FIR filter frequency response

The frequency response of the FIR filter repeats at the sampling frequency
Fs and its multiples. 
Im confused about how this affects the filter output, if the input signal
has frequencies at Fs, 2 Fs, 3Fs?This doesn't seem right to me, that a low
pass filter will pass higher frequenies too, since the frequency respone
repeats at multiples of Fs.What am I not understanding correctly?

In frequency domain, all information is contained within 0 to Fs range.

Now when we say that "somethings repeat at multiple of Fs" it essentially
means that you are traversing the unit circle again and again and again.

So whether you traverse once or you traverse N number of time's it is
same.

Hence all information is contained within 0 to Fs range. where Fs/2 is
called mirror image symmetry point.

Even fin = 0 hz and Fs point are same. For example if you were to sample
8000 hz signal at Fs = 8000 hz, then you would get only 1 point per cycle
and if you position it to be peak or any other point other then 0, then
you would get a DC output.

Hope this clarifies some clutter.

Rgds
Bharat Pathak

Arithos Designs
www.Arithos.com

DSP Design Consultancy and Training Company.





>The frequency response of the FIR filter repeats at the sampling
frequency
>Fs and its multiples. 
>Im confused about how this affects the filter output, if the input
signal
>has frequencies at Fs, 2 Fs, 3Fs?This doesn't seem right to me, that a
low
>pass filter will pass higher frequenies too, since the frequency respone
>repeats at multiples of Fs.What am I not understanding correctly?
>

On Wed, 22 Oct 2008 22:54:53 -0500, Zeph80 wrote:

> The frequency response of the FIR filter repeats at the sampling
> frequency Fs and its multiples.
> Im confused about how this affects the filter output, if the input
> signal has frequencies at Fs, 2 Fs, 3Fs?This doesn't seem right to me,
> that a low pass filter will pass higher frequenies too, since the
> frequency respone repeats at multiples of Fs.What am I not understanding
> correctly?

This may shed some light:  http://www.wescottdesign.com/articles/Sampling/
sampling.html.

As Bharat said, in the sampled-time domain Fs "means" the same thing as 
0Hz; this doesn't make a whole lot of difference in the sampled-time 
domain, but it makes a huge difference in going from continuous-time to 
sampled-time, or back.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Do you need to implement control loops in software?
"Applied Control Theory for Embedded Systems" gives you just what it says.
See details at http://www.wescottdesign.com/actfes/actfes.html

>The frequency response of the FIR filter repeats at the sampling
frequency
>Fs and its multiples. 
>Im confused about how this affects the filter output, if the input
signal
>has frequencies at Fs, 2 Fs, 3Fs?This doesn't seem right to me, that a
low
>pass filter will pass higher frequenies too, since the frequency respone
>repeats at multiples of Fs.What am I not understanding correctly?
>


Im not sure Im following your replies. I guess what Im asking is, when you
sample a signal for instance , the frequency response repeats at multiples
of Fs, and those need to be filtered, or if you uponvert you see all those
frequencies 2Fs, 3Fs etc. Those frequecies actually exist.
If that actually exists, then doesnt this mean even the frequency response
of a FIR filter, allows frequecies at 2Fs, 3 Fs, etc.since its response is
repeating at those frequencies?

>In frequency domain, all information is contained within 0 to Fs range.
>
>Now when we say that "somethings repeat at multiple of Fs" it
essentially
>means that you are traversing the unit circle again and again and again.
>
>So whether you traverse once or you traverse N number of time's it is
>same.
>
>Hence all information is contained within 0 to Fs range. where Fs/2 is
>called mirror image symmetry point.
>
>Even fin = 0 hz and Fs point are same. For example if you were to sample
>8000 hz signal at Fs = 8000 hz, then you would get only 1 point per
cycle
>and if you position it to be peak or any other point other then 0, then
>you would get a DC output.
>
>Hope this clarifies some clutter.
>
>Rgds
>Bharat Pathak
>
>Arithos Designs
>www.Arithos.com
>
>DSP Design Consultancy and Training Company.
>

Im not sure Im following your replies. I guess what Im asking is, when
you
sample a signal for instance , the frequency response repeats at
multiples
of Fs, and those need to be filtered, or if you uponvert you see all
those
frequencies 2Fs, 3Fs etc. Those frequecies actually exist.
If that actually exists, then doesnt this mean even the frequency
response
of a FIR filter, allows frequecies at 2Fs, 3 Fs, etc.since its response
is
repeating at those frequencies?

