Hi GNURadio fellows,
considering that this list has grown to something highly relevant in
Software Defined Radio I thought it would have been a good idea to share
here a few thoughts I’ve been having since long and as well as a result
that
was just achieved.
Since a few months after my first approach to SDR in 2006,
I thought I picked up two major facts about the technology:
.:. SDR infinite potential lying for sure in its flexibility but, even
more
relevantly, in its ability to bypass
the costly HW-level design stage which is embedded in any
traditional
radio design/production process
.:. Its equally infinite power-inefficiency compared to traditional,
HW-implemented competitor technologies.
In fact, ease of development as well as flexibility appear to be
inversely proportional to power efficiency.
The latter being in my opinion the reason for which SDR has been growing
for
ages up to now but has never “exploded” as we could expect from a
technology
cutting away a conspicuous part of the design costs of any radio system.
Actually, flexibility and cost-efficiency, though considerable, do not
appear to be sufficient motivation for accepting to upscale power
requirements (at a given computational cost yielded by the implemented
wireless standard) by a factor which typically is in [100 ; 300].
Whether right or wrong, by working with these thoughts in mind, during
the
research I’m carrying on at the University of Pisa, Italy while doing my
PhD
here, I developed a novel implementation technique targeted at
software-implemented Signal Processing over General Purpose CPUs or DSPs
which we (at DSPCoLa lab, http://dspcola.iet.unipi.it ) call “MA”.
Current research results have shown that MA was able to increase by
slightly
more than one order of magnitude the power efficiency of a traditionally
implemented (MA-free) SDR.
By applying such “MA” technology to the ETSI DVB-T receiver chain with
the
help of:
Mario Di Dio (former master thesis student, now PhD Student at DSPCoLa)
Luca ROSE (former master thesis student at DSPCoLa, now PhD student
at
Supélec Paris)
we obtained the receiving companion of Soft-DVB: SR-DVB.
Standing for Software Receiver - DVB,
SR-DVB is a fully software (all signal processing is done in pure C++
over
the host computer) ETSI DVB-T receiver capable of running realtime
while providing 11.612 Mbps throughput
and absorbing less than 50% of computational resources available over an
Intel Q9400, 2.66 GHz CPU.
As long as MA was applied only to the two computationally-heaviest
blocks of
the receive chain (i.e. Viterbi Decoding and OFDM synch), we believe
that
considerable margins for improvement of the presented result do exist.
They
will be explored in the next months.
SR-DVB will be presented in Karlsruhe at WSR 10
as the article:
“A Fully Software ETSI DVB-T Receiver Based on the USRP”
during such presentation also MA technology will be briefly outlined.
A demo video of our proof-of-concept receiver is available at
www.legalepellegrini.it/ing/SR-DVB_demo_long.VRO
as usual, mplayer or VLC wil play this camcorder mpeg2 viedo easily.
Best regards to all writers and readers of the list
vincenzo
