that the A-weighting curve over corrected S&V OBSERVER the low weighting, resulting in artificially low numbers. Even the linear The Difficulties in Evaluating level measurements (a 2 dB differ- ence) did not really reflect what I was A-Weighted Sound Level Measurements hearing. John M. Masciale, Data Physics Corporation, San Jose, California Further analysis showed that in look- ing at below 2 Bark the first It has been known for a long time that octave bands. Method A calculates loud- snow blower had a loudness of 6.8 , the frequency response of human hearing ness using the Stevens method. Method while the second snow blower had a is anything but linear. In 1933 Fletcher B (which is the more commonly used loudness of 2.8 Sones. The first snow and Munson generated their Equal Loud- method) performs the calculations based blower was nearly two and one half times ness Curves. These curves compare on the work of Zwicker, and is referred to as loud in the lower frequency range. sound at different and levels, as Zwicker loudness. Looking above 17 bark I found that the and are plotted for different Phon levels. The calculation of loudness is based on first Snow Blower had a loudness of 16.5 Each Phon level is defined by the deci- measurements in critical bands (denoted Sones, whereas the second snow blower bel level at 1000 Hz. Hence the 40 Phon in Bark). Figure 2 is a graph of critical had a loudness of 13.8 Sones. From these curve is a plot of the amplitude of single band number versus log frequency. The numbers, I can see that the low frequency tones at different frequencies that sound resulting amplitudes for each critical noise content was the most markedly dif- as loud as a tone at 1000 Hz and 40 dB. band are usually denoted in Sones/Bark. ferent and significant, which agrees with To improve on acoustic measurements, The area under the specific loudness what I was hearing. weighting curves were created to com- curve results in a total loudness number Conclusion. There is a place for the use pensate for sound at different levels. The in Sones. The convenience of measuring of A-weighted measurements. A-weight- A-weighting curve was created to com- in Sones is that, unlike dB, they correlate ing is a convenient standardized proce- pensate for sound along the 40 Phon con- to human hearing. Therefore, if measure- dure that may be used for comparing tour, the B weighting curve was created ment A is 20 Sones and measurement B simple measurements. It is unrealistic, to compensate for sound along the 70 is 40 Sones, measurement B will sound however, in most circumstances to expect Phon contour, and the C weighting curve twice as loud as measurement A. A-weighted measurements to correlate was created to compensate for sound Example Measurement and Compari- well with a subjective evaluation of loud- along the 100 Phon contour. son. The best way to illustrate some of the ness. Loudness is one of several psychoa- To properly take a measurement using difficulties in comparing A-weighted coustic parameters that can be used to these weighting curves, the user was sup- measurements is to compare measure- quantize the subjective evaluation of posed to look at the overall sound pres- ments of similar levels. I noise, and is readily available as a mea- sure level, and then select the appropri- recently purchased a new snow blower. surement in current acoustical instru- ate weighting curve. This is not a bad In comparing the sound of the old one mentation. approach for highly tonal noise, or flat with the new, I found that the older ma- random noise, but there can be a lot of chine seemed to be significantly louder 140 pitfalls trying to measure sound with dif- than the new. The sound of the old ma- ferent amplitudes at different frequencies chine seemed to go right through me. Yet 120 (which is the case for most noise mea- when I measured the A-weighted sound surements). pressure level of the two machines there 100 was only a 1 dB difference. 90 Phons When trying to standardize acoustic 80 noise measurements, people found that Figure 3 shows the 1/3 octave spectra 70 Phons this approach led to a lot of confusion. As of the sound pressure levels of the two 60 a result, most noise measurement stan- snow blowers. Snow blower 1 resulted in Equivalent Loudness, dB 40 dards settled on the A-weighting curve to a sound pressure level of 85 dBA. Snow 40 Phons use in taking a measurement. There might blower 2 resulted in a sound pressure 20 be good accuracy in measuring things level of 84 dBA. A 1 dB difference in 10 100 1k10k Frequency, Hz that are operating in a quiet room. But for measurement is typically not considered any sound pressure measurements at or perceivable in terms of overall loudness, Figure 1. Comparison of three equal loudness above the sound pressure level of a typi- yet my ears were telling me otherwise. contours with the A-weighted frequency re- cal conversation, this can lead to mea- Looking at the frequency content you can sponse network. surements that have little to do with per- see that snow blower 1 has considerably ceived noise levels. more low frequency energy than snow A Weighting Versus Equal Loudness. blower 2. Figure 4 shows the same mea- 24 Figure 1 shows the inverted A weighting surements with A-weighted 1/3 octave 20 contour plotted on top of the equal loud- bands. Because of the propensity of the 16 ness curves at 40 Phons, 70 Phons, and A-weighting curve to attenuate low fre- 12 100 Phons. It is plain to see that at 40 quency noise, the dBA calculation is 8 Phons there is fair correlation, but the dominated by the higher amplitude infor- 4 higher you go in loudness level the mation at the higher frequencies.

Critical Band Numbers, Bark 0 poorer the correlation, especially at lower I then decided to take a look at the spe- 10 100 1k10k frequencies. cific loudness of the two machines (Fig- Frequency, Hz In an effort to better correlate human ure 5). It is plain to see that the most sub- hearing with measured sound pressure stantial difference between the two Figure 2. Critical band number vs. frequency. levels, various techniques of measuring machines is in both the lower and higher loudness have been developed. ISO stan- frequency range. The total loudness of 71 dard 532 Acoustics – Method for Calcu- Sones versus 63 Sones indicates that the lating Loudness Level outlines two differ- first snow blower is perceptibly louder ent methods for calculating the human than the second snow blower. The reason perceived loudness of complex sound that the loudness measurement corre- that has been measured in octave or 1/3 lated better with what I was hearing is

2 OBSERVERLIB 90 Snow Blower 1 -- 88 dB, 85 dBA Snow Blower 2 -- 86 dB, 84 dBA 80

70

60

50

Sound Pressure Level, dB 40 32 125 500 2000 8000 A L 1/3 Octave Band Center Frequency, Hz

Figure 3. Sound pressure levels of two snow blowers.

Figure 4. A-weighted sound pressure levels of two snow blowers.

4.5 4.0 3.5 3.0 2.5 2.0 Snow Blower 1 -- 71 Sones 1.5 Snow Blower 2 -- 63 Sones 1.0

Magnitude, Sones/Bark 0.5 0 0 2 4 6 8 10 12 14 16 18 20 22 24 Frequency Band, Bark

Figure 5. Specific loudness of two snow blow- ers.

OBSERVERLIB 3