[hpsdr] Blackfin 32*32 bits multiply

[Phil Harman]

Hi Ramakrishnan,

The Pic-A-Star is basically a conventional superhet front end with a DSP IF. The main selectivity is done using a crystal filter (which has lots of bits!) with subsequent filtering done using DSP.  Since you have already provided most of your protection selectivity before the DSP stages you can get away with using less bits in the DSP. In addition, you can apply AGC to the stages before the crystal filter and hence increase the dynamic range so you can remain within a 16 bit boundary.

The problems start when your crystal filter is say 2.4kHz wide and your DSP filter is 500Hz. A signal then at 1kHz spacing many not be detected by your AGC if this is after the 500Hz DSP filter. If you monitor AGC in the 2.4kHz bandwidth and reduce the gain then the S/N in the 500Hz bandwidth can be degraded.   You can compensate for this loss of gain and that was the approach Doug Smith used in the Kachina 505DSP.

Hope that helps!

73's Phil...VK6APH 




 On Thu Aug 23  4:10 , Frank Brickle  sent:

>
>Ramakrishnan Muthukrishnan wrote:
>
>> Why do we need a floating point processor for narrow band SDR?
>
>Algorithmic generality and flexibility.
>
>> Yeah, I
>> know the answer "more dynamic range". I am looking for a more specific
>> answer. Where exactly are the dynamic range issues coming in? Energy
>> calculations? Can't we do with 16 bit samples?
>
>IIR filter coefficients, for example. There are certain classes of 
>Butterworth filters you can't realize using 16-bit fixed point because 
>of the dynamic range of the coefficients. This is especially unfortunate 
>when the processor forces you to rely *more* on IIR and *less* on FIR 
>filters.
>
>73
>Frank
>AB2KT
>
>
>_______________________________________________
>HPSDR Discussion List
>To post msg: hpsdr at hpsdr.org
>Subscription help: http://lists.hpsdr.org/listinfo.cgi/hpsdr-hpsdr.org
>HPSDR web page: http://hpsdr.org
>Archives: http://lists.hpsdr.org/pipermail/hpsdr-hpsdr.org/


 1187830622.0