Could someone please tell me what the accuracy of the FreqPeak in
the PowerFrequencyEstimate function is?
We are developing a vibration analysis system that uses a pcmica card,
developed by us, for data collection.
Problem: Calculating the FFT using the AutoPowerSpectrum() and the
PowerFrequencyEstimate() to
Calculate the FreqPeak the results are not consistent. Results range from
..01Hz to .75Hz and randomly 1 to 1.5 Hz inaccuracy. Without changing the
signal source.
The above also results in inappropriate power being displayed in the
spectrum. E.g. bin 40 varies from 60 to 95% power, and a wide spread in
spectrum when measuring 20.7 Hz (signal source)
Measurement Frequencies:
20Hz
30Hz
60Hz
160Hz
PCMCIA Card Statistics:
1) 8 MHz. Clock. Ability to set sample rate to within .03 Hz.
2) A/D Converters: 12-bit sampling (successive approximation) A/D. 72dB S/(N
+ D) and 80 dB THD @ 100kHz, +/- 0.35LSB INL.
3) Filter: 8th order elliptic, clocked variable break frequency. (This is
not being used for the testing)
Signal Source for Testing:
Digital Function Generator, Tektronix Model AFG 310.
Pseudo-Code
1) Set Sample Rate. 2.56 * FrequencyOfInspection*10. Frequency must be
displayed in 40th bin in spectrum, 10 orders.
2) Collect 1024 points at above sample rate.
3) Hand points collected to AutoPowerSpectrum().
dt=1/Fs as above
CWDSP1.AmpPhaseSpectrum FFTXRawDataHold, 0, dt, FFTYRawDataHold,
FFTZRawDataHold, df.
4) Hand Results to PowerFrequencyEstimate().
No Windowing
CWDSP1.PowerFrequencyEstimate FFTYRawDataHold, df * 40, 0, 0, df, 20,
FreqPeak, PowerPeak.
5) Set new sample rate based on FreqPeak
6) Collect 1024 points again.
7) Display graphically results. Di
splay actual calculated frequency.
Also, Jeremiah Where is Multifunction IO (DAQ) forum
Thanks
Rod
