• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

"Bias" of some members towards headphone measurements?

The bass and treble are most likely not emphasized at all in the Harman transfer function preference target.
This is how I perceive H2018 tuned headphones, but not H2019 tuned IEMs. I'm not sure why. This also happens to be a common anecdote on the internet.

I'm often reminded of something of a divergence trend that can be seen among different types of playback systems. Ones that are lower-SPL(and lower-attentiveness?) applications like car audio systems and home stereos vs. and others like studio and stage monitors. In my estimation this difference is pretty close to an equal-loudness differential EQ. Could it help explain pervasive playback EQ divergences to some extent?
1737585265941.png


Two very different targets in Sean Olive listening test were statistically tied. https://aes2.org/publications/elibrary-page/?id=22696
This might reflect the range HRTF variation like you rightfully bring up. OTOH is my SPL-driven speculation any less valid?
 
Last edited:
This is how I perceive H2018 tuned headphones, but not H2019 tuned IEMs. I'm not sure why. This also happens to be a common anecdote on the internet.

I'm often reminded of something of a divergence trend that can be seen among different types of playback systems. Ones that are lower-SPL(and lower-attentiveness?) applications like car audio systems and home stereos vs. and others like studio and stage monitors. In my estimation this difference is pretty close to an equal-loudness differential EQ. Could it help explain pervasive playback EQ divergences to some extent?

Two very different targets in Sean Olive listening test were statistically tied. https://aes2.org/publications/elibrary-page/?id=22696
This might reflect the range HRTF variation like you bring up. OTOH is my SPL-driven speculation any less valid?
I think in a large enough group (slice by demographics and training too, why not, we're not paying for it) you can probably estimate the actual underlying variance in HRTF by finding the maximum deviation in EQ that still results in a statistical tie. The confidence window would correspond to how much people's hearing varies. You'd have to compare it to the confidence window for speakers to account for randomness in preference, i.e. preference would be expected to vary by the same amount for both.
 
This is how I perceive H2018 tuned headphones, but not H2019 tuned IEMs. I'm not sure why. This also happens to be a common anecdote on the internet.

I'm often reminded of something of a divergence trend that can be seen among different types of playback systems. Ones that are lower-SPL(and lower-attentiveness?) applications like car audio systems and home stereos vs. and others like studio and stage monitors. In my estimation this difference is pretty close to an equal-loudness differential EQ. Could it help explain pervasive playback EQ divergences to some extent?
View attachment 423094

Two very different targets in Sean Olive listening test were statistically tied. https://aes2.org/publications/elibrary-page/?id=22696
This might reflect the range HRTF variation like you rightfully bring up. OTOH is my SPL-driven speculation any less valid?
IEM's will have a different target function than circumaural type cans.The application is different. One fires through your pinna at 90 degrees inducing your pinna's acoustic filter at a 90 degree azimuth and the other bypasses that filter altogether. You cannot ignore these filtering effects. IEM's and circumaural headphones are two different animals. Although it is possible to get a target for IEM's , they will be generalized above 2K and it will only be possible to make it completely accurate for one defined set of Pinna's. With circumaural it's already using your ear (to some degree) so it can be closer to accurate to a reference set of loudspeakers in a room. But I do not believe we are there yet. I still believe that our individual HRTF's are only relevant for spatial cues. The spectral balance can be corrected for if we apply the science. If we include the cross-feed function it opens up pandoras box. The Smyth realizer can simulate everything you need, but it's only for one person and you need a home reference system to begin with. We can build a set of loudspeakers that a majority of people agree that sound "good" using science. I am certain we can do this with headphones at some point, we are just not doing it yet as an industry. You also have the effect of people having no reference of what an "accurate" system sounds like. That ends up as a crap shoot. You like vanilla and I like chocolate. That may be some peoples opinion but it does not resolve the question of what is closer to the source. Sound reproduction is a science.
 
I'm often reminded of something of a divergence trend that can be seen among different types of playback systems. Ones that are lower-SPL(and lower-attentiveness?) applications like car audio systems and home stereos vs. and others like studio and stage monitors. In my estimation this difference is pretty close to an equal-loudness differential EQ. Could it help explain pervasive playback EQ divergences to some extent?
One possible explanation is natural hearing loss. As the class two group (28%) were a different demographic and wanted less bass and more treble. This is typical with people who have hearing loss.
 
One possible explanation is natural hearing loss. As the class two group (28%) were a different demographic and wanted less bass and more treble. This is typical with people who have hearing loss.
"If you like transducers that target the Harman curve, your hearing is fine. If you don't like those transducers, you probably have hearing loss." This is a very common "bias" here.

