• 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!

Master Complaint Thread About Headphone Measurements

Maybe just including another line in the graph, with quieter measurements of IEM's to prevent those pitfalls, would be everything needed?
 
I don't agree with your assessment here, because if what we wanted was to single out response to "peak" levels, we would be looking into transient response (I don't know if there's an objective measurement for this, is there?).
The transient response can be deduced from the impulse response which can be obtained from the Fourier transform of the magnitude and phase response. Room EQ Wizard makes it pretty easy to retrieve that transient response or go the other way and calculate the magnitude and phase from a direct impulse response measurement.

2023-12-13 - Meze Elite hybrid L 4M8R - magnitude and phase response.jpg

Figure 1: Meze Elite hybrid pads with "V3.1 PEQ" left ear blocked canal magnitude and phase response.

2023-12-13 - Meze Elite hybrid L 4M8R - impulse and step response.jpg

Figure 2: Impulse and step response derived from the same.

What I don't know is whether dynamic compression or changes in FR at higher playback levels accurately translate; i.e. whether the same bass decrease seen in a sine sweep would be measured from directly measuring the equivalent loud Dirac impulse provided that the microphone, amp, and audio interface can handle it.

I would concur that taking a white spectrum distortion sweep at such loud playback levels would be quite unrealistic, whereby applying something like a pink spectrum sweep if that were possible, or like I had suggested in https://audiosciencereview.com/foru...phone-measurements.18451/page-61#post-1956895 (post #1,210), a pink spectrum multi-tone, might better simulate the realistic distribution of SPL levels for the frequency content within music.
 
I don't agree with your assessment here, because if what we wanted was to single out response to "peak" levels, we would be looking into transient response (I don't know if there's an objective measurement for this, is there?).
They are actual measurements of the voltage on a driver and the SPL (excursion) of the transducer, assuming there is no compression, is real.
Music consists mostly of transients.
Low frequencies require larger excursions to reach the same SPL. For 20kHz to be reproduced at say 100dB SPL a much smaller excursion is needed.
To reproduce a square-wave of say 1kHz increasingly smaller amounts of higher frequencies are needed so when a headphone can reproduce 20kHz at say 110dB SPL and also can do 1kHz at 100dB SPL more 'speed' is not needed.

Here is the impulse response for MDR-7506
step-7506.png


Below the step response for Sennheiser HD580 Precision.
step-hd580.png


One can also measure the a square-wave (40Hz and 440Hz) and needle (impulse) response
hd580-sq-kl.png


Below the Hifiman HE400SE:
sq-he400se.png



Since what we're looking for are best "average" response representation, then it makes sense to use the most "average" of levels, which is also the level defaulted to by most mastering engineers to judge a master, and that is hovering around 85db.

Actually, averaging 85db is already quite loud, and louder than most people I worked with monitor for long periods, 85db is a level to raise up to as you work, then back down to "rest." We're talking about a level that's safe for 8 hours a day and may cause hearing loss (still average, so the fluctuations are implied, but nonetheless represented in this average). Talking about a constant sweep of frequencies for a response, a sweep of 85db should represent that average appropriately.
Yep, average 85dB is quite loud. 85dB peaks are not and that's what we listen to so one should measure at that peak level in order to ensure that lower levels are O.K. too.

That's exactly my point... I didn't measure the Reds, but I heard this over the course of a few days, and it's kind of reasonable to assume that that happened in a very similar manner even if the headphones depicted are different.

Given the distortion plots of the red even at 114dB there is no hint of distortion in the bass so there is no dynamic compression. The FR at 114dB thus is the same as at 70dB SPL.
Considering both that the test results Amir posted had nothing to do with what I heard, and the differing resistance values on the bass response of the iem's, I assume in your graph that the lines at 90 and 100db would've been closer to Amir's measurement, and since I rarely go much past the purple 80db line, I'm stuck with a "bloated" experience in real life, nothing like the loud tests...
When measuring at lower SPL you run into noise floors of the mic (test fixture) and surrounding sounds.
One can measure the FR at many different levels if the the distortion plots indicate there is compression (increased odd harm. distortion at higher SPL).
Not many people do this or publish that.
 
