dbd wrote:> On Jun 13, 8:32 am, Vladimir Vassilevsky <antispam_bo...@hotmail.com> > wrote: > >>There are systematic methods for designing FIR filter to an arbitrary >>response, and the FIR doesn't have to be linear phase. There are no such >>methods for IIRs. Hence in some cases the non-symmetric FIR could be >>very convenient. >> > > See > author = "Ricardo A. Vargas and C. Sidney Burrus", > title = "On the Design of L_p IIR Filters with Arbitrary Frequency > Response", > url = "citeseer.ist.psu.edu/663472.html" } > > And Matlab's iirlpnorm >Those are the "adaptive iterative" methods of tomtomes and tamburines. It is about finding the global maximum of a function of many variables by some sort of brute force search. I wouldn't call this a deterministic algorithm. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
non symmetric filter using park mcclelans
Started by ●June 12, 2008
Reply by ●June 13, 20082008-06-13
Reply by ●June 13, 20082008-06-13
On Fri, 13 Jun 2008 07:39:43 -0700, "Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote:> >"Eric Jacobsen" <eric.jacobsen@ieee.org> wrote in message >news:var2545aoc7gjtip8pdoh4btb5grphelf5@4ax.com... >> On Thu, 12 Jun 2008 02:12:09 -0500, "chivak" <cd_prasad@hotmail.com> >> http://www.dspguru.com/comp.dsp/tricks/dsn/nlp_fir.htm >> >> Eric Jacobsen >> Minister of Algorithms >> Abineau Communications >> http://www.ericjacobsen.org >> >> Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php > >Eric, > >I'm embarrassed to note that I'd seen *dsprelated* mentioned many times but >had never checked it out. Very nice! Was it done from scratch or did you >use a tool or a service provider / toolset? >...> >Fred >Fred, DSPRelated isn't mine, but Stephane Boucher's. I'm just an invited blogger there, so I can't take credit/blame for anything but my articles. Credit for the site or questions regarding it should be directed to Stephane. Eric Jacobsen Minister of Algorithms Abineau Communications http://www.ericjacobsen.org Blog: http://www.dsprelated.com/blogs-1/hf/Eric_Jacobsen.php
Reply by ●June 13, 20082008-06-13
On Jun 13, 11:56 am, Vladimir Vassilevsky <antispam_bo...@hotmail.com> wrote:> >... > Those are the "adaptive iterative" methods of tomtomes and tamburines. > It is about finding the global maximum of a function of many variables > by some sort of brute force search. I wouldn't call this a deterministic > algorithm. > > Vladimir Vassilevsky > DSP and Mixed Signal Design Consultanthttp://www.abvolt.comWe are talking about 'arbitrary response', not 'tomtomes and tamburines'. In what sense do you think Parks McClellan is a deterministic algorithm? How well would it do if it didn't search after making an initial estimate? Dale B. Dalrymple
Reply by ●June 14, 20082008-06-14
"dbd" <dbd@ieee.org> wrote in message news:18a5e59a-46dc-466a-b9f3-cf69d5298727@a9g2000prl.googlegroups.com...> On Jun 13, 11:56 am, Vladimir Vassilevsky <antispam_bo...@hotmail.com> > wrote: >> >>... >> Those are the "adaptive iterative" methods of tomtomes and tamburines. >> It is about finding the global maximum of a function of many variables >> by some sort of brute force search. I wouldn't call this a deterministic >> algorithm. >> >> Vladimir Vassilevsky >> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com > > We are talking about 'arbitrary response', not 'tomtomes and > tamburines'. > > In what sense do you think Parks McClellan is a deterministic > algorithm? How well would it do if it didn't search after making an > initial estimate? > > Dale B. DalrympleParks-McClellan is deterministic in that it finds the global optimum and is guaranteed to converge - subject to numerical roundoff issues with any particular implementation. Theoretically it converges. Thus, it's not a "search" in the classical sense but, rather, iterative. It's very closely related to linear programming. On the other hand, search methods that have no such capability can only do their best to find an optimum. Often what's found is a local optimum. Or, exhaustive search can be used to find a global optimum. Now *that's* a search! Fred
Reply by ●June 14, 20082008-06-14
