Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Are all headphones equal? (Read 4686 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Are all headphones equal?

OK, that question is of course a wild exaggeration posed to catch your attention ;)

However: I recently spent some to to fine-tune different headphones to a near-perfect, small room corrected diffuse-field response curve in order to answer that question for myself. Given that headphone drivers are expected to very closely approximate a minimum phase system, any decent pair of cans (decent translate to sufficiently low distortion levels any [say < 0.2% at 1kHz and < 1% at 100 Hz @ 90dB] and a relatively smooth frequency response up to 10 kHz) should be able to almost perfectly reproduce the sound sound signature and thus perceived quality [insert all that audiophile jargon here] of any other pair of headphones. At least this is what the research reported in this article at InnerFidelity strongly suggests. Fascinated by this approach, I tried to "validate" the findings in a heavily biased and statistically utterly insignificant n=1 experiment and got a pair of Sennheiser HD598 with the plan to equalize them such that they sound like a pair of HD650. To do so, I gathered the measured frequency response graphs from three trusted sources (InnerFidelity, Headphone.com and GoldenEars) and analyzed them for common characteristics. I concluded that I need a least a 30 band (1/3 octave-width) graphical EQ, or optimally, a parametric EQ to manipulate the response curve with a sufficient level of precision. I quickly decided to give up on corrections aiming for more than 1/3 octave precision, as the variations between my sources were to large to allow for that. What I ended up with*, though, employing only 4 bands of Rockbox' parametric EQ, has really convinced my of the validity of this approach, indeed giving the HD598 a sound very close to an unequalized HD650 or even better (more "open" towards the high end). Have any of you made similar experiences or theoretically examined this subject?

*For those of you who are interested in the exact settings: Pregain -6 db, low-shelf 30 Hz +12 db q=0.7,  peak 2 kHz +2db q=1.4, peak 9 kHz -3db q=4.3, peak 12 kHz +6db q=4.3. Note that when listening to movies or bass heavy electronic music one may have to set the pre-amp even lower to avoid clipping. Most naturally recorded music however won't have sub-bass content at anything close to -6dB FS.

Re: Are all headphones equal?

Reply #1 –
Frequency response is by far the largest component of how headphones sound, so adjusting it can make one device sound like another.  It's not the only one though, distortion and even background isolation also contribute to the sound of a pair of headphones.

Re: Are all headphones equal?

Reply #2 –
Agreed, but if distortion is low enough to be entirely inaudible (I think the figures mentioned should fall well below the threshold of audibility), what remains? I can only think of two factors: 1) "Openness" (the sound pressure level on the outside divided by the level of isolation), providing both natural crossfeed and ambience and 2) the angle of the driver relative to the ear (what Sennheiser calls E.A.R. technology). I'm not sure whether this is more marketing than anything else but I would assume that drivers that are angled towards the front allow for more interaction with the earlobes, thus shaping the signal in a way that improves the sound stage and helps to shift it to the front. Interestingly, I found the sound of the HD598 too harsh when I completely compensated the dip at 1-3 kHz, and this dip corresponds perfectly with the delta of free field vs diffuse field response.

Btw. you are among the people I hold responsible for the fact I have become totally spoiled. :) When I tried to switch my home listening over to DLNA I went almost mad because of the lack of custumization, (properly implemented) gapless playback, crossfeed and equalization even with the most "advanced" apps and devices. It felt like one step forward and several steps back. I've managed to get at least gapless FLAC playback working with BubbleUPNP, but AFAIK there is still not a single DLNA app available providing a usable equalizer, let alone a crossfeed. I still have to use my Sansa Clip for headphone listening. Yet I'm affraid that one day such devices will become basically extinct and rockbox with them... unless it will be reborn in some way as an DLNA-capable app.

Re: Are all headphones equal?

