Buy < Sell > Trade

Reference Series

Dick Olsher, Stereophile 9/86:

The Model R107 represents the flagship of KEF's Reference Series, and is second only to the Professional Series KM-1 in KEF's product line. Anatomically, the 107 resembles a person. Beneath a decorative "hat", there's a special head assembly akin to the head on the old Model R105. This head assembly contains the brains of the 107, namely a T33 ferrofluid-cooled tweeter and an improved version of the classic B110 midrange driver, featuring a better voice-coil and a new polypropylene cone. The nerve center is also here, in the form of two passive dividing networks and load-impedance equalizing network. Level equalization of the drivers is performed actively within the KUBE, the second brain of the 107−about which you'll hear more shortly.

The head assembly swivels at the neck, allowing the mid and treble units to be toed-in to maintain a good stereo image when the main enclosure is placed square against a wall. The rounded shape of the 105 head has been retained in order to minimize treble diffraction effects, which can smear or defocus the stereo image. Otherwise, the head enclosure has been re-engineered to increase its rigidity, and to slightly increase the cavity volume. A mineral-loaded polymer "goop" is injected into the cavity walls of the head enclosure for damping purposes.

The guts of the 107 are housed in the main low-frequency enclosure that forms the body trunk. The bass loading used by KEF has the intimidating name of "twin coupled-cavity bass loading" and was first seen in the Model 104/2. (In retrospect, the 107 may be seen as a hybrid of the 104/2 and 105 technologies, with a few improvements thrown in). The "twin" in the title refers to the use of two 10" paper-cone woofers, each working in its own sealed enclosure and firing into a common central cavity. Acoustic energy from the cavity is vented via a tuned port that radiates from the top face of the LF enclosure adjacent to, or at the neck of, the head assembly.

This is where my anthropomorphic analogy breaks down. After all, who's ever heard of a bottom end venting at the neck? Positioning the head assembly close to the upward-vented bass energy enables the entire audio bandwidth to be, as KEF puts it, "directed towards the listener from an area little larger than the human head". Output integration is thus optimized, while low-frequency directional effects are minimized. Also, because of the height of the bass vent above the floor, reinforcement effects are minimized. All of which should add up to an exceptionally smooth in-room response. Does it? Stay tuned.

The bass driver system is effectively an extension of the bass reflex principle, but with the port supplying all the output rather than reinforcing the main driver above resonance and cancelling its output below. The air mass in the port, driven by the concealed woofers, acts as a long-throw, low-mass drive-unit with an intrinsic bandpass response tuned to 90Hz. The woofers' magnet structures are rigidly linked by a force-cancelling rod that minimizes vibration transfer to the cabinet from the driver frames or baskets. The cabinet itself is very nicely finished and exudes elegance, as an expensive speaker should. Gold-plated binding posts are provided on the rear that will accept bare wire, banana plugs, or spade lugs. These posts feature large, easy-to-use knurled nuts that can be finger-tightened for a really tight connection. However, the posts are not spaced on 0.75" centers, so dual banana plugs cannot be used.


That is the question, and the 107 answers in the affirmative. The KUBE cuts through several vexing problems that have traditionally compromised subwoofer designs. Even my six-year-old, Dahlia, associates bass sensitivity and extension with large enclosures. And, indeed, the laws of physics conspire to make it so, as Richard Small (now working as Head of Research at KEF in the UK) and A.N. Thiele so aptly demonstrated almost 20 years ago. I'll get a bit didactic and elaborate:

In order to achieve flat LF response to, say, 20Hz, you have to start with a woofer whose free-air resonance is no higher than 20Hz, and preferably even a bit lower. Remember that below the driver's resonant frequency, the output will droop as the driver is no longer mass-controlled and cannot maintain a constant acceleration as frequency decreases. Next, it turns out that a large woofer is much more adept at pressurizing a given-size box than a small woofer, and a very large air volume is needed to prevent the air stiffness of the enclosure from significantly raising the system's resonant frequency. Reduce the volume and low bass can still be squeezed out of the design, but only at the expense of severely curtailing sensitivity.

If you want low bass and a usable sensitivity, therefore, what you end up with is a very large box that is not only expensive to construct and ship, but also, most importantly, almost impossible to make rigid enough to minimize colorful cabinet radiations. Have you ever seen the WAMM subwoofer? Do you know how much carpenters get paid these days? Now you know why that system costs over $40k. So it appears that, at least for realistically sized production-line speakers, the hopes of realizing a 20Hz response using a natural alignment are doomed.

The intelligent engineering solution is to use active equalization between the pre- and power amplifiers to "artificially" extend and flatten the LF response. There has been a recent trend to do just that: recall for a moment the Enigma subwoofer, or, more recently, the free-standing Celestion woofer (footnote 1). (Of course, the latter two designs minimize cabinet colorations by doing away with the cabinet altogether, another plus). KEF's strategy is to align the coupled-cavity for optimum upper-bass sensitivity, then boost the frequencies below the natural turnover to give a flat extended response.

(Don't run out now and buy yourself an equalizer to do just that with your speakers; odds are that you'll simply succeed in destroying the woofers. You should know that woofers are unhappy about being pushed into the deep bass because cone velocity doubles and excursion quadruples for each lower octave. Any design that uses bass boost, therefore, must necessarily include woofers that can cope with the excursion demands).

Footnote 1: Actually, the only people who can "recall" the Celestion must have been in attendance at the last CES, since the woofer is only now in production. −Larry Archibald

KUBE stands for KEF Universal Bass Equalizer and takes the form of a separate solid-state component that connects in the tape monitor loop or between the preamp and power amp. The KUBE supplied with the 107 includes a speaker-specific equalization module that provides level equalization for the tweeter and mid driver, as well as an inverse of the 107's LF response. Furthermore, the KUBE allows the user to control several parameters not previously accessible. First, the LF cutoff is adjustable in four steps: 50, 35, 25, and 18Hz. Second, "Q" of the bass response is continuously variable from 0.3 (overdamped) to 0.7 (maximally flat). The Q parameter describes the sharpness of a speaker's bass resonance. As the Q increases, bass damping decreases until ultimately a response peak develops. The Q parameter also correlates well with bass transient behavior, high-Q designs resulting in underdamped, boomy bass with a lot of overhang. For example, jukebox bass may have a Q as high as 3. That's real, state-of-the-art boom.

The last control on the KUBE shelves the response below 160Hz up or down by as much as 3dB to suit one's tastes or room requirements. The bottom line (no pun intended) is that the KUBE offers extremely flexible control of the in-room bass quality, magnitude, and extension, that I for one appreciate very much.

The Case of the Sick Cube

Even KUBEs can get sick, and mine certainly wasn't well right out of the box, as I discovered when my Boulder 500 amplifier indicated "DC offset" and refused to play. I measured an abnormally hefty DC voltage of 700mV at both KUBE outputs. The day after I described the problem to Fred Yando at KEF America, I received a phone call from England. The KEF engineer with whom I spoke assured me that the KUBE parameters for my pair (SN 1407) were on file (footnote 2), and that a new KUBE would be on its way to me as soon as possible. A week later the new KUBE showed up at my doorstep via Federal Express, and I'm happy to report that I couldn't measure any DC offset this time around! KEF acquitted itself very well throughout this episode, providing efficient and courteous service. And, as Mr. Yando put it, KEF guarantees this level of support not just to reviewers, but to anyone who invests in the R107s.

Sonic Impressions

Whereas most dynamic speakers struggle to integrate their driver outputs into a semicoherent blend in the far field, the 107 is seamless. What struck me immediately about the 107s, when I heard them first under show conditions, was their wonderfully cohesive and effortless sound. That impression has not diminished in the comfort of my own home. The transition from the bass to the mids is magically integrated, and the conviction with which I perceive instrumental space and focus across the soundstage is heightened considerably.

Music emerges from the 107 like ripples in a clear mountain lake. There is no sense of strain or stress, even on very loud passages. The vocal-range magic of the Spendor BC1 and SP1 is present here in full force. Voice is reproduced very naturally, without any coloration. Neither are instrumental timbres slighted in any way, and harmonic accuracy is preserved up and down the audio bandwidth. String fundamentals and overtones are convincingly reproduced, and woodwinds possess excellent clarity and detail, while brass has its proper bite, without gratuitous brightness or harshness. With the right electronics, musical textures are fleshed out with a convincing mixture of soft and hard, sweet and brash.

This is not a hi-fi-ish sort of speaker; it doesn't overdo the music with impressive special effects or extraneous boom and sizzle. It is very enjoyable, always listenable, and soundstage imaging is excellent, with width and depth perspectives as good as anything I've had in the house. The illusion of a continuous soundstage independent of the speakers is very strong with the 107. Instrumental focus within this stage is very good; perhaps not as holographic as I've heard with the Apogees, but extremely good compared with the conventional competition.

