That would depend on whether the pulses were shaped for optimimum
bandwith or just keyed on and off with (relatively) fast rise and
fall times. I'd have to check the text book, but the first ought
to put nearly everything into about +- 7.5Hz and the latter
would generate a signal with significant power to several 10s
of Hz and a theoretically infinite bandwidth.
> but an FFT of 0.1 seconds indicates a binwidth of 10Hz. It
That could be the worst possible bin width for this signal in search
mode, as the sidebands wouldn't show up at all if you were in phase with
the modulation (FFTs and DFTs in general do have some artefacts due to
their finite nature). It would be the optimimum width for recovering
the modulation, but you couldn't use any integration, and you would have
to synchronise the start of the FFT with the start of the bits.
> The 2.8 minute message was only transmitted once. Averaging
> longer than 2.8 minutes would be counter productive as it
> would degrade the SNR.
Just for clarification, we are talking here from a hindsight point of view
as far as the receiver is concerned. 2.8 m is rather a short integration
time for a general search.