in cascade or
parallel are used
13
2004 Olli Simula
14
Gain responses of a 5-th order elliptic

lowpass filter with unquantized and
quantized coefficients
Fullband Gain Response
Passband Details
original - solid line, quantized - dashed line
-20
-2
Gain, dB
Gain, dB
-40
-4
-6
-60
-80
-8
0
0.2
0.4
0.6
0.8
-10
0.1
0.2
0.3
0.4
Coefficient Quantization Effects

On a Direct Form IIR Filter
Pole and zero locations
of the filter with
(denoted by x and o)
and those of the filter
with unquantized
coefficients (denoted by
+ and *)
1
Imaginary Part

On a Direct Form IIR Filter
0.5
0
-0.5
-1
-1
-0.5
0
0.5
Real Part
0.5
2004 Olli Simula
15
2004 Olli Simula
16

On a Cascade Form IIR Filter
Passband Details

0.5

On A Direct Form FIR Filter
Gain responses of a 5-th order elliptic

lowpass filter implemented in a cascade form
with unquantized and quantized coefficients
Gain responses of a 39-th order equiripple

lowpass FIR filter with unquantized and
Passband details
1

0
0
Gain, dB
-40
-0.5
-1
-1.5
-60
0
0.2
0.4
0.6
0.8
-2
2004 Olli Simula
0.1
0.2
0.3
0.4
0.5
17

Olli Simula
Gain, dB
-20
Gain, dB
Gain, dB
-20
-40
-60
-1
-2
-3
0.2
2004 Olli Simula
0.4
0.6
/
0.8
0.2
0.4
0.6
/
18

Filtering
28.9.2004
Example of Coefficient Quantization

in 6th Order Direct Form Realization
Amplitude responses
2004 Olli Simula
Pole-zero locations
19
Example:
Amplitude responses
2004 Olli Simula
Pole-zero locations
20
Coefficient Quantization in FIR Filters
6th
order bandstop
filter with unquantized
coefficients
Consider an (M-1)th order FIR transfer function

M 1
H ( z ) = h[n]z n
n =0
Quantization of the filter coefficients results in a new

transfer function
Cascade form with

coefficients quantized
to 6 bits
M 1
M 1
n=0
n =0
H ' ( z ) = h'[n] z n = (h[n] + e[n])z n

H ' ( z) = H ( z) + E ( z)
Parallel form with

coefficients quantized
to 6 bits
2004 Olli Simula
Example of Coefficient Quantization in

6th Order Cascade Form Realization
H(z)
+
E(z)
Linear phase: h[n] = + h[N-1-n]

Symmetry of the impulse response not affected by
quantization
21
A/D Conversion Noise Analysis
2004 Olli Simula
22
Quantization Noise Model

Twos complement
representation
Analog input
Input analog
sample x[n]
Quantized
input sample
xeq [n] =
Quantization of the input signal introduces error at

the input of the filter
This error is propagated through the filter together
with the input signal
Affects the signal-to-noise ratio of the system
2004 Olli Simula
1 xeq [n] < 1
Binary equivalent
of quantized input

Olli Simula
2 x[n]
RFS
RFS = 2b +1
Input signal is assumed to be scaled to be in the
range of +1 by dividing its amplitude by RFS/2
23
2004 Olli Simula
24

Filtering
28.9.2004
Model of the Quantization Error
Quantization Error
x[n ]
The quantization error e[n]:
e[ n] = Q( x[n]) x[ n] = x[ n] x[n]
Assumptions:
2 < e[n] 2
For twos complement rounding:
e[n] is called
granular noise
Outside RFS the error increases linearly; e[n] is called
the saturation error or the overload noise
The output value is clipped to the maximum value
2004 Olli Simula
25
me =
e[n ]
The assumptions hold in most practical situations with rapidly

changing input signals
2004 Olli Simula
26
Signal-to-Noise ratio
( / 2) ( / 2) = 0
2
(( / 2) ( / 2))2 = 2
e2 =
12
12
0
=
2
2
2
(
0 ) 2
2
e =
=
12
12
me =
(b) Twos
complement
truncation
x[n] = x[n] + e[n]
1) The error sequence {e[n]} is a sample sequence of a widesense stationary (WSS) white noise process, with each sample
e[n] being uniformly distributed over the quantization error
2) The error sequence is uncorrelated with its corresponding input
sequence {x[n]}
3) The input sequence is a sample sequence of a stationary
random process
Quantization Error Distributions

