Quad ESL-63 loudspeaker Measurements
As I had dragged my motley collection of test equipment (footnote 1) over to Larry Archibald's listening room to carry out a set of measurements on the Mirage M1 loudspeaker he has reviewed elsewhere in this issue, I thought it might be a good idea to quickly take a look at the measured performances of the latest, "US Monitor" version of the $3995/pair Quad ESL-63, which had so impressed Larry Greenhill in February (Vol.12 No.2). Sam Tellig also writes about the Quad in this month's "Audio Anarchist" column, and the particular sample measured has been serving as Dick Olsher's primary reference for the last six months or so.
The Quads were held clear of the floor with Arcici's excellent dedicated stands and had their cloth "socks" rolled down. Fig.1 shows the speaker's modulus of impedance, measured with an average 10mV drive level. Dropping below 6 ohms below 50Hz and between 5kHz and 18kHz, with minima of 3.5 ohms at 10Hz and 13.85kHz, this represents a reasonably demanding load for an amplifier to handle. Tube amplifiers would be best used from their 4-ohm tap (if they have one), though any solid-state amplifier that more or less doubles its power output into 4 ohms compared with 8 ohms would have no problems driving the Quad.
Fig.1 Quad ESL-63, electrical impedance magnitude (2 ohms/vertical div.).
Unusually, the position and height of the impedance peak in the bass is level-dependent. I am not sure to what cause the resonant peak at 22.4kHz can be attributed; perhaps it is the Helmholtz resonance of the airspaces in the plastic grid used to resistively load the diaphragm. As it will only be audible to dogs and very young children, however, neither of whom have sufficient disposable income to be able to afford the Quads, it is presumably benign (though the two small impedance peaks at 8.3kHz and 12kHz do tie in with audible amplitude-response peaks).
Turning to the spatially averaged 1/3-octave response at a 3m listening position (fig.2) (footnote 2), this is one of the smoothest through the midrange that I have ever measured. With the exception of the slight dip in the 250Hz region due to interference between the direct sound and the reflection from the floor between the speaker and the measuring microphone, ±1dB limits cover the deviation from a flat in-room response between 160Hz and 5kHz. The top two octaves appear to be somewhat depressed on the spatially averaged response, though the region between 12.5kHz and 16kHz can be seen to be a little lively, but this is more a function of the speaker's very limited dispersion in this region than of a true lack of on-axis energy. The individual measurements used to derive fig.2 show that in the vertical plane, the top two octaves drop by 6dB when the listener is level with the top of the speaker, while below the central axis, there is a 3dB drop in the treble. The same applies to off-axis lateral listening.
Fig.2 Quad ESL-63, spatially averaged, 1/3-octave, in-room response in Larry Archibald listening room.
The Quad US Monitor must be listened to on its optimum axis level with the speaker's center, therefore, if the treble is not to be excessively dulled, which means that only two listeners, at best, will be able to get full HF measure. For a normal listening position, it will therefore be essential to use stands; as implied above, the Arcicis, which are excellent value at $175/pair, are the best we have experienced. Exacerbating this sweet-spot characteristic is the fact that the speaker's limited treble dispersion and dipole nature will rob the room's reverberant field of HF, which will make the Quad sound a little lifeless, particularly in overdamped rooms. On the bright side, however, the dipole nature means that bass room resonances will be less easily excited.
Note, for example, that neither the 31.5Hz or 63Hz peaks that appear on the Mirage or Infinity response measurements, due to the two major resonant modes in Larry's room, appear in fig.2. Of course, to be honest, the lower resonance is hardly likely to be excited anyway by a speaker that starts to roll off below 55Hz. In-room, the low-frequency response was 6dB down at 40Hz, though in the nearfield, the response didn't reach -6dB until 32Hz.
The design of the Quad, with its concentric diaphragm regions driven with signals featuring successively increased time delay, is intended to synthesize the acoustic equivalent of a point source positioned some 12" behind the diaphragm's midpoint. Its impulse response should therefore be excellent. Fig.3 shows the 55µs-long unidirectional rectangular pulse that I use as an analytical test signal; fig.4 shows the output of the Quad US Monitor over a 5ms window with the measuring microphone on the optimum axis at a distance of 1m. Pretty damn good, I'm sure you will agree; in fact, it is the best impulse response I have ever seen from a speaker, with a clearly defined, single main spike, with then a small negative undershoot on its return. The slight ringing overlaying the tail of the pulse has a period of just under 0.1ms, equivalent to a frequency of approximately 12kHz. This resonance is probably due to the dust cover. Fig.5 shows the step response calculated from this impulse response. It has an excellent, time-coherent, right-triangle shape, with a lazy overshoot resulting from the bass alignment.
