L to R: Dr. Sean Olive; Blair Francey (Chair AES Toronto) — Toronto AES Section -- October 22 2013
Blair Francey introduced Dr. Sean Olive.
Sean began by explaining the reason for conducting headphone testing, primarily being a lack of meaningful standards on headphone sound quality or target function.
His presentation, conducted with the aid of slides, consisted of in depth summaries of four previously published AES papers, including the background, methodology, and conclusions. A fifth and final paper will be presented in China in two weeks.
The first paper concerned tests of a series of controlled double-blind listening tests on popular headphones using trained listeners. The goal was to better understand the relationship between their perceived sound quality and acoustic performance. Conclusions showed good agreement among listeners regarding preferences. Preferences favored "neutral, accurate sound" ie: flat frequency response.
The second paper tested measured listener preferences of different headphones equalized to different target curves responses including the recommended diffuse and free-field target curves. The problem with both diffuse and free-field targets is that the room is neither diffuse nor free-field but somewhere in between! There is a combination of direct sound and reflected sound. The other problem is that most rooms have bass reinforcement from standing waves and boundary effects, neither of which is accounted for in diffuse and free-field calibrations. Headphones calibrated to these targets will sound too bright and too thin in the bass. Realizing stereo recordings are made to sound good played through speakers in a room, that to make those sound best through headphones, one has to simulate the in-room response of a well-designed speaker system calibrated in a reference listening room. Ultimately, conclusions showed that this target response was strongly preferred to both diffuse and free-field targets; and participants stated these targets were the most neutral in terms of spectral balance.
In the third paper, the tests used an adjustment method whereby listeners directly adjusted bass and treble levels of a headphone and loudspeaker equalized to the same in-room target response to determine their preferred targets. This way, the variation in individual listeners' taste in headphone spectral balance, and the extent to which the preferred headphone target response simulates the response of an accurate loudspeaker in a reference room could be measured and determined. Results showed listeners' preferences towards a boost in the bass.
The fourth paper concerned virtual headphone technology developed by Harman which allowed different brands of headphones to be simulated through one a single replicator headphone. This removes any biases test subjects may have towards specific brands. The only downside is it doesn't simulate leakage effects, and only includes linear distortions. The results showed that when influences of brand, fashion and celebrity endorsement are removed, both trained and untrained listeners generally agree on which headphones sound best. Contrary to conventional perception, listeners prefer good quality sound (ie: CD) over lossy formats (ie: mp3) with a generally strong correlation between trained and untrained listeners. There was no evidence listeners preferred boomy sound.
Responding to an audience question, cultural backgrounds of listeners do not greatly affect the consistency of the results although that is a target for future testing. Other targets include in ear headphones, and headphone quality preferences for younger people.
Another question concerned whether background noise could influence preferences. Sean replied masking could have an affect. For example, when listening in a car there may be a preference for more bass. All tests for these papers were done in quiet environments.
He discussed the current tests conducted previously in the day at Harris Institute for the Recording Arts which involved testing different virtualized headphones. The test also included the option for participants to adjust tonal preferences. Sean was very complimentary of the Harris facility and proposed returning in June 2014.
Some responses to questions:
Sean believes there's a problem measuring fidelity in general because the recordings are the unknown variable; there are no standards that define where they're made, what the speakers are calibrated to, and what the in-room target is. "We're only guessing at what fidelity means." This is why preferences are used in testing. He noted seeing variations in adjusting bass and treble preferences from different music programs.
When asked what the preferred listening method would for determining flaws in recordings, Sean replied: "In the absence of knowing what the room is doing and what the speakers are, headphones are a good alternative. At least it's a more controlled environment."
Regarding in-ear testing and the differing individual ear canal resonances affecting results, Sean said there hasn't been enough testing yet to determine what the in-ear target should be. "We're born with our ears and we learn to adapt to sounds that we hear." Many argue we should just leave that element alone and try to deliver good sound to the ear. Any individual characteristics will be ignored because one has adapted to that.
Money doesn't necessarily correlate with good sound.
Comfort was considered as part of the data collection.
The iPhone 5 earbuds have "improved."
The presentation went quickly and smoothly. There was no break, so socializing occurred after the meeting's end.
Blair thanked Sean for his presentation, and the audience for attending, inviting them to stay for refreshments and socializing. Dr. Olive was presented with a Toronto AES certificate of appreciation and coffee mug.
