I posted an article on Multistage decimators yesterday.  I decided it needed some tweaking, so today I updated it to show the stopband requirements in a couple of the plots.

regards,

Neil

Hi Neil,

Nice topic but you might have to tweak it more.

Normally I analyse it by direct modelling as follows:

 b1= [-1 0 9 16 9 0 -1]/32;                            % fs = 1600 Hz

 b2= [23 0 -124 0 613 1023 613 0 -124 0 23]/2048;      % fs/2 = 800 Hz

 b3= [-11 0 34 0 -81 0 173 0 -376 0 1285 2050 1285 0 -376 0 173 0 ...

        -81 0 34 0 -11]/4096;                            % fs/4 = 400 Hz

x =[zeros(1,2^12),1, zeros(1,2^12)];

y1 = filter(b1,1,x);

y1 = y1(1:2:end);

y2 = filter(b2,1,y1);

y2 = y2(1:2:end);

y3 = filter(b3,1,y2);

y3 = y3(1:2:end);

Hi Kaz,

Your code computes the impulse response of the decimator at the output sample rate of fs/8.  My post computes the impulse response and frequency response at the input sample rate of fs.

regards,

Neil

Hi Neil,

OK different perspectives. 

I am normally interested in output. In effect you are looking ahead expecting decimation further down.