Harman Curve Poll (Public)

Yup, that's true. It's not the Harman target per se as we're discussing here. But the general framework they use seems very similar between loudspeakers and headphones. "The less accurate and neutral loudspeakers" (as per the paper) mean "loudspeakers lower on the preference scale". Both quoted from the manuscript. And that loudspeaker preference scale was derived in 2004 study (mainly favoring in-room flat "plus" preference-based boosting) -- also note 2004/2012 both tested with headphones mimicking loudspeaker reproductions (they call this method 'virtual testing').

Actually, the whole framework doesn't look any different in their headphones research. Just way more sources of variability added in.

The points I try to deliver is neither the result, the quality, nor the usefulness of his research. He tends to present his result in misleading (I'd call that out as 'not ethical') ways.

 

I've participated in these "what you like best" target curve tests and my instinct always tells me - screw what I like, give me what's correct!

P.S. What's the best way to actually hear the pure unadulterated Harman curve?

 

How do we know the records in question weren’t recorded and mixed on cocaine? To be truly scientific, wouldn’t they need to offer lines to a trained control group? Coke was a variable that Harman neglected to account for. Playing them for a sober audience without a shitty subwoofer is not how they were intended to be played back. The same with clipped and heavily distorted (more imd than anything from the past) modern pop produced for phone speakers and sound bars. It’s meant for systems that make NS10s and JBL 305s sound high-fidelity.

 

Tube eq and a wooly woofer. Ignite amps Pultec is free. http://www.igniteamps.com/#pteq-x That way it will at least be soft and won’t turn a kick drum into a car crash.

Fuse and Variety of Sound (Quit the plugin game) have some cool free easy to use Baxandall shelves:
https://fuseaudiolabs.com/#/pages/product?id=300965965
https://varietyofsound.wordpress.com/2011/09/26/baxtereq-–-released-today/

 

An emphasis around 12 khz, would give far more natural clarity, than a boost at 3 - 4khz.

Such a boost in the upper-mids, is asking for fatigue and peakiness.

Atleast the HD650 pulls the upper-mid emphasis off with great refinement, and even then the 650 could get a bit peaky there at times without mods.

 

3-4khz boosts basically want to turn guitar into slayer solos, where streams of random notes jump out of the speakers. They basically want everything to be an amusement park ride without any dynamics. A roller coaster that is always going down hill. If there is no contrast, there is no excitement. You can’t detect that if there is no contrast and external stimuli. Einstein was right.
They just want everything to be fake and harsh. Like modern pop bullshit. Kick drums in front of the singer with the snare 10 feet behind. Everything is leveled to be the same volume. Nothing has any enveloping resonances. Nothing stands out. Nothing sounds good. Everything sounds bad.

The general public thinks Ariana Grande records sound good and Marvel movies are well shot.

 

Wasn't Olive's effort the first of its kind? I'd love to see more groups work independently on neutral average frequency responses. We might inch closer to something universal. At the end of the day though, it's a highly subjective thing. If you have a headphone that people say isn't neutral, but reference recordings sound just right with, then it's neutral for you.

 

Right now partly agree and partly disagree add up to... 64%... Maybe that number is somewhat true

 

I think you're talking about slide 10 of this pack (also slides 18, 19):

http://www.juloaudio.sk/Umiestnenie_reprosustav/History of Harman Target Curve.pdf

For earlier work on listener preferences vs loudspeakers in an anechoic room, this one:

https://pearl-hifi.com/06_Lit_Archive/15_Mfrs_Publications/Harman_Int'l/AES-Other_Publications/Trained_vs_Untrained_Listeners.pdf

Result: Preference of both trained & untrained listeners was for anechoic flat (top graph of fig. 9, p. 823).

 

Posted this on @spwath 's profile comment, but would probably be worth putting here too. It looks like the conversation has already gone down this route:

- Does this 64% include those with tinnitus, hearing loss, etc? Or on the flip side maybe trained musicians, golden ears, etc?
- Another topic that I find slightly curious is how the Harmon curve slightly resembles the average "equal-loudness" curve (ignoring degradation with age).

