Randy Yates wrote:> It's a damn good article - one of the best I've ever read within the > class of engineering magazine articles.Hear, hear! I found a typesetting error, though. Somewhere near the bottom, F_s turned into >_s. I figures it out, though! :-) Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
CIC Group Delay
Started by ●October 12, 2005
Reply by ●October 14, 20052005-10-14
Reply by ●October 14, 20052005-10-14
Rick Lyons wrote:> On Tue, 11 Oct 2005 22:00:37 -0700, Bob Cain >>What is a CIC filter? >> >>Bob > > Hi, > ya' have a look at: > > http://www.embedded.com/shared/printableArticle.jhtml?articleID=160400592Thanks, Rick. Excellent article. Bob -- "Things should be described as simply as possible, but no simpler." A. Einstein
Reply by ●October 16, 20052005-10-16
Thanks, Rick. That is a very helpful article. (Jerry, you didn't catch
the missing minus sign in the Euler's identity preceding equation 9?).
I have also been refering to Harris's Multirate Signal Processing.
My simulation results suggest I will need more supression around the
zeros of the filter before I decimate. Otherwise, the residual around
that zero will fold into my band of interest at a level I can't
tolerate.
I am considering cascading (I think this is called "sharpening"?)...
Hk(Z)=[(1-Z^-M)/(1-Z^-1)]^K
But I don't see how to take advantage of the Noble Identity that is
obtained in the Hogenauer method. The advantage I am refering to is the
reduction of memory requirements by decimating immediately after the
integrator. It's not a big deal, since the cost of a few registers
isn't too high, but I would like to understand what optimizations I can
obtain while cascading.
Jim
Reply by ●October 19, 20052005-10-19
On 14 Oct 2005 07:13:54 -0700, "Randy Yates" <yates@ieee.org> wrote:>It's a damn good article - one of the best I've ever read within the >class of engineering magazine articles. > >--RandyHi Randy, Thanks. Actually, Ray Andraka help me an awful lot with the text of that article. He's a true "CIC expert". Wonder what's "happening" with Ray. [-Rick-]
Reply by ●October 19, 20052005-10-19
On Fri, 14 Oct 2005 11:10:01 -0400, Jerry Avins <jya@ieee.org> wrote:>Randy Yates wrote: >> It's a damn good article - one of the best I've ever read within the >> class of engineering magazine articles. > >Hear, hear! I found a typesetting error, though. Somewhere near the >bottom, F_s turned into >_s. I figures it out, though! :-) > >JerryHi Jerry, yea, I see that typo now. That ">_s" is supposed to be "Fs,in" where the "s,in" characters are subscripted. My original manuscript didn't have that error. (It makes me think that all italicized characters in a manuscript must be "re-typed" by some magazine production person.) See Ya', [-Rick-]
Reply by ●October 19, 20052005-10-19
On Fri, 14 Oct 2005 11:10:26 -0700, Bob Cain <arcane@arcanemethods.com> wrote:> > >Rick Lyons wrote: >> On Tue, 11 Oct 2005 22:00:37 -0700, Bob Cain >>>What is a CIC filter? >>> >>>Bob >> >> Hi, >> ya' have a look at: >> >> http://www.embedded.com/shared/printableArticle.jhtml?articleID=160400592 > >Thanks, Rick. Excellent article. > > >BobHi Bob, Thanks for sayin' "Thanks". [-Rick-]
Reply by ●October 19, 20052005-10-19
On 16 Oct 2005 10:04:10 -0700, jim_nospam_beasley@yahoo.com wrote:>Thanks, Rick.Hi Jim, you're most welcome.>That is a very helpful article. (Jerry, you didn't catch >the missing minus sign in the Euler's identity preceding equation 9?).I also see that typo now that you point it out. My original manuscript didn't have that error. Again, it makes me think that all italicized characters in a manuscript must be "re-typed" by some magazine production person.>I have also been refering to Harris's Multirate Signal Processing.Good idea. Harris a "Master of Sample Rate Change".