Re: [Discuss-gnuradio] SDR question

[Marcus D. Leech]

On 28/10/2011 5:31 PM, Andrew Rich wrote:
> Thanks Marcus
>
> So you can go outside the useable bandwidth, you just need to 
> understand that you will loose something as you move to the next chunk 
> of RF ?
Generally, the hardware samples at a fixed rate (USRP1 samples at 
64Msps, and USRP2/N2XX sample at 100Msps for example).  From
  the hosts point of view, if the tuned frequency is X, and the 
requested bandwidth is Y, your baseband extends from X-(Y/2) to X+(Y/2).
  Not sure what you mean by "loose something as you move to the next 
chunk of bandwidth".
> I saw an image of several MHz and a little decode window, but I guess 
> that is a decoding window, smaller than the SDR sampling window.
It is often the case that an actual application will bring in more 
bandwidth than it strictly needs--usually because the application-specific
  bandwidth is a little bit smaller than one of the integer fractions 
of the samplers input bandwidth.  So you typically bandpass-filter in 
software
  to whatever the bandwidth of your application is, and then possibly 
down-sample, or not, depending on the application.

Let's say your sampler front-end runs at 100Msps, and you really only 
want 75Khz of bandwidth.

On a USRP2 or N210, the maximum decimation value is 512, which produces 
a sample rate of 195312sps, (100e6/512) since this is
  complex-baseband, that 195312sps is *also* 195312Hz of usable 
bandwidth (in a sense, complex sampling "cheats" Nyquist).

So, you'd place a bandpass filter in the signal path to filter it down 
to 75Khz, before you did anything else with it in the signal flow.

> I want to use SDR for satellites and packet radio
>
> Does it meet a tnc / analogue radio specs ?
An SDR-based platform would have no trouble doing at least the 
modulation and radio parts of a TNC, although implementing something
  like AX.25 in Gnu Radio is likely a poor architectural choice.