Additionally, I would say SBAF and similar online communities are also far from the average listener and asking here produces a VERY biased sample. Not knocking on anyone for any of the below - just pointing out that most things we take for granted here on SBAF are probably not "normal" things for the population at large:

- Most of my IRL friends who have looked at and tried my gear usually can't tell the difference between 320kpbs and various lossless formats and as far as I know, they don't have explicit hearing loss or damage.
- Most of them don't know why something sounds good to them or not ("it's clear and I can hear the details and I think the bass is good" or some other similarly generic answer). This includes some of them who are "trained" musicians and can hear things in music that I have a very hard time with - mostly psychoacoustic stuff (I've also received some musical training but I'm not that good of a musician w/ regards to anything psychoacoustic).
- Most can't tell two noticibly different DACs apart - I tried this with my SDAC and Yggdrasil once (whether you subscribe to SBAF, ASR-fi, HF, etc. philosophy - I think most in all camps would agree that these are very different DACs.) Some can't tell the sound of tube amps apart from SS unless the tubes are blatantly coloring the sound, etc. (at the level that many people on SBAF would call broken)
- Most of the ones w/o technical backgrounds would have a hard time glancing at a FR curve, much less a suite of measurements such as CSD, THD/THD+N, distortion, etc etc. plots and immediately knowing what is going on - i.e. they'd have to look at it for a while, think about it, try to make sense of it, etc.
On and on and on. Should also clarify that I do not think the above isn't something that can't be developed in a lot of people, just that they have never gone down the road so to say.

 

To me, I’m not interested in my headphones sounding like a pair of speakers in a room. I’d prefer it to sound like the live, original sound.

It would be interesting to compare these EQs using a live feed: Original sound source in one room which you can hear directly live. In that same room is a microphone (perhaps stereo pair) feeding their signal into headphones in an adjacent room with selectable EQ. Which headphone-EQ combo sounds more like the original live source?

@purr1n: I have 1/2” B&K mic & preamp we can use for such a test. Could be fun to setup at a future mini-meet

 

The big wrinkle here is that the headphone preference is wrt the preferred corrected response of the Harman listening room. That corrected response not flat in that room (see slide 5 of the Harman pack), and I can't find a freely-accessible source that correlates the response of that room vs an anechoic chamber. So we don't know how the reference used for the headphone correction correlates to flat itself.

Here's a ppt on how they reached the preferred correction for the room:

http://www.petoindominique.fr/pdf/The Subjective and Objective Evaluation of Room Correction Products.pdf

See slide 25: perceived flat (or at least audience preferred) = 10 dB decline *of the listening room* (not an anechoic chamber) from 20-20kHz.

Edit: I suppose it could be assumed from the earlier result (preference for loudspeakers that measured anechoic flat) that the correction to the Harman room would more or less correlate to anechoic flat - but yes, your comment about the studies being some years apart and preferences having changed might be a factor...

 

One more rhetorical question:
Who were the people polled, and in which country or countries?

I suspect country is important for 2 reasons:
1) Population pyramid distribution. For example, the US population mostly resembles an inverted pyramid. Meaning if those polled were truly representative of the US population, we are looking at A LOT of hearing loss by default already simply due to age.
2) Language/s spoken and music people listen to in that country - phonetic vs tonal, what sounds are most prominent in the language, frequency density of the language and the music in that country, what instruments are most present, etc. If we polled a bunch of young people listening to electronic, pop, and hip-hop music in the US the preference would be very different vs say parts of Mainland China (probably the 2nd or 1st biggest audio market) where they listen to a lot of ballards, etc.

 

https://www.aes.org/e-lib/browse.cfm?elib=17500

Edit: But see also https://www.jstage.jst.go.jp/article/ast/38/5/38_E1714/_pdf

Japanese love bass, Dutch hate treble!

 

For #1, the target (as well as general measurement) does not look at anything beyond 10khz. White age may still matter, I believe we can roughly take out hearing losses from the equation.

#2 could be academically interesting, but as already pointed out by many (and you), we are far away from the ±0.9 standard deviation (roughly 64%). Music preference and listening habits seem no longer similar with the general public regardless of the living country -- e.g., classical/jazz are already super outlier..

 

It's been a few years since I dug into the methodology, but I do believe they screened the participants for self-reported hearing loss/medical issues.