Program for Active Analog Filter Selection and Computation

Thread Starter

MikeElec

Joined Jul 13, 2020
4
Hi,
I have many active analog filters to design, mainly bandpass an lowpass.
I am tired to write pages and pages of computations, try to find the best architecture, select between Butterworth Tchebytcheff, Cauer etc..
Would you know any intelligent program which would help me to choose the best solution and to compute it from the global parameters : gain, frequencies, bandpass, etc...
Yes, I am a little lazy, but I am getting old now ...
Thank you for your help
MikeElec
 

DickCappels

Joined Aug 21, 2008
10,153
From reading your post I get the idea is that you are not only asking about something to speed up calculations but are hoping to find an expert system that will help you figure out what kind of filter you should use in this project or that. I think you will find that you can discuss if you post your questions about you applications on this forum you will find some very capable help in performing that analysis.
 

jrb_sland

Joined Dec 24, 2021
24
Hi,
I have many active analog filters to design, mainly bandpass an lowpass.
I am tired to write pages and pages of computations, try to find the best architecture, select between Butterworth Tchebytcheff, Cauer etc..
Would you know any intelligent program which would help me to choose the best solution and to compute it from the global parameters : gain, frequencies, bandpass, etc...
Yes, I am a little lazy, but I am getting old now ...
Thank you for your help
MikeElec
I learned my analog active filter stuff in the 1980s. Two references which {may or may not} still be available:

Active Filter Cookbook by Don Lancaster (c) 1975 by Howard W. Sams & Co, Inc.

Active Filter Design Handbook by G. S. Moschytz & Petr Horn (c) 1981 by John Wiley & Sons Ltd.

In my experience, the most commonly used low-pass filter is designed around the simplest circuit topology, i.e. the Sallen & Key structures, with a Butterworth {maximally flat amplitude} response. Minimum parts count, easy to implement.

Beware designing high-Q sections - you'll need increasing tightly-toleranced components. Prepare to sort both resistors & capacitors to 0.1% or better, & worry about temperature compensation. Pairing PPS & PP dielectrics works pretty well.

Also in my experience, the quasi-optimum bandpass filter is based on a multiple-feedback topology with added positive feedback. Several of these stages can then be connected in tandem with overall negative feedback to control the bandpass shape. For instance, in the late 1970s I worked up a two opamp bandpass design with deliberate underdamped behaviour to place the two peaks in its amplitude response at 1000 Hz & 1200 Hz to enhance detection of 1 Hz "ticks" from WWV {5 cycles @ 1000 Hz} & WWVH {6 cycles @ 1200 Hz}. Timing only good to ~10 ms, depending on how many hops the signals take by reflections from the ionosphere in their trip from transmitter to your receiver...
Search for WWV{H} for more info on those almost obsolete standard time transmitters in the USA. I now depend on Garmin OEM GPS receivers to get 1 Hz time pulses good to much better than 1 us.

While I have your attention, I should also mention that the field of analog wave filters is broad & confusing to the uninitiated. Be prepared for some intellectual effort if you want more than canned filter designs. Just for fun I have designed & built 8th order low-pass filters with -3 dB corner frequencies near 1 Hz with superb {< 1 % amplitude ripple, yet with low-Q (<5ish} individual sections} performance for geophysical applications, but I've been at this game for many decades.
 

MrAl

Joined Jun 17, 2014
11,389
Hi,
I have many active analog filters to design, mainly bandpass an lowpass.
I am tired to write pages and pages of computations, try to find the best architecture, select between Butterworth Tchebytcheff, Cauer etc..
Would you know any intelligent program which would help me to choose the best solution and to compute it from the global parameters : gain, frequencies, bandpass, etc...
Yes, I am a little lazy, but I am getting old now ...
Thank you for your help
MikeElec
Hi there,

Active filters are incredibly easy to design. All you have to know is some Bode analysis and you can almost draw the response by hand without any calculator.
That comes as a technique from the distant past when slide rules ruled the world and everything in it. Today we have such modern devices as the home PC computer and that can do so much for us that we really dont have to rely on an approximation technique like Bode. If you like though i can show you how easy it is or you can look it up on the web.
Just for a quick example, because of the way op amps work the frequency of interest can come from a simple knowledge of how a resistor R and capacitor C work together to create frequency sensitive circuits. The key is this:
w=1/(R*C), w=2*pi*f.
If you can understand that simple formula you can design a lot of active filters of various orders. Of course it also depends on how sharp you want the filter to be.

If you use an online calculator though you will miss all the fun of designing active filters :)


What i was wondering though was what kind of filters are we talking about here. Sometimes a set of filters falls into a general category and because of that there are generalizations that lead to very simple formulas.
 

Ian0

Joined Aug 7, 2020
9,667
The best type of filter (Butterworth, Bessel, Chebyshev, etc.) depends on the application. Do you need maximally flat response? Do you need constant group delay? Does neither of those matter and you need the steepest roll-off? There's no way to calculate which you need.
When you have decided on it, this website
http://sim.okawa-denshi.jp/en/Fkeisan.htm
will calculate the values for you (the same one @crutschow suggested)
If you need more than third order, then this book
https://www.analog.com/media/en/training-seminars/design-handbooks/basic-linear-design/chapter8.pdf
will give you give you the f and Q so that you can implement it in cascaded second-order sections.
For low-pass filters I would suggest you look at both MFB and Sallen-and-Key designs. MFB isn't great for high-pass.
If you need both low-pass and high-pass output from a single 2nd order section then look at State Variable.
 

LvW

Joined Jun 13, 2013
1,752
Hi Mike Elec,

There are numerous programs to support the design and dimensioning of passive and active filter circuits. The use of such programs can significantly reduce the amount of work up to the final dimensioned circuit and is therefore quite recommendable, provided that one makes sensible use of them.
For example, one should not expect a design program to take over the decision from the user as to which transmission characteristic and which degree of filtering will best meet the selectivity or attenuation requirements.
And the determination of a suitable circuit and the associated impedance level is also left to the user.

Before selecting a particular program for filter synthesis, it should be noted that the various programs differ significantly in their range of functions and offer only a more or less small selection of filter types, synthesis methods and circuit structures.
For example, the most powerful filter circuit of the cascade technique - the GIC stage - is only considered by very few programs. The same is true for the circuit-wise somewhat more demanding functions with transmission zeros (approximations according to Chebyshev/invers and Cauer) as well as for circuits of delay elements and all-pass filters.

With many products, the user then often only has the choice between the two standard circuits:
(1) Structure with dual negative feedback or (2) Sallen-Key topology.

In addition, these programs are limited to voltage amplifiers (OPV) when implementing the active elements. However, for many applications the amplifiers with current output (OTA) represent an attractive alternative (keywords: IC technlogy, tunability).

Another significant limitation of most programs is that filter design is only possible according to the principle of cascade synthesis. However, especially in case of increased requirements for selectivity, stability and accuracy, the methods with active emulation of passive RLC networks - active inductors, FDNR technique, leapfrog structures - have significant advantages.

There are further differences in the maximum possible filtering degree and in the formulation of the requirements for the filter - either via attenuation values (tolerance scheme) or via a specification of filtering degree and passband limit. In both cases, the desired characteristic (approximation) must then be selected from the variants offered in each case.

In general, one should be warned against an uncritical adoption of the solutions suggested by the programs. In order to find an "optimal" filter for certain requirements, additional boundary conditions and restrictions often have to be considered, which the program cannot take into account.
 
Top