In reality, Harman's own research, where they gave listeners tone controls and had them listen to the same tracks, reported that listeners who were pre-determined to have "normal hearing" demonstrated distinct differences in treble and bass preferences based on their gender, age, and country of residence.
 
Last edited:
"If you like transducers that target the Harman curve, your hearing is fine. If you don't like those transducers, you probably have hearing loss." This is a very common "bias" here.

In reality, Harman's own research, where they gave listeners tone controls and had them listen to the same tracks, reported that listeners who were pre-determined to have "normal hearing" demonstrated distinct differences in treble and bass preferences based on their gender, age, and country or residence.
I agree.
It largely comes down to musical taste and listening experience, much more than the common notion that deviations in preference curves are solely due to hearing loss. Personal preferences are strongly influenced by the person's exposure to music and training in critical listening, which shape their perception of what sounds "natural" or "balanced."

Additionally, hearing loss can influence preferences in different ways. Some people may compensate by increasing the SPL at the affected frequencies, while others might prefer those frequencies attenuated, as they can sometimes become overly sensitive or fatiguing in that range.
 
The quicker people get over the fact that FR measurements of speakers/headphones aren't as predictive of SQ as they wish they were, the better the discussions will be. Transducers are not solid state devices like DACs and amplifiers. Amplifiers, for instance, output out a 2-dimensional voltage vs time. Simple. Measureable. Transducers, OTOH, output a 3-dimensional polar radiation of sound pressure vs time, further influenced by room/pinna interference in the time and amplitude domains. And we somehow think the way we measure DACs and amplifiers translates well to transducers? Asinine.
 
The quicker people get over the fact ...

What follows in your post is your opinion, not a fact. There is a difference.

... FR measurements of speakers/headphones aren't as predictive of SQ as they wish they were,

You can't see into other people's minds. You don't know the nature or degree of what they wish for.

... the better the discussions will be

The discussions seem to be just fine.
And we somehow think the way we measure DACs and amplifiers translates well to transducers?

That depends on your meanings. If you are criticizing the particular methods of measuring electronic circuits vs. transducers, then yes, you are correct. If you mean that measurements of transducers using instruments is inferior to some other method, such as uncontrolled listening, then no.


This language is unnecessary.
 
The quicker people get over the fact that FR measurements of speakers/headphones aren't as predictive of SQ as they wish they were, the better the discussions will be. Transducers are not solid state devices like DACs and amplifiers. Amplifiers, for instance, output out a 2-dimensional voltage vs time. Simple. Measureable. Transducers, OTOH, output a 3-dimensional polar radiation of sound pressure vs time, further influenced by room/pinna interference in the time and amplitude domains. And we somehow think the way we measure DACs and amplifiers translates well to transducers? Asinine.
One would imagine that this is why FR is not the only thing measured when reviewing loudspeakers - dispersion patterns, waterfall, resonances, and distortion are all measured as well.

That being said, it's almost impossible to get a clean, nearly-flat frequency response plot from a speaker that measures poorly in these other respects. How often have we seen a speaker with close to flat frequency response but terrible resonances, or awful distortion (excepting smaller speakers made to be played at lower volumes that may have low distortion at 86dB and rapidly rising distortion over 90dB)?

Building a speaker that will produce flat, neutral frequency response isn't a matter of simply picking flat drivers and sticking them in a box - there's a lot of engineering going on to achieve those results, and pretty reliably, a well-engineered speaker that produces flat FR also produces lower distortion, fewer resonances, and more even dispersion than speakers with coloured frequency response.
 
we somehow think the way we measure DACs and amplifiers translates well to transducers? Asinine.
Who thinks that? The Klippel NFS measures the output of a speaker in 3D. The graphs used to report this data are cosmetically similar to those used for electronics but the measurements are pretty different.
 
Another forum member's opinion.
View attachment 423356
Spin data "predict highly rated loudspeakers," when analyzed via complex regression formulas on a computer. How often do you see folks applying the regression formula to spin data to make comparisons?

And even then, the predictive value isn't perfect, and primary validation of the Harman-funded research was done using single musical genre (rock). Even Toole, himself, admits that the Circle of Confusion introduces variation in loudspeaker preferences.
 
Last edited:
Who thinks that? The Klippel NFS measures the output of a speaker in 3D. The graphs used to report this data are cosmetically similar to those used for electronics but the measurements are pretty different.
I'm specifically referring to the practice of looking at a single 2-D FR measurement of a transducer (headphones, in particular) and attempting to make strong predictions about SQ. I'm sure you've undoubtedly seen headphones "measured" as a single 2-d freq vs amplitude measurement with a superimposed Harman curve meant for "comparison."
 
