Jackson Labs PHASE STATION 53100A, User Manual

Page: 48 / 225

PhaseStation 53100A User’s Manual V1.04 25-Apr-2022 48
Common artifacts in ADEV and related measurements
High-resolution plots captured by the PhaseStation 53100A can reveal artifacts that
don’t seem to be present when the same measurement is made by other instruments.
For example, spurs due to AC power coupling or ground loops, discussed below, appear
very different when acquired at high sample rates and rendered with 20 or more ADEV
bins per decade. Line-related spurs may not have been resolvable before, but that
doesn’t mean they weren’t present!
Measurement bandwidth and τ
0
One artifact that
does
originate in the 53100A
driver software is shown in the plot at right. Four
successive ADEV acquisitions are shown, each
conducted at a different noise-equivalent
measurement bandwidth (ENBW) setting. Near
the beginning of each trace, some “droop” is
present that doesn’t accurately reflect the
behavior of the DUT.
What causes this artifact? Stein
7
has shown that the optimum ENBW for a given ADEV
τ
0
8
is simply the 1/(2 * τ
0
) Hz Nyquist rate. This bandwidth is unachievable with a non-
ideal antialiasing filter, so the 53100A driver internally acquires oversampled phase data
that corresponds
to an artificially short τ
0
interval relative to the selected ENBW. If
allowed to appear on an ADEV plot, the antialiasing filter’s response is easy to mistake
for a real effect.
TimeLab supports two different ways to avoid displaying invalid data near τ
0
. First,
Trace → Clip xDEV traces by noise bandwidth ( Ctrl-b ) is enabled by default. (This option
was turned off to demonstrate the issue above.) With noise bandwidth clipping
enabled, as seen in the figure at left,
TimeLab does not render any portion of
an xDEV trace at taus shorter than
1/(2 * ENBW) seconds.
7
The Allan Variance – Challenges and Opportunities , Samuel R Stein, Symmetricom, Inc., Boulder, Colorado USA
8
τ
0
refers to the very first tau at the left end of an xDEV trace, which typically corresponds to the period of the phase or frequency samples
being analyzed.
Figure 12: Artifact near beginning of unclipped traces
Figure 13: Traces clipped by bandwidth