The Nature of headphone Measurements

[markanini]

Could you elaborate??? Erroneous ?

I'd say erroneous because they assume it doesn't beg the question. But consistency says it should beg the question. And if you try to find out more you learn that that measuremt systems for headphones and IEM represent something approximate and higly variable. This should negate a lot of scritiny that's sometime naively expressed by audio hobbyists about granular FR deviations.

 

[Jaxx1138]

I'd say erroneous because they assume it doesn't beg the question. But consistency says it should beg the question. And if you try to find out more you learn that that measuremt systems for headphones and IEM represent something approximate and higly variable. This should negate a lot of scritiny that's sometime naively expressed by audio hobbyists about granular FR deviations.

Well, I did work for Sean Olive for 9 years and I have been an acoustic engineer for 25 years. I have conducted many measurements over those years of IEM.s , microphones, loudspeakers, headphones, multi-media devices and worked for several audio companies. I do have some background in the field of acoustic measurements. Having said that, I do believe that our current system of metrics for headphones is greatly flawed. We have no definitive metric for an objectively flat headphone response. I do we believe we can determine one.

 

[Jaxx1138]

There are standards.
The test fixtures are built to the specifications outlined in that standard and have to comply to them within the given tolerances.
What comes out of them is dependent on several conditions as mentioned above.

HATS are used for measuring acoustics for example. For speakers you just need to measure SPL (in certain conditions and axis) and really don't want to modify the sound (HATS) and (try) to undo that later on to get intuitive plots.

Speakers also don't interact with HATS so there is no reason to use them where with headphones the interaction is clear but also is headphone and positioning dependent.
Also with measuring microphones you can measure pretty reliable and well above 8kHz which you can't with HATS.

The current target curves are derived from speakers in (standard) rooms by measuring speakers in a certain relative position and then trying to match the found response with those of headphones mounted on them.
Harman curve is the most popular one but many people are now proponents of tilted DF (diffuse field with a downward slope).

I believe the target curves derived by Sean Olives research , were developed utilizing preference EQ filters from random trained listeners. Not from loudspeakers in rooms. Although I believe this data is necessary to determine a more accurate response curve.

 

[ZolaIII]

@Jaxx1138 heheh it's +12 dB SPL to equal loudness not compensated in room speaker response in bass LV from Harman over the ear as that's 76 dB SPL (as much as I remember from paper). ELC scales invented when you cross 88 dB programme. Now we are talking about huge boost (hard encoded+5.5 dB) on high SPL where it should be minus). Sometimes good intentions go rogue really, really bad. Don't get me wrong my intentions aren't bad.

 

[Jaxx1138]

@Jaxx1138 heheh it's +12 dB SPL to equal loudness not compensated in room speaker response in bass LV from Harman over the ear as that's 76 dB SPL (as much as I remember from paper). ELC scales invented when you cross 88 dB programme. Now we are talking about huge boost (hard encoded+5.5 dB) on high SPL where it should be minus). Sometimes good intentions go rogue really, really bad. Don't get me wrong my intentions aren't bad.

The Fletcher /Munson curves have nothing to do with the objective metrics of a headphone, Those are related to human perception.

 

[ZolaIII]

The Fletcher /Munson curves have nothing to do with the objective metrics of a headphone, Those are related to human perception.

And based on that perception participants actually adjusted those simplified controls in tone control manner doing the listening test experiment on fixed mid SPL from which averaged you got "Harman curve". Is that correct? It's all based on perception, we need to make it more accurate. Believe me that's only secure source to start building from. You are a hard core engineer... You can't separate them (hardware/psy) but you can make former more accurate and explain why is that so including stand out cases (even to the margine of error 0.1% scale). If you can't value quantities measurements they aren't very useful to us (humans).

