GP 503: Geophysical Signal Processing

Lecture – 5

Instructor: E. Chandrasekhar
Department of Earth Sciences
Indian Institute of Technology Bombay
E-mail: [email protected]
Application of Z-transform: Digital Filters
Z-transform is mainly used in digital filters to determine the transfer function and ‘Poles’
and ‘zeros’ of the digital filters. The ratio of signal output to input signal is defined as
transfer function.

Since ‘Z’ is known as unit-delay operator, let’s try to understand the following notations,
before proceeding further.
zyn = yn −1 i.e., zy2 = y1 ; zy1 = y0 ; z 2 yn = yn − 2 ; zy0 = y−1 = 0, etc.
If yn is the output of an input series xn, given by

10 yn = 6 xn − 5 xn −1 + xn − 2 + 7 yn −1 − yn − 2 (4)

In Z-transform notation, equation (4) can be written as

( z 2 − 7 z + 10) yn = ( z 2 − 5 z + 6) xn (5)
The transfer function for the above filter (equation (5)) is given by
yn ( z 2 − 5 z + 6) z − 3
= 2 =
xn ( z − 7 z + 10) z − 5 (6)

A few observations:
• Equation (4) explains that, the output of the given digital filter contains the present
input, xn, previous inputs and previous outputs (yn-1 and yn-2). Such a filter is known as
‘Recursive filter’. It is also known as Feedback filter.
• If a filter contains only present input and previous inputs, then such a filter is known as
‘Non-recursive filter’. It is also known as Feedforward filter.
• The maximum number of previous inputs OR previous outputs in a given recursive filter
defines the order of the digital filter. What is the order of the above digital filter?
• In equation (6), the values of ‘z’ at which the numerator becomes zero are known as
ZEROs of the digital filter. Similarly, the values of ‘z’ at which the denominator becomes
zero are known as POLEs of the digital filter.
contd……..
A few observations:
• In the Pole-Zero plot of the digital filter, the Poles and Zeros signify the region
of stability of the designed filter.

• Poles and Zeros are also useful in determining the filter response.

• The poles and zeros of a transfer function may also be complex, and thus may
be represented graphically by plotting their locations on the complex plane,
whose axes represent the real and imaginary parts of the complex variable.

• Such plots are known as pole-zero plots. It is usual to mark a zero location by
a circle (o) and a pole location a cross (×).
Example of a pole-zero plot for a linear digital filter, having
A zero at z = -2 and Poles at z = -1 and z = -1 ± 2i:

Fig. 3
Example -1:
Determine the ‘Poles’ and ‘Zeros’ of a digital filter: yn =
1
[xn + 2 xn−1 + 5 y n−1 − y n−2 ]
6

The transfer function of the above filter is given by

The digital filter has:

yn (1 + 2 z ) 2( z + (1 / 2)) (i) a single real zero at z = −1/2.
= 2 = (ii) a pair of real poles at z = 3 and z = 2 and
x n ( z − 5 z + 6) ( z − 3)( z − 2)
(iii) a gain factor of 2.

The poles and zeros signify the properties of the digital filter describing the dynamics of the input-
output system. Together with the constant gain factor, K, they provide a complete description of
the filter design.
Example -2:
Determine the linear digital filter, if it has ‘Poles’ and ‘Zeros’ as shown in the figure below.

This filter has a zero at z = -2

and
Poles at z = -1 and z = -1 ± 2i:
Filtering and Windowing
 Different types of digital filters that are commonly used
in geophysical signal analysis:

1. Low pass filter

2. High pass filter
3. Band pass filter
4. Band reject filter
 Different types of digital filters that are commonly used
in geophysical signal analysis:
1. Low pass Filter

A low pass filter ‘filters’ (or passes) the signals having frequencies ’lower’ than the
selected cut-off frequency and attenuates the rest of the signals, which will have
frequencies higher than the specified cut-off frequency.

The shape of an ideal low-pass Filter is 1

Filter gain (dB)

Any digital filter is characterized by its transfer function,
which is expressed in generalized form in z-domain as 𝑓𝑓𝑐𝑐 𝑓𝑓 →
𝑦𝑦(𝑧𝑧) 𝑎𝑎0 + 𝑎𝑎1 𝑧𝑧 + 𝑎𝑎2 𝑧𝑧 2 + 𝑎𝑎3 𝑧𝑧 3 + ⋯ … … … 𝑎𝑎𝑁𝑁 𝑧𝑧 𝑁𝑁
𝐻𝐻 𝑧𝑧 = =
𝑥𝑥(𝑧𝑧) 𝑏𝑏0 + 𝑏𝑏1 𝑧𝑧 + 𝑏𝑏2 𝑧𝑧 2 + 𝑏𝑏3 𝑧𝑧 3 + ⋯ … … … 𝑏𝑏𝑀𝑀 𝑧𝑧 𝑀𝑀
1. Low pass Filter

where,, 𝑎𝑎0 , 𝑎𝑎1 , 𝑎𝑎2 , 𝑎𝑎3 , … … … . . 𝑎𝑎𝑁𝑁 and 𝑏𝑏0 , 𝑏𝑏1 , 𝑏𝑏2 , 𝑏𝑏3 , … … … . . 𝑏𝑏𝑀𝑀 are known as
filter coefficients, which will be estimated using least-squares technique. N or M
designates the order the digital filter.

