This measurement column addresses frequently asked questions received by our customer support center and provides answers to those questions.
Motors and gears can emit unusual noises that exhibit repeated changes in volume. To investigate the characteristics of these repetitions, it is common practice to extract the envelope and perform FFT analysis to examine the magnitude of the repeating frequency components.
Frequency Analysis from the Basics (28) - "Equipment Diagnosis of Rotating Machinery"
This time, we will perform a 1/3 octave analysis and extract repeating frequency components from the fluctuations in sound pressure levels.
For simplicity, we will show an example of extracting the fluctuating component from a waveform with a known repetition frequency using 1/3 octave analysis.
Original waveform
A signal that fluctuates from 2000 Hz to 10 Hz.
A signal that fluctuates from 5000 Hz to 30 Hz.
Random signal
We add the three signals together to create a composite waveform.
From the composite waveform (purple waveform) in the diagram below, it's unclear whether or not there are repetitions.
We will perform an FFT analysis on the composite waveform and examine its power spectrum.
Peaks are visible at 2000 Hz and 5000 Hz. We've determined that the frequency components at 2000 Hz and 5000 Hz are larger compared to other frequencies.
However, it's not possible to see how much the size of this peak is fluctuating.
The synthesized waveform will be analyzed by 1/3 octave. For this analysis, the time constant (time weight) will be set to NONE (none) and the time interval to 1 ms.
Be careful when multiplying by the time constant, as it smooths out the fluctuations and causes them to disappear.
You can see a repeating pattern of 2000 Hz and 5000 Hz color variations with changing sizes. Check this while adjusting the color scale (95-60 in the diagram).
A repeating color change can be observed at 2000 Hz and 5000 Hz.
Let's look at the time trends (time changes) for 2000 Hz and 5000 Hz.
This is the trend over a 1-second period.
You can see that it varies within the range of 80 dB to 95 dB.
Due to this repetition frequency, we will perform an FFT analysis on the trend waveforms at 2000 Hz and 5000 Hz.
Save the trend data in CSV format. Right-click on the screen.
Import the saved CSV file using the data file import function, specifying it as an ASCII file.
Since we are calculating 1/3 octave intervals at 1 ms intervals, the sampling frequency is 1000 Hz.
Importing will add more files to your data manager.
Track 5 shows a trend waveform at 2000 Hz, and Track 6 shows a trend waveform at 5000 Hz.
We will perform a fundamental frequency analysis.
Since the data is collected over a 1-second period and the sample frequency is low at 1000 Hz, be careful not to shorten the frame length, otherwise you will end up with insufficient data.
Peaks appeared at 10 Hz and 30 Hz.
We were able to confirm that there are components that repeatedly fluctuate in the 2000 Hz and 5000 Hz trends.
If you hear a sound that seems to be repeating or changing, you can try performing a 1/3 octave analysis, exploring the changes using a color pattern, and then performing an FFT analysis on the trend waveform to reveal the frequency. Please use this as a reference.
(Excerpt from the email newsletter issued on May 20, 2020)
