in August 2007
The Sound of Silence
By Robert C. Chanaud, PhD
Exploring advanced functions in sound masking.
Editor’s Note: Following is a list of major manufacturers offering products with advanced functions, referred to in the text by number: Atlas Sound (Mitek Corp.) (1); Dynasound, Inc. (2); K.R. Moeller Associates, Ltd. (LogiSon Acoustic Network) (3); Lyntec (4); SoftdB (5).
Early sound masking systems consisted of small boxes powered by a low-voltage source. As commercial use expanded, centralized systems evolved, giving rise to more control opportunities. As manufacturers learned more about listener responses to masking, these opportunities developed into expanded functions for the equipment. Some functions improve listener acceptance, while others improve system performance.
Because most are time-based, the development of cost-effective DSP technology has permitted integration of these functions without significant cost increases. Surprisingly, acoustical consultants have been slow to include these functions in specifications.
The purpose of the Initial Ramp-Up time-based function is to avoid a perceived rapid increase in background sound level at the initiation of a sound masking system. Ramp-up is based on the adage that everyone is in favor of progress, but against change. People tend to have a negative view of a new, unfamiliar environment and, because masking often is a new ingredient, it is important to introduce it slowly. (The auto industry learned this lesson in the 1930s with introduction of the Chrysler Airflow.)
The initial ramp-up function is controlled by software. It initially sets the masking at background levels and then slowly raises the level over a number of days. The required inputs are the start date, the initial level relative to the final level and the number of days to the final level. A typical function is shown in Figure 1. Experience has shown that there is positive feedback from listeners when this function is used. Expectations of improved privacy are not met initially, but privacy improves gradually and noticeably over time. Several manufacturers have integrated this function into their equipment [1,2,3,4,5: Numbers refer to companies listed in the Editor’s Note].
The purpose of the Fast Ramp-Up function is to eliminate rapid changes in the sound masking level. With the loss and regain of building power, the masking system must turn off and on, violating the reasoning used for the initial ramp-up function. The ramp for this function is completed in minutes instead of days .
Programmed Level Control
The purpose of Programmed Level Control is to create a balance between the requirement for privacy and the need for a sense of community in commercial open offices. When a room is fully occupied and active, an individual wishes to maintain a high degree of privacy and reduce his or her exposure to noise. However, as activity levels decline and the population density decreases, an individual wants to be aware of the presence of surrounding people. Also, security personnel do not want privacy during nighttime hours.
An early attempt at creating this balance was to switch the system on and off at appropriate times, but often, the person in control neglected to turn the switch on or off at correct times, or not at all. The real failing, however, was the presumption that the need for privacy changed abruptly. Experience has shown that smooth and slow changes are necessary, and that the time and amount of level change must vary with each user.
This writer held the original patent on this function; the masking level varied smoothly throughout the day and permitted different maximum levels depending on the day-of-week. An example of this early application is shown in Figure 2.
Nighttime levels were down 10dB from daytime levels and would rise slowly, starting about one hour before working hours, and would reach full level about one hour after the start. The level decrease was similarly programmed. Maximum level on weekends is lower than that for weekdays because occupancy is lower. Newer designs permit more complex variations in the level.
This function lends itself to software control; several manufacturers have integrated it into their equipment [1,2,3,4,5]. Most allow for arbitrary time-level histories, but care must be taken to control the rate of level change. This function is implemented best in DSP-based systems; standard clock level controllers do not have the required versatility. When paging is integrated into the system, the paging level rides a constant amount above the changing masking to maintain good intelligibility.
Adaptive Level Control
The purpose of Adaptive Level Control is to adapt the masking level to the activity sound level in order to maintain a constant degree of privacy as opposed to a constant, or programmed, level of masking. One weakness of programmed level control is that the person setting it must make presumptions about work hours, occupancy and activity levels. Another weakness is that it does not accommodate changes from week to week.
Adaptive level control detects activity sounds in real time and automatically adjusts the sound masking level to maintain the desired degree of privacy. Most programmed level control settings are based on speech sounds; this particular function responds to all sounds. The advantage of adaptive control has been obvious for many years, but the technology to sense sound has been too costly to implement. That is no longer the case.
There has been some confusion related to automatic gain control. Gain controls raise a signal (speech) above the noise (background), while the adaptive controls raise the noise (masking) to inhibit the signal (speech). These are opposite aims. Gain controls sense the overall noise level, and adjust the output of the desired signal (speech) to override the noise (background).
For adaptive controls, the overall sound level (both speech and masking) is separated into the two components by use of percentile levels. Typical implementations use the 10th percentile level to represent the transient activity sounds and the 90th or 99th percentile levels to represent the masking. The function acts to keep the difference between them constant.