>My simulation results suggest I will need more supression around the >zeros of the filter before I decimate. Otherwise, the residual around >that zero will fold into my band of interest at a level I can't >tolerate. > >I am considering cascading (I think this is called "sharpening"?)... > > Hk(Z)=[(1-Z^-M)/(1-Z^-1)]^KYep, cascaded CIC filters give increased attenuation around the zeros. Commercial chips that use CIC filters often have the capability to cascade 3-5 CIC filters. (However, such simple cascading isn't called "sharpening", at least not that I know of. There a filtering technique called "filter sharpening", but it's not just simple cascading of some single filter. Matt Donadio wrote an article about this for my "DSP Tips & Tricks" column in the Sept., 2003, issue of the IEEE Sig. Proc. magazine.)>But I don't see how to take advantage of the Noble Identity that is >obtained in the Hogenauer method. The advantage I am refering to is the >reduction of memory requirements by decimating immediately after the >integrator. It's not a big deal, since the cost of a few registers >isn't too high, but I would like to understand what optimizations I can >obtain while cascading.Humm, I'm guessing that you're talking about the N = D/R shown in Figure 9 of my article (although, I'm not sure). Remember, in most applications that I've seen D = R, so N = 1. So instead of needing D storage locations for the comb filter, you only need one location. Good Luck Jim. See Ya', [-Rick-]
Reply by ●October 19, 20052005-10-19
jim_nospam_beasley@yahoo.com wrote:> But I don't see how to take advantage of the Noble Identity that > is obtained in the Hogenauer method. The advantage I am refering > to is the reduction of memory requirements by decimating > immediately after the integrator.If your word width allows for multiple integration, you could push the comb filters in at least the last few CIC blocks to the end of the chain and run them all at the output sample rate. Martin -- Quidquid latine dictum sit, altum viditur.
Reply by ●October 19, 20052005-10-19
On 19 Oct 2005 20:24:54 GMT, Martin Eisenberg <martin.eisenberg@udo.edu> wrote:>jim_nospam_beasley@yahoo.com wrote: > >> But I don't see how to take advantage of the Noble Identity that >> is obtained in the Hogenauer method. The advantage I am refering >> to is the reduction of memory requirements by decimating >> immediately after the integrator. > >If your word width allows for multiple integration, you could push >the comb filters in at least the last few CIC blocks to the end of >the chain and run them all at the output sample rate. > > >Martin > >-- >Quidquid latine dictum sit, altum viditur.Hi Martin, just out of curiosity, what does "Quidquid latine dictum sit, altum viditur." mean in English? Thanks, [-Rick-]
Reply by ●October 20, 20052005-10-20
Rick Lyons wrote:> On 19 Oct 2005 20:24:54 GMT, Martin Eisenberg > <martin.eisenberg@udo.edu> wrote: > > >>jim_nospam_beasley@yahoo.com wrote: >> >> >>>But I don't see how to take advantage of the Noble Identity that >>>is obtained in the Hogenauer method. The advantage I am refering >>>to is the reduction of memory requirements by decimating >>>immediately after the integrator. >> >>If your word width allows for multiple integration, you could push >>the comb filters in at least the last few CIC blocks to the end of >>the chain and run them all at the output sample rate. >> >> >>Martin >> >>-- >>Quidquid latine dictum sit, altum viditur. > > > Hi Martin, > just out of curiosity, what does > > "Quidquid latine dictum sit, altum viditur." > > mean in English?"Whatever is said in Latin, seems profound." But I prefer "Quidquid latine dictum sit, altum sonatur" which means "Whatever is said in Latin, sounds profound." -- Jim Thomas Principal Applications Engineer Bittware, Inc jthomas@bittware.com http://www.bittware.com (603) 226-0404 x536 Never ascribe to malice that which is adequately explained by incompetence. - Napoleon Bonaparte
