> The whole point of a PLL is that the oscillator is free running, not a
> crystal one, this allows a large frequency swing. This signal is then
> divided down and compared to a reference crystal oscillator (also
> usually divided down!), this process produces an error signal, which is
> used to correct the free running oscillator which locks the oscillator
> solidly to the crystal reference. The output frequency can be changed by
> altering the division ratio in the divider chain, the actual resolution
> is dependent on how this is done - basically (and simplified) if your
> crystal reference is divided down to 1000Hz you will have 1000Hz
> resolution.

One thing I cannot seem to get my mind around is why... Why lock a an
oscillator's phase to a crystal reference. Why not just use the crystal
frequency and divide it as necessary? Is it the shape of the output wave
or something?