Audio Physic Caldera III loudspeaker Measurements
Like the Peak Consult Empress reviewed last month, the Audio Physic Caldera was too bulky to lift on to a high stand for the acoustic measurements. The farfield measurements therefore inevitably suffer from a limited midrange resolution.
My estimate of the Caldera's voltage sensitivity was slightly higher than specified, at 90dB(B)/2.83V/m. This is also a useful 3dB higher than the average of the more than 500 speakers I have measured over the past 15 years. The Caldera's impedance (fig.1) varied between 3.5 and 6 ohms between the upper bass and upper midrange, and between 6 and 14 ohms in the upper midrange and low treble. This will tend to give a tilted-up balance with, say, a tube amplifier's high source impedance. However, I don't believe this to be the cause for MF's dissatisfaction with the Caldera's treble balance, given his use of solid-state Musical Fidelity amplifiers for much of his auditioning.
Fig.1 Audio Physic Caldera, electrical impedance (solid) and phase (dashed). (2 ohms/vertical div.)
I found three primary vibrational modes present on the Audio Physic's cabinet surfaces. Fig.2 shows a cumulative spectral-decay plot calculated from the output of a simple plastic-tape accelerometer fastened to the curved front baffle beneath the lower-midrange unit. Ridges of delayed energy can be seen at 219Hz, 310Hz, and 380Hz. The last was strongest on the top panel, but none of these modes is high enough in level to lead to problems in midrange clarity, I feel.
Fig.2 Audio Physic Caldera, cumulative spectral-decay plot calculated from the output of an accelerometer fastened to the center of the front baffle below the lower-midrange unit (MLS driving voltage to speaker, 7.55V; measurement bandwidth, 2kHz).
The saddle in the impedance-magnitude trace at 21Hz implies that this is the tuning frequency of the twin passive radiators, which in turn suggests excellent low-frequency extension. However, the measured output level of these units seemed lower than I was anticipating, while the woofer's output peaked significantly in the 40–80Hz octave. As a result, the Caldera's overall low-frequency output, shown to the left of fig.3, features significantly more midbass energy than expected from the nearfield measurement technique. The crossover to the midbass unit is specified as 150Hz, which is where my own measurements would place it.
Fig.3 Audio Physic Caldera, anechoic response on tweeter axis at 50", averaged across 30° horizontal window and corrected for microphone response, with the complex sum of the nearfield midrange, woofer, and port responses, taking into account acoustic phase and distance from the nominal farfield point, plotted below 300Hz.
Higher in frequency in this graph, the Caldera's output is smooth and even in both the midrange and in the mid-high treble, but the latter is 3–5dB higher in level than the former. Accentuating this imbalance is a lack of energy in the crossover region between the midrange units and the ring-radiator tweeter. MF talked about a "peak at around 6kHz" that he found "impossible to ignore." I suspect that MF's ears locked on to the speaker's general output in the 1kHz region as his reference level, with then the energy excess in the upper treble standing isolated by the relative lack of energy in the low treble.
Though the lack of on-axis presence-region energy will be compensated for, to some extent, by the off-axis flare in the same region, the Caldera's horizontal radiation pattern (fig.4) will add to this perception by accentuating the mid-treble energy in the room's reverberant field. This graph also reveals that while the ring-radiator tweeter has impressive on-axis extension above the audioband, its output falls quite rapidly to the sides. Vertically (fig.5), the Caldera's balance is maintained over quite a wide listening window, presumably a result of the coincident topology of the tweeter and the upper midrange driver.
Fig.4 Audio Physic Caldera, 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 Audio Physic Caldera, 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 5–10° below axis.
Fig.6 shows how this quasi-anechoic behavior adds up in MF's listening room. (To produce this graph, I averaged 120 spectral measurements for each speaker individually in an array centered on the position of Mikey's ears in his listening seat: 36" above the floor.) From both MF's auditioning comments and the quasi-anechoic measurements, I was expecting the powerful bass, the suppressed low treble, and the elevated treble, all of which can be seen in this graph. What I was not expecting from his comments was the degree to which the upper bass and lower midrange were suppressed. I can only assume that this results both from a lack of integration in-room between the woofers and the lower-midrange unit and from the "Allison Effect": destructive interference between the direct sound and the early reflections from room boundaries, which can become significant when a drive-unit is raised high above the floor, as are the Caldera's midrange units.
Fig.6 Audio Physic Caldera, spatially averaged, 1/3-octave response in MF's listening room.
In the time domain, the decay of the tweeter's step response doesn't properly integrate with the onset of the midrange step (fig.7), which correlates with the poor frequency-domain integration between the units seen in fig.3. Similarly, the decay of the midrange step doesn't quite match the onset of the woofer step. The Caldera's cumulative spectral-decay plot (fig.8) is clean in the high treble—this is indeed a good tweeter—but the decay lower in frequency is marred by early reflections.
