Hi Guys,
I've been asked to review a manuscript for possible
publication in an IEEE journal. The material concerns
the design of IIR filters using an "Evolutionary Algorithm".
In the text the author states that the "classic" (i.e.,
traditional) approach to IIR filter design is:
"... a digital filter is transformed to a corresponding
analog low-pass (LP) filter. Then, well-known LP filter
design methods, such as Butterworth, Chebyshev Type I,
and Chebyshev Type II, can be used to accomplish the
design of the analog LP filter. Finally, the analog LP
filter is transformed back to the digital filter by
again using a bilinear transformation."
Stated in different words, it seems that this design
method comprises:
1. Start with a digital filter,
2. Convert that digital filter to an analog
(continuous) filter,
3. Optimize the analog filter in some way, and
4. Convert the analog filter back to a digital
filter.
I've never heard of this "classic" design method before.
Have any of you heard of this IIR digital filter design
method? I must be missing something here.
Thanks Much,
[-Rick-]
IIR Filter Design Question
Started by ●December 22, 2010
Reply by ●December 22, 20102010-12-22
On Dec 22, 3:23�pm, Rick Lyons <R.Lyons@_BOGUS_ieee.org> wrote:> > � �I've been asked to review a manuscript for possible > publication in an IEEE journal. �The material concerns > the design of IIR filters using an "Evolutionary Algorithm". > > In the text the author states that the "classic" (i.e., > traditional) approach to IIR filter design is: > > � �"... a digital filter is transformed to a corresponding > � � analog low-pass (LP) filter. Then, well-known LP filter > � � design methods, such as Butterworth, Chebyshev Type I, > � � and Chebyshev Type II, can be used to accomplish the > � � design of the analog LP filter. Finally, the analog LP > � � filter is transformed back to the digital filter by > � � again using a bilinear transformation." > > Stated in different words, it seems that this design > method comprises: > > � 1. Start with a digital filter, > � 2. Convert that digital filter to an analog > � � �(continuous) filter, > � 3. Optimize the analog filter in some way, and > � 4. Convert the analog filter back to a digital > � � �filter. > > I've never heard of this "classic" design method before. > > Have any of you heard of this IIR digital filter design > method? �I must be missing something here. >perhaps the author means that he/she: 1. starts with the *parameters* (like cutoff frequency, transition band width, allowable deviations in either the passband or stopband, whatever) of the digital filter in the domain of the digital filter. 2. knowing Fs or T (or assuming it's normalized), covert those parameters to the corresponding parameters in the analog filter domain. (if bilinear transform is used, this would be the "pre- warping" compensation that we do. then 3. and 4. is what the author said. other than that, Rick, i sure as hell dunno why someone would buy this product called "dehydrated water" where you add water to get water. (i sorta remember a cartoon with that concept way back when i was a kid.) or maybe a better "analog" is start with a dehydrated vegetable (or some food), reconstitute it however you do with added water, then dehydrate it again and that's your end product. come to think of it, whenever one buys orange juice (this would not be Tropicana) that is "Made from concentrate", i wonder why not just buy the frozen concentrate? oh well. enough stream-of-concentrate (err consciousness) thinking. y'all have a nice holidaze of whatever is your chosen substance. r b-j
Reply by ●December 22, 20102010-12-22