There are two other prize-winning aspects to the 107. First, the bass. The quality and quantity of the bass octaves are exceptional−the best I've heard from a conventional full-range system. Bass power, impact, and clarity leave one breathless on orchestral spectaculars. Second, headroom is sufficient to duplicate, at a typical listening position (even in a large room), the full dynamic range of any orchestral work−even a Respighi. The 107 gracefully complies with the dynamic demands of the music−it's capable of blooming from soft to very loud without a trace of compression or distortion. For example, it is hard to think of choral music with more electrifying intensity and grandeur than Sir William Walton's Belshazzar's Feast (EMI SAN324). Andre Previn is able to get the most out of the London Symphony Orchestra, and the English baritone John Shirley-Quirk is particularly effective in the haunting "King's Feast" movement. This work demands a full-range system with excellent dynamics, and the 107 is equal to the task, handling the full orchestra and chorus with aplomb.

Midrange transparency, and resolution of low-level detail, are very good but not quite state-of-the-art; the Apogees and the really good full-range electrostatics better the 107 here. The Apogees in particular possess an outstanding transparency that I find addictive. These are areas where I believe conventional speakers will always lag behind high-tech designs. As with the Frieds reviewed in this issue, you'll need to juggle your priorities and decide for yourself.

Although the highs are very extended, they don't have the delicacy and speed of the best tweeters money can buy. I'm thinking specifically of the Plasmatronics helium-plasma design and the much older Dukane Ionovac tweeter. In fact, KEF's T33 tweeter doesn't even sound as good as the Celestion copper-dome tweeter in the SL600. This is most obvious in the detail of brushed cymbals: the 107 is very listenable up here, but unable to resolve all of the available transient detail with sufficient intertransient silence. There's a little more mud and less clarity with the T33. This really is the only significant reservation I have about the 107, and I hope that KEF pays the treble a bit more attention in future models.

The impedance of the 107 is an amplifier-friendly resistive load of 4 ohms, made possible by adding a "conjugate load" to the crossover, which compensates for inductive and capacitive impedances. However, the LF boost fed the amp by the KUBE will make life more difficult for tube amplifiers. And, although I had pretty good results with my modified Michaelson and Austin TVA-10, I suspect that such really top-notch solid-state amps as the Boulder 500 will work best with the 107.

Overall, I consider the KEF 107s to be an engineering marvel and, by a clear margin, the best full-range conventional speaker money can buy. This is a smart speaker (two brains, remember), so it's not surprising that−in my opinion−it blows away all other conventional speakers in our Class B recommendations. They are that good! I have enjoyed the 107 over the past several weeks, and urge you to audition them promptly at your nearest KEF dealer.

Footnote 2: The use of the term "Reference" for this range of KEF loudspeakers means that both pair-matching and matching to the engineering prototype are exceptional; drivers, crossover components, and the complete loudspeaker are measured, and the details kept on file. Blow a driver and KEF will be able to supply you with a replacement that will match the sound of the original. −Dick Olsher

I decided to measure the 107 before doing any serious listening (something I usually do not do) so I could sort out the KUBE options and optimize the 107's response for my favorite placement. With the bass cutoff set at 18Hz (the setting I used for my auditioning), the 107 puts out energy down to 20Hz, which is as low as I can measure. In fact, bass output is so extended that the 107 can readily overload a listening room. Standing-wave resonances were excited in my room like never before, including an especially nasty one at 60Hz. The contour control cannot cope with such narrow-band peaks or dips. It certainly works as advertised, shelving the response below 160Hz, but that only lowers and raises the peaks and dips without flattening them out. What is needed, in the absence of a digital time-domain equalizer, is a half- or third-octave equalizer that can work over narrow enough bandwidths. I was hoping to try Audio Control's "Richter Scale", but it did not arrive in time to meet my copy date.

The Q control does affect the prominence of the bass line, and is best adjusted for a given placement by ear (using, say, double bass pizzicato) until the proper heft and decay are achieved. Ultimately, the choice of proper Q is a matter of taste and room acoustics. In my room I preferred a Q of 0.7 for placements far removed from room boundaries and a Q of 0.5 for near-wall placements. From about 200Hz to 20kHz, the in-room response is exceptionally flat and easily meets specification. This sort of response smoothness is quite an accomplishment for a conventional speaker; only large-area electrostatics in the near field have measured better in this respect. −Dick Olsher

Martin Colloms, Stereophile 3/87:

Dick Olsher first reviewed the R107 in Stereophile in Vol.9 No.7, and readers should read that review for a complete description. Briefly, bass is handled by two 10" pulp cone woofers mounted in small infinite-baffle enclosures within the R107's main box. The chassis of these drivers are coupled by a bracing bar to reduce the mutual vibration contribution to the enclosure. Their acoustic outputs are summed in a central chamber with the filtered "bandpass" energy emerging from the large port/grille on the enclosure top surface, just under the head assembly. The air mass in the port is, in effect, the actual bass driver.

The head assembly itself is a molded and sculpted casing with a low diffraction shape. It is double-walled, and heavily loaded with damping mastic; it is thus highly nonresonant, and, with appropriate drive-units, allows for a very smooth frequency response. These units comprise a 5" polypropylene-cone unit based on the classic B110, and the T33 doped-fabric dome tweeter, the latter a recessed design with a short horn/phase plate. The R107 is essentially driver time-delay compensated, allowing the 24dB/octave crossover to operate with an in-phase electrical connection. The system is well styled and finished, requires only a small floor "footprint", and the main listening axis is at a good height.

The sophisticated crossover is fully compensated in terms of impedance amplitude and phase, the entire system presenting a constant 4-ohm resistive load to the amplifier. Electrical connection is via knurled, gold-plated 4mm binding posts on the rear of the bass enclosure. These are deliberately spaced too far apart for double-plugs to be used.

The KUBE electronics box connects either in the preamplifier tape loop or between pre- and power amplifiers, and includes a remote-cord power supply. The circuitry is based on TL0 series IC op-amps, with rather a large number required to implement all the equalization functions. Some measure of mono bass-blending is applied electrically at very low frequencies to reduce the audibility of out-of-phase rumble noise on recordings, such as background traffic, air conditioning, pressing rumble, and the like. These often become all too obvious when an 18Hz limit woofer with a healthy output is in operation. At present, the KUBE comes with permanently connected audio cables, terminated in gold-plated phono plugs, but production from early 1987 onward will be fitted with RCA sockets, allowing the user to optimize cable choice for his or her system.

The KEF features a very high standard of construction, engineering, and finish, and a helpful operating manual is supplied.

Sound Quality

It so happened that I first tried this speaker with a fine, budget-priced, integrated amplifier, the Audiolab 8000A, and achieved remarkable results for the total cost of the system. The R107 immediately provided the rewarding experience of a big, focused stereo soundstage offering good depth and perspective. Resolution of fine detail was most promising, the tonal balance was essentially neutral, and the bass extension and power placed it in a rarefied league commensurate with some of the largest and most costly models in the business. Outweighing the importance of these primary qualities, however, was a sense of reserve and ease pervading the subjective dynamics. Even with this modest amplifier (typically 70W/channel), the volume control was set well down, and the maximum level attainable seemed unbounded.

Old warhorse demonstration records were brought out and belted for all they were worth, to see just how well the '107 could rattle the windows! I had no doubts, even at this early stage in the auditioning, that here was a formidable product. Other amps−a pair of Krell KMA-100s on temporary loan, as well as my Audio Research M100s driven by an SP-11−were pressed into service. These provided more performance in quality terms, but the improvement was not as great as anticipated. The small integrated amplifier−admittedly a good one−had, in fact, taken the R107 close to the limit of its performance, and in the main, larger amplifiers only showed the 3-4dB maximum-level improvement that was possible with their higher voltage drive.

The various KUBE settings were the subject of much experiment, though ultimately the manufacturer's suggested adjustments of 20Hz bandwidth, Q of 0.4-0.5, were preferred, in this instance with a touch (+1dB) of contour lift.

Various room positions were also tried: while the sound was clearly optimal in a free-space position, 0.9m from the rear and 1.1m from the side walls and floor-spiked, this speaker could also survive a position close to the back wall surprisingly well. Some bass rolloff (30Hz), a reduced Q (0.3-0.4), and some contour cut (-1.5dB) helped it to balance pretty well against the wall, though there was a noted impairment in the stereo focus and depth, with some increased wall-reflection coloration also apparent.

As listening proceeded, it became clear that this big KEF was fundamentally "correct" in a manner more reminiscent of the R105 II than the more exuberant R104/2. Checking my notes on the R105, I found the R107 to be much more sensitive and capable of maximum sound levels some 5dB higher, as well as producing still better stereo imaging and a faster, yet more extended, bass.

Taken to the limit, it did prove possible to fault the '107 in relatively minor areas. There was a hint of an "enclosed" quality in the high treble, like a dulling of "air". The midrange could sound a touch lean at times, this characteristic also noted with church choir, where upper-range voices could mix and harden together. Care was taken to ensure that a similar effect in the ear itself, resulting from excessive level, was not occurring.