Fig.7 Audio Physic Caldera, step response on tweeter axis at 50" (5ms time window, 30kHz bandwidth).
Fig.8 Audio Physic Caldera, cumulative spectral-decay plot at 50" (0.15ms risetime).
Especially considering the Caldera's high price, I was disappointed in its measured performance. Though with some kinds of music the speaker's shelved-up highs will be balanced by its heavy bass, the mid-treble emphasis is hard to hear through, as MF found. The overall poor integration of the drive-units suggests that more work is required to produce a neutrally balanced design.
New samples: As Michael Fremer explained, Audio Physic was horrified when they received the preprint of the original review of the Caldera III, and were convinced that the sample I had measured, serial number 002A, must have been damaged in transport. Accordingly, I performed a set of measurements on the other of the pair, serial number 002B, repeating some measurements with a different microphone (an EarthWorks QTC-40) in case my main mike, a calibrated DPA 4006, was misbehaving. Both microphones were within specification. The resultant response measurements of the two speakers are shown in fig.9 (original sample plotted in blue, second sample in red).
Fig.9 Audio Physic Caldera III, anechoic response on tweeter axis at 50", averaged across 30° horizontal window and corrected for microphone response, with the complex sum of the nearfield midrange, woofer, and port responses, taking into account acoustic phase and distance from the nominal farfield point, plotted below 300Hz. Serial numbers 002A (blue) and 002B (red).
As can be seen, the two samples measured almost identically. I had not expected anything different—when I'd measured the speakers in MF's room, they had sounded identical. So given that the speakers had almost certainly not been damaged but concerned that we had not been sent representative samples, I felt it appropriate to delay publication of the review by one month, so that Audio Physic could submit a second pair of loudspeakers, ones that they assured us were performing to specification when they left the factory.
MF reports above on his reaction to the sound of this second pair. When he finished his auditioning, I again performed a set of in-room response measurements, again centered on the position of his ears in his listening chair, then took the pair home for more detailed analysis.
Fig.10 shows the impedance of one of the new pair, serial number 057A. In its broad outline, it looks very similar to the impedance of serial number 002A. But it differs in detail: The impedance peak at 775Hz is lower, at 9.25 ohms rather than 11.1 ohms, while the peak at 4.3kHz is higher, at 16.3 ohms compared with 14.8 ohms. In fact, the impedance of this new sample is overall a little higher in the treble than the original sample's.
Fig.10 Audio Physic Caldera III, s.n. 057A, electrical impedance (solid) and phase (dashed). (2 ohms/vertical div.)
Fig.11 plots the frequency response of the new sample (red trace) and the older sample (blue). Below 300Hz in this graph, the responses shown are calculated from individual nearfield measurements of the woofers, passive radiators, and midrange units. However, because of rain and strong winds on the only day I had available to measure the second sample's acoustic behavior, I had to do this indoors instead of outdoors, which limited the microphone distance to 39" (1m) rather than my usual 50". Even so, the responses of the two speakers are effectively identical in the midrange and bass. In the treble, however, while the region covered by the tweeter is still a little elevated, its output is a couple of dB lower in level than the original's and better integrated with the output of the midrange units.
Fig.11 Audio Physic Caldera III, anechoic response on tweeter axis at 50", averaged across 30° horizontal window and corrected for microphone response, with the complex sum of the nearfield midrange, woofer, and port responses, taking into account acoustic phase and distance from the nominal farfield point, plotted below 300Hz: serial number 002A (blue), 057A (red).
This was also the impression I had gotten when I auditioned the new pair of Caldera IIIs in MF's listening room. The speakers' treble now didn't sound disconnected from the rest of the range, and I was struck by how clean and smooth it sounded. This can be seen in the new pair's spatially averaged response (fig.12; old samples in blue, new samples in red). Between 315Hz and 16kHz, the new samples' in-room output meets extraordinarily good, ±1.1dB limits, with just a trace of extra energy in the middle of the midrange. Although there is a slight lack of energy in the presence region, this is not to anything like the extent with the original samples. The lower frequencies are still uneven, however, which could be heard as a lack of power in the lower mids and a slight midbass boost.
Fig.12 Audio Physic Caldera III, spatially averaged, 1/3-octave response in MF's listening room of original samples (serial numbers 002A/B, blue) and new samples (serial numbers 057A/B, red).
If this new pair of Caldera IIIs, with their much better treble balance, are typical of Audio Physic's current production, then the speaker can be given a guarded reservation—guarded because the woofer alignment probably needs a room larger than MF's so that the bass won't sound disconnected. But the fact that the tweeters in both of the original pair were identically out of specification does raise questions about the company's QC procedures.—John Atkinson