Sonus Faber Extrema loudspeaker Measurements

Sidebar 3: Measurements

As far as possible, my tests conform to established Stereophile practice, DRA Labs' MLSSA being the main analytical tool. The 88dB/W (8 ohm, 2.83V) sensitivity claim was met within 0.5dB, using broad averaging over the axial reference response. This above-average sensitivity, in conjunction with the 200W maximum program input, will provide up to 111dB at 1m, or a typical value of 106dBA maximum for a stereo pair in a typical 80m3 listening volume. While not deafening, this level is well above the norm for compact two-way systems. The BBC LS3/5a sticks at 95dBA, the SL700 at around 99dBA.

Of these modest-sized boxes only the WATT, with its 91dB sensitivity and 100W power handling, can match the Extrema. (With the Puppy, the WATT will take an increase to 200W to give 109dBA in-room.)

The impedance rating is a smooth 5 ohms with 4 ohm minima of low reactive content. Tube amplifiers rated over 150W, 8 ohms are advised, set to their 4 ohm taps if available. Solid-state models of good peak-current capacity (10A or more) will drive the Extrema well, given 8 ohm ratings of 75W or more. The 225W clip power of the KSA-150 caused no audible distress with high-quality program. The impedance curve (fig.C) shows that the Extrema's impedance holds within 4 and 9 ohms from 10Hz-2.5kHz before rising gently in the high frequencies.

At 1m the Extrema's reference response (fig.1, middle solid trace) shows a mild trough ahead of the crossover point. While this feature was characteristic, it was less severe at a 2m microphone distance (top trace), and still less so at the 3-5m listening position. For 6dB/octave systems, some distance is required for full integration of the driver outputs. On the reference response, the low-frequency response taken in the nearfield has been added. This gives a -6dB point at a relatively low 40Hz—good extension for the speaker's size.

Fig.1 Sonus Faber Extrema, anechoic response on tweeter axis at 1m (solid, bottom) and 2m (solid, top), with the individual drive-unit responses (dashed), and theoretical first-order crossover slopes.

The reference response suggests a 3dB trough an octave wide centered on 2kHz; perhaps the crossover is slightly "gapped" in this design, since the 2kHz interchange appears at typically -7dB on the driver rolloffs.

Overall, the midrange has a broad, mild prominence. The treble is slightly recessed 1dB or so relative to the mean, but is particularly smooth, ±1dB limits applying from 3 to 13kHz. A mild 2dB peak appears at 16kHz before the high-frequency output rolls away without further resonance.

The individual driver responses are also plotted in fig.1 (dashed traces) together with the theoretical 6dB/octave electrical slopes (dotted traces). The mid rollout aligns well with the theoretical up to 6kHz, beyond which falling cone output dictates a faster decay. The tweeter's high-pass function was rather steeper than 6dB, in fact nearer to 10dB/octave, and rather faster than required for a "linear phase" design. This is because the tweeter's overdamped rolloff at low frequencies adds to the electrical 6dB/octave feed, giving the steeper acoustic slope. The single crossover is not as simple as it seems. The optimum listening axis was found to be slightly high of the median between the mid and treble drivers, horizontally directed. If the stand height is fixed, then the adjustable floor spikes may be used to provide some tilt as required.

Listening tests gave the best results with the grille off. The effect of adding the grille can be seen in fig.2, which just shows the response errors introduced by the grille, running +2, -10dB from 800Hz-12kHz. [Sonus Faber points out that the grille is only provided for drive-unit protection.—Ed.]

Fig.2 Sonus Faber Extrema, difference in on-axis response made by grille (2dB/vertical div.).

I analyzed the LF response in some detail, in particular the nearfield response of the ABR (fig.3). The specific response extends up to 150Hz, while some undesirable higher-frequency transmission modes can be seen between 200Hz and 800Hz. With the mass-loading slug added by Sonus Faber, the main breakthrough peak is the first bending modes of the B139 wedge diaphragm, here lowered to 650Hz rather than the usual 800Hz. This suggested that the extra mass would have been better dispersed as a stiffening/damping layer rather than concentrated at one point. That main breakthrough peak was only 6dB below the main ABR output at the "middle" or "0" setting, a total of 12dB below the normal output at 1m.

Fig.3 Sonus Faber Extrema, nearfield ABR response, 1/3-octave smoothed, showing effect of LF Contour control (2dB/vertical div.).

The "Q" of the ABR output varied with setting, from a minimum of 0.86 fully damped through 1.2 at "mid" to 3.1 undamped. The boost bandwidth naturally becomes narrower with increasing Q. The mid setting of 1.2 sounded right; in practice, the ABR output at this stage was already recessed relative to the main output, resulting in an effective system Q of 0.6—close to the ideal. This was why the level steps were quite large at over 2dB when seen in isolation on the ABR curve, but if married up to the main response were diluted to around 1.5dB increments centered on the 40Hz region. These are shown to the left of the forward response family of curves (fig.4), +2, 0, and -2 settings being plotted.