(a) Rounding
Additive quantization noise e[n] on the signal x[n]

Signal-to-quantization noise ratio in dB is defined as
2
SNR = 10 log10 x2 dB
e
where
x2 is the signal variance (power) and
e2 is the noise variance (power)
The variance represents the noise power

2004 Olli Simula
27
Signal-to-Noise Ratio
2
12
R
= 6.02b + 16.81 - 20log10 FS
dB
Thus, SNR increases 6 dB for each added bit in the

wordlength
2004 Olli Simula

Olli Simula
28
Let the input scaling factor be A with A>0

The variance of the scaled input Ax[n] is A2x2
The SNR changes to
2
2
2 2 ( b +1) RFS
2 2b RFS
=
12
48
48 2
SNR A / D = 10 log10 2b x2
2 RFS
Effect of Input Scaling on SNR
A/D conversion:
(b+l) bits: = 2-(b+1)RFS , where RFS is the full-scale
range
e2 =
2004 Olli Simula
29
R
SNR A / D = 6.02b + 16.81 - 20log10 FS
A x
= 6.02b + 16.81 20 log10 ( K ) + 20 log10 ( A)
where RFS=Kx (x is the RMS value of the signal)
Scaling down the input signal (A<1) decreases the SNR
Scaling up the input signal (A>1) increases the possibility
to exceed the full-scale range RFS resulting in clipping SNR
2004 Olli Simula
30

Filtering
28.9.2004
Propagation of Input Quantization

Noise to Digital Filter Output

The mean and variance of v[n] characterize the output
noise
The mean mv is:

Due to linearity of H(z) and the assumption that x[n] and
e[n] are uncorrelated the output can be expressed as a
linear combination (sum) of two sequences:
The noise variance v2 is:
v2 =
y[n] = h[n] x[n]
= h[ n] [x[ n] + e[n]] = h[n] x[n] + h[n] e[n]
e2
2
H (e
) d
The output noise power spectrum is:
Pvv2 ( ) = e2 H (e j )
e[m]h[n m]
v[n] =
The output noise is:
mv = me H (e j 0 )
m =
2004 Olli Simula
31
2004 Olli Simula
32

The normalized output noise variance is given by
v2 1
=
e2 2
which can be written as:
v2,n =
v2,n =
H (e
Analysis of
) d
Arithmetic Round-Off Errors
1
H ( z ) H ( z 1 )z 1dz
2j C
An equivalent expression is:
v2,n =
2004 Olli Simula
h[n]
n =
33
Quantization of Multiplication Results

The quantization model can be used to analyze the
quantization effects at the filter output
Assumptions:
Quantization before
summation
The number of
multiplications kl at adder
inputs
The rth branch node with
signal value ur[n] needs
to be scaled to prevent
overflow
1) The error sequence {e[n]} is a sample sequence of a

stationary white noise process, with each sample e[n]
being uniformly distributed
2) The quantization error sequence {e[n]} is uncorrelated with
the signal {v[n]}, the input sequence {x[n]} to the filter, and
all other quantization errors
The assumption of {e[n]} being uncorrelated with

{v[n]} holds for rounding and twos complement
truncation
2004 Olli Simula

Olli Simula
35
2004 Olli Simula
36

Filtering
28.9.2004

Statistical model of the filter:
fr[n]
Impulse response from
filter input to branch
node r
gl[n]
Impulse response from
input of lth adder to
filter output
2004 Olli Simula
37

L
l =1
1
2j C
Branch nodes to be scaled

lead to multipliers and are
outputs of summations:
Gl ( z )Gl ( z 1 ) z 1dz
l =1
n =0
2004 Olli Simula
vl [n]
Let 02 be the variance of each individual noise

source; then kl02 is the noise variance of el[n]
The output noise variance is:
[(
02 kl
G ( z)G ( z
1
2j C
2004 Olli Simula
)]

) z 1dz = 02 kl 21

G(e
2
) d
38
The Output Quantization Noise

Quantization of
multiplication results
after summation
reduces the number of
noise sources to one
The variance of the
noise source el[n] is
now 02
DSP processor carry out multiply-accumulate

operation using double precision arithmetic
39
Dynamic Range Scaling

Olli Simula
40
Three different conditions to ensure that ur[n]

satisfies the conditions:
1) An absolute bound
2) Linfinity -bound
3) L2 -bound
Different bounds are applicable under certain input
signal conditions
The rth node value ur[n] has to be scaled

