<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000066">
David McQuate wrote:
<blockquote cite="mid:4A2D8936.70204@sonic.net" type="cite">
<pre wrap="">***** High Performance Software Defined Radio Discussion List *****
If I have a block of samples (say N = 4096, fs = 122.88 Msps),
a straightforward FFT produces data with frequency resolution of 30kHz.
Is there a technique that will allow me to obtain information on a finer
frequency grid (say 1kHz or 300Hz or ...)--other than "simply" acquiring
additional data, extending the length of the time record?
I've looked at Chirp-Z transforms and polyphase decimation, but neither
seems to produce higher frequency resolution. I've also looked into DttSP,
but not found a solution there.
Am I hoping for the impossible?
Suggestions please.</pre>
</blockquote>
What if you take the time series data _before_ the 4096 points FFT, and
multiply it by a complex<br>
exponential so to bring the spot of interest at zero Hz, then you
low-pass filter and decimate it by <br>
a convenient factor. At this point you have a much reduced sampling
rate and doing now an FFT<br>
with the same number of points will give a much greater resolution.<br>
<br>
When you want to move up or down in frequency, you just change the
value of the complex exponential,<br>
which can be easily done in a semi-continuous way, given the high
sampling rate (122.88 MHz) you start with.<br>
<br>
73 Alberto I2PHD<br>
<br>
</body>
</html>
1244641857.0