[hpsdr] Blackfin

[Frank Brickle]

On 7/11/07, Ramakrishnan Muthukrishnan <vu3rdd at gmail.com> wrote:
>
>
> I believe dttsp is definitely much more amenable to fixed point
> conversion than the above stuff I worked on, where FFT outputs went
> thru a series of operations resulting in values of very high dynamic
> range.


Many months ago, Bob went through the painstaking exercise of normalizing
the gain through both the entire signal path on both RX and TX. It was a
noble job and somebody had to do it, but now it's in the situation where
*all* of the nonlinear amplitude stuff lives either at the very front or the
very end of the path. The remainder of the path is all linear and normalized
to unity.

I can't emphasize enough how important and painful that was. You will be
able to see the benefits if you start looking at where scaling needs to be
done in a fixed point version. It isn't something that gets a lot of
attention but a lot of the current progress hinges on what he did in a
critical way.

73
Frank
AB2KT
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openhpsdr.org/pipermail/hpsdr-openhpsdr.org/attachments/20070711/f91b989a/attachment-0004.htm>