FFT SIZE DETECTION

hi jim,
        nice to see your reply..

somebody has taken fft (128,.256 or 512) of the complex signal Y(n).

assume that all these selection of fft size (128, 256 or 512) is
random.

i mean for pick 256 samples out of 10,000 and then do fft and store it

second pick 128 samples of remaining (10,000-256) and then do fft and
append to above result

and the thing goes on..

now i have to dynamically detect as what size of fft was taken (i mean
did the person took 256 fft or 128 fft or 512???? ) and then indicate
the FFT size taken..

this is all what the problem is ?. hope somebody hints or give
solutions

particlereddy

On Tue, 18 Jul 2006 20:45:13 +1000, Andrew Reilly
<andrew-newspost@areilly.bpc-users.org> wrote:

>Hi Par,
>
>On Mon, 17 Jul 2006 21:26:59 -0700, PARTICLEREDDY wrote:
>>                           seems everyone has trained guns against
>> me.doesnt matter. for first and last no apologies to anyone.
>
>Don't sweat it.  It's too hard to infer tone of voice or attitude on
>Usenet, so it's best not to infer either.  It didn't seem to me that
>anyone here had guns trained.  They just didn't understand your question. 
>Since most of the regulars here have been DSP professionals and/or
>professors for a very, very long time, that might indicate something to
>you about the way you phrased your question...
>
>> problem stated here;
>> 
>> X(n) ---- > complex random signal.
>> 
>> W(n) ------ > Noise vector.
>> 
>> Y(n)  = X(n) + W(n).  (obviously Y(n) is complex vector)
>> 
>> assume that Y(n) is of 10000 samples length.
>> 
>> i will be taking fft of Y(n) in chunks of (choice --- >128, 256 and
>> 512), now i have the FFT of the Y(n).
>
>How is this statement compatible with question one?  Doesn't the fact that
>you've taken the FFT indicate that you know how long it is?
>
>> 1. I Dont know what is the length of the FFT TAKEN.
>
>Maybe "taken" means something special when shouted?
>
>> 2. I DONT KNOW WHAT IS THE LENGTH OF THE SIGNAL TAKEN.
>
>Didn't you just say that it was 10000 samples?  Maybe this question has
>special meaning because it's been shouted, too.
>
>>                   i have to find out size of the fft taken on the
>> signal ???
>
>You seem to have done the taking.  How can you not know?
>
>> any more clarifications please do ask
>
>Just the questions above.
>
>> ANY FURTHER AGAIN NO APOLOGIES..i didnt do any mistake. if people
>> presume my behaviour as rude, its their mistake.
>
>Sure.  Just remember that replying to questions on comp.dsp is a
>recreation.  How participants react to you is naturally up to them.
>
>Cheers,

I'm with Andrew on the above clarifications.

Particle, the fact that several people found your posts rude should be
a hint to you.   I found your response to Andor quite rude, so
whatever your intent you may want to put some thought into how best to
respond to people.

It's generally best to not offend people from whom you're asking help.

And your subsequent clarifications have added to the ability to
understand your problem from what you first stated, but it's still not
very clear and I'd say is quite ill-posed.

The first step to getting a solution is understanding the problem, and
it appears that we're not yet to that step.

Eric Jacobsen
Minister of Algorithms, Intel Corp.
My opinions may not be Intel's opinions.
http://www.ericjacobsen.org

Hi Par,

On Tue, 18 Jul 2006 07:36:52 -0700, PARTICLEREDDY wrote:
> somebody has taken fft (128,.256 or 512) of the complex signal Y(n).

Aah.  Someone else.  And they haven't been decent enough to include the
size information.  That's kind of obscure.  Was there a good reason for
them to behave that way?

> assume that all these selection of fft size (128, 256 or 512) is random.
> 
> i mean for pick 256 samples out of 10,000 and then do fft and store it

Do you know for sure that one of the supposed frames of transformed data
are aligned with that end of the signal?  If not, you might have to do
some tests to get an initial algnment.