Assume that the input sequence is bounded by
unity, i.e., |x[n]| < 1 for all values of n
The objective of scaling is toensure that
|ur[n]| < 1 for all r and all values of n
2004 Olli Simula
Dynamic Range Scaling
Digital filter
2004 Olli Simula
Scaling transfer function: Fr(z)

Noise transfer function: Gl(z)
where L is the number of summation nodes to which

noise sources are connected
The noise variance can also be written as
2 = 02 kl g l[n]
u r [n]
The amount of noise depends on the implementation
The total output noise variance:
2 = 02 kl
41
2004 Olli Simula
42

Filtering
28.9.2004
An Absolute Bound
An Absolute Bound
Assuming that x[n] satisfies the dynamic range
constraint |x[n]| < 1
Digital filter
u r [n] =
f r [n]
2004 Olli Simula
k =
43
2004 Olli Simula
f r [k ]
k =
45
Scaling Norms: L-Bound
1
u r [ n]
2
F (e
F (e
r
) X (e j )e jn d
) X ( e j ) d
Fr (e j )
2004 Olli Simula
X (e
X ( e j )
2004 Olli Simula
= max F (e j )

46
If the mean absolute value of the input spectrum is

bounded by unity, then there will be no adder
overflow if the peak gains from the filter input to all
adder output nodes are scaled satisfying the above
bound
The scaling rule is rarely used since with most input
signals encountered in practice ||X||1 < 1 does not hold
1
Fr (e j )
2
If ||X||1 < 1, then the dynamic range constraints

satisfied if
F 1
An inverse Fourier transform
1
p
p
=
F (e j ) d
p
2
L2-norm, ||F||2, is the root-mean-square (RMS) value

of F(ej), and
L1-norm, ||F||1, is the mean absolute value of F(ej)
over
Moreover, limp-> ||F||p exists for a continuous F(ej)
and is given by its peak
Scaling Norms: L-Bound
U r (e j ) = Fr (e j ) X (e j )
1
2
44
The scaling rule based on the absolute bound is too

pessimistic and reduces the SNR significantly
More practical and easy to use scaling rules can be
derived in the frequency domain if some information
about the input signal is known a priori
u r [ n] =
2004 Olli Simula
f r [k ] 1 for all r
Define the Lp-norm of a Fourier transform F(ej) as
If the dynamic range constraint is not satisfied the

filter input has to be scaled with the multiplier K
f r [k ]
Scaling Norms
Scaling with the Absolute Bound
max
k =
This is both necessary and sufficient condition to

guarantee that there will be no overflow
K=
k =
f [k ]x[n k ]
k =
The node value ur[n] now satisfies the dynamic

range constraint, i.e., | ur[n]| < 1 if
Fr(z) is the scaling transfer function

The node value ur[n] is determined by the
convolution
u r [ n] =
f [k ]x[n k ]
) d

Olli Simula
47
2004 Olli Simula
48

Filtering
28.9.2004
A General Scaling Rule
Scaling Norms: L2-Bound
1
u r [ n] =
Fr (e j ) X (e j )e jn d
2
Applying Schwarz inequality
1
2
ur [n]
2
1
j 2
Fr (e ) d 2
u r [n] Fr (e j )
or equivalently
A more general scaling rule is obtained using

Holders inequality
ur [n] Fr (e j )
2
X (e j ) d
X ( e j )
for all p,q > 1, with
If the filter input has finite energy bounded by unity, i.e., ||X||2 < 1,
then the adder overflow can be prevented by scaling the filter
such that the RMS value of the scaling transfer functions are
bounded by unity:
2004 Olli Simula
1, r = 1,2,..., R
49
i =1
where H i ( z ) =
Scaling transfer functions: Fr ( z ) =
Bi ( z ) 1 + b1i z 1 + b2i z 2
=
Ai ( z ) 1 + a1i z 1 + a2i z 2
K ' r 1
H 'l ( z) , r = 1,2,..., R
Ar ( z ) l =1
Fr(z) can be expressed by poles and zeros of the original H(z)