>
>
>
>
>>The frequency response of the FIR filter repeats at the sampling
>frequency
>>Fs and its multiples. 
>>Im confused about how this affects the filter output, if the input
>signal
>>has frequencies at Fs, 2 Fs, 3Fs?This doesn't seem right to me, that a
>low
>>pass filter will pass higher frequenies too, since the frequency
respone
>>repeats at multiples of Fs.What am I not understanding correctly?
>>
>

>On Wed, 22 Oct 2008 22:54:53 -0500, Zeph80 wrote:
>
>> The frequency response of the FIR filter repeats at the sampling
>> frequency Fs and its multiples.
>> Im confused about how this affects the filter output, if the input
>> signal has frequencies at Fs, 2 Fs, 3Fs?This doesn't seem right to me,
>> that a low pass filter will pass higher frequenies too, since the
>> frequency respone repeats at multiples of Fs.What am I not
understanding
>> correctly?
>
>This may shed some light: 
http://www.wescottdesign.com/articles/Sampling/
>sampling.html.
>
>As Bharat said, in the sampled-time domain Fs "means" the same thing as 
>0Hz; this doesn't make a whole lot of difference in the sampled-time 
>domain, but it makes a huge difference in going from continuous-time to 
>sampled-time, or back.
>
>-- 
>
>Tim Wescott
>Wescott Design Services
>http://www.wescottdesign.com
>
>Do you need to implement control loops in software?
>"Applied Control Theory for Embedded Systems" gives you just what it
says.
>See details at http://www.wescottdesign.com/actfes/actfes.html
>

Im not sure Im following your replies. I guess what Im asking is, when
you
sample a signal for instance , the frequency response repeats at
multiples
of Fs, and those need to be filtered, or if you uponvert you see all
those
frequencies 2Fs, 3Fs etc. Those frequecies actually exist.
If that actually exists, then doesnt this mean even the frequency
response
of a FIR filter, allows frequecies at 2Fs, 3 Fs, etc.since its response
is
repeating at those frequencies?

On 23 Okt, 21:49, "Zeph80" <surabhi_tal...@hotmail.com> wrote:
> >The frequency response of the FIR filter repeats at the sampling
> frequency
> >Fs and its multiples.
> >Im confused about how this affects the filter output, if the input
> signal
> >has frequencies at Fs, 2 Fs, 3Fs?This doesn't seem right to me, that a
> low
> >pass filter will pass higher frequenies too, since the frequency respone
> >repeats at multiples of Fs.What am I not understanding correctly?
>
> Im not sure Im following your replies. I guess what Im asking is, when you
> sample a signal for instance , the frequency response repeats at multiples
> of Fs, and those need to be filtered, or if you uponvert you see all those
> frequencies 2Fs, 3Fs etc. Those frequecies actually exist.
> If that actually exists, then doesnt this mean even the frequency response
> of a FIR filter, allows frequecies at 2Fs, 3 Fs, etc.since its response is
> repeating at those frequencies?

You are correct, both in your observations and in the
concern about the ambiguity. There is nothing you can
do about such ambiguities in discrete-time domain.

That's why one always should use an analog anti-alias
filter prior to the ADC in sampled systems. The anti-
alias filter should only allow one of the 'mirrors'
through (it needs not be the baseband 'mirror') so that
one always knows the 'true' bandwidth of the signal.

Rune

>The frequency response of the FIR filter repeats at the sampling
frequency
>Fs and its multiples. 
>Im confused about how this affects the filter output, if the input
signal
>has frequencies at Fs, 2 Fs, 3Fs?This doesn't seem right to me, that a
low
>pass filter will pass higher frequenies too, since the frequency respone
>repeats at multiples of Fs.What am I not understanding correctly?
>


I finaly found something which answers my question. In this link
http://www.hunteng.co.uk/pdfs/tech/ddctheory.pdf on page 3 he clearly
shows the frequency response of the LPF filter repeating at Fs,and he shows
that if the input to the filter has frequencies at Fs and 0, both will be
passed through.

Zeph80 wrote:

   ...

> Im not sure Im following your replies. I guess what Im asking is, when
> you
> sample a signal for instance , the frequency response repeats at
> multiples
> of Fs, and those need to be filtered, or if you uponvert you see all
> those
> frequencies 2Fs, 3Fs etc. Those frequecies actually exist.
> If that actually exists, then doesnt this mean even the frequency
> response
> of a FIR filter, allows frequecies at 2Fs, 3 Fs, etc.since its response
> is
> repeating at those frequencies?

You do understand that before sampling an analog signal at a sample rate 
fs, you must low-pass filter it to remove components at and above fs/2? 
(Bandpass sampling is an extension of this, not an exception to it.) 
Since the input to the filter can be no higher than fs/2, what is there 
to repeat?

A more subtle way to say the same thing is that every component you feed 
the system that lies above fs/2 is translated by the physics of the 
situation to one below according to some simple rules. Keyword: alias.

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

surabhi_talwar@hotmail.com wrote:
 >  visited DSPRelated.com and clicked on your name from this page:
 > 		http://www.dsprelated.com/showmessage/104558.php
 > 		to contact you.  His message follows:
 > 		
 > 				I finaly found something which answers
 > my question. In this link
 > http://www.hunteng.co.uk/pdfs/tech/ddctheory.pdf on page 3 he clearly
 > shows the frequency response of the LPF filter repeating at Fs,and he
 > shows
 > that if the input to the filter has frequencies at Fs and 0, both will
 > be passed through.
 >
 > I guess I didnt frame my question well, but this is what I was asking.


Chopping a signal by sampling it creates images. When those images 
overlap, we call them aliases. The important point to keep in mind is 
that for discrete-time signals, 0 and Fs are one and the same. The 
continuous-time frequency line becomes a frequency circle with 
discrete-time signals. On that circle,  0 and Fs overlap.

Please don't bring newsgroup issues to me at home. Aside from any 
possible intrusion, you get the benefit of the group's collective wisdom 
by posting publicly there.

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