Forums

CIC filter question

Started by AAA January 17, 2006
I try to understand how CIC decimator works in presence of DC offset of any
kind at its input? Filter of that tipe passes through all low frequencies
including zero, but integrator stages may be overflowed by DC or low
frequencies. How to prevent this in practice?

Excuse me my english, please help!
Thank you in advance.
Alex.




AAA wrote:
> I try to understand how CIC decimator works in presence of DC offset of any > kind at its input? Filter of that tipe passes through all low frequencies > including zero, but integrator stages may be overflowed by DC or low > frequencies. How to prevent this in practice? > > Excuse me my english, please help! > Thank you in advance. > Alex.
CIC filters use modulo arithmetic. The integrators can overflow without problem, as long as they have enough bits that each increment is less than the size of the integrator. Regards, John McCaskill

AAA wrote:

> I try to understand how CIC decimator works in presence of DC offset of any > kind at its input? Filter of that tipe passes through all low frequencies > including zero, but integrator stages may be overflowed by DC or low > frequencies. How to prevent this in practice? >
As long as you are using the same width nonsaturating ariphmetics in the comb and in the integrator sections, the intermediate overflows are mutually canceled. You don't have to worry about the intermediate overflows if the expected output result can be represented with the given number of bits. Vladimir Vassilevsky DSP and Mixed Signal Design Consultant http://www.abvolt.com
> >AAA wrote: >> I try to understand how CIC decimator works in presence of DC offset of
any
>> kind at its input? Filter of that tipe passes through all low
frequencies
>> including zero, but integrator stages may be overflowed by DC or low >> frequencies. How to prevent this in practice? >> >> Excuse me my english, please help! >> Thank you in advance. >> Alex. > > > >CIC filters use modulo arithmetic. The integrators can overflow >without problem, as long as they have enough bits that each increment >is less than the size of the integrator. > >Regards, > >John McCaskill > >
Hi, John! Thank you for reply! I think what will happen with the comb output at the very overflow moment? Even though each increment may be much less than the size of the integrator, its output will "rotate" by a full scale or so and will harm the comb's output. Isn't it? Would you like to explane me this in some details. I have learned about this integrator's property but didn't undestand it. Thank you. Alex.
Hi Alex,

Rick Lyons had an article in some on-line magazine
somewhere on CICs which is well-worth reading, but
I forgot where it was. Maybe someone can give the
pointer? 

--Randy

"Randy Yates" <yates@ieee.org> wrote in message
news:1137521663.277432.123680@o13g2000cwo.googlegroups.com...
> Hi Alex, > > Rick Lyons had an article in some on-line magazine > somewhere on CICs which is well-worth reading, but > I forgot where it was. Maybe someone can give the > pointer? > > --Randy
Here you go http://www.eetimes.com/showArticle.jhtml?articleID=160400592 The complete thread on comp.dsp which includes some discussion on overflow in the CIC is here http://tinyurl.com/c4uzb Cheers Bhaskar
Randy Yates wrote:

> Rick Lyons had an article in some on-line magazine > somewhere on CICs which is well-worth reading, but > I forgot where it was. Maybe someone can give the > pointer?
I can't find it either, but here's another intro article: http://users.snip.net/~donadio/cic.pdf Martin -- Quidquid latine scriptum sit, altum viditur.
Randy Yates wrote:
> Hi Alex, > > Rick Lyons had an article in some on-line magazine > somewhere on CICs which is well-worth reading, but > I forgot where it was. Maybe someone can give the > pointer?
The link is at http://www.dspguru.com/info/tutor/cic.htm, -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
>Randy Yates wrote: >> Hi Alex, >> >> Rick Lyons had an article in some on-line magazine >> somewhere on CICs which is well-worth reading, but >> I forgot where it was. Maybe someone can give the >> pointer? > >The link is at http://www.dspguru.com/info/tutor/cic.htm, >-- >Engineering is the art of making what you want from things you can get. >&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095; >
Thank you all! I need some time to think about... Alex.
junkmail@fastertechnology.com wrote:
> AAA wrote: > >>I try to understand how CIC decimator works in presence of DC offset of any >>kind at its input? Filter of that tipe passes through all low frequencies >>including zero, but integrator stages may be overflowed by DC or low >>frequencies. How to prevent this in practice? >> >>Excuse me my english, please help! >>Thank you in advance. >>Alex. > > > > > CIC filters use modulo arithmetic. The integrators can overflow > without problem, as long as they have enough bits that each increment > is less than the size of the integrator. > > Regards, > > John McCaskill >
This assumes fixed point processing, correct? Furthermore, I think it assumes twos-complement representation. I don't think a Hogenaur(sp?) representation of a CIC filter is appropriate in floating point. Anyone disagree? -- Mark Borgerding