by D. B. KEELE, JR.
Focal, S.A. is a French company known for its high-quality speaker drivers; JMlab is its speaker system division. Focal sells many drivers for use in home construction projects, and many more as an OEM supplier to companies all over the world, including such well known American manufacturers as Avalon, Legacy, Snell, Swans, Thiel, Westlake Audio, and Wilson Audio. However, finished speakers actually comprise over 70% of Fo cal's business. The American sales arm, JM lab America (a division of the Prism Audio Group), markets more than 30 models in the U.S., ranging in price, per pair, from $495 for a small two-way unit up to $65,000 for the extreme high-end Grande Utopia. The 913.1 comes near the top of the Spectral line in the U.S. Only three JMlab systems exceed it in price: The Antea ($6,495), the Alcor ($11,000), and the aforementioned Grande Utopia.
The 913.1 is a fairly large and heavy three-way speaker using dual 8-inch woofers in a rear-ported vented box. The other drivers are a 5-inch midrange and a 1-inch, titanium inverted-dome tweeter. One of the two parallel-connected woofers is mounted near the top of the front panel; the other is near the middle of the panel, below the midrange and tweeter.
The 913.1's enclosure is a hefty double-sided design consisting of a main cabinet constructed of 3A-inch MDF covered on the sides by 1-inch slabs of African anigre wood. The cabinet contains two equal-volume enclosures for the woofers, each vented to the rear through a tube 2 1/2 inches in diameter and 6 inches long. The midrange is mounted in its own sealed chamber, a 6-inch diameter tube which connects the front and rear of the enclosure. All drivers are mounted to the front panel with socket-head machine screws. The systems are sup plied with spikes that can be screwed into the bottom of the cabinet. The 913.1's grille is made from 1/2-inch MDF and is attached to the front panel with six plastic projections that mate with plastic fittings in the cabinet.
The system's high-excursion 8-inch woofers have diecast baskets and Polykevlar cones. Polykevlar, a Focal exclusive, is a sandwich of resin-bound hollow micro spheres between two sheets of Kevlar. This combination is said to have light weight, rigidity, and excellent damping, but with- out the coloration of conventional Kevlar construction. The woofer's voice-coil is o flat ribbon wire wound on a Kapton former, 1 1/2 inches in diameter.
The 5-inch midrange also uses a Polykevlar cone and features a stationary conical phasing plug. A 1-inch voice-coil, wound on a Nomex former, surrounds the phasing plug and vibrates in relation to it. No dust cap is used or required.
The 913.1's tweeter utilizes an inverted metal dome made of a material that Focal calls Tioxid, titanium that is covered with a thin film of titanium dioxide, and suspended with a flexible, flat foam surround. Un like the voice-coils of conventional convex dome tweeters, which are larger than their domes, the inverted-dome tweeter's voice coil is significantly smaller than its dome, and drives it from its center. The tweeter uses a large magnet, 33A inches in diameter, that raises the tweeter's overall weight to a hefty 3 pounds.
The system's crossover is constructed on a large, 6 x 10-inch p.c. board that is attached to a removable MDF panel mounted on the bottom of the cabinet. The parallel-connected woofers are driven by a third-order, 18-dB/octave, low-pass filter (two series inductors with a series capacitor and resistor to ground). The midrange is fed through a modified bandpass network formed of a third-order, 18-dB/octave, high-pass network plus a high-frequency rejection circuit. The latter is a parallel-resonant LC tank, tuned to 7.9 kHz, in series with the input. (For you electrical engineers out there, it's a third-order, elliptic low-pass filter with a zero in the stopband at 7.9 kHz, which provides a rapid roll-off with a slope approaching 24 dB/octave between 4 and 7 kHz.)
The crossover contains 20 components (five resistors, five inductors, and 10 capacitors, all premium quality) that function as 16 distinct parts, considering paralleled units. The three largest inductors have ferrite cores. Connections between crossover and speakers use 14-gauge stranded wire, soldered at both ends. All drivers are connected in positive polarity.
Input connections are via two bi-wirable pairs of gold-plated, heavy-duty binding posts at the bottom rear of the cabinet. Terminal spacing is a nonstandard 40 mm (15/g inches), so standard double-banana plugs will not work. For normal wiring, the two pairs of input terminals are connected together with large brass bus bars, 0.2 inch in diameter and 2 1/2 inches long. These bus bars go into the holes in the terminal posts, where they will prevent normal-gauge speaker wire from being inserted, so single banana plugs must be used.