Last edited:
Another forum member's opinion.
View attachment 423356
The only issue I have with these points from a very knowledgeable individual is the way I interpret their claim -that EQ isn’t useful unless your speakers have a particularly well-behaved on-axis response.

In my experience, the response of any speaker can be improved by EQ. While a perfect on-axis response is desirable, EQ can make meaningful improvements to any speaker system.

While I partially agree that offering active speakers with various curves tailored to taste might be a marketing promise, this holds true only if you take those preset curves at face value. Naturally, you can’t rely on those curves as absolute since the response at the listening position in my room will differ from that in someone else’s. The purpose of these options is to allow users to experiment with preset EQ settings to suit their preferences. It’s a beneficial feature and not as malicious or misleading as suggested in point no. 10.

Think of it as the modern graphical EQ. The manufacturer could have easily included a graphical equalizer as back in the day, but these preset curves are simply a modern equivalent.
While they don’t give you as much detailed control, they make it easier for people who aren’t familiar with EQ settings to customize the sound to their room and tase.

1737694226845.png
 
In my experience, the response of any speaker can be improved by EQ. While a perfect on-axis response is desirable, EQ can make meaningful improvements to any speaker system.
Sure.... but when doing so with 'predicted well performing' and 'predicted poor performing' loudspeakers would the result be equally 'well sounding' ?
 
The quicker people get over the fact that FR measurements of speakers/headphones aren't as predictive of SQ as they wish they were, the better the discussions will be. Transducers are not solid state devices like DACs and amplifiers. Amplifiers, for instance, output out a 2-dimensional voltage vs time. Simple. Measureable. Transducers, OTOH, output a 3-dimensional polar radiation of sound pressure vs time, further influenced by room/pinna interference in the time and amplitude domains. And we somehow think the way we measure DACs and amplifiers translates well to transducers? Asinine.

In what sense is voltage 2-dimensional?
 
Sure.... but when doing so with 'predicted well performing' and 'predicted poor performing' loudspeakers would the result be equally 'well sounding' ?
Maybe, maybe not. The point is, there’s no reason to discourage people from trying it and simply telling them to "get better speakers." EQ could very well make even a less-than-stellar speaker sound quite good. I noticed that Toole concludes his point no. 10 by essentially suggesting EQing to taste, which, in my opinion, contradicts some of his earlier points. That said, I I think I might be misinterpreting it?
 
The only issue I have with these points from a very knowledgeable individual is the way I interpret their claim -that EQ isn’t useful unless your speakers have a particularly well-behaved on-axis response.

In my experience, the response of any speaker can be improved by EQ. While a perfect on-axis response is desirable, EQ can make meaningful improvements to any speaker system.

While I partially agree that offering active speakers with various curves tailored to taste might be a marketing promise, this holds true only if you take those preset curves at face value. Naturally, you can’t rely on those curves as absolute since the response at the listening position in my room will differ from that in someone else’s. The purpose of these options is to allow users to experiment with preset EQ settings to suit their preferences. It’s a beneficial feature and not as malicious or misleading as suggested in point no. 10.

Think of it as the modern graphical EQ. The manufacturer could have easily included a graphical equalizer as back in the day, but these preset curves are simply a modern equivalent.
While they don’t give you as much detailed control, they make it easier for people who aren’t familiar with EQ settings to customize the sound to their room and tase.

Point #10, the one that Toole highlighted, is what you are disagreeing with. And while it may seem possible to EQ every speaker to better performance, that's just not the case, as explained in point 5. You really can't EQ a resonance. Perhaps you can EQ one aspect of it's radiation pattern, like an on-axis peak at a given frequency, but you will then introduce other on- and off-axis artifacts. Even if you can get one slice of the field EQ'ed with on-axis performance that is OK in one position in the room, you are left with other a host of other problems, sometimes worse than the on-axis performance you EQ'ed.

Resonances are hard to deal with in any system, not just speakers. They uniformly defy equalization, require good design up front. I tested an old Yamaha speaker with a rigidly mounted Styrofoam woofer that operates as a resonant membrane, and is a good example of a speaker that cannot have it's issues EQ'ed.

I recommend thinking deeply about point # 5 in Toole's post, and then about EQ.
 
the practice of looking at a single 2-D FR measurement of a transducer (headphones, in particular) and attempting to make strong predictions about SQ
With headphones 2D is actually pretty predictive (maybe not 100.0%) since there isn't a 3D sound field reverberating around the room and averaging out to something different than the on-axis measurement.

The high frequencies are pretty hard to measure accurately but other than that... FR is definitely still the most important measure I'd use to consider headphone sound quality if I could only have one, would you not agree?
 
Back
Top Bottom