See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/283348303

The Time Domain Performance of Chebyshev Filters

Conference Paper · March 2009

CITATIONS READS

0 671

1 author:

Pieter Willem Van der Walt

Stellenbosch University
59 PUBLICATIONS 254 CITATIONS

SEE PROFILE

All content following this page was uploaded by Pieter Willem Van der Walt on 01 November 2015.

The user has requested enhancement of the downloaded file.

Time Domain Performance of
Chebyshev Filters

PW van der Walt

RRS
University of Stellenbosch
Introduction 1
 In modern multi-mode radars pulse
compression, spectrum shaping and detection
is performed digitally
 With waveform libraries the transmitted signal can
be optimized for different circumstances
 The analogue signal paths are required to
transport signals of different bandwidths
 They are not primarily used for shaping the signal
spectrum
 They must preserve signal fidelity
Introduction 2
 Filters in the up and down conversion signal
paths of pulsed systems must be designed
with care to ensure signal fidelity
 Ringing filters produce time side-lobes in the
matched filter response
 This may lead to non-detection of small targets in the
vicinity of targets with large RCS
 Ringing in the transmit path can delay the
operation of the receiver protector
 This may degrade performance and prevent detection of
close-in targets
Introduction 3
 Traditionally filters approximating the gaussian
response of the Ideal Time Domain Filter (ITDF) are
used for pulse signals
 Examples are the Gauss filter, the Gauss-to-12 dB filter,
maximally flat phase, equiripple (Chebyshev) phase, limited
overshoot filter.
 Filters approximating the brick-wall response of the
Ideal Frequency Domain Filter (IFDF) are usually not
considered
 Examples are the Butterworth, Chebyshev and Elliptic
function filters
 The next slide shows why…
 Responses Impulse and
Frequency
 Impulse responses of n=5 Gauss-to-12 dB, Chebyshev Phase 0.5° and
Chebyshev 0.1 dB ripple filters with ω-3dB=1. Note logscale for impulse
response.
Discussion
 The Chebyshev filter is clearly not suitable!
 However:
 the comparison is unfair!
 The ITDF group of filters have poor attenuation
performance
 The superior attenuation performance of the
Chebyshev filter is not taken into account
 The real life input signal is not an impulse
Another approach
 Filters are required to reject unwanted
signals, such as images, and pass wanted
signals
 In real life a filter must always meet a stop-
band specification
 An alternative approach is to compare the
time domain performance of filters with
equivalent stop-band performances
 A shaped pulse is a more appropriate input
signal than an impulse or a rectangular pulse
 Shaped pulse characteristics
Cosine-squared pulse and the equivalent rectangular pulse with their spectra
Performance Comparison
 n=5 Gauss-12 dB, Chebyshev Phase, Chebyshev and Lossy Chebyshev
filters response to sin2(t) shaped pulse of 3 dB duration 3.6 s with the
same 50 dB cut-off frequency
Conclusions 1
 On this basis, the time domain performance
of the Chebyshev filter is comparable with
that of the IFDF approximations
 When element losses are taken into account,
the non-predistorted Chebyshev filter
produces a shorter pulse and the close-in
ringing dies out faster than that of the ITDF
filters
Conclusions 2
 The Chebyshev approach is attractive
 Chebyshev filters are easier to realize in
coupled resonator form than IFDF filters
 The lossy Chebyshev filter is well matched
to source and load
 The Chebyshev filter is easy to tune and
test
 Band-pass example
Band-pass filter with n=10, transmission zeros at zero (1), infinity (5)
and finite imaginary axis (4) and Chebyshev pass-band response.
 Bandpass Example
 Filter centered on 125 MHz with 30 MHz ripple bandwidth and stop band starting at 85 MHz

CH1 S11 dB MAG 5 dB/ REF 0 dB 3: -10.59 dB CH1 S21 LIN DLY 5 ns/ REF 42.4 ns 3: 52.34 ns
CH2 S21 dB MAG 10 dB/ REF 0 dB 3: -5.971 dB
140.075 MHz
67.4 ns 160 MHz
140.075 MHz
10 dB
20 160 MHz 1: 32.39 ns
1: -3.016 dB
125 MHz
125 MHz
2: 47.24 ns
2: -3.938 dB
109.925 MHz
109.925 MHz
4 5 4: -103.4 ns
2 6 1 7 4: á 0 dB
-70.38 dB
83.15 MHz
3
83.15 MHz
3 5: -59.92 ns
5: -69.70 dB
167.075 MHz
167.075 MHz
CAL 6: 36.71 ns CAL
6: -3.118 dB 2
115.1 MHz
3
115.1 MHz
7: 39.58 ns
2 7: -3.832 dB
134.9 MHz
6 134.9 MHz
5 ns/
5 dB/
10 dB/
1 á 42.4 ns
7 7

6
CPL CPL

FIL FIL
10k 10k

4 5

-40 dB
-80 4
17.4 ns 5

START 80 MHz 10 MHz/ STOP 170 MHz START 80 MHz 10 MHz/ STOP 170 MHz

Date: 17.MAR.09 12:07:16 Date: 17.MAR.09 13:07:50

Envelope responses
 Conclusion
 Full time domain simulation of filter
responses with real input signals should
be done when selecting filter prototypes
 You may be surprised by the outcome!

View publication stats