Skullcandy's Director of Electrical & Acoustical Engineering, Dr. Tetsuro Oishi Visits InnerFidelity! Page 2

Tet Teaches Me About Headphone Measurements
Most interesting of all, to me, was our dialog about headphone measurements and what they mean. First, it's terribly complicated with many variables coming into play, so looking at a set of measurements will never let you diagnose exactly what's going on. But there are a few things that tend to show up on the graphs, which, depending on where they are, can indicate the likelihood of particular problems.

120305_blog_tetsvisit_graph_frcompareMain Spring - Well ... that's my word for it. The "springiness" of the earcushions and the enclosed air volume allows for a low frequency resonance to develop. Think of it like the earphones bouncing gently on and off your ears. This will typically appear as some frequency response feature between 50Hz and 150Hz. In the Frequency plots to the right, the initial feature in the lows of each headphone is probably this "main spring" effect. You can see in the incredibly well engineered Sennheiser HD 800 the merest blip at 60Hz; all heaven and earth moves between 60Hz and 80Hz on the DT48.

Poor Ear-pad Seal - If the ear-pad is not sealing well, you will see a second-order (dramatic) drop-off in the lows. In the headphones to the right, the AKG K272 is likely not sealing properly. I'll probably have to go through my measurements and identify poorly sealing cans for re-measurement using my improved headphone positioning methods.

Voice Coil Wobble - If the magnetic field strength or voice coil weight isn't perfectly distributed around the circle of the voice coil, it will be accelerated with more force on one side than another. This might be from misaligned pole pieces or an eccentric voice coil, for example. At some frequency, this wobble will hit a resonance and create a blip in the frequency or impedance response, typically somewhere between 300Hz and 1kHz. In the frequency response plots to the right, the features above 300Hz on the T5p and K272 are likely candidates. Tet said coil wobble features are typically high-Q--meaning they are typically quite spiky looking.

Internal Driver Resonances - Behind the diaphragm is the magnet assembly. Small volumes of air can be partially trapped in this space creating the opportunity for resonance. Because these volumes are very small, the resonant frequency is quite high--typically between 2kHz and 8kHz. Headphone frequency response tends to become noisy in this range, so it may be difficult to separate internal driver resonances from resonances elsewhere in the headphone. But because these resonances exert a direct load on the driver diaphragm, and therefore voice coil, the reactive load of these resonances will readily appear on the impedance response plot. It is likely the fine features above 2kHz in the impedance response plots to the right are due to internal driver resonances.

There are, of course, many other characteristics of headphones that may manifest as wiggles in plots. The magnitude and combinations are endless, and as Tet looked at the measurements there was always great caution in his judgments. He'd always suggest some sort of test we could try to affect the plotted measurement to identify its origin. That's his job. Unfortunately, it's not mine, and while I'll continue to learn, my time with Tet tells me I'll really never be able to diagnose accurately what's going on in a particular pair of headphones by measurements alone.

There is good news though: while we can't know for certain the cause of particular bumps and wiggles, we can say for certain that less is more. A well designed headphone that is well manufactured with quality parts will have fewer of these bumps and wiggles than a poorly-executed headphone of the same type. It's important to note here that some headphone types will naturally have more features in their plots than others. A well designed circumaural, open headphone will always be flatter than an equally well designed supra-aural sealed headphone.

Thanks for the Visit, Tet!
Funny story: Tet and I are going through the book of graphs just for giggles and to point stuff out, and we stumble upon the Beyerdynamic DT48. He says, "Oh stop! Look at that: that has got to be a sealed circumaural headphone with a very small and tightly sealed chamber behind the driver."

"How do you know," I wondered?

"Well, it's virtually a text book plot. The huge main spring dip and peak at 60Hz and 80Hz means its a tightly sealed enclosure between the driver and the side of the head, and that it has very springy cushions. And the fact that the bass recovers flat, but very low in amplitude means the enclosure behind the driver is tightly sealed and small."


I am familiar with the headphones, and Googled a picture of the DT48. Of course, they fit Tet's description to a T.

