[Top][All Lists]

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

Re: [Discuss-gnuradio] typical bandwidth?

From: Dave Emery
Subject: Re: [Discuss-gnuradio] typical bandwidth?
Date: Fri, 22 Mar 2002 03:25:20 -0500
User-agent: Mutt/1.2.5i

On Thu, Mar 21, 2002 at 04:16:44AM -0800, John Gilmore wrote:
> Thanks, Dave, for a great and detailed summary of the kinds of signals
> people might want to mess with.        I have just a tiny addendum.
> While most individual signals only use 30-50kHz, software-defined
> radios also let you build a "scanner" that can watch a whole band at
> once.  Unlike usual scanners which retune themselves frequently, a
> wideband scanner would be able to digitize a wider chunk of spectrum,
> and detect and extract multiple signals from it simultaneously, saving
> each to disk if desired.  E.g. one of the demo applications that comes
> with GNU Radio is the ability to tune two FM stations at once (you get
> mono sound from each one, out the PC's left and right audio channels).
> Such applications are limited by the incoming bandwidth available, CPU
> power available, and the differences in signal gain required on the
> various channels.

        Oh there is no doubt at all in my mind what can be done with
wideband A/Ds and fast processors,   One of the most interesting
applications in oonventional radio listening is very very fast scanners
that can  search a wide spectrum for intermittant brief signals, capture
their exact frequencies and other relevant characteristics and even
record the traffic, all in milliseconds.   One of the real limitations
of current commercial scanners and scanning transcievers is that the PLL
settling time and squelch response time greatly limits the rate the
radio will search a range of frequencies.  Few commercial radios do this
much faster than 100 to 200 frequencies a second  (and many are much
slower than that) and this greatly limits the ability to locate and
identify signals that last for a very few seconds or less, particularly
in large frequency ranges such as some used for military aircraft
communications and the like.

        Obviously a high res FFT coupled with wideband downconversion
and sampling of broad swaths of spectrum by using a fast A/D allows
search and identification of signals in very wide chunks of spectrum
with only a few PLL retunings at rates of tens of thousands of
frequencies a second.  And this is potentially very useful.  Probability
of intercept of brief burst (and sometimes covert) signals goes way up...

        Another interesting application is digitizing of chunks of spectrum
during particularly interesting events.  This might range from digitizing
part of 20 meters during a contest or 2 Meters during a band opening to
digitizing the 800 mhz public safety band in its entirety and saving
it during a particularly intense public emergency.

        And there are many times when one might wish to be able to
record all the traffic on all the trunk channels of a public safety
trunking system rather than trying to decide which talkgroups carried
the most interesting stuff and listen only to them - most of these
systems live in bandwidths comparable to what the current gnuradio is
playing with.

        Of course on a technical note - much of this sort of wideband
spectrum swath digitization and search and signal extraction requires
pretty damn good A/Ds in order to support the dynamic range and
intermodulation noise floor requirements of dealing with the world of
real rf spectrum with very strong close by signals as much as 80-100 db
over the minimum detectable signal at the noise floor in some cases and
quite typically 60-70 db above the weakest signals of interest.   It is
much easier to just use DSP in the last IF stage of a receiver where 
(in most cases, but of course not all) signal levels are more controlled
by the AGC and roofing filters and fewer intermodulation components show
up in band.

        Thus gnuradio is likely to be limited (at least in the short
range) by the affordability of fast high resolution A/Ds as much as
anything else - the ideal A/D is probably 18 or 20 bits with 75 db or
more dynamic range to the intermod floor and so far these are very very
expensive at interesting sample rates for wideband stuff.  It obviously
takes a much less sophisticated A/D than that to digitize a signal 
that has been filtered out from the morrass of other signals by analog
IF and roofing filters and adjusted in level by AGC, but then you
cannot be demodulating four other signals in other places in the band
at the same time.

        But yes, there are myriad reasons why wideband radios are

        Dave Emery N1PRE,  address@hidden  DIE Consulting, Weston, Mass. 
PGP fingerprint = 2047/4D7B08D1 DE 6E E1 CC 1F 1D 96 E2  5D 27 BD B0 24 88 C3 18

reply via email to

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