Forums

Yet another article claiming Goertzel k must be integer

Started by Jon Harris January 16, 2004
http://www.embedded.com/showArticle.jhtml?articleID=17301593


On Fri, 16 Jan 2004 10:49:24 -0800, "Jon Harris"
<goldentully@hotmail.com> wrote:

>http://www.embedded.com/showArticle.jhtml?articleID=17301593 >
Hi Jon, could you expand a little upon what is the point you're making?: Thanks, [-Rick-]
On Fri, 16 Jan 2004 10:49:24 -0800, "Jon Harris"
<goldentully@hotmail.com> wrote:

>http://www.embedded.com/showArticle.jhtml?articleID=17301593 > >
Ah, I see what you're saying now. The Embedded Sys Mgazine has had two recent DSP articles, but the quality of those articles hasn't been good. The magazine doesn't have an editor that's well-versed in DSP theory. Those kinds of editors are hard to find. [-Rick-]
That's it basically.  Sorry I was so brief.  I was referring to the fact that
there have been several articles written about the Goertzel algorithm that
either imply or state outright that the "K" parameter much be an integer.  As
we've discussed here on comp.dsp, this simply isn't the case.  This myth seems
to continue to propagate, probably as new authors read the previous
publications.

Rick Lyons <r.lyons@REMOVE.ieee.org> wrote in message
news:400aaeba.98391468@news.west.earthlink.net...
> On Fri, 16 Jan 2004 10:49:24 -0800, "Jon Harris" > <goldentully@hotmail.com> wrote: > > >http://www.embedded.com/showArticle.jhtml?articleID=17301593 > > > > > > Ah, I see what you're saying now. > > The Embedded Sys Mgazine has had two > recent DSP articles, but the quality > of those articles hasn't been > good. The magazine doesn't have an editor > that's well-versed in DSP theory. > Those kinds of editors are hard to find. > > [-Rick-]
Hi,
  yep, I agree with you.  I fell into that 
"k must be an integer" trap and stayed there for a 
long time until (thankfully) you guys straightened 
me out.  I mentioned the "k need not be an 
integer" idea in both the 2nd edition of my 
DSP book and in a very short article in the 
Jan. 2004 issue of the IEEE Sig. Processing 
magazine.  So I've done might part to dispel 
the "myth".

What I found interesting about Goertzel is that 
the frequency response of the Goertzel filter 
(the filter's structure) and the frequency 
response of the Goertzel algorithm are 
not the same.

See Ya',
[-Rick-]  

------------------------------------------------
On Sun, 18 Jan 2004 20:50:32 -0800, "Jon Harris"
<goldentully@hotmail.com> wrote:

>That's it basically. Sorry I was so brief. I was referring to the fact that >there have been several articles written about the Goertzel algorithm that >either imply or state outright that the "K" parameter much be an integer. As >we've discussed here on comp.dsp, this simply isn't the case. This myth seems >to continue to propagate, probably as new authors read the previous >publications. > >Rick Lyons <r.lyons@REMOVE.ieee.org> wrote in message
>> >> Ah, I see what you're saying now. >> >> The Embedded Sys Mgazine has had two >> recent DSP articles, but the quality >> of those articles hasn't been >> good. The magazine doesn't have an editor >> that's well-versed in DSP theory. >> Those kinds of editors are hard to find. >> >> [-Rick-] > >
Guys,

I didn't see the article in question, but I believe if they are talking
about a "sliding Goertzel", then k must be an integer, otherwise you don't
get a good zero/pole cancellation and the filter described by the "sliding"
approach will go unstable.  By sliding I mean, after collecting N initial
samples, you can get a new Goertzel output every sample.

I think I saw a sliding approach described in the March 2003 issue of IEEE
Signal Processing magazine.

I've used the fractional k for a DTMF decoder that was running at a
nonstandard sample rate, it let me place the bin EXACTLY where I needed it.

