AES Headphone Technology Conference: Stuff Possibly Useful for Headphone Enthusiasts

Most papers presented were pretty esoteric stuff—interesting, for sure, but not necessarily practically informative for enthusiasts. I found the following papers had information that was more directly relevant for enthusiasts.

Headset EMI Verification and Measurement Method

Electro Magnetic Interference (EMI) performance is, other than electro acoustic performance, one of the most important aspects of designing headsets specifically for cellular phones and other portable handset applications. A system to measure the EMI performance of headsets was studied, built and tested with almost two dozen headsets. The results of tests show that the differences of cabling methods greatly influenced both EMI and crosstalk performance. The impact of different cable shielding conditions was simulated and is documented herein

Here's one for all you DIYers out there who like to re-cable your headphones. In this case, we're talking about headphones with a mic—a headset. The paper is a deep dive into the various cabling configurations and their various effectiveness in rejecting electromagnetic interference (EMI)—focussing in particular on the grounding and shielding scheme.

The not necessarily obvious problem with wiring up headphones—particularly the shields—is that in a portable audio device there is no earth ground connection, and in a floating system the problem of shielding becomes a bit pesky. The current paper goes into great lengths about the nature of this problem, creating a SPICE model (Simulation Program with Integrated Circuit Emphasis; a program that can model the performance of an electrical circuit) based on the electrical aspects of the Human Body Model and the circuitry of a headset and handset (headphone and portable player). (See image at top of page.)


Also developed for the research was a test fixture that allowed headsets (earbud and IEM types) to be immersed and accurately measure in a controlled field of various types of electro-magnetic interference.

A total of 21 popular commercially available headsets were tested. Looking at the grounding schematic at top of this section: Eight of the headsets tested were in configuration A. In five of those eight, a simple ground conductor is used and the cables had no shields! Three headsets were in configuration B. Five headsets in configuration C. And five headphones were in configuration D, though a mix of shielded and unshielded combinations existed.

A number of EMI signals were used during the test, but it should be noted that radio frequency interference was not a focus of the test, which was more concerned with things like switching power supply noise, AC 60Hz interference, and parasitically capacitive interactions between the human body and the devices. The paper also addresses some issues concerning variance of output and input impedance measures of headset transducers, player headphone amp, and microphone input impedance's. Two case of particular difficulty were also considered: effects of using the headset with a switching power supply attached for charging; and players in which the headset is used as an FM antenna.

Let's cut to the chase. First, that fact that 5 of the 21 headsets tested had no shielding at all seemed to leave the researchers shaking his head in understated disgust in his summary comment, "This is beyond the norm of standard audio practice." Not shielding the mic cable!? Ridiculous!

During the research process a baseline pass/fail mask (a target line) for EMI rejection was developed. Only 9 of the 21 passed; worst case failures had interference measures 30dB higher than the mask. Best case products beat the mask by 10dB.

It's going to come as no surprise to DIY enthusiasts that the [D] cable grounding circuit diagram that independently runs all grounding and shielding to a common connection at the 3.5mm ground connection was the best performer. What was a bit of a surprise is that the research found that in the [D] configuration, even a 3mm gap in shielding will move a device from near ideal performance up to the "just acceptable" target EMI rejection mask.

Your take-away as a DIY re-cabler when working on headphones with a mic? Make damn sure your shields are continuous and cover as much as possible...even the mic circuit board if possible.

The Preferred Low Frequency Response of In-Ear Headphones

A series of controlled listening tests were conducted to determine the preferred low frequency response of in-ear (IE) headphones. Using a method of adjustment ten trained listeners adjusted the bass level and frequency of a 2nd order low shelving filter applied to a high quality IE headphone that was calibrated to the preferred target response of a circumaural headphone [5]. The adjustments were done for three different music programs, and repeated with and without loudness normalization and control of leakage effects. The influence of program, individual taste, and loudness normalization and leakage effects on preferred low frequency response are presented and discussed.

In this paper, Todd Welti continues the Harman research developing target response curve for headphones, extending the research done on over-ear types with subjective tests, to now develop a target response curve for IEMs—in particular, the preferred amount and cut-off frequency for bass emphasis.