The bass had a quick, fluid quality and appeared to pump effortlessly from the large port. The upper bass was clean and articulate, setting high standards, but ultimately the low bass was a little overblown, as if the 20-30Hz band was "heavy". On some material, it produced a sort of low rumble, an exaggeration of program faults sufficient to require resetting the bandwidth control to 40Hz. The bass balance was fine down to 40Hz, but the extra plumpness below this range could not be ameliorated without drying up the next immediate curve.

A final and still moderate criticism concerns two other areas−the treble and general transparency. In general terms, the R107 sounded clear and open but did not fully resolve detail far in back in the depth plane, despite a decent retrieval of the recorded acoustic. This was a broad-band effect which was a little poorer in the treble. Good forward treble detail was available, but there was also muted "grain", with a loss of real transparency; just how much this matters will depend on the quality of the program source and, of course, on the transparency of the matching electronics.

Next, we removed the KUBE from the chain. Yes, we were aware that the treble would then be 2dB brighter, the output peaking a little at the upper crossover point, and the bass not optimally aligned. Nevertheless, the "naked" speaker sounded quite well balanced, while the change in bass performance was less than might have been expected. Although this was admittedly an academic exercise, the sound quality was briefly evaluated under these conditions: subjective dynamics (a sense of "liveness") and transparency were both significantly improved, and rather more than any superficial change in response might have suggested. The KUBE electronics are appropriate and well matched to the basic R107, but appear to bound its performance, preempting its potential challenge to the next rank of quality. This might otherwise have been achieved via significant expenditure on the matching power amplifier.


This remarkable speaker represents a challenge to the increasing performance standards being set by large panel speakers, matching the best of them in terms of both bass power and extension, while at the same time conceding little in terms of coloration levels. In fact, the KEF beats a number of the panel models in such fundamental areas as tonal balance and response uniformity.

The R107 is a refined system, with no obvious flaws or rough edges. In no way does it represent a "first prototype"−its designer is not going to have second thoughts during the next six or twelve months! The R107 is clearly a bighearted speaker offering good engineering value and a high overall performance. The bass extension is remarkable, with a genuine 18Hz possible at respectable levels, as well as the ability to produce respectably high maximum sound levels appropriate to larger rooms. Its KUBE provides control of the bass and tonal balance to suit a wide range of locations and specific acoustics; the extended bass is allied to a high standard of sound quality throughout, and has not been obtained at the expense of the mid or treble quality.

The R107 is clearly a most effective flagship for KEF, and shows off the company's engineering strengths to the full. −Martin Colloms


As part of the preliminary test work, the KUBE was subjected to some basic lab tests for crosstalk, headroom, distortion, and noise. These were all passed with no trouble at all; we then ran off some response curves at the various settings. For example, fig.1 (one horizontal division equals 5dB) shows the effect of the bass-extension control for a constant Q of 0.5, Contour set at Zero. Given that the speaker has a falling low-frequency response, increasing its extension requires various degrees of bass boost. On line 1, with the minimum extension of 50Hz, the boost is negligible at 20Hz, while for the most extended 18Hz, -3dB response, the 20Hz point is lifted by 11dB and the 10Hz point by 17dB, relative to the midband. (Note that two constant equalization features are present on the graph, namely the dip at 50Hz to account for the underdamped intrinsic bass resonance cancelled by the KUBE, and the small dip near the upper crossover point at 2kHz, together with an associated 2.5dB of treble shelf cut).

In practice, the bass power and equalization demands are generally lower than this, at typically +6dB at 30Hz, this providing little restriction on maximum level. The R107 load impedance checked out as an essentially pure 4 ohms resistance, within 5% from 20Hz to 20kHz−a relatively straightforward amplifier load. This resistive loading is compatible with many tubed amplifiers connected on their 4 ohm taps; eg, the ARC D115 or a Quicksilver mono. With a verified 90dB/W (8 ohm, 2.83V) sensitivity (within ±0.5dB), and a 200W (8 ohm) power capacity on program, the R107 will provide high maximum sound levels of up to 110dBA in typical rooms. Amplifiers as small as 25W will still offer quite high levels.

A plot of the forward frequency characteristics (semi-anechoic third-octave) was taken at 1 meter (fig.5). This showed the speaker to have a superbly uniform axial response, typically holding within ±2dB limits over the 40Hz-20kHz test range. Given its physical height, the 15 degrees off-axis result in the vertical plane above this speaker is an unlikely one, and shows some irregularity around the crossover point. Over the 30 degrees and 40 degrees lateral off-axis responses, the main frequency range is extremely well controlled. Some minor off-axis unevenness exists in the upper treble.

Fig.1 KEF R107 KUBE, Extension at Q = 0.5, Contour = 0 (5dB/vertical div.).

Fig.2 shows the effect of the Contour control at Q 0.5, Extension 18Hz (Setting 4). Operating below 300Hz, the contour applied a gentle, broad lift to the bass/low midrange, allowing control of the tonal balance to account for taste and variation in room acoustics. Typically, a range of ±3dB is available relative to the mean setting. In practice, no more than ±1.5dB should be required for such adjustment, so the range KEF offers is satisfactorily generous.

Fig.2 KEF R107 KUBE, Contour at Q = 0.5, Extension = 18Hz (5dB/vertical div.).

Fig.3 explores the effect of the synthesized system resonance Q over the allowed range of 0.3 to 0.7, with Extension set to 18Hz and Contour set at the mid or zero position. At the maximum Q of 0.7, the 50Hz box "excess" is barely nulled, while the rate of boost is at its most extreme, with +18dB at 10Hz. For a Q of 0.5, close to the optimum, the rate of boost is more gentle, with increased control at 50Hz. For Q of 0.3, the slope is shallower still, approximately 7dB/octave at the transition, while the 50Hz lift is fully cancelled by 3-4dB of equalization dip.

Fig.3 KEF R107 KUBE, Q at Extension = 20Hz, Contour = 0 (5dB/vertical div.).

Given the wide combination of equalization options, fig.4 was prepared to show the maximum and minimum limits. With a minimum 50Hz extension, an overdamped Q of 0.3, and Contour at -3, the 50Hz point is subject to a substantial cut of 12dB. Conversely, with everything at "full wick" the 10Hz boost rises to 22dB, requiring considerable extra headroom in the matching amplifier. Suppose a disc with relatively heavy bass requires an spl at 30Hz 3dB below the peak program level. In the fully equalized position, the bass requires 10dB of lift, less 3dB, for the equivalent bass level. Thus a 100W amplifier clipped in the bass at this 30Hz level could only drive the mid to a level 7dB lower, or approximately 20W. This is where the well-above-average sensitivity of the R107 performs a service, still managing to generate a substantial 98dB-plus for a stereo pair in a typical room from this "20W" midband input.

Fig.4 KEF R107 KUBE, Maximum.minimum equalization possible (5dB/vertical div.).

Fig.5 KEF R107, forward frequency characteristics (semi-anechoic 1/3-octave) at 1 meter.

Fig.6 shows the predominant axial frequency response at a rather close 0.5m measuring distance. This fine result also indicates the excellent integration of the driver outputs in this system.

Fig.6 KEF R107, 1/3-octave axial frequency response at 0.5m.

Fig.7 gives the averaged room response, with the KUBE omitted−a little bright, with a lumpy low bass, but with sufficient extension to 25Hz, in my room at least.

Fig.7 KEF R107, spatially averaged, 1/3-octave, free-field response in MC's listening room, no KUBE.

Fig.8 shows the computed room response with the standard settings of Q 0.5, Extension 20Hz, and with the Contour at zero. The low bass is seen to be a touch excessive, from 25Hz to 50Hz, while the upper bass is muted at 80Hz.

Fig.8 KEF R107, spatially averaged, 1/3-octave, free-field response in MC's listening room. KUBE settings: Q = 0.5, Extension = 20Hz, and Contour = 0.

Fig.9 shows the result of one of my attempts to tame the bass, using a Q of 0.4 and an extension of approximately 40Hz. Subjectively, this was considered respectably "dry" in the low bass, but somewhat lightweight in terms of midrange balance.

Fig.9 KEF R107, spatially averaged, 1/3-octave, free-field response in MC's listening room. KUBE settings: Q = 0.4, Extension = 40Hz, and Contour = 0.

Fig.10 shows the result of a good compromise in terms of bass balance and midrange tonality, though with some inevitable excess in the low bass. This result employed a reduced Q of 0.3, full 20Hz extension, and some +1 of contour lift to  warm  the tonal balance. The 50Hz point is several dB high, but not unacceptably so, while the overall frequency balance is pretty good.

Fig.10 KEF R107, spatially averaged, 1/3-octave, free-field response in MC's listening room. KUBE settings: Q = 0.3, Extension = 20Hz, and Contour = 1.

