Floor Loudspeaker Reviews

As with the Magico A5 loudspeaker that Jim Austin reviewed in the July issue, I drove my test gear to his apartment to perform the measurements of this large, heavy speaker. I used DRA Labs' MLSSA system, an Earthworks microphone preamplifier, and a calibrated DPA 4006 microphone to measure the Audiovector R 8's behavior in the farfield. (We maneuvered one of the speakers onto a dolly and aimed it across a room diagonal so that it was maximally distant from the sidewalls.) I used an Earthworks QTC-40 mike for the nearfield and spatially averaged in-room responses. (For the latter, the loudspeakers were in the positions where JCA had auditioned them.)

The R 8's sensitivity is specified as 92.5dB/W/m. My estimate was slightly lower, 90dB(B)/2.83V/m, but this is still usefully higher than average. Audiovector specifies the R 8's impedance as 8 ohms. Using Dayton Audio's DATS V2 system, I found that the impedance magnitude (fig.1, solid trace) was less than 8 ohms across the audioband and remained below 4 ohms for almost the entire bass and midrange. I checked this measurement with the other R 8 sample; the impedance values were identical. The minimum magnitude was 2.52 ohms at 43Hz and 2.54 ohms at 250Hz. The electrical phase angle (dashed trace) is generally low. However, the EPDR (footnote 1) does drop below 2 ohms between 29Hz and 42Hz, with minimum values of 1.4 ohms at 36Hz and 1.63 ohms between 510Hz and 530Hz. The R 8's demand for current will be ameliorated by its high sensitivity, but it should be used with amplifiers that don't have problems driving 4 ohm loads.

The traces in fig.1 are free from the small discontinuities that would imply resonances of some kind. When I investigated the enclosure's vibrational behavior with a plastic-tape accelerometer, I did find some resonant modes on the sidewalls and on the front baffle below the panel on which the forward-firing drive-units are mounted. The most significant mode lay at 375Hz (fig.2), with other, lower-level modes present at 336Hz, 508Hz, and 891Hz. However, all the modes are relatively low in level and have a high Q (Quality Factor), which will work against their having audible consequences.

With the R 8 supported on its spiked feet, its tweeter is 52.5" from the floor. This is well above the ear height of a typical seated listener, which a survey undertaken for Stereophile in the 1990s by Thomas J. Norton indicated averages 36". According to an email JCA received from Audiovector, the R 8's recommended listening axis is between 100cm and 110cm high (39.4"–43.3"). Accordingly, for the quasi-anechoic farfield measurements, I positioned the microphone level with the junction between the two upper woofers, which was 43" from the top of the dolly.

The black trace and the green trace above 400Hz in fig.3 show, respectively, the farfield responses of the tweeter and the lower-frequency drive-units on the recommended listening axis. The specified crossover frequency is 3kHz, but the tweeter rolls off sharply below 4kHz. Although the output of the lower-frequency drivers slopes down above 1.5kHz, there is significant energy up to 10kHz, with narrow peaks at 4.3kHz and 9.1kHz. The former peak is higher in level than the tweeter's output at the same frequency.

The green trace below 400Hz in fig.3 shows the summed nearfield output of the three midrange/woofers on the front baffle. (The top midrange/woofer's output extends into the treble; the middle woofer's output starts to roll off above 650Hz, the bottom woofer's above 300Hz.) While their nearfield output is relatively flat above what appears to be a high-pass corner frequency of 50Hz, the usual boost in the upper bass that occurs with nearfield measurements (footnote 2) is absent. Also absent is the notch at the port tuning frequency, which the low-frequency saddle in the impedance magnitude trace in fig.1 suggests is 43Hz. The blue trace in fig.3 shows the nearfield output of the second port from the bottom on the R 8's rear, which I was told loads the two lower woofers. (I have truncated this trace at 22Hz and 400Hz, as below and above those frequencies the measurement was contaminated with crosstalk.) This port doesn't extend the woofers' low-frequency response. However, Audiovector's R&D manager and founder, Ole Klifoth, mentioned in an email that they "try to avoid any compression build up. ... [T]he drivers become able to deal with more power with very low distortion."

The isobaric woofers fire downward into the R 8's vented base. Their output (fig.3, red trace; truncated below 30Hz and above 300Hz because of crosstalk) is specified as operating below 100Hz, which was confirmed by the measured nearfield response. This rolls off below 50Hz, no lower in frequency than the front-firing woofers. The amber trace in fig.3 shows the nearfield response of the bottom port on the R 8's rear panel, which reflex-loads the isobaric woofers. The port response peaks between 30Hz and 60Hz, extending the isobaric woofers' output. The port behind the top woofer is where a rear-firing midrange unit vents; I haven't shown it in this graph.

