Headphone Reviews

Sidebar: Measurements

Measuring headphones is fraught with practical problems, mainly due to the fact that the target response for a drive-unit that fires straight into the ear canal is anything but flat, given the frequency-response–modifying natures of the pinnae and inner ear canal which, as mentioned earlier, are unique for each person. Fig.1, extracted from an excellent 1980 AES paper (footnote 1) from Audio Contributing Editor Jon Sank, shows the envelope of responses he found desirable for conventional headphones to sound flat.

For this review, I measured the responses of the three sets of headphones and my pair of workhorse Sennheiser HD420SLs both in free space, with a B&K 4006 ½" microphone approximately 1" away, and pushed against a flat wooden baffle by a G-clamp with the same mike set flush. Neither situation duplicates the environment faced by a headphone forced against the head and firing into the ear canal. In addition, the ½" microphone capsule is not that much smaller than the headphone drive-units, and would thus be expected to influence the local acoustic. Nevertheless, as I regard review measurements as primarily providing support for the listening comments, I felt it sufficient to use a setup that would produce relative, rather than absolute, data. The following curves should be viewed in this light.

With the exception of the Stax, the headphones were driven by a Linn Intek integrated amplifier for the measurements, the test signal provided by the DRA Labs MLSSA system. All response curves shown are composites, the data below 1kHz being computed by the MLSSA set to a 5kHz bandwidth, above 1kHz with it set to a 20kHz bandwidth.

For reference, fig.2 shows the measured response of the Sennheiser '420 in free space and how it is modified when this supraaural headphone is pressed against the flat baffle. In free space, the response can be seen to slope down below 2.5kHz or so; with the baffle, it hinges up as expected, to provide a basically flat response from 100Hz to 2.5kHz, with the slight upper-bass boost and the low-frequency rolloff point both very dependent on the pressure with which the pad is pressed against the baffle. Above 2.5kHz, the response against the baffle shelves down, though it retains the distinguishing signature of peaks and dips. The '420's impulse and step responses in free space—a clean spike with only a slight degree of treble ringing—can be seen in figs.3 & 4, while fig.5 shows the cumulative spectral-decay plot corresponding to that impulse response—note the slight resonant ridge at 4760Hz.

These three measurements, with fig.2 compared with the target responses in fig.1, correlate quite nicely with the '420's perceived sound quality: rather dull in tonal balance, with a lightweight bass and rolled-off highs, though very clean-sounding apart from a slight degree of nasality and residual brightness. These measurements do not, however, indicate that the Sennheisers significantly lack transparency compared with the other three headphones reviewed.

The Grado HP 1 behaves very similarly to the Sennheiser in how its free-space response is modified by the baffle, as can be seen in fig.6. Again, the exact degree of upper-bass boost and LF extension will be affected by the pressure with which the headphone is held against the baffle. Nevertheless, the HP 1 offers a significantly more generous bass response than the German 'phone, with usable extension to a low 30Hz, while comparing fig.6 with fig.2 shows that the Grado will have a brighter treble presentation than the Sennheiser, though it will not sound as lively as the Staxes or AKGs. Note also the generally exaggerated lower midrange, which correlates with the HP 1's warm tonal balance and which will also contribute to the subjectively rather depressed top octaves.

Looking at the HP 1's impulse response in free space (fig.7), the initial pulse is well-presented, with a somewhat ragged tail overlaid with a degree of ultrasonic ringing. The step response (fig.8) has an excellent right-triangle shape but indicates that the Grado headphones invert abolute polarity. The associated cumulative spectral-decay or "waterfall" plot (fig.9) shows three residual resonances at 3kHz, 10kHz, and 24kHz, though these are mild in degree. No wonder the Grados sounded smooth—this is one well-behaved diaphragm!

My measurement of the Grado's impedance in free space agreed with GAG's in his review: ca 40 ohms across the band, with only a very slight peak at the well-damped free-air LF resonance frequency of 138Hz, and a rise above 10kHz due to the voice-coil's inductance (fig.10).

