[time-nuts] Notes on tight-PLL performance versus TSC 5120A

[John Miles]

> 3) To look at several different noise *slopes*, looking for specific cases
> where filtering or integration might cause the instruments to respond
> differently depending on the spectral characteristics of the noise in a
> given tau region.
>
> Bruce's concern is largely that of (3)...

One other point that might not be clear: the 5062C does exhibit a wide
variety of noise types, but not at short timescales near that of either
Warren's RC time constant or my low-pass decimation filter.  As this PN plot
shows, the 10811 is transitioning from 1/f^2 to 1/f noise (i.e., white
frequency to flicker phase) by the 100-150 Hz offset where Warren's RC
integrator is doing its thing, and the 5062C is, too, since its disciplining
loop BW is somewhat lower:
http://www.thegleam.com/ke5fx/pn_slopes.gif

The medium-term ADEV measurement with the 5062C tells us that there are
probably no time-variant effects that would jeopardize measurements at
longer timescales regardless of their slope, but -- since the 5062C's noise
statistics are so similar to the 10811's at 100 Hz -- it *doesn't* tell us
as much about the quality of the integration or filtering as I may have
suggested.

As Warren says, at short timescales the technique's accuracy depends on
oversampling, which I tend to agree has an effect similar to higher-quality
integration.  To put this assertion to a real-world test, one might use the
tight PLL to observe both an oscillator which is still in 1/f^3 territory at
t=0.01s, and one which is already well into its white-phase noise region by
that point.

-- john, KE5FX