Specific distortion tests were not conducted; however, no perceptible distortion was noted during the tests, except in the case of gross bass overdrive. −Martin Colloms

John Atkinson, Stereophile 5/91:

Bass. A small word. But a profound one: bass is the base of all Western music, the underpinning that organizes the overlying structure. Basically speaking, music without bass is music without roots or soul. Yet how many loudspeakers do justice to that concept? "You don't understand bass", said a St. Louis reader in a recent letter, who concluded that I (once a professional bass player) was "either deaf below 60Hz, didn't have room for decent-sized speakers, or never listened to bass-heavy rock". (Not one of the three, as it happens, is true). But did he understand? Did he have a true feel for the concept of "bass", citing as he did in his letter a veritable litany of best-selling thunderboxes −loudspeakers that are only loud; underdamped speakers that, in general, I have found to confuse the musically aware listener with a vague, monochromatic sea of low-frequency nonsense that obscures rather than clarifies harmonic structure.

My abhorrence of such proletarian boomboxes explains why over the years I have resisted the blowzy, bosomy beckonings of "full-range" speakers, why I have tended to stick with musically accurate, tonally precise minimonitors like the Celestion SL600 and SL700. While sacrificing the true pedal weight of the bass register, these thoroughbreds present sufficient of its outline and pitch to preserve musical sense.

Either of these superb Celestions represents the triumph of quality over quantity...yet, at regular intervals, I forget my asceticism. My monkey bone hankers for music's full measure. If only it wasn't so rare to find a full-range speaker that hung onto the delicacy of the leading edges of bass notes with the pedal pressed to the metal.

Occasionally−very occasionally−I do find such a paradigm: "There was suddenly a moment in the Ashkenazy Sibelius 5−the double-bass discord in the last movement before the final flare-up−when I realized that I was hearing low-frequency reproduction quite an order of magnitude more transparent than I had previously experienced from an apparently conventional loudspeaker", read the start of a loudspeaker review I wrote almost five years ago (footnote 1). That discord, underlying the symphony's sense of rhythmic and tonal uncertainty, concerns two low notes a major second apart, E-flat and F, which are almost identical in frequency but far enough apart in pitch to be distinguished. Unfortunately, distinguishing those notes is a hurdle at which just about every dynamic loudspeaker to feature true 20Hz extension falls. But the speaker that was the subject of that review, the original KEF R107, was hardly ordinary; it handled its task with harmonically accurate aplomb.

It was therefore with considerable anticipation that I awaited the arrival of the R107/2 in my listening room...

Means of Ascent

In a 1979 AES paper entitled "A bandpass loudspeaker enclosure" (footnote 2), KEF's Laurie Fincham offered a theoretical analysis of a different kind of woofer loading. In a normal reflex enclosure, a port added to a sealed box allows a cavity resonance to extend the woofer's response downward by a useful amount. But what, asked Fincham, if the woofer were mounted so that it had no direct connection with the outside world, instead firing into an internal cavity which communicates with the listening room via a port? In this manner, the air in the port would act as the bass driver proper, having very low mass and capable of very high excursions. In an ideal world, this will result in lower distortion and a higher sensitivity than an equivalent sealed-box alignment using the same drive-unit (equivalent, that is, in that it features the same lower cut-off frequency and that the total internal enclosure volume is the same). The unit's intrinsic response will resemble a band-pass electrical filter centered on the port-tuning frequency, with smooth out-of-band behavior.

In the real world, you will not be surprised to learn that things are not so simple, the bandpass enclosure suffering a propensity for coloration-inducing high-Q resonances above its passband "at frequencies where the wavelength of the sound is less than eight times the smallest dimension of the second cavity", found Mr. F. But hey, if you can arrange for those resonances to be pushed into a corner, you've got one hell of a woofer! Which is why the '80s saw a proliferation of such devices, both from KEF in the form of otherwise conventional loudspeakers and from the manufacturers of those cute little three-way systems that promise everything but deliver...well.

The original R107 and the new R107/2 implement the bandpass woofer with two 10" pulp-cone units mounted vertically in separate sealed internal enclosures and firing in push-pull into a cavity. The driver frames are linked by a nonmagnetic tie-bar, and as the reaction motions in the frames are identical and out of phase, they cancel, leading, it is to be hoped, to minimal excitation of the enclosure walls. The bandpass cavity communicates with the outside world via a port sited on the top of the bass enclosure to position the effective bass driver very close to the midrange unit. (The vent opening is covered with black wire mesh to prevent animals and small children from venturing inside).

The R107/2 reveals its evolution from KEF's R105 of 1977 with its swiveling treble/mid "head" unit, which connects to the bass bin via a gold-plated XLR plug. The head is constructed from a honeycomb plastic material, its cavity filled with a mineral-loaded polymer damping compound. However, whereas the 105 used a B110 driver, the R107 head uses a polypropylene-cone midrange driver, this mounted on rubber O-rings to give a degree of compliance between the driver chassis and the enclosure to eliminate specific resonances. The original R107 used KEF's T33 tweeter, a large soft fabric-dome unit which I and other reviewers felt not to be equal in terms of clarity to the rest of the drivers used. Perhaps not uncoincidentally, the revised 107 features a metal-dome tweeter, this featuring a short horn flare around the dome and constructed on a cast front-plate with "KEF" prominently displayed. Surprisingly, the tweeter magnet is still marked "T33". More surprisingly, some of the bolts holding the tweeters in place were a little loose, presumably from shipping vibration. (The midrange bolts, which I assume have to be torqued correctly to give the right behavior with the O-ring gaskets, were all screwed tight).

Footnote 1: HFN/RR, July 1986, p.65. −John Atkinson
Footnote 2: Presented at the 63rd Audio Engineering Society Convention, May 1979. Preprints are available from the Audio Engineering Society, 60 East 42nd Street, New York, NY 10165. Tel: (212) 661-8528. Fax: (212) 682-0477.
−John Atkinson

Apart from the new tweeter, other changes between the original 107 and the 107/2 involve a completely new crossover (although this still features KEF's Conjugate Load Matching, where additional circuit elements are used to make the unit's impedance fundamentally resistive and constant in nature), said to eliminate a slight thinness in the earlier model's lower midrange, and a new KUBE equalizer.

The 107/2 KUBE is less complex than the one supplied with the original version of the speaker, which offered control of both LF Q and extension. The new equalizer has just two rotary controls, marked "LF Contour" and "HF Contour", the former offering tailoring of the lows to best match the room acoustics and needs of the program material, the latter optimum voicing of the mid-treble to suit the "liveness" of the room. A button marked "Bypass" switches on an amber LED when pressed; the naive audiophile (or reviewer) might think that it also removes the equalization. Actually, it only switches the unit's rotary controls out of circuit, allowing their effect to be quickly evaluated. It also led to minutes of fun as, not realizing this−I hadn't read the comprehensive handbook at this point−I switched this button out with the rotary controls at their center position and, perplexed, couldn't hear any difference at all!

The KUBE, which has a separate remote power supply, sports enough pairs of in/out RCA jacks that it can be used either between pre- and power amplifiers or in the preamp tape loop, where it replaces the preamp's tape monitor functions. It also offers a duplicate set of adjustable-level equalized outputs, to facilitate bi-amping of the 107/2s, and a set of fixed-level unequalized outputs so that the system can also drive a second, different amplifier/speaker combination in another room.

Internally, the KUBE's circuitry is neatly laid out on a single double-sided printed circuit board. Component quality is high, with a reasonably sized, regulated power supply and Signetics NE5532 dual op-amp chips used to implement the EQ and output drive. However, given the success aftermarket manufacturers have had in the US offering equalizers for the B&W 801 (Denver dealer Listen Up's sonically superb MaughamBox comes to mind); wonder how long it will be before KUBE substitutes appear for the R107/2.

The review sample bass bins were finished in a grain-matched walnut veneer, with electrical connection via two pairs of gold-plated Michell binding posts on their rears. (As supplied, these are joined by jumpers, easily removed for bi-wiring). I've never been happy with these posts, finding that the knurled nuts tend to work loose over time with vibration. With the review 107s, fully tightening the nut actually started to rotate the entire post, but didn't affect the auditioning. These posts also accept 4mm banana plugs.

The Path to Power

Source components used during this review consisted of a Linn Sondek Lingo/Ekos/Troika setup sitting on an ArchiDee table to play LPs, a Revox PR99 to play 15ips master tapes, and either a Meridian 208 (Bitstream) CD player or the Stax DAC-X1t processor driven by the Meridian 602 transport to play the little silver devils. Line-level sources were fed straight into my Mod Squad Line Drive Deluxe (this recently updated to superb early-1991 standard, with solid-core internal wiring and Cardas jacks), which in turn fed the KEF KUBE equalizer via 1m of AudioQuest Lapis interconnect. Phono preamplification consisted of the Expressive Technologies transformer hooked into a Mod Squad Phono Drive EPS. Power amps used included a Mark Levinson No.23.5, an Audio Research Classic 60, and a pair of VTL 500W monoblocks, all connected to the preamplifier via 15' lengths of AudioQuest Lapis unbalanced interconnect. Speaker cable was 5' bi-wired lengths of AudioQuest Clear.

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.