> second pick 128 samples of remaining (10,000-256) and then do fft and
> append to above result

Or 256 or 512, one presumes...

> and the thing goes on..
> 
> now i have to dynamically detect as what size of fft was taken (i mean
> did the person took 256 fft or 128 fft or 512???? ) and then indicate
> the FFT size taken..
> 
> this is all what the problem is ?. hope somebody hints or give solutions

In general, if you don't know anything about the input signal, or if what
you know isn't helpful (say it's known to be pure (iid) complex random
values, or say it's all zero-valued), then you're just stuck: make up some
sequence of supposed frame sizes and claim that you're done: no-one will
be able to say differently.

If you do know something  about the original signal, like band-limitedness
(leading to smoothness tests, perhaps), then you have a chance of finding
the boundaries by computing *all* of the possible inverse FFT frames
(i.e., all sizes at all plausible offsets), and use the known constraints
on the original signal to stitch together frames that fit those
constraints.  If you have ambiguity, you'll need to either maintain
multiple alternative answers or discover a stronger constraint about the
original signal.

Once you've got a plausible-looking reconstruction of the original, you
also have the frame boundaries and sizes.

Seems like a fairly unlikely scenario, though.  I suspect that there are
still aspects of the problem that you aren't telling us, and they may
contain the seeds of a more efficient answer.

Cheers,

-- 
Andrew

Hi Particle,

On Tue, 18 Jul 2006 07:36:52 -0700, PARTICLEREDDY wrote:
> somebody has taken fft (128,.256 or 512) of the complex signal Y(n).
> 
> assume that all these selection of fft size (128, 256 or 512) is
> random.

Since you know so little about the data stream that results, are you sure
that you can be sure that the supposed FFTs are only computed at those
three lenghts?  Not longer or shorter?  Not non-power-of-two?  That
they're engineering or mathematical FFTs?  That the data has been streamed
in "natural" order or in bit-reversed order, or some other permutation
(djbfft produces an interesting output order, for example).  If you don't
know the answer to *any* of these issues, then any search-based approach
is doomed to failure.

Just looking for more possible clarification about the nature of the real
problem...

Cheers,

-- 
Andrew

On Tue, 18 Jul 2006 09:13:57 -0700, Eric Jacobsen
<eric.jacobsen@ieee.org> wrote:

  (snipped)
>
>Particle, the fact that several people found your posts rude should be
>a hint to you.   I found your response to Andor quite rude, so
>whatever your intent you may want to put some thought into how best to
>respond to people.
>
>It's generally best to not offend people from whom you're asking help.
>
>And your subsequent clarifications have added to the ability to
>understand your problem from what you first stated, but it's still not
>very clear and I'd say is quite ill-posed.
>
>The first step to getting a solution is understanding the problem, and
>it appears that we're not yet to that step.
>
>Eric Jacobsen

Hi Eric,
  remember my recent post about "manners" here on comp.dsp?
Ha ha ha.

I'll bet Particle doesn't talk to his professor, or 
his colleagues the way he talked to Andor. 
Particle deserves no help.

See ya',
[-Rick-]

On 17 Jul 2006 05:24:51 -0700, "Andor" <andor.bariska@gmail.com>
wrote:

>
>PARTICLEREDDY wrote:
>> hi jerry,
>>            thanks for your interest. the question is like that.
>> again reiterating i need to dynamically detect size of fft taken on the
>> signal..
>>
>> i dont have any knowledge of signal nor length of fft taken..
>>
>> this is my question. no who's and no what's
>
>That's a funny question.
>
>The output of an FFT is usually stored in an array. If you don't know
>the length of the array (and with it the length of the FFT), how do you
>know when the array stops and other data begins?

Hi Andor,

   I was shocked to read Particle's reply to your 
post.  What incredibly bad manners by Particle.
Astounding!   Truly astounding.

Anyone who does not listen to you, and carefully consider 
your words, is making a very BIG mistake.
In the U.S., people that make very big mistakes 
are called "stupid". 

Particle deserves no help.
What he does deserve I will not say.

