The Science and Marketing of Sound Quality
By Sean Olive, Director of Acoustic Research, Corporate Engineering
A recent Consumer Electronics Association (CEA) survey found that, for 90% of consumers, sound quality is the most important factor in their audio experiences, and 39% were willing to pay more for high-quality audio products. This is good news for HARMAN, which sells premium audio products with best-in-class sound. The bad news is that many consumers may not be aware of this, and the technical specifications of our products don’t reflect how good they truly sound. While HARMAN has already developed much of the science for measuring the perceived sound quality of audio, we continue to use industry specifications that keep consumers in the dark.
Audio performance specifications haven’t changed much since Thomas Edison. Today’s loudspeaker and headphone spec sheets only guarantee the products make sound, with no guarantee they produce quality sound. As Dr. Floyd Toole aptly puts it, “There is more useful information on the sidewall of a car tire about its performance than what you find on a loudspeaker specification sheet.” Many loudspeaker and headphone specifications give only a vague idea of the device’s real capabilities.
While consumers want better audio specifications that identify the products that deliver the truth, some detractors argue that consumers can’t handle the truth. The excuses range from claims that sound quality is “too complex for consumers to understand,” to “it’s too expensive for manufacturers and reviewers to measure,” to “it’s too subjective and personalized according to demographics, culture, prior audio experience, or the type of music you like.” Another reason why many audio companies don’t want an accurate sound quality specification is that would easily differentiate the good-sounding products from the duds. A sound quality specification would uncover that audio emperor(s) is/are wearing no clothes. Better sound through B.S. (“before science”) would no longer be a viable option.
Most of these arguments don’t hold up well when tested against current scientific knowledge about the perception and measurement of audio sound quality.
THE SCIENCE OF MEASURING LOUDSPEAKER SOUND QUALITY
HARMAN scientists have been at the forefront of research in loudspeakers and sound reproduction. The work began in the 1980s at the National Research Council of Canada, where Dr. Floyd Toole’s pioneering work provided the first correlations between subjective and objective loudspeaker measurements. In the early 1990s, this work continued at Northridge, where we developed improved listening test facilities and methods that provided better control of nuisance variables, including loudspeaker position. Two state-of-the art calibrated anechoic chambers were also added so that engineers could make accurate measurements that characterize how loudspeakers sound.
Over the next 10 years, trained listeners evaluated hundreds of HARMAN and competitors’ loudspeakers as part of the routine competitive benchmarking process. The tests were conducted double-blind so that the listeners had no knowledge of the brand, price or visual appearance of the products being tested. A consistent finding was that trained listeners generally agreed on which loudspeakers they preferred.
However, there remained a question of whether the sound quality preferences of trained listeners were similar to those of untrained listeners who represent the targeted customer for the product.
In 2004, an elaborate study involving over 300 untrained subjects compared the performances and loudspeaker preferences of. trained versus untrained listeners. The loudspeaker preferences were essentially the same for both listening groups, except the trained listeners tended to give lower, more discriminating, and consistent ratings.
More recently, Los Angeles— area high school and college students were tested to see if their sound quality tastes had become corrupted through prolonged listening to low-quality MP3 devices, as reported in the media. On average, the students preferred CD-quality music files to the lower-quality MP3 versions in 70% of the trials. None of the students preferred the lowerquality MP3 audio. The conclusion: kids can hear just fine and prefer quality sound.
Figure 1. The mean loudspeaker preference ratings and 95%
confidence intervals are plotted for the high school and
college students from LA, Japan and the HARMAN listeners.
Click for larger image.
The same students participated in a double-blind loudspeaker test in which comparative preference ratings were given to four different loudspeakers: a “well-balanced, neutral model”, (Infinity Primus 362), one with “hyped bass and treble” (Polk RTi 10), another with “forward midrange” (Klipsch RF35), and an expensive model with “colored, unbalanced sound” (Martin Logan Vista). The test was repeated using 149 visiting college students to see if their tastes in sound quality were different from their American counterparts. Figure 1 plots the average preference ratings given to each loudspeaker by the different listening groups, including 12 trained HARMAN listeners. It is reassuring to see that the most preferred loudspeaker is the Infinity model (the most technically accurate and least expensive model), and that there is good agreement in sound quality preference among the students from the United States, Japan, and the trained HARMAN listeners. As observed in previous studies, the better trained listeners tended to give lower, more consistent ratings distributed over a wider range of the preference scale.