Psychotic Reactions

With the help of Sitting Duck Software's "Listening Room" program (footnote 3), I ended up with each speaker some 60" away from its sidewall and 37" from the front of the LPs that line the rear wall, giving a speaker-to-listener distance of almost 9'. The sidewalls of my dedicated listening room also have almost floor-to-ceiling bookshelves at the points where the speakers would otherwise produce strong reflections, and the room features 16"-diameter ASC Tube Traps sited at strategic points to even out its upper bass and lower midrange response. A recent addition to the room's acoustic treatment is a pair of RPG Abfusors behind the listening chair, which both improves the perceived depth of recorded image and ameliorates a slight tendency toward brightness with some loudspeakers. Having found the optimum positioning for the 107/2s, I removed the covers from the front spikes but placed two brass cones from German Acoustics under the rear feet of each speaker so that the reference axes were aimed at my ears−you should be able to look the name badges in the face.

First things first: I thought I would assess the optimum LF Contour setting with the bass warble tones on the Stereophile Test CD. Into the Meridian 602 went the CD. "Oh-oh, a rattle". The fix proved easy. The midrange unit and tweeter are covered by curved wire-mesh grids push-fitted into slots around the driver front-plates. It was one of these that was buzzing slightly; no problem to remove all the grids. (The tweeter remains protected from nosey fingertips by an additional fine-wire mesh). The smoothest, most musically appropriate lows were obtained in my room with the LF contour all the way up for classical music, though I tended to back this off to the 3 o'clock setting for rock music, so that recorded kickdrum retained its transient snap. With the HF Contour at 12 o'clock, the tonal balance was a little reticent. Setting this control was a little tricky, the exact amount of boost depending on whatever record was playing. Ultimately, I decided on 2 o'clock, which added a suitable but subtle smidgin of "szing" to the mid-treble. (Without the Abfusors, a 10 o'clock setting was more appropriate).

I decided to run the speakers in with pink noise overnight for two nights before proceeding with any serious listening. I couldn't help copping a listen, however. Now pink noise is a cruel test of a speaker's midrange smoothness. Yet the 107/2s offered a smooth, characterless presentation that boded well for the moment when I would actually play music through them. Leaving the pink noise playing and reaching for the stethoscope revealed the bass bin to emit very little of anything at all, the rear wall vibrating between 160Hz and 190Hz and at 270Hz and the side walls between 90Hz and 170Hz, but this was all to a very small degree. Promising. The head unit was more lively, a deep "aww" sound audible particularly from its rear panel through the stethoscope due to cabinet modes between 150 and 300Hz, but this was with the pink noise playing at a high level. It remained to be heard whether this would have an effect on music.

Footnote 3: Reviewed in December 1990 by Thomas J. Norton, this is an inexpensive ($29.95) program for IBM PCs and compatibles. −John Atkinson

Time for truth. I reached for that same Sibelius 5 recording (London 410 016-2). Have you noticed that when something is working right in the system, it seems hard to concentrate on the hardware? Such was the case here. By the end of the symphony, I had forgotten what exactly it was I was supposed to be listening for.

Oh yes, that bass discord. I put the third movement on again, this time trying to shut out the music. Mmmmm. Just as I remembered it from the original 107, the E-flat and F close but far, melded but differentiated−but there's so much more going on, so much wider a window into the music...

Bass instruments in real life have so much more complexity to their sound than the mere fact that they pump out low frequencies. The sound of the double-bass has a woody formant structure surrounding the underlying tubbiness that is hard for full-range speakers to reproduce in the right proportion; 16' organ pipes should sound like their sound is produced by a vibrating air column, not by a sinewave oscillator; the sound of the bass drum should remind you that a beater or pedal striking a plastic or calfskin diaphragm is responsible for its sonic weight. Even the sound of the electric bass should reveal the intrinsic nature of the loudspeaker through which its sound is reproduced−even if it is an underdamped thunderbox with severe resonant modes in the presence region, like my Fender Bassman cabinet.

The R107/2 easily allowed you to hear all that richness. From the dry kickdrum and bass guitar on the B52's remix CD (footnote 4) ("Party Mix", Reprise 9 26401-2) to the magnificent thunder of Terry Bozzio's bass drum on Jeff Beck's Guitar Shop (Epic EK 44313), the woody weight of the double-bass on the superb new Oxnard Sessions, Vol. 1 album from pianist Mike Garson (Reference Recordings RR-37), and the deep barking nature of the big pipes on the Jean Guillou Pictures at an Exhibition transcription (Dorian DOR-90117), bass instruments reproduced via the KEFs with both their essential characters and the differences between those characters preserved intact. Weight, power, and definition. In fact, the only dynamic loudspeaker that I have heard to equal the 107/2 in this respect is the Thiel CS5, and that with the right amplifier, right cables, and most importantly, in the right room. The 107/2 seems much less fussy with respect to such demands.

One of the most adrenalin-raising rock concerts I ever went to was the Grateful Dead at London's Rainbow Theater back in the Spring of '81. Bassist Phil Lesh−one of the most intelligent bass players to emerge from the '60s with brain cells intact, to judge by his interesting choices of exactly what notes (footnote 5) to slot into the overlying harmonic structure−was producing a pretty neutral sound from his stack, with the exception of his low E-string register when the speakers, hit with pure power at their LF resonant frequencies, produced a throbbing pulse that rolled forward from the stage and crawled across the floor toward you. The San Francisco and New York concerts from that Dead tour were preserved on the Dead Set album (Arista ARTY 11); while minimonitors almost always fail to reproduce any of that uncontrolled flatulence, most full-range speakers obscure it with their own problems in the same region. The R107/2s get it just right: under Jerry Garcia's and Bob Weir's meandering guitar chopping in the center of "Friend of the Devil," Mr. Lesh's formidable formants can be heard to perfection.

But I mentioned earlier the window the 107s open into the music. If you're to believe the technologists, the reason planar speakers produce impressive soundstage depth is because of the presence of strong reflections from the wall behind the loudspeakers. Neatly put; sounds convincing; end of story.

But if that's truly the case, Mr. Engineer, then why should the R107/2s also throw a soundstage so deep that it extends way behind the rear wall of my listening room? Indeed, the second "Promenade" on the Dorian Pictures album starts with a vibrato reed stop carrying the melody set so far behind my listening room wall that it's actually in Colorado! I have rarely heard such excellent layering of depth from recordings that have the stereophonic sonic wherewithal. Again, Professor Johnson's miking on The Oxnard Sessions album must have been inspired. In its solo outings−"Lady Be Good," for example−the piano hangs between and behind the KEFs, not just palpable but "material," "substantial," "tangible," "tactile," to give my thesaurus a work-out. And the drums behind the piano, the bass to its side and in front, the alto sax to the other−there's enough "there" there for it to hurt.

But I gibber. Is this sense of presence exaggerated? I put on my own recording of Anna-Maria Stanczyk performing Chopin on the Stereophile Test CD. Again, the piano image was set well back; again the sense of palpability, of solidity. But as this sound was what I had hoped to capture when I made the recording, I conclude that the R107/2s are, in fact, merely accurate: merely reproducing what was there to be reproduced.

This precision of imaging I noted with music recordings was confirmed by the tests on the Chesky Test CD (JD-37). The vertical imaging signal test reproduced better than I have heard it from any other speaker in my system, the sampled and psychoacoustically processed cabasa sound riding smoothly above the loudspeaker positions almost to the ceiling, with only a slight lateral wavering noticeable, while the "Over" signal reproduced as a smooth arch between the speakers. Laterally, too, the Chesky signals precisely mapped out a beautifully defined stage, though not one that extended much beyond the speaker positions. I also noted this with the Q-Sound-processed Soul Cages album (A&M 75021 6405 2) from Sting (another thinking man's rock bassist, particularly to judge from this album). With one or two exceptions, such as the final, whispered "Good Night," which appeared to come from somewhere in the vicinity of my left ear, the much-vaunted ability of Q-Sound to produce steady images way beyond the speaker positions was not particularly apparent via the KEFs. Are the speakers right and Q-Sound wrong? I don't know. Do you?

Reviewers are supposed to split the tonal balance of components into neat digestible bands in their reviews; I guess I should do the same before you lose patience. The midband of the 107/2 was free from coloration−no "aww," an absence of "eee"; instruments and, more importantly, voices sounded correct. The treble, too, seemed clean and unexaggerated, without any spit or spitch. Although in absolute terms, there is a lack of air in the top octave, this is one neutral-sounding speaker.