Last edited:
They are actual measurements of the voltage on a driver and the SPL (excursion) of the transducer, assuming there is no compression, is real.
Music consists mostly of transients.
Low frequencies require larger excursions to reach the same SPL. For 20kHz to be reproduced at say 100dB SPL a much smaller excursion is needed.
To reproduce a square-wave of say 1kHz increasingly smaller amounts of higher frequencies are needed so when a headphone can reproduce 20kHz at say 110dB SPL and also can do 1kHz at 100dB SPL more 'speed' is not needed.

Here is the impulse response for MDR-7506
step-7506.png


Below the step response for Sennheiser HD580 Precision.
step-hd580.png


One can also measure the a square-wave (40Hz and 440Hz) and needle (impulse) response
hd580-sq-kl.png


Below the Hifiman HE400SE:
sq-he400se.png




Yep, average 85dB is quite loud. 85dB peaks are not and that's what we listen to so one should measure at that peak level in order to ensure that lower levels are O.K. too.



Given the distortion plots of the red even at 114dB there is no hint of distortion in the bass so there is no dynamic compression. The FR at 114dB thus is the same as at 70dB SPL.

When measuring at lower SPL you run into noise floors of the mic (test fixture) and surrounding sounds.
One can measure the FR at many different levels if the the distortion plots indicate there is compression (increased odd harm. distortion at higher SPL).
Not many people do this or publish that.

Thank you!! That sums up a lot of what we were talking on the other thread as well, I suspect my new LCD-2 Classic to perform better than all the headphones pictured, and the HE400SE and 580 prints definitely seem like what I was listening in my head that made my life harder when judging higher frequency percussion and some other high frequency content with the Sundara, and definitely has to do with the plot you posted of the resonances and poor damping altogether. There's "speed" (or lack of) depicted for sure...

Now my point with the Reds, and I may or may not be correct, but it's a hunch, is that the crossover and varying resistance has an effect when more current comes to play, and that wouldn't necessarily be perceived as compression. Not just Fletcher-Munson while listening (although it would be an effect for us) because the measuring rig isn't a human ear. The higher you go in volume with a headphone that has different impedance across the spectrum, the less increase you will get where it's harder to drive. So the Reds might start with way more bass than its given balance point in loudness, Amir's test might have gone past (that's why he measured "less" bass), while I was residing underneath (and had overwhelming bass). The fact that they play with that with the dongle should be an indication that the unit doesn't behave consistently across different loudness levels. I assume, because I haven't measured, but...
 
Due to its impedance it is highly likely the Red will sound incorrect from higher output impedance sources.
Most interfaces have a too high output resistance and some studio gear may also have that. The reason for that is usually that the headphone out of such devices is often an afterthought and just an op-amp with some 'safety' resistors in the output.
It is also the reason why the bass boost using an in-line resistor works.

The rest of the measurements do not suggest tonal changes caused by level differences.
Differences can come from equal loudness contours and could come from a too high output impedance from the source is is connected to. But this is not level dependent.

IMHO Harman bass is too much bass boost for critical listening. It is fine for listening to music while commuting or enjoying bass-light recordings or listening at low listening levels.
Just my personal opinion.
 
Last edited:
Due to its impedance it is highly likely the Red will sound incorrect from higher output impedance sources.
Most interfaces have a too high output resistance and some studio gear may also have that. The reason for that is usually that the headphone out of such devices is often an afterthought and just an op-amp with some 'safety' resistors in the output.
It is also the reason why the bass boost using an in-line resistor works.

The rest of the measurements do not suggest tonal changes caused by level differences.
Differences can come from equal loudness contours and could come from a too high output impedance from the source is is connected to. But this is not level dependent.

IMHO Harman bass is too much bass boost for critical listening. It is fine for listening to music while commuting or enjoying bass-light recordings or listening at low listening levels.
Just my personal opinion.

