44
Time1
Time2 Time3
…
FFT
1
FFT
2
FFT
3
time
τ
τ
−
t
e
[Fig. V-2-10] Exponential Averaging
2. 3. 3 Peak Hold Averaging
This averaging keeps the biggest peak value of spectrum at the same frequency to repeat with
averaging number. This theory is shown in [Fig. V-2-10] and <Eq. V-2-5>.
[
]
( )
( )
i
FFT k
MAX FFT k
=
,
0
1
a
i
n
≤ ≤
−
,
0
2
N
k
≤ ≤
<Eq. V-2-5>
where
a
n
is averaging number(or iteration number for averaging).
Time1
Time2 Time3
…
FFT
1
FFT
2
FFT
3
Instant FFT
FFT
1
FFT
2
FFT
3
Averaged FFT
[Fig. V-2-10] Peak Hold Averaging
2. 4 Picket Fence Error
The resolution of frequency is defined as dividing the sampling frequency by buffer size. Therefore,
if the peak of spectrum exists in the smaller region than resolution, the peak was not shown with real
value. The energy of peak divides into both side spectrums. In this situation, we can’t get a real
frequency and magnitude of peak. This is called ‘Picket Fence Error’. This error is shown in [Fig. V-
2-11]. If you want to get the exact frequency and magnitude of spectrum, you have to compensate
both parameters with some theory.
Summary of Contents for SLA-PA201
Page 64: ...64 Appendix 1 The Theory for calculating RMS by ISO2631 and ISO8041 ...
Page 67: ...67 Appendix 2 Frequency Response of Filter in the Vibration Level Meter ...
Page 75: ...75 Appendix 3 Theory of WBCombined Filter and Frequency Response ISO6954 Filter ...
Page 79: ...79 Fig 2 Lowpass Filter with 100Hz Cutoff Fig 3 a v transition Frequency Response ...
Page 88: ...88 Fig 12 Error between theoretical value and measured value ...
