May 2014

Brent ButterworthThe new $1099 USD Oppo PM-1 planar-magnetic headphones have headphone enthusiasts more excited than they’ve been in months. My test sample just arrived, and from what I gather we’ll be reading a lot of reviews of the PM-1 in the coming weeks. We’ll also be seeing lab measurements of the PM-1, and reading debates about what those measurements mean.

I thought this would be a great time to discuss how and why interpreting headphone measurements isn’t like interpreting speaker measurements.

Thanks mainly to work started at the Canadian National Research Council (where the SoundStage! Network does its speaker measurements) and continued at Harman International, we know a lot about what speaker measurements mean. Even most casual audio enthusiasts understand that a relatively flat on-axis response in a loudspeaker is a good thing. The equipment and techniques for speaker measurement are well standardized, so any two competent engineers are likely to get similar results, even if one’s working in an anechoic chamber and the other’s working in a basement.

With headphones, it’s different. The measurement standards that exist are out of date and flawed in many ways, a situation that requires technicians to develop their own methods. The equipment is well standardized, but due to the high cost of headphone measurement gear, some reviewers and manufacturers use improvised setups with, say, a microphone flush-mounted in a plate instead of one of the ear/cheek or head/torso simulators made by Brüel & Kjær or G.R.A.S. Sound & Vibration.

There are also multiple ways of processing headphone frequency-response measurements in an attempt to make them correlate better with what we hear. These processing methods take into account the acoustical effects of the ear canal, the pinna, and the head. You can calibrate your measurements for drum reference point (DRP), which is the response right at the eardrum; ear entrance point (EEP), which is the response at the entrance to the ear canal; or ear reference point (ERP), the response at a point in space roughly defined by the intersection of the axis of your ear canal and your palm when your palm’s pressed against your ear.

Some technicians who measure headphones employ a diffuse-field compensation curve, meant to counter the diffuse-field equalization used in many headphones. These headphones typically have, for example, a response peak of about +15dB around 3kHz, so if you process the measurements with a reverse compensation, that +15dB peak will now appear flat -- which, in theory, is the way your ear perceives it. However, due in part to recent research by Sean Olive and Todd Welti at Harman International, the validity of the diffuse-field equalization curve has come into question.

Comparison chartComparison chart of three headphone models

The ability to interpret a headphone frequency-response curve requires that you know what compensation curve was applied to the measurement. At SoundStage! we tell you exactly how we do our headphone measurements (which, by the way, is similar to the way they do it at Harman). If you read headphone measurements that don’t include an explanation of how they were done, you have little idea what you’re looking at.

Then we have the problem that there is currently no standardized average response curve for the ear canal at frequencies above 10kHz. That’s partly because most measurement gear used for headphones was actually designed primarily for use with hearing aids, in which such high frequencies are not a concern. It’s also because ear canal response varies so dramatically above 10kHz that we may never develop a useful model of average response.

Even if you know how the measurements were done, interpreting the curves remains a combination of art and science. During the last few years, I’ve been fortunate enough to be able to conduct many headphone tests that incorporated my own measurements as well as the subjective impressions of several experienced listeners. I was sometimes shocked to see the measured response of the headphones disagree with the listeners’ subjective perceptions -- and even with my own subjective impressions.

I was also dismayed to hear how much individual listeners’ impressions varied. With speakers, experienced listeners tend to agree.

Olive, Welti, PSB’s Paul Barton, and other researchers are working to better correlate headphone measurements with subjective data. At least until more such research is completed, I’m reluctant to judge a headphone’s quality by its frequency-response measurements.

However, after a couple of years of comparing measurements with subjective data, I have a few general rules of thumb. These usually hold true for my measurements, and I suspect they’re probably applicable to most headphone measurements no matter how they were conducted or processed.

Audeze LCD-31) If a headphone shows a strong roll-off in the treble between about 2 and 10kHz, with the response decreasing by -6dB/octave or more, it’s almost certainly going to sound dull. (I know, this seems like a no-brainer, but in headphone measurements, no-brainers currently don’t exist.)

2) Most headphones show a bump in the low-frequency response, but if that bump’s centered up around 200Hz instead of somewhere down between 40 to 80Hz, the headphone’s likely to have bloated, poorly defined bass -- which may also have the psychoacoustic effect of dulling the treble.

3) Headphones that show wild response swings in the midrange, between about 400Hz and 2kHz, usually sound bad. (How did I determine this? Because this is how a lot of those hip-hop headphones measure, and most experienced listeners don’t like those.)

4) The headphones that I’ve found to be the most popular among listeners show a flat response from about 100Hz to 1.5kHz, with a shallow bass roll-off below 100Hz, and somewhat elevated response (say, averaging +6dB relative to the response at 1kHz) between 2 and 10kHz. This is the response I’ve measured from most of the planar-magnetic headphones I’ve tested, including the Audeze LCD-3 (shown right), which is probably as universally acclaimed as any headphone in history.

I’ll close by acknowledging that there are plenty of audio pundits out there who argue that measurements are useless for judging headphones -- or for judging pretty much any other type of audio equipment, for that matter. All of these pundits have one thing in common: none, to my knowledge, has significant experience in performing and evaluating audio measurements.

. . . Brent Butterworth