High Fidelity, The Miraculous Invention
Audio or high fidelity has been a miraculous invention. If you need to understand why, read Steppenwolf. We audiophiles are at the forefront of a revolution written about by Larry Klein, in Stereo Review, over a quarter century ago. We believe that, if we can create the right conditions, the original reality can be recreated at will. Today we recreate auditory information, tomorrow information for all the senses. Ultimately, our revolution leads to a version of "The Matrix". Understanding this, we have pushed the envelope for developing better and better audio equipment.
This push for furthering the goal has led to two camps: those that rely on measurements to determine accuracy and those that rely on their hearing. Let's briefly examine both.
The Measurements Camp
Relying on measurements implies that the measuring equipment is at least as good as our hearing in ascertaining differences. The advantage is that the equipment does not contain a bias. In addition, it can perform the task over and over again without tiring and is therefore far less biased and far more consistent than a human.
The Hearing Camp
Those that rely on hearing believe that there is no proof the "machines" are as refined as human hearing in distinguishing differences. In addition, there is no proof the "machines", even if they were capable of determining differences that the human ear cannot, can discern all the differences the human ear can discern. Therefore, there is no proof that test equipment can test for a duplication of reality as determined by human senses and experience.
We certainly must agree that a lack of bias is a desired goal in testing audio components, as it is in all scientific inquiry. If we say that bias is desirable then the testing has lost its validity, as would any and all claims arising from the biased tests. Therefore, whether we use test equipment for measuring, or we use humans for listening, eliminating any and all possible bias is a desirable goal.
The history of science is filled with trends. Science follows trends because a certain form of thinking "takes hold" within the scientific community. Some obvious examples are that during the age of the machine, man was compared to the machine and during the age of the computer, man is compared to the computer. Neither comparison is valid. It is precisely because neither comparison is valid that I must side with those that want to rely on hearing to determine how accurate an audio system is for recreating the original event.
Test equipment does have a purpose in the above scenario however. It should be used to determine whether or not perceived differences in audio equipment are measurable. We'll call this "eliminating the charlatan factor". To understand this I'll use interconnect cables as the example. If, for instance, I am using an analog interconect cable between my CD drive and my processor, and I hear a difference when I change cables, I should measure the cables to be sure there isn't a measurable reason for the difference. If one of the cables has significantly higher R,L,C attributes, I should be able to measure a difference at the frequency extremes. In this case I know that someone is attempting to sell me a "better" cable by claiming it is "smoother", when in fact all they've done is sold me a cable with a higher impedance attributes in order to roll off the top end. So we use test equipment to keep the manufacturers, that claim what they have developed is better, honest. If I hear a difference, but I can't measure a lack of accuracy, then I know we may actually have an improvement in reproduction.
Elimination or Reduction of Bias
The most important step in the reduction or elimination of testing bias is to know what you are testing for but to not know what you are testing. In the case of audio components or cables that would mean the listener should know he is listening to hear a difference, but should not know what product he is listening to.
This is accomplished by having someone that switches the equipment and someone else that is the "golden ear" listener. The equipment and the person switching it should both be concealed from view. So if, for example, we are testing cables a curtain can be set up which blocks the view of the listener. The curtain can be set up behind the speakers and therefore the actions of the person switching the cables and the appearance of the cables is shielded from the listeners view.
The listener is free to choose whatever music or test discs they want. They can have selections played to their hearts content. There is no time limit. They can choose the listening level. They can determine when they want the cables switched. If the listener believes they hear a difference we can commence with the actual testing.
The listener chooses a disc or discs and a set volume level. The volume level is confirmed by an SPL meter measuring output at a fixed position from the loudspeakers. When the listener wants, the cables are changed and the volume level is confirmed to be identical by the SPL meter. The listener writes down whether they believe this to be product A or product B. This process is repeated several times. To further eliminate "luck" and bias affecting the statistics, on some occasions, when the listener asks for the cable to be switched it is made to appear the switch has taken place but in fact it has not been done. The listener is made aware that this non-switching can randomly take place. To be fair to the listener, the pattern of switching and non-switching should be determined in advance and written down so the person switching also does not present a bias as to when they, in fact, switch. This switching pattern, of a minimum of 6 switches, is not shown to the listener. The occasional non-switching is equivalent to the sugar pill in pharmaceutical testing. The listener writes their impressions as to whether it is product A or B after each time a switch is requested.
At the conclusion of the testing if, for example, the listener shows that they can hear a difference (getting it correct a minimum of 5 out of 6 times) the cable is now tested for a measurable difference. The testing is done with the same audio components used for the listening tests. The reason for this, is equipment impedance differences. For example an amplifier with high output impedance could show a measurable difference at the end of a speaker cable whereas a low output impedance amp would not. These can be tests for such things as frequency and square wave response. If a lack of accuracy is measured we know the difference is merely the same as using tone controls and a "better" cable has not been developed. A cable with measurable ringing for instance might be heard as having better detail because things are heard which are not heard with the accurate cable. However the ringing is not present in the original signal and is, in fact, an artifact added by the system that should not be there.
Engineering theory plays no part in the testing described above. In fact, knowledge of the theory could only serve to bias the listener. The engineer's theory, as to why a product they developed is better, only matters if the test proves there is an audible difference that is not due to a measurable lack of accuracy.
updated 03/09/03: Corrected comments about cable impedances