-Shawn

"Rick Lyons" <r.lyons@REMOVE.ieee.org> wrote in message
news:400bc326.10250203@news.west.earthlink.net...
> > Hi, > yep, I agree with you. I fell into that > "k must be an integer" trap and stayed there for a > long time until (thankfully) you guys straightened > me out. I mentioned the "k need not be an > integer" idea in both the 2nd edition of my > DSP book and in a very short article in the > Jan. 2004 issue of the IEEE Sig. Processing > magazine. So I've done might part to dispel > the "myth". > > What I found interesting about Goertzel is that > the frequency response of the Goertzel filter > (the filter's structure) and the frequency > response of the Goertzel algorithm are > not the same. > > See Ya', > [-Rick-] > > ------------------------------------------------ > On Sun, 18 Jan 2004 20:50:32 -0800, "Jon Harris" > <goldentully@hotmail.com> wrote: > > >That's it basically. Sorry I was so brief. I was referring to the fact
that
> >there have been several articles written about the Goertzel algorithm
that
> >either imply or state outright that the "K" parameter much be an integer.
As
> >we've discussed here on comp.dsp, this simply isn't the case. This myth
seems
> >to continue to propagate, probably as new authors read the previous > >publications. > > > >Rick Lyons <r.lyons@REMOVE.ieee.org> wrote in message > > >> > >> Ah, I see what you're saying now. > >> > >> The Embedded Sys Mgazine has had two > >> recent DSP articles, but the quality > >> of those articles hasn't been > >> good. The magazine doesn't have an editor > >> that's well-versed in DSP theory. > >> Those kinds of editors are hard to find. > >> > >> [-Rick-] > > > > >
On Tue, 20 Jan 2004 08:51:40 -0600, "Shawn Steenhagen"
<shawn.steenhagen@NOSPAMappliedsignalprocessing.com> wrote:

>Guys,
Hi Shawn,
>I didn't see the article in question, but I believe if they are talking >about a "sliding Goertzel", then k must be an integer, otherwise you don't >get a good zero/pole cancellation and the filter described by the "sliding" >approach will go unstable. By sliding I mean, after collecting N initial >samples, you can get a new Goertzel output every sample.
You're right. By the way, the article was primarily on the "Chirp Z Transform" but the author mentioined a few things about the standard Goertzel algorithm (not the sliding Goertzel).
>I think I saw a sliding approach described in the March 2003 issue of IEEE >Signal Processing magazine.
You most certainly did, & it was written by two of your DSP pals here on comp.dsp. Hey Shawn, could I bug you to take a look at a Jan. 19, 2004 post here on comp.dsp? It's Subject: Important! Do you want more DSP articles?
>I've used the fractional k for a DTMF decoder that was running at a >nonstandard sample rate, it let me place the bin EXACTLY where I needed it.
Ha. Great. [-Rick-]
Hi Rick,

Thank you for asking me to post for job in sacramento but I m H1B
status so i do not qualify.I am DSP software engineer mainly done
integration of algorithms in assembly language(ADSP-21x81).
Meanwhile I want to do good use of my time could you suggest projects
i can do at home (I have matlab) that i can add on my resume also if
any suggested reading.
Regards,
Asha 
r.lyons@REMOVE.ieee.org (Rick Lyons) wrote in message news:<400e66c1.183214140@news.west.earthlink.net>...
> On Tue, 20 Jan 2004 08:51:40 -0600, "Shawn Steenhagen" > <shawn.steenhagen@NOSPAMappliedsignalprocessing.com> wrote: > > >Guys, > > Hi Shawn, > > >I didn't see the article in question, but I believe if they are talking > >about a "sliding Goertzel", then k must be an integer, otherwise you don't > >get a good zero/pole cancellation and the filter described by the "sliding" > >approach will go unstable. By sliding I mean, after collecting N initial > >samples, you can get a new Goertzel output every sample. > > You're right. By the way, the article was > primarily on the "Chirp Z Transform" but the author > mentioined a few things about the standard Goertzel > algorithm (not the sliding Goertzel). > > >I think I saw a sliding approach described in the March 2003 issue of IEEE > >Signal Processing magazine. > > You most certainly did, & it was written by > two of your DSP pals here on comp.dsp. > Hey Shawn, could I bug you to take a look > at a Jan. 19, 2004 post here on comp.dsp? It's > > Subject: Important! Do you want more DSP articles? > > >I've used the fractional k for a DTMF decoder that was running at a > >nonstandard sample rate, it let me place the bin EXACTLY where I needed it. > > Ha. Great. > > [-Rick-]
On 23 Jan 2004 09:15:37 -0800, asha_rathni@yahoo.com (Asha Rathni)
wrote:

