[hpsdr] Future projects?

Murray Lang murray.lang at metoceanengineers.com
Mon Mar 12 15:48:08 PDT 2007


Naylor Jonathan wrote:
> ***** High Performance Software Defined Radio Discussion List *****
>
> [snip] 
> A comment was made about making frequency domain information available,
> I disagree. The problem comes down to what sort of frequency
> information? If we assume tha output of an FFT, which we use for the
> pretty graphics on an SDR console, then it really isn't good enough.
> Different programs need different information, for example a program
> for VLF would require FFT bins down to the mHz level, while WSJT
> requires 6 Hz bins. Then the question exists of how many FFT's to
> compute per second. For data reception, if an FFT is used, typically it
> is optimal to align the FFT calculation with the beginning of each
> symbol and to have the FFT size match the parameters of the signal. So
> I would strongly oppose the shipping of frequency information over the
> network, instead provide the raw audio so that the receiving
> application can do with it what it likes, which may not require an FFT
> at all. [snip]
>   
I see the FFT being the same as used in an SDR design such as described 
in "SDR for the Masses". In other words the SDR receiver is distributed. 
The reason I'm suggesting this approach is to reduce the load on both 
the network and the server CPU. The network is spared because rather 
than send the entire bandwidth of the acquired I/Q signal to each 
connection, only the segment of interest needs to be sent. This can't 
happen in the time domain unless the server also performs the IFFT, but 
the idea here is also to reduce the work of the server down to an 
absolute minimum so as to maximise the number of possible connections - 
apart from the initial FFT it's just pumping data around.
I suppose doing the IFFT at the server and sending the reduced bandwidth 
audio isn't that much of an extra drain on the server, but the 
client/stub is probably going to want to get it back to the frequency 
domain at some point anyway, so nothing is gained.

Murray VK6HL



 1173739688.0


More information about the Hpsdr mailing list