Fig.4 Sonus Faber Extrema, horizontal and vertical response families at 2m: reference response (solid); response ±15 degrees off-axis vertically (dotted, dip is below axis); response 30 degrees off-axis (short dashed); response 45 degrees off-axis (long dashed); response 60 degrees off-axis (dotted/dashed). Effect of LF Contour control is shown in +2, 0, and -2 positions below 100Hz.

The 1/3-octave-smoothed response family curves in fig.4 reveal good integration. The only anomaly concerns the dip at crossover seen with the microphone 15 degrees below axis—for listeners lying on the floor. The smoothness and low decay rate off-axis were thanks to the Extrema's narrow vertical profile and low-diffraction geometry. The Extrema is not strongly axis-dependent.

These responses were highly integrated in the room averaged response (fig.5). Here a graph was generated by averaging 24 responses taken over the listening region for left- and right-hand loudspeakers, a summation of the room acoustic energy with an on-axis contribution. Despite the usual floor dip in the 100Hz region, the Extrema, on its middle-bass setting, gave a respectable in-room response of ±4dB from 27Hz-10kHz, with the normal treble rolloff beyond. Certainly the bass stands out a little in fig.8, the mid remaining broadly prominent; by contrast the treble is a touch reticent, but the result is most satisfactory when looked at overall. The extension to 27Hz was real; checks with a sinewave generator confirmed this.

Fig.5 Sonus Faber Extrema, spatially averaged, 1/3-octave response in MC's listening room.

By virtue of the magnet design, third-harmonic distortion should be quite low; the sweet-sounding midrange supported this. (Increased third harmonic can give the midrange a harder, more nasal sound.) At 90dB spl, the Extrema's distortion was pretty low for a dynamic (moving-coil) design. Above 100Hz the third harmonic was better than 0.1%, and did not exceed 0.15% through the treble. Less subjectively harmful second-harmonic held in the range between 0.35% and 0.15% right down to 50Hz; quite an achievement. Even at 30Hz, second- was a very reasonable 1.5%, third- better than 1%.

Limited by audible distortion, the Extrema could sustain the equivalent of 102dB at 25Hz, 105dB at 30Hz, and its normal maximum of 110dB from 40Hz and above: pretty impressive results.

Distortion tests were also made at 2kHz to check for treble overload. I observed nothing untoward throughout this range, and the distortion through the treble remained at a fine 0.3% level over a wide dynamic range.

The Energy-Time Curve is shown with the preferred Blackmann-Harris weighting (fig.6). This was excellent over the early portion of the decay, but rose back after 2 milliseconds. I suspected that the cause was the low crossover rate taken in conjunction with the tweeter's 500Hz fundamental resonance returning at 45dB down—not very serious, I think. A standard impulse response is also shown (fig.7), this notably clean with a fast decay. A plot of the acoustic phase response, corrected for delay, showed a most uniform trend with much of the main response varying almost linearly with frequency (fig.8). This was a good result.

Fig.6 Sonus Faber Extrema, Energy-Time Curve on tweeter axis at 1m (5dB/vertical div., 10ms time window).

Fig.7 Sonus Faber Extrema, impulse response on tweeter axis at 1m (8ms time window, 20kHz bandwidth).

Fig.8 Sonus Faber Extrema, acoustic phase response on tweeter axis at 1m (corrected for time of flight, 50 degrees/vertical div.).

The first of the two "waterfall" spectral-decay plots (fig.9) uses the processing parameters typical of Stereophile's plots and concentrates on the speed of early decay. It has a 30dB amplitude window. The Extrema was very well-behaved here, with that desirable white space clearly seen at the back of the "waterfall." The midrange was not so quick to clear as the treble, which might be connected with the mild veiling heard in the lower midrange.

Fig.9 Sonus Faber Extrema, cumulative spectral-decay plot at 1m (0.1ms risetime, 30dB dynamic range).

The second "waterfall" (fig.10) changes the filter risetime to a slower 0.2ms, which gives the graph greater frequency resolution. The dynamic window was doubled to 60dB, to allow the decays to grow in the graph. Something interesting is going on here after a millisecond or so. The good clearing can still be seen at the back, but the lower treble shows a dominant feature which itself only fully decays after 3ms. This was the upper-range output of the bass/mid unit present at a reduced level, turning up delayed and requiring its own good time before finally shutting up. Its shape correlates with the ripples seen in the mid-driver rolloff on the first graph of the series. One could argue that this system's transient performance might well be improved by a higher-slope crossover for the mid, pushing these effects out of earshot and out of mind.

Fig.10 Sonus Faber Extrema, cumulative spectral-decay plot at 1m (0.2ms risetime, 60dB dynamic range).

Such considerations aside, these technical results were very good.—Martin Colloms

Sonus Faber
Sumiko - distributor
2431 Fifth Street
Berkeley, CA 94710
(510) 843-4500
Share | |

Enter your username.
Enter the password that accompanies your username.