This time, following on from the previous article (Sound Measurement Examples - Part 5 "FFT Analysis and Octave Band Analysis (Part 1)"), we will introduce the analysis results using two methods: octave band analysis and bundled octaves using FFT analysis.
The subjects of this analysis are two sounds: orchestral music and the sound of an excavator. Orchestral music is a fluctuating sound with a sound pressure level ranging from 66 dB to 80 dB. The sound of the excavator is a steady-state sound of approximately 74.4 dB.
The sounds of an orchestra and an excavator are represented by the following signals:
-
Figure 1: Musical sounds (top) and excavator sounds (bottom)
Comparison of analysis results using a color map (musical tones)
Figure 2 shows the results of real-time octave analysis, and Figure 3 shows the bundled octave results of FFT analysis. Real-time octave analysis was performed using 1/3 octaves, fast dynamic characteristics (time weighting characteristics) (125 ms), and frequency weighting characteristics of Z (FLAT). FFT analysis was performed with a frequency range of 18.75 kHz, 16384 sample points, a Hanning window, and frequency weighting characteristics of Z (FLAT).
Musical tones are fluctuating sounds, but as long as the horizontal axis scale is displayed over a span of this length (approximately 18 seconds), the two analysis results show almost the same trend. However, as will be shown later, the measured values themselves show an overall difference of about 4 dB.
-
Figure 2 Real-time octave analysis results (musical tone)
-
Figure 3: FFT analysis results for bundled octaves (musical tones)
Comparison of analysis results using a color map (excavator noise)
Figure 4 shows the results of real-time octave analysis, and Figure 5 shows the bundled octave results of FFT analysis.
This will be shown. Real-time octave analysis is 1/3 octave, and dynamic characteristics (time-weighted characteristics) are
The analysis was performed using a fast (125 ms) frequency weighting characteristic Z (FLAT). The FFT analysis was performed using frequency
The range is 18.75 kHz, the number of samples is 16384, the Hanning window is, and the frequency weighting characteristics are
I did it using Z (FLAT).
Since the sound of the excavator is a constant sound, the analysis results were similar. However, in reality, individual values
There are differences when you look at them.
-
Figure 4 Real-time octave analysis results (excavator sound)
-
Figure 5: FFT analysis results for bundled octave (excavator sound)
Comparison of time trends for all-pass overalls (musical sound)
The all-pass value of the real-time octave analysis results is measured with a sound level meter (noise meter).
This is the value measured using the same method as the displayed sound pressure level value. Sound pressure level from a sound level meter or
If you do not have equipment or software to measure the all-pass value, you can instead use FFT analysis over
Sometimes all values are measured. By adjusting the number of sample points in the FFT analysis, a certain value may be obtained.
You can obtain results that are similar to a certain extent.
The time difference between the all-pass value of the real-time octave analysis result and the overall value of the FFT analysis.
The trend is shown in Figure 6. Real-time octave analysis is 1/3 octave, dynamic characteristics (time
The weighting characteristics were fast (125 ms), and the frequency weighting characteristics were set to A. The FFT analysis was performed as follows:
Frequency range 18.75 kHz, sample count 16384 points, Hanning window, frequency weighting.
The characteristics were analyzed using method A. Figure 7 shows the dynamic characteristics at 10 ms and the analysis results for 2048 sample points.
The results are also shown, along with an enlarged portion of the graph.
Figure 6 shows the effects of dynamic characteristics (125 ms) and FFT frame time length (approximately 340 ms under these conditions).
There is a time lag between the two analysis results, but if we ignore this time lag, the time trend is roughly
The results match, with a maximum difference of approximately 4 dB. As shown in Figure 7, changing the dynamic characteristics or the number of sample points alters the smoothness of the time variation, resulting in significantly different outcomes.
-
Figure 6. Time Trend 1 of Allpass Overall (Musical Tone)
-
Figure 7 Time Trend 2 of Allpass Overall (Musical Tone)
Comparison of time trends for all-pass overall (excavator noise)
The time difference between the all-pass value of the real-time octave analysis result and the overall value of the FFT analysis.
The trend is shown in Figure 8. Real-time octave analysis is 1/3 octave, dynamic characteristics (time
The weighting characteristics were fast (125 ms), and the frequency weighting characteristics were set to A. The FFT analysis was performed as follows:
Frequency range 18.75 kHz, sample count 16384 points, Hanning window, frequency weighting.
The characteristics were analyzed using method A. Figure 9 shows the dynamic characteristics at 10 ms and the analysis results for 2048 sample points.
The results are also shown, along with an enlarged portion of the graph.
Figure 8 shows that the range of variation differs depending on the analysis method, dynamic characteristics, and number of sample points, but all are approximately 74 dB.
This shows that a steady-state sound can be measured.
-
Figure 8. Time trend 1 of all-pass overall (excavator noise)
-
Figure 9. Time trend 2 of all-pass overall (excavator noise)
Comparison of equivalent sound levels and the summation average of bundled octaves
Equivalent noise level (average value over 15 seconds) from real-time octave analysis results and bundle of FFT analysis results
Figures 10 and 11 show the averaging results of the octaves (average value over 15 seconds).
Octave analysis is 1/3 octave, dynamic characteristics (time-weighted characteristics) are fast (125 ms), frequency
The weighting characteristic was set to A. The FFT analysis was performed with a frequency range of 18.75 kHz and the number of samples was [number missing].
16,384 points were used, with a Hanning window and frequency weighting characteristics set to A.
The dynamic characteristics (fast (125 ms)) and the FFT frame time length (approximately 340 ms under these conditions) are both sufficiently long.
When averaging over time is performed, the differences due to the analysis method are almost eliminated.
Regarding the discrepancy, there is a difference of about 1.7 dB in the frequency band below 160 Hz, but in the frequency band above 200 Hz...
The difference in the region was less than 0.6 dB.
-
Figure 10 Equivalent noise level - averaging of bundled octaves (musical tone)
-
Figure 11 Equivalent noise level - averaging of bundled octaves (excavation noise)
summary
This time, we will discuss octave band analysis and how to analyze bundled octaves using FFT analysis.
We then presented the results of our analysis of the excavator noise and showed the differences between the two analysis results.
If a real-time octave analyzer is not available, perform FFT analysis using an FFT analyzer.
Band data may be obtained by bundling octaves, but the analysis method is different.
Generally, the results do not match, and the differences vary depending on the nature of the signal being analyzed and the analysis conditions.
When comparing data, it is generally best to analyze it using the same analytical method, but if unavoidable,
As introduced here, we compared the results of the two analysis methods to see how much difference there is.
It is necessary.
In the next installment (Sound Measurement Examples - Part 7: "FFT Analysis and Octave Band Analysis (Part 3)"), we will introduce the results of impact sound analysis.
(Excerpt from the email newsletter issued on June 20, 2013)
