Sidebar 3: Measurements
I used DRA Labs' MLSSA system and a calibrated DPA 4006 microphone to measure the GoldenEar Triton Reference's frequency response in the farfield, and an Earthworks QTC-40 for the nearfield and in-room responses. The GoldenEar's sensitivity is specified as a very high 93.25dB/2.83V/m; my estimate was close to that at 92.5dB(B)/2.83V/m, which places the Triton Reference in exalted company. Speakers this sensitive are rare. Fig.1 shows how the impedance and electrical phase vary with frequency. The magnitude does dip below 6 ohms in the midrange and high treble, with a minimum value of 3.4 ohms at 320Hz and a combination of 4 ohms and –45° phase angle at 73Hz. Though GoldenEar specifies the Triton Reference's impedance as "compatible with 8 ohms," I would use an amplifier with this loudspeaker that is comfortable driving 4 ohms.
The traces in fig.1 appear free from the glitches that would suggest the presence of panel resonances, but under magnification a slight discontinuity can be seen in both traces between 600 and 700Hz. While the Triton Reference's side panels seemed generally inert, the rear panel was relatively lively between 500 and 800Hz, with resonant modes at 550, 665, and 760Hz. Fortunately, the areas affected are small and face away from the listener; these frequencies are also sufficiently high that they should have no audible effects.
The Triton Reference's plot of lateral dispersion, normalized to the tweeter-axis response (fig.4), reveals that the speaker has smoothly controlled off-axis behavior up to the cursor position just below 9kHz. Above that frequency the GoldenEar becomes more directional throughout the treble, though the dispersion doesn't narrow significantly until above 15kHz. In the vertical plane (fig.5), a suckout develops more than 10° above or below the tweeter axis, which is 41" from the floor. (Work by Thomas J. Norton 20 years ago for Stereophile found that 36" was the average height of a seated listener's ears.) This is why I tilted the speaker forward by placing cones under the rear of its base, to get the optimal treble balance at my listening position.
Fig.6 compares the spatially averaged response of the Triton Reference (red trace) with that of the KEF Reference 5 (blue), which I reviewed in October 2017. (Using SMUGSoftware's FuzzMeasure 3.0 program and a 96kHz sample rate, I average 20 1/6-octave–smoothed spectra, individually taken for the left and right speakers, in a rectangular grid 36" wide by 18" high and centered on the positions of my ears.) While a flat treble response is not what you want to see in a graph like this—a room's furnishings are more absorbent at high frequencies than they are lower in the audioband—this graph explains both why I felt the Reference 5 sounded a little sweet, and why the Triton Reference's treble balance sounded more natural in my room. The GoldenEar has a bit more midrange energy than the KEF and definitely more bass, especially in the 60Hz region, where the nearfield response peaked. (The volume controls for the active woofers were set midway between their maximum and minimum positions.)
Fig.1 GoldenEar Triton Reference, electrical impedance (solid) and phase (dashed) (2 ohms/vertical div.).
The green trace in fig.2 shows the response of the midrange units below 312Hz, measured in the nearfield. It rolls off smoothly below 100Hz, and I've plotted the nearfield response of the woofers (red trace) at a level that suggests that the crossover occurs at 80Hz. This is an arbitrary decision, however, as the woofers' level control has a range of 28dB between its minimum and maximum positions. The output of the woofers rolls off rapidly above and below a relatively narrow passband, as does that of the passive radiators (blue trace). The slight notch at 22Hz in the woofers' response suggests that this is the tuning frequency of the radiators, but peculiarly, their output has a peak at the same frequency at which the woofers' output is greatest. The black trace below 300Hz in fig.2 shows the complex sum of the midrange, woofer, and passive radiator outputs, summed in the ratio of the square roots of their radiating areas. The large peak at 63Hz appears to be due to crosstalk from the woofers, with this nearfield measurement. It was absent when I repeated the measurement using an accelerometer rather than a microphone (fig.3). The Triton Reference's low-frequency output extends to a low 25Hz with then a very steep, sixth-order rolloff.
Fig.2 GoldenEar Triton Reference, anechoic response on tweeter axis at 50", averaged across 30° horizontal window and corrected for microphone response, with nearfield responses of midrange unit (green), woofers (red), and passive radiators (blue), respectively plotted in the ratios of the square roots of their radiating areas below 312Hz, 200Hz, and 170Hz, with complex sum of nearfield responses plotted below 300Hz.
Fig.3 GoldenEar Triton Reference, output of passive radiator measured with an accelerometer taped to the diaphragm (purple) and measured in the nearfield with a microphone (blue).
Above 300Hz, the black trace in fig.2 shows the Triton Reference's farfield response, averaged across a 30° horizontal window centered on the tweeter axis. It is impressively flat, though the mid-treble is shelved down by a dB or so. The speaker's upper-frequency response is disturbed by narrow peaks and dips. These won't be audible as such, but result, as I explain below, from reflections of the tweeter's output from the rather bulky grille. I repeated the tweeter-axis measurement without the grille; the response ripples were absent, and there was about 0.5dB more energy present above 1.2kHz. However, the small, narrow suckout at 5kHz was unaffected by removing the grille.
Fig.4 GoldenEar Triton Reference, lateral response family at 50", normalized to response on tweeter axis, from back to front: differences in response 90–5° off axis, reference response, differences in response 5–90° off axis.
Fig.5 GoldenEar Triton Reference, vertical response family at 50", normalized to response on tweeter axis, from back to front: differences in response 15–5° above axis, reference response, differences in response 5–15° below axis.
Fig.6 GoldenEar Triton Reference, spatially averaged, 1/6-octave response in JA's listening room (red); and of KEF Reference 5 (blue).
Turning to the time domain, the GoldenEar's step response on its tweeter axis (fig.7) indicates that its tweeter, midrange drivers, and woofers are all connected in positive acoustic polarity. The decay of the tweeter's step smoothly blends with the start of the midranges' step, and the decay of their step in turn smoothly blends with the rise of the woofers' step, all of which suggests optimal crossover design. However, there is a strong reflection of the tweeter's output 600µs after its initial arrival at the microphone, which must be from the bulky grille. This reflection gives rise to the ripples in the response seen in figs. 2 and 4.
Fig.7 GoldenEar Triton Reference, step response on tweeter axis at 50" (5ms time window, 30kHz bandwidth).
Finally, the Triton Reference's cumulative spectral-decay plot on the tweeter axis (fig.8) shows an initially clean decay, but a lot of low-level hash develops, particularly in the high treble. I conjectured that this is connected with that strong reflection of the tweeter's output, but while the initial decay was somewhat cleaner when I repeated the measurement without the grille, there were still some low-level artifacts in the treble. These could, therefore, be due to reflections of the tweeter's output from the edges of its chassis. Apart from that quibble, the Triton Reference's measured performance reveals excellent engineering.—John Atkinson
Fig.8 GoldenEar Triton Reference, cumulative spectral-decay plot on tweeter axis at 50" (0.15ms risetime).