2004 Olli Simula
51
( 1 p )+ ( 1q ) = 1
F ' p 1, r = 1,2,..., R
In many structures the scaling multipliers can be

absorbed to the existing feedforward multipliers
2004 Olli Simula
50
Scaling - Back-Scaling
1
H ( z ) = K H i ( z ),
X (e j )
After the scaling the transfer functions become ||F||p

and the scaling constants should be chosen such that
Scaling of a Cascade Form IIR Filter
The nodes
(*) need to
be scaled
FILTER
The effect of input scaling is compensated by back-scaling at

the output of the filter
Scaling block-by-block in cascade realization forms
H1(z)
1
H2(z)
2
R1
HR(z)
Each second order block is scaled individually

The scaling coefficients between the blocks contain the backscaling of the previous block and the scaling of the the next
block
2004 Olli Simula
52
Scaled Cascade Form IIR Filter Structure

Scaling transfer
functions:
Fr ( z ) =
K r 1
H l ( z),
Ar ( z ) l =1
r = 1,2,..., R
R
H ( z ) = K H l ( z ),
l =1

Olli Simula
Pole-Zero Pairing of a Cascade

Form IIR Digital Filter
1 + b1i z 1 + b2l z 2
where H l ( z ) =
1 + a1l z 1 + a2l z 2
The scaled structure has new values of the coefficients in the

feed-forward branches
Only one critical branch node in each second order block has
to be checked for overflow
2004 Olli Simula
Optimum Section Ordering and
Ordering of second-order sections as well as

pairing of poles and zeros affects the output noise
power of the filter
53

Filtering
28.9.2004
Noise Transfer Functions

The output noise power spectrum due to product
round-off is given by
2
R +1
Pyy ( ) = 02 kl Gl (e j )
l =1
and output noise variance is

R +1 1
l =1 2
y2 = 02 kl
The noise transfer functions can be expressed using the

transfer functions of the cascaded second-order blocks
The scaled noise transfer functions are given by
R
R
Gl ( z ) = K H i ( z ) = i Gl ( z ) , l = 1,2,..., R ; and GR +1 ( z ) = 1
i =l
i =l
2004 Olli Simula
55
Noise Model of Second-Order Blocks

The noise model introduces noise sources to the
input/output summation of each block
The number of elementary noise sources, kl, has
different values depending on the location of
rounding (before or after the summation) and
depending on the block (first, intermediate, last)
Let kl be the total number multipliers connected to
the lth adder
Rounding before summation:
k1 = kR+1 = 3,
kl = 5, for l = 2, 3,...,R
Rounding after summation: kl = 1, for l = 1,
2,...,R+1
2004 Olli Simula
57
Minimizing the Output Round-Off Noise

The scaling transfer function Fl(z) contains sections
Hi(z), i = 1, 2,..., l-1
The noise transfer function Gl(z) contains sections
Hi(z), i = l, l+1,..., R
Every term in the sum for the noise power or the
noise variance includes the transfer function of all R
sections in the cascade realization
To minimize the output noise power the norms of Hi(z)
should be minimized for all values of i by
appropriately pairing the poles and zeros
2004 Olli Simula

Olli Simula
59
G (e
2
2
R +1
) d = 02 kl Gl 2
l =1
where the integral in the parenthesis is the square

of the L2-norm of the noise transfer function
2004 Olli Simula
56

The scaling coefficients are
R
i =l
Fl
l
=
R +1 H
p
p
The output noise power spectrum of the scaled filter is

Pyy ( ) =
02
k R +1 H
R +1
+ kl Fl
l =1
2
Gl (e j )
and output noise variance is
y =
2004 Olli Simula
02
k R +1 H
R +1
+ kl Fl
l =1
2
p
2
Gl 2
58
Pairing the Poles and Zeros

Poles close to unit circle introduce gain and zeros (on
the unit circle) introduce attenuation
1) First, the poles closest to the
unit circle should be paired
with the nearest zeros
2) Next, the poles closest to the
previous set of poles should
be paired with the next
closest zeros
3) This process is continued
until all poles and zeros are
paired
2004 Olli Simula
60
10

Filtering
28.9.2004
Section Ordering
Section Ordering
A section in the front part of the cascade has its
transfer function Hi(z) appearing more frequently in
the scaling transfer functions
A section near the output end of the cascade has its
transfer function Hi(z) appearing more frequently in
the noise transfer function expressions
=>
The best location for Hi(z) depends on the

type of norms being applied to the scaling
and noise transfer functions
2004 Olli Simula
61
Error Spectrum Shaping
L2 scaling:
The ordering of paired sections does not influence too much
the output noise power since all norms in the expressions are
L2-norms
L scaling:
The sections with poles closest to the unit circle exhibit a
peaking magnitude response and should be placed closer to
the output end
=> The ordering should be from least-peaked to most-peaked
On the other hand, the ordering scheme is exactly opposite if
the objective is to minimize the peak noise ||Pyy()|| and L2scaling is used
The ordering has no effect on the peak noise with L-scaling
2004 Olli Simula

