1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.

    Dismiss Notice

Headphones Frequency Response: Challenges & Solutions

Discussion in 'Sound Science' started by samvafaei, Jun 20, 2017.
  1. samvafaei
    Thanks for all the comments guys! :)

    The idea comes from the assumption that the perfect headphone sounds like the perfect loudspeaker in the perfect room, whatever perfect means*. And this was largely confirmed in a few studies where listener's preference was measured.

    *Harman took Floyd Toole's research for loudspeaker target response as the starting point. Toole's research wasn't really earth shattering, in the sense that it didn't contradict the already accepted consensus that the ideal loudspeaker measures flat in an anechoic room.

    You are indeed correct that some headphone companies are following the direction of Harman (or have done their own studies that reached a similar conclusion). Check out BeatsX and Solo2/3 for example, they are a massive improvement over previous Beats headphones! It's also interesting to see how some very popular older headphones end up lining up pretty well against our/Harman curve. DT990 and M50x for example.

    I'm not sure about the Golden Ear's curve, but in my personal opinion which could obviously be wrong, we have figured out the ideal headphones response up to 1KHz, give or take a dB or two for personal preference and program genre. It's the treble part that needs some work. The JBL Everest Elite 700 that is based on the Harman target measures basically flat on our website up to 1KHz. What I believe adds to the circle of contusion, is that the tracks people use as reference may not be mixed properly, or that they may not be trained audio engineers. For example, if I listen to some older 70's rock our/Harman curve would not have nearly enough bass. Just because older tracks do not enough bass! For a true control test, the reference tracks should be mixed in the same environment where the target curve was derived from.

    Research also has shown that untrained listeners do prefer the Harman-based targets, given the chance. They just may not be able to express their perception as well, and are less critical of poorly balanced sound signatures.

    As I mentioned in the video research has suggested that humans perceive more or less the same thing, especially if they are trained engineers. It seems we do have an inverse filter of our own hardware. But the thing I'm not quite sure about, is whether headphones behave differently on different people in the treble range depending the shape and size of the head/ear. That would be a different story.

    Our scoring system does give less weight to low-bass and high-treble for example, since we are less sensitive to changes in that range. But there's always room for improvement and we have already some ideas in place for future improvements.

    Doing things subjectively goes against the philosophy behind the website, partly because it's not reliable, scalable or repeatable. For example, we do score the Comfort section subjectively (since clamping force is not all that there is to comfort). We do it in a team of 3 and average our scores. But as our number of headphones grow, we find it harder and harder to do this test subjectively and have to constantly go back and double check our work. We also keep noticing our personal biases come into play. Talking yourself into giving a poorly sounding headphone a stellar comfort score is harder than you think!
    Last edited: Jun 22, 2017
  2. Strangelove424
    I totally respect that. I can understand the value in having a site online that just serves to measure and keep data on headphones. I was going to suggest a way for site visitors to organize comparison charts using their own parameters, but I see you have this feature too. I might only suggest making the Build Your Own Comparison more prominent. Headroom really drove a lot of site views based on their custom graphing capability. I think if you guys expanded on the ability to compare data, with the possibility of exporting results to a forum link, you would become a major resource for HeadFi members.
  3. samvafaei
    Thanks for your suggestion. Noted!
  4. aleksanderp
    Apology if this doesn't makes sense: if a good headphone should sound like a flat speaker in a good room, wouldn't the fr of a flat speaker measured through your specific dummy head be a better curve, rather than taking the curve from somewhere else? Since every dummy head (and every person) has a different HRTF, if you measure the speaker and headphone using the same dummy head, wouldn't the effect of pinna/ear cannal interaction etc cancel out each other?
  5. bigshot
    The thing is, I'm not sure that an anechoic chamber is what I would describe as a "good room". It may be "accurate", presenting the recording and nothing more, but I believe that the sound of the room itself is a big part of sound quality. Sometimes I think that a lot of what audiophiles strive after has more to do with accuracy than sound quality. Perhaps the best sound would be to take a totally dry recording of a violin and play it through a single full range speaker four feet above the stage of Carnegie Hall and listen to it from a seat in the middle of the 13th row. I bet that would sound better than an audiophile quality recording of a violin played back through the best headphones made.
  6. aleksanderp
    By good room I don't mean an anechoic chamber, but whatever room that sounds natural (yet another source of subjectivity :) )
  7. bigshot
    It also involves things that headphones just can't do... like reflection creating an ambience of scale and directionality. This makes me think of an interesting question for you. I see tons of discussion of "soundstage" as it relates to various brands and models of headphones. How is that quantified and measured? Distortion and frequency response are things I understand, but most of the time, I have no idea what people are talking about when they talk about the soundstage of headphones. Is there an ideal measurement for soundstage?
  8. samvafaei
    Exactly, that's why the Treble of the Harman curve didn't quite work for us and we used the Diffuse Field measurement of our HMS instead. I also did try making an in-room curve with our HMS in two different rooms, but for some reason I still preferred the sound of the diffuse field. Sean Olive has asked to use our curve in an upcoming study to see which method of making the curve (In-room or DIffuse with added bass) is more preferred. I'm not quite sure which one will come out on top because it seems headphones react quite differently in the Treble range depending on the shape and size of the individual. What I can tell for sure is that most of the well-liked headphones on the market measure very similar to the diffuse curve in the treble and not the in-room curve that I made (check these out on our website: HD 800 S, HD 600, DT 990, BackBeat Pro 2).

