Sidebar 3: Measurements
Because of the Klipschorn AK6's bulk—each weighs 220 lb—I drove my test gear the 177 miles to Art's place and measured the speaker sitting on a furniture dolly in his driveway. I used DRA Labs' MLSSA system and a calibrated DPA 4006 microphone to measure the Klipschorn's behavior in the farfield and an Earthworks QTC-40 mike for the nearfield responses.
Klipsch specifies the Klipschorn's sensitivity as 105dB/2.83V/m, which is extraordinarily high. My estimate was lower, at 101.1dB(B)/2.83V/m, but this is still the second-highest sensitivity of all the speakers I have measured over the past 30 years. (The highest was the Auditorium 23 Hommage Cinema, which features a measured voltage sensitivity of 102dB(B)/2.83V/m.) The Klipschorn's sensitivity is a whopping 18.6dB higher than the sensitivity of the BBC LS3/5a I always measure at the same time I test a speaker (to ensure that I haven't made an error in setup). This speaker will play loudly even with flea-powered amplifiers driving it. And at typical listening levels, the drive-unit diaphragms will hardly be moving, which implies low distortion.
Klipsch specifies the Klipschorn's nominal impedance as "8 ohms compatible." This is optimistic: Not only does the speaker's impedance magnitude (fig.1, solid trace) drop to 3 ohms in the midbass and 2.7 ohms in the upper bass, but the electrical phase angle (dotted trace) is sometimes extreme. There are current-demanding combinations of 4.1 ohms and –43° phase angle at 44Hz and 4.1 ohms and +43° phase angle at 139Hz. Despite its very high sensitivity, the Klipschorn will not be at its best with amplifiers that are not comfortable driving 4 ohm loads.
While there are discontinuities in the impedance traces that would imply the presence of resonances of various kinds, with a horn speaker these might also be due to the effect of reflections from the horn's mouth. The only part of the two enclosures where panel resonances might give rise to audible problems is the front of the woofer bin. When I investigated this panel's vibrational behavior with a plastic-tape accelerometer, I found a fairly strong resonant mode at 316Hz, with others almost as strong to either side of it (fig.3). There were also some low-level modes in the upper bass. I could just hear these modes with the noise-like MLSSA signal and my ear close to the panel. They might not be significant with music, therefore, other than perhaps adding some midrange congestion with some kinds of music.
The difference in arrival times of the tweeter's output and that of the midrange unit can also be seen in the Klipsch's cumulative spectral-decay or waterfall plot (fig.9). This graph is difficult to interpret, but the different arrival times make it possible to calculate individual cumulative spectral decay plots. Fig.10, therefore, shows the midrange unit's waterfall plot. It is fairly clean, though with several spikes of low-level delayed energy visible above the crossover frequency to the tweeter. The tweeter's waterfall plot (fig.11) reveals a clean initial delayed energy, but again there are some ridges of delayed energy at the top of its passband.
Multiway loudspeakers with horn-loaded drivers but a flat baffle can be made time-coincident with digital signal processing. As the Klipschorn has separate input terminals for each of its three drive-units, I can't help wondering what a fully DSP-corrected, tri-amplified version of this loudspeaker, with the high sensitivity coupled with optimized step and frequency responses, would sound like.—John Atkinson
Footnote 1: See the diagram at 14:40 in my lecture on loudspeaker measurements. Footnote 2: See wikipedia.org/wiki/Precedence_effect. Footnote 3: See villchurblog.com/inventing-the-speaker/ and www.stereophile.com/interviews/105villchur/index.html.
Fig.1 Klipsch Klipschorn, electrical impedance (solid) and phase (dashed) (2 ohms/vertical div.).
In addition, the very large difference between the average impedance in the bass and the rise to 20.7 ohms at 1.5kHz means that the speaker's perceived tonal balance will be very dependent on the amplifier's output impedance. For example, as I always do, I used my solid-state Krell KSA-50 amplifier for the acoustic measurements, which has a measured output impedance of 0.13 ohms from 20Hz to 20kHz. To investigate the interaction between the Klipsch's impedance and the amplifiers used to drive it, I measured the speaker's farfield response on its tweeter axis with the Krell (see later), then repeated the measurement with the tubed Shindo Haut-Brion amplifier that Art Dudley used for much of his auditioning. The difference between the two responses, which were normalized at 1kHz, is shown in fig.2. I don't know what the Shindo's output impedance is, but it must be high, as this amplifier tilts down the Klipschorn's response above 2kHz, with the difference between the response with the Shindo and the Krell reaching –10.7dB between 11kHz and 13kHz.
Fig.2 Klipsch Klipschorn, difference in the anechoic response, 300Hz–30kHz, on tweeter axis at 50" due to substitution of a Shindo Haut-Brion amplifier for a Krell KSA-50 (5dB/vertical div.).
This might seem less-than-optimal performance on the Shindo's part. However, when I asked Art to let me listen to music with the Klipschorns using first the Air Tight ATM-300 amplifier that he reviewed in February 2019, which has a low output impedance of 1.2 ohms from its 8 ohm output transformer tap, then the Shindo Haut-Brion, the latter gave the best sound. Yes, the balance with the Shindo lacked top-octave air, but the high frequencies were exaggerated with the Air Tight, there was no low bass, and the midbass sounded lightweight. Low-frequency extension was still absent with the Haut-Brion, but the midbass now sounded in better balance with the midrange. The sound of this loudspeaker will be very dependent on the amplifier it is used with.
Fig.3 Klipsch Klipschorn, cumulative spectral-decay plot calculated from output of accelerometer fastened to center of woofer bin front panel (measurement bandwidth, 2kHz).