Footnote 4: I reach for this album, particularly "Mesopotamia," whenever I feel that I am starting to take high-end hi-fi just a bit too seriously. His subject was Jethro Tull, but the late Lester Bangs may well have been talking about the yet-to-be-conceived High End in 1973 when he wrote, "You can reach a point...when the trappings and tinsel and construction become so important that it really doesn't matter what's inside." −John Atkinson

Footnote 5: Try fooling around on a handy keyboard or guitar, underlaying an F-major triad with a pedal C or G, then a C-major chord with a pedal B, E, or D, to get the effect Lesh produces from what are supposed to be orthodox changes. The "wrong-note" bass notes suggest suspensions and tensions, producing paradoxes that add depth to the mundane, interest to the ordinary. Indeed, in addition to providing the rock upon which the music sits, one of the rock bass player's essential tasks is to reveal unsuspected facets to the music in the manner in which he (or she, not forgetting Carol Kaye) gets from tonic here to dominant there, and back again. −John Atkinson

The adjective "neutral," however, while held to be a positive quality in a hi-fi component, can also imply a lack of commitment, a shrinking from extreme positions. Thus it was with the KEFs. "They're a little polite?" ventured Tom Norton after I forced him to listen to the Dorian Pictures. Yes. They are. Compared with, for example, the Avalon Eclipses or Wilson WATT 3/Puppy combination, the R107/2s lose a little immediacy, offering a less dynamically vivid sound picture. They are also less forward-balanced than the similar-in-concept B&W 801 Matrix. But they still present all of the four-in-a-bar rhythmic drive on the Reference Recordings Garson album, and all the power of the live Jimi Hendrix (footnote 6) recording at the LA Forum in the new Lifelines box set (Reprise 9 26436-2).

"They don't quite have the seamless midband magic of the Apogee Stages on voice," added Tom, warming to his subject after I had subjected him to "Suo Gan" from the Empire of the Sun soundtrack (Warner Bros. 9 25668-2). That's true, too. There's a residual grain in the midband, at times almost a very slight hardness, that you don't get with the Apogees.

But, Major Tom, this is a moving-coil speaker. And it has bass definition and extension to wipe the Apogees' noses with. With a soundstage presentation to match.

Perhaps I'm getting old, but I can only take so much vividness at one time. Take the Avalons that I reviewed in January−they would make Ted Koppel talking about the savings and loan scandal sound exciting. But with the 107/2s, I spent hours digging out recordings that I had hitherto found unlistenable. I haven't played the 1983 Chandos recording of the Shostakovich G-minor Piano Quintet (CHAN 8342) since I bought it, for example, so offputting is the hard, steely sound of the augmented Borodin Trio within the bathroomy acoustic of London's Church of St. George the Martyr. But via the KEFs, their ability to decode space made sense of the reverberation and their smooth tonal balance tamed the sound, allowing me to appreciate the fact that all these years, I have had a superb performance sitting unheard on my shelves. Better the vividness be in the music than in the speakers, say I!

All the King's Men

Toward the end of Robert Penn Warren's perfect prose poem on the clash between reality and purity, he writes: "You meet a corner at a party, while the glasses clink and somebody beats on a piano, you talk with a stranger whose mind seems to whet and sharpen your own and with whom a wonderful new vista of ideas is spied." To me, when your system is singing, when it features well-matched high-end components that are musically (as opposed to technically) accurate, this is the true impact of recorded music. A stranger to the music when you buy the LP or CD, the mere act of putting your new purchase in the player, or on the turntable, opens you up to new vistas that lead to new and unimagined depths of intellectual emotion plumbed, to new communions with the music's creators, that lead in turn to new recordings, until you can't believe how late the hour, how unfatigued the brain−how wiped-out the psyche.

But then, the next morning, the apprehension, the anxiety. Was it just that you'd been both receptive and forgiving, your selective deafness tuning out the hi-fi, focusing in on the music? As Mr. Warren subsequently wrote, "Then afterward...when you meet again...something happens, or almost always happens, to the gaiety, the brilliance, the communion...You remember the steps of the dance but the music isn't playing any more." It's the same with your system: you put one of the previous evening's records on−the magic's gone. The system wheezes and gasps, booms and buzzes−the music isn't playing any more. The record sounds like nothing at all, especially it doesn't sound how it did the previous evening. Musical misery! Audiophile angst! (Stereophile's search for recommended components could be summed up as our looking to minimize such common complaints).

But that's with lesser speakers than KEF's latest incarnation of the R107. Throughout the time I spent with these speakers, record led to record, CD to CD. Sure, the 107/2's balance is polite, its dynamics a little restrained. But not at the expense of the musical sweep, the emotional dance. The speaker's big-boned sound, raw when required, civilized when appropriate, generous always, effortlessly opened a deep, well-defined passageway to the heart of whatever record I played. Recommended as one of the few full-range dynamic loudspeakers that I have experienced to touch the soul−not just of the music, but of the listener. −John Atkinson

Footnote 6: Hendrix aficionados have no option but to check out Charles Shaar Murray's Crosstown Traffic (1989, Faber & Faber) for what is possibly the most insightful analysis of post-1950 popular music since Charlie Gillett's The Sound of the City. −John Atkinson


Looking first at the performance of the KUBE, its distortion and input and output impedances measured to spec, while fig.1 shows its fundamental response with both contour controls set to their 12 o'clock positions. Note that the treble is shelved down by 1dB while the gentle, 2dB boost below 1kHz is presumably to compensate for the midrange driver's diffraction loss. The sharp rise in output level below 50Hz is to extend the speaker's bass below the bandpass enclosure frequency (where a slight, 2dB suckout compensates for what must be a slightly underdamped intrinsic tuning).

Fig.1 KEF R107/2 KUBE, frequency response with variable controls bypassed (2dB/vertical div.).

Plotted with a slightly different vertical scale, fig.2 shows the KUBE's effect with both controls set to their maximum and minimum positions. The HF contour can be seen to boost or depress the speaker's output over an octave or so in the presence region, centered on 4.5kHz, while the action of the LF contour control is a little more complex, the position of the upper-bass peak changing with control position as well as the amount of low-frequency boost altering. Note that with the control set to maximum, there is almost 20dB of gain at 10Hz: the R107/2 owner is advised to use a turntable with low rumble and a well-matched arm/cartridge if his amplifier is not to spend most of its resources amplifying LF mud. Or do KEF owners just use CD sources in these days of analog twilight?

Fig.2 KEF R107/2 KUBE, frequency response with variable controls at maximum and minimum settings (5dB/vertical div.).

Turning to the R107/2, its impedance is shown in fig.3. The minimum value is 3.6 ohms at 940Hz. Note how uniform both impedance amplitude and phase are from 20Hz to 2kHz, the former varying less than half an ohm from the 4 ohm specification due to KEF's Conjugate Load crossover. Even moderately priced amplifiers, provided they have no trouble driving a 4 ohm resistive load, should be capable of raising high spls with the 107/2, particularly with a measured sensitivity (using a 1/3-octave warble tone centered on 1kHz) of 89dB at 1m with 2.83V input. (Note that the speaker's 4 ohm impedance means that it is drawing 2W from the amplifier at this drive voltage rather than 1W) .

Fig.3 KEF R107/2, electrical impedance (solid) and phase (dashed, not valid above 2kHz). (2 ohms/vertical div).

Looking at the speaker's impulse response (fig.4), this was taken 44" away on the HF axis (ie, just above the reference axis) with the KUBE controls centered and with the speaker on a high stand to move the first boundary reflection (from the floor) a little farther back in time. This floor reflection can be seen just after the 7ms mark, but note also a couple of reflections about 1ms after the initial impulse. These could well be from the top edges of the woofer enclosure; looking at the Energy-Time Curve (not shown) reveals these reflections to be 25dB down in level, however. Other than that, there is very little to note from the impulse response other than a small amount of ultrasonic ringing from the tweeter, this well-damped in comparison with other metal-dome units, and a rather lazy decay of low-frequency energy from the high-order crossover filters employed. The step response is shown in fig.5.

Fig.4 KEF R107/2, on-axis impulse response at 50" (5ms time window, 30kHz bandwidth).

Fig.5 KEF R107/2, on-axis step response at 50" (5ms time window, 30kHz bandwidth).

Transforming the impulse response to the frequency domain courtesy of Monsieur Fourier and averaging five such measurements on the HF axis across a 30 degrees horizontal window gives the quasi-anechoic response shown in fig.6. Commendably smooth, there is a slight lack of energy in the lower part of the tweeter's passband which might correlate with my feeling that the speaker's sound lacked immediacy with the KUBE's HF contour set to 12 o'clock. But a look at fig.7 shows the way in which the 107's response changes from directly on the HF axis to 30 degrees horizontally off that axis−not much! (The rearmost curve is a straight line because, rather than show the actual responses in this diagram, I have normalized them to the on-axis curve so that only the changes in response will appear).

Fig.6 KEF R107/2, anechoic response on-axis at 44", averaged across 30 degrees horizontal window and corrected for microphone response (KUBE controls centered).

Fig.7 KEF R107/2, lateral response family at 44", normalized to response on tweeter axis, from back to front: differences in response 30-7.5 degrees off-axis, reference response, differences in response 7.5-30 degrees off-axis (KUBE controls centered).

