The report several times also mentions IM as part of the noise.
Raw Data -- Spectra With CW Keying – Clicks + IMD + Phase Noise
ts From Electronic Data: Post-processing of the raw data (Figs 3-5) provides far better resolution, and clearly shows the contribution of phase noise and keying waveshape. Averaging essentially makes the random component of the phase noise disappear. Differences between the rigs remain about the same.
For CW, these spectral plots tell the complete story for what is happening close to the carrier, because they include the effects of keying transients, IMD and phase noise. Because the spectral distribution of pink noise band limited to 100 Hz to 5 kHz closely approximates the spectrum of speech, a comparable, but wider, plot for SSB modulated by that test signal would provide a similar result for SSB.
iPad truncation of copy. Sorry.