Blind Listening Page 6
I hope that you will agree from Will's analysis that there did seem to be a slight audible difference between the two power amplifiers. With so many trials, the statistics give a high degree of confidence in the results. Are there obvious measured differences that would explain the blind identification?
Absolute phase differences, though often audible under blind conditions, were not a factor here as both amplifiers are non-inverting. Level differences also did not contribute. However, there were slight frequency-response differences between the two.
Fig.3 shows the measured impedance characteristic of one of the B&W 801 loudspeakers: It features a well-damped enclosure resonance at 35Hz with the port tuned to a very low 17Hz or so, and two peaks at 280Hz and 1745Hz, presumably due to the high-order crossover design. Measured at the speaker terminals using 1/3-octave tones on a test CD played on the Marantz CD-94 via the Iris—the complete chain used in the listening tests—the response of the Adcom was ostensibly flat throughout the audio band into this load, with a very slight droop in the top octave (-0.1dB at 12.5kHz, -0.2dB at 16kHz, -0.4dB at 20kHz), presumably due to the CD player. There was considerably more interaction between the VTL monoblock's output impedance and the B&W 801, however, as can be seen from fig.4, which shows the difference in response between the two amplifiers measured at the speaker terminals. Reflections of the peaks in fig.3 can be seen, as can a slight (0.5dB) excess of energy in the presence region. There are also broad depressions in the midbass, lower midrange, and high treble, reaching a maximum of -1.5dB at 80Hz, 500Hz, and 16kHz, respectively. (All figures indicate the VTL level with respect to that of the Adcom when made equal at 1kHz.)
Fig.4 B&W 801 matrix, modulus of impedance
Fig.3 VTL 300 response ref. Adcom 555 (measured at 801 terminals)
Many prominent engineers—Stanley Lipshitz, the current President of the Audio Engineering Society, for example—have stated that if two amplifiers are found to be audibly different, then simple linear errors, such as a frequency-response difference, will be the cause. Having listened at length to the two amplifiers during the four days of preparation and testing, as well in Gordon's and my systems in Santa Fe, and therefore become familiar with their sounds, I must admit that part of the reason for any identification will be due to the amplitude-response difference. (Though it then becomes hard to explain why the engineers at the Los Angeles AES Convention failed to hear what I believe to be much greater response differences. It also raises the question as to why the listeners in the highly publicized Quad and HFN/RR tests in the UK in the late '70s also failed to detect response differences that were as great as those here.)
I consistently identified the VTL as being slightly more forward in the upper midrange, though less "fizzy" in the extreme highs than the Adcom, for example, which does appear to correlate with the measured response differences. The deep bass on the Adcom also had noticeably more energy, though it was also tighter, being better controlled than the tube amplifier.
Overall, however, there is no doubt in my mind that the VTL is more pleasant to listen to for long periods, something that would not be suggested by the slight response aberrations seen in fig.2. Despite the slight forwardness to its sound, the VTL lacked the bothersome "dry" quality of the Adcom's sound that, for me, was most irritating on the King's Singers track. Individual images within the soundstage, too, were more "rounded" with the tube amplifier than the solid-state, where the totally artificial soundstage on the "Penny Lane" track took on more of a cardboard-cutout characteristic.
That there is more to the difference than simple amplitude differences must remain conjecture, although I must point out that J. Gordon Holt founded this magazine back in the early '60s on the grounds that there was more. The next opportunity we have to stage such a test, therefore, I will choose two amplifiers that have identical output impedances, hence frequency responses, into the same loudspeaker load, a B&K ST140, say, and a pair of Mark Levinson No.20.5s.
To sum up:
• When considered overall, the results indicate to a high degree of confidence that a small but real audible difference exists between the Adcom GFA-555 and VTL 300W monoblocks when used to drive B&W 801 loudspeakers.
• That the amplifiers could be distinguished by ear under blind conditions could be due to the interaction between the amplifiers' different output impedances and the loudspeakers in use. Further testing is required to reveal whether there is more to it than this.
• More than 500 people took part in these tests, each suffering seven trials. When small subjective differences are involved, these results confirm that it is essential to use a very large number of presentations in a blind test if the possibility of the results featuring a Type 2 error—erroneously concluding that there was no difference when in fact there is one—is to be avoided. (See "The Highs & Lows of Double-Blind Testing" and "The Double Blind & the Not So Blind," Vol.9 Nos.2 & 5 respectively, for a full discussion.)
• Different kinds of music vary considerably in their abilities to reveal differences between amplifiers. The Fauré piece for choir and orchestra was the best in this respect, with the naturally miked drum recording totally inadequate (despite its fulfilling J. Gordon Holt's criteria as a suitable test program for revealing departures from "fidelity"). That there was no original for the Jennifer Warnes or Flim & the BB's tracks didn't prevent them from being almost as efficacious as the solo piano recording in revealing the differences between the amplifiers. Both were even a shade better at revealing when the amplifier remained the same.
• Despite the visitors to the show being keen audiophiles, over half those who took part in these listening tests were unable to reliably hear the small differences between the two power amplifiers. Part of the reason must be the high level of sound breakthrough from the adjacent room, as well as the overcrowded conditions—it would be hard to imagine someone either sitting on the floor to the side of one of the speakers, or standing next to the rear wall behind eight rows of seats, getting much of a good sound. But in my opinion, the main reason for the lack of identification was that even keen audiophiles rarely perform the kind of concentrated listening that I was asking from them in these tests. I was not surprised to note that some of the high scorers were in fact active in the high-end industry. Tony Di Chiro of Kinergetics, for example, scored 6 correct out of 7, as did Jon Iverson, the retailer whose comments on the test appear in this month's "Letters" section [and now webmaster for the Stereophile website—JA].
I would think it obvious that those professionally involved in listening at this level of concentration get better at being able to discriminate very small differences between nominally identical components. Isn't it reasonable to expect that J. Gordon Holt or Harry Pearson, for example, who have been professionally listening to high-end components for three and two decades, respectively, should have developed a considerable degree of skill in this area? The question then should be whether it is worth designing and manufacturing components that only a favored few will be able to distinguish.
But that's the raison d'être of high-end audio.