The Klipschorn's impedance- magnitude plot has a single peak in the bass centered on 36Hz, suggesting that this is the drive-unit's fundamental tuning frequency. The red trace in fig.4 was taken with the microphone close to one of the low-frequency horn openings; it indicates that the woofer's output rolls off sharply below the tuning frequency. There are also significant peaks in the woofer's output at 103Hz and 190Hz, which I could hear with the MLSSA signal as a "hooty" quality.
Fig.4 Klipsch Klipschorn, acoustic crossover on tweeter axis at 50", corrected for microphone response, with the nearfield woofer response (red) plotted below 675Hz.
I plotted the output of the woofer in fig.4 at a level where it crossed over to the farfield response of the midrange unit (blue trace) at the specified 450Hz. The woofer rolls off rapidly above that frequency, but this graph does suggest that the woofer's level is somewhat lower than the midrange's. The midrange unit crosses over to the tweeter (green trace) a little lower in frequency than the specified 4.5kHz; while the rolloff is fast, a couple of resonant modes are visible. Driven by the Krell amplifier, the tweeter appears to be balanced on average 3dB too high in level, with a 6dB-higher peak between 8kHz and 11kHz and a sharp rolloff above 18.5kHz.
Fig.5 shows the Klipschorn's farfield response, averaged across a 30° horizontal window centered on the tweeter axis. Again, the tweeter appears to be too high in level. Some sharply defined suckouts are visible between 2.5kHz and 4.5kHz. The farfield responses in figs.3 and 4 were taken with the grille in front of the midrange and treble horn openings, as that is how Art Dudley performed his auditioning. Removing the grille and repeating the farfield measurement suggested that these suckouts filled in to some extent. Note that I haven't plotted the response below 350Hz in fig.5, as the horn-loaded woofer's output in free space wasn't fully captured within the gated time window. (For this kind of measurement, the FFT is applied to the portion of the impulse response before the first reflection of the sound, which in this case will be from the ground in front of the loudspeaker (footnote 1.)
Fig.5 Klipsch Klipschorn, anechoic response on tweeter axis at 50", averaged across 30° horizontal window and corrected for microphone response.
The plot of the Klipschorn's horizontal dispersion (fig.6), normalized to the response on the tweeter axis, is complex. Disregarding the off-axis peaks, which will be due to on-axis suckouts filling in to the speaker's sides, overall the speaker is more directional through the midrange and treble than a conventional, direct-radiator design. This graph also suggests that in free space the woofer horn offers a higher output to the speaker's sides than it does on the tweeter axis. However, that suggestion is incorrect. What is happening is that the time delay affecting the woofer's output (see later) is reduced to the speaker's sides, meaning that more of it is captured in the gated time window. In the vertical plane (fig.7), the on-axis response is maintained over a wide ±10° window. This is a good thing as the tweeter is 51" from the floor and the average listener's ears are 36" high. I do note, however, that AD's listening chair places his ears closer to 45" from the floor.
Fig.6 Klipsch Klipschorn, 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.7 Klipsch Klipschorn, vertical response family at 50", normalized to response on tweeter axis, from back to front: differences in response 10–5° above axis, reference response, differences in response 5–15° below axis.
In the time domain, as with the Auditorium 23 horn-loaded loudspeaker, the Klipschorn's step response (fig.8) is complicated. The problem with a loudspeaker using horn-loaded drive-units but with the horn openings in the same plane is that the outputs of the drive-units arrive at the microphone or the listener's ears at different times, due to the different lengths of the horns. While all three drive-units are connected in positive acoustic polarity, the tweeter's output—the upward-moving spike at 4ms in fig.8—arrives first at the microphone. The output of the midrange unit doesn't arrive at the microphone for another 1.5ms, while the woofer's output is too late to be shown in this graph. (Separate measurements of each drive-unit indicate that the woofer's output arrives 6ms after the midrange unit's.) Does this matter? In theory, even the woofer's output is within the hearing system's tolerance for different arrival times (footnote 2).
Fig.8 Klipsch Klipschorn, step response on tweeter axis at 50" (5ms time window, 30kHz bandwidth).
But from my own experience with truly time-coincident, multiway loudspeakers like Quads and Vandersteens, where the outputs of all the drive-units arrive simultaneously at the ear, I feel such time delays smear and obscure stereo imaging precision.
Fig.9 Klipsch Klipschorn, cumulative spectral-decay plot on tweeter axis at 50" (0.15ms risetime).
Fig.10 Klipsch Klipschorn, midrange unit only, cumulative spectral-decay plot on tweeter axis at 50" (0.15ms risetime).
Fig.11 Klipsch Klipschorn, tweeter only, cumulative spectral-decay plot on tweeter axis at 50" (0.15ms risetime).
The Klipschorn's measured behavior reveals that the performance parameters that are generally held in the 21st century to correlate with good sound quality in both the time and frequency domains have been compromised to achieve that astonishingly high sensitivity. Its sound will also be heavily dependent on the amplifier with which it is driven. I don't recommend using a typical solid-state design with the Klipschorn because the region covered by the tweeter will be too hot.
I wasn't too surprised by the Klipschorn's limited low-frequency extension despite its size. I was reminded of the impact Acoustic Research's first loudspeaker, with its relatively small sealed enclosure and "acoustic suspension" woofer, made in the mid-1950s. "My measurements showed that my little prototype had better bass and less distortion than anything on the market, yet it was one quarter the size," wrote AR's founder, Edgar Villchur, adding "I thought, 'This has got to be the future of loudspeakers.'" (footnote 3) It was.
Footnote 1: See the diagram at 14:40 in my lecture on loudspeaker measurements. Footnote 2: See wikipedia.org/wiki/Precedence_effect. Footnote 3: See villchurblog.com/inventing-the-speaker/ and www.stereophile.com/interviews/105villchur/index.html.
