Stanford Research Systems SR785, Operating Manual And Programming Reference

Page: 105 / 546

FFT Averaging 2-25
SR785 Dynamic Signal Analyzer
Since signed values are combined in the mean, random signals tend to average to zero.
This reduces the noise floor since random signals are not phase coherent from
measurement to measurement. Only signals with a constant phase have real and
imaginary parts which repeat from time record to time record and are preserved. Vector
averaging can substantially improve the dynamic range of a measurement as long as the
signals of interest have stable phases.
For single channel measurements, vector averaging requires a trigger. The signal of
interest MUST be phase synchronous with the trigger to have a stable phase.
For a two channel measurement, the phase is relative between Channel 2 and Channel 1.
As long as the signals of interest have stable relative phases, triggering is not required for
vector averaging. Triggering is still required to isolate time records which contain the
signals of interest.
Peak Hold Averaging
Peak Hold averaging is similar to rms averaging, in that the rms values of quantities are
calculated, but instead of combinging the rms values, rather the magnitude of the new
data is compared to the magnitude of the averaged data, and if the new magnitude is
larger, then the new data becomes the averaged data. This is done on a frequency bin by
bin basis. The result is averaged data with the largest magnitudes which occurred over a
number of measurements. Peak Hold can compare a fixed number of measurements or
run continuously.
Linear Weighting
Linear weighting combines N (Number Of Averages) measurements with equal
weighting in either RMS or Vector averaging.
While Linear averaging is in progress, the number of averages completed is shown in the
Horizontal Scale Bar below the graph. When the Number Of Averages has been
completed, the measurement stops and ‘Done’ is displayed to the left of the graph.
Waterfall Storage
If Waterfall Storage is on, the waterfall buffer only stores the completed linear averages,
not each individual measurement. Each time the linear average is done, the result is
stored in the waterfall buffer and the average is reset and started over (instead of
stopping). Each completed average counts as a single waterfall record.
Exponential Weighting
Exponential weighting weights new data more than old data. For RMS and Vector
averaging, weighting takes place according to
Average
N
= (New Data • 1/N) + (Average
N-1
) • (N-1)/N
where N is the number of averages.
While Exponential averaging is in progress, the number of averages completed is shown
in the Horizontal Scale Bar below the graph. The displayed number stops incrementing at
the Number of Averages while the averaging continues.