Headphones Frequency Response: Challenges & Solutions

Boom tizz boom tizz. Measurements do tell us certain things.

 

Yeah, and that engineer guy that works in the Sony studio think the Z1R sounds good, sooooo...

Your HD650 results don't look all that correct. For one, no one should ever, ever hear them as that rolled in the bass. I mean, maybe from the worst possible, weakest amp...maybe. Because amps do matter. If anything, you should see a big, fat emphasis around 100hz with good extension to maybe 30Hz. This should be obvious compared to good speakers + a sub if needed.

The rest of the response doesn't look as weird. I could see that depending on the head, mic, and compensation. But what does bother me is the HD650 is known to have very well matched drivers. The fact your results don't is troubling.

 

Naming dropping in Hollywood gets you the job. The next step is that you actually hire the people who know what they are doing (this isn't easy and should not be discounted). The people who actually know what they are doing have their names buried in the credits. No one knows who they are.

This is what happens when you construct a curve by taking poll of random people, trained ears or not. Even if they went through the two hour JBL ear training program, they are still randos, unless they have experience actually getting their hands dirty with this stuff. When one "trained" listener prefers a 12db boost in the bass, and another 7db boost in the bass, something is very wrong. They should have failed the training. This is like giving a drivers license to someone who ignores stop signs.

A 1db/octave natural room gain or B&K target would have been more suitable than the rando derived Olive-Welti curve. Reference curves are already inherently understood by the sound engineers, especially the "Mr. Scruffy"* (equipment) engineers who have been around a long time, the guys that fix the consoles, wire up the systems, and calibrate the systems. I am not comfortable about using curves based on egalitarian principles, at least not comfortable on this site. However, JBL should certainly use this "marketing research" to appeal the the football players, wannabe gangstas. Absolutely nothing wrong with that.

Only Optimus' Prime's voice is supposed to be boomy. Sam Witwicky's is not.

 

Academic to the left: He's has lots of degrees and certifications and learned under big names.
Mr. Scruffy to the right: He started as a teen as a intern sound engineer on the movie Cleopatra starring Liz Taylor.

Who do you trust to work on the soundtrack of your film?

 

I will read this thread again later. Too much information and at work I didn't have access to some of the plots and even discussion. Here is what I came up with for the same headphones:

HD650 (@anetode)



HD600s (@ultrabike - similar cans - careful about the y-scale since it's 15 dB more zoomed out)



HD580 (@anethode - similar cans - interestingly these do roll off more like what you have - possible they need pad replacements or something else is going on)


HD580 (a friend from work - similar cans - very old set)


Some measurements I made of HD600 with a coupler, and w/o a coupler (someone in another thread asked for something like this I think)

 

@ultrabike. Your measurements are all wrong. The problem with your results is that they have not been calibrated to the average of the opinions of 11 randos: 4 of them total noob randos, and 2 others having failed the ear training for preferring +7db and +12db bass boost.

 

LOL! Like @Hands, the problem I see with the measurements at http://www.rtings.com, is that many of the measurements seem to be rolling off too much from 100 Hz down. I checked the AKG 702, AKG 701, SR-60, Beats Studio, DT-990, ... you name it. I do have some of those measurements from my POS rig as well, and I feel rtings measurements roll off way too much in the bass area.

Possible they have too much leakage in their coupler. But this could also explain the +7 dB and +12 dB bass boost in their target.

There seems to be a lot of octave averaging as well.

 

Yesh!

 

Alright. I went through your video and made some notes.

First of all, it is to be expected that different heads will have different absolute impact on the sound. This is to be expected on headphones, speakers or any real life sound application.

Think about the headphone as a transmitter, and the subject as a receiver. Think of the subject ears and other stuff as antennas. The receiver antennas will have gain across a particular set of frequencies, and if one changes the geometry of the antenna, the tuning will be different. However, we are not measuring the receiver, we are measuring the transmitter. Our goal is for the transmitter to behave the same, or as close to, the original signal. This is sound reproduction, not reception and receiver processing which should be the same for reproduced signals and live signals.

With that in mind, it is desirable to remove the effects of the ear and head. This is why there are compensation curves. To remove the gain of the outer ear of dummy heads or actual heads. Unfortunately, every ear is different and tuned differently. So the compensation curve would have to be tuned to the specific ear used for measurements.

In headphones, unlike speakers, measuring fully w/o head or ear interactions is difficult because while speakers drive a room, headphones drive a very small volume around the ear. So:

1. BOTH open and closed headphones rely on seal and in both cases, treble may be dependent on coupler.
2. You have made a great case for why the HMS is an expensive piece of shit. I agree.
3. I see still too much bass loss in your own head measurements. I suspect your compensation curve is not properly matched to your head, and you have some sort of leakage somewhere.
4. Matching left and right ear is difficult on a dummy or real life head. One of the reasons I prefer a flat coupler.
5. Positioning does indeed affect sound. But I think it's more of a mess up on direction and placement, which is pretty screwed up with headphones anyway. And indeed we have way to much receiver factor on a measurement intended for transmitter (headphone) performance only.
6. Coupling and seal affects bass response. And the effects might also be driver dependent. From what I've seen, the thinner the diaphragm, the more the seal dependence. There might be other factors as well.
7. Again, remove the ear gain (and head). The ear acts as a frequency selective antenna. Remove it from the measurement. And yes, use seal, you are not driving a room with a headphone.

BTW, great avatar! Me likes.

 

The compensation based on the Olive-Welti curve is wrong. Olive-Welti used the opinions of people who on average liked a feckton of bass and a smidgen of treble. Olive did not use not neutral speakers as a reference, which would have measured similar to a 1db/octave downward slope or B&K 1974 on a microphone.

As benevolent dictator, I decree that use of the Olive-Welti curve as a target curve on SBAF be a ban-able offense. A reference curve should rely on neutral speakers in a controlled environment, what sound engineers use in the studio. A reference curve should result in a product that reflects the intent of the sound engineers, and the artists, producers, and stakeholders who come into the studio to give their final approval.

 

I agree with what you are saying. But I see it from a different perspective.

We, as receivers, rely on the pinna to provide gain particular in frequencies around 2 kHz to about 5 kHz. From what I understand, these frequencies give speech "presence". There is also gain from about 500 Hz to 2 kHz for intelligibility. Possible the extra gain around 2-5 kHz helps on sound localization which may be important for survival.

A headphone producing "flat" frequency response at the source will experience the same pinna gain as real life sounds would. And in that sense, one could argue that the "flat" frequency response headphone will sound like real life sounds. Further deviations from "flat" by the headphone will result in perceived coloration. Exactly the same argument could be made about speakers (minus the head and cross channel which are different between speakers and headphones).

Another argument one could make is that a recording might be too bright or too colored. And that a headphone with particular characteristics would help alleviate a particular recording problem. But then, recordings do not share the same acoustic characteristics. I would argue then that it is likely that a "neutral" headphone will have a tendency to sound great with most recordings. But that would be colored by personal opinion and preferences.

Anyhow. That's how I currently understand things.

 

Interactions with cup, pads, pinna, etc. should not affect the compensation or target curve because what the dummy head microphone hears is what the dummy head microphone hears.