Editor: Since J. Gordon Holt's description of the SWDT (straight wire differential test), there have been many comments and inquiries directed both to Stereophile and directly to me concerning this test. I will try to answer most of these in this catchall commentary.
The starting point is that it is very difficult to compare amplifier A with amplifier B to make a quality assessment. The conventional A/B test, or any comparison, may indicate that the amplifiers are different, but how can one determine which is "best"? "Best" is a matter of judgment, and even though many golden-eared experts think their ears are infallible, they disagree among themselves as to amplifier quality. What is needed is an accurate reference against which an amplifier can be judgeda reference which does not involve the entire listening chain, from source to loudspeaker, but which is the equivalent of a perfect amplifier. That reference is the straight wire.
If readers recall the "Carver challenge," they read where Bob Carver put two amplifiers into a bridge configuration and adjusted one to match the other on a differential listening test. In analogous fashion, the SWDT matches an amplifier to a straight wire rather than to another amplifier. This is precisely a comparison of the input of an amplifier with its output. If the input and output are the same, the amplifier is accurate. If they differ, the amplifier is inaccurate. This has nothing to do with the subjective reaction of whether the amplifier has "pleasant" or "natural" sound. It has to do with correct sound.
To carry out the SWDT, the input signal is applied to one terminal of a headphone or monitor loudspeaker; through a variable attenuator, the same signal is fed to the input of the amplifier under test. The output of the test amplifier is connected to the other terminal of the monitor transducer. The normal loudspeaker is also connected to the output so that its load becomes part of the test, but it must be moved to a separate area so that its sound will not interfere with the sound through the monitor transducer. The test signal for this arrangement must come from a low-impedance source: this can most simply be one half of the stereo amplifier. The attenuator is adjusted for minimum sound output in the monitor speaker while the input signal is set to normal listening levels in the load loudspeaker. The input and output conditions are precisely those which prevail in normal listening operation.
When the attenuator is adjusted for minimum sound, the best "null," the signal in the monitor represents the difference between input and output. If input and output are the same, there is no voltage potential across the monitor, and consequently no sound in the monitor. When the null is inaudible, it means that all of the distortions in the amplifier are inaudible; at that point, there cannot be improvement in the audible performance of the amplifieruntil, perhaps, some new program sources come along which will be more difficult to handle.
The logic is irrefutable: if the input and the output of the amplifier are the same, the audible performance of the amplifier cannot be faulted.
Now let's consider the questions which have been raised about the SWDT.
Mr. Babrauskas' convoluted arguments in Vol.10 No.3 against low-source impedance amplifiers take me back some 40 years. He has discovered interface intermodulation distortion at a time when it is no longer of consequence. All present loudspeaker manufacturers use amplifiers which are essentially constant-voltage devices to design and test their loudspeakers, and modern amplifiers are all designed to give uniform voltage output into speaker systems in which the impedance varies with frequency. With properly designed amplifiers with very low impedance output, the back-EMF from the loudspeaker has no detrimental effect and does not affect the SWDT. An accurate amplifier can still achieve a deep "null" despite Mr. Babrauskas' contentions.
Reg Williamson, whose opinions I very much respect, says that what we have done is not new but is the reinvention of the wheel. I agree that differential testing is not newI used it myself back in the 1960s to test components. However, has anyone else applied the SWDT to amplifier design to make an amplifier which has no audible aberrations? I have not seen any signs of that. I am aware, of course, of the excellent work done by Peter Walker of Quad in testing amplifiers with a differential bridge. Quad's approach was a laboratory approach, not readily applicable by the hi-fi hobbyist. The SWDT can be set up and checked with a minimum of equipment. Quad also introduced phase and amplitude compensation in the straight-wire path to improve the null by making the amplifier and the wire more alike. This alteration"bending," if you willof the straight wire was done on the assumption that phase and amplitude errors are inaudible. I prefer to make no assumptions as to audibility, thus making the test more stringent.
Mr. Williamson incorrectly faulted our arrangement in claiming that it was not operating under "normal working conditions of source signal and load." If he examines the arrangement carefully, I believe he will see that his conclusion was not correct. He also criticizes our tweaking of the amplifier to make it have a better match with the straight wire. I feel that it is standard good engineering practice to trim an amplifier for minimum distortion under actual working conditions. Matching the straight wire is exactly the same as trimming for minimum distortion.
There were several readers who have questioned the use of a driving amplifier to supply signal for the SWDT. They mistakenly believed that the characteristic of the driving amplifier is to provide a low-impedance signal at a level equal to the level desired in the test amplifier. As long as the drive amplifier supplies a wide-band, fast-risetime signal, its own distortion characteristics do not matter. It could be a noise generator or a spark gap, and the SWDT would still be valid. If the drive amplifier has distortion, then the test amplifier is being tested as to its capability to handle a distorted signal without adding further distortion.
Several people misanalyzed the circuit as having positive feedback and assumed that this would affect the results of the SWDT. The connection of the monitor speaker from input to output of the test amplifier superficially appears to be a positive feedback connection. However, if there were positive feedback, it would be reduced to insignificance by the low output impedance of the drive amplifier. Further, with an accurate amplifier, there is little or no current in the monitor speaker; it is, effectively an open circuit, not making a feedback connection.
The question of phase shift and time delay from input to output has been raised by several readers. That difference between input and output will prevent a good "null" and will produce residual sound. Obviously, a constant time delay (uniform across the band) will not produce an adverse audible effect. However, the residual from phase shift might mask other distortion components in the SWDT. I prefer to see, therefore, an amplifier design which is sufficiently broadband to minimize time delay in the audio band. This can be accomplished with some phase compensation.
While many people do not believe that phase shift is audible, there is one place where it definitely has an effect, and that is between the two stereo channels. Unless the two sides track phase precisely, there will be problems of stereo imaging. The most practical way to have the left and right channels track is for each of them to have little or no phase shift. The design efforts which make the Hafler XL-280 amplifier perform well on the SWDT give a side-to-side differential null of about 70dB. This preserves the stereo imaging better than on any other units we have examined.
The question has been raised as to why some subjective reviews have not always been consistent with the rankings of the SWDT. Preferences do not necessarily have to correlate with objective tests. A deep wideband "null" may not be necessary for euphonious sound, but it is sufficient to establish that the amplifier has no audible aberrations. Although there is high correlation between a deep "null" and good (accurate) sound, it is conceivable that there can be a poor "null" with pleasant sound. This can happen if the distortions are of a non-offensive nature, such as a small high-frequency roll-off. It can also come from a listener's inaccurate mental reference standard as to what is accurate sound.
There is a philosophical question here. Should an amplifier be pleasant sounding, or should it be accurate even if accuracy is not as pleasant? The SWDT reveals accuracy or the lack of it. It is the only way to assess accuracy directly and objectively with the amplifier operating under normal signal and load conditions. It does this in real time, with a dynamic signal source. It encompasses all forms of distortion. It requires no special instrumentation or skill. If the residual "null" is inaudible, the conclusion that the amplifier is accurate is unequivocal.
Can anyone suggest a more meaningful or practical test of amplifier accuracy?David Hafler, The David Hafler Company, Pennsauken, NJ