Reply #3 –
Quote from: Northpack on
OK, that question is of course a wild exaggeration posed to catch your attention ;)
We can take it but in an audiophile forum you would have caused a few heart attacks. :P

Adding to what saratoga said, you have to keep mechanical limitations in mind. Even if a "simple" (single-driver, no multi-driver/crossover stuff) headphone approximates a min. phase system well enough in its stock form that does not mean that it still will do so after equalization. Some resonance may be hard to fix with an EQ, as will be e.g. bass roll-off.

The frequency response measured on a dummy head will look differently on your own head, especially at high frequencies. The magnitude of this inconsistency also depends on the headphones. I think that the HD5/6xx series is more consistent in this regard.

"I hear it when I see it."

Re: Are all headphones equal?

Reply #4 –
Quote
. I concluded that I need a least a 30 band (1/3 octave-width) graphical EQ, or optimally, a parametric EQ to manipulate the response curve with a sufficient level of precision.
I doubt you need 30 bands, but parametric EQ may be helpful in some cases.  Remember you're not trying to match the measurements, you're just trying to match the sound.   (And, you were reasonably successful with 4 bands.)

And of course, the real difficulty with this experiment is the listener.  Since different headphones usually feel  different, it's very difficult to do a blind ABX test.

Re: Are all headphones equal?

Reply #5 –
Quote from: xnor on
We can take it but in an audiophile forum you would have caused a few heart attacks. :P
I know. This is why I don't post there anymore. Too much colateral damage :))

Quote from: xnor on
Adding to what saratoga said, you have to keep mechanical limitations in mind. Even if a "simple" (single-driver, no multi-driver/crossover stuff) headphone approximates a min. phase system well enough in its stock form that does not mean that it still will do so after equalization. Some resonance may be hard to fix with an EQ, as will be e.g. bass roll-off.
Could you elaborate on that? Why and when does equalization (linear phase, of course) create nonlinear effects? I agree that there's a mechanical limit for the amount of bass a driver is able to reproduce such that equalizing the rolloff may in some cases cause distortion to become audible even at moderate listening levels. Resonances in the higher frequencies can indeed be hard if not impossible to fix if they produce too narrow spikes and dips in the frequency response, going beyond the limit of what you can target even with a full-blown parametric EQ, if you can measure them accurately enough in the first place. What I don't understand is whether you expect resonances to cause nonlinearities other than in the frequency domain?

Quote
The frequency response measured on a dummy head will look differently on your own head, especially at high frequencies. The magnitude of this inconsistency also depends on the headphones. I think that the HD5/6xx series is more consistent in this regard.

I also think so, or so it seems from the frequency response graphs. This is probably why Olive et al. chose the HD 518, which has the same driver as the HD598 I made my experiment with. By the way, this is a good example for the dilemma headphone manufacturers are confrontend with, and that manufacturers of ICs have long known: that once the science and know-how have been established, the unit price for mass-producing even the most excellent headphones drivers eventually comes down to a level where you could theoretically sell nearly perfect sounding headphones for cheap (see the Superlux phenomenon). In order to maintain the established price differentation with higher margins for "hifi" products, the companies either have to add some bling factor (e.g. leather headband and plastic that looks like burl wood as in the case of the original HD598) or they have to cripple down the performance of the drivers by deliberately adding dampeners to the case that diminish the high frequency response (as in the case of the HD 518 and 558).

Quote from: DVDdoug on
I doubt you need 30 bands, but parametric EQ may be helpful in some cases.  Remember you're not trying to match the measurements, you're just trying to match the sound.  (And, you were reasonably successful with 4 bands.)
Certainly not, but you need the 1/3 octave width only 30 band graphic equalizers provide in order to make sufficiently precise corrections in the uppermost frequency range, even for the relatively well-behaved HD5x8 series.

Re: Are all headphones equal?

Reply #6 –
So, you only need to carry a 30-band EQ with your headphones?

Re: Are all headphones equal?

Reply #7 –
Quote from: theriverlethe on
So, you only need to carry a 30-band EQ with your headphones?

