Reply by navaneetha krishnan ●January 16, 20032003-01-16
Hi,
Setting format long does not affect the precision of
the coefficients. It affects only how it is displayed
on the screen. By default the coefficients will be in
double precision (64 bits). To do finite precision
analysis, you can use either the fdatool or qfilt.
Both are available in filterdesign toolbox.

Regards
Navan

--- Appalayagari Sreedhar <> wrote:
> Dear all,
>
> my problem straight away,
>
> i am doing recursive frequency sampling filter
> design,
>
> in matlab i design the filter coefficients, i gave
> the format long at the top and received the 15
> digits for representation after decimal point.
>
> does this give the precision value for how many bits
> is it 64 or 32 bits i want to do finite wordlength
> effects, in my project i hope u all might have
> worked on this i will be expecting your valuable
> suggestions,
>
> regds,sree.


__________________________________________________


Reply by Appalayagari Sreedhar ●January 12, 20032003-01-12
Dear all,

I am working on the recursive frequency sampling filter design.

My confusion is, I read the theory saying that as it contains the poles.and also
exactly pole zero cancellation occurs to give the stability.

in case of fir filter using remez or any,

we get only the b coefficients and a coefficients are assumed to be[1 rest all
zeros];

using tf2zp we can get zeros.

in the recursive frequency sampling filter case how can i get the zeros and
poles in the frequency sampling filter design .because i dont know how to get
the b and a coefficients.

i want to know how can i calculate the numerator and denominator
coefficients.using matlab.

and also to do the quantisation i needed to reduce the radius(r) to make exact
pole zero cancellation.

I will be waiting for all your responses.

regds,sree.