Re: [Discuss-gnuradio] USB2 <-> fast ADC & DAC

[Dave Emery]

On Tue, May 27, 2003 at 02:11:17PM -0700, Chris Albertson wrote:
> 
> There are two areas were software defined radio is usful
> The requirements for ADC are different
> 
> 1) Wide bandwidth, strong signals like broadcast HDTV
> 2) Narrow bandwidth, very weal signals. One example of 
>    this is HAM weak signal CW.  The bandwidth is only about
>    500Hz but the signal/noise ratio is very low requiring
>    a 24 bit ADC and nearly 100db dynamic range if any signal
>    is to be captured at all
> 
> So it's pretty clear you can't build a totally generic
> ADC. You need to choose your application area
> 

        I would generalize this a bit more (something I think this
group has sometimes glossed over, in fact).

        There are really two places one might put a DSP in a receiver.
One is before the channel selectivity, expecting the DSP to provide
the channel filtering for each channel (of several perhaps). And 
the second is after most of the channel filtering (or at least much
of it).
        In the second case the DSP is mostly involved in demodulation
and maybe some final filtering but in the first case, particularly
for fast wideband digitization of a whole band of frequencies and
all included signals, the DSP provides the channel selectivity.

        And in the first case the dynamic range of the conversion 
process must match that of ALL signals within the WHOLE band of interest
(being sampled and digitized) , which unfortunately is often a very tall
order even for the best of technology available today.

        What this means is that using a wideband downconversion of
say the entire 2 meter band or the entire 800 mhz public safety band
or all of 20 meters and digitizing it all in one converter with a suitable
sample rate expecting to split out individual narrowband signals on
one or more frequencies doesn't work so well unless the A/D has
lots of spurious free dynamic range and resolution.   This is because
the gain has to be set so as not to clip on the strongest signals in
the band and that implies that weak signals which may easily be 80-100
db below the strongest signal but still well above the noise floor in
their channel have to still be reproduced accurately.  Thus a big SFDR
is required.

        This of course is not the case - usually - with wideband signals
such as HDTV which may (one signal) occupy a similar bandwidth but which
do not contain signal components of importance that need to be accurately
reproduced 80-100 db below the composite signal power.

        Thus if one's goal is to demodulate a wideband signal of some
sort, and one can provide basic channel selectivity in hardware (IF filters
or whatever) one can get by with modest SFDR in most cases provided one
has some kind of gain control/AGC to ensure the signal level going into
the converter is right.

        And if ones goal is to demodulate a weak narrowband signal and
one has the IF filtering there, there may be no need of great SFDR either.

        But the dream application of digitizing an entire band filled with
narrow signals of interest (or wide signals of greatly varying amplitude)
pushes the limits of the A/D much much harder and 12 bit solutions
look much less interesting in real signal environments.

        Also another issue is whether the input of the ADC contains
baseband signals going all the way to DC (as in I and Q baseband down
conversion) or just IF.  Transformer coupling IF is great and gets
around a lot of problems, but may or may not be suitable for direct
conversion to DC type applications.

-- 
        Dave Emery N1PRE,  address@hidden  DIE Consulting, Weston, Mass 02493
PGP fingerprint 1024D/8074C7AB 094B E58B 4F74 00C2 D8A6 B987 FB7D F8BA 8074 C7AB