Re: Confusion about FFT resolution - PLEASE HELP
- From: amerdsp <amerdsp@xxxxxxxxxxx>
- Date: Mon, 28 Apr 2008 18:40:44 -0700 (PDT)
N = 1024 in your case...
You do not care about the bytes after you assemble your audio
vector...
On Apr 28, 8:03 pm, "cpptutor2...@xxxxxxxxx" <cpptutor2...@xxxxxxxxx>
wrote:
Could some DSP guru please clarify the following ? I have a simple
audio application, in which at records time about 20000 bytes from the
received audio signal. I am using PCM encoding with sampling frequency
16000, 16 bits, mono channel, little endian and signed.
I wish to do FFT with the collected data. Since I collect bytes, I
iterate through the buffer, collecting two bytes at a time, and with
simple processing convert these byte pairs to shorts. I collect
altogether 1024 shorts, by processing the first 2048 bytes of the raw
data.
From basic DSP principles we know that if 'N' data samples have been
collected at a sampling frequency of 'fs' Hz, then the FFT resolution
is:
fs/N.
So, in my case, is the resolution 16000/20000 or 16000/1024 ? I am
confused. Please let me know.
.
- References:
- Confusion about FFT resolution - PLEASE HELP
- From: cpptutor2000@xxxxxxxxx
- Confusion about FFT resolution - PLEASE HELP
- Prev by Date: normalised cross correlation by FFT
- Next by Date: Re: Is octave good to do image processing?
- Previous by thread: Confusion about FFT resolution - PLEASE HELP
- Next by thread: normalised cross correlation by FFT
- Index(es):