Friends,

I read the application note about The MSP430 Hardware Multiplier:

http://www.microcontrol.cn/datasheet/MSP430/MSP430AN/slaa042.pdf

I understand many things but I have a question: what is the FFT
results format? The data results are in the format Tension x Freq?

I need to know what are the predominant frequencies of the signal.

Please help a desperate student,

Thanks,
------------------------------------



(You need to be a member of msp430 -- send a blank email to msp430-subscribe@yahoogroups.com )

The input to FFT is a discrete quantized waveform in the time domain.
The output from FFT is a discrete quantized frequency spectrum in the
frequency domain. The predominant frequencies are the frequencies of
the predominant peaks in the frequency spectrum.

--- In m...@yahoogroups.com, "keller.lima" wrote:
>
> Friends,
>
> I read the application note about The MSP430 Hardware Multiplier:
>
> http://www.microcontrol.cn/datasheet/MSP430/MSP430AN/slaa042.pdf
>
> I understand many things but I have a question: what is the FFT
> results format? The data results are in the format Tension x Freq?
>
> I need to know what are the predominant frequencies of the signal.
>
> Please help a desperate student,
>
> Thanks,
>

------------------------------------



(You need to be a member of msp430 -- send a blank email to msp430-subscribe@yahoogroups.com )

Hello.

As a student, you might be interested in reading a little bit about what
is an FFT. Google keys: Cooley, Sand, Tuckey (I am not sure of the spelling
of this one).

That said, the FFT transform a set of time samples (power of 2), into
another set of frequency samples (same power of 2). So if you FFT your
input signal, you will get exactly what you want: the amplitude of each
of the frequency components. Then you look for the strongest sample to
have the predominant frequency.

The result? If you are sampling a voltage (V(t)) you will get the same
voltage as a function of the frequency (V(f)). Basically, you get what
you input, under another representation.

Suppose that your sampling frequency is 1000 Hz and that you are
sampling a sinewave of amplitude 1 at exactly 250 Hz, you will get
2 spikes only in the output signal, ona at 250 Hz, and one at 750 Hz.
Just ignore the second (1) one which is an alias of the first one, and that's
about it, the result tells you that you have only 250 Hz, what you already
knew in this case.

(1) as I assume your signal is purely real (output of an ADC),
you can just ignore the upper half of the output set (samples from N/2
to N-1). Hint: you don't have to calculate what you will finally ignore.

Last thing (that should come first): Before sampling, make sure that the
max frequency of what you want to measure is less than half of the
sampling frequency, otherwise you will get unwanted results.

Pascal

--- In m...@yahoogroups.com, "keller.lima" wrote:
>
> Friends,
>
> I read the application note about The MSP430 Hardware Multiplier:
>
> http://www.microcontrol.cn/datasheet/MSP430/MSP430AN/slaa042.pdf
>
> I understand many things but I have a question: what is the FFT
> results format? The data results are in the format Tension x Freq?
>
> I need to know what are the predominant frequencies of the signal.
>
> Please help a desperate student,
>
> Thanks,
>

------------------------------------



(You need to be a member of msp430 -- send a blank email to msp430-subscribe@yahoogroups.com )

Sorry friend,
I think that my question was wrong. I know a little about FFT.
Actually I want to know how I can read the results of the FFT
according to the frequency. The results is a sequence of bits,
right? So, how I show the results voltage x frequency? For example,
if I could measure the output with an oscilloscope, as I see this
sign?
In my project I need to discovery two predominat frequency in one
signal, but I don't know what I do after FFT.

Did you understand my questions? Sorry, my english is poor.
Thanks for all,
--- In m...@yahoogroups.com, "p_murayama" wrote:
>
> Hello.
>
> As a student, you might be interested in reading a little bit
about what
> is an FFT. Google keys: Cooley, Sand, Tuckey (I am not sure of the
spelling
> of this one).
>
> That said, the FFT transform a set of time samples (power of 2),
into
> another set of frequency samples (same power of 2). So if you FFT
your
> input signal, you will get exactly what you want: the amplitude of
each
> of the frequency components. Then you look for the strongest
sample to
> have the predominant frequency.
>
> The result? If you are sampling a voltage (V(t)) you will get the
same
> voltage as a function of the frequency (V(f)). Basically, you get
> you input, under another representation.
>
> Suppose that your sampling frequency is 1000 Hz and that you are
> sampling a sinewave of amplitude 1 at exactly 250 Hz, you will get
> 2 spikes only in the output signal, ona at 250 Hz, and one at 750
Hz.
> Just ignore the second (1) one which is an alias of the first one,
and that's
> about it, the result tells you that you have only 250 Hz, what you
already
> knew in this case.
>
> (1) as I assume your signal is purely real (output of an ADC),
> you can just ignore the upper half of the output set (samples from
N/2
> to N-1). Hint: you don't have to calculate what you will finally
ignore.
>
> Last thing (that should come first): Before sampling, make sure
that the
> max frequency of what you want to measure is less than half of the
> sampling frequency, otherwise you will get unwanted results.
>
> Pascal
>
> --- In m...@yahoogroups.com, "keller.lima" wrote:
> >
> > Friends,
> >
> > I read the application note about The MSP430 Hardware
Multiplier:
> >
> > http://www.microcontrol.cn/datasheet/MSP430/MSP430AN/slaa042.pdf
> >
> > I understand many things but I have a question: what is the
FFT
> > results format? The data results are in the format Tension x
Freq?
> >
> > I need to know what are the predominant frequencies of the
signal.
> >
> > Please help a desperate student,
> >
> > Thanks,
>
------------------------------------