----------The terminals can be bi-wired but cannot be used with standard
double-banana plugs.
Measurements
==========
SPECS
Type: Three-way, floor-standing, vented-box system.
Drivers: Two 8-in. cone woofers, 5-in. cone midrange, and 1-in. titanium inverted-dome tweeter.
Frequency Response: 35 Hz to 25 kHz, +0,-3 dB.
Sensitivity: 93.5 dB at 1 meter, 2.83 V rms applied.
Crossover Frequencies: 500 Hz and 3.5 kHz, with 18-dB/octave slopes.
Impedance: 8 ohms average, 4 ohms minimum.
Maximum Amplifier Power: 175 watts continuous, 250 watts program power.
Dimensions: 43 3/8 in. H x 11 1/4 in. W x 14 3/8 in. D (111 cm x 28.5 cm x 36.5 cm).
Weight: 86 lbs. (39 kg) each.
Price: $4,295 per pair.
Company Address: do Prism Audio Group, 14038 Tanglewood Court, Dallas, Tex. 75234, USA.
==========
Figure 1 shows the on-axis anechoic frequency response of the Spectral 913.1. Measurements were taken at a distance of 2 meters from the front of the cabinet, on the midrange axis (as recommended by the factory). A voltage of 5.66 V rms was applied (equivalent to 8 watts into the rated 4-ohm minimum impedance) and then referred back to 1 meter with a 2.83-V rms in put (equivalent to 2 watts into 4 ohms). A combination of ground-plane and elevated free-field measurements was used to derive the curve, which was then averaged with a tenth-octave filter.
If you exclude a high-frequency rise above 15 kHz, the on-axis curve is quite flat and fits within a tight, 4.5-dB window from 50 Hz to 18 kHz. Distinguishing features include a moderately elevated low end be tween 60 and 500 Hz, roughness above 800 Hz, a slight dip at 11 kHz, and a rising high end above 12 kHz. A separate frequency response measurement extending to 30 kHz (not shown) revealed a high-frequency dome-resonance peak at 21 kHz (which reached a level of 97 dB) and then a rapid roll-off at higher frequencies.
Averaged from 250 Hz to 4 kHz, the sensitivity of the 913.1 measured a high 90.0 dB, which is 3.5 dB below JMlab's rating.
The right/left match of the speakers was poor, with the left being some 0.5 to 1 dB above the right below 2 kHz, and below the right by about 1 to 2 dB above 3 kHz. The major difference occurred in the high-frequency range, above 11 kHz, where the left unit was some 2 dB below the right. The grille caused moderate response deviations above 2 kHz, with maximum deviations of about +2 and-3.5 dB in the range from 3 to 9 kHz.
A frequency response check, taken with the woofers re versed at the rear terminals, revealed that the lower crossover was at about 320 Hz rather than the factory-rated 500 Hz. With this reversal, there was a very sharp null, about 50 dB at 320 Hz. The sharpness of this null indicated that the woofer and midrange are solidly in phase throughout their crossover region when connected normally, an ideal situation.
The phase and group-delay responses of the 913.1, referenced to the tweeter's arrival time, are shown in Fig. 2. Also shown is waveform phase, a measure which predicts waveshape fidelity in specific frequency ranges. The phase curve is quite well behaved but rotates 300° between 1 and 10 kHz. When averaged above 1 kHz, the group delay indicates an approximate time offset of about 0.2 mS between the midrange and woofer, with the midrange output delayed behind the tweeter's. The waveform phase curve indicates that wave-shapes will not be preserved within any frequency range, because the phase values are not at or near either 0° or ±180° over any significant frequency band.
Figure 3 shows the Spectral's 1-meter, on-axis, 2.83-V rms energy/time response. The test parameters were chosen to accentuate the system's response between 1 and 10 kHz, which includes the up per crossover region. The main arrival, at 3 mS, is quite compact but is followed by one significant response, down about 20 dB and delayed about 0.76 mS after the main arrival.
The unsmoothed horizontal off-axis responses of the 913.1 are shown in Fig. 4. (The bold curve at the rear of the graph is the on-axis response.) These responses are quite similar over the whole frequency range, particularly in the important main coverage window (±15° of the axis), where no narrowing of high-frequency response is evident.