>Hi Rick, > >Thank you for asking me to post for job in sacramento but I m H1B >status so i do not qualify.I am DSP software engineer mainly done >integration of algorithms in assembly language(ADSP-21x81). >Meanwhile I want to do good use of my time could you suggest projects >i can do at home (I have matlab) that i can add on my resume also if >any suggested reading. >Regards, >Asha
Hello Asha Whew. I'm not sure what to say. If you have MATLAB at home, start using it to design various filters, perform spectrum analysis on signals, model various DSP algorithms (like Goertzel and the sliding DFT from this thread). Search the Internet for "MATLAB exercises" & "MATLAB assignments". That'll take you to college classes (from various universities) that have MATLAB homework problems, and hopefully their solutions. With lots of practice with MATLAB, you can tell your perspective employer that you have assembly language experience and "MATLAB modeling" experience. I like the MATLAB code from Dr. Bonnie Heck (Georgia Tech) at: http://users.ece.gatech.edu/~bonnie/book/M-files.html Have you seen the following website? http://wwweng.uwyo.edu/electrical/dsp_audio/ Asha, have you seen this webpage: http://www.redcedar.com/learndsp.htm Employers don't care what books you've read, but they might care what DSP-specific training you've had. There's a very good DSP class for engineers being held in Sunnyvale California in June of this year. (Hopefully you'll be working by then.) How strong are you in the theory of DSP, ie., filters, spectrum analysis, understanding and mitigating the effects of finite-word widths, etc.? Do you know all the theory that you'd like to know at this time? The second edition of a good DSP book for beginners is coming out in March/April of this year by a guy named Lyons. ;-) Asha, you should post a message to all the guys here asking something like "How do I prepare for a DSP job?" Tell of your experience and then ask for advice. These guys have lots of experience and sensible advice. Best of Luck Asha, [-Rick-]
Hi Rick ,

Thank you for the website info.
I m suprised ,fact is I have already bought your book!In fact I
prepared for some  portions  of  the brainbench certification from
this book
I did not know it was you.
    Understanding DSP by Richard G Lyons .Like u suggested I will
start working on Matlab excercises and hopefully I will get a job
soon.


Bye,
Asha Rathni



r.lyons@REMOVE.ieee.org (Rick Lyons) wrote in message news:<40141ffd.125207718@news.west.earthlink.net>...
> On 23 Jan 2004 09:15:37 -0800, asha_rathni@yahoo.com (Asha Rathni) > wrote: > > >Hi Rick, > > > >Thank you for asking me to post for job in sacramento but I m H1B > >status so i do not qualify.I am DSP software engineer mainly done > >integration of algorithms in assembly language(ADSP-21x81). > >Meanwhile I want to do good use of my time could you suggest projects > >i can do at home (I have matlab) that i can add on my resume also if > >any suggested reading. > >Regards, > >Asha > > Hello Asha > > Whew. I'm not sure what to say. > > If you have MATLAB at home, start using it > to design various filters, perform > spectrum analysis on signals, model various DSP > algorithms (like Goertzel and the sliding > DFT from this thread). Search the Internet > for "MATLAB exercises" & "MATLAB assignments". > That'll take you to college classes (from > various universities) that have MATLAB > homework problems, and hopefully their > solutions. With lots of practice with MATLAB, > you can tell your perspective employer that > you have assembly language experience and > "MATLAB modeling" experience. > > I like the MATLAB code from Dr. Bonnie Heck > (Georgia Tech) at: > > http://users.ece.gatech.edu/~bonnie/book/M-files.html > > Have you seen the following website? > > http://wwweng.uwyo.edu/electrical/dsp_audio/ > > > Asha, have you seen this webpage: > http://www.redcedar.com/learndsp.htm > > Employers don't care what books you've read, > but they might care what DSP-specific training > you've had. There's a very good DSP class for > engineers being held in Sunnyvale California > in June of this year. (Hopefully you'll be working > by then.) > > How strong are you in the theory of DSP, ie., > filters, spectrum analysis, understanding and > mitigating the effects of finite-word widths, etc.? > Do you know all the theory that you'd like to know > at this time? The second edition of a good DSP book > for beginners is coming out in March/April of this > year by a guy named Lyons. ;-) > > Asha, you should post a message to all the guys > here asking something like "How do I prepare for > a DSP job?" Tell of your experience and then > ask for advice. These guys have lots of > experience and sensible advice. > > Best of Luck Asha, > [-Rick-]