How to select offset frequency for phase noise measurement?

When reviewing datasheets or papers on oscillators, phase noise is obviously an important parameter, but I see that the phase noise is highlighted at one offset frequency from the carrier in particular, usually selected to be somewhere between about 1 kHz out to 10 MHz.

If you have an oscillator, how you pick the offset frequency at which to specify your phase noise performance? Do you aim to provide a measurement in the $$\1/f\$$ or $$\1/f^2\$$ regions, for example (or one of the other regions illustrated on this page)?

For many systems (such as GSM, where an integrated phaseNoise of 8 degrees may be the system bit_error budget for the LocalOscillator under Synthesizer feedback control), the entire phase noise curve, from DC out to Symbol Rate, may be documented.

===================================

Additionally, the wideband noise, out past symbol rate, will FM/PM modulate the transmitter power, and cause de_sensing of close_in channels.

For most free running oscillators, especially VCOs, the phase noise profile can be approximated by three main regions.

a) There's a 1/f region, which covers many decades, and extends from close in, to far out
b) There's a flat noise floor far out
c) There's a region close to carrier which rises more steeply than 1/f

For synthesiser manufacturers wanting to purchase VCOs, they are only really interested in the first two

a) Anywhere on the 1/f region. It's not too important where, as it can be roughly scaled to any offset by knowing it's 1/f. This governs the gain your loop is going to have to provide to pull the VCO noise down to your target.
b) The noise floor. It doesn't matter where, as it's flat. This cannot be controlled by a PLL, so it's going to govern the ultimate noise floor of the synthesiser.
c) Generally they're not too fussed about close in, as if the PLL can handle the 1/f region, it will certainly handle close in.

In the days of audio modulation, 1 kHz was a common offset for specification for obvious reasons. As digital systems emphasise larger offsets, VCOs will often include a specification at wider offsets, often 100 kHz. It may make a few dBs difference, as the 1/f region is never exactly 1/f.

On the other hand, when you specify a synthesiser, the entire noise profile may well be significant, depending on what system your synth is going to be driving. Integrated phase deviation across the demodulation bandwidth is important for demodulation accuracy. Phase noise at channel and system spacings is important for avoiding inter-channel and inter-system interference, both as victim and aggressor.

Close-in performance gets to be important when you're specifying references for systems and synthesisers.

• Thank you for your helpful answer, Neil_UK. From what I can understand from most references, the 1/f region corresponds to the 6 dB/octave or 20 dB/decade region of the phase noise curve, is that correct? Which is interesting, because that corresponds to 1/f^2 in Leeson's equation... is that because it's a 1/f voltage noise and gets squared to plot phase noise in dBc? Nov 14, 2020 at 14:23