Given the complex nature of sound, it's not surprising that considerable work has been done to develop an effective system of single-number descriptors. With such a system, "quality sound" targets can be established for different building environments. These targets aid designers in specifying appropriate acoustical requirements that can be substantiated through measurement. For example, a designer can specify that "the background sound level in the Acme theater shall be X," where X is a single-number descriptor conveying the desired quality of sound.

The most frequently used single-number descriptors are the A-weighting network, noise criteria (NC) and room criteria (RC). All three share a common problem: they unavoidably lose valuable information about the character or "quality" of sound. Each of these descriptors is based on octave band data which, as noted earlier, already masks tones. The process of converting from eight octave bands to a single number overlooks even more sound data. Despite this shortcoming, the single-number descriptors summarized below are valuable tools for defining sound and are widely used to specify acoustical requirements.

"A" Weighting

One simple method for combining octave band readings into a single-number descriptor is A-B-C weighting. Represented by the curves shown in Figure 2, these weighting networks compensate for the ear's varying sensitivity at different frequencies. "C" weighting is applied to high-volume (loud) sound levels where the ear's response is relatively flat, while "B" weighting is applied to medium-volume sound levels. "A" weighting, which is used for low-volume (quiet) sound pressures, best approximates human hearing levels in the comfort range where no protection is needed.

The following steps describe how to calculate an A-weighted (dBA) descriptor.

1. Subtract these decibel values from the octave band cited: 26 dB from 63 Hz, 16 dB from 125 Hz, 9 dB from 250 Hz, and 3 dB from 500 Hz.
2. Add 1 dB each to the 2000-Hz and 4000-Hz octave bands.
3. Logarithmically add all eight octave bands together to arrive at an overall A-weighted sound level (dBA).

Data about the relative magnitude of each octave band is lost with the completion of Step 3. So, even though the target dBA level is achieved, an objectionable tonal quality or spectrum imbalance may exist.

Most sound level meters automatically calculate and display A-weighted sound values, providing a simple and objective means of verifying acoustical performance.

"A" weighting is often used to define sound in outdoor environments. For example, local sound ordinances typically regulate dBA levels at property lines. Hearing-related safety standards written by such bodies as the Occupational Safety and Health Organization (OSHA) also commonly refer to A-weighted sound readings.

Note: As a rule, "A" weighting is applied to octave-band sound pressure data and combined into a single number ... but an exception exists. ARI Standard 270 recommends the use of A-weighted sound power. To avoid confusion with A-weighted sound pressure values, A-weighted sound power is expressed as bels rather than decibels. Ideally, both "A" weighting of sound pressure while displaying all eight octave bands and any A-weighting of sound power (except in accordance with ARI Standard 270) should be avoided.

Noise Criteria

"Noise criteria" or NC curves are probably the most common single-number descriptor used to define the sound quality of indoor environments. Like the equal loudness contours (Figure 1) on which they're based, the loudness along each NC chart curve is about the same. Each NC curve also slopes downward to reflect the ear's increasing sensitivity at higher frequencies.

Determining the NC value for a given set of octave band data is easy. Simply plot the octave band data on the NC chart ... the highest NC curve crossed by the data curve determines the NC rating. Of course, this strategy still doesn't account for the tonal nature and relative magnitude of each octave band even though it avoids logarithmic addition.

Why is this "lost" information so critical? The answer is best explained with an example. Figure 3 shows octave band data measured in an open-plan office area and plotted on an NC curve. Notice that the resulting value, NC 39, is acceptable for this environment. Also observe that the NC level is set by the 63-Hz octave band, and that the sound in the upper bands quickly drops off.

In this particular example, sound produced by the air handling unit travels through the ductwork and radiates into the office area through the duct wall. To achieve the desired NC level, two layers of sheet rock were added to the duct exterior to sufficiently block the low-frequency sound.

Unfortunately, because high-frequency sounds are much more easily attenuated than low ones, the upper octave bands are now overattenuated. Although an objective analysis deems the resulting NC 39 sound level acceptable, most listeners probably wouldn't as the unbalanced spectrum produces an annoying rumble. Interestingly enough, quality sound could be achieved in this example by adding sound to the space. Placing speakers in the room (or above the ceiling tile) to introduce sound in the upper bands would balance the sound spectrum. The subjective analysis of the office occupants would then agree with the objective acoustical data.

Room Criteria

Sound spectrums can be unbalanced in other ways that result in poor acoustical quality. While a lot of low-frequency sound results in a rumble, too much high-frequency sound produces a hiss. Room criteria (RC) curves provide a means of identifying these imbalances. Calculating an RC value from a set of octave band data isn't quite as easy as determining an NC value. Yet, it's still a simple process that yields a single-number descriptor followed by one or more letters indicating sound character:

• N identifies a "neutral" or balanced spectrum.
• R indicates "rumbly."
• H represents "hissy."
• RV denotes "perceptible vibration."

To aid system designers, ASHRAE recommends target RC ratings for various types of spaces ( Table 1) and encourages use of the RC noise-rating procedure "whenever the quality of the space dictates the need for a neutral, unobtrusive background sound."

If we plot the acoustical data for our example open-plan office on an RC chart (Figure 4), we find that it results in a rating of RC 31 (R). This time, our objective and subjective analyses lead to the same conclusion: though "quiet" enough, the background sound in the space is rumbly. Similarly, a sound spectrum curve falling into the RC "neutral" category would be judged as excellent by most observers. It's this conformity of analysis results that makes the RC noise rating method a better tool than its predecessors for specifying acoustical requirements.

Despite this advantage, the RC rating system is less widely used than other single-number descriptors. Perhaps system designers are unfamiliar with its benefits or are comfortable with the more easily calculated NC rating. They may also question the usefulness of the RC rating system's letter descriptors which identify the nature of a sound quality problem, but don't convey its magnitude.

Continue on to Specifying Quality Sound.