discuss-gnuradio
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Discuss-gnuradio] gnuradio based spectrum analyzer


From: Ettus, Matt
Subject: RE: [Discuss-gnuradio] gnuradio based spectrum analyzer
Date: Thu, 25 Sep 2003 16:44:10 -0700

> 
> OK -- I need to explain a bit more what I'm up to.  There's a 
> ham radio 
> operating event where a station (W1AW, the headquarters 
> station of the US 
> amateur radio organization) transmits a series of 60 second 
> carriers on 
> several frequency bands.  The challenge is to measure those 
> signals with as 
> much accuracy as possible.  The signal is not incredibly 
> strong, and it's 
> subject to the vagaries of propagation and interference from 
> other sources 
> (and things like power line noise, etc.).  My technique is to 
> tune in the 
> unknown signal and inject a known reference frequency a few 
> tens of Hz 
> away, and then measure the delta between the two in the audio 
> passband -- 
> this way any drift in the receiver is common mode and affects the two 
> signals equally.

That is a good way of performing the measurement if you can't lock the rig
to your good oscillator.

> Last year I used the baudline spectrum analyzer software and 
> it worked 
> pretty well, but the maximum FFT size is 32768 which led to a 
> couple of 
> limitations.  I had to slow down the sample rate to 
> 400/second to get a 0 
> to 200Hz displayed bandwidth, which yielded a resolution of (I think) 
> 0.16Hz but required me to tune the notes to an awfully low 
> frequency, which 
> presents a few problems with the receiver interface.

There really aren't any limits to the fft width that gnuradio can use.
FFTW, which provides the routines, can do ffts up to millions of points.

It would be very easy to run all of the data through 2 or more ffts of
different resolutions in parallel.

> The waterfall display does a good job of letting you find the 
> main signal 
> against the noise, and also lets you track the propagation 
> variations and 
> pick the best spot in the history to do the measurement.

We don't have a waterall display yet, but it would just involve a little bit
of graphics hacking (in python if you wanted) to the specrtum display block.

Matt





reply via email to

[Prev in Thread] Current Thread [Next in Thread]