(You need to be a member of msp430 -- send a blank email to msp430-subscribe@yahoogroups.com )

OK, this is an EXTREMELY simplified explanation, but probably what you
are looking for. When you perform an FFT you are basically calculating
the energy present for a specfic frequency within the composite signal
over a fixed time period. FFts are said to have a certain number of
'taps', which are sometimes called 'buckets'. a signal sampled at 8000
samples a second will have a useful rage of 0-4000hz,. analysign this
signal with a 100 tap signal will split the original signal into 40 hz
intervals. The results you get back are typically amplitude and phase.
each pair will allow you to create a sinusoial signal at the centre
frequency of that 'bucket'. Summing these signals will result in the
reconstruction of the original signal. So, very simply looking for the
two highest amplitudes will give you the two dominant signals.

remember this is a crude explanation of what is happening. there is much
more to it, and I suggest that you study this practically, as well as
from reference books. for example:-

1. create a few waveforms from known sinusoidal signals then run an FFt
on them with 'buckets' that match the original freqiencies.
2. Now repeat this with a differnet set of 'buckets'
3. analyse a complex waveform and then reconstruct it. see how accurate
it is. (HINT: beware of the DC component)
4. analyse a complex waveform such as an audio signal, then reconstruct
it using only every 2nd, 3rd or 4th 'bucket', keep getting 'thinner' wit
your data and see what happens.
5. recostruct using a fixed value for phase, and see what happens to
your audio signal.

These are justa few fun things to do with FFT's.

By the way EXcel also can do FFT's. it is interesting to load a table
with a set of data froma composite waveform and then play aroun din Excel.

Al

keller.lima wrote:

>Sorry friend,
>I think that my question was wrong. I know a little about FFT.
>Actually I want to know how I can read the results of the FFT
>according to the frequency. The results is a sequence of bits,
>right? So, how I show the results voltage x frequency? For example,
>if I could measure the output with an oscilloscope, as I see this
>sign?
>In my project I need to discovery two predominat frequency in one
>signal, but I don't know what I do after FFT.
>
>Did you understand my questions? Sorry, my english is poor.
>Thanks for all,
>--- In m...@yahoogroups.com, "p_murayama" wrote:
>
>
>>Hello.
>>
>>As a student, you might be interested in reading a little bit
>>
>>
>about what
>
>
>>is an FFT. Google keys: Cooley, Sand, Tuckey (I am not sure of the
>>
>>
>spelling
>
>
>>of this one).
>>
>>That said, the FFT transform a set of time samples (power of 2),
>>
>>
>into
>
>
>>another set of frequency samples (same power of 2). So if you FFT
>>
>>
>your
>
>
>>input signal, you will get exactly what you want: the amplitude of
>>
>>
>each
>
>
>>of the frequency components. Then you look for the strongest
>>
>>
>sample to
>
>
>>have the predominant frequency.
>>
>>The result? If you are sampling a voltage (V(t)) you will get the
>>
>>
>same
>
>
>>voltage as a function of the frequency (V(f)). Basically, you get
>>you input, under another representation.
>>
>>Suppose that your sampling frequency is 1000 Hz and that you are
>>sampling a sinewave of amplitude 1 at exactly 250 Hz, you will get
>>2 spikes only in the output signal, ona at 250 Hz, and one at 750
>>
>>
>Hz.
>
>
>>Just ignore the second (1) one which is an alias of the first one,
>>
>>
>and that's
>
>
>>about it, the result tells you that you have only 250 Hz, what you
>>
>>
>already
>
>
>>knew in this case.
>>
>>(1) as I assume your signal is purely real (output of an ADC),
>>you can just ignore the upper half of the output set (samples from
>>
>>
>N/2
>
>
>>to N-1). Hint: you don't have to calculate what you will finally
>>
>>
>ignore.
>
>
>>Last thing (that should come first): Before sampling, make sure
>>
>>
>that the
>
>
>>max frequency of what you want to measure is less than half of the
>>sampling frequency, otherwise you will get unwanted results.
>>
>>Pascal
>>
>>--- In m...@yahoogroups.com, "keller.lima" wrote:
>>
>>
>>> Friends,
>>>
>>> I read the application note about The MSP430 Hardware
>>>
>>>
>Multiplier:
>
>
>>>http://www.microcontrol.cn/datasheet/MSP430/MSP430AN/slaa042.pdf
>>>
>>> I understand many things but I have a question: what is the
>>>
>>>
>FFT
>
>
>>>results format? The data results are in the format Tension x
>>>
>>>
>Freq?
>
>
>>> I need to know what are the predominant frequencies of the
>>>
>>>
>signal.
>
>
>>>Please help a desperate student,
>>>
>>>Thanks,
>>>
>>>
>>>------------------------------------



