Unfortunately there's no readily available description of
how BPL works, so none of us can really claim he knows exactly
what he's talking about. My general understanding is that
BPL uses a set of spaced carriers, each modulated at some
reasonably low data rate. These carriers are arrayed, I think,
over a range like 2MHz to 75 Mz or maybe 2 - 40 or something
of that order of magnitude.
The transmitted signal is not a pure digital baseband signal -
anytime any wideband signal has to propagate over any appreciable
distance this can't be used because the medium be it wire or "ether"
has a response which varies in frequency, and is not flat enough
over the passband of the entire signal bandwidth. There are two
broad classes of approach to combatting this (very oversimplified).
A single carrier approach takes the baseband signal and uses
it to modulate a carrier. This moves the spectrum away from DC
which most media don't like (including power lines as you note in the
paragraph I've quoted below). The resulting signal would have a
natural bandwidth of a modest multiple of the symbol rate
as you have noted in another post. Even this is usually too wide
in most applications, so the baseband pulses are shaped in time
to concentrate them (around DC at baseband, but around the carrier
when using the signal to modulate one). This pulse shaping causes the
pulses to have very long time durations compared to the signalling
interval, but a matched filter is used at the receiver, to
re-concentrate the energy temporally.
The other broad approach is to use a bunch of evenly spaced RF
carriers (as Frank has suggested, and is what I think they actually
do). The tones are spaced at frequency intervals of 1/(signalling
interval), and the tradeoff between length and frequency spacing is
a design decision depending on medium - it does not affect the overall
data rate. You can key twice as fast but then carry only half as many
tones in a given bandwidth. This is OFDM. In this case, since the
signalling rate of each tone is only a small fraction of the total
bandwidth the fact that the effective bandwidth of a rectangular pulse
is 3 or 5 or whatever times the signalling frequency doesn't effect the
total bandwidth by much since the total bandwidth is that of hundreds or
thousands of tones. Again, like the case of single carrier, since each
tone modulates an RF carrier, there is no LF energy in the resulting
signal. Now the medium is roughly flat over the effective BW of each
individual tone, and all is basically well.
Lastly there was a confusion in the thread between harmonics and excess
bandwidth. A randomly modulated square wave does not have discrete
harmonics, because the modulation eliminates the periodicity. It does
have excess BW as per your discussion of the need for sharp edges.
Again in practice the pulses used are not rectangular so the spectrum
does not have the ideal (sin(x)/x)^2 roll-off, but something that falls
off much more quickly, but there is still significant energy beyond
the Nyquist frequency.
> I haven't read how BPL is supposed to work but is it reasonable to
> expect that a encoding scheme would be used that would shift the
> spectrum requirements downward so that increased coupling would be
> needed across the transformers in the power system?