How the filter coefficients affect the filter response? Optimum number of filter
coefficients should be decided by the user, by trial-and-error method, depending
upon the type of application and till all the required frequencies are filtered out
from the signal.

Generally, the order of a digital filter should be reasonably high, such that, with the
increase in the order of the digital filter, the filter attenuates the high frequencies
with a steeper roll-off.
An example of Butterworth filter with different orders with a cut-off frequency of 𝜔𝜔0 = 1:

(Source: Wikipedia)
 Steeper roll off


An example of comparison of responses of different standard low pass filters

https://en.wikipedia.org/wiki/Chebyshev_filter
Response functions for low pass filters
where,
1
𝐻𝐻(𝜔𝜔) = 𝜔𝜔 = 2𝜋𝜋𝜋𝜋 = frequency
1. Butterworth filer 2𝑛𝑛
𝜔𝜔 𝜔𝜔0 = cut-off frequency
1 + 𝜀𝜀 2
𝜔𝜔0
n = order of the filter
1
2. Chebyshev filter (Type-I) 𝐻𝐻(𝜔𝜔) = 𝜀𝜀 = ripple factor
𝜔𝜔
1 + 𝜀𝜀 2 𝑋𝑋𝑛𝑛2
𝜔𝜔0
1 𝑋𝑋𝑛𝑛 = Chebyshev polynomial
3. Chebyshev filter (Type-II) 𝐻𝐻(𝜔𝜔) =
1 of order, n
1+ 𝜔𝜔
𝜀𝜀 2 𝑋𝑋𝑛𝑛2
𝜔𝜔0
1
4. Elliptic filter 𝐻𝐻(𝜔𝜔) = 𝑅𝑅𝑛𝑛 = Elliptic rational function
𝜔𝜔
1 + 𝜀𝜀 2 𝑅𝑅𝑛𝑛2 𝛽𝛽,
𝜔𝜔0 𝛽𝛽 = selectivity factor
 Different types of digital filters that are commonly used
in geophysical signal analysis:
2. High Pass Filter

A high pass filter ‘filters’ the signals with a frequency ’higher’ than the selected cut-off
frequency and attenuates the rest of the signal, which will have frequencies lower
than the specified cut-off frequency.

The shape of an ideal high-pass Filter is 1

Filter gain (dB)

𝑓𝑓𝑐𝑐 𝑓𝑓 →
The order of a digital filter should be reasonably high, such that, with the increase
in the order of the digital filter, the filter attenuates the low frequencies with a
steeper roll-off.
 Different types of digital filters that are commonly used
in geophysical signal analysis:
2. High Pass Filter: First differencing
 2. High Pass Filter

After applying First

difference filter on the
data

(Chandrasekhar et al., 2003, EPS)

 Different types of digital filters that are commonly used
in geophysical signal analysis:
3. Band Pass Filter
A band pass filter ‘filters’ (or passes) the signals corresponding to a particular range of
frequencies of interest at the chosen band of cut-off frequencies and attenuates the
rest of the signal, which will have all the frequencies above and below than the
specified cut-off frequencies.
The shape of an ideal band-pass Filter is 1

Filter gain (dB)

Here again, the order of a digital filter should be reasonably
high, such that, with the increase in the order of the digital 𝑓𝑓𝑐𝑐𝑐 𝑓𝑓𝑐𝑐𝑐 𝑓𝑓 →
filter, the filter attenuates all the frequencies above and
below the specified cut-off frequencies with a steeper roll-
off.
(a) Fourier spectrum of the raw
TEC time series (black line). Note
the well-resolved 1-day
periodicity, depicting the
fundamental period of TEC. The
spectrum of the filtered time
series of the one-day period (blue
line) is superposed. To highlight
the effectiveness of the Type-I
Chebyshev filter, the enlarged
portion of the spectra of raw and
filtered data of one-day
periodicity are shown in (b).
Other well-known periods of
solar origin, viz., 27-day, 13.5-day
and 9-day have also been shown
in the inset in (a).
(Chandrasekhar et al., JASTP, 2016)
 Different types of digital filters that are commonly used
in geophysical signal analysis:
3. Band reject filter
A band reject filter ‘filters’ (or rejects) the signals corresponding to a particular range
of frequencies of interest at the chosen band of cut-off frequencies and passes the
rest of the signal, which will have all the frequencies above and below than the
specified cut-off frequencies.
The shape of an ideal band-reject Filter is 1

Filter gain (dB)

Here again, the order of a digital filter should be reasonably
high, such that, with the increase in the order of the digital 𝑓𝑓𝑐𝑐𝑐 𝑓𝑓𝑐𝑐𝑐 𝑓𝑓 →
filter, the filter passes all the frequencies above and below
the specified cut-off frequencies with a steeper roll-off.
Any Questions?