Hi Sascha, On Sat, Jul 24, 2010 at 08:41:41PM +0200, sascha wrote: > > As mentionned OpenBTS laurent's decomposition demod seems to be way > > better than the current one (from the limited testing I did). Another > > benefit is that you can exploit CUDA _a_lot_ for the first stage of a > > multi ARFCN receiver. (when you do the math you'll see that things fit > > together nicely and you end up with a bunch of complex MACs and at the > > output you have N channels of I/Q samples pre-multiplied for > > laurent's). > > i am currently implementing a freq_xlating_fir_filter_ccf and a > fractional_interpolator_cc in cuda.
fractional resampling is one of the steps required if your sample clock
is truly free-running and not tunable - like in a typical "true" SDR
receiver like USRP1/USRP2. It is the most cpu-intensive operation
of the receive process, and having that in CUDA (or the ATI equivalent)
is definitely going to help.
> next would be the viterbi decoder.
that one is also important, but it is only the last step. Make sure you
implement a soft-input viterbi decoder.
> is this not the optimal algorithm to implement?
What is also needed is the actual demodulator ahead of all this,
i.e. what you do after the resampling and before the viterbi decoder.
Something that gives you the 'soft output' bits that you feed into
the 'soft input' viterbi. And that is where OpenBTS is much better
than what you see in the current airprobe code.
Regards,
Harald
--
- Harald Welte <[email protected]> http://laforge.gnumonks.org/
============================================================================
"Privacy in residential applications is a desirable marketing option."
(ETSI EN 300 175-7 Ch. A6)
signature.asc
Description: Digital signature
_______________________________________________ A51 mailing list [email protected] http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51
