Which processor for decimation?

Hi,
i have to decimate my input data (16 bit resolution) from 10Khz to 10Hz 
using a 60db/dec or 80db/dec butterworth digital filter, i need to know if 
this is possible using a standard MCU (for example 16 bit PIC) or is 
necessary to use a DSP core.
Thanks in advance,
Emiliano. 

Any processor can do this. I would choose one that has a math library
available as you may want to use floating point to make the implementation
simple.

You want a CIC filter followed by halfband compensation filter. Then your
final butterworth filter.


"Emiliano" <fiorenza.emiliano@tiscali.it> wrote in message
news:43d59238$0$345$5fc30a8@news.tiscali.it...
> Hi,
> i have to decimate my input data (16 bit resolution) from 10Khz to 10Hz
> using a 60db/dec or 80db/dec butterworth digital filter, i need to know if
> this is possible using a standard MCU (for example 16 bit PIC) or is
> necessary to use a DSP core.
> Thanks in advance,
> Emiliano.
>
>

Hi,

> Any processor can do this. I would choose one that has a math library
> available as you may want to use floating point to make the implementation
> simple.

Yes, but I have to use not too much power, I think to use a MCU such a 16 
bit PIC.

> You want a CIC filter followed by halfband compensation filter. Then your
> final butterworth filter.

I found this document about CIC filter:
http://www.embedded.com/showArticle.jhtml?articleID=160400592

CIC filter use X(n-d) so in my application I have to store more then 1000 
samples in memory to implement this filter? In this case a PIC MCU is not 
able to store large number of data in memory.
I don't understand why is not more easy to use a simple butterworth filter 
for decimation, I know that in this case I have to compute more multiply. My 
second question is: is possible to have flat phase response (or quasi-flat) 
using a butterworth/elliptical digital filter?
Thanks, and sorry for my English,
Emiliano.

"Emiliano" <fiorenza.emiliano@tiscali.it> writes:

> Hi,
> i have to decimate my input data (16 bit resolution) from 10Khz to 10Hz 
> using a 60db/dec or 80db/dec butterworth digital filter, i need to know if 
> this is possible using a standard MCU (for example 16 bit PIC) or is 
> necessary to use a DSP core.
> Thanks in advance,
> Emiliano. 

Emiliano,

You are mis-specifying this filter. An IIR filter (which is what a digital
Butterworth filter is) cannot take advantage of polyphase decomposition
due to the requirement to compute the intermediate internal states. 

However, I used Mitra's section on windowed FIR filter order
estimation (see table 7.2 in chapter 7) to estimate that it would
require about 7000 taps for a passband edge frequency of 9 Hz, a
stopband edge frequency of 11 Hz and a minimum stopband attenuation of
75 dB (using the Blackman window). 

However, if you use a polyphase decomposition of this filter, this
reduces to 7 taps per sample. At 10 Hz, that's about 70 multiplies and
70 adds per second - pretty easy. And, this is a MUCH better filter
than the one you specified. It would take a significant amount of
storage (7K) for the taps, however.

This is probably not the most efficient implementation, either: a
multistage approach would probably be even less computations, but it
goes to show that the goal is easily accomplished with a
low-horsepower processor like a PIC.
-- 
%  Randy Yates                  % "Bird, on the wing,
%% Fuquay-Varina, NC            %   goes floating by
%%% 919-577-9882                %   but there's a teardrop in his eye..."
%%%% <yates@ieee.org>           % 'One Summer Dream', *Face The Music*, ELO
http://home.earthlink.net/~yatescr

Randy Yates <yates@ieee.org> writes:
> [...]
> However, if you use a polyphase decomposition of this filter, this
> reduces to 7 taps per sample. At 10 Hz, that's about 70 multiplies and
> 70 adds per second - pretty easy. 

Sorry, Emiliano. This is NOT correct. It is about 70,000 multiplies
and 70,000 adds per second. Still well within reach of a PIC, but not
nearly as efficient as I stated.
-- 
%  Randy Yates                  % "So now it's getting late,
%% Fuquay-Varina, NC            %    and those who hesitate
%%% 919-577-9882                %    got no one..."
%%%% <yates@ieee.org>           % 'Waterfall', *Face The Music*, ELO
http://home.earthlink.net/~yatescr