The research was designed to answer four questions:

  • What is the preferred low frequency target response of IE headphones?
  • Is the preferred target response the same for IE headphones as an OE (over-ear) headphone?
  • What influence does loudness normalization have on the preferred low frequency response?
  • What influence does leakage have on the preferred low frequency response?

The last two items didn't seem to have much of an effect. Loudness normalization is accomplished by a system that automatically readjusts the overall loudness of the signal as subjects increased and decreased bass response. It was shown to have very little effect.

Regarding bass leakage, in the first wo test runs, subjects inserted headphones without any monitoring of whether or not they achieved a good seal. Researches expected to see increased bass levels desired in this condition as some subjects would likely not get a good seal at times and would dial in more bass as a result. They did find subject desired more bass overall in this condition, but the result was smaller than expected.

It is posited in the paper that these two finding may be small because trained listeners were used in the study, and it's possible that they are simply more conscientious inserting the IEMs and are able to listen through changes in overall level during bass adjustment for tonal correctness.

The research did show a clear results for the first two questions. In the plot at the top of this section we can see the results for two studies of preferred response for over-ear headphones. The red dotted line is a large study with many diverse participants, largely untrained. The black line is a smaller study using trained participants, five of whom also participated in the present study. Welti suggests in this paper that readers should likely focus on the difference between the black line (preferred over-ear target) with the blue line (the preferred IEM target developed in this paper), as participants in both studies were of similar listening skills and demographic.

Essentially what they found is that listeners preferred about 5dB more bass in IEMs than OE headphones, and the shelving frequency was about the same. Makes sense to me. I have a suspicion that even though it's much less than with speakers, some bone conducted information is available with full size headphones that simply doesn't exist with IEMs. I'm curious to see if they work on an on-ear target response and find a similar need for increased bass boost—on-ear headphones won't deliver much bone conducted info, I would suspect.

So, with IEMs, people want about a 10dB boost in the lows below 150Hz. The next question becomes, "What are they getting now?" The answer is pretty pathetic.


The above plot shows the IEM target response curve developed in the present research against 15 popular commercially available IEMs costing from $10 to $400. As you can see, almost all of them start to rise in response somewhere around 800Hz, and most have bass elevated in excess of the boost described in this paper. If you look through InnerFidelity's IEM measurements in the AllGraphs.pdf you'll see this shape over and over and over again!

My assumption is that it's simply what you get when you slap a dynamic driver into a generic IEM shell (with a punk Hello Kitty stamped on it) and shove it into some kid's ear canal. It's analogous to putting speakers in a room and getting a low frequency gain once the sound's wavelength becomes longer than the room's dimensions and you begin to start pressurizing the room with the speakers at very low frequencies. Well, the ear is much smaller than a room, and therefor you get this pressurization up to higher frequencies. Makers don't seem to care much and probably just slap the things together and call it good without concern for this over-pressurized bass.

I've had this thought for quite a while now, and while I don't think I buy into the way it's being described, it may be that the ADEL (Ambrose Diaphonic Ear Lens) may be an avenue forward to reduce this so called "pneumatic pressure." 1964 Ears had added this technology to their IEMs, and now appears to have moved on to a second generation of something similar to it with their Apex technology.

Bottom line: as stated by Harman researchers in the paper, "Listeners reported that vocals and other instrument became muddy, muffled and colored when the low shelving filter was set to frequencies above 200-300 Hz. This suggests that there is much room for improvement in how IE headphone manufacturers optimize the sound quality of their products."

Yeah, IEMs are generally pretty sad and it would be nice to see better tonal profiles among them. I'll be looking closely into the Etymotic ER4XS and PSB M4U 4 soon in that regard.

Refinements in the Electroacoustic Testing of Headphones

A number of measurement methods, processing techniques, and data presentation guidelines for improving the standardized testing of headphones are presented. Selected measurements and specifications from the published standards are reviewed. Areas for appropriate modifications, simplifications, and improvements are identified and explained. Relevant new metrics are also introduced. The rationale for each of these changes is described and examples of the new tests are shown.