Fig.3 55µs test pulse (5ms time window).
Fig.4 Quad ESL-63, impulse response on listening axis at 1m (5ms time window, 12kHz bandwidth).
Fig.5 Quad ESL-63, step response on listening axis at 1m (5ms time window, 12kHz bandwidth).
Time-coherent impulse and step responses imply an inherently good squarewave response. Fig.6 shows the Quad reproducing a 1kHz squarewave with the microphone 1m away, a hair above the optimum listening axis. Again, this is superb performance, the wave having beautifully defined leading and trailing edges, with the tops broken up by ringing at the same frequency featured in the impulse response, approximately 12kHz. The slope of the top of the squarewave is dependent on microphone position. Moving the mike down a little flattens the top at the expense of increasing the initial overshoot. The waveshape in fig.5 implies a slight low-bass exaggeration on this particular axis, which correlates with the relatively severe treble beaming found earlier in the vertical plane, while the increased overshoot on the optimum axis confirms the steady-state implication that there are slight resonant peaks in the top octave.
Fig.6 Quad ESL-63, 1kHz squarewave response on listening axis at 1m (5ms time window).
Looking at the Discrete Fourier Transform of the US Monitor impulse response (fig.7), this time taken over a 10ms time window and plotted up to 10kHz, reveals a reasonably flat amplitude characteristic, broken up with what I would imagine are mainly interference peaks and dips. It would have been best to take the impulse response at a 2m distance, where the sound from the individual driver sections would have integrated better, but the measurement would then have been rendered invalid by early room reflections of the pulse emitted by the speaker. The low treble seems dominated by two distinct regions, one between 2 and 5kHz and the other centered on 8kHz. Above 10kHz but not shown on this figure, the on-axis response features a slight peak in the 12-14kHz region.
Fig.7 Quad ESL-63, anechoic response on tweeter axis at 1m.
To conclude, the Quad US Monitor features excellent time- and frequency-domain performance, which will render its sound neutral and unexaggerated throughout the midband and low treble, coupled with a freedom from transient smearing. Above that range, some listeners may be bothered by a slight degree of liveliness in the high treble due to the small resonant peaks, but to put these in perspective, they are considerably lower in degree than with almost any dynamic speaker you can name. Low frequencies are adequate at reasonable levels, but the Quad will never be a loudspeaker for bass-drum freaks or Telarc fanciers (footnote 3).
To get the best balance between mids and highs, the Quad's dispersion pattern indicates that it should be listened to on its optimum axis, which will effectively rule out its use for more than two listeners. Those more than a little off-axis will hear a sound significantly depressed in the top two audio octaves. This sensitivity to listening axis will also make it mandatory to raise the Quad off the floor on suitable stands. (With it on the floor, I find the quality of the Quad's upper bass to become, for want of a better word, "puddingy.")—John Atkinson
Footnote 1: Audio Control Industrial SA-3050A 1/3-octave spectrum analyzer with calibrated microphone and integral pink-noise source; Heath/Zenith PC-controlled 8-bit storage 'scope; Fluke 87 true-RMS multimeter; homebrewed computer-controlled sinewave, squarewave, and pulse generators.—John Atkinson
Footnote 2: Individual 1/3-octave response curves are taken for both left and right loudspeakers at 14 positions covering an approximately 8' by 3' spatial "window" centered on the listening position. The curve shown in the figure is the average of these responses, weighted toward the actual listening position. I have found it to give a benchmark response that eliminates the effects of LF room resonances to a large degree, and includes enough of the room sound with that heard direct from the loudspeaker for it to correlate very well with a loudspeaker's subjectively perceived tonal balance. It is not intended to be accurate in absolute terms, but as a consistent basis for comparison with other loudspeakers reviewed in Stereophile, it has proved very reliable.—John Atkinson
Footnote 3: Does anyone else get irritated with Telarc's editorializing when it comes to bass drums? Take the opening of their Maazel Tchaikovsky Symphony 4, when an unscored bass-drum thwack is added to the opening in measure 13, presumably on the grounds that one is present at the analogous moment in the fourth movement, and that any bass drum is a good bass drum.—John Atkinson