On 12/22/2010 12:40 PM, robert bristow-johnson wrote:> On Dec 22, 3:23 pm, Rick Lyons<R.Lyons@_BOGUS_ieee.org> wrote: >> >> I've been asked to review a manuscript for possible >> publication in an IEEE journal. The material concerns >> the design of IIR filters using an "Evolutionary Algorithm". >> >> In the text the author states that the "classic" (i.e., >> traditional) approach to IIR filter design is: >> >> "... a digital filter is transformed to a corresponding >> analog low-pass (LP) filter. Then, well-known LP filter >> design methods, such as Butterworth, Chebyshev Type I, >> and Chebyshev Type II, can be used to accomplish the >> design of the analog LP filter. Finally, the analog LP >> filter is transformed back to the digital filter by >> again using a bilinear transformation." >> >> Stated in different words, it seems that this design >> method comprises: >> >> 1. Start with a digital filter, >> 2. Convert that digital filter to an analog >> (continuous) filter, >> 3. Optimize the analog filter in some way, and >> 4. Convert the analog filter back to a digital >> filter. >> >> I've never heard of this "classic" design method before. >> >> Have any of you heard of this IIR digital filter design >> method? I must be missing something here. >> > > perhaps the author means that he/she: > > 1. starts with the *parameters* (like cutoff frequency, transition > band width, allowable deviations in either the passband or stopband, > whatever) of the digital filter in the domain of the digital filter. > > 2. knowing Fs or T (or assuming it's normalized), covert those > parameters to the corresponding parameters in the analog filter > domain. (if bilinear transform is used, this would be the "pre- > warping" compensation that we do. > > then 3. and 4. is what the author said. > > other than that, Rick, i sure as hell dunno why someone would buy this > product called "dehydrated water" where you add water to get water. > (i sorta remember a cartoon with that concept way back when i was a > kid.) > > or maybe a better "analog" is start with a dehydrated vegetable (or > some food), reconstitute it however you do with added water, then > dehydrate it again and that's your end product. come to think of it, > whenever one buys orange juice (this would not be Tropicana) that is > "Made from concentrate", i wonder why not just buy the frozen > concentrate? > > oh well. enough stream-of-concentrate (err consciousness) thinking. > y'all have a nice holidaze of whatever is your chosen substance.Can you ask for a clarification? R-B-J's version seems to be what the author must have meant. I don't design many IIR filters. When I do, it's for a control system, and there's a lot of hand work involved. What's left, I either use the above method, or I get desperate and try to do something by blind optimization in Scilab. -- Tim Wescott Wescott Design Services http://www.wescottdesign.com Do you need to implement control loops in software? "Applied Control Theory for Embedded Systems" was written for you. See details at http://www.wescottdesign.com/actfes/actfes.html
Reply by ●December 22, 20102010-12-22
On Dec 22, 3:23�pm, Rick Lyons <R.Lyons@_BOGUS_ieee.org> wrote:> Hi Guys, > � �I've been asked to review a manuscript for possible > publication in an IEEE journal. �The material concerns > the design of IIR filters using an "Evolutionary Algorithm". > > In the text the author states that the "classic" (i.e., > traditional) approach to IIR filter design is: > > � �"... a digital filter is transformed to a corresponding > � � analog low-pass (LP) filter. Then, well-known LP filter > � � design methods, such as Butterworth, Chebyshev Type I, > � � and Chebyshev Type II, can be used to accomplish the > � � design of the analog LP filter. Finally, the analog LP > � � filter is transformed back to the digital filter by > � � again using a bilinear transformation." > > Stated in different words, it seems that this design > method comprises: > > � 1. Start with a digital filter, > � 2. Convert that digital filter to an analog > � � �(continuous) filter, > � 3. Optimize the analog filter in some way, and > � 4. Convert the analog filter back to a digital > � � �filter. > > I've never heard of this "classic" design method before. > > Have any of you heard of this IIR digital filter design > method? �I must be missing something here. > > Thanks Much, > [-Rick-]Interesting!! If one starts with a digital filter why do a bunch of conversion work to arrive at a digital filter? Are they somehow reducing the order of the filter? Or are they taking a complicatedly designed digital filter (designed via who knows which method) and finding a lower order approximation. In terms of "classical" design that involves both digital and analog forms I can really only think starting with an analog prototype and transforming it to a digital form with the bilinear transform being one among several options. It seems like almost any method would appear better when compared to a digital - analog - digital transformation of filters. Clay
