Quad ESL-989 electrostatic loudspeaker Measurements
Although the different kind of interaction between a panel speaker and the room can make it sound louder than its measured sensitivity would suggest, the Quad ESL-989 is still on the low side, at an estimated 83dB(B)/2.83V/m. Given its strict 100W power handling, this puts quite a low ceiling on its maximum loudness capability, as LG found. Over most of the audioband the Quad is not a difficult load for an amplifier to drive, the impedance (fig.1) remaining above 6 ohms from 32Hz to 3.7kHz and from 18kHz to 32kHz. And though the magnitude drops to below 4 ohms at 10kHz, the electrical phase angle is close to 0 degrees at the same frequency, which mitigates the drive difficulty. The impedance does drop to a very low value above 50kHz. Fortunately, there will not be any significant musical energy in this region, even with SACD and DVD-A.
Fig.1 Quad ESL-989, electrical impedance (solid) and phase (dashed). (2 ohms/vertical div.)
Interpreting the measured frequency response of a fairly large panel speaker like the ESL-989 is not straightforward. First, the assumption that the microphone is in the speaker's farfield at my routine 50" distance is no longer true, which means that there will be a slight downward trend with frequency, due to the proximity effect. Second, my usual nearfield measurement of the low frequencies will not show the effect of the dipole cancellation, as the antiphase backwave increasingly wraps around to cancel the speaker's direct output with decreasing frequency. The peaked-up bass shown in fig.2 will therefore tend to be more flat in the farfield than it appears in this graph, with useful extension reaching below 40Hz, as LG found in his auditioning.
Fig.2 Quad ESL-989, anechoic response on mid-panel axis at 50", averaged across 30 degrees horizontal window and corrected for microphone response, with the nearfield response plotted below 300Hz.
Looking higher in frequency, there is a broad depression in the midrange compared with the low treble, while the high treble is first a little peaky, then rolls off sharply above 15kHz or so. Figs.3 and 4 show the '989's lateral dispersion, fig.4 having the off-axis responses normalized to the response on the mid-panel axis. Below 10kHz or so, the speaker shows classic dipolar dispersion, with a deep null to its sides. The contour lines are evenly spaced with an even characteristic off-axis, which correlates with stable, precise stereo imaging. The picture looks less good above 10kHz, with a dramatic rolloff evident even 5 degrees away from the central axis, and a series of peaks rather than smooth off-axis behavior. In terms of the listener getting a full measure of top-octave energy, the ESL-989 is definitely a one-person design.
Fig.3 Quad ESL-989, lateral response family at 50", from back to front: responses 90 degrees-5 degrees off-axis, reference response on mid-panel axis, responses 5 degrees-90 degrees off-axis.
Fig.4 Quad ESL-989, lateral response family at 50", from back to front: differences in response 90 degrees-5 degrees off-axis, reference response on mid-panel axis, differences in response 5 degrees-90 degrees off-axis.
The picture is similar in the vertical plane (fig.5). In order to get full top-octave energy, the listener's ears need to be level with the center of the panels and the speakers toed-in to the listening chair. Even then, the '989 may well sound too mellow in a large room.
Fig.5 Quad ESL-989, vertical response family at 50", from back to front: differences in response 15 degrees-5 degrees above mid-panel axis, reference response, differences in response 5 degrees-15 degrees below mid-panel axis.
The Quad's impulse response on the mid-panel axis (fig.6) suggests a time-coherent presentation, with some high-frequency ringing evident. The step response (fig.7) has an almost perfect right-triangle shape, disturbed by what must be a reflection of some kind about 300;us after the initial arrival of the step, and again some high-frequency perturbations. These show up as ridges of delayed energy at 8kHz and above in the cumulative spectral-decay plot (fig.8). The initial dieaway of the sound in this graph is fairly clean at lower frequencies but rather hashy at higher ones, perhaps due to the presence of that early reflection. This reflection might well emanate from the dustcover or the grillecloth—many of my most enjoyable musical experiences with Quad's earlier ESL-63 were with these removed, though the speaker then becomes very vulnerable to environmental damage.
Fig.6 Quad ESL-989, impulse response on mid-panel axis at 50" (5ms time window, 30kHz bandwidth).
Fig.7 Quad ESL-989, step response on mid-panel axis at 50" (5ms time window, 30kHz bandwidth).
Fig.8 Quad ESL-989, cumulative spectral-decay plot at 50" (0.15ms risetime).
You can find my 13-year-old measurements of the ESL-63 online. Though the measurements were performed with completely different hardware to those for the '989, those graphs are almost identical to and as enigmatic as are these measurements of the ESL-989. All I can say is that the reasons for this speaker's undoubtedly superb sound quality are not readily apparent from its measurements. I hope to explore this subject in more depth in a follow-up.—John Atkinson