How can you make bad headphones sound great? What determines the sound quality of headphones according to research?

I’ll start with a small illustrative example. By

this link

you can download and listen to recordings of three headphones of different price categories (HiFiMAN Sundara – $350, Creative Aurvana Live! SE – $60, Takstar PRO82 – $100) made with the help of an artificial ear. The distortion it introduces is similar to that of a real human ear. With just one equalizer, these recordings have been corrected back to the original playable file. Try to determine which recording corresponds to which headphones and where the digital original is

. For these headphone recordings, only one type of distortion was corrected – linear (or, in other words, only their frequency response was corrected – the amplitude-frequency response). All other distortions introduced by the artificial ear remained uncorrected. Among the files there is also the original playback file (digital source of recordings). It is present both explicitly and hidden among 4 files – A, B, C, D. There are also non-equalized records (in the unequalized folder) at the link.

Were you able to find the hidden original? How big is the difference between the recordings and the original? Despite the fact that the recording was made on a cheap stand, and the equalization was done after recording, which reduced the already low dynamic range and increased distortion, the sound between all the files after equalization turned out to be very similar. Moreover, the difference in sound is caused, first of all, not by the difference in the equalized headphones, but by the shortcomings of the implementation of the method: the low signal-to-noise ratio of the recording setup and the correction of the frequency response after recording, and not before it. In a more qualitative experiment, using a simulation of a human ear of a higher level, you can achieve even more sound similarity.

The question may arise, is such a test generally correct? After all, you listen to the recordings through another pair of headphones or speakers, introducing their own distortions. To eliminate this problem, the original recording is present in the test. When you listen to it, you hear only the distortion of your headphones and your reproducing path. When you listen to recordings, you hear the distortion of your headphones and your playback system PLUS the distortion of the headphones being recorded. So you are comparing

“your playback system distortion” vs. “your playback system distortion + recording headphone distortion”

. That is, in this case,

the audible difference will primarily reflect the distortion of the recorded headphones

. Of course, ideally, the recordings should be listened to on the highest quality audio equipment available to you.

Sean Olive, ex-president of The Audio Engineering Society, the largest community of scientists and engineers in the field of audio, over the past 20 years, along with colleagues, has done a lot of research on the topic of various sound systems in general, and headphones in particular. Research includes objective and subjective ways to evaluate sound quality, determine its main factors, etc. Sean Olive profile at AES:

www.aes.org/aes/seanolive

As a result of research, Sean Olive and colleagues came to the following conclusion:

sound quality majority Modern headphones are determined exclusively by their frequency response – amplitude-frequency response. The subjective sound quality of headphones is the higher, the better they reproduce the frequency response that perceived specific human ear when listening to good monitors in an acoustically prepared room. We are talking about exactly the frequency response that occurs at the level of the eardrum – and it is very different from linear.

The frequency response is the very thing that can be changed with the equalizer.



FIGURE 1. The initially flat frequency response of the acoustic system (bottom of the figure), having interacted with the ear and ear canal, turns at the level of the tympanic membrane into approximately the one shown at the top of the figure.

Hundreds of listeners took part in these studies and experiments – ranging from those trained in the Harman laboratory according to a special program, and ending with ordinary people “from the street.” It turned out that non-linear distortions, for most modern headphones, are not a significant factor, since, most often, on a “live” musical signal, they are beyond audibility. Thesis, in the form of a presentation, the key works on this topic are collected in this document.

about these studies

The paper collects and summarizes the results of numerous studies, including with the participation of trained listeners who not only have objectively good hearing, but were able to show that they can pick up subtle differences in timbre, distortion and other nuances, according to a special program of the Harman Research Laboratory:

harmanhowtolisten.blogspot.com

Also, there is a video with Sean’s interview on the topic of headphone sound, which can be found (in English):

https://habr.com/en/post/566846/image

But what about the “detail”, “scene”, “drive” and other properties that some audio enthusiasts characterize the sound of headphones?

Sometimes it’s just audiophile fiction, and sometimes it’s just frequency response side effects.

. So, for example, in the case of headphones, the “scene” is determined primarily by the activation of resonances and interferences inherent in a particular auricle. These are, after all, just peaks and dips in the frequency response that, if individually measured, can be emulated by an equalizer.