Reply by ●December 22, 20102010-12-22
On Dec 22, 2:23�pm, Rick Lyons <R.Lyons@_BOGUS_ieee.org> wrote:> Hi Guys, > � �I've been asked to review a manuscript for possible > publication in an IEEE journal. �The material concerns > the design of IIR filters using an "Evolutionary Algorithm". > > In the text the author states that the "classic" (i.e., > traditional) approach to IIR filter design is: > > � �"... a digital filter is transformed to a corresponding > � � analog low-pass (LP) filter. Then, well-known LP filter > � � design methods, such as Butterworth, Chebyshev Type I, > � � and Chebyshev Type II, can be used to accomplish the > � � design of the analog LP filter. Finally, the analog LP > � � filter is transformed back to the digital filter by > � � again using a bilinear transformation." > > Stated in different words, it seems that this design > method comprises: > > � 1. Start with a digital filter, > � 2. Convert that digital filter to an analog > � � �(continuous) filter, > � 3. Optimize the analog filter in some way, and > � 4. Convert the analog filter back to a digital > � � �filter. > > I've never heard of this "classic" design method before. > > Have any of you heard of this IIR digital filter design > method? �I must be missing something here. > > Thanks Much, > [-Rick-]Could (s)he be converting a FIR filter into an IIR filter. Maurice
Reply by ●December 22, 20102010-12-22
On Wed, 22 Dec 2010 12:58:23 -0800 (PST), Clay <clay@claysturner.com> wrote:>On Dec 22, 3:23�pm, Rick Lyons <R.Lyons@_BOGUS_ieee.org> wrote: >> Hi Guys, >> � �I've been asked to review a manuscript for possible >> publication in an IEEE journal. �The material concerns >> the design of IIR filters using an "Evolutionary Algorithm". >> >> In the text the author states that the "classic" (i.e., >> traditional) approach to IIR filter design is: >> >> � �"... a digital filter is transformed to a corresponding >> � � analog low-pass (LP) filter. Then, well-known LP filter >> � � design methods, such as Butterworth, Chebyshev Type I, >> � � and Chebyshev Type II, can be used to accomplish the >> � � design of the analog LP filter. Finally, the analog LP >> � � filter is transformed back to the digital filter by >> � � again using a bilinear transformation." >> >> Stated in different words, it seems that this design >> method comprises: >> >> � 1. Start with a digital filter, >> � 2. Convert that digital filter to an analog >> � � �(continuous) filter, >> � 3. Optimize the analog filter in some way, and >> � 4. Convert the analog filter back to a digital >> � � �filter. >> >> I've never heard of this "classic" design method before. >> >> Have any of you heard of this IIR digital filter design >> method? �I must be missing something here. >> >> Thanks Much, >> [-Rick-] > >Interesting!! If one starts with a digital filter why do a bunch of >conversion work to arrive at a digital filter? Are they somehow >reducing the order of the filter? Or are they taking a complicatedly >designed digital filter (designed via who knows which method) and >finding a lower order approximation. > >In terms of "classical" design that involves both digital and analog >forms I can really only think starting with an analog prototype and >transforming it to a digital form with the bilinear transform being >one among several options. > > >It seems like almost any method would appear better when compared to a >digital - analog - digital transformation of filters.Or (maybe reading too much into it): 1. I want a digital filter with such and such characteristics. 2. This classical analog filter would do the trick... 3. ... with some tweaking. 4. Then, convert it to the digital realm and tadaaaa. -- Rich Webb Norfolk, VA
Reply by ●December 22, 20102010-12-22
On �ro, 22 gru 2010 21:23:45 in article news:<gil4h6deoomm3pgjc1sk6oljnbi2ngnqbj@4ax.com> Rick Lyons wrote:> 1. Start with a digital filter, > 2. Convert that digital filter to an analog > (continuous) filter, > 3. Optimize the analog filter in some way, and > 4. Convert the analog filter back to a digital > filter. > > I've never heard of this "classic" design method before.Google shows that isn't unknow :( http://www.fh-sw.de/sw/fachb/et/labinfo/v/vundv/designofiir-filters.pdf The design is realized by four steps. The desired squared magnitude characteristic via frequency is defined for the digital filter, first. Then the prescription is transformed to the analogue domain by inverse bilinear transform, this is followed by an optimization procedure of the analogue frequency response. Finally the optimized result is transformed back to the digital domain yielding the desired IIR-filter. Page 20 - the same idea: http://www.commsp.ee.ic.ac.uk/~tania/teaching/dsp/Lecture%2010%20Digital%20Filter%20Design.pdf Page 5: https://engineering.purdue.edu/~bouman/ece438/lecture/module_1/1.7_dig_filter_design/1.7.3_iir_filter_design.pdf
Reply by ●December 22, 20102010-12-22
> > >Hi Guys, > I've been asked to review a manuscript for possible >publication in an IEEE journal. The material concerns >the design of IIR filters using an "Evolutionary Algorithm". > >In the text the author states that the "classic" (i.e., >traditional) approach to IIR filter design is: > > "... a digital filter is transformed to a corresponding > analog low-pass (LP) filter. Then, well-known LP filter > design methods, such as Butterworth, Chebyshev Type I, > and Chebyshev Type II, can be used to accomplish the > design of the analog LP filter. Finally, the analog LP > filter is transformed back to the digital filter by > again using a bilinear transformation." > >Stated in different words, it seems that this design >method comprises: > > 1. Start with a digital filter, > 2. Convert that digital filter to an analog > (continuous) filter, > 3. Optimize the analog filter in some way, and > 4. Convert the analog filter back to a digital > filter. > >I've never heard of this "classic" design method before. > >Have any of you heard of this IIR digital filter design >method? I must be missing something here. > >Thanks Much, >[-Rick-]I think people tend to jumble analogue vs digital with continuous time vs discrete time. There is plenty of filter optimisation done by playing with continuous time functions that will ultimately be implemented as discrete time digital filters. I am interested in where the initial digital filter comes from, though. Is that handed down by the Gods? Steve
Reply by ●December 23, 20102010-12-23
On Dec 22, 9:23�pm, Rick Lyons <R.Lyons@_BOGUS_ieee.org> wrote:> Hi Guys, > � �I've been asked to review a manuscript for possible > publication in an IEEE journal. �The material concerns > the design of IIR filters using an "Evolutionary Algorithm". > > In the text the author states that the "classic" (i.e., > traditional) approach to IIR filter design is: > > � �"... a digital filter is transformed to a corresponding > � � analog low-pass (LP) filter. Then, well-known LP filter > � � design methods, such as Butterworth, Chebyshev Type I, > � � and Chebyshev Type II, can be used to accomplish the > � � design of the analog LP filter. Finally, the analog LP > � � filter is transformed back to the digital filter by > � � again using a bilinear transformation." > > Stated in different words, it seems that this design > method comprises: > > � 1. Start with a digital filter, > � 2. Convert that digital filter to an analog > � � �(continuous) filter, > � 3. Optimize the analog filter in some way, and > � 4. Convert the analog filter back to a digital > � � �filter. > > I've never heard of this "classic" design method before.Stu(pi)dents neither learn their craft nor language these days. Add the term 'specification' in steps 1 and 2, and 'coefficients' in steps 3 and 4, and the paragraph almost makes sense. Not very well written, though. Rune
Reply by ●December 23, 20102010-12-23
On Dec 22, 9:40�pm, robert bristow-johnson <r...@audioimagination.com> wrote:> other than that, Rick, i sure as hell dunno why someone would buy this > product called "dehydrated water" where you add water to get water. > (i sorta remember a cartoon with that concept way back when i was a > kid.)I remember somebody - a Swede? - say something like "we get all this hot water as by product from our process. It's free energy to be used at will later - just freeze it for storage and heat it back up when you need it!" Rune
