KEF Q900 loudspeaker Measurements
I measured the KEF Q900's frequency response in the farfield with DRA Labs' MLSSA system and a calibrated DPA 4006 microphone. The Q900's voltage sensitivity is specified as 91dB/2.83V/m. My estimate was slightly lower, at 90dB(B)/2.83V/m, but this is still usefully higher than average. The speaker's impedance is specified as 8 ohms, but as fig.1 shows, the impedance drops to 4 ohms in the top octaves and to below 4 ohms in the lower midrange, reaching a minimum value of 3.17 ohms at 160Hz. There is also a combination of 5 ohms magnitude and 42° electrical phase angle at 80Hz, meaning that the KEF does need to be used with a good 4 ohmrated amplifier or receiver.
The traces in fig.1 are free from the small wrinkles that would imply the existence of panel resonances of various kinds. However, investigating the cabinet's vibrational behavior with a simple plastic-tape accelerometer did uncover some pumping of the sidewalls at the tuning frequency of the passive radiator, as well as, on all surfaces other than the baffle, resonant modes at 375, 453, and 512Hz (fig.2). The higher-frequency modes might be of sufficiently high Quality Factor (Q) to have no audible consequences; certainly, Kal Rubinson did not comment on any midrange congestion that might have resulted from their presence.
As KR did mention, the Q900 has four 8" cones covering the lower frequencies: an active midrange driver loaded by the passive radiator below it, and an active woofer loaded by a second passive radiator at the bottom of the front baffle. The saddle at 50Hz in the impedance-magnitude graph (fig.1, solid trace) suggests that this is the tuning frequency of both passive radiators, but looking at the individual nearfield responses (fig.3), the Uni-Q midrange unit's minimum-motion notch lies at 46Hz (black trace), the woofer's at 43Hz (blue trace). In each case, the output of the passive radiator (green and red traces) peaks significantly higher than the active unit's minimum-motion notch. The woofer's output gently rolls off above 120Hz or so, meaning that the midrange driver alone handles the musically critical midrange region (200Hz2kHz). But the midrange unit's response does also extend down to the upper bass, which is why the speaker is referred to as a "two-and-half-way" design.
Note, by the way, that the ultimate low-frequency roll-off of each radiator in fig.3 is closer to 18dB/octave rather than the usual 12dB/octave. This suggests the presence of a large series capacitor in the feed to both the midrange unit and the woofer, presumably to limit cone excursion at subsonic frequencies.
Usually, I calculate the nominal low-frequency farfield response of a speaker with multiple diaphragms by adding the nearfield responses of those diaphragms and weighting each response in the ratio of the diaphragm's diameter (or, if the diaphragm is not circular, the square root of the radiating area), taking into account each unit's acoustic phase. As I always say, this gives rise to a boost in the speaker's apparent output in the upper bass, due to an assumption made by the nearfield measurement technique: that the diaphragm is mounted in a true infinite baffle; ie, one that extends to infinity on all direction.
With the Q900, this calculation gave a large boost centered on the frequency of the peak outputs of the passive radiators (fig.4, green trace below 300Hz). This raised my eyebrows, so I attempted to measure the speaker's true farfield low-frequency response by placing it 8' above the groundeasier to say than do! in the center of my back yard, and positioning the microphone to get the maximum path-length difference between the Q900's direct sound and the reflections from the walls and ground. This response is shown, in fig.4, as the red trace below 300Hz. Though accurate down to only 70Hz, it suggests that the sum of the nearfield responses is over-generous. In the listening room, due to boundary reinforcement, the Q900's low-frequency performance will be somewhere between the two extremes shown in fig.4, depending on the size of the room, but this does suggest that Kal's initial impression, that the balance was bit lightweight, was correct.
Higher in frequency in fig.4, the crossover between the Q900's midrange unit and woofers (green trace) and its coaxial tweeter (blue trace) lies approximately at the specified 1.8kHz, but there is broad overlap between the outputs, suggesting the use of low-order crossover filters. There are some peaks in the midrange unit's output between 5 and 10kHz; while these were audible on pink noise with the tweeter disconnected, they may well be sufficiently low in level when the tweeter is operating to add little to no treble coloration. However, when KR wrote "It was only when I really stressed the Q900 with very high levels that the Uni-Q tweeter became somewhat intolerant," I wonder if it was actually this misbehavior of the midrange driver that contributed to what he heard. The Q900's overall response on the tweeter axis is relatively even, but with a broad excess of energy in the low treble that will add to the sense of recorded detail. Overall, however, the treble is smoother and more uniform than that of some of the earlier Uni-Q speakers I have measured, suggesting that KEF's Tangerine waveguide works as advertised. The tweeter does have some very high-Q peaks present just below 30kHz, though of course these will be subjectively benign.
All the responses mentioned so far were taken with the Q900's grille removed. The effect of that grille is shown in fig.5: It introduces some small ripples in the response, and without it the Q900 does perform better, with a smoother treble. The Tangerine waveguide contributes to well-controlled dispersion in both the horizontal (fig.6) and vertical (fig.7) planes, the former correlating with the excellent stereo imaging noted by KR. The tweeter has wider top-octave dispersion than you'd expect from a 1.5" dome, and this will tend to compensate for the slight shelving down of the output in the same region in fig.4. Fig.7 suggests that the Q900 will not be fussy about the exact listening axis.
In the time domain, the Q900's step response on the tweeter axis (fig.8) suggests that the tweeter is connected in inverted acoustic polarity, the midrange and woofer in positive polarity. The fact that the decay of the tweeter's step blends smoothly with the start of the low-frequency units' step suggests optimal crossover design. Other than the low-level peaks in the mid-treble, the cumulative spectral-decay or waterfall plot on the tweeter axis (fig.9) is superbly clean throughout the midrange and treble.
Despite its affordable price, KEF's Q900 offers superb measured performance, and the new Uni-Q driver, with Tangerine waveguide, gives much more uniform dispersion than have earlier Uni-Q designs.John Atkinson