Designer interconnects for audio have, in many ways, become something of a cult industry in recent years. Many golden ear audiophiles claim to hear audible differences between ordinary cables and those of exotic construction and price. Sometimes the listeners can confirm the differences in controlled double-blind listening tests, but more often, they cannot. In some cases, big differences that were described in great detail were reported when listeners were led to believe they were comparing two cables but nothing was actually changed [1]. In other cases, listeners were unable to tell whether their own designer speaker cable or a piece of 16-gauge "zip cord" was in place while listening, without time limits, to program material of their own choosing reproduced on their own (expensive) sound systems [2]. Such double-blind tests are mainstays of real science and are used, for example, to separate the real from the imaginary effects of drugs in medical trials. When applied to audio, such tests prove that even highly trained and experienced listeners can be strongly biased by their own expectations and beliefs.
Some audio experts believe audio is too important to be trusted to technology, and consequently, they dismiss all scientific methods, including double-blind tests. This attitude, combined with the widespread notion that more expensive products must be better, has opened the door to a flood of marketing hype and misinformation. Promotional white papers abound with pseudo-science buzz words, theoretical explanations based on absurd and fanciful physics, and new proprietary measurement techniques replete with previously unknown units of measure.
Simple interactions vs. solutions to nonexistent problems
When audible differences truly do exist, most are easily revealed with
either traditional frequency response or distortion tests and are the
result of simple interactions between the cable and the equipment's input
and output circuitry. At audio frequencies, the most significant parameter
of unbalanced line level interconnect cable is its parallel capacitance,
shown as a single equivalent capacitor C in Figure 1. For most ordinary
cable, this is about 50 pF per foot. This capacitor and the line driver's
output impedance Zo form a simple low pass filter (Zi has no significant
effect). Because output impedance Zo can be 1 kV or higher in
consumer/audiophile gear, long- and/or high-capacitance cables can
significantly degrade high frequency response. The Zo of some vacuum tube
equipment is so high that it can drive only a few feet of typical cable.
With a typical Zo of 1 kV and a 50 pF per foot (305 mm) cable, 20 kHz
response will be -0.5 dB for 50 feet (15.2 m), -1.5 dB for 100 feet (30.4
m), and -4 dB for 200 feet (60.8 m) of cable. Some designer cable has much
higher capacitance and will produce these responses with much shorter
cables. For long cable runs, consider a low-capacitance cable, such as
Belden #8241F. Although listed as a video cable, its 17 pF per foot
capacitance allows about three times the lengths for the listed responses.
It also has a heavy gauge, low-resistance shield, which minimizes
common-impedance coupling (more on this later).
At audio frequencies, speaker cables can be represented by series
resistance R and inductance L, shown in Figure 2. The parallel capacitance
is normally insignificant. The Zo of a typical power amp is less than 0.1
V, while speaker impedance Zs will vary with frequency but have a typical
minimum value of 5 V. The most significant effect of R and L is on the
frequency response, which is caused by the loudspeaker's impedance
variations. Generally, using a wire size that keeps R under 5% of Zs will
hold response variations to less than 0.5 dB and have no significant effect
on actual damping factor [3]. Depending upon cable length, the inductance L
of simple "zip cord" may introduce some measurable, but most likely
inaudible, response roll-off and phase shift at 20 kHz. However, inductance
can be made negligible by simply using 40 (or more) conductor flat computer
ribbon cable with alternate conductors paralleled at each end. In addition
to being low in cost, this cable is ideal for routing under carpeting.
Some designer speaker cables have very high capacitance, creating instability problems for poorly designed power amps. Of course, there's an expensive solution for this, too-cables with built-in LC or RC termination networks. Properly designed power amps have these well-known "Zobel" networks (or their equivalents) inside where they belong in the first place.
Some cable manufacturers would have you believe that audio cables are transmission lines, but in the engineering meaning, cables are not transmission lines unless they are "electrically long" (about a quarter wavelength) at the highest frequency of interest. At 20 kHz, this is almost two miles (3.2 km)of cable. For audio cables less than 1,000 feet (305 m), transmission line effects are not an issue. But because of the much higher frequencies involved, most video, RF and data cables are transmission lines and require termination. The response of an audio cable to nanosecond impulses is irrelevant because audio signals contain no significant or intentional energy at those frequencies.