The black trace below 300Hz in fig.4 shows the complex sum of the nearfield responses, each weighted in the ratio of the square root of the radiating areas and compensating the acoustic phase for the differences in distance from a nominal farfield microphone position (footnote 3). It peaks in the region covered by the internal isobaric woofers; the rolloff below that region has an approximate slope of 18dB/octave. Above 300Hz in fig.4, the trace shows the R 8's farfield response, aver aged across a 30° horizontal window centered on the recommended axis. The balance is relatively even. The depression in the presence region may well have contributed to my estimate of the loudspeaker's sensitivity being slightly lower than the specification. The output between 100Hz and 450Hz is also a little lower than that between 500Hz and 1.6kHz.

Fig.5 shows the R 8's horizontal radiation pattern, normalized to the response on the recommended axis, which thus appears as a straight line. (The physical limitations of performing the measurements in JCA's room restricted the off-axis measurements to 45° to the sides rather than my usual 90°.) The loudspeaker's dispersion is relatively even, with apparent peaks and dips off-axis compensating for dips and peaks in the on-axis output. The tweeter starts to become directional in the top audio octave, as expected. Fig.6 shows the Audiovector's vertical dispersion, again normalized to the response on the recommended axis. Usefully, the response 5° below the recommended axis, which will be close to JCA's listening axis, is not dissimilar to the response on the recommended axis, though the depression in the mid-treble deepens a little. A large suckout develops at 3.55kHz more than 5° below that axis. The output in the tweeter's passband increases a little as you move above the recommended axis.

The red trace in fig.7 shows the Audiovector R 8s' 1/10 octave-smoothed, spatially averaged response in Jim Austin's room. (The spatial averaging (footnote 4) tends to average out the peaks and dips below 400Hz that are due to the room's resonant modes.) The blue trace shows the spatially averaged response of the Magico A5s taken under identical conditions, other than the presence of subsonic noise from JCA's building's heating/ventilation system, which could not be turned off on the morning that I performed the A5 measurements. The heating system was not operating when I measured the Audiovectors; the red trace in fig.7 therefore plots their output down to a lower frequency than the Magicos'. Because the loudspeakers have different sensitivities, I have normalized their outputs in the lower midrange and mid-treble in this graph.

Both pairs of speakers have an excess of midbass energy in-room, which I suspect will be due, at least in part, to the excitation of the low-frequency modes in JCA's room. Compared with the A5s, which have an impressively even balance at the listening position, the Audiovectors produce too much output in the upper midrange and slightly too little in the presence region. With an unflat response like this, whether the upper mids will be heard as exaggerated or the lower mids and treble will be heard as suppressed depends on the music being played.

I listened to one of my recordings on the R 8s—"In Paradisum" from the Portland State Chamber Choir's Translations album—and was impressed not only by the well-defined stereo imaging but also by the loudspeakers' excellent midrange articulation. This may well have been the result of the balance shown in fig.7.

The R 8s have a little higher output than the A5s in the top two audio octaves, but if you take as a reference the level at 1kHz, the in-room response slopes down in a generally smooth manner. A speaker that has a flat measured top-octave output in an in-room measurement will sound as if the highs are tilted up.

In the time domain, the R 8's step response on the recommended axis (fig.8) reveals that the tweeter and woofers are all connected in positive acoustic polarity. The decay of the tweeter's step smoothly blends with the positive-going start of the top woofer's step, but there is a second arrival, presumably from the middle woofer, 200µs later. The slight rise just before 6ms in this graph might be the output of the isobaric woofer—looking at its nearfield output revealed that it is also connected in positive acoustic polarity. Other than ridges of delayed energy at the frequencies of the treble peaks in the on-axis farfield response, the R 8's cumulative spectral-decay plot (fig.9) is relatively clean.

With its multiple drive-units, each covering a different passband but with significant overlap, the Audiovector R 8 is the most complex loudspeaker I have ever measured. Whether it is more complex than it need be can only be determined by listening, and for that I refer readers to JCA's auditioning comments.—John Atkinson

Footnote 2: A nearfield measurement assumes that the radiators are mounted in a true infinite baffle, ie, one that extends to infinity in both planes.

Footnote 3: See stereophile.com/content/measuring-loudspeakers-part-three-page-6.

Footnote 4: Using MLSSA to generate white noise, I averaged 20 1/10-octave–smoothed power spectra, individually taken for the left and right speakers in a rectangular grid 36" wide by 18" high and centered on the positions of Jim Austin's ears, which were 34" from the floor.