 

[10PSPMemSticksForMiku2007]

In my experience it's true that the frequency response of different ears coupled with IEMs can vary wildly. I've also been using the 7hz Zeros lately. To me they don't sound at all like what the measurements posted in OP would indicate. They were ridiculously bright, painfully tinny, and absolutely horrible without EQ. I tried watching some shows with them on my phone, but the video player only has a graphical EQ and the upper treble spike was still quite noticeable.
I bought this product since so many people on this site, who are very educated and know what good sound is, were so positive in its thread. I'm not doubting their experience or measurements at all, but I hope you can believe me. Since I was young, family doctors have told me that I have unusually shaped ears, and my audiologist told me the same as an adult. Maybe there is some odd resonance inside my ear. As far as I know, my experience with headphones hasn't been so different to everyone else's.
I made a PEQ with the aid of sine sweep videos on youtube. They're very helpful for finding irregularities. The IEMs sound normal to me when I use this.

 

[Jaxx1138]

And based on that perception participants actually adjusted those simplified controls in tone control manner doing the listening test experiment on fixed mid SPL from which averaged you got "Harman curve". Is that correct? It's all based on perception, we need to make it more accurate. Believe me that's only secure source to start building from. You are a hard core engineer... You can't separate them (hardware/psy) but you can make former more accurate and explain why is that so including stand out cases (even to the margine of error 0.1% scale). If you can't value quantities measurements they aren't very useful to us (humans).

No. That's not how acoustic transfer functions work. You are attempting to simulate a timbre that closely resembles a pair of reference loudspeakers in a room. That transfer function can be characterized through a series of measurements. The Schroeder frequency most likely accounts for the bass increase that you often see in preference tests. Personal perception is not related to the performance of the loudspeaker or the headphone otherwise known as the DUT (Device Under Test).

 

[Jaxx1138]

In my experience it's true that the frequency response of different ears coupled with IEMs can vary wildly. I've also been using the 7hz Zeros lately. To me they don't sound at all like what the measurements posted in OP would indicate. They were ridiculously bright, painfully tinny, and absolutely horrible without EQ. I tried watching some shows with them on my phone, but the video player only has a graphical EQ and the upper treble spike was still quite noticeable.
I bought this product since so many people on this site, who are very educated and know what good sound is, were so positive in its thread. I'm not doubting their experience or measurements at all, but I hope you can believe me. Since I was young, family doctors have told me that I have unusually shaped ears, and my audiologist told me the same as an adult. Maybe there is some odd resonance inside my ear. As far as I know, my experience with headphones hasn't been so different to everyone else's.
I made a PEQ with the aid of sine sweep videos on youtube. They're very helpful for finding irregularities. The IEMs sound normal to me when I use this.

Looks normal to me. This is not uncommon.

 

[ZolaIII]

No. That's not how acoustic transfer functions work. You are attempting to simulate a timbre that closely resembles a pair of reference loudspeakers in a room. That transfer function can be characterized through a series of measurements. The Schroeder frequency most likely accounts for the bass increase that you often see in preference tests. Personal perception is not related to the performance of the loudspeaker or the headphone otherwise known as the DUT (Device Under Test).

Yes but you qualify actual measurements based upon it. If you want better measurements you need Q preference to get more accurate as one of the things and simple to actually do, other one is ear model classification which won't be easy. Now I don't understand. Isn't that what you wish to accomplish?

 

[Jaxx1138]

And based on that perception participants actually adjusted those simplified controls in tone control manner doing the listening test experiment on fixed mid SPL from which averaged you got "Harman curve". Is that correct? It's all based on perception, we need to make it more accurate. Believe me that's only secure source to start building from. You are a hard core engineer... You can't separate them (hardware/psy) but you can make former more accurate and explain why is that so including stand out cases (even to the margine of error 0.1% scale). If you can't value quantities measurements they aren't very useful to us (humans).

That is an assumption on your part not a statistically definitive determination. You have no way to determine why the participants chose their preferences. You are simply inferring that they chose it based on a phenomena that may or not be related in any way. Psychoacoustics is a perceptual based function. It only exist within the perception of a biological function or a subjective experience. Whereas acoustics can exist in a purely objective physical function separate and independent of any biological entity. All objective measurements are intended to be useful in order that any device can be designed to better serve its function in the most efficient and effective manner. That is the purpose of quantifying physical phenomena.