With such a wide-dispersion design, it will be essential to keep the speakers away from sidewalls (or to arrange for those sidewalls to be diffractive) if the imaging is not to become smeared. Vertically, fig.8 shows the anechoic response 7.5 degrees above the HF axis (rear), on the midrange axis, and 7.5 degrees below the midrange axis (front), again all normalized to the response on the HF axis (straight line). You can see that above and below the preferred axis, notches develop at the crossover frequency, with that above being more severe, corresponding to a hollowness to the sound heard on pink noise on this axis. The listener should arrange to sit with his or her ears approximately level with the speakers' name badges to get the most neutral tonal balance. Note, however, that the response on the midrange driver axis is almost identical to that on the HF axis, suggesting that the optimum vertical window is not too critical.

Fig.8 KEF R107/2, vertical response family at 44", normalized to response on tweeter axis, from back to front: difference in response 7.5 degrees above axis, reference response, differences in response 7.5 degrees-15 degrees below axis (KUBE controls centered).

Next I looked at the individual responses of the bass bin (taken in the nearfield below 150Hz) and the midrange/treble head (taken again at 44" on the tweeter axis). These are shown in fig.9 with the respective levels arbitrarily plotted to agree with the specified 160Hz crossover frequency. Again, the midrange and treble are smooth, with a shallow but broad suckout in the presence region. In the bass, the more extended trace shows the response with the KUBE in circuit (with the LF contour set to 12 o'clock). This is ostensibly flat to 25Hz or so, gently rolling off below that frequency to -6dB at 19.5Hz. The other bass trace shows the woofer response with the KUBE transfer function subtracted from it by the MLSSA software. (Remember that the KUBE's bypass switch does not eliminate the low-frequency equalization.) The intrinsic woofer-port response can be seen to peak slightly at 48Hz and roll out at approximately 12dB/octave, much as if it were a sealed box.

Fig.9 KEF R107/2, individual anechoic responses of head unit and bass bin at 44". The latter is plotted in the near field below 160Hz, with KUBE controls centered (extended trace) and out of circuit (less-extended trace).

In-room, these quasi-anechoic measurements translate to a flat frequency balance, with excellent low-frequency extension, as shown in fig.10. This measurement, which I have found to correlate well with the perceived tonal balance of a speaker in my room, shows the 1/3-octave in-room response spatially averaged across a 6' by 2' window at the listening position to minimize the effect of room standing waves. No wonder the R107/2 sounds so smooth, so neutral!

Fig.10 KEF R107/2, spatially averaged, 1/3-octave, free-field response in JA's listening room.

Finally, the cumulative spectral decay plot for the anechoic part of the 107/2's impulse response (fig.11) shows a relatively clean decay, although there is a little more hash in the presence region than I would have expected. (The black ridge at 16kHz is due to the computer monitor's line scanning and should be ignored.) The small reflections I mentioned in my discussion of the speaker's impulse response (fig.4) manifest themselves in this plot as the time-axis ripples in the upper midrange, around the cursor position at 1240Hz. Their effect is hard to predict, however. Certainly they do not compromise the superb imaging results I heard with the Chesky Test CD.

Fig.11 KEF R107/2, cumulative spectral-decay plot at 50" (0.15ms risetime). −John Atkinson

Thomas J. Norton, Stereophile 10/95:

It's a little mind-bending to recall that KEF first introduced their Model 107 loudspeaker in 1986. KEF, like most British loudspeaker manufacturers, tends to be conservative when it comes to its high-end models. Why mess with a good thing? The Model 107 has never been a radical design"that wouldn't be KEF's style. But the original 107 built on a solid background of innovation from KEF products in the immediately preceding years.

Take the twin coupled-cavity enclosure. First used by KEF in 1984 for the Model R104/2 and generically known as a bandpass enclosure, the concept was originally patented by Andre d'Alton in 1934. Interest in the technology has exploded in the past 15 years, arguably triggered by Laurie Fincham's (then of KEF) "A Bandpass Loudspeaker Enclosure" paper, which was presented at the 63rd AES convention in 1979.

Simply described, a bandpass enclosure has two separate compartments, with the driver mounted on the internal wall that divides them. The compartment behind the driver acts more or less as a conventional cabinet, bass-loading the woofer. It may be sealed or ported. (The Bose Corporation owns a 1985 patent on the front- and rear-ported bandpass enclosure.) The enclosure in front of the driver is always ported by a large duct to the exterior of the cabinet. One important characteristic that results from this arrangement is that the high-frequency output of the driver that reaches the outside is rolled-off acoustically. This greatly simplifies the design of (or even eliminates the need for) a conventional electrical low-pass filter for the woofer.

While it's not without problems (notably the potential for resonances to find their way out of the port and into the audible output of the loudspeaker), proponents of the system argue that its positive attributes greatly outweigh the drawbacks. Claimed gains include all the usual suspects: higher efficiency, greater extension into the deep bass, lowered cone excursion for a given sound pressure level, and improved power-handling.

In the 107's implementation of the bandpass system, two 10" woofers are each mounted in their own separate, sealed chambers. The front of each driver feeds into a third chamber, which is in turn vented to the outside at the top of the enclosure, immediately adjacent to the head-unit that contains the midrange unit and tweeter. A rigid rod−which runs through the voice-coil centers−links the two woofers to cancel any differential movement between their frames.

Two other features of the Model 104/2 found their way into the original 107: conjugate load-matching and electronic bass equalization. The former is a crossover configuration that results in a load that remains very close to the target impedance of a resistive 4 ohms across the frequency range; the latter uses an outboard, powered network (the KUBE) to tailor and extend the low-frequency response. Two controls−one for bass, the other for treble−were provided on the KUBE as a supplement to its fixed-response contouring. These were included to provide for a limited degree of room compensation. The KUBE was designed to be placed either in the tape loop (an alternate tape loop was provided on the KUBE itself) or between the pre- and power amplifiers.

The 107's 5" polypropylene-cone midrange and 1" dome tweeter were mounted in a separate enclosure atop the woofer cabinet. This small enclosure, contoured to reduce diffraction and damped with a mineral-loaded polymer, could be rotated separately from the bass cabinet to facilitate proper setup.

Updated in 1990 to Model 107/2 status, with a revised crossover, a metal-dome tweeter, and biwiring terminals, the 107 sailed into the mid-'90s facing increasingly stiff competition−not only from other manufacturers but also from other KEF models, such as the new Reference Model Four. Termination (or, as the British might put it, redundancy) seemed an imminent possibility.

But with further refinements to the midrange and tweeter balance, along with changes in the head-assembly damping, KEF readied the 107−now known as the Reference Series Model 107/2 Raymond Cooke Special Edition−to do battle with the forces of 1995. Is the result as impressive as the length of its model name? Does it do justice to the name it bares−that of KEF's late founder and president, Raymond Cooke? And does it justify the KEF's $5900/pair price tag?

Murphy Rules

I set the 107/2s up in my 18' by 26' by 11' listening room, well out from the walls and firing down the long dimension of the room. The speakers come with large grillecloth-covered hoods that may be used to cosmetically conceal the mid-tweeter modules; I used them grille-less. Before doing any listening, I ran them in for a long weekend with pink noise played at moderate levels.

My first listen wasn't too encouraging. The sound was rather bloated and forward, with little transparency. I fiddled with the KUBE controls, but they didn't solve the problem. The timbre seemed reasonably consistent from left to right (the importance of this observation will soon be clear), but a day or two of listening convinced me that I needed to try a change of positioning.

This wasn't the first time a loudspeaker has been unhappy with this room arrangement, though my recent experiences with the Thiel CS7 (reviewed elsewhere in this issue) confirmed that it could work well with the right loudspeaker. The next day I put the 107s in the cross-room (diagonal) position that has worked well for me with numerous other loudspeakers. This arrangement has proven to slightly lean-out the mid- and upper bass"a balance change that seemed to be what the KEFs required.

"Whoa! What happened?" I blurted aloud to no one in particular as the first bars of music rolled toward me. The left-right spectral balance was now all wrong. One channel appeared to have lost its top end, with a resultant skewing of the sound to the unaffected side. I listened more, to other material. Same result. I pulled the speakers out and set them close together. No change. I reversed the midrange/tweeter heads (they come packed separately and may be easily installed or removed by the user). The problem moved with the change in heads, indicating that it lay with the tweeter itself. Swapping amplifiers left to right only confirmed this observation. Somewhere between the initial listening test and the new setup, one of the tweeters had lost most of its output. Why this happened remains a puzzle−one that will likely go unresolved. Perhaps it was a flawed sample that, while it appeared fine on the assembly line, was weakened by the break-in and finally failed at the very end of my first listening session. Possible, I suppose, but certainly a first in my experience.

In any event, KEF's preferred solution was to send me an entirely new pair of loudspeakers. As deadline time was rapidly approaching, and there were apparently no 107/2s in stock either in the US or the UK, the factory had to build us a pair on short notice and air-freight them from England to Santa Fe. They were so new I could still smell the finish when they arrived. Fortunately, it was dry. Murphy just wouldn't quit, though. When I hooked the new pair up, they sounded out of phase. My external hookup was correct. Furthermore, listening to the bass alone revealed that the woofers were in phase. Apparently, the midrange/tweeter phase was internally reversed on one of the loudspeakers. But which one? Reverse the wrong one, and the relative phasing between the midrange/tweeter and woofer would be wrong on both channels, even while the left and right woofers and left and right midranges/tweeters were in proper phase with each other.

