FIR filters

Hi all

Im currently trying to design a FIR filter in Matlab (fdatool) v7.04
and im experiencing a few problems.  A list of these have been included
below:

1) Im currently trying to design a filter with a flat passband and
therefore i have choosen to use a butterworth - filter slope 12dB/ Oct.
 However when trying to select this option, Matlab blocks you.  Is the
reason for this FIR filters cannot use butterworth filters?  If not
what do you recommend to provide a smooth passband.

2) I currently understand the sampling frequency is twice the highest
frequency.  At this point my missunderstand starts.  If you have 3
filter in a design (e.g. a graphic equaliser - fstop - 100, 1000, and
4000), should the sampling frequency be 8000 for all filters or should
the sampling frequency be relavent in all three instances ( 200, 2000
and 8000)?

look foward to your reply

Tuurbo46

tuurbo46@yahoo.co.uk wrote:
> Hi all
> 
> Im currently trying to design a FIR filter in Matlab (fdatool) v7.04
> and im experiencing a few problems.  A list of these have been included
> below:
> 
> 1) Im currently trying to design a filter with a flat passband and
> therefore i have choosen to use a butterworth - filter slope 12dB/ Oct.
>  However when trying to select this option, Matlab blocks you.  Is the
> reason for this FIR filters cannot use butterworth filters?  If not
> what do you recommend to provide a smooth passband.

Butterworth, Chebychev, and all the other filters you are familiar with 
from the analog world are digitally rendered as IIRs. With an FIR, you 
specify lower and upper passband edges, upper and lower stopband edges, 
the stopband attenuation, and the amount of passband ripple. (The lower 
passband and upper stopband edges are not needed for a low-pass filter.) 
The tighter the specs, the longer the filter. A stopband that begins 
close to the passband, lots of stopband attenuation, and very little 
passband ripple are all tight specs.

> 2) I currently understand the sampling frequency is twice the highest
> frequency.  At this point my missunderstand starts.  If you have 3
> filter in a design (e.g. a graphic equaliser - fstop - 100, 1000, and
> 4000), should the sampling frequency be 8000 for all filters or should
> the sampling frequency be relavent in all three instances ( 200, 2000
> and 8000)?

The sampling frequency must, in theory, be /more than/ twice the highest 
frequency in the analog signal being sampled if aliasing is to be 
avoided. In practice (without infinitely sharp filters), you need to 
sample faster. Sampling at 8000 Hz is normal for telephone quality, with 
an upper frequency limit of about 3500 Hz. I suspect that you want your 
equalizer for high-quality audio; CDs are sampled at 44,100 Hz.

It is simplest to run all the filters at the same rate. That way you can 
easily merge the filtered signals.

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