You are close to what a waveform analyzer does.  It is not mathematically very clean, 
but it is useful for watching frequency evolution over time.  But you need a time 
averaging buffer.
For each of the 6 segments, do an FFT and get the Mag and phase.   Have a buffer that 
is the same length as your phase and magnitude and as each segment is acquired (B is 
buffer and N is new signal)

B' = x*B + (1-x)*N

This gives a time average.  Where the time constant x goes from 0 to 1.  If X is zero, 
then you are only looking at the new signal.  If x is 1 then you are only looking at 
the past.  Usually it is expressed as a Number to average as an exponential average.  
But this way you can update the average in real time.

BTW, this is hardly a windows question!  The math involved in FFT is mostly platform 
independent!  :-)

-Scott


At 15:50 +0100 02/03/2004, Roman Golubovski wrote:
>I need help/guidance on how to use "FFT Spectrum (Mag-Phase).vi" to perform
>RMS averaging on sampled strip(s). DAQ produces strip of up to 120.000
>samples (max 120sec at max 1kS/s). After that I am asked to do the following:
>
>1) segment the strip into overlaping segments (equally sized)
>2) perform RMS averaging on them
>
>I am not familiar with this particular technological background and the
>"customer" also cannot explain how it is being done - he just wants me to
>perform what his signal analyzer does! I am not even sure the segmentation
>is compatible with the RMS averaging process.
>
>Anyway, lets assume the strip length is 200 samples, the egment size is 40
>samples and the overlap is 10 samples. Than obviously I can segment the
>strip into 6 segments starting at strip positions 0, 29,59,89,119,159.
>
>What do I do now? I gues I have to turn the segments into waveforms with
>dt=1/fs and t0=0? I guess I should put the "FFT Spectrum (Mag-Phase).vi" in
>a loop with 6 iterations feeding a segment-waveform into the "time signal"
>connector each iteration? I expect the output is NOT auto-indexed?
>
>How to apply it properly actually?


Reply via email to