 

[Jaxx1138]

Yes but you qualify actual measurements based upon it. If you want better measurements you need Q preference to get more accurate as one of the things and simple to actually do, other one is ear model classification which won't be easy. Now I don't understand. Isn't that what you wish to accomplish?

The idea of modeling human anatomy is simply a series of assumptions that have been utilized early in acoustic studies in order to quantify how humans hear. It has little to do with quantifying a headphone's acoustic performance. The Klipple NFS makes none of these assumptions and does not adhere to this notion in order to quantify a loudspeaker's performance.

 

[ZolaIII]

@Jaxx1138 not assumption, hypothesis. It's not one of the 100 it's 101 fo 101. Fletcher Munson been one of the hundred and you being hundred and first. If you don't see causal link which is that obvious I can't help you there.
Same goes regarding modeling, ear chenel will be tough and will require consultations with medical branch eventually to be used as data source but it's on going effort. Like I said I don't believe any statistical significant sample will be done anytime soon. Best such effort all together represent equal loudness compensation studies summed together. Klipple NFS? Weren't we speaking about earphones/headphones measurements? Seams you run off doing studies from methodology lectures (which is actually quite usual). I couldn't as I had three two semester one's.

 

[Jaxx1138]

@Jaxx1138 not assumption, hypothesis. It's not one of the 100 it's 101 fo 101. Fletcher Munson been one of the hundred and you being hundred and first. If you don't see causal link which is that obvious I can't help you there.
Same goes regarding modeling, ear chenel will be tough and will require consultations with medical branch eventually to be used as data source but it's on going effort. Like I said I don't believe any statistical significant sample will be done anytime soon. Best such effort all together represent equal loudness compensation studies summed together. Klipple NFS? Weren't we speaking about earphones/headphones measurements? Seams you run off doing studies from methodology lectures (which is actually quite usual). I couldn't as I had three two semester one's.

In order for something to be correlated to something else you would have to show at least through some impirical evidence that the two are connected. You have not shown this but rather simply made the claim that they are connected. You have not shown any connection to what Fletcher /Munson loudness contour has to do with how a person's preference for any specific timbre in a headphones target response. How a human hears is not relevant to how a headphone functions, but only that one is designed to reproduce music for a human. A piano makes sound and when someone plays that piano it PRODUCES music. When a loudspeaker reproduces a music signal it REPRODUCES music. What the piano player is doing with the piano has nothing to do with what the loudspeaker is doing other than the fact that eventually both make sound vibrations that are harmonious (in most cases). You would never design a loudspeaker to vibrate like a piano, as that would not achieve its design goal. Sound reproduction is science. Music creation is subjective. The headphone is the transceiver and the human hearing mechanism is the receiver. We need to measure the headphone and not the acoustical filters of the human hearing mechanism. You are conflating the two which seems to be the pattern for an entire industry and the enthusiast who follow it.

 

[LevityProject]

That makes little sense: it's either meaningful data or not. Price has nothing to do with it, apart from subjective expectations towards performance perhaps. There's a reason sites like this review products of all price ranges, as they should.

If I am buying an expensive pair of headphones, I want the measurement data to be as precise as possible across the entire spectrum so I know what I'm getting into.

For lower priced stuff, some anomalies in the data isn't as important as long as I know the nature of the anomalies. Like I said before, the measurements of the zeros show a problem that actually does exist, it's up to the listener to figure out where precisely that problem lies when they put the headphones in their ears.

The measurements show something real happening ... But due to the nature of headphone measurements, the problem isn't necessarily exactly where the measurement say it is, but it's also not far off either. So it's useful information.

 

[markanini]

The measurements show something real happening ... But due to the nature of headphone measurements, the problem isn't necessarily exactly where the measurement say it is, but it's also not far off either. So it's useful information.