This paper was presented by Christopher Struck of CJS Labs, a headphone measurement consultant to industry. He, like I (but with much more experience and authority), have numerous problems with the current IEC 60268-7 standard for testing headphones. In the present paper he outlines some of the areas the he would like to see changed in the current specification. More importantly for InnerFidelity readers, this paper offers a few insights for enthusiasts with measurement rigs and how they might produce more informative measurements. These are a few of his relevant recommendations:

Source Impedance - The IEC spec calls out that headphone measurements should be made with an amplifier that has a 120 Ohm output impedance. Any experienced headphone enthusiast will tell you that under most circumstances the output impedance of the amp should be at least eight times lower than the headphone impedance. A "good" amp should be under one Ohm; most solid-state headphone amps will be under 2 Ohms; and even iPhones and cheap portable players will usually be under 10 Ohms. The paper recommends the spec should include at least a check of measurements taken with an amp of low impedance output to ensure measurements are reliable with most consumer devices.

Sensitivity - Another of my pet peeves. It's expressed in a rather more complex way in the paper, but the main gist is this: IEC spec cause sensitivity/efficiency measurements to use measures of power, i.e., dB/mWatt. It's simply much more reasonable to measure headphone efficiency in dB/Volt—given how easy it is to drive headphones and the lack of need to worry about power limitations of driving amplifiers.

Look, the knob on your headphone amp is changing the voltage at the output. What you want from a manufacturer is a number that's going to let you know how high you're going to have to turn up that knob. Or, you want to know that if the dB/Volt is too low, you're not going to be able to drive it with your iPhone. A dB/Volt spec is just way more useful.

Positioning Error and Test Repeatability - Chris here advises the IEC to include a requirements for frequency response measurements that headphones should be measured at least five times and be removed and re-fitted before each measurement. I totally agree; InnerFidleity measurements are an average of five placements in slightly diferent positions. This requirement should be in the IEC spec, and I would suggest enthusiasts at least measure cans a few times to see if they vary much before taking a representative plot.

Swept Two-tone Distortion Measurements - Chris has found in his work as a design consultant that swept two-tone intermodulation distortion measurements end up being much more telling of headphone acoustic performance and problems than single tone or fixed frequency two-tone measurements. The current IEC spec calls for intermodulation distortion measurements with a fixed two-tone method.


The paper recommends intermodulation distortion be measured with a fixed frequency tone at 70Hz, and a 12 dB lower in level swept tone beginning at 200Hz plotting +/-2nd and +/-3rd order distortion products.


He also describes his preferred difference frequency distortion measurement, which is in the spec. Two tones at equal level separated by 80Hz are swept and resulting distortion products plotted.

In a stroke of good luck, I got to enjoy some time at lunch with Chris and chat about my recent adventures to Harman Labs to measure my head and try to develop a headphone compensation curve. It was a fascinating conversation, but one fraught with much more difficulty than I had expected. Which turns out to be the subject of my "Conference Highlight Paper" tomorrow.

See you then.

tony's picture

the more you don't know.

I've been reading along, feeling like I'm getting to the deep end of the pool, now getting the feeling that the pool keeps getting deeper with the idea that we're not gonna find out just how deep it's getting.

Which has me looking to Noam Chomsky and the digging into human cognitive abilities.

I'm getting lost.

I better get out of the pool for a while, I'm past hoping for clarity.

Tony in Michigan

lmsinn's picture

Hi Tyll, you are correct about less bone conduction with in-ears (IE) compared to circumaural (OE), however there is more bone conduction with on-ear (supraaural) compared to IEs. The more surface area contact with the head/skull, the more bone conduction. I am in a doctorate program for audiology and since we test each ear separately, we care a lot about “crossover,” or the amount of sound that travels from one side of the head/ear to the opposite side/cochlea. If we suspect that the sound in the “test” ear can be heard by the opposite ear due to bone conduction, we “mask it out” by presenting narrow band noise into the non-test ear. Our masking norms are dependent on transducer type. If you use IEs, (or IP which is our acronym) there is much less crossover compared to on-ears, and less with on-ears compared to OEs. Granted, different studies in the audiology world resulted in slightly different norms, but overall you absolutely can assume less crossover due to bone conduction, and of course, the amount is frequency dependent but you already knew that : )