Stand Loudspeaker Reviews

Sidebar 3: Measurements

I used DRA Labs' MLSSA system and a calibrated DPA 4006 microphone to measure the KEF LSX's frequency response in the farfield, and an Earthworks QTC-40 microphone for the nearfield responses. After I connected the Master LSX to my network, it checked with KEF that its firmware was up to date—it was (v.3.0). I then opened the Roon app on my iPad Mini and the speaker was recognized by Roon as "KEF LSX" and successfully handled data sampled up to 192kHz via the network. The KEF's volume control was active in Roon as well as in KEF's own Control app, which I used to select inputs. The Master has both analog and digital inputs and I performed most of the measurements using its analog Aux input, which appears to have an input impedance of 6.8k ohms. I repeated some testing via the speaker's optical S/PDIF digital input, which locked reliably to data with sample rates up to 96kHz. I checked with the KEF app how the speakers were set before I started the measurements: Desktop and Wall modes off; Treble Trim, 0dB (the range is ±2dB); Phase Correction, on; Bass Extension, Standard; Subwoofer Output: off (the control panel offers subwoofer polarity, gain, and low-pass frequency adjustments).

One problem reared its head when I began testing: Like the Kii Three loudspeaker, which we reviewed in September 2017 and which is also digitally corrected, there was a significant latency when the LSX was fed an analog signal. I usually set MLSSA to examine the first 10 milliseconds of the sound emitted by a speaker, but in this case all I got was background noise. I increased the time window to 100ms and there was the speaker's impulse response, at the 52ms mark. This means that, fed an analog signal, the LSX delays its output by about 48ms. Peculiarly, when fed S/PDIF digital data, the latency was a more manageable 5ms, which means there will be no synchronization problems when the KEFs are connected to a video player's optical digital output.

There is no convention for specifying the sensitivity of a powered speaker, but when I set the LSX's volume control to its maximum of "100" and fed the speaker's Aux input pink noise at 300mV, the resultant SPL was 90dB(C) at 1m on the tweeter axis. With pink noise at –20dBFS fed to the optical digital input, the SPL at 1m was the same 90dB(C). The enclosure seemed relatively inert with the knuckle-rap test, though when I investigated its vibrational behavior with a plastic-tape accelerometer I found a moderately strong resonant mode at 258Hz on the sidewalls (fig.1), and a higher-level mode at 727Hz on the top panel. While I could hear these modes through a stethoscope when playing a recording of a man speaking, they were inaudible at the listening position, presumably because of the small radiating areas of the affected surfaces.

The blue trace in fig.2 shows the output of the LSX's woofer measured in the nearfield. The minimum-motion notch in its output lies at 58Hz, revealing that this is the tuning frequency of the flared port on the rear panel—a relatively high frequency, but to be expected in a small speaker such as this. The port's output (red trace) peaks between 40 and 80Hz, but a sharply defined resonant mode is visible at 800Hz. Again, I couldn't hear this mode at the listening position, presumably because the port is on the speaker's rear panel. Both the port's and woofer's low-frequency outputs roll off faster than the 12dB/octave slope that is usual in a reflex design. I suspect that this is to protect the tiny woofer from subsonic overload, but it means that the complex sum of the nearfield woofer and port responses (fig.2, black trace below 300Hz) rolls off with a sixth-order, 36dB/octave slope below the port tuning frequency.

The black trace above 300Hz in fig.2 shows the LSX's farfield response, averaged across a 30° horizontal window centered on the tweeter axis. It is astonishingly flat, though some suckouts can be seen in the top two octaves, these interference effects resulting from the coaxial tweeter being mounted at the center of the woofer cone. One surprise in this graph is that the speaker's output drops like a stone above 23kHz, suggesting that the LSX digitizes its analog input to a 48kHz sample rate, perhaps because the Phase Correction was enabled. This might also contribute to the latency I noted earlier. Repeating the measurement with optical digital data (not shown) indicated that the LSX's metal-dome tweeter has a significant oil-can resonance at 37.6kHz, well above the audioband.

The plot of the LSX's horizontal dispersion (fig.3) indicates a well-controlled radiation pattern. The evenly spaced contour lines suggest that a pair of these speakers will offer stable stereo imaging. Fig.3 shows that the on-axis suckout at 8.9kHz tends to fill in to the speaker's sides, but also that the speaker becomes very directional in the top octave, a region where the on-axis response also lacks energy. I suspect that the LSX is balanced to be listened to in the nearfield—on a desk, say—rather than at a distance. I haven't shown the LSX's vertical dispersion, as it was basically identical to the behavior in the horizontal plane.

Turning to the time domain, the LSX's use of digital correction can be seen in its step response (fig.4). The outputs of the tweeters and woofer arrive at the microphone at the same time, resulting in an almost perfect positive-polarity, time-coincident, right-triangle shape. Other than a ridge of delayed energy at the frequency of the on-axis suckout, the KEF's cumulative spectral-decay plot (fig.5) was superbly clean.

I was impressed by the KEF LSX's measured performance, and will explore its sound quality further in a Follow-Up.—John Atkinson