Re: Subject: [Discuss-gnuradio] APCO 25

[Dave Emery]

uOn Tue, Jan 21, 2003 at 02:59:28AM -0000, Matt Ettus wrote:
> 
> > >  This sounds like a better route. I can take my IF from the output
> > > of the WFM ceramic filter rather than its input so I will then only
> > > have a bandwidth of 200 KHz or so. I have 10.8 MHz xtal somewhere and
> > > have some SBL-1 mixers in my junkbox. Now you have set me thinking ..
> > >
> > > Regards
> >
> >
> >     You might have some aliases here I'm afraid.   Most receivers
> > (R-7000, R-7100, R8500, AOR5000-3) that have 10.7 mhz outputs have very
> > wideband 6 mhz or so wide outputs designed to even handle NTSC/PAL video
> > (ICOM eveb makes/made a video demod for this output).   So you obviously
> > need serious added selectivity if you are going to digitize the IF
> > directly and aviod terrible aliases.
> 
> Why?  If you are digitizing with a soundcard, it has antialiasing filters
> already.  Besides, you can get narrow 10.7 MHz filters that are the size of
> small tantalum caps for less than a dollar from digikey.
>
        What I was refering to here is taking the unfiltered 10.7 mhz,
mixing it with 10.8 mhz and feeding the result into a soundcard.
Depending on your use of terms the other sideband on the other side of
the 10.8 mhz LO (the image) is properly called an alias in that
situation, though in fact you may not be seeing the more classic alias
due to downconversion of energy on the other side of the sampling clock
because of the low pass filtering in the soundcard.

        And I will readily agree that if one uses a decent 10.7 mhz
bandpass filter of a bandwidth selected to be compatible with  the
sampling rate used, one might be able to reject enough of the image
(10.8 mhz and up in this example) to not have problems.

        But this project presumably involves working with real tetra
signals in a (potentially) dense rf environment, so one would certainly
like to have lots of rejection of signals 200 khz away and that in this
example is entirely provided by the skirts of the 10.7 mhz bandpass
filter.

        And my original comment was primarily directed at Ian's proposal
to use the existing wfm mode ceramic or xtal filter - my guess would be
that its rejection 150-200 khz from the center of its passband would not
be all that great.   But I could be wrong.      
 
        And indeed, using a better filter would likely improve things
(or even just another one cascaded with the first).  And they are
available though in what exact bandwidths I don't immediately know.

> >     But these wfm filters aren't all that sharp (probably with a
> > LO at 10.8 mhz the close in part of the other sideband (the image)
> > would be down only 15-25 db or so).  I guess you could use an
> > image rejecting mixer, but these are less common in junk boxes.
> 
> see above
> 
> >     The only problem with doing this with a stock FM broadcast bw
> > filter (~150 khz) is that you have to be able to digitize say 200 khz
> > of bandwidth with the center at around 100 khz and this rather
> > implies 400 ksamples per second sampling.
> 
> Not true.  Soundcards have their own antialiasing.  Very good antialiasing, I
> might add, since they use delta-sigma converters with very high oversampling
> ratios.
> 

        I must be missing something here.  Perhaps I am. I am getting
pretty senile these days I'm afraid...  The normal Nyquist criterion
would require a 0-200 khz signal centered on 100 khz (what Ian was
proposing) to be sampled at 2F where F is the highest frequency
component of interest (in this case 200 khz or so).   This is pretty
fundemental sampling theory I think.   Now I recognize that delta-sigma
type converters can provide a good intrisic low pass filtering by
oversampling which moves the ailiases way up in frequency, but still we
are talking about a 200 khz bandwidth here centered at 100 khz.  And Mr.
Nyquist had a point...

        Which brings up an ignorant question - of the available high
end sound cards with 96 khz sampling, do any of them use external
(not part of an ASIC) delta sigma converters, and if so are they clocked
when running at 96 khz at the full spec'd max of the chip or could one
hack the card to further up the sample clock rate.   And would the 
USB or PCI logic support this at all if you did.   Yes, I know it
is a dumb question, but I thought I'd ask it anyway.  

        Certainly for gnu-radio one might get by with somewhat reduced
specs but a higher sampling rate quite nicely if the PCI or USB logic
could handle faster sample streams and the converter could be clocked
faster too.   And there are significant numbers of signals that 200 to
400 k samples/sec would digitize that 96 k samples/sec does not handle.



-- 
        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