Forums

Odd Wave Sections

Started by phan...@stonewolfsoft.co.uk September 13, 2011
Greetings,

This might seem like a daft question to some as it imvolves a pretty fundamental bit of theory, but for for the intended application (which I cant go into at present), I need everything intact. No filters, window functions etc. are allowed.

If you have a 2^n sample section containing a pure sine wave which has an arbritrary frequency (ie. you have an incomplete number of cycles), is there a way to guarantee that no stray frequencies turn up in an FFT? It is pretty obvious to us a humans that we are dealing with a single frequency, and we dont care about the accidental edge that occurs, but is there a way to realise this algorihmically?
Phan-

> This might seem like a daft question to some as it imvolves
> a pretty fundamental bit of theory, but for for the
> intended application (which I cant go into at present), I
> need everything intact. No filters, window functions etc.
> are allowed.
>
> If you have a 2^n sample section containing a pure sine
> wave which has an arbritrary frequency (ie. you have an
> incomplete number of cycles), is there a way to guarantee
> that no stray frequencies turn up in an FFT? It is pretty
> obvious to us a humans that we are dealing with a single
> frequency, and we dont care about the accidental edge that
> occurs, but is there a way to realise this algorihmically?

We do care. If you string together such wave sections and play out, people will perceive a buzz or "rough edge" along
with the basic tone.

The human ear is very sensitive to even tiny deviations from pure tone. You can put just one altered sample in 10,000
samples of a 1 kHz sine wave (8 kHz Fs) and hear it without problem.

-Jeff
Hi all

Phan

I'm not sure if I understood right, but since you are dealing with just only
one frequency at a time can you take the FFT and consider only the strongest
frequency (discard any frequencies except for the peak)?

Hugs
Ricardo Baptista