I used a Fireface UCX-2, which has an 1 Ohm output with 210mW power per channel. So I assume the sound was correct (didn't get so far as plugging the Reds into my Adi-2 FS because that's where I run the Audeze/Sundara).

Maybe it was just an equal loudness thing then, or maybe they have bigger unit variance than we assume...

As for Harman being too much, I am ok with that amount of bass on big over-ears since I have the feeling most modern music is produced to play on relatively bass-boosted systems. That provided the high-mids are nicely filled out too to compensate for that... Somehow nothing can be too scooped out otherwise there's only bass left.

So If I go through my references and have the feeling they sound "right", then I'm good with it. It only bothers me when we get to notoriously bassy tracks (like a few Run The Jewels tracks, vs Aquemini from Outkast which sounds "right" everywhere even if it's clipping like there's no tomorrow).

Btw the only tracks that sounded "correct" in the Reds to me were a few of the hardest mid-pushing pop tracks on my list... And I must say the past 10 days were of intense listening while producing because of the change up, the Audeze with sunken low-mids and boosted high-mids and low-lows made most of my playlists make sense, so I started trusting it quick, the 7Hz Zeros too! Just the Reds were far off...
 
And to follow on from above, here’s the talk by Blaine that Andrew refers to in above video

Good video by Blaine giving a talk at the recent CanJam regarding the variability of headphone/iem measurements and how that relates to a clusterf*ck of misinterpretation regarding the known data

(With occasional interruptions and clarifications courtesy of Oratory)

TLDR/TLDW - Headphone data does not currently portray the range of potential sources of variation perception between listeners

 
A nice compact overview presentation of the currently still existing limitations and problems when measuring, listening and evaluating headphones:

And to follow on from above, here’s the talk by Blaine that Andrew refers to in above video

Good video by Blaine giving a talk at the recent CanJam regarding the variability of headphone/iem measurements and how that relates to a clusterf*ck of misinterpretation regarding the known data

(With occasional interruptions and clarifications courtesy of Oratory)

TLDR/TLDW - Headphone data does not currently portray the range of potential sources of variation perception between listeners

I think it's probably true that a single line on a graph does not totally explain a headphone's performance for any given individual, but the core "problem" remains of "what you going do about it". That single line on a graph (GRAS frequency response) is a pretty strong predictor of preference, and to go beyond that in terms of providing useful information to prospective buyers would mean research/studies/listening testing to prove that other repackaged information from different measurements (whatever they may be) is proven useful. So the various ideas are largely theoretical at the moment, it would now be down to whoever to do the work to prove the various mechanisms & effects & somehow package that information in reviews to make it useful for prospective buyers.

Given that we are all individuals with our own anatomy I suppose the thrust of any work that is done would be to somehow prove & assess the degree of variability that any given headphone has for different people, and the buyer would choose a headphone that has low variation to ensure that they don't get something that sounds widely different to what is expected from the published frequency response measurement. Some websites like RTings already do something like this. How far would a website/reviewer need to go to prove that it's important? How would they do their variation measurements, how many different heads would they use? Would they subtract the individuals HRTF from the measured HpTF to try to get at the theoretical heard variations for that individual? Would that website/reviewer have to prove with listening tests somehow that subtracting HRTF from HpTF is a valid thing to do? Just how far & deep would these reviewers/websites need to go in proving that their "variation measurements" are valid & representative & repeatable? How extreme do you go with choosing different anatomy extremes for the people & heads you have in your variation testing? How many individuals & different heads should you use in your testing of the headphone to create your variation data? I suppose there's a lot to consider, and it'll only be as good (& useful) as the depth/validity of the work involved. I suppose the challenge is working out how deep you need to go with this in order to output something that is really genuinely useful to headphone buyers in terms of a predictor of their experience.
 
Last edited:
I am interested in learning about specific headphones that are light with the bass frequency. I am one of those individuals who prefer MUCH LESS amount of bass. Full disclosure, I have a ski-slope audiogram, with a great ability to hear the bass but not the high frequency. I am hard of hearing and find that most headphones emphasize too much bass for my preference. I have preferred AKG 701/702 over the currently popular AKG 371 that follows the Harman curves. Thanks in advance.
 
