simple average filters...

Consider i have a simple average filter(FIR) of length N. Can I find its
cut-off frequency(rad/sample), if at all this much data is sufficient to
find it, without using any equations?
Regards,
Rakesh.

On Sep 28, 11:19 am, "Rajoshi" <bld....@gmail.com> wrote:
> Consider i have a simple average filter(FIR) of length N. Can I find its
> cut-off frequency(rad/sample), if at all this much data is sufficient to
> find it, without using any equations?
> Regards,
> Rakesh.

Does asking someone who does use equations count?

On Sep 28, 11:19 am, "Rajoshi" <bld....@gmail.com> wrote:
> Consider i have a simple average filter(FIR) of length N. Can I find its
> cut-off frequency(rad/sample), if at all this much data is sufficient to
> find it, without using any equations?
> Regards,
> Rakesh.

Yes, you can connect the FIR with an A/D and a D/A, and use a network
analyzer to measure the frequency response between A/D and D/A.

Rajoshi wrote:
> Consider i have a simple average filter(FIR) of length N. Can I find its
> cut-off frequency(rad/sample), if at all this much data is sufficient to
> find it, without using any equations?
> Regards,

You would have to define what you mean br "cutoff frequency". The 
response is a sinc which becomes narrower as N increases.

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

Rajoshi wrote:

> Consider i have a simple average filter(FIR) of length N. Can I find its
> cut-off frequency(rad/sample), if at all this much data is sufficient to
> find it, without using any equations?

Yes.

> Regards,
> Rakesh.

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

>
>
>Rajoshi wrote:
>
>> Consider i have a simple average filter(FIR) of length N. Can I find
its
>> cut-off frequency(rad/sample), if at all this much data is sufficient
to
>> find it, without using any equations?
>
>Yes.
>
>> Regards,
>> Rakesh.
>
>Vladimir Vassilevsky
>DSP and Mixed Signal Design Consultant
>http://www.abvolt.com
>
>
True, the response would be a sync and would get narrower as the filter
length starts growing. My query was can we intutively guess the cutoff(3db
low) freq of the filter? We can get hold of the frequencies which are not
passed thr the filter by merely comparing the window length to the period
of a sine wave.Does that help in finding the cut off freq?
Regards,
Rakesh

Rajoshi wrote:

> ... My query was can we intutively guess the cutoff(3db
> low) freq of the filter? ...

There is a simple relation between the 3-dBpointand N, normalized to the 
sample rate, but you need an equation to find it. Intuition won't do.

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

Or could it be a simple scaling factor?
If I change the length of the FIR, the curve just stretches on the sample
axis => the ratio between zero crossings and 3 dB cutoff frequency remains
the same.

I'd say it's around 0.4:
http://www.elisanet.fi/mnentwig/sinc.png

A related question: 
Is there an analytical solution to sin(x)=a x , or at least a built-in
function to Matlab etc like for example Bessel functions?

-mn

Rajoshi wrote:
> Consider i have a simple average filter(FIR) of length N. Can I find its
> cut-off frequency(rad/sample), if at all this much data is sufficient to
> find it, without using any equations?
> Regards,
> Rakesh.
> 
> 

An approximation for the -3dB point is ~ pi*0.88/N

On Sep 28, 8:19 am, "Rajoshi" <bld....@gmail.com> wrote:
> Consider i have a simple average filter(FIR) of length N. Can I find its
> cut-off frequency(rad/sample), if at all this much data is sufficient to
> find it, without using any equations?
> Regards,
> Rakesh.

If it's a software routine, you could run a large
number of test vectors of various frequencies through
the routine, plot the results, and find a spot on the
plot that looks like a reasonable cut-off frequency.

But even creating test vectors or a plot of the results
requires the use of some equations, unless you can find
turn-key packages which just happen to have the correct
default or auto-adaptive settings.


IMHO. YMMV.
--
rhn A.T nicholson d.0.t C-o-M