Search This Blog

Saturday, July 14, 2012

Do headphones really need a bass boost?


Die somniantes
SRSLY??? R U **NG KDDNG ME?

For God's sake, it seems Goldenears.net has officially implemented a low-frequency boost in their headphone measurement data, +9 dB for earphone, and +6 dB for headphones respectively, caused by the missing 6 dB effect. And they verified the claim with a comprehensive listening test as well. Wait a minute, the effect has been debunked already!

Previously, GE's contradiction on mannequin measurement data has been discussed. While their idea of a high-frequency attenuation is a reasonable method to minimize possible perceived tonal error for reproducing loudspeaker materials, correlating such subjectivism with mannequin data simply does not work, considering how IEC 60268-7 clearly defines the purpose of a mannequin use.

Basically with the claimed target curve, its "ideal" headphone should have a frequency response with a sound pressure attenuating as the frequency arises, as seen above. It is absolutely true that a negatively tilted room response is ideal for a hi-fidelity loudspeaker reproduction. And although headphone manufacturers know it is not a definitive solution to diminish the association principle mismatch between loudspeaker-headphone acoustics, they still tune their headphones in a similar way, so that their headphones do not sound too sharp and dry. According to Dr. Sear Olive's research from 2009:
"A flat in-room target response is clearly not the optimal target curve for room equalization. The preferred room corrections have a target response that has a smooth downward slope with increasing frequency. This tells us that listeners prefer a certain amount of natural room gain." from: http://seanolive.blogspot.com/2009/11/subjective-and-objective-evaluation-of.html

And GE's target curve closely matches the above description quite well. Well, there is always a catch:
"If the room is acoustically dead with few reflections and/or the directivity of the loudspeaker is quite high, the in-room response will represent a higher proportion of the direct sound, which should be flat. Using a target curve with large downward tilt will make the loudspeaker sound too dull...The only case where a flat in-room response would be desirable was if the room were the acoustic equivalent of an if anechoic chamber and the listener is hearing just the direct sound."
from: http://seanolive.blogspot.com/2009/11/subjective-and-objective-evaluation-of.html

Not to mention even a slight offset of a driver placement on the head causes a night and day difference in the frequency response(extremely directive), considering headphones' transient properties mostly derive from the drivers' own sole mechanical performance, it is only logical to assume the (compensated) frequency response must remain flat on headphones. Should GE's claim verified to be true, typical headphone measurement data must be as complex as in-room loudspeaker measurement data, with various reflective/reverberant characteristics skewing the direct sound. Unfortunately, that is just not the case with headphones; even cogitating a very-reverberant SONY MDR-CD3000's measurement data.


A compensation to no avail

Then how about the listening test? GE say they were able to verify their claim under a reference test condition. As mentioned before, to normal untrained listeners, the low frequency error is still prevalent due to the association principle mismatch. The bottom line is that there is absolutely no way to correctly measure the definitive number for compensation, since it is deriving from an auditory illusional phenomenon after all.


from Olive and Rumsey


One may argue that there is a ~6 dB bass level increment even in IEC 60268-13 conditions, headphones must reflect it as well. According to the above data, this argument seems sound. However, it must not be forgotten that below the Schroeder frequency, at which the room acoustics start to kick-in, there is no absolute 'response target' in regards to 'an ideal bass':


"it is necessary to diminish the importance of the curve at these low frequencies. In fact, we can ignore these effects because they will be merged with, and indeed swamped by, the standing-wave/room resonances active in this frequency region." F. Toole, "Sound Reproduction", Focal Press (2008).

Moreover, there is nothing known about GE's room condition, so it can't be assumed that the sound pressure level difference perceived between speakers and headphones was correctly measured. In other words, even if the frequency responses of both speakers and headphones were well-matched, other factors might have altered the perceived in-room response; there are just too many acoustic variables affecting this frequency region.

"there was evidence of variations in judgments due to irregularities in bass performance. Again, humans seem to be able to adapt to a certain amount of bass misbehavior, but the more extreme the problem, the more difficult it is to completely ignore...there is nothing that can be done in advance of building and setting up a room that has a high degree of certainty of achieving good bass. Probabilities may be improved in some respects of acoustical performance, but not all respects. There is no “magic bullet.” F. Toole, "Sound Reproduction", Focal Press (2008).