    Not sure what you are referring to. I agree with you that an anechoic room is not a "good room" for critical listening. If you are referring to what I say in the video, you may have misunderstood me. What I mean by a "good room" is Harman's critical listening room.

    We have our own ideas about localization effects at rtings. I haven't written a full article on what we call Soundstage and what we call Imaging yet. But I'm going to copy/paste what we already have on the website. If you click on the blue question marks in our reviews, you get a short description of what each test is.

    Soundstage: Soundstage qualities are not inherent to the audio content, the headphones have to 'create' them rather than 'reproduce' them. They determine whether the sound is perceived to be coming from inside or in front of the head, how open and spacious the soundstage is, how much the headphones acoustically interact with the environment, and how strong the phantom center is.

    So basically the combination of HRTF, room/reverb effects and crosstalk is what I put under Soundstage. In other words: the difference between a pair of loudspeakers in a room and a pair of headphones. We have 4 tests for Soundstage: PRTF, Openess, Acoustic Space Excitation, and Correlated Crosstalk. You can read more about them on the website.

    Imaging: Imaging qualities are inherent to the audio content, the headphones have to 'reproduce' them rather than 'create' them. They determine how accurately the objects are positioned in the stereo image, and how transparent the imaging is.

    In other words, what I call imaging are the localization cues that are already in the music. That is the amplitude and phase qualities in the music which result in panning and depth effects (Look up ITD and IID on Wikipedia under the Sound Localization article). Unlike Soundstage, there's not much difference between Imaging capabilities of speakers and headphones.

    Since our Gold Standard for headphones sound is the "perfect loudspeaker, in the perfect room" (whatever perfect means!), then headphones that have a speaker-like Soundstage would get full score from our Soundstage test. It's a bit more complicated than that though! :p
    Last edited: Jun 23, 2017
  9. bigshot

    I've heard a lot of cans, and I own some very good ones, but I have yet to hear any headphone that even remotely resembles the soundstage of a good speaker system in a good room. That's why I do most of my listening on my speaker rig and only use cans for portable or editing where I don't want to bother anyone else.

    You say soundstage is something created by the
    headphones themselves, not depth cues embedded in the mix... Sound in front of the head, open and spacious, interacting with the environment and phantom center... With me, every headphone has sound that goes right through my ears to the middle of my head. The phantom center is what it was recorded to be and no more. If I switch to mono, everything is smack dab in the center of my skull, no matter what headphone I use. I never hear anything in front of my head. Even with binaural recordings, which are imaging not soundstage by your definition, any sound outside my head snaps from front to back without warning. I actually find binaural recordings irritating because I have trouble controlling localizing the direction of the sound. I can perceive "open" or "closed" sound where sound is either over my ear or cupped around it, but that doesn't relate to the "best speaker in a good room" goal much. I'm not sure what you mean by interacting with the environment.

    Am I "color blind" to soundstage in headphones? I've never experienced it myself and I've always chalked it up to headphone people who just don't know what speaker soundstage actually is. But perhaps it exists, the factors that affect it can be defined and it can be measured precisely, and I am just personally unable to perceive it.

    When it comes to speaker soundstage, I am VERY good at perceiving and adjusting that. When I had a regular 2 channel stereo speaker setup, that was a breeze... just follow the basic principles of positioning everything in the room and it worked great. But with my 5.1 system it was much more complicated because I wasn't just creating a meshed sound field between two speakers, I had to do that with each speaker and all four other speakers. If my math is correct, that's 16 pairs that all had to be balanced to form a solid phantom center. The triangle theory and the placement of the rears was a good start, but because of furniture, room reflections and wall coupling, it became much more complicated. I found I had to bend rules here and there and create my own pattern of speaker placement to get it to work the way I wanted it to. My goal was to create a natural scale soundstage up front- meaning the width and height that say a jazz quartet would occupy on a stage sitting from the same distance in a club, or the width and height of an orchestra in a concert hall from the 13th row. I discovered that height was the area that required the most deviation from the accepted rules of speaker placement, and it depended a lot on the dispersement pattern of the speakers I was using. I did a lot of experimenting, but I was able to create a soundstage that felt like it was about 8 to 10 feet tall and about 20-25 feet wide. When I use recordings with natural placement of the instruments, I can close my eyes and point exactly to them in a natural scale.