Looking next at the Stax Lambda Pro, which has pretty much been Stereophile's reference headphone since we first reviewed it in 1984 (Vol.7 No.5, p.45), the free-space response in fig.11 reveals a strong LF resonance at 205Hz. Against the baffle, however, this peak flattens out, with the bass then extending down to 50Hz with a mild degree of pressure. Note that the response then also hinges down above 2kHz, but with the various interference peaks and dips preserved. The free-space impulse and step responses (figs.12 & 13) are inverting, as has been noted elsewhere ("Manufacturers' Comments," March 1991, p.222), and is considerably more ragged-looking than the Grado, with some hash noticeable in the lower treble in the waterfall plot (fig.14). (Ignore the black ridge just under 16kHz, which is due to the computer monitor.) Note the waterfall plot's crinkly appearance, something that always seems to be a feature of planar, surface-driven drive-units, and the long decay in the bass due to the undamped free-space LF resonance.

The AKG K-1000 measured in free space gave an average impedance of 120 ohms (fig.15) and a very flat amplitude response, as might be expected given that its design principle differs from those of the other 'phones reviewed here. Fig.16 shows its free-space impulse response—clean, marred by only a very small amount of HF ringing—and fig.17 the step resposne. Fig.18 shows its waterfall plot in free space. The latter reveals rather more hash between 2 and 4kHz than the Stax, but a smooth, flat midrange and treble response; if this were a loudspeaker, it would be very neutrally balanced one. Note also the languid decay in the bass due to a rather underdamped LF resonance. That the free-space response is flat is confirmed by the full-range frequency response shown in fig.19.

But note from fig.19 what happens when the K-1000 is placed near the flat baffle: while the bass response is effectively unchanged (as might be expected), a significant peak develops at just over 2kHz, with a just-noticeable accentuation of the upper-frequency peaks in the free-space response. The impulse response also changes considerably facing the baffle (fig.20)—you can see the treble ringing extending for 3ms or so, very different from fig.11—while the waterfall plot (fig.21) now features an extended resonant ridge at 2.3kHz. In addition, calculating the Energy-Time Curve (fig.22), which gives an idea of the times at which the sound arrives at the measuring microphone, appeared to show a regular series of arrivals at a spacing consistent with the 2.3kHz "resonance," resembling nothing so much as flutter echo! This behavior was not significantly changed by angling the K-1000's drive-unit so that the sound struck the flush-mounted microphone at a glancing angle. Note also that none of the other headphones measured this month showed this kind of problem.

My first thought was that this behavior must surely correlate with the nasality Bill Sommerwerck mentioned in his March review. Well, maybe it does, maybe it doesn't. As I remarked earlier, measuring headphones is fraught with difficulty. You can never be sure you're measuring what you think you're measuring; the behavior of the headphone in the vicinity of a flat baffle approximates only very crudely the real-life situation. AKG's own measurements, taken with a sophisticated dummy-head technique (March 1991, p.222), do show the shelving down above 2kHz, generally agreeing with my fig.19, but are apparently free from any such resonant or flutter effect.

So what's going on? From my auditioning, I had found the K-1000's nasal coloration to be accentuated when its drive-unit was not facing directly into my ear canal. I conjecture that there is a complicated reaction between the drive-unit and the side of the listener's head, similar to that between the headphone and my flat baffle, which results in this anomalous behavior around 2kHz becoming audible when the headphone is not accurately centered on the ear canal.

Finally, I ran a response sweep on the Stax ED-1 equalizer, which modifies the headphone's frequency response to account for the fact that the sound is firing straight into the ear canals rather than through a 90° angle via the pinna. The EQ, shown in fig.23, revealing a complicated pattern of boost and cut in the treble region, is intended to synthesize the effect of the outer ear on frontal sounds. It should be noted that in addition to changing the amplitude response, the ED-1 inverts signal polarity when the EQ button is pushed. (It's polarity-correct in bypass mode.)

Auditioning the effect of the $800 ED-1 is awkward, as it takes a while to get used to its tonal balance; then when you bypass it, the music seems colored by its inverse response. (This is particularly disturbing with recorded tape hiss.) As GAG found, the equalization is perhaps a bit too extreme in the upper midrange: there's a degree of emphasis in this region, giving a slight "eee" coloration (heard as a complementary "aww" character when the EQ is switched out). Nevertheless, the ED-1 tames the forward treble of the Lambdas, giving a more relaxed, more musically natural sound overall, but still with some steeliness apparent. (It actually worked better with the Grados in this respect.)—John Atkinson

---

Footnote 1: "Improved Real-Ear Tests for Stereophones," Jon R. Sank, JAES, April 1980. See also "On the Standardization of the Frequency Response of High Quality Studio Headphones," Günther Theile, JAES, December 1986; and "Headphones: As Close As You Can Get," Edward M. Long, Audio, April 1991.