The attached PowerPoint presentation has been presented as part of National Instruments' 2011 RF Roadshow.
Here are some of the key takeaways:
'Phase-coherent' is not a strictly defined term. In this presentation, it is interpreted as signals/systems operating at the same frequency with a consistent phase relationship.
Phase-coherency can be achieved for RF analyzers/generators by sharing a common reference clock amongst all channels and having each channel synthesize its own LO signal, or by sharing a single LO signal amongst all channels directly.
Sharing a common reference clock results in each channel using a PLL to synthesize that channel's LO signal.
Each channel's LO will subsequently have unique phase-noise. Even if the phase-noise profile/performance is similar on each channel, the instantaneous phase noise on each channel will not be the same.
Part of each channel's PLL is the divider, and usually there is no mechanism in place to keep a consistent divder state for all of the channels, resulting in possible changes in the mean phase delta between various channels. This makes it difficult/impossible to calibrate channel-to-channel phase relationships and have that calibration stay valid from run to run, and across reboots/power cycles of the equipment.
Sharing a common LO signal directly results in one channel using a PLL to synthesize all channels' LO signal.
All channels receive the same LO signal, so all channels will see the same instantaneous phase noise. The net effect of this is that the instanatanoues phase changes due to phase noise are the same on every channel, and therefore cancel out.
The uncertainties introduced by the PLL divider is still there, but since each channel is using the same LO, all channels would move together so again, the net effect is that these changes cancel out.
Two proposed metrics for measuring phase-coherency performance are:
The mean channel-to-channel phase delta - What is the average phase offset between channels, and how does it change over time/temperature and from run-to-run/reboots/power cycles?
The channel-to-channel phase delta variance/standard deviation - How is the phase offset between channels varying around the mean in the short-term?
Sharing a common 10 MHz reference, with each channel using its own PLL for LO synthesis, results in unique phase noise on each channel that adds together and results in greater channel-to-channel phase delta variance in the short term.
Sharing a common 10 MHz reference, with each channel using its own PLL for LO synthesis, results in changes to the mean phase delta between channels unless the PLL divider states are somehow managed and kept consistent, which is atypical.
Sharing a common LO directly results in superior short-term channel-to-channel phase delta variance performance, and will not have the mean phase delta vary based on PLL dividers.
Sharing a common LO directly also has the benefit of making short-term channel-to-channel phase delta variance immune to carrier frequency.
Phase noise of signal generators (i.e. the LO used in a generator or analyzer) typically degrades as you go higher in frequency.
If you are sharing a common 10 MHz reference clock and each channel is individually synthesizing an LO, the greater, unique phase noise on each channel will add together with the greater, unique phase noise of other channels to create overall greater channel-to-channel phase delta variance.
When sharing an LO directly, although the LO signal may have degraded phase noise performance at higher frequencies, all channels see the same degradation and see the same instanatanoues phase noise, and therefore cancel out. This results in the channel-to-channel phase delta variance being immune to degradation in phase noise performance at higher and higher frequencies.
When sharing an LO signal directly, the driving factor behind instantaneous channel-to-channel phase delta variance becomes SNR.
As SNR on each channel degrades, phase delta variance around the mean increases.
Since the phase noise is the same on each channel, thermal noise remains as the main contributor to unique differences between channels.
Several plots in the presentation serve to illustrate this by lowering signal level and/or raising noise level.
Temperature considerations are the driving factor in mean phase delta performance over time.
Sharing a 10 MHz reference will cause each channel's temperature gradients to move its phase around.
Sharing an LO directly has the benefit that if temperature changes cause slight phase shifts in the LO signal phase, all channels see this change together and it cancels out.
This, combined with the compact form factor of the PXI platform and the resulting temperature stability within the chassis, results in very well behaved mean phase delta stability over time, with plots showing less than a few tenths of a degree of wander over periods of time as long as 10 hours.
In summary, sharing an LO signal directly has many advantages over sharing a common reference clock, though the instruments must be designed to accomodate this by having the ability to import/export LO signals.