Various reference targets do not reflect such hump in the sub-bass region: from B&K, Olive, and Toole.



And how about the discrepancy with earphones? It has been found that the eardrum over-excursion caused by a pressurized ear canal increases the lower frequency response. Since this type of excursion error causes a boost of ~15 dB @ 20 Hz, GE's "test-confirmed" compensation figure of as much as 9 dB can be explained. As the perceived bass of earphones mostly depends on the interpersonal variance of the eardrum's compliance & the amount of air pressure built-up in the ear canal, GE's approach can be interpreted as a case of haste & immature generalization.


Do headphones really need a bass boost?

With all the data I've discussed put into consideration, I would say the answer is definitely 'NO'. A simple amplification in the sound pressure level may reduce the error for normal listeners, but does not solve the source of the problem. Also, it is not simply possible to turn headphones into loudspeakers either, unless an accurate binaural synthesis is implemented.

I think, ironically, headphone enthusiasts' best bet lies in loudspeaker acoustics, by extending the low frequency bandwidth:
"The average bass extension - the low-cutoff frequency - progressively decreases as the fidelity rating increases. The listeners liked low bass - not more bass, in the sense that it is boosted, but bass extended to lower frequencies." F. Toole, "Sound Reproduction", Focal Press (2008).

Then why flogging a dead horse?

I've been wondering, why such objectivism-oriented websites continue to implement rather 'unobjective' test methodologies, when the resulting data can seriously mislead the end-users? Such target curve turns typical headphones' frequency response more favorably to the manufacturers, who simply rely on a pressure level compensation due to the lack of research on their ends. As long as the graph looks good, the product should sell. Since commercial-oriented websites depend on sponsorships, their relationship with headphone manufaturers can be easily understood; They do this for living, in the end. (wait, that is why I can't get a single test sample from any manufacturers! LOL)

3 comments:

  1. 별 의사소통도 없었는데 읽어본 내용들이 거의 비슷해보이는 이유는 뭘까요 ㅋㅋ 역시 닥치는대로 뒤지면 읽게 될 내용은 거의 비슷한 듯 함다. (...)

    그나저나

    http://www.aes.org/e-lib/browse.cfm?elib=15253
    http://mirlab.org/conference_papers/International_Conference/ICASSP%202011/pdfs/0000289.pdf

    요 논문은 당연히 읽어보셨으리라 믿습니다. RECD 문제를 종결하는 아주 간단명료한 설명인듯 한 느낌이... ㅎㅎ

    ReplyDelete
  2. 솔직히 골든이어스의 청취 실험은 매우 미심쩍은 부분이 있고-타겟 커브 산출씩이나 되는 일인데 너무 실험 설계를 약하게-_-한 면이 강함다-기껏해야 받아들일 수 있는 것은 룸 게인(room gain)에 의한 저역 증가분인데, 이 정도는 표준 룸 정도의 규모라면 어느 정도 합리적이지 않나 싶긴 합니다만, 좀 더 조심스럽게 접근해야 할 것 같기는 합니다. (다수의 룸 응답 측정 및 적절한 설계의 리스닝 테스트 수행)

    또한 이어폰에서 9dB씩이나 증가시킨 것은 언급하신 논문에 나와 있는대로 등골근 반사에 의한 미드레인지 이상 대역의 다이나믹 레인지 저하를 초래함으로써, 그리 바람직한 해결책은 아니라고 봅니다. 더불어 골귀 자체 청취 실험 마저도 너무 애매한데 말이죠. (사실 저런 실험은 학부생 리포트 용으로 쓰기에도 좀 부끄럽슴다;;)

    여하튼 적어도 라우드니스만큼이라도 과연 시간 응답의 영향을 얼마나 받는지 싶어서 계속 파고 있습니다만, 찾고 있던 내용과 거의 근접한 요 논문

    Loudness calculation for individual acoustical objects within complex temporally variable sounds
    http://www.aes.org/e-lib/browse.cfm?elib=14624

    마저도 꽤 애매한 것 같습니다. (아시는 내용이 있다면 제보를 부탁드립니다. ㅎㅎ)


    아무튼 애매한 부분에 대해서는 좀 확실히 유보하고 신중한 자세를 보이는 게 학술적인 걸 자시고 떠나서 '프로'로서의 태도가 아닌가 싶은데, 확실히 근래 골귀의 행보는 좀 무책임한 면이 있지 않나 싶네요.


