[hpsdr] DttSP ovsv.c possible optimization?

[Chris Albertson]

--- Philip Covington <p.covington at gmail.com> wrote:

> ***** High Performance Software Defined Radio Discussion List *****
> 
> On 8/13/07, Robert McGwier <rwmcgwier at gmail.com> wrote:
> > Phil:
> >
> > Thank you.  I moved the scaling by p->scale into the constructor
> and
> > just commented out the scaling in the running routine (with a
> notation)
> > in ovsv.c in both the DttSP in PowerSDR and in ab2kt/dttsp-ng
> branch in
> > sdr_linux.
> >
> > PLEASE anytime someone finds something like this, point it out.
> >
> > Bob
> 
> For whatever reason, I woke up for a few minutes at about 3AM on
> Saturday morning with the idea that doing the pre-normalization to
> the
> filter was equivalent to doing the normalization after the inverse
> fft.   I guess I have been reading too much DSP stuff lately...

I once worked on an imaging radar system.  It was a down looking
system that was to fly on an aircraft for mapping puposes.  One
someone figured out that FFT was linear, they designed the system
such that the FFT and FFT-1 were in hardware.  (Special build FFT
chips)  This makes the software nearly trivial because usable image
data scrolls right off the FFT-1 output.

I said "nearly" because we had to rapidly re-compute the filter because
the impulse respose of the system changed in real-time as the aircraft
bounced around in the air.  This was 20 years ago so we had to resort
to some custom parallel hardware for this but today an old Pentium II
would do fine.

This was a very expensive system but now only could make FFT/FFT-1
chips using an FPGA.  After doing this a ham=style SDR could be run
with even a low-end micro controller. Simply pre-store all the filters
The linear nature of the FFT really does mean you can but
__everything__ into just that one step.

One more idea.  I think many people doing ham-SDR today are modeling LC
filters in software.  Basically building a software version of a
hardware radio.  I think you can do better by modeling what the signal
should look like at the receiver.  For example in CW mode once you have
the first hundred code elements you can average then you then know the
exact spectra of the signal and compute a "perfect" filter customized
for that transmitter and signal path.

In our radar project we computed what the return echo should look like
if it were aimed at a small spherical reflector in free space with the
ccurrent target to antenna geometry and relative motion.  We computed
FFT this and fed it into the multiply hardware along withthe FFT of
the actualy signal.

Lots of fun work to do....

Chris Albertson
  Home:   310-376-1029  chrisalbertson90278 at yahoo.com
  Office: 310-336-5189  Christopher.J.Albertson at aero.org
  KG6OMK


      ____________________________________________________________________________________
Shape Yahoo! in your own image.  Join our Network Research Panel today!   http://surveylink.yahoo.com/gmrs/yahoo_panel_invite.asp?a=7 



 1187027393.0