Even on an old ARMv5 processor, one parametric EQ band at 48 KHz stereo is only about 2 MHz of CPU.  So 30 bands is only 60 MHz.  On a cell phone with NEON instructions its probably less than half that, so you could easily do this on a modern cell phone and never notice the CPU hit. 

Re: Are all headphones equal?

Reply #8 –
Hmmm, Neutron offers four bands of parametric EQ.  This might be a better option than a graphical equalizer.

Anyway, does perception of "transient response" correlate with FR?  What about spatial perception and "soundstage?"

Re: Are all headphones equal?

Reply #9 –
A company called Sonarworks makes a headphone calibration VST plugin with quite a few stock profiles, and will even measure your sample (for a small fee).  Their house curve sounds quite a bit like HD650's, but it is user configurable.  Are there any mobile players that support VST?

Re: Are all headphones equal?

Reply #10 –
Quote from: theriverlethe on
Anyway, does perception of "transient response" correlate with FR? 

The impulse response of a device is just the fourier transform of the frequency response, so one determines the other.

Quote from: theriverlethe on
A company called Sonarworks makes a headphone calibration VST plugin with quite a few stock profiles, and will even measure your sample (for a small fee).  Their house curve sounds quite a bit like HD650's, but it is user configurable.  Are there any mobile players that support VST?

I don't know about VST, but on Android Viper Audio is fairly popular:

http://vipersaudio.com/blog/

Ignoring the more gimmicky effects, it can convolve audio with an arbitrary impulse response.  You can find impulse response files for many models of headphones. 

Re: Are all headphones equal?

Reply #11 –
Northpack, very cool! Also check out Rin Choi's review of Accudio: http://rinchoi.blogspot.ca/2013/04/accudio-accurate-audio-player-by.html

Quote
Some resonance may be hard to fix with an EQ, as will be e.g. bass roll-off.
about resonances, I've always wondered how to interpret CSD waterfall plots of headphones - do they contain info beyond what is in the FR plot? e.g. the Denon D2000 measurements here: http://en.goldenears.net/4170

Re: Are all headphones equal?

Reply #12 –
I think a 4-band parametric is way too few bands to tailor the sound of one headphone to another... even if it's parametric.
Quote
The frequency response measured on a dummy head will look differently on your own head, especially at high frequencies. The magnitude of this inconsistency also depends on the headphones.

That's why I've hardly ever bothered with instrumental measurements and always went with perceptual sweep testing :D
http://www.head-fi.org/t/794467/how-to-equalize-your-headphones-2016-update

I don't know exactly what the ideal frequency response of a pair of earphones should be as perceived by my eardrums, but removing those high frequency resonance peaks and dips via EQ has always resulted in a much more pleasurable, "open" sound for me.

In my case I'm not trying to turn one earphone into another, but simply to optimize the sound of all the earphones I own.  And I find that once optimized, one pair sounds pretty much like another, especially IEMs.

Viper4Android is indeed the real deal--it lets you import stereo impulses for convolution, so you can have arbitrary EQ curves for the left and right ear.

For creating the curves in the first place, the over-ten-year-old Electri-Q free edition by Christian Budde is still THE go-to app for me.  It's an unlimited-band parametric EQ (with arbitrarily wide or narrow bandwidth per band) with instant graphic feedback!

Attached is a pair of real EQs I use for the left and right channels of a real pair of IEMs.  20 bands per channel... I could simplify the EQ if I had to export it to other apps, but within Electri-Q who cares? :D

Re: Are all headphones equal?

Reply #13 –
Quote from: Joe Bloggs on
That's why I've hardly ever bothered with instrumental measurements and always went with perceptual sweep testing :D
Measurements are actually fine until you get above a few kHz.
Sweeps are not that helpful to equalize tonality but yeah, as you write later, they certainly help to locate (individual) resonance peaks.
"I hear it when I see it."