Stanford Research Systems SR785, Operating Manual And Programming Reference

Page: 223 / 546

Display Setup Menu 4-27
SR785 Dynamic Signal Analyzer
signal and noise amplitudes. The power spectrum is a real quantity and contains no phase
information.
The precise definition of the Power Spectrum 1 measurement for all averaging modes is
as follows:
No Average
Power Spectrum1 = FFT1*• FFT1
Vector Average
Power Spectrum1 = <FFT1>*• <FFT1>
RMS Average
Power Spectrum1 = <FFT1*• FFT1>
Peak Hold Average
Power Spectrum1 = MAX(FFT1) • MAX(FFT1)
Time Record
A time record is simply a sequence of data samples. The length of the time record is the
FFT resolution/span.
For baseband spans (spans which start at DC), the time record resembles a digital
oscilloscope display. Signals at frequencies above the span have been filtered out.
Baseband time records are entirely real, they have no imaginary part.
For zoomed spans (spans which start above DC), the time record does NOT resemble the
original data. The data has been frequency shifted. Signals at the center of the span
appear at DC while frequencies at both edges of the span appear as high frequencies.
Zoomed time records are complex, they have both a real and an imaginary part. The
sampling rate is always half of the equivalent baseband span.
Remember, the time record is not a continuous representation of the input signal. The
data is sampled and has a time resolution of 1/(sample rate). High frequency signals will
appear distorted in the time record. However, ALL of the spectral information is
preserved by the Nyquist sampling theorem as long as the value of each sample is
accurate. A triggered time record will always jitter by 1 sample. This jitter is removed in
the computation of the phase of the spectrum relative to the trigger.
Averaging does not affect the time record. Averaging is performed on the FFT spectrum
and not on the time data.
Amplitude calibration is also performed in the frequency domain. Hence, the time data
amplitudes are not calibrated.
Windowed Time Record
The FFT operates on windowed time records. The window function is applied to the time
record immediately before the FFT. Most window functions taper off to zero at the start
and end of the time record. If a transient signal occurs at the start of the time record, the
corresponding windowed time record and FFT may not show anything because the
window function reduces the transient to zero.