Forums

about cic filter problem

Started by liux...@yahoo.cn April 8, 2009
Hello all,

In the book of "understanding of digital signal processing", it is said that cic filter has a inherent problem which is cic filter can not remove the outband imgage clearly and the image will be folded into the real signal when decimation especially when stage of cic filter is very small.
But there is an integrater ahead of decimation. Is the integrater a good low pass filter enough which can remove the outband image clearly?
Hello all,
>
>In the book of "understanding of digital signal processing", it is said that cic filter has a inherent problem which is cic filter can not remove the outband imgage clearly and the image will be folded into the real signal when decimation especially when stage of cic filter is very small.
>But there is an integrater ahead of decimation. Is the integrater a good low pass filter enough which can remove the outband image clearly?

In my opinion, the cic filter constructure used for decimation is as the following description. integrater is firstly calculated and then decimate followed by comb filter. Integrater is basically a low pass filter, but it still can not filter out the outband images. The amplitude-to-frequency feature is H(w)=1/(1-z^-1). In the dc frequency, the signal can be infinite and in fs/2 feqeucncy point, signal will be attenuated by 1/2 which is only -6db. So in my opinion, integrater is not a good low pass filter which can not filter out all the outband images.
I don't know whether my conclusion is correct or not.
I really appreciate your help.
--- In m..., liuxinzhu20080904@... wrote:
>
> Hello all,
> >
> >In the book of "understanding of digital signal processing", it is
said that cic filter has a inherent problem which is cic filter can not
remove the outband imgage clearly and the image will be folded into the
real signal when decimation especially when stage of cic filter is very
small.
> >But there is an integrater ahead of decimation. Is the integrater a
good low pass filter enough which can remove the outband image clearly?
> >
> > In my opinion, the cic filter constructure used for decimation is as
the following description. integrater is firstly calculated and then
decimate followed by comb filter. Integrater is basically a low pass
filter, but it still can not filter out the outband images. The
amplitude-to-frequency feature is H(w)=1/(1-z^-1). In the dc frequency,
the signal can be infinite and in fs/2 feqeucncy point, signal will be
attenuated by 1/2 which is only -6db. So in my opinion, integrater is
not a good low pass filter which can not filter out all the outband
images.
> I don't know whether my conclusion is correct or not.
> I really appreciate your help.
>

The essence of CIC filter doesn't lie in the fact that they effectively
perform the anti aliasing but they perform anti aliasing with relatively
smaller hardware when very large decimation is required. Decimation with
FIR filters would require a very large number of taps (i.e.
multiplications) for the same job and would be practically unrealizable.
So dont consider integrator being used here as a very efficient
realization of antialiasing filter (in the alias rejection sense).
Moreover, CIC filters are suitable for narrow band low pass filtering.
This means, when aliases are folded back to the real spectrum, their
power would be reduced when the bandwidth of the signal is small.

Still, you can improve the CIC filter antialiasing attenuation by
increasing the stages i.e. order of the CIC filter. But this comes with
the penalty of additional hardware complexity and pass band droop.

Ubaid Abdullah
http://dspdotcomm.blogspot.com
Hello all,
>
>In the book of "understanding of digital signal processing", it is said that cic filter has a inherent problem which is cic filter can not remove the outband imgage clearly and the image will be folded into the real signal when decimation especially when stage of cic filter is very small.
>But there is an integrater ahead of decimation. Is the integrater a good low pass filter enough which can remove the outband image clearly?
The essence of CIC filter doesn't lie in the fact that they effectively
perform the anti aliasing but they perform anti aliasing with relatively
smaller hardware when very large decimation is required. Decimation with
FIR filters would require a very large number of taps (i.e.
multiplications) for the same job and would be practically unrealizable.
So dont consider integrator being used here as a very efficient
realization of antialiasing filter (in the alias rejection sense).
Moreover, CIC filters are suitable for narrow band low pass filtering.
This means, when aliases are folded back to the real spectrum, their
power would be reduced when the bandwidth of the signal is small.

Still, you can improve the CIC filter antialiasing attenuation by
increasing the stages i.e. order of the CIC filter. But this comes with
the penalty of additional hardware complexity and pass band droop.

Ubaid Abdullah

To Ubaid Abdullah

Thank you so much for your reply. I agree with you totally. Can I ask another question in your link http://dspdotcomm.blogspot.com
Liuxin-

>>In the book of "understanding of digital signal processing", it is said that cic filter has a inherent problem which
>> is cic filter can not remove the outband imgage clearly and the image will be folded into the real signal when
>> decimation especially when stage of cic filter is very small.
>>But there is an integrater ahead of decimation. Is the integrater a good low pass filter enough which can remove the
>> outband image clearly?
> The essence of CIC filter doesn't lie in the fact that they effectively
> perform the anti aliasing but they perform anti aliasing with relatively
> smaller hardware when very large decimation is required. Decimation with
> FIR filters would require a very large number of taps (i.e.
> multiplications) for the same job and would be practically unrealizable.
> So dont consider integrator being used here as a very efficient
> realization of antialiasing filter (in the alias rejection sense).
> Moreover, CIC filters are suitable for narrow band low pass filtering.
> This means, when aliases are folded back to the real spectrum, their
> power would be reduced when the bandwidth of the signal is small.
>
> Still, you can improve the CIC filter antialiasing attenuation by
> increasing the stages i.e. order of the CIC filter. But this comes with
> the penalty of additional hardware complexity and pass band droop.

Just want to say that's a really well-said answer.

-Jeff

> Ubaid Abdullah
>
> To Ubaid Abdullah
>
> Thank you so much for your reply. I agree with you totally. Can I ask another question in your link
> http://dspdotcomm.blogspot.com