Man, I am so glad to have had the opportunity to talk to a real expert on headphone design and learn this stuff. I'll no doubt be rummaging through all my measurements to start a list of plots that I think I can improve on; and I'll also be looking at data in the future with wiser eyes. I'm sure all you headphone geeks out there will feel the same after reading this far.

Thanks so much, Tet!


RPGWiZaRD's picture

No but seriously, I really like that Skullcandy finally tries to improve the audio quality of their headphones and hired Tets, I have much respect for this person. I may even go out and buy a Hesh 2 haha despite my basshead needs is wonderfully covered in M-Audio Q40 headphones (A seriously overlooked headphone. One person commented as the bass beating Ultrasone PRO900 in quality with similar quantity and having way better mids and having less stringent and neutral highs, not bad for 3.5x less cost).

plin's picture

Thank you both for sharing your knowledge with us, very interesting indeed!
I am curious about the flat impedance curve in orthodynamic headphones. Does it mean that there is no internal drive resonances, in the whole of frequency spectrum?

Tyll Hertsens's picture
That's a very interesting question, one that I have been thinking about since it's related to why the impedance response remains apparently wholly resistive no matter what. I don't know the answer however. I'll report if/when I learn more.
John Grandberg's picture
...and for many young folks Skullcandy is likely the place. So increasing the quality of those entry level options can only be a good thing. Hiring Tet says a lot about their intentions - they really don't NEED to do anything to improve their sound, because people will buy no matter what. But they seem genuinely interested in stepping up their game. I'm impressed.
arnaud's picture

Tyll, you get to hang around cool people, this was a neat article. The first discussion about mass / air spring resonance is particularly interesting because we were discussing this on head-fi in regards to inconsistencies between different lcd3 responses. My take on this was it's not that simple to diagnose because I assume it depends on several variables:
1. if the baffled plate is sealed or not
2. If the headphone is closed or open back
2. If the earpead is impervious (leather) or porous type
3. If the diaphragm is suspended (electrodynamic) with very low mechanical resonance or a tensioned membrane (electrostats, orthos to some extent it would seem).

Anyhow, Tets discussion toward the end about the dt48 got me curious and I wish I can find time to play around with Vibro-Acoustics simulation tool to get some more insight about what matters and what doesn't... Thanks for sharing, the geek in me much appreciated ;).

hydrocarbon's picture

Good technical content there, thank you! It's always worthwhile to to talk shop with someone who's a genuine expert in his field.

Regarding the DT-48s, "tightly sealed" is an understatement. They are airtight like no other headphone (or industrial hearing protection headset) I've ever used. They're essentially suction cups, particularly right getting a haircut. In fact, I suspect that taking them off too abruptly could damage eardrums!

(Frankly, the reason I got them is because I like German-made metal things. The fact that they sound quite good to me with certain types of music is just a bonus; I was mainly concerned with their build quality, history and aesthetics.)

miceblue's picture

I am quite pleasantly surprised with Skullcandy's recent headphones. I owned the Hesh (1st gen) until they broke after 6 months of use. Perhaps I'll give the Hesh 2's another chance. The graphs look really impressive when compared against HeadRoom's Hesh (1st gen) measurements.

zobel's picture


Looks like most headphones you measure show a dip in sound pressure level at around 6 or 7KHz.

The wavelength of sound at 7KHz is 50mm. One half wavelength is 25mm.
The ear canal on average is 25mm. Could this cause out of phase cancellation from the reflection off the stiff microphone diaphragm?

As you know,the human ear isn't really the same as the test head's, since the eardrum, middle, and inner ear have different acoustic compliance and resonant frequency than the microphone at the end of the dummy ear canal.

Do headphones have to be made non-flat in response above 1KHz to correct for the resonances created by closing off (more or less depending on type of headphone) the ear canal with the driver?

I'm wondering what the high frequency portions of your charts show. I was wondering if Tet and you talked about about your raw data and the correction curve you have been using?

