 | | From: | cbull | | Subject: | Frequency domain filters? | | Date: | Tue, 18 Jan 2005 13:53:37 +0100 |
|
|
 | Hello,
I'm a newbie to dsp and would like to ask you some questions.
Does any filter in frequency domain has its equivalence in time-domain
filters (IIR or FIR)?
Is there something that can be done in frequency domain (after FFT) that
cannot be done in time-domain or vice-versa?
Or maybe doing it in frequency domain (although more expensive) is more
accurate?
Please forgive me any stupidity I have made wirtting these questions and
answer in simple words :)
Best wishes,
Tomasz
|
|
| | From: | Richard Owlett | | Subject: | Re: Frequency domain filters? | | Date: | Tue, 18 Jan 2005 13:07:01 -0600 |
|
|
| cbull wrote:
> Hello,
>
> I'm a newbie to dsp and would like to ask you some questions.
>
> [ snip specific question ]
>
> Please forgive me any stupidity I have made wirtting these questions and
> answer in simple words :)
>
Don't apologize.
The only "dumb" question is one you don't bother asking.
I've asked many naive and confused questions on this forum and have
always received patient educational replies.
|
|
| | From: | Fred Marshall | | Subject: | Re: Frequency domain filters? | | Date: | Tue, 18 Jan 2005 08:08:14 -0800 |
|
|
|
"cbull" wrote in message
news:csj19t$8ol$1@nemesis.news.tpi.pl...
> Hello,
>
> I'm a newbie to dsp and would like to ask you some questions.
>
> Does any filter in frequency domain has its equivalence in time-domain
> filters (IIR or FIR)?
All linear, time-invariant filters have a Fourier Transform and. It is a
1:1 transform so the transforms are unique. So, the answer is YES.
Now, if you mean a finite frequency domain as with an FFT, then any filter
defined in frequency is defined in time. But, any filter defined in time is
not defined in finite frequency. For example, a continuous filter or an
infinite duration filter.
> Is there something that can be done in frequency domain (after FFT) that
> cannot be done in time-domain or vice-versa?
> Or maybe doing it in frequency domain (although more expensive) is more
> accurate?
Well, yes in a way: you can multiply in frequency instead of convolving in
time.
Then there are a bunch of nonlinear things you can do. See homomorphic
filtering and cepstrum.
Fred
|
|
| | From: | cbull | | Subject: | Re: Frequency domain filters? | | Date: | Tue, 18 Jan 2005 17:12:12 +0100 |
|
|
|
>
> > Is there something that can be done in frequency domain (after FFT) that
> > cannot be done in time-domain or vice-versa?
> > Or maybe doing it in frequency domain (although more expensive) is more
> > accurate?
>
> Well, yes in a way: you can multiply in frequency instead of convolving in
> time.
Well that is what I was thinking of when asking this question. What are the
interesting effects I could achieve? I'm mostly interested in sound
processing.
Tomasz
|
|
| | From: | Jerry Avins | | Subject: | Re: Frequency domain filters? | | Date: | Tue, 18 Jan 2005 11:59:20 -0500 |
|
|
| cbull wrote:
>>>Is there something that can be done in frequency domain (after FFT) that
>>>cannot be done in time-domain or vice-versa?
>>>Or maybe doing it in frequency domain (although more expensive) is more
>>>accurate?
>>
>>Well, yes in a way: you can multiply in frequency instead of convolving in
>>time.
>
>
> Well that is what I was thinking of when asking this question. What are the
> interesting effects I could achieve? I'm mostly interested in sound
> processing.
>
> Tomasz
Time domain and frequency domain views represent the same signal and
operations. Although some operations are faster or simpler in one domain
or the other, whatever can be accomplished in either can be done by
both. They are entirely equivalent.
Jerry
--
Engineering is the art of making what you want from things you can get.
ŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻŻ
|
|
| | From: | Fred Marshall | | Subject: | Re: Frequency domain filters? | | Date: | Wed, 19 Jan 2005 09:17:01 -0800 |
|
|
|
"Jerry Avins" wrote in message
news:354tjaF4ik5ajU1@individual.net...
> cbull wrote:
>> Well that is what I was thinking of when asking this question. What are
>> the
>> interesting effects I could achieve? I'm mostly interested in sound
>> processing.
>>
>> Tomasz
>
> Time domain and frequency domain views represent the same signal and
> operations. Although some operations are faster or simpler in one domain
> or the other, whatever can be accomplished in either can be done by
> both. They are entirely equivalent.
>
> Jerry
> --
>
This is an essential point for linear filtering and Jerry has said it more
directly than I did.
Now, if you want to consider nonlinear filtering then that's a different
matter:
e.g. if you hard limit in the time domain then you will generate a lot of
distortion that may or may not be associated with what's called "fuzz". I
think a better forum might be rec.music.makers or rec.audio.misc or
alt.sci.physics.acoustics. The latter has some overlap with comp.dsp and
any number of experts. The other two I've not read but they pop up out of a
Google search on "audio fuzz".
Fred
|
|