I am interested in learning about specific headphones that are light with the bass frequency. I am one of those individuals who prefer MUCH LESS amount of bass. Full disclosure, I have a ski-slope audiogram, with a great ability to hear the bass but not the high frequency. I am hard of hearing and find that most headphones emphasize too much bass for my preference. I have preferred AKG 701/702 over the currently popular AKG 371 that follows the Harman curves. Thanks in advance.
Instead of searching for headphones that match your specific requirements out of the box, why not use EQ to make any headphone suitable?

With EQ, you can dial in your preferred sound signature precisely, often completely for free.
 
Instead of searching for headphones that match your specific requirements out of the box, why not use EQ to make any headphone suitable?

With EQ, you can dial in your preferred sound signature precisely, often completely for free.
I have done that in the past, with ASUS equalizer setup...which is the best for me. But I want to use a headphones without having to rely on an equalizer
 
I am interested in learning about specific headphones that are light with the bass frequency. I am one of those individuals who prefer MUCH LESS amount of bass. Full disclosure, I have a ski-slope audiogram, with a great ability to hear the bass but not the high frequency. I am hard of hearing and find that most headphones emphasize too much bass for my preference. I have preferred AKG 701/702 over the currently popular AKG 371 that follows the Harman curves. Thanks in advance.
How about trying Oratory's Optimum Hifi Target Curve. It has no bass shelf, just linear extension. It's probably a bit misleadingly titled because I really don't think it should normally be considered "optimum", but in your case you might like it given it has no bass shelf. Following is an example of the HD560s EQ'd to Oratory's Optimum Hifi target:
Optimum Hifi Curve.jpg

So you could ask Oratory (over on his reddit) to do an Optimum Hifi Target EQ for your K702, and following is your K702 EQ'd to Harman just as a comparison to the above:
K702 Oratory example.jpg
 
I have done that in the past, with ASUS equalizer setup...which is the best for me. But I want to use a headphones without having to rely on an equalizer
This is going off-topic… Sorry!
A potential solution is to look for a relatively inexpensive “EQ-er” and leave it attached to the headphones.

For example I have a Qudelix 5K “permanently” attached to my Hifiman Sundara, EQ’ed to what works for me. It makes my Sundara an—almost-universal—USB-C HP… and gives me BT as a bonus.

I also have a Moondrop FreeDSP cable “permanently” attached to an Aune AR5000 HP, also EQ’ed to what works for me. It also makes the AR5000 into a USB-C HP.

The point here is to factor the cost of that “EQ-er” into the HP: Neutron DAC V1, Qudelix 5K, FiiO KA15, FiiO BTR13, Moondrop FreeDSP, FiiO JA11… all support advanced PEQ features and can drive adequately many good headphones.

EDIT: forgot to mention this… By removing the HP freq. response out of the equation (it is EQ’ed to whatever works for you), you can focus on other important characteristics of the HP, such as fit and comfort…
 
Last edited:
I have done that in the past, with ASUS equalizer setup...which is the best for me. But I want to use a headphones without having to rely on an equalizer
That's understandable but hear this, getting bass correctly equalized is MUCH easier than fiddling with the higher frequencies. As per above post, the K702 can have linear extension with just two simple negative filters that won't mess up with phases or have to deal with weird cup resonances.

If you're still set on getting a decent sounding headphone out of the box though, I'd stray you towards most of the open ear models and away from the closed ones like the K361. Getting a bass shelf or even linear bass is easier with a chamber for the waves to bounce back from, so most open-eared headphones "linearity" it's most often a inefficiency turned into feature.

In this context, I'd recommend you seek out the open-eared planars recommended in ASR from Hifiman, like the Sundara or the HE400se

1731446974915.png
1731447000395.png


In an even more unorthodox approach, considering your ski-slope audiograms, and assuming you don't have tinnitus or anything inducing high frequency hyperacusis, you can try to "compensate" your hear loss with artificially bright and vocal centric headphones like the oval Hifimans such as the Anandas and Edition XS's (or even, on the cheaper side, the Moondrop Para):

1731447195315.png


In an ending note, Sai's squig.link has a plethora of headphones measured AND targets like the old Harmans (that lean into the brigther spectrum), Linear Harman, optimum hifi etc, and even Diffuse Field, which is the target your AKGs are based around (bright leaning and very vocal forward).
 