(You need to be a member of msp430 -- send a blank email to msp430-subscribe@yahoogroups.com )

Try this document:
http://focus.tij.co.jp/jp/lit/an/spra417a/spra417a.pdf
It is an implementation example of a spectrum analyser using 128pt FFT in a=
TI's DSP. It is an assembly program (:P) but can easily be converted to MS=
P's code (assembly, of course... You may need to make some routines to perf=
orm functions that are natural to DSP's architecture).=20
Then just summ this paper with all others you have, along with explanations=
you've got here, and you will be able to implement your project.
In a matter of fact this paper does exactly what you need. Just forget the =
part where it puts the results in a DAC and stays where it scans the result=
s to find the biggest value...
-Augusto
De:m...@yahoogroups.com

Para:m...@yahoogroups.com

C=F3pia:

Data:Thu, 04 Sep 2008 14:01:26 -0000

Assunto:[msp430] Re: Doubts about FFT

Sorry friend,=20
I think that my question was wrong. I know a little about FFT.=20
Actually I want to know how I can read the results of the FFT=20
according to the frequency. The results is a sequence of bits,=20
right? So, how I show the results voltage x frequency? For example,=20
if I could measure the output with an oscilloscope, as I see this=20
sign?
In my project I need to discovery two predominat frequency in one=20
signal, but I don't know what I do after FFT.

Did you understand my questions? Sorry, my english is poor.

Thanks for all,=20

--- In m...@yahoogroups.com, "p_murayama" wrote:
>
> Hello.
>=20
> As a student, you might be interested in reading a little bit=20
about what
> is an FFT. Google keys: Cooley, Sand, Tuckey (I am not sure of the=20
spelling
> of this one).
>=20
> That said, the FFT transform a set of time samples (power of 2),=20
into
> another set of frequency samples (same power of 2). So if you FFT=20
your
> input signal, you will get exactly what you want: the amplitude of=20
each
> of the frequency components. Then you look for the strongest=20
sample to
> have the predominant frequency.
>=20
> The result? If you are sampling a voltage (V(t)) you will get the=20
same
> voltage as a function of the frequency (V(f)). Basically, you get=20
> you input, under another representation.
>=20
> Suppose that your sampling frequency is 1000 Hz and that you are
> sampling a sinewave of amplitude 1 at exactly 250 Hz, you will get
> 2 spikes only in the output signal, ona at 250 Hz, and one at 750=20
Hz.
> Just ignore the second (1) one which is an alias of the first one,=20
and that's
> about it, the result tells you that you have only 250 Hz, what you=20
already
> knew in this case.
>=20
> (1) as I assume your signal is purely real (output of an ADC),
> you can just ignore the upper half of the output set (samples from=20
N/2
> to N-1). Hint: you don't have to calculate what you will finally=20
ignore.
>=20
> Last thing (that should come first): Before sampling, make sure=20
that the
> max frequency of what you want to measure is less than half of the
> sampling frequency, otherwise you will get unwanted results.
>=20
> Pascal
>=20
> --- In m...@yahoogroups.com, "keller.lima" wrote:
> >
> > Friends,
> >=20
> > I read the application note about The MSP430 Hardware=20
Multiplier:=20
> >=20
> > http://www.microcontrol.cn/datasheet/MSP430/MSP430AN/slaa042.pdf
> >=20
> > I understand many things but I have a question: what is the=20
FFT=20
> > results format? The data results are in the format Tension x=20
Freq?
> >=20
> > I need to know what are the predominant frequencies of the=20
signal.
> >=20
> > Please help a desperate student,=20
> >=20
> > Thanks,
>

=20
[Non-text portions of this message have been removed]
------------------------------------



(You need to be a member of msp430 -- send a blank email to msp430-subscribe@yahoogroups.com )

Great!

It is exactly what I wanted! Thank you very much. I think that I am
ready to start my project. You changed my project and for the better.

Thank you.

Kel

--- In m...@yahoogroups.com, Onestone wrote:
>
> OK, this is an EXTREMELY simplified explanation, but probably what
you
> are looking for. When you perform an FFT you are basically
calculating
> the energy present for a specfic frequency within the composite
signal
> over a fixed time period. FFts are said to have a certain number of
> 'taps', which are sometimes called 'buckets'. a signal sampled at
8000
> samples a second will have a useful rage of 0-4000hz,. analysign this
> signal with a 100 tap signal will split the original signal into 40
hz
> intervals. The results you get back are typically amplitude and
phase....................

------------------------------------



(You need to be a member of msp430 -- send a blank email to msp430-subscribe@yahoogroups.com )