Without error-feedback
the error signal e[n] is
the pure quantization
error, i.e.,
e[n] = y[n] - x[n]
In the compensated
structure the error signal
is the difference
between the output y[n]
and the compensated
input signal
Quantization error can

be compensated using
the so called errorfeedback (or error
spectrum shaping)
The filtered error signal
is added to the signal
branch before
quantization (Q[.]).
2004 Olli Simula
62
63
2004 Olli Simula
64
e[n] = y[n] w[ n]
Solving y[n] - x[n]:
y[n] x[n] = e[n] + ae[n 1] + be[n 2]
Substituting w[n]:
e[n] = y[n] x[n]
Taking the z-transform:
ae[n 1] be[n 2]
Total error between
output and input is still:
e[n] = y[n] x[n]
Y ( z ) X ( z ) = E ( z ) + az 1 E ( z ) + bz 2 E ( z )
= 1 + az 1 + bz 2 E ( z ) = G ( z ) E ( z )
where G(z) is the error shaping transfer function
w[n] = x[ n] + ae[n 1] + be[n 2]

2004 Olli Simula

Olli Simula
65
2004 Olli Simula
66
11

Filtering
28.9.2004

Example: a=-2 and b=1
(
= (1 2 z
G ( z ) = 1 + az 1 + bz 2
1
+z
) = (1 z )
1 2
Double zero is at z=1

Noise spectrum is modified by attenuating noise at
low frequencies
2004 Olli Simula

Olli Simula
67
12

Mitra-Ch9 (1) CH 5 Adsp

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Mitra-Ch9 (1) CH 5 Adsp

Загружено:

Авторское право:

Доступные форматы

T-61.

246 Digital Signal Processing and

Finite wordlength effects are caused by:

Rounding / truncation of multiplication

Quantization of the filter coefficients

Quantization of the input signal

T-61.246 / Mitra: Chapter 9

Analysis of Finite Wordlength Effects

Analysis of Finite Wordlength Effects

Ideally, the system parameters along with the

These nonlinear equations, in principle, are

2004 Olli Simula

2004 Olli Simula

T-61.246 / Mitra: Chapter 9

T-61.246 / Mitra: Chapter 9

Copyright 2001, S. K. Mitra

Copyright 2001, S. K. Mitra

Analysis of Noise Properties and

Analysis of Finite Wordlength Effects

2004 Olli Simula

T-61.246 / Mitra: Chapter 9

Using the statistical model, it is possible to

2004 Olli Simula

T-61.246 / Mitra: Chapter 9

Copyright 2001, S. K. Mitra

Mitra: Chapter 9 / Copyright 2002

T-61.246 Digital Signal Processing and

Example: First Order IIR Filter

The Quantization Process and Errors

y[n] = y[n 1] + x[n]

Quantization of input x[n]:

v '[n] = v[n] + e [n]

Output y[n] with finite wordlength: y '[n] = y[ n] + [n]

T-61.246 / Mitra: Chapter 9

The Quantization Errors

T-61.246 / Mitra: Chapter 9

Quantization of Floating-Point Numbers

2004 Olli Simula

T-61.246 / Mitra: Chapter 9

T-61.246 / Mitra: Chapter 9

2004 Olli Simula

T-61.246 / Mitra: Chapter 9

T-61.246 / Mitra: Chapter 9

Mitra: Chapter 9 / Copyright 2002

T-61.246 Digital Signal Processing and

The actual frequency response H (e j ) is

T-61.246 / Mitra: Chapter 9

Direct form realizations are more sensitive

2004 Olli Simula

T-61.246 / Mitra: Chapter 9

Copyright 2001, S. K. Mitra

Gain responses of a 5-th order elliptic

original - solid line, quantized - dashed line

Coefficient Quantization Effects

Coefficient Quantization Effects

Copyright 2001, S. K. Mitra

2004 Olli Simula

T-61.246 / Mitra: Chapter 9

2004 Olli Simula

T-61.246 / Mitra: Chapter 9

Copyright 2001, S. K. Mitra

Coefficient Quantization Effects

Fullband Gain Response