L-R: David K. Bialik, Dr. Sean E. Olive
Approximately 60 attendees gathered at the Harman Flagship Store on Madison Avenue in New York City to hear Dr. Sean E. Olive, Director of Acoustic Research for Harman, and current President of the AES speak
about Harman's intensive research on headphone design. Headphone use has really taken off in recent years
largely due to the increased use of tablets and smart phones. Users of these devices seem to be looking for
an improved listening experience, and Harman discovered several problems relating to headphone design:
(a) lack of meaningful standards for measurements and quality evaluation;
(b) inconclusive results from poorly controlled blind listening tests;
(c) lack of comparisons of preferred frequency response of headphones vs loudspeakers
(d) how do listening preferences vary by age and cultural/national background?
The result of examining these parameters were five AES papers, based on listening tests covering six
popular headphones which ranged in price from $115 to $995. A series of double-blind tests vetted four
headphone models at a time. Two models were strongly preferred. Comparisons were made between the perceived spectral balance graphs drawn by trained listeners (JBL employees) vs actual measurements of the headphone made on a GRAS ear simulator. The preferred headphones tended to have the flattest perceived spectral balance and the smoothest measured frequency response curves.
To further study the preferred headphone target response, double-blind comparisons were made using two different headphones equalized to different target responses: three different diffuse-field (DF) calibrations, a free-field (FF) calibration, and a new target response based on the measured in-room response of an accurate loudspeaker in a reference listening room. The unequalized headphone was also included as a hidden reference. Listeners strongly preferred the headphone target response based on the in-room response of the loudspeaker to the DF, FF and unequalized choices. This makes sense since most stereo recordings are monitored through loudspeakers in rooms whose acoustic properties are neither DF nor FF, but rather somewhere in between. Moreover, most listening rooms reinforce frequencies below 200 Hz that are not accounted for in both DF and FF headphone calibrations
Do listeners agree on a preferred spectral balance for both headphones and loudspeakers?
Subjects were first presented with a flat signal and then provided with continuous-turn bass and treble controls
(Griffen PowerMate) to set their preferred frequency curves. The panel consisted of 8 trained and 3 untrained
listeners. All had their hearing acuity measured prior to the listening sessions. A number of programs were played and repeated, with half the group testing loudspeakers first, then headphones, then the reverse.
Eventually a virtual headphone listening test method was devised to simulate the spectral response of headphones, and remove potential errors or biases from listener preference ratings related to headphone brand/model, visual appearance, comfort/fit and leakage.
Another series of tests established that college students, carefully studied, preferred CD quality vs MP3 quality by up to 86% of the test trials. The next tests measured four consumer loudspeakers listened to by 58 U.S. college and high school students, 149 Japanese college students and 12 Harman trained listeners. A randomized double-blind protocol was provided by placing the speakers behind a curtain and having automated equipment silent move the speaker under test to the front center position. All of the different listening groups preferred the most accurate loudspeaker (as measured in an anechoic chamber) that yielded the widest, flattest and smoothest spectral response of all the speakers.
A comparison was also made between around the ear and in-ear headphone using both Harman trained listeners and 18 untrained college students, and again the highest quality modes were preferred.
A final comparison to determine any cultural differences in preferred headphone sound quality was made using a special iPad application that allows controlled blind comparisons of virtualized headphones. To date at least 219 trained and untrained listeners from Canada, the U.S., Japan and Germany prefer the spectral balance of the Harman Target Curve, and ranked the Beats Studio headphones (which exhibit a broad band bass boost) in last place.
Several thoughtful questions were discussed after the main talk was concluded.
