Between the Ears: the art and science of measuring headphones Page 2
As Toole went on to say, it isn't just that a representative artificial ear should be shaped like the real thing; it also needs to have similar mechanical properties. Circumaural and supraaural headphones typically crush the pinna to some degree, altering its shape and thereby its acoustical interaction with the headphone. So an artificial pinna should have mechanical properties similar to the real thing, as well as a realistic shape.
Twenty-four years after Toole's clarion call, such artificial pinnae are readily available, the best probably being those originally produced by Knowles Electronics for the KEMAR manikin, the most anatomically realistic HATS ever made. Now sold by Danish acoustics company GRAS, the KEMAR ears are available in a range of sizes and Shore hardness. Granted, they don't completely mimic the mechanical properties of a real ear, which has fleshy tissue overlaying cartilage, but they do have some "give" when a circumaural or supraaural headphone is placed on them.
GRAS makes a complete ear-and-cheek simulator incorporating the KEMAR pinna, designated the 43AG and shown in fig.1. Some of its elements, specifically the clamp and stand, are unnecessary for measuring hi-fi headphones, so I bought only the key components for my own artificial-ear setup: the cheek plate, two pinnae (left and right to accommodate "handed" headphone designs), and the IEC 60711 eardrum simulator, which incorporates the measurement microphone. The whole assembly is mounted in a "head" I constructed by bonding together MDF sheets. The result isn't pretty (it will look better with a lick of paint), but it does provide an acoustically inert housing across which circumaural and supraaural headphones can be placed as they would in normal use, thereby applying a representative clamping force.
Fig.1 The 43AG ear-and-cheek simulator from GRAS.
This setup is about as good as commercially available headphone-measurement solutions get, but equipping yourself with such a device is just the beginning of your difficulties. As already alluded to, the frequency response you measure from a headphone using an artificial ear usually isn't flat, nor should it be. So a correction needs to be applied to generate a more familiar response trace, in which flat is nominally correct. This requirement may sound simple enough, but it projects us into the middle of a controversy that has raged for decades: Just what should the ideal headphone frequency response be?
FF or DF?
In what follows, I exclusively consider headphones for reproducing stereo signals originally intended for replay over loudspeakers. It is important to specify this because the frequency-response requirements for headphones intended to reproduce dummy-head (binaural) recordings are differentsomething commonly forgotten when the same headphones are used to listen to both types of recordings.
If we are using headphones in place of speakers, then it might seem logical that the headphone's frequency response should imitate that of a sound source at head level, 30° off the median plane ie, where the loudspeaker would be in a conventional stereo setup. In headphone parlance, this is termed the free-field or FF response assumption. Two such responses are shown in fig.2: one from research conducted in the 1970s by Shaw (blue trace), the results of which were later published in numerical form (footnote 3); and the second, from the publicly available HRTFs measured at MIT using the KEMAR manikin (red). There are significant differences between the two, as there are in the results of other researchers, but a big peak in response between 2 and 3kHz is a consistent feature, as is a progressive decline thereafter except for a smaller peak at around 13kHz. (Note that these are eardrum responses, such as we expect to measure using an artificial ear with an eardrum simulator. Some HRTFs and headphone frequency responses are so-called blocked meatus measurements, obtained with the ear canal closed off. It is important to distinguish which type of response you are looking at; they are not equivalent.)
Fig.2 Frequency responses at the nearer eardrum for a sound source at head height, 30° off the median plane, according to Shaw (blue trace) and the KEMAR manikin (red).
Self-evident as the correctness of the FF-response assumption may seem, it came under concerted attack in the 1980s, principally through the work of Günther Theile at the Institut für Rundfunktechnik (IRT) in Germany (footnote 4). Using a Gestalt model of auditory perception, Theile argued that a free-field headphone frequency response would be appropriate only if the stereo image were perceived to be forward of the listener, as it is when reproduced over loudspeakers. As everyone who has used headphones knows, this is not the casethe image is generally perceived to be either inside or close around the head. Because of this, Theile claimed, a headphone with a free-field frequency response is perceived as spectrally colored.
Footnote 2: F.E. Toole, "The Acoustics and Psychoacoustics of Headphones," Second AES International Conference (May 1984).
Footnote 3: E.A.G. Shaw and M.M. Vaillancourt, "Transformation of Sound-Pressure Level from the Free Field to the Eardrum Presented in Numerical Form," Journal of the Acoustical Society of America, Vol.78 No.3 (September 1985).
Footnote 4: G. Theile, "On the Standardization of the Frequency Response of High-Quality Studio Headphones," Journal of the Audio Engineering Society, Vol.34 No.12 (December 1986).