LOS ANGELES -- In his lab at Harman International in Southern California, Director of Acoustic Research Dr. Sean Olive leads me around tables full of loud speakers and transducers, soldering equipment and wiring, past a wood shop with prototype speaker cabinets, and into a room lined with massive baffles. In fact, we're standing on a wire grid, as these wedge-shaped baffles cover the floor.
This is one of Harman's anechoic chambers, a room designed for such complete silence that I begin to get a headache from the sound of my own veins.
Olive uses this room to test Harman's speaker systems, running set tones through them, recording the output from every angle, and using a computer to analyze the frequencies. Olive and his engineering team can then analyze output graphs and get a good idea of the sound quality.
But graphs don't tell the whole story, so Harman uses a more qualitative means of testing the sounds quality of its systems: human focus groups. To train the human test subjects, mostly culled from Harman's own employees, to understand what they are hearing and be able to communicate with a common vocabulary, Olive came up with an app that asks its users to identify what frequencies in a music track have been boosted or reduced. Harman coaches its listeners in a vocabulary to describe when, for example, a peak has been applied to the 25 hertz frequency of a track.
I give the app a try at the lab, first listening to a track with no frequency manipulation. Afterward, Olive hits the EQ button on the track and I can hear a difference in the sound. But then I have to look at a graph giving me four options as to what has been changed in the sound output, a boost or reduction at 100 hertz or 5,000 hertz, equivalent to general bass and treble frequencies. Listening to the test track a few times, and being coached by Olive to listen for specific instruments, I form an opinion and actually get it right.
That sample represented just one test in the app of the 91 that Harman's listeners have to pass to become part of a focus group. I take a look at a more advanced test in the series. It splits the frequency range into seven dips and peaks, asking the test subject to choose which frequency of the seven has been boosted or reduced. Harman requires that its listeners reach level seven in the app, but the levels continue on up from there, splitting the frequencies with ever more granularity.
Harman offers the app for Windows and Mac OS on its Web site. It looks a bit rough, as it was designed for internal testing, but does an excellent job training you to identify different frequencies in sound output.
During this visit, Jonathan Pierce, Harman Senior Engineer of Acoustic Systems for automotive, tells me that the company can't do the same sort of anechoic chamber testing on car sound systems as it does for home speakers. But it does use these trained listening groups to rate the systems it builds into new cars. For each new automotive system that Harman designs, a minimum of 8 listeners, and usually 10, sit in the car and rate the audio output using a standardized set of high-quality recordings. These listeners can point out overly bright treble, identifying the approximate frequency range, or muddy mids.
Pierce team then analyzes the human ratings and determines if the car's system reaches Harman's -- and the automaker's -- standards, or if it needs further tuning.
As Harman supplies audio systems to more than 35 percent of the world's cars, there is a good chance you've heard the results of this testing. And when you look at premium audio brands, Harman's percentage likely goes way up. The company produces Mark Levinson for Lexus, Lexicon for Hyundai, Revel for Lincoln, Bowers and Wilkins for Jaguar, and Infinity and JBL for a variety of automotive brands. Most recently, Harman acquired Bang and Olufsen's automotive audio business, further extending its reach into the premium market.
Click here to download Harman's app, and train yourself to be a better listener.