On Jun 14, 1:20 pm, "Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote:> "dbd" <d...@ieee.org> wrote in message > > news:18a5e59a-46dc-466a-b9f3-cf69d5298727@a9g2000prl.googlegroups.com... > > > > > On Jun 13, 11:56 am, Vladimir Vassilevsky <antispam_bo...@hotmail.com> > > wrote: > > >>... > >> Those are the "adaptive iterative" methods of tomtomes and tamburines. > >> It is about finding the global maximum of a function of many variables > >> by some sort of brute force search. I wouldn't call this a deterministic > >> algorithm. > > >> Vladimir Vassilevsky > >> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com > > > We are talking about 'arbitrary response', not 'tomtomes and > > tamburines'. > > > In what sense do you think Parks McClellan is a deterministic > > algorithm? How well would it do if it didn't search after making an > > initial estimate? > > > Dale B. Dalrymple > > Parks-McClellan is deterministic in that it finds the global optimum and is > guaranteed to converge - subject to numerical roundoff issues with any > particular implementation. Theoretically it converges. Thus, it's not a > "search" in the classical sense but, rather, iterative. It's very closely > related to linear programming. > > On the other hand, search methods that have no such capability can only do > their best to find an optimum. Often what's found is a local optimum. Or, > exhaustive search can be used to find a global optimum. Now *that's* a > search! > > Fred
Reply by ●June 14, 20082008-06-14
On Jun 14, 1:20 pm, "Fred Marshall" <fmarshallx@remove_the_x.acm.org> wrote:> "dbd" <d...@ieee.org> wrote in message > > news:18a5e59a-46dc-466a-b9f3-cf69d5298727@a9g2000prl.googlegroups.com... > > > > > On Jun 13, 11:56 am, Vladimir Vassilevsky <antispam_bo...@hotmail.com> > > wrote: > > >>... > >> Those are the "adaptive iterative" methods of tomtomes and tamburines. > >> It is about finding the global maximum of a function of many variables > >> by some sort of brute force search. I wouldn't call this a deterministic > >> algorithm. > > >> Vladimir Vassilevsky > >> DSP and Mixed Signal Design Consultanthttp://www.abvolt.com > > > We are talking about 'arbitrary response', not 'tomtomes and > > tamburines'. > > > In what sense do you think Parks McClellan is a deterministic > > algorithm? How well would it do if it didn't search after making an > > initial estimate? > > > Dale B. Dalrymple > > Parks-McClellan is deterministic in that it finds the global optimum and is > guaranteed to converge - subject to numerical roundoff issues with any > particular implementation. Theoretically it converges. Thus, it's not a > "search" in the classical sense but, rather, iterative. It's very closely > related to linear programming. > > On the other hand, search methods that have no such capability can only do > their best to find an optimum. Often what's found is a local optimum. Or, > exhaustive search can be used to find a global optimum. Now *that's* a > search! > > FredI agree that Parks-McClellan is deterministic in that it's search is not random and it finds an optimum in the polynomial search space it operates in. The approach in the article I cited is also not a random search. Because it does not substitute a polynomial search space in designing an 'arbitrary response' it is also a more 'brute force' approach. But they both design useful 'arbitrary response' filters. Dale B. Dalrymple
Reply by ●June 14, 20082008-06-14
>On Thu, 12 Jun 2008 02:12:09 -0500, "chivak" <cd_prasad@hotmail.com> >wrote: > >> >> Greetings All, >> >> Anybody know how to implement non symmetric filter with park >>mcclelans. >> >> For instance, the firpm() (as far as my knowledge goes) outputssymetric>>filters with even or odd symmetry depending on the Order of the filter. >> >> Any suggestion or link is appreciated. >> >> Thanx, >> chivak > >Hi, > I was wonderin'. Do you really want a >non-symmetric (nonlinear phase) FIR filter? >I ask that because the advantage of FIR filters >is that they are linear phase when their >coefficients are symmetric. The disadvantage >of FIR filters is that they require more computations >(per filter output sample) than nonlinear phase >IIR filters. > >Wouldn't it be better for you to use an IIR filter? >Just wonderin'. > >[-Rick-] >Thanx for the suggestion. You are correct about using an IIR filter like an all pass filter which is what some people use for phase equalization. I am trying to build a programmable filter which comprises of an FIR ,a CIC compensation and equalizer. Hence , i thought using an FIR would be more easy on the implementation side. -Chivak
Reply by ●June 15, 20082008-06-15
On Jun 12, 2:48�pm, Vladimir Vassilevsky <antispam_bo...@hotmail.com> wrote:> Eric Jacobsen wrote: > >> �Anybody know how to implement non symmetric filter with park > >>mcclelans. > > >http://www.dspguru.com/comp.dsp/tricks/dsn/nlp_fir.htm > > Designing Re and Im parts separately is the neat trick, however the > resultant filter will be neither equiripple nor optimal.oh, come on, Vlad! there's the friggin' triangle inequality. so define your ripple tolerance for the Re and Im parts to be 1/2 or maybe 1/sqrt(2) of the ripple in the mag. so it's not quite optimal. it at least can be designed to restrain the max ripple error to a known quantity. this nit is too small to pick. r b-j
