Avalon Eclipse loudspeaker Measurements
I use a mixture of nearfield, in-room, and quasi-anechoic FFT measurement techniques (using primarily DRA Labs' MLSSA system with a B&K 4006 microphone, but also an Audio Control Industrial SA-3050A 1/3-octave spectrum analyzer with its calibrated microphone) to investigate objective factors that might explain the sound heard. The speaker's nearfield low-frequency responses and impedance phase and amplitude were measured using Stereophile's Audio Precision System One.
The first thing I wanted to examine was why the sound of the Eclipses seemed so dependent on the amplifier with which it was driven. Accordingly I measured the frequency response of four of the amplifiers I used it with while hooked up to the speakers (footnote 1). The results are shown in fig.1, which plots, from top to bottom, the responses of the Rowland Model One, Mark Levinson No.23.5, Audio Research Classic 60, and VTL Compact 160 monoblock. Note that both the solid-state amplifiers acquire a quarter-dB or so of top-octave boost when hooked up to the Eclipse. More important, note how the relatively high output impedances of the two transformer-coupled tube amplifiers results in a significant modification of their frequency responses, due to the voltage-divider interaction between their output impedance and the speaker impedance.
Fig.1 From top to bottom, frequency response of Jeff Rowland Model One, Mark Levinson No.23.5, Audio Research Classic 60 (4 ohm tap), and VTL Compact 160 monoblock, all measured at their output terminals when loaded by the Avalon Eclipse connected by 2m of bi-wired Cardas Hexlink. (1dB/vertical div.)
Fig.2 shows the Eclipse's impedance amplitude and phase; note that it follows the shapes of the lower two curves in fig.1 exactly. This effect is nothing new. But the Avalon's basic sound changes with the tube amplifiers in that a welcome degree of midrange emphasis is added, about 1dB with the Classic 60 and about 1.5dB with the VTL, which more successfully balances the speaker's treble. The tube amplifiers also slightly but usefully fatten the low bass while reducing the energy in the upper bass, a region where the Eclipse tends to be a little overgenerous, at least in my room. Used with the Audio Research and VTL amplifiers, the Eclipse is altogether a more finely balanced loudspeaker, in my opinion.
Fig.2 Avalon Eclipse, electrical impedance (solid) and phase (dashed) (2 ohms/vertical div.).
Returning to fig.2, the sealed-box woofer tuning results in a high peak of 34.5 ohms at 36Hz. Though the spec implies an impedance not varying much from 6 ohms, there is a similar peak in the midrange due to the crossover, this set to quite a low frequency (below 1kHz, in fact). The high impedances of these peaks will result in more of an interaction between the Eclipse and amplifiers with a highish output impedance, as shown in fig.1. As the impedance doesn't drop below 7 ohms, with high phase angles only present at frequencies when the value is high, the Eclipse represents an easy load for any amplifier to drive, more like 8 ohms than the specified 6. Its lowish sensitivity—it measured 4dB louder than the LS3/5A with a 1/3-octave-wide warble tone centered on 1kHz, approximately equivalent to 87dB/W/m—means that low-powered amplifiers must be avoided.
To look at the Eclipse's anechoic frequency response, I hoisted it onto a 40"-high platform in my yard, well away from the adobe walls, placing the first reflection of its sound, that from the lawn, more than 7ms after the arrival of the direct sound at the microphone, which was 55" away from the tweeter. This 7ms time window means that the MLSSA-derived quasi-anechoic response will be accurate down to a frequency of 140Hz. Below that frequency, I plotted the Eclipse's nearfield response with the mike almost touching the woofer's dustcap. This is shown to the left of fig.3, and shows the Eclipse to dig down reasonably deep in the bass, being -6dB down (with respect to the level at 100Hz) at the sealed-box tuning frequency of 36Hz, comfortably below the bottom note of the double-bass and bass guitar. (The Eclipse was driven with a Mark Levinson No.20.5 for all the MLSSA frequency-response measurements.)
Fig.3 Avalon Eclipse, anechoic response on listening axis at 55", averaged across 30 degrees horizontal window and corrected for microphone response, with the nearfield woofer response plotted below 300Hz.