Figure 5 shows the vertical off-axis curves of the 913.1. (The bold curve in the middle of the graph is the on-axis response.) Within ±10° of the axis, the curves are fairly uniform and symmetrical (on an up/down basis) through the upper crossover range (2.5 to 5 kHz).
Although exhibiting some curve-to-curve variation, the overall vertical coverage does not show any dramatic peaks or nulls in the off-axis response. On average, the vertical off-axis response is quite uniform.
Figure 6 shows the Spectral's impedance magnitude versus frequency. In the bass range, below 100 Hz, the characteristic two peaks and a dip of the vented box are evident. The 4.7-ohm dip at about 27 Hz indicates the approximate location of the vented-box tuning. In the audible range (20 Hz to 20 kHz), an impedance maxi mum of 14.6 ohms is reached in the bass range, at 50 Hz, and a minimum of 3.0 ohms at the slightly higher frequency of 90 Hz.
Fig. 1-One-meter, on-axis frequency response.
Fig. 2--On-axis phase response, group delay, and waveform phase.
Fig. 3--Energy/time response.
Fig. 4--Horizontal off-axis frequency responses.
Fig. 5--Vertical off-axis frequency responses.
The system's max/min impedance variation from 20 Hz to 20 kHz is a fairly high ratio of 4.9 to 1 (14.6 divided by 3.0). This means the cable series resistance should be limited to a low maximum of about 0.044 ohm to keep cable-drop effects from causing response peaks and dips greater than 0.1 dB. For a typical run of about 10 feet, 14-gauge (or larger), low-inductance cable should be used with this speaker.
Figure 7 shows the complex impedance of the 913.1, plotted over the range from 5 Hz to 30 kHz. Although well behaved, the curve exhibits many major and minor loops. The two largest loops correspond to the two impedance peaks of the vented box.
The next smaller loop corresponds to the impedance rise at about 700 Hz. Two very small loops occur at 148 Hz and 4.4 kHz for unknown reasons. A single 913.1 will not be a bad load for any competently designed amplifier, although its low 3-ohm minimum and fairly large impedance variations may cause response aberrations when connected to amplifiers having low damping factors, such as tube units.
A high-level sine-wave sweep revealed a quite rigid cabinet with very low activity of the sides and top at any frequency. The 8 inch woofers have a healthy travel capability of about 0.6 inch, peak to peak, and over load quite gracefully. A moderate amount of dynamic offset was noted at sine-wave levels above 14 V rms. The vented enclosures work very well and reduce the cone excursion at box resonance by a significant three-fourths (x 0.25), comparing response with port open to port closed. Minimum woofer excursion occurred at 33 Hz, the system's vented-box resonance. Port wind noise at high power levels was very low at and near box resonance.
Fig. 6--Impedance.
Fig. 7--Complex impedance.
Fig. 8--Three-meter room response.
Fig. 9--Harmonic distortion for E1 (41.2 Hz).
Figure 8 shows the 3-meter room response of the 913.1, with both raw and sixth-octave smoothed data. The system, in the right-hand stereo position, was aimed laterally at the test microphone, at ear height (36 inches) at the listener's position on the sofa. The speakers were driven with a swept sine-wave signal of 2.83 V rms (corresponding to 2 watts into the 4-ohm impedance). The direct sound and 13 mS of the room's reverberation are included.
If you exclude dips in the floor-bounce range, between 200 and 500 Hz, the averaged curve is well behaved and fits a fairly tight, 8-dB window. With these dips included, the averaged curve still fits within a looser but not-so-bad window of about 12.5 dB. Distinguishing features include a peak at 800 Hz and a general downward trend in response from 1 to 10 kHz, followed by a dip at 12 kHz and a rise at higher frequencies.
Figure 9 shows the E, (41.2-Hz) bass harmonic distortion of the Spectral 913.1, with input power ranging from 0.1 to 100 watts (20 V rms into 4 ohms). The second harmonic reaches only a low 2.2%, while the third rise only to a moderate 3.1%. Higher harmonics are 0.65% or lower at full power. At 1 meter in free space with a 100-watt input, the system generates a fairly loud 100 dB SPL at 41.2 Hz.