Arbitrary application of adaptive controls can lead to accidental misuse. No one expects the masking to increase in the evening when a vacuum cleaner is operated. Nor would one desire the masking to rise because of a very loud, brief, transient sound, such as a shout. Similarly, it is not desirable to have the masking disappear when there is a temporary lull in activity sound.
The solution has been to put upper and lower limits on the adaptive levels in the same way as is done for the programmed level function. An example of the adaptive function operating within limits is shown in Figure 3. The level rise and fall times within these limits are sufficiently slow to prevent notice.
One manufacturer has incorporated this function into its equipment and reports customer acceptance . The company apparently has solved the problem of adequate spatial averaging.
Because masking covers a broad area, a sufficient number of sensors is required to provide adequate averaging. An example of an adaptive control in use is shown in Figure 4. It shows the 10th percentile level (activity sounds) and the 90th percentile level (masking) as a function of time. While it is clear that the difference between these levels is not constant, it is also clear that the need for masking varies considerably during the workday.
The purpose of Rapid Equalization is to reduce the time for, and the errors in, creating the desired masking spectrum of a room. A 1/3 octave band equalizer is used in high quality systems. Most equalizers have either mechanical or software controls. The usual equalization process is to get a sample of the existing masking spectrum, and then make manual adjustments. It is a time-consuming, iterative process.
The rapid equalization method can be accomplished in several ways. The first is to sample the existing spectrum with a real time analyzer and store it to a file. Software then reads the file, compares it to the desired spectrum and reports the corrections necessary to match them. The corrections can be transferred manually or automatically [1,5]. The second method is to download the existing spectrum by connection to the computer and manually or automatically set corrections . The third method is to download the existing spectrum by wireless method and automatically set corrections . The third approach generally is the fastest method but requires more equipment. If the spectrum measurements are properly averaged spatially, equalization can be done in one step.
Figure 5 shows a screen in which the two spectra have been mean square fit and the equalizer corrections listed.
The purpose of an addressable masking system is to permit rapid, remote and detailed changes to the functions of the system. One feature of such a system is to monitor functioning. Earlier methods used a monitor panel in a centralized rack. Each channel of the monitor sampled either the amplifier outputs or the zone control outputs. With development of modern electronics, it is now possible to monitor down to speaker level. Masking systems, even those with paging, are designed very conservatively, so most equipment is reliable and sound masking applications usually are not critical.
Experience suggests that monitoring for most systems is optional, and monitoring to speaker level is not needed. However, one benefit of monitoring centralized outputs is to permit sound levels to be set electrically should some unexpected output change be made. (This can also be done with a voltmeter.) It is advantageous in large, multi-floor systems, provided the relationship between the voltage and the sound level meter is known.
The advent of versatile and comprehensive hardware and software has led to the possibility of addressing each generator, zone (group of speakers) or speaker, individually or in groups. Standard systems usually permit the masking/paging spectrum or the zone levels to be set centrally and have a limited number of audio functions (masking, masking plus paging, paging).
The addressable function extends that capability. Levels at individual speakers can be altered; the masking spectrum and level in any zone can be tailored individually; a speaker can be changed to another zone; the audio function in any zone or speaker can be altered. In standard systems, these tasks must be accomplished by rewiring. The possibility of addressing the system remotely, either through a dedicated phone line or through the internet, makes addressable systems quite attractive.
A few manufacturers have developed systems that include this function to a lesser or greater extent [1, 3]. One system can accept the signals from almost any number of masking, paging or music sources, and distribute them arbitrarily. It also has the capability of incorporating the adaptive function. The more complexity and versatility built into the system, of course, the more expensive it will be.
Standard masking signals are from either analog or digital noise sources. The analog source creates a truly random signal, while the digital source creates a repetitive, but seemingly random, signal. The digital source is simpler to implement and repeat times now are sufficiently long that the repetition is not detectable to the ear. It is exceedingly difficult for a listener to tell the difference between these two sources.
However, both create what is called “stationary noise.” When a signal (speech) is buried in this noise, sophisticated eavesdroppers can use the properties of the noise to reduce its level relative to the signal (speech) and thus improve speech recovery. If the source is made “non-stationary,” the spectrum and level are changing continually and this inhibits the recovery process.
A generator of this type is used for protection of secure facilities from deliberate listeners, particularly those with sophisticated listening devices such as laser or fiberoptic mics. There is one manufacturer of non-stationary masking generators .
Dr. Robert C. Chanaud received his BS from the US Coast Guard Academy, his MS from the University of California and his PhD from Purdue University. Active in the field of acoustics since 1958, he taught at Purdue University and the University of Colorado and, in 1975, founded Dynasound, Inc. He has developed software programs to facilitate the design and equalization of masking systems and recently completed a book about sound masking. A consultant to several manufacturers of masking systems, he currently is working with Atlas Sound.