This sort of sounds like wishful thinking without further context. You don't specify which kind of deviations you deem more problematic vs. less problematic and why. The why is the important part because some people online do this despite not being guided by any proven protocol and assume the legitimacy of what they are doing just by other hobbyists doing the same thing.

 

[LevityProject]

This sort of sounds like wishful thinking without further context. You don't specify which kind of deviations you deem more problematic vs. less problematic and why. The why is the important part because some people online do this despite not being guided by any proven protocol and assume the legitimacy of what they are doing just by other hobbyists doing the same thing.

Above 8khz = More problematic. I thought that was implied by the limitations of headphone measurements.

 

[Jaxx1138]

This sort of sounds like wishful thinking without further context. You don't specify which kind of deviations you deem more problematic vs. less problematic and why. The why is the important part because some people online do this despite not being guided by any proven protocol and assume the legitimacy of what they are doing just by other hobbyists doing the same thing.

This is true. This has been an ongoing problem with headphone measurements. I believe new protocols which are more repeatable and less confusing are in order. We should be able to look at the metrics for headphones and at the very least be able to discern a similar sort of characteristics that we do with a full series of Klippel NFS data for a loudspeaker. High frequency data on headphones has always been an issue due to the small nature of the space in which the transducer operates. B&K has offered up some band-aides in their latest offering but not a clear solution. My hope is that more people will start thinking about it snd perhaps look into some alternatives that make more sense acoustically.

 

[storing]

If I am buying an expensive pair of headphones, I want the measurement data to be as precise as possible across the entire spectrum so I know what I'm getting into.

For lower priced stuff, some anomalies in the data isn't as important as long as I know the nature of the anomalies

Do I understand correctly that for you there's a threshold expressed in $ below which you don't care about precise measurements anymore? Even though this site itself has shown that cheaper gear - including IEMs and headphones - can measure very good, if not as good, as expensive gear. And obvisouly to do that the exact same type of measurements are needed for all gear. Wouldn't you care that people might be missing out on non-expensive but good gear if we'd start skipping (or doing different) measurements on cheaper gear?

But due to the nature of headphone measurements, the problem isn't necessarily exactly where the measurement say it is, but it's also not far off either. So it's useful information.

This is a bit confusing You first said that not being able to determine the exact location of high frequency issues is a problem. Which lead me to ask: is that really a problem, or is it good enough if it's not too far off? To which you said knowing the exact location is useful 'from a data perspective' and 'really important for expensive gear'. But now you seem to use my reasoning as well, saying it's also useful if it's not too far off. So we actually seem to agree on that point?

 

[LevityProject]

Let's recap: With headphone measurements, above 8khz isn't as precise as below 8khz (there was some talk about IEM measurements being more precise than Over ears). From a purely scientific standpoint, it's still not completely "reliable" for telling the true exact frequency response of the headphones above 8khz.

This whole thread is about that fact, and also the fact that even though the measurements above 8khz aren't as precise as below 8khz, if the measurements are done well, it's still useful data ... The measurements and real world tone of the Salnotes zero are a pitch perfect example.

The measurements of the beyerdynamic DT990 pros is another perfect example. The issues that show up in the measurements above 8khz are REAL.

The zeros: We've got a problem around 12.5khz in those headphones. This ballpark is accurate enough to make consumer related decisions .. eg if you have good hearing above 8khz, don't want a 10-13khz ring and don't know how to use wavelet to tune the ring out, don't get these because that ringing is REAL. It's precise location might different from person to person due to limitations in the science and personal hearing characteristics, but it's THERE.

That's it.

Again: the measurements of the Salnotes zeros show a ringing at around 12.5khz. that ringing is REAL, though it's exact location can vary from person to person because of variances in hearing, ear shape, and the imperfect science of headphone measurements. But it doesn't mean the ringing isn't there or that (as some would say) the science is useless.