It kind of looks like a 10dB/octave roll off from 2 to 6 or 7KHz, combined with a steep 20dB gain from 6 or 7KHz to around 10KHz, followed by a roll off of about 25dB from 10KHz to 20KHz indicates a flat perceived frequency response.

Very interesting! Please keep talking to us about this stuff.

Tyll Hertsens's picture
Actually, the head is designed to have the same ear canal resonances as the average human head. Also, the acousic impedance/compliance of the microphone diaphragms at the end of the ear canal are designed to mimic the human ear. So the head and it's hearing system is designed to achieve very near the average human in as many was as possible. Not perfect of course, and it's only the average of humans, so won't be exactly like yours or mine.

Most of the spikes and valleys above 6-8kHz, or so, are not from the driver, but are the result of various resonances of the ear canal, concha cup, and size and shape of the enclosed volume between the driver and ear, and behind the driver.

Tet and I did talk about the HRTF compensation used, and I think he agreed that the Independant of Direction Curve was a good one to use.

I think the peak at 8-10kHz is problematic, but I do think it's likely optimal for there to be a slight bump there on the graphs I make. The leading edge of the 300Hz squarewave having some, but not too much, overshoot is related, and there should be a slight overshoot for an ideal percieved response ... still thinking about that one, though.

yuriv's picture

If the head could talk, he (she?) would complain about that. You can see it in the graphs of so many cans, and it's also in the raw measured response. So it's not a result of the ID HRTF compensation. It's really there, and it's probably the result of the shape of the dummy's ears. In fact, almost none of the IEM measurements show that 5.5 kHz null.

Perhaps the reflected sound arrives at the opening at an odd multiple of half a wavelength later after the direct sound. If you could post photos of the head simulator's ears next to a ruler, that would be instructive. So would headphone measurements with different-shaped ears.

There is so much disagreement about which headphones sound correct, and--average ears on the head simulator or not--this is a big factor. For example, I almost never hear a null between 5-6 kHz. After playing with Sinegen with many, many cans, I noticed that for me, the null happens just under 4 kHz. I never hear that with IEMs or loudspeakers.

Jazz Casual's picture

Not playing devil's advocate here but given the variables at play, is it possible that we're attaching too much weight to headphone measurements as a reliable predictor for what we're supposed to be hearing?

Tyll Hertsens's picture
I guess that depends on how much weight you're actually putting on it. I certainly don't believe you can tell exactly what kind of listening experience you'll have from measurements, but I do think one can tell, in gross terms, wether it's got too little or too much bass; whether the headphones are carefully designed so as to have an even response; and whether they'll be "bity" sounding. But only when these features are strong and readily apparent.

However, once it appears the headphones are within a certain range, you simply must do a listening test to tell what they really sound like --- and, of course, then matters of individual ear variations and taste come into play.

ultrabike's picture

Amen Tyll

RPGWiZaRD's picture

One often overlooked thing is that only a very small portion of people are experienced enough and so deeply involved in the hobby for starters to be able to translate and separate their own personal taste from "objective evaluation" of a headphone. Rarely people want a perfectly flat looking response in a frequency response graph but many people think that's what to strive for while disregarding your own personal taste of how it SHOULD sound according to your ears/preferences and therefore some people may say one headphone has recessed highs when it doesn't just because your preference wants a 3dB boost...

Jazz Casual's picture

I make a point of studying headphone frequency response measurements. I welcome an objective point of reference in such an inherently subjective hobby. Frequency response graphs can't plot the intangibles of what makes a headphone appealing or unpleasant to my ears, but they have helped me to make sense of what I'm hearing across the frequency spectrum. I find them useful for comparative purposes and when considering headphones that I can't audition prior to buying.

Tyll Hertsens's picture
Glad to be of service. :)

(Jots down yet another note to measure PS1000.)

Jazz Casual's picture

That would be great thanks Tyll. :)

The Monkey's picture
Let's get Tet to some meets!
Tyll Hertsens's picture
I'll tell him about the NY meet.
mrspeakers's picture

Thanks for another interesting post.

It'd be fun if Tet came to New York.