To the right of fig.3 is shown the anechoic response on the tweeter axis at 55" averaged across a 30 degrees horizontal window. The response trend is pretty smooth, apart from the slight excess of energy in the very top octave, most of which will be above the typical listener's range of hearing. The sharp spike at 25kHz is due to the tweeter's "oil-can" resonance and will also be inaudible. Some minor blips can be seen around 2kHz, but these are not nearly pronounced enough to indicate why the Eclipse's sound should be bright. For comparison, fig.4 shows Avalon's own quasi-anechoic curves for this speaker, measured with the Crown TEF system from 20Hz to 350Hz with 9Hz resolution and from 350Hz to 20kHz with 200Hz resolution. (This is the curve I whistled at earlier.) Broadly similar to fig.3 in that both feature a degree of boost between 1 and 2kHz, it shows more of a problem between 3 and 5kHz, though its top octave is somewhat depressed by comparison, which could be due to a difference in microphone response.
Fig.4 Avalon Eclipse, quasi-anechoic frequency response supplied by manufacturer, lefthand 20-350Hz, righthand 350Hz-20kHz (10dB/vertical div.).
Looking at the manner in which the Eclipse's output changes with listening height, the speaker's balance remains pretty constant as the listener's ears move from about 27" to 40" from the ground, indicating the presence of a high-order crossover. It is only when the listener stands that a significant notch appears in the speaker's output around the crossover frequency.
To look at how these quasi-anechoic responses translate to what would be heard in a room, I take 20 1/3-octave spectral responses of left and right speakers individually across a 72" by 20" window centered on the listening seat and average them with a slight weighting toward the listening position. This has proven to give quite good correlation with the subjective balance in my room, which has reasonably live acoustics. This spatially averaged response for the Eclipse, again with the speakers driven by the Mark Levinson amplifiers, is shown in fig.5. Note the excess of energy in the upper bass, the exact region which fig.1 indicated would be subdued a little by the tube amplifiers.
Fig.5 Avalon Eclipse, spatially averaged, 1/3-octave response in JA's Santa Fe listening room.
Note also the broad excess of energy in the lower treble. Here, I suspect, is one of the reasons why the Eclipse can sound lean with solid-state amplification and why it is so revealing of preamp and CD-player treble problems. Despite its having a fundamentally flat on-axis response, the speaker's wide dispersion in the treble results in too much lower-treble energy in the room, right in the frequency region where the ear is most sensitive. With a typical tube amplifier, the slight rise in midrange energy will balance the treble to what subjectively turns out to be quite a significant extent, though the speaker still could never be called reticent. The typical upper-midrange depression offered by most moving-coil pickups will also make the Eclipse's sound better balanced when compared with CD sources.
The curve shown in fig.5 also explains the results of my bi-amping experiments with the Rowlands. The exaggerated region in fig.5 is just above the speaker's crossover point. Lowering the tweeter level by even a couple of dB usefully hinges down the response, achieving a better match between the midrange and the low treble. However, this will do nothing about the in-room upper-bass thickness I noticed during the bi-amped auditioning.
Turning to the time domain, the Eclipse's impulse and step responses (figs.6 & 7) are typical of a design using a high-order crossover and a metal-dome tweeter. Looking at the individual step responses of the tweeter and woofer (fig.8) reveals that they are connected with the same polarity. (Repeating the measurements for the other speaker of the pair revealed identical behavior, proving that the wide, deep soundstage thrown by the Eclipses was not due to any drive-unit misconnection.)
Fig.6 Avalon Eclipse, impulse response on listening axis at 55" (5ms time window, 30kHz bandwidth).
Fig.7 Avalon Eclipse, step response on listening axis at 55" (5ms time window, 30kHz bandwidth).
Fig.8 Avalon Eclipse, step response of tweeter (red) and woofer (blue) on listening axis at 55" (5ms time window, 30kHz bandwidth).
Footnote 1: As I didn't want to schlepp the Audio Precision System One home, I carried out these measurements in a somewhat different manner. I used the DRA Labs MLSSA system to capture each amplifier's impulse response, first with the MLSSA bandwidth set to 1kHz, then to 35kHz. Performing a Fourier Transform on each impulse response gave the frequency response, which was then corrected to account for the frequency response of the MLSSA's anti-aliasing filter. I then joined the 4Hz-resolution data from 10Hz to 1kHz to the 103Hz-resolution data from 1kHz to 33kHz to give the full-range curves shown in fig.1.—John Atkinson