Fortunately, with direct access to the midrange/tweeter phasing via the bi-wire inputs, this was an easy problem to solve. Reverse the wrong midrange/tweeter, and the whole sound was simply wrong, with a lower midrange suckout that resulted in a clear hollowness−particularly noticeable on male voices. The Orson-Welles-as-Don-Knotts effect.

As to any quality-control problems this internally mis-wired loudspeaker might imply to the reader, I must emphasize, in fairness to KEF, that this second pair of 107/2s was assembled in haste. If I have to get a set of loudspeakers with a problem, I'll take a problem with an easy, on-site solution every time. Once this was all ironed out, the second pair of KEFs functioned flawlessly.


The Model 107/2s were auditioned in a system fronted by a Denon DP-S1 transport and Mark Levinson No.36 D/A converter, linked by Kimber AGDL digital coaxial cable. The amplifier was a Carver Lightstar. TARA Labs Master RSC (unbalanced) interconnect tied the Levinson converter to my reference Rowland Consummate preamp, with the preamp-to-KUBE and KUBE-to-power-amp interconnections via Cardas Hexlinks. Loudspeaker cables were Monster M1.5s.


"Whoa," I blurted again, only this time I was smiling. Not only was everything finally functioning, but functioning in a sit-up, take-notice kind of way. No, the sound didn't blow me out of the room, make my jaw drop, or generate any other kinetic activity. Remember, I had just finished evaluating another terrific set of loudspeakers: the Thiel CS7s. The sound from the KEFs was noticeably different than the Thiels, in ways that I hope will become clear shortly. The important point, however, is that I was attuned to a high standard, and the KEFs did not disappoint.

Gone was the rather bloated quality I had heard from the first pair of 107/2s in the first setup. The sound was now well-balanced, dynamic, and three-dimensional. A minor lack of soundstage imprecision was quickly cured by a slight repositioning of the right-channel loudspeaker. (Turns out it was a couple of inches closer to the listening seat than the left.) A trace of edge in the mid-treble quickly receded into the background as the speakers began to break-in.

Nothing in the 107/2's performance immediately called attention to itself. This loudspeaker was easy to listen to−perhaps even a bit old-fashioned in its lack of any spectacularly flashy attributes. The 107/2 let the music speak for itself.

The closest the 107/2 came to having a distinctive character was in its immediate, just-slightly-forward, rounded yet richly detailed midrange. Vocals were particularly well-handled−rich and vibrant, with a natural bloom. Coloration was very low, with no trace of nasality, boxiness, or chestiness. The modular cabinet appeared to be doing its job.

Both male and female vocals were equally well-served. From Gordon Lightfoot to Custer LaRue−with stops along the way with Mary Black, The Chieftains, David Wilcox, and others−nothing really caught it out or sounded wrong. Sibilants seemed to be those of the recording; if there was excess fizziness in the program material, the KEF neither covered nor exaggerated it.

Vocal reproduction is very important to me; the performance on vocals of the 107 was first-class. If I had to make a call, however, I'd say that the mids on the Thiel CS7 sounded even less colored than those on the 107/2; but the "colorations" that existed in the KEF's midrange are hard to pin down in words. They were in no way unnatural. In broad terms, the midrange of the 107/2 struck me as perhaps a little more romantic than clinically accurate−certainly a valid design choice.

That quality was enhanced by a bottom end that tilted more toward a full, rich quality than punchy tightness. In this respect, it also differed from the Thiel CS7. But when I say "tilted," I really mean it. In my large listening room, the 107/2 was full-bodied through the bass range, but in no way overblown. Bass definition was good. There was no muddle or confusion. The richness did reduce the overall transparency of the sound compared with that from a loudspeaker with a vise-like grip−particularly through the mid- and upper bass.

I did note a slight sonic fog in this region, particularly when the going got heavy, which seemed to relate to balance rather than congestion−the latter was admirably low, even at uncomfortably high levels. At least some of this may have originated elsewhere in the system; the Denon transport, Rowland preamp, and Carver amplifier all tend more to a sweet, slightly warm rather than analytical sound.

That said, the 107/2's balance was nevertheless particularly flattering to large-scale orchestral music; the character often referred to as the "hum" of the orchestra was definitely there. Combine this with the 107's prodigious dynamic range and big, open quality, and you have a loudspeaker that can definitely sound convincing on the most challenging material.

The 107/2's impressive dynamic range extended all the way to the lower bass. My first reaction to the KEF at the extreme bottom was, "So where is it?" With a rated -6dB point of 18Hz, I was somehow expecting more. Perhaps I've been listening to too many subwoofers recently. But the more I listened to the 107/2, the more I appreciated its power and bottom-end reach. No, it definitely did not equal the "feel" of the very best subwoofers in either extension or the capacity to roll down your socks. But it definitely did go deep, with plenty of punch.

And it handled every musical bass test I could throw at it, from Dafos (Reference RR-CD12) to the opening notes on the Jurassic Park soundtrack (MCA MCAD-10859). It also passed the crunch test with video-based material, emitting no protest when asked to reproduce the falling rock from Aladdin and the dino stomp from Jurassic Park.

But while it did a satisfactory job with the deep bass on the video material, it did not really produce the sort of room-shaking "feel" that gives this sort of material its full emotional impact. I have yet to find a single, full-range loudspeaker that does, however. Even expensive, high-end loudspeakers can use a little added bass oomph with video material. With music, however, a pair of 107/2s will more than hold its own. They beat the Thiel CS7s, in my room, in both bass dynamic range and subjective extension. The Thiels' bass was definitely tighter, however, and their overall sound more open and transparent−at least partly because of this tighter, cleaner mid- and upper bass.

I haven't said much about the 107/2's top end, because there's little to say. It was clean, sweet, and slightly soft at the very top, but in no way dull. Everything was there, but in proper proportion. My listening height put my ears about level with the midrange, which put me just a few degrees off-axis to the tweeter. I experimented briefly with tilting the cabinet forward slightly to listen on a more direct line to the tweeter axis, and while this resulted in a somewhat brighter, tighter sound, some of the 107/2's relaxed quality disappeared in the process. I quickly returned to the standard cabinet setup.

Other observations? The KEFs' soundstage was right up there with the best large loudspeakers. Depth was good, though in my room I didn't find it to stand out in any particular way from that of other good designs, such as the Thiels or the Energy Veritas v2.8s.

As to the KUBE, which is an integral part of the system: some may balk at its combination of ICs and a minimalist power supply (fed by one of those ubiquitous wall-warts), but it did appear to be well-made. Since the system is only properly balanced with the KUBE, I made no attempt to audition the 107/2s without it, as that would prove little.

I did almost all of the listening for the above observations with the KUBE controls set on flat and saw no pressing reason to alter them. I did experiment briefly with them, however, and found that their effect was very subtle−far more evident on pink noise than on music. The changes weren't insignificant, but their usefulness in correcting for room-matching problems appeared to be very limited.

The KUBE, incidentally, has only unbalanced inputs and outputs. This didn't bother me in the least−I have never found significant practical advantages in the home with balanced electronics. But readers who have gone to considerable effort and expense to set up a balanced system may not warm to the 107/2.


When I heard the original KEF Model 107/2s in JA's listening room a few years ago, I found them a little polite and dynamically restrained. I have no such reservations about the newer versions. I suspect the difference is due less to any change in the loudspeaker than to a change to a larger, slightly livelier room and a longer exposure to a wide range of program material.

Polite? Well, if polite means lacking in unnatural vividness and an unnatural "hi-fi" quality, then the 107/2 qualifies. JA subsequently reviewed the 107/2 in May '91 (Vol.14 No.5) and found them to be "one of the few full-range dynamic loudspeakers that I have experienced to touch the soul−not just of the music, but of the listener." A strong endorsement. And, based on my auditions of the latest version, one that I would not argue with. −Thomas J. Norton

−  Established 1973  −
"love what you listen with"
Los Angeles, California USA
FOR SALE - TRADE              

  HOME / Master-Index       

* AllegroSound et al assumes no responsibility for the use or misuse of products, services, info, etc.

design & original content ©AllegroSound, Los Angeles USA. All rights reserved.

−  Established 1973  −
"love what you listen with"
Los Angeles, California USA
telephone 323.960.5014

> home < > home < > home < > home < > home < > home < > home <

design & original content ©AllegroSound, Los Angeles USA. All rights reserved.

click here to contact AllegroSound                click here to contact AllegroSound                click here to contact AllegroSound                click here to contact AllegroSound                click here to contact AllegroSound                click here to contact AllegroSound                click here to contact AllegroSound                click here to contact AllegroSound                click here to contact AllegroSound                click here to contact AllegroSound