How to implement an efficient ALL-PASS decimation?

Hi all:
I want to implement a Decimation with D=6 to decimate from 5Mhz to 
833Khz, but all the 833Khz content is with interest, so the priori 
low-pass filter seem to have a very sharp transition width, which is
hard to implement. so anyone got some ideas?

liubenyuan wrote:

> Hi all:
> I want to implement a Decimation with D=6 to decimate from 5Mhz to 
> 833Khz, but all the 833Khz content is with interest, so the priori 
> low-pass filter seem to have a very sharp transition width, which is
> hard to implement. so anyone got some ideas?

There is no silver bullet. The reasonable upper band for the decimation 
filter is no higher then 0.8...0.9 of Nyquist. So, with 833kHz sample 
rate the cutoff will be at 375kHz.

Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

liubenyuan wrote:
> Hi all:
> I want to implement a Decimation with D=6 to decimate from 5Mhz to 
> 833Khz, but all the 833Khz content is with interest, so the priori 
> low-pass filter seem to have a very sharp transition width, which is
> hard to implement. so anyone got some ideas?

You want a signal that's not oversampled at all. That's bad news.

Jerry
-- 
Engineering is the art of making what you want from things you can get.
�����������������������������������������������������������������������

liubenyuan wrote:
> Hi all:
> I want to implement a Decimation with D=6 to decimate from 5Mhz to 
> 833Khz, but all the 833Khz content is with interest, so the priori 
> low-pass filter seem to have a very sharp transition width, which is
> hard to implement. so anyone got some ideas?
> 
> 
> 

Use a long comb filter first, taking advantage of the fact that most of 
the coefficients are zero.

Then use a relatively short LP filter.  All it has to do is to attenuate 
the hgher-order responses, so can have a fairly wide transition band.

Then decimate.

Regards,
John