Subjectivity in Audio

Headfi member, sforza, published a concise treatise (Google Docs) against the lickety-split labelling of headphones as 'V-shaped'. These are the headphones he refers to: 

Sennheiser HD600
Koss KSC75
Mr. Speakers Alpha Dog
Beats studio

Each is accompanied by frequency response graphs taken by several equally well-known and much-respected objective headphone sites. sforza writes:

Below are two graphs of averaged human voice frequency ranges. Singing would fall under “normal, raised and loud”. Meaning females would have a peak of 1.6khz, and males would have a peak at just 500hz on most tracks, reaching 1.2khz in some tracks with loud vocals. This means that as long as a headphone is neutral until 1.6khz for females, or 1.2khz for males, we can safely say that the headphone is neutral for vocals and not “V-shaped”. Treble peaks are something very different from V-shaped vocals, as shown in the graphs of the alpha dog and KSC75.

As far as I can make out, sforza believes that singing voice, which mimics the frequency curve of a shout, is the optimal band for evaluating vocal sound pressure, and therefore, the best metric against which to determine the linearity of a frequency response. Analysing vocal music on a shout index is wrong. I have next to no range to speak of, but my singing voice hits higher frequencies than my screaming voice. 

Check it:


That topped out at about 800Hz. And now I've got a sore throat. 

[Three glasses of wine later]


That topped out at 1,2kHz, or an effective range gain of 50%. (It might also have been the wine.)


Of course, I could be wrong. And so could sforza. Subjective reviewers have blind spots. Objective reviewers have blind spots.

My objection to this article is simple: equating the range of a screaming/shouting voice to a singing voice could be inaccurate. Further, there is more to measuring vocals than just a microphone, a good audio interface, and software.

That said, sforza is correct: sans qualification, labelling the sound of a headphone as v-shaped, or u-shaped is a bit unfair. But then again, so is the following statement: gooey American style cookies taste like shit.

That is called an opinion. And, in my case, it is right. Gooey American style cookies are pointless pieces of fat. And if you or I interpret that a different part of the sound spectrum contains the important vocal bits, we probably won't agree on what is or is not v-shaped. No scientific test, standardised or other, will align our opinions. Your cookies are nasty pieces of fat that require a serviette. Mine are crunchy, and require a plate.

There is one other problem: objective, standardised headphone measurement systems do not exist. Differences in recording equipment, in product positioning, in room acoustics, not to mention post-production, and much much more, force incongruence on even the most pious of measurements. 

Of course, subjective listening has no standards. It's your word against my sensitivities. And the sensitivity of our ears changes throughout the day, and is subject to many stimuli. Coffee. Wine. The train. The idiotic 自民党 and 民主党 politicians shouting through over-amped microphones at Nagareyama corners from 6:30 AM. The idiotic construction raging till 12:00 AM. On and on it goes.

There is no perfect review. There is no perfect reviewer, objective or otherwise. And barring 100% repeatable and standardised objective testing, I see no problem with subjective ejaculations of v-shaped, u-shaped, and so on, especially in light of personal preferences.

Imagine a world where reviewers discussed headphone sound signatures based on a standardised measurement system. Reviews would be shorter. Dime-a-dozen audio blogs would be fewer. Less time would be wasted online. Bring it on, I say.

Humans aren't wired for objectivity. We group. We clump. We rave. 

No matter the output, one person's full bass is another person's anaemic thrum. And your cookies suck.

Subjectivity in Audio, and the importance of Accuracy in Frequency Descriptions