(Sorry to break the threading, I just subscribed. I’ve been a long
time reader of this list but this is my first post.)
I’m pretty interested to see d-star being discussed here, since this
has been a topic of interest for me for a while.
It is my position that in the US AMBE is already prohibited on amateur
bands under the FCC part 97.113 (a) (4) prohibition against
encryption, and that any codec which is incompletely documented,
patented, incredibly expensive, prohibited for software implementation
could be described no more accurately than a “codes or ciphers
intended to obscure the meaning”.
I’ve drafted a complaint to the FCC to that effect, but have not yet
fired it off because I do not yet have a have a good alternative.
To that end I’ve been working on creating a fairly narrowband OFDM
based modem for ham use using gnuradio. The intention was to define a
technique which used OFDM, QAM16, LDPC, and Speex to deliver better
than better than typical narrow FM audio quality under signal
conditions which would not otherwise grant copyable speech.
But I’ve run into issues with managing peak power: Naive OFDM has a
very significant difference between peak and average power and since
much of the interesting target equipment is power limited that is a
serious issue. The significant gap between peak and average power also
makes level adjustment for users using soundcard-based software modems
much harder. There are a number of approaches to controlling OFDM
peak power, but everything I’ve tried that works well has been
computationally expensive, or not especially robust against weak
signal conditions.
I had not previously considered forking d-star’s signaling and simply
using Speex instead of ambe, as has been suggested here.
Speex is a truly excellent codec for speech. At its target bitrates it
is arguably one of the the best performing voice codec available. It
is also fully open, not patent encumbered, and the freely licensed
reference implementation has been ported to a number of DSPs and low
power CPUs. At 8-32kbit/sec (32kbit for wide band) speex is hard to
argue with… For higher bitrates and more general purpose audio Xiph
now has a new ultra low latency free codec called CELT in development
(http://www.celt-codec.org/).
However at the bitrate that AMBE is used at in d-star speex’s
delivered quality is not good at all. The only reason speex has any
support for those bitrates is to encode background noise during quiet
times in VBR applications. This may present a challenge to anyone
trying to stuff speex into d-star. The free world may yet have a
speech codec for AMBE-like bitrates coming, but one does not publicly
exist yet.
If anyone takes any action on this front I’d be interested in helping
out. If something picks up off this list please keep me in the loop.
Thanks!