The A2 (110-Hz) bass harmonic distortion (not shown) was below the floor of my measuring gear and hence was below 0.33% at full power at all measured harmonics.
The A4 (440-Hz) harmonic distortion is shown in Fig. 10. The second harmonic rises only to 1.3% at full power. Higher harmonics were below the floor of my analyzer at all power levels up to 80 watts. However, at higher power levels the speaker's output exhibited a harsh sound that corresponded with a sudden appearance of the third and fifth harmonics (at 2.0% and 4.0%) at the 100-watt level. Investigations described later revealed that a saturating inductor in the midrange part of the crossover caused sym metrical rounding of the mid range's waveshape.
The IM versus power, created by tones of 440 Hz (A4) and 41.2 Hz (E1) of equal power, is not shown. At 100 watts, the IM rose only to 1.3% at full power, a very low value.
This low IM value is a direct result of the low woofer-to-midrange crossover frequency, which occurs between the frequencies of the test signal. The test tones are handled separately by the woofer and tweeter.
The short-term peak-power input and output capabilities of the Spectral 913.1 are shown in Fig. 11 as a function of frequency.
The test signal was a 6.5-cycle, third-octave-bandwidth tone burst. The peak input power was calculated by assuming that the measured peak voltage was applied across the 4-ohm impedance.
The peak input power starts very strong (240 watts at 20 Hz), rises to a peak of 700 watts at 31 Hz (near the box tuning), stays within a narrow range until 250 Hz, falls to a minimum of 270 watts at 400 Hz, and then rises rapidly into the range of 6 to 8 kilowatts above 2 kHz. A reduction in power handling is noted above 13 kHz due to distorted waveshapes.
Between 200 and 800 Hz, the peak input and output were limited primarily by the midrange's generating a harsh, hollow sound. Severe symmetrical rounding of the output acoustic waveshape was noted in this range. Suspecting inductor saturation in the crossover, I drove the midrange direct, bypassing the internal network. This resulted in considerably more input power handling and much higher clean acoustic output, as shown.
Fig. 10--Harmonic distortion for A4 (440 Hz).
Fig. 11--Maximum peak input power and sound output.
At 400 Hz, when the 913.1 was connected normally, its clean peak maximum output sounded rather anemic. When the midrange was driven directly, the output was much louder and sounded quite robust and clean. The cause of this waveform distortion was saturation of the core of a 2 mH shunt inductor in the midrange circuit.
Saturated shunt inductors cause symmetrical rounding or clipping of the waveform, while saturated series inductors cause triangularization of the waveform. The distorted waveshape driving the midrange was con firmed by observing its voltage drive with an oscilloscope.
With room gain, the maximum peak output SPL of the system starts at a healthy 104 dB at 20 Hz, rises rapidly to a peak of 123 dB at 80 Hz, and then, after maintaining levels of about 121 dB to 250 Hz, falls to 115 dB at 400 Hz. It then rises into the loud range of 125 to 129 dB above 1.25 kHz. The dip in maximum output at 400 Hz coincides with the limited power handling noted in the same range. The peak output crosses the 110-dB SPL level at a low 25 Hz and then crosses the 120-dB SPL level at a significantly low 50 Hz. Its strong low-frequency output places the 913.1 in the top 25% of speakers I have tested.
Use and Listening Tests
When I unpacked the Spectral 913.1s, I was very impressed with the design and quality of the shipping containers.
The speakers were packed in no fewer than three separate boxes cocooned inside each other! Often, the shipping containers I receive, which have passed through the New York offices of Audio and have been unpacked and repacked once, are in shambles. I frequently have to request new containers so that equipment can be returned safely to the manufacturer after I review it. This was not the case with the 913.1s. After unpacking them, I thought that if these speakers were designed and manufactured half as well as they were packed, they must be very good indeed. They did not let me down.
The 913.1s are very handsome. The African anigre wood of the side panels is somewhat similar to a medium or light oak in grain and coloring, but with a slight orange tint. The appearance, fit, finish, and workmanship of the cabinet are excellent. All parts fit together well, and the total assembly is very solid. The speakers are quite heavy for their size and bulk, but fairly easy to move around. The grille is essentially captured between the wood side panels. Tolerances are so tight that the grille is somewhat difficult to remove.
