[hpsdr] Blackfin

[Lyle Johnson]

> All of the code I wrote for the AEA DSP1232/2232 in 1989 and the TAPR 
> DSP project before that was fixed point.  It had all of the 
> normalization stuff done.  After the FFT is optimized,  I do NOT view 
> this as the end of the world to accomplish.   The need for more dynamic 
> range than 16 bits in some applications and depending on the hardware 
> design (where is the agc done for example) may mitigate in favor of 
> using a wider fixed point or floating point.  The AEA box was 24 bit 
> DSP56001.

Wider fixed point is not readily available except from Freescale (ex 
Motorola).  TI has some 32-bit fixed point, (TMS320F280x and 281x 
series)  but the accumulators are only 32 bit, not the 72 to 80 bits 
that you'd probably want.  TMS320VC33 is floating point, cheap, low 
power, packaged with pins you cans older and probe, and free tols.

And I see TI has just announced 32-bit floating point "digital signal 
controllers" - TMS320F283x series.  Interesting parts...

Lyle KK7P


 1184185575.0