Sean Olive

Sean Olive

@seanolive.bsky.social

i am a former Senior Fellow at Harman doing research in perception and measurement of sound quality including loudspeakers and headphones. Interests include music, hockey, wine, cycling, politics, science and technology. A Canadian who is also American.

314 followers112 following991 posts

Sean Olive
@seanolive.bsky.social

A birthday present I received last evening. The first open-back gaming headphone by HyperX. All the benefits of an open-back headphone but this one has bass

Sean Olive
@seanolive.bsky.social

We just launched the website today for our new startup venture in spatial audio called Hear3D. 6 of 7 the Founders are former Harman audio research scientists and engineers with decades of experience in scientific research and audio product development. You can learn more @ www.hear3D.com

Sean Olive
@seanolive.bsky.social

The @JBL Endurance Run has always been a decent wired in-ear headphone for $19. I just ordered 3 pairs Endurance Run 3 and measured one sample and compared to Run 2. Both measured on the GRAS45-CA10 with the IE target curve plotted for comparison.

Sean Olive
@seanolive.bsky.social

Five reseats of a TWS in a GRAS-45-CA10 measured with an AP x517. I don't think I've seen such poor repeatability in a TWS in some time. Something isn't quite right.

Sean Olive
@seanolive.bsky.social

There was an audio reviewer in the 1980s from TAS named Enid Lumley who advocated evaluating audio equipment naked as she believed clothing caused electrostatic forces and absorption that colored the sound. We can be thankful that this was before YouTube was invented.

View full thread
Sean Olive
@seanolive.bsky.social

1-5% is not a huge differences in preference rating within the margin of error in measuring preference. These peaks and dips are very high Q and we know are difficult to detect with music. This example is from Toole and Olive Resonance Detection study (1988).

Sean Olive
@seanolive.bsky.social

Of course, whether these measured deviations from a personal target matter based on in-ear measurements remains an open question. It needs further research based on controlled studies. So I urge caution in coming to conclusions until we have such evidence.

View full thread
Sean Olive
@seanolive.bsky.social

If you apply the SD and slope to the predictive model (Olive et al, 2018) you can predict the preference scores from an ideal target. Strictly speaking this model was based on the Harman Target not a personal target but for the sake of illustration let's go with it.

Sean Olive
@seanolive.bsky.social

These are the predicted scores based on deviations from their personal targets. You can see that for Resolve and Golden Ear the deviations from their personal target would only reduce the preference score less than 1% and 5% with the biggest drop from Listener (13% drop)

Sean Olive
@seanolive.bsky.social

These curves have been normalized to their personal DF HRTFS. A perfect match would be a flat line for each reviewer but presumably a 1 dB/octave downward tilt for the DF HRTFS is a good approximation of what would sound good. Therefore, a regression line shown is a good approximation.

Sean Olive
@seanolive.bsky.social

I calculated deviations from the target from 50 Hz to 10kHz and plotted the error curves and the absolute slope for a regression line that best fits the error. Shown here

Sean Olive
@seanolive.bsky.social

This headphone had the widest distribution of preference ratings among the 32 headphones tested. It had significantly less bass and a big peak at 2 kHz and only pleases a small segment who like little bass and apparently coloration. It’s a polarizing headphone in terms of sound quality

Sean Olive
@seanolive.bsky.social

In the last 2 days I have re-analyzed 6 papers we published from 2013-2025. This shows the distribution of preferred headphone curves in the 3 preference segments discussed in the 2019 paper based on the deviations from the Harman AE Target Curve (2018) -- the so-called error curve.

Sean Olive
@seanolive.bsky.social

This is the same plot with the three curves normalized to mid-band level. I think it more clearly shows the delta in bass preference among the people who like more and less bass.

Sean Olive
@seanolive.bsky.social

In the last 2 days I have re-analyzed 6 papers we published from 2013-2025. This shows the distribution of preferred headphone curves in the 3 preference segments discussed in the 2019 paper based on the deviations from the Harman AE Target Curve (2018) -- the so-called error curve.

Sean Olive
@seanolive.bsky.social

In the last 2 days I have re-analyzed 6 papers we published from 2013-2025. This shows the distribution of preferred headphone curves in the 3 preference segments discussed in the 2019 paper based on the deviations from the Harman AE Target Curve (2018) -- the so-called error curve.

Sean Olive
@seanolive.bsky.social

What facilitated this analysis was the use of AI --and I would never attempted doing this without it. For inquiring minds AI is more of a curse than savior in terms of giving you more free time.

Sean Olive
@seanolive.bsky.social

That’s not a modern Ford Lincoln tuned by Beats but a JBL premium sound system 40 years ago. Bass has suffered from inflation like everything and that may be a good thing

Sean Olive
@seanolive.bsky.social

I revisiting some data from this 2015 study to analysis and plot the mean preferred bass and treble settings for 256 listeners. Olive, Sean; Welti, Todd; 2015; Factors That Influence Listeners’ Preferred Bass and Treble Levels in Headphones Paper 9382; Available from: aes.org/publications...

Sean Olive
@seanolive.bsky.social

The first graph shows the mean bass and treble settings for individual listeners. The 2nd graph are the quartiles and 5% and 95% percentiles. The 3rd graph are the statistics applied to the 2018 AE target. You could consider these preference bounds with curves falling within 25 -75 Q satisfying 50%

Sean Olive
@seanolive.bsky.social

The graph is the average of left+ right channels based on the average of 5 reseats. I don't know what the consensus is for these but to me -- they sound very muffled in vocals with muddy bass.

Sean Olive
@seanolive.bsky.social

Been looking for a cheap open back headphone for psychoacoustic experiments that is consistent across samples and can be easily equalized to a target. Harder to find. Here are 6 samples of a $150 Sennheiser HD560 S that I think meet that criteria. Measured on a GRAS 45CA-10 with an AP517x

Sean Olive
@seanolive.bsky.social

This shows the variation in FR of a sample with 5 reseats on the fixture (graph 1), the mean left and right response (graph 2) and the standard deviation of left and right as a function of frequency (Graph 3).

Sean Olive
@seanolive.bsky.social

I just received the Final Tonalite TWS that has technology to personalize the timbre to match your DF HRTF and taste. Several AES papers have been presented on the research behind these including one at 160th Copenhagen

Sean Olive
@seanolive.bsky.social

Some positive feedback about the new Dan Clark Aeon Core headphone the tuning influenced by some research we did together. It's funny how science can positively influence products and some reviewers -- but not others. headphones.com/blogs/announ...

headphones.comDan Clark Audio announces ÆON CORE—an $899 closed-back planar magneticDCA's new planar magnetic headphone launches at $899, tuned to an updated reference curve and built around DCA's most efficient driver to date.