The owner's manual is, of course, printed in both French and English and consists of a large, somewhat odd-sized (8 1/4 x 11 3/4-inch) triple fold-up of six pages. Although generic to all the JMlab products, the manual contains a lot of useful information.
The translation from French is a bit awkward in spots, however, with sentences like:
"This parameter is directly related to the linearity of restitution of your listening room." JMlab suggests a live-end/dead-end style of room, but with the speaker at the live end of the room and the dead end be hind the listener.
The speakers are quite easy to hook up, due to their very accessible terminals. Unfortunately, as mentioned earlier, the terminals' nonstandard spacing precludes use of standard, 3/4-inch-spaced double-banana plugs, and the connection straps used when not bi-wiring preclude the use of bare wire, tips, or spade lugs.
Listening was done with the Spectrals set up in my usual locations, aimed in toward my listening position and placed 8 feet apart and far from the rear and side walls. I listened from a point equidistant from the speakers, 10 feet away. Listening gear included Onkyo and Rotel CD players, a Krell KRC preamp and KSA250 power amp, Straight Wire cabling, and B & W 801 Matrix Series 3 reference speakers.
First listening exhibited lively dynamics, a powerful and extended low end, and a smooth, even character with wide dispersion. Further listening revealed some minor flaws but, everything considered, the Spectral 913.1s did a very credible job.
The jazz piano on The Wonderful Sound of Three Blind Mice (Three Blind Mice GS CD004) was reproduced with quite energetic dynamics and a smooth, open sound.
The acoustic bass sounded quite even and clean, with minimal emphasis or de-emphasis of particular notes. Some slight forwardness was noted, as compared to the B & W speakers.
The dynamic range and high output capability of the 913.1s was demonstrated very well with a recording of Beethoven's "Wellington's Victory," complete with can non and musket shots, on The Cincinnati Symphony's Battle Music of Beethoven and Liszt led by Erich Kunzel (Telarc CD-80079, a favorite of mine). The Spectral 913.1s shine on this kind of material, because their high sensitivity and power handling result in clean reproduction of the high peak pressures required when this music is played at high levels. Here, the JMlab speakers did just about as well as my reference B & W 801s, exhibiting only slightly less bass on the cannon shots. Actually, " Wellington's Victory" describes a battle between the French and the British, which the British won decisively. In my own French-versus-British showdown (JMlab versus B & W), the French do much better! On older, not-so-clean analog vocal recordings, the 913.1s exhibited some high-frequency emphasis and harshness on sibilants that the B & Ws did not exhibit. On clean vocal recordings, they did just fine.
On pink noise, the 913.1s did show significant midrange tonal changes when I stood up. Octave-to-octave spectral balance on pink noise was quite acceptable; I heard minimal tonality. These speakers did, how ever, sound somewhat different than the B & Ws on pink noise. The differences were not bad, just different, and are hard to describe.
On third-octave, band-limited pink noise, the 913.1s generated some usable output in the 20-Hz third-octave band, somewhat more usable output in the 25-Hz band, and very strong and clean output at all higher bands. Port wind noise was quite low. The JMlabs' low-frequency output, al though quite robust, was still not quite as clean as the B & W systems' when reproducing the same levels.
In a second listening session, after I took the measurements, I couldn't find any particular music material that demonstrated the reduced maximum output of the JMlab speakers in the 400-Hz range. What did re veal the problem was a recently recorded CD of my special tone-burst signals.
On large-scale symphonic music, such as Latin American Ballets (Dorian DOR 90211), the 913.1s demonstrated lively dynamics and a wide and accurate soundstage while being quite neutral and well balanced, with an extended frequency response. Tip top information on percussion was reproduced with a bit more emphasis than I heard with the B & W 801s. The JMlab speakers also did very well on less dynamic classical music, such as string quartets, where the 913.1s' stereo focus was very good while preserving the recorded ambience of the source material.
Although $4,295 a pair is a significant amount of money, the Spectral 913.1s did quite well in both my measurements and listening tests, and they compete very well with other systems in their price range.
They exhibited a very dynamic, even, and smooth sound, with an extended bass response. The speakers sounded very good on all types of music, including material that profits from being played very loud. Definitely consider them.
(Adapted from Audio magazine, Nov. 1995)
Also see:
Acoustic Research 338 Speaker (Nov. 1995)
JBL Ti Series Loudspeakers (Auricle, Aug. 1985)
= = = =