In the previous article, we showed the possibility of evaluating differences in timbre by using roughness and fluctuation intensity, which are indices extracted from the temporal fluctuation components. While these temporal characteristics-based indices do not specify all aspects of timbre, they can explain some aspects of timbre, such as the perceived differences between comparisons where the differences in loudness and pitch (frequency characteristics) are small.
This time, we will introduce an example of using roughness as an evaluation metric to pass/fail judgment of a developed product, in response to auditory impressions such as "roughness" and "muddiness."
First, let's compare the sounds of a faulty (NG) and a working (OK) motor in automotive electrical components.
Automotive electrical component motor noise
You probably got the impression that the perceived difference was more about the qualitative difference in the "roughness" of the sound than the quantitative difference in the "loudness" of the sound. Now let's look at the results of the 1/3 octave analysis of these two sounds.
-
Figure 1: 1/3 Octave Analysis (Blue: NG, Red: OK)
Figure 1 shows the comparison. In the high-frequency range, particularly in the 10 kHz and 12.5 kHz bands, the defective product is slightly less than 10 dB louder, showing a difference between the two. The difference in the high-frequency range is also significant in the loudness spectrum comparison shown in Figure 2.
-
Figure 2. Loudness spectrum (blue: NG, red: OK)
To examine the impact of these high-frequency components on the overall sound impression, we will compare the original sound with a version where frequencies above 8 kHz have been cut using a graphic equalizer.
Automotive electrical component motor noise
There is a slight difference in sound quality, but the unique "roughness" remains, and this high-frequency component is the "roughness".
It is believed that this does not contribute to the goal.
Next, Figure 3 shows the results of the roughness analysis performed to extract the sense of "roughness".
For detailed information on the calculation process for roughness, etc., please see the technical report "Sound Quality" on our website.
Please refer to section 9, "Variability and Roughness," of the evaluation section.
-
Figure 3. Roughness spectrum (blue: NG, red: OK)
These results show that in each critical band from 1 kHz to 4 kHz (excluding the 2 kHz and 3.15 kHz bands), the difference in roughness between the rejected and rejected products was significant. This indicates that while it was difficult to differentiate the perceived differences in sound intensity using loudness spectra in each critical band, calculating roughness makes it possible to differentiate these differences. For sounds above 8 kHz, the difference in "loudness" was clear, but it did not contribute to "sound roughness," and there was almost no difference between the rejected and rejected products.
In summary, we have demonstrated that roughness is useful for evaluating and pass/fail judgment the quality of sounds generated from machinery with rotating mechanisms, specifically focusing on characteristics such as "roughness" and "muddiness."
Next time, I would like to explain useful indicators for the process of developing countermeasures, based on this case study.
• All data in this measurement column was obtained using our OS-2740 sound quality evaluation pack.
Ono Sokki- OS-2740 Sound Quality Evaluation Pack (Discontinued)
(Excerpt from the email newsletter issued on November 19, 2009)
