FIR-design with maximally flat passband response
Besides classical methods FIWIZ implements the innovative filter design by Differential
Evolution (DE). This approach designs filters by using poles and zeros
as parameters which are automatically evolved to meet the constraints
set forth by the user. There are several types of constraints that can be
defined, allowing for the following filter properties:
Arbitrary magnitude constraints
This feature allows for multiband filters, differentiators,
sinc compensated filters, Jakes-filters for wireless channel simulation, gaussian filters, and others.
See a multiband filter and a differentiator as
examples below. The y-axis is the magnitude in dB and the x-axis denotes the normalized
If the upper and lower constraint are equal then a least squares fit will be computed.
For a simple example see the next picture:
Arbitrary group delay constraints
Applications are mainly classical lowpass, highpass, bandpass, and bandstop filters
which should exhibit approximately linear phase in the passband(s) but not necessarily
in the stopband(s). This way the filter degree can often be reduced considerably compared
to exactly linear phase FIR filters. See an IIR graphics codec as an example:
FIWIZ allows to include coefficient quantization in the filter design, i.e.
quantization is incorporated into the design as opposed to quantizing the coefficients
after the filter has been designed with high precision coefficients. This feature, which
is often crucial for FPGA-based filter implementation, is
only available for the innovative, Differential Evolution based Z-plane design. For the classical filter design methods
(butterworth, chebyshev, inverse chebyshev, elliptical) the quantization is applied
after the filter has been designed (as most common off-the-shelf filter design
See an example screen shot of the corresponding input screen (This time with Windows XP®):
Definition of a prefilter with constant coefficients
Defining a prefilter has many applications like presetting specific zeroes to suppress DC or the
50/60Hz powerline frequency, accomodating filters which are already in a design and cannot be removed,
or setting a frequency response to equalize. A well-known example for the latter is sinc compensation
needed for D/A-conversion. An example for the input screen is shown below:
Minimum phase filters
Some applications don’t require any specific phase response, and hence the filter
degree can be minimized by using minimum phase filters. Minimum phase can be easily
enforced by constraining the zero radii to be inside or on the unit circle.
Minimum delay and fractional delay filters
By allowing to freeze the group delay constraints FIWIZ assists in the design of
minimum delay or fractional delay filters.
The phase of an existing IIR-filter can be linearized with an allpass approach.
IIR-filters with reduced impulse response length
By constraining the pole radii the impulse response length of an IIR filter can be reduced. It also assists
in designing IIR-filters with maximally flat frequency response by forcing the poles back into the unit circle.
Output of poles and zeroes
The results file of FIWIZ contains not only the filter coefficients of the direct form 1
(or 2) or first and second order sections, but also the pole and zero radii as well as angles.
MATLAB friendly output format
FIWIZ's output can be directly posted on to MATLAB's command line interface for further
Storage and retrieval of configuration files
The settings of constraints and design parameters can be stored and retrieved so that
there remains only little retyping if a previous filter design shall be altered.
Platform independence through JAVA technology
FIWIZ is an application that has been written completely in
JAVA, and hence it runs on any platform which
supports the JAVA virtual machine (e.g. Windows 7,8,10, Solaris, or Mac OS).
First you have to get the Java runtime environment (JRE) for your platform.
It allows so-called Java Bytecode to run on your machine (this is the format which
Fiwiz comes in).
For convenience FIWIZ is also available as a Windows® .exe
version which runs on Windows 2000, XP, Vista, 7, 8, and 10.
Wizard based approach
FIWIZ's wizard based approach makes using FIWIZ almost self-explanatory. The sequence of
operations is evident.
Thanks to the richly featured plotting engine
PtPlot by the University
of Berkeley you can watch various data online while the filter is designed. Resizing of the
plots as well as zooming in and out is possible.
If you want to know more about Fiwiz you can download the user manual in
A4 format. If you want to try the code you can
download the full version of FIWIZ3.1 for the Java virtual machine, 32 bit,
or the Windows® .exe version. It includes the Java sources, the user manual, and the design specification.
For more information or suggestions for improvement send me
Many people have asked for the configuration
file format of FIWIZ because they want to automatically generate a configuration file via MATLAB or some other program
rather than typing the data into the GUI of FIWIZ. To inspect the file format click here.
As a further convenience a Microsoft Excel ® frontend is now provided.
More information on FIWIZ
The book "Differential Evolution - A Practical Approach to Global
Optimization" by Ken Price, Rainer Storn, and Jouni Lampinen (Springer,
ISBN: 3-540-20950-6) provides the latest findings concerning DE. It also
contains a chapter about the inner workings of FIWIZ when it comes to
Z-plane filter design as described above.
Other Books by Rainer Storn
The book "How to Become an A-Student in Science and Engineering" is supposed to help students to improve their efficiency in their science or engineering studies. It contains
many study tips that have far-reaching impact, also into their professional lives.