So what sort of symbol rate becomes theoretically possible within a HF frequency band?
Of course, the 2.8 kHz bandwidth rule prohibits that in the ham bands. The FCC had originally declined to set a maximum bandwidth in their first ruling, but that caused everyone's head to explode (and that's one of the reasons this ruling has taken so long).
I'm ignorant of what the ramifications would be of this - can you explain why people would have responded like that?
The reason the FCC didn't want to set a limit is because it's arbitrary. For example, MIL-STD-188-110D (which is an open standard) has a 2400 baud 3.24 kHz mode. But now it's illegal.
I'd still call this a win. The 300baud limit in USA was keeping the whole field behind by preventimg world wide adoption of more efficient modems.
But I agree. Even with the 2.8 kHz limit, it's still a big step forward.
The bandwidth of a single carrier signal is symbol rate * (1 + RRC excess bandwidth). A typical excess bandwidth is 0.35, so 2000 syms/s * 1.35 = 2700 Hz.
You can use a tighter roll-off, like 0.2. 2400 syms/s * 1.2 = 2880 Hz. Unfortunately, the tighter roll-off increases the PAPR (Peak to Average Power Ratio) of the transmitted signal.
If you have more than 2 possible symbols, you can do much more than that; 256QAM could be getting 8 bits/second/Hz under very strong signal conditions.
This theorem states that the theoretical maximum channel capacity ("bandwidth" as measured by bits per second) is dependent on analog bandwidth (the other "bandwidth" lol) and signal to noise ratio.
Now that's the information theoretical maximum, but stuff like QAM as people mentioned, and other more advanced techniques like OFDM and "next layer" stuff like FEC can get you very close to the theoretical maximum.