    So I do understand what soundstage is and how it works on at least a basic level. I just have no clue how any of this relates to headphones. Soundstage with a good 5.1 system is like 2D with left/right and front/back. Add a set of Atmos speakers and you get 3D (left/right, front/back, up/down). Headphones to me sound like 1D- a straight line right through the middle of the head.

    I can see what you're talking about with imaging. That's the localization cues built into the music. That isn't a function of the playback equipment. It's all in the mix. Playback equipment either reproduces it or it doesn't, and that would already be covered in response, distortion and dynamics. There isn't much point even mentioning imaging in a review where you're already discussing detail under other categories.

    One more question on a different subject. I find that auditory masking as it relates to frequency response has a huge impact on sound, especially in the upper mids and treble where a lot of the detail is defined. Do you know if masking caused by frequency response spikes acts exactly the same for everyone, or does it vary from person to person? I've always wondered that. My guess is that it's pretty much the same for everyone.
    Last edited: Jun 23, 2017
  10. samvafaei
    You're not blind to Soundstage, it's just the none of the currently available headphones do all those 4 things that we test for right! But if they do it right (like what the Smyth Realiser tries to do), then except for the visceral bass effect, there shouldn't be much of a difference between what you perceive with headphones vs loudspeakers.

    Yes, Imaging is all in the mix, and it's the job of a good headphone not to screw it up. If there's a mismatch in L/R frequency, amplitude and phase responses then they are messing up the Imaging information that's in the music. And messing up the single channel phase affects the depth cues that are in the mix.
  11. bigshot
    So without a Smyth Realiser, there really is no such thing as headphone soundstage? Because of your four things you're using to define soundstage in cans, two of them are things I've never heard a headphone do at all (sound in front, variation in phantom center) and two are things that are pretty much irrelevant to soundstage in a loudspeaker rig (open/closed, interacting with environment). Why even talk about soundstage if no headphones can do it without a computer head tracking system?

    For speaker soundstage I would say there are three aspects... distance from the listening position to the speakers, meshing of the phantom center between each speaker, and height. Headphones can't do any of this.
    Last edited: Jun 23, 2017
  12. bigshot
    I've been thinking about soundstage this morning, and I've been mulling over how I would go about rating it. With speakers, the easiest way to do that is to play a totally dry (no reverb or room reflection cues) mono recording. That would put the perception of depth and the strength of the phantom center in the most clear spot, and it would eliminate the subjective bias effects of secondary depth cues baked into the mix.

    Now comes the $64,000 question... you've heard a lot of different models of cans. Have you ever heard a pair that is able to create the effect of depth, width or soundstage with a dry mono recording? Every one I've ever heard puts mono in the middle of my skull.
  13. bigshot
    I have another question... Sorry to be a pest! You say that you are using the sound of good speakers in a good room as the ideal model for what headphones should sound like. I'm curious what kind of speakers in what kind of room you are using as your model? Do you have a reference rig that you are referring to? I'd like to see photos if you do.
  14. TYATYA
    I usually dont base on FR curve found online for my listening but myself hearing and adjustment. A parametric eq is a must to do this.( I never use a graphic eq.)
    What I need is a best response from my certain equipment into my ears.
    Sound equipment makers always show nice spec of their products, example +/-0.
    3dB (for amp/dac) or +/-3dB(for headphone) in 20-20.000hz range.
    Do you belive? I don't.

    With my certain audio paths (1) : Jriver mc22 - usb link - hdvd800 amp- hd800s hp.
    (2) same but insert AK120ii between Mc22 and hdvd800.
    In both case, heard "metalic" in sound (sound like lean to "e" spelling make it unnatural) and sibilance.
    Reason is imbalance fr in a range of 5khz - 9khz.
    Found àfter using test tones :
    5khz : valley
    6khz : 0
    7khz: high peak
    8khz : deeppppp valley
    9khz : deepppp valley
    10khz: high peak again

    I get much better SQ after compensate : (FR/Q/Gain dB)

    and extra adjustment, sometime On/Off both okay : Hi shelf fillter 9khz/1.0/+2
    Last edited: Jun 25, 2017
  15. bigshot
    Graphic equalizers are better for narrow band spikes and dips. Parametric is better for overall curves. The more bands the better. I'm currently working with a five band parametric, so I have to leave the bumps alone, but thankfully there aren't a lot of them and they aren't very serious.

Share This Page