MODELING AND PREDICTING LISTENER LOUDSPEAKER PREFERENCES BASED ON OBJECTIVE MEASUREMENTS
So far, we’ve only presented perceptual evidence that listeners prefer more accurate, neutral loudspeakers based on their comments and sound ratings However, there remained a question of whether the sound quality preferences of trained listeners were similar to those of untrained listeners measured in controlled listening tests. In this section, we will show that the perceptual evidence presented so far can be confirmed and actually predicted based on a set of technical loudspeaker measurements.
Figure 2. The family of anechoic frequency response
measurements of the loudspeakersis shown to illustrate the
relationship between preference and technical performance.
The preferred loudspeakers had flatter, smoother, and
extended curves perceived as accurate and neutral sound
quality. Click for larger image.
Figure 2 shows the mean preference ratings and 95% confidence intervals for the four loudspeakers in Figure 1 along with the technical measurements made over 70 different angles in the large anechoic chamber in Northridge. From top to bottom, the spatiallyaveraged frequency response curves represent the on-axis sound, the listening window, the first reflections, the total radiated sound power, and the directivity indices based on the first reflections and the sound power. There are clear correlations between physical features of the loudspeaker curves, and what listeners prefer. The more preferred loudspeakers tend to have flat, smooth, and extended on-axis and listening window curves combined with smooth and well-behaved early reflection, sound power, and directivity index curves. In other words, the more linear and accurate the loudspeakers measure, the more listeners will like them. Listeners prefer to hear the truth.
The final piece of the sound quality puzzle is a patented mathematical model that analyzes the technical loudspeaker measurements and predicts how a trained listener would rate its overall sound quality on a scale from 0 to 10 The loudspeaker preference rating prediction is 86% accurate based on 70 different loudspeaker models. Even at 86% accuracy, the model tells us more about the perceived sound quality of a loudspeaker than current loudspeaker specifications, which only tell us that the loudspeaker makes sound.
WHAT ABOUT HEADPHONE SOUND QUALITY?
HARMAN has initiated scientific studies to study the relationship between the perception and measurement of headphone sound quality. So far, we are finding similar trends in listener sound quality preference as were observed in the past 30 years of loudspeaker research. When the influences of brand, price, fashion, and celebrity endorsement are removed from the tests, listeners prefer the headphones that produce the most neutral, well balanced sound. The commercially successful Beats by Dre Studio model was rated by both trained listeners and 71 untrained Japanese college students near the bottom of the scale due to its “boomy, muffled, and colored sound.” Listeners apparently want to “hear the truth,” whether the speakers are in the room or strapped onto the sides of their heads.
More investigations are underway using both trained and untrained listeners from different demographics to define a headphone target response that satisfies these consumers.
MARKETING SOUND QUALITY AS PRODUCT FEATURE
Market research indicates audio consumers want better sound quality experience, but they lack the right information and tools to find them. For audio companies like HARMAN that sell premium audio solutions, there is an opportunity to show consumers the path towards achieving better sound. HARMAN could take a leadership role in this mission since it has developed much of the science related to the perception and measurement of sound quality.
ARTICLE REFERENCE LINKS
1 CEA Market Research Report “Notions of Sound Quality: Consumer Expectations,” July 2011.
2 Floyd Toole, Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers and Rooms, Focal Press (2008).
3 F.E. Toole and S.E. Olive, “Hearing is Believing vs. Believing is Hearing: Blind vs. Sighted Listening Tests and Other Interesting Things”, 97th Convention, Audio Eng. Soc., preprint no. 3894 (1994 Nov.).
4 S.E. Olive, “Differences in Performance and Preference of Trained versus Untrained Listeners in Loudspeaker Tests: A Case Study”, J. Audio Eng. Soc., Vol. 51, no. 9, pp. 806-825, (2003 Sept.).
5 S. E. Olive, “Some New Evidence that Teenagers and College Students May Prefer Accurate Sound Reproduction”, 132nd Convention, Audio Eng. Soc.,
preprint no. 8683, (2012, April).
6 S.E. Olive, “A Multiple Regression Model for Predicting Loudspeaker
Preference Using Objective Measurements: Part II – Development of the Model”,
117th Convention, Audio Eng. Soc., preprint no. 6190, (2004 October).
Tagged in acoustic, audio, quality, sound