    덧. 골귀 사무실의 룸 컨디션은 요 링크에 나와 있긴 합니다.

    http://goldenears.net/board/index.php?mid=ST_KB_byGE&document_srl=2056288

    얼핏보면 125Hz 이하의 저역이 플랫한 듯 보입니다만, 사실 전반적으로 좀 강조되어 있지 않나 싶슴다. 그리고 전달함수의 주파수 응답(magnitude) 측면에서 생각해봐도 리스닝 룸의 청취각 60도 조건과 DFE를 그대로 비교하는 건 결과를 호도할 측면이 강하죠.

    따라서 (솔직히 영 못미더운 실험이긴 한데 그래도) 골귀의 실험 결과 요인을 해석해보자면 스피커/룸 시스템과 이어폰/헤드폰이란 재생 시스템 차이말고도

    1) 적절하게 컨트롤되지 못한 스피커/룸 응답
    2) 적합하게 등화되지 않은 전달함수

    라는 요인이 있기 때문에 당연히 재생 시스템 차이로 인한 게 가장 주요한 원인이라고 단정지을 수는 없겠슴다.

    특히 아시겠지만, 40~80도 정도의 청취각에서 HRTF를 측정하게 되면 1~4kHz 인근과 9kHz 인근에 딥이 생기는만큼 상대적으로 저역이 부각되는 것처럼 느낄 가능성은 매우 큽니다.

    그런 면에서 룸 코렉션 시스템의 도입, FIR 필터를 사용한 등화가 추가적으로 실험에 수반되어야 하지 않나 싶군영. 솔직히 그렇게 했을 때, 잘 댐핑된 룸에서 과연 얼마나 차이가 날지 잘 모르겠습니다.

    ReplyDelete
    Replies
    1. 헉, 당근님이 리플을 달아주셔서 영광입니다 굽신굽신 (응?)

      RECD에 관련해서는 전기와 음향의 기본 물리적 상관 관계만 알고 있어도 쉽게 풀리는 문제입니다. 그에 대한 원리를 논파하다가 축출(?)되었다는 점이 참으로 촌극이 아닐 수 없지요.

      말씀 해 주신 내용 전부 동의합니다만, 저역 증가분은 본문에서 적었듯이 단순히 룸특성을 따라가는 것이 아니라, 특정 헤드폰이 가지는 저역의 모자른 부분을 보충하여 대역폭을 늘리는 성격을 띈다면 의미가 있다고 봅니다. 실제로 ER4만 하더라도 초저역이 아주 플랫하지는 않으니까요. 일반 청취자가 가지는 6dB 상실 특성 보상은 청취자 간 편차가 상당하므로 개개인의 선택에 맡김이 옳다고 봅니다.

      링크 해 주신 Loudness calculation for individual acoustical objects within complex temporally variable sounds에서는, 멜로디와 리듬 파트를 각기 다른 acoustic object로 간주 하였을 때, 2~3 kHz 대역에서 3 dB 정도 차이가 발생한다고 합니다.

      그리고 http://quod.lib.umich.edu/cgi/p/pod/dod-idx?c=icmc;idno=bbp2372.1994.013 에서는, 더 큰 공간이 같은 loudness intensity에서 시간축 특성에 의해 더 크게 들린다지요.

      그리고 이는 관련 서적 http://books.google.com/books?id=eGcfn9ddRhcC 에서도 확인할 수 있는 부분입니다.

      정말 제대로 청취 실험을 한다면, 이어폰-헤드폰의 주파수 특성(1/20 oct 이상의 해상도로 매칭) 뿐만 아니라 시간축 에서 기인하는 지연 특성까지 올바르게 매칭해야죠. 현재 GE의 objectivism을 대하는 안일한 태도는 오만에 가깝지 않나 생각 해 봅니다. 9 dB라는 무서운 에너지량(?)까지 나온 마당에 이제 GE 쪽엔 신경을 끄고 싶긴 합니다만, 저런 일반화의 오류가 공론화되는 것을 좌시할 수만도 없는 입장입니다;;

      Delete