The Unseen Variable Page 2
One of the problems for someone auditioning a good SE amplifier for the first time is that all audiophiles have a psychological need to hear a difference. Some audio differences, while important, are quite subtle. Their analysis requires fine equipment, good circumstances, and, in some cases, extended listening. However, when compared with a conventional, solid-state, or push-pull low-impedance tube amplifier, a single-ended tube amplifier will always sound substantially different. This is due to the complex interaction of an SE design with typical loudspeaker loads—as well as their nonlinear behavior at high listening levels. Unfortunately, this provides immediate vindication for the listeners, reassuring them that they are skillful, sensitive, and capable.
There is therefore the attendant danger in such comparisons of confusing the acknowledged positive aspects of SE tube amplifiers with other changes. In fact, there is a distinct danger of making a virtue out of the entire sonic package.
Accordingly, I designed a very simple test to see whether just one of the variables in this subjective problem can be evaluated in isolation. That variable is source impedance.
In the days when "specmanship" was the name of the game, amplifier manufacturers boasted of very high damping factors—even claiming figures of several thousand in some cases. Damping factor is the ratio between the speaker load impedance—typically 8 ohms—and the output impedance of the driving amplifier.
It's a trivial matter for amplifier designers to include a small fraction of the output current in the feedback loop: they can set the theoretical damping factor to infinity, or even make it negative. Even with a zero output impedance, practical damping factors are typically no better than 50 for an 8 ohm speaker (25 for a 4 ohm model), due to speaker-cable and contact resistances. (The load impedance in this discussion is defined as that existing at the loudspeaker terminals. I am treating the loudspeaker as an electro-acoustic "black box" here, not delving into its inner structure, which includes several finite resistances.)
As single-ended tube amplifiers have output resistances in the 2-5 ohm range, I settled on a median value of 3.3 ohms for this investigation. This results in a damping factor (8 ohms) of just 2.4.
One detail required resolution: Should the listening test be conducted with the volume setting unchanged, or should the average midrange loudness be adjusted to take into account the attenuation introduced by the resistor? If the speaker system used could be considered as an 8 ohm design in the region responsible for perceived loudness, then the drop in level would be exactly 3dB.
Clearly the power level should be raised to take into account this loss. But by how much? A simple rise in level is hard to judge, since the loudness gain will be uneven over the frequency range. In practice, as with loudspeaker listening tests, all the assessors can do is to judge loudness by ear (footnote 1).
Two loudspeakers were auditioned for the listening tests: the Wilson WATT V and the Audio Physic Tempo. The former is a nearfield monitor, consequently showing a forward mid-treble and a shy, well-damped bass. Used in a free-field position, its quality and character are instantly recognizable. By contrast, the Tempo (reviewed in Vol.17 Nos.8 and 11, pp.103 and 169, respectively) has a full, well-developed bass octave and generally good mid-treble balance—but is perhaps a little rich in the low-treble range. Fig.1 shows the Tempo's impedance characteristic, fig.2 its on-axis frequency response when driven by amplifiers with source impedances of 0.1 ohms and 3.3 ohms, respectively.
Fig.1 Audio Physic Tempo, electrical impedance magnitude (2 ohms/vertical division).
Fig.2 Audio Physic Tempo, on-axis frequency response when driven by an amplifier with an output impedance of 0.1 ohms (top) and 3.3 ohms (bottom). Note overall drop in level and significant change in midrange balance (5dB/vertical division).
Footnote 1: In Stereophile's comparative loudspeaker listening tests, we have found that equalizing the B-weighted spls does give subjectively similar loudnesses.—JA