That's understandable but hear this, getting bass correctly equalized is MUCH easier than fiddling with the higher frequencies. As per above post, the K702 can have linear extension with just two simple negative filters that won't mess up with phases or have to deal with weird cup resonances.

If you're still set on getting a decent sounding headphone out of the box though, I'd stray you towards most of the open ear models and away from the closed ones like the K361. Getting a bass shelf or even linear bass is easier with a chamber for the waves to bounce back from, so most open-eared headphones "linearity" it's most often a inefficiency turned into feature.

In this context, I'd recommend you seek out the open-eared planars recommended in ASR from Hifiman, like the Sundara or the HE400se

View attachment 405924View attachment 405925

In an even more unorthodox approach, considering your ski-slope audiograms, and assuming you don't have tinnitus or anything inducing high frequency hyperacusis, you can try to "compensate" your hear loss with artificially bright and vocal centric headphones like the oval Hifimans such as the Anandas and Edition XS's (or even, on the cheaper side, the Moondrop Para):

View attachment 405927

In an ending note, Sai's squig.link has a plethora of headphones measured AND targets like the old Harmans (that lean into the brigther spectrum), Linear Harman, optimum hifi etc, and even Diffuse Field, which is the target your AKGs are based around (bright leaning and very vocal forward).
Thanks for pointing out various ways to get what I want. I agree that the open headphones worked better for me, after having tried several closed headphones that amplify the bass, disrupting the overall spectrum of sounds. A good example of the extreme basshead was JVC's HA-RX900 headphones. WOW...avoiding that one!

Thanks to all of those who shared their thoughts on this matter. Most appreciative.
 
Within https://www.head-fi.org/threads/stax-sr-x9000.959852/post-18447647 (post #3,323) I present a follow-up to the multi-tone distortion findings from https://www.audiosciencereview.com/...phone-measurements.18451/page-61#post-1956895 (post #1,210). Here, the Stax SR-X9000 and EQed Meze Elite had comparable harmonic distortion levels, whereby it appears that the Meze Elite's lower bass distortion correlates with its ultimately exhibiting lower multi-tone distortion than the X9000 despite said headphone being driven by the powerful Carbon CC.

Thus, multi-tone distortion measurements can help further discern the excellence of certain drivers, and the outcome can likely be predicted from the bass distortion insofar as higher harmonic distortion of lower frequencies incurs higher intermodulation distortion with the upper frequencies. Hence, for the purpose of ASR, the practical point of action should multi-tone distortion measurements not be feasible or economical would be to lower the noise floor of the measurement setup or environment or take advantage of measurement averaging, allowing for the resolving of lower bass distortion levels.
 
@amirm One measurement that I would absolutely love to see in more reviews (and one you used to include sometimes) is the physical size of headphone ear pads. As someone with larger and very touch-sensitive ears, many otherwise good headphones are too small. Like nozzle size on IEMs, interior pad dimensions are something that really should have been a specification shared by the manufacturer, but since very few manufacturers share either it's up to you and other reviewers. Height/width measurements should be fairly easy to take, and although depth measurements won't be perfect due to variations in how much pads compress it's still better than nothing.
 
@amirm One measurement that I would absolutely love to see in more reviews (and one you used to include sometimes) is the physical size of headphone ear pads.
I started to measure these early on. It is a bit of pain and hard to characterize with odd shaped pads. I just didn't enjoy doing it so gave up. :)
 
I started to measure these early on. It is a bit of pain and hard to characterize with odd shaped pads. I just didn't enjoy doing it so gave up. :)
Understandable, but if you ever decide to start again it would be highly appreciated! Also IEM nozzle measurements, which should be much easier as they're nearly always round.
:)
 
Back
Top Bottom