Good headphones will activate these individual resonances without the need for EQ.

.



Figure 2. The shape of the auricle determines the individual resonances and frequency response drops perceived by the tympanic membrane.

In addition to distortion, on some resources you can find phase measurements, impulse response, step response, and more. The truth is that the impulse response and step response is just another way of displaying the frequency response and phase response – in the time domain instead of the frequency domain. I.e,

knowing only the frequency response and phase response, you can build the response of the system to a single impulse or step without any measurements at all

. Well, headphones are a phase-linear system – changing their frequency response changes the phase, and changing the phase changes the frequency response. That is, in their case, the shape of the impulse response or response to a step can be recreated using only a phase-linear equalizer and nothing more by applying the measured uncompensated frequency response to the original ideal signal (step or impulse) in an audio editor. That is, in the case of headphones, as a phase-linear system, all these measurements do not carry any information at all and only display the frequency response in a different way. Examples where the shape of the impulse and other characteristics were accurately predicted with the help of an equalizer alone:

Example 1

Example 2

Example 3

And what about the so-called “waterfalls”? They, in the case of a phase-linear system, also only reflect the frequency response. Remove the peak – remove the “ringing” of the waterfall. Article on this topic. And there is no correlation between the sound quality and the type of waterfall. This topic in detail researched by Floyd E. Toole many years ago.

In general, according to real research and mathematics, the answer is unambiguous:

other factors besides the frequency response, most often, are not significant for the sound quality of headphones. Only the frequency response and, occasionally, distortion are really important, if the latter exceed the threshold of audibility.

It should be noted here that even $10 headphones can have a very low level of distortion, beyond audibility on a music signal. For example:

Panasonic RP-HJE120 measurements

SONY MH750 measurements

It turns out that if you accurately equalize the headphones to one frequency response, then they will also sound indistinguishable? For most headphones, regardless of price range, the answer is clear:

yes they sound the same

. More precisely, different models of headphones will sound just as similar as the same headphones worn differently. Because even the same headphones change the sound when they move a few millimeters relative to the ear, or when they are worn on long hair, or on the headband of glasses, or, for in-ear headphones, are inserted a little deeper into the ear canal. That is, every time you put on the headphones, you will perceive a slightly (and sometimes significantly!) Different sound – this is a fact that is easily measured.

But the main problem lies in the fact that we are talking about the perceived frequency response of the headphones, that is, the one that occurs as a result of the interaction of the sound emitted by the headphones with your particular auricle and ear canal, and the eardrum caught.

It is not easy to measure such an individual characteristic. Manufacturers measure their products by simulating the “average” human auditory system. For example such or such. At the same time, it should be understood that individual features (ear shape, hair, headphone pressure, tightness, and so on) can make their own adjustments to the final frequency response, sometimes very significant. However, even the result of equalization according to measurements on such “average” professional systems roughly predicts how certain headphones will sound for most people (more precisely, for the “average” person).



Figure 3. Hearing simulation systems used by manufacturers to develop and test headphones.

If their frequency response fluctuations are not too large and there are no narrow peaks and dips, then headphones, most often, can be successfully equalized to the desired frequency response, and demonstrate good sound quality. To do this, for example, you can find your model among those measured on a professional simulation of the auditory system on these resources and equalize it:

one. www.reddit.com/r/oratory1990/wiki/index/list_of_presets
2. crinacle.com/graphs

or, find headphones in a complete collection of presets based on these and other resources:

3. github.com/jaakkopasanen/AutoEq/tree/master/results

In point 3 you will also find instructions and settings for various equalizers for Windows, Android, macOS that can make your headphones sound either great, or at least better than they already are. Problems will arise if the headphones have too narrow peaks and dips, or deviations of too large amplitude, or they have too high a level of distortion – the equalizer simply cannot fix it all.

If the use of these presets does not give you the desired quality (because the measuring stand-imitation can be very different from your particular ear, especially for frequencies above 2 kHz), that is, another, individualized method based on your own perception of the frequency response. With it, you can make almost any headphones sound exactly the same to you, well, or very similar. It allows you to individually copy the sound from one headphone to another. I may describe it another time…

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *