CF-5200 Frequently Asked Questions

The data length (the time T required to capture 2048 sample data points) changes depending on the frequency range. Therefore, changing the frequency range changes the data time for FFT, which in turn results in a difference in energy.

To average and normalize the energy, we consider the power spectrum divided by T, that is, the power spectrum per unit time length of 1 s (unit energy), and the power spectral density is defined as the power spectrum per 1 Hz.
With random signals, even the same signal will produce different power spectrum values depending on the FFT resolution (corresponding to bandwidth). By normalizing the power spectrum resolution (bandwidth) to per Hz and displaying it, the effects of differences due to frequency range can be minimized.

Energy spectral density is defined as the power spectral density multiplied by the data length T.
For random signals, power spectral density is used for averaging, but for single-event phenomena such as shock waveforms, the energy is finite in duration, so the energy spectral density, which takes into account the duration of the phenomenon, is used.

The relationship between them is as follows: Let T be the data length (=time) and Δf be the frequency resolution.

• Power spectrum = spectrum obtained using 2048-point FFT
  (Displays the effective amplitude values of each frequency component)
• Power spectral density = Power spectrum ÷ Δf
  (Displays power spectrum per unit frequency)
• Energy spectral density = Power spectral density × T
  (Displays the energy spectrum per unit frequency)

T = 2048 ÷ frequency range ÷ 2.56
Δf = frequency range ÷ 800

Related instruction manual chapters

5.8.3 Power spectral density / Energy spectral density