From what I've read, the spectral properties of the modulation largely depend on the rate of change of the signal phase - "a signal in which the phase changes stepwise has a broad spectrum, well beyond the allocated radio channel". I don't know if I understand this correctly.
An example of the spectrum of QPSK modulation with a random signal causing step changes in the output signal is shown in the figure below:
The book I am using says that "if we assume that the bandwidth of the radio channel allocated to this emission is \$B_k = 1/T\$, then a significant portion of the signal power (approximately a quarter) is emitted outside the bandwidth of that channel. The appearance of such strong interference in other radio channels is unacceptable (moreover - the band in which 99% of signal power is transmitted is approximately ten times wider than \$B_k\$)"
What does it result from? Why ~1/4 of the signal power is emitted outside the bandwidth of this channel? How do phase spikes affect this? And why is the bandwidth in which 99% of the signal power is transmitted approximately ten times wider than \$ B_k \$?