Take care Andor,
[-Rick-]

On 17 Jul 2006 21:26:59 -0700, "PARTICLEREDDY"
<particlereddy@gmail.com> wrote:


>
>ANY FURTHER AGAIN NO APOLOGIES..i didnt do any mistake. if people
>presume my behaviour as rude, its their mistake.
>
>particlereddy

Hello Particlereddy, (it's too bad you chose not to 
give your real name.  However, I don't blame you---with 
behavior like yours, I would also be too embarressed to 
give my real name.)

What you must realize is that the fantastically 
talented guys here on comp.dsp are not your Mother, 
who thought anything you did was simply wonderful.

When you crapped in your diaper and smeared it on the 
walls your Mother thought that was really great.  
In that way, she gave you the false impression that 
anything you do is beautiful.  She deluded you into 
thinking that you arwe always correct.  
Well Particle, you are in the real world now.

If you want help from experts, show a little respect.
And if an expert says your question (which was truly weird, 
profoundly weird) was hard to understand then, instead of 
insulting the expert, you should re-phrase your question.

If you came into my office and talked to me the way you 
talked to the experts here, I would have grabbed you by 
the throat, squeezed until your eyeballs popped out, and 
thrown you out of my office.

Please notice how I did not insult you or curse you in 
any way.  I merely stated facts that I believe are true.
You have an opportunity now to better 
yourself and perhaps learn something from the 
gentle-mannered experts here about DSP.  
Don't screw it up!

[-Rick-]

Andor wrote:
> [...]
> It does seem odd that you have data, but don't know how much of it you
> have.

Yes, quite 2*n + 1.

--Randy

Jerry Avins wrote:
> PARTICLEREDDY wrote:

>> ok andor,
>>              before saying it as a funny question, try to put your
>> brain into new technologies then you will know, where exactly this
>> question has arised. i am sorry, but your idiotic answer is provoking
>> me to say this..have brain and think and then only answer..

>> and further behave has an engineering professional before giving
>> USELESS and WORTHLESS ANSWERS.

> Your frustration is evident, but it doesn't justify rudeness.

I was about to say the same thing before reading your answer.
(I usually read newsgroups in reverse order, which complicates
replies sometimes.)

> Your question as stated is, if not funny, decidedly peculiar. Two 
> possibilities come to mind: either you just don't know what you're 
> asking about, or you're expressing yourself very poorly. I've been 
> trying to elicit the meat of the matter from you, but your rudeness put 
> me off. I'll await your apology before communicating with you again.

With additional explanation it makes slightly more sense, but still
not much.  If the goal is an FFT of 10000 points, partial FFTs,
other than those done by the FFT algorithm itself, are not really
a partial solution.

With a lot more detail of the source of the data it might be that
someone could come up with a reasonable answer.  Otherwise, I
think the answer is no, you can't do that.

-- glen

hi rick
         really pity that i didnt even bothered about the answer or any
questions i asked. seems you are hovering in a fantastic world that i
will come to your office and that you do all the so said stuff. To tell
frankly, you are no engineer by profession. you did talked all the dirt
and saying that you didnt insult anyone. i think you are trying to
impress upon the experts in this group. first and last thing, i dont
involve in tussle with brutes (YOU) (SORRY, I DIDNT INSULT YOU). I

points i would like to expand

1. when somebody called my question funny, i replied that useless
fellow aptly in a befiting manner, he could have asked "particle,
please do provide more details instead of calling it as funny question
and andor i answered well with full respect, i know him very well."

2. second i didnt insult any one..

3. i wont get provoked for unecessary and unwanted matters from others
(as a matter of fact from you). (The say when elephants are marching
royally, darks will be barking, so you are barking when i am not
involved in tussle at all ...)

4. i have respect for each and everybody in this comp.dsp.

Also you post my reply in this group also.

5. lastly you ask your mom and dad as why didnt they teach YOU
manners,respect and common sense to you to not interefere in other's
business.

6. TO all dear respected members, i didnt involve in insult to anyone.

7. Lastly this will be my last post in this topic, i solved my problem.


---------------------BYE-----------------------------------