AES Section Meeting Reports

Chicago - October 12, 2017

Meeting Topic:

Speaker Name:

Meeting Location:


Dr. Olive reviewed several studies he has undertaken with his colleague Todd Welti. The principle goal of the research was to determine if the subjective preference of a given pair of headphones or earphones could be predicted based on objective measurements, especially frequency response.

An initial effort to do a blind comparison of headphones was implemented by having an assistant remove and replace headphones on the listener's head without their knowledge of which model they were listening to. This proved problematic because there were inevitable differences in the weight and fit of the various models, and thus the evaluations were not truly blind.

The team next moved to simulating various models of headphones by measuring the frequency response of each model and recreating this response by applying a filter that would adapt the response of the physical headphones to that of the intended test pair. This proved to be very effective, and comparisons of preference between a simulated headphone and the actual physical unit showed very good correlation. Using this approach, the comparison of virtual headphones could be done in a truly blind fashion as there was no change in the user's experience other than the sound.

Olive and company determined the ideal frequency response by taking measurements with a dummy head. The head was placed at the ideal listening position in front of a reference pair of loudspeakers in an acoustically tuned room.

Fine tuning of the curve was done by allowing trained listeners to (blindly) make adjustments to the level of treble and bass frequencies. The results were generally consistent, with variations of a few dB between listeners and also variations based on the program content.

Tests were also taken across age groups and across cultures. The results were remarkably consistent, despite what one might guess from anecdotal evidence. Tests done using in-ear phones proved to be consistent as well, although listeners tend to prefer more bass for in-ears than for headphones. It is unclear why this is the case, but some speculation from the audience suggested it was to compensate for the lack of physical sensation of bass in the chest.

The ultimate goal of Olive's research was to create a model that predicts a preference rating based on the measured frequency response alone. By assigning a point system based on closeness of fit to the idea frequency response, an effective model was created.

Written By:

More About Chicago Section

AES - Audio Engineering Society