Then we have oxygen-free, high-conductivity, and linear-crystal copper and combinations thereof. To the best of my knowledge, oxygen-free copper was originally developed for the Navy to reduce flexure (work hardening) failures of cables. It had no special electrical properties until the designer cable people developed their speculative theories about the difficulty of current flow in ordinary copper. I have never seen any scientific evidence that the kind of metal used in audio cables makes any difference beyond the expected (and usually negligible) differences in resistance. Long flex life, however, is a good thing, especially for such applications as headphone or patchbay cords.
Skin effect is another problem exaggerated by hype. As frequency increases, current flow in a wire tends to concentrate toward its outer surface, causing an increase in AC resistance. At 20 kHz, for example, skin depth is 0.018 inch (0.46 mm), where 63% of current flow is between the surface and this depth [4]. For line-level audio interconnects, the effect is absolutely inconsequential even if center conductor resistance doubles or triples, because it accounts for less than 0.01% of total circuit resistance anyway. For loudspeaker cable, there may be some measurable effects because total circuit impedance is lower, but we're still talking about increases in resistances that, if the wire gauge is properly chosen for reasonable losses, are negligible in the first place.
Noise coupling in unbalanced interconnect cables
The biggest real problem with unbalanced interconnects is that they fall victim to common-impedance coupling. This coupling causes more than just audible hum and buzz; the coupling of ultrasonic and RF interference causes subtle degradations of audio quality. This degradation mechanism has been largely ignored by the industry for many years. It's the "dirty little secret" of consumer (and "semi-pro") audio. This may also explain why the designer cable fad has relatively few followers in professional audio where interconnects are balanced.
Briefly, noise currents are coupled from the power line by each piece of
equipment and then flow through any wires that connect the two pieces of
equipment. As shown in Figure 3, the ground conductor or shield of an
unbalanced interconnect becomes the path for the noise current. As the
noise current flows, it causes a noise voltage to be developed across the
length of the cable which adds directly to the signal at the receive end.
The coupling is worsened by longer cables or those with higher shield
conductor resistance.
There's a widespread misconception that hum, buzz and other interference arrives through the air and that it is picked up as a result of inadequate cable shielding. This idea results in a variety of cables with 100% coverage (foil) or double- and triple-shielded cables touted to solve noise problems. For the vast majority of systems, however, the most effective way to reduce noise coupling is simply to choose cables with the lowest shield resistance and keep them as short as possible. Although a designer cable might slightly alter the nature of the coupling and its subsequent audio effects, it cannot eliminate it. Power conditioning products, despite their claims, cannot overcome this inherent weakness of unbalanced signal interconnects either. Only complete disconnection from the power line (as with battery power) could eliminate the coupling. Where cables must be long, consider an audio transformer-based ground isolator, which can effectively stop the interference/noise current coupling by preventing current flow in the shield.
Ultrasonic noise coupling, bandwidth and spectral contamination
The audio signal degradation caused by ultrasonic and RF interference coupling is easier to explain than measure. Typically, it reveals itself only with rather sophisticated tests such as spectral contamination, which was proposed by the late Deane Jensen in a 1988 paper [5]. Basically, this test applies a large number of simultaneous tones to a device under test. Any non-linearity in the device under test will create complex intermodulation products at new frequencies, collectively called spectral contamination. Because the new frequencies are usually not harmonically related and appear only when audible signals are also present, they behave more like distortions than noise. Generally, listeners describe the audio as "veiled," "grainy" or "lacking detail and ambience." I think such distortion tests may have the ability to correlate laboratory measurements to real listener experiences.
Systems designed to produce subjectively pleasing and distinctive coloration appropriately belong in the recording studio where, like musical instruments, they can be manipulated by artists and producers to generate the desired effects. I believe, however, that the ultimate goal of a music reproduction system is as much transparency and neutrality as science will allow. Further, real science requires skepticism, especially if an observation suggests violation of well established physical law. Remember cold fusion?
References:
[1] John Dunlavy, "Cable Nonsense," letter, 5 Nov 1996 (available at www.solutions.com/verber/cables.html).
[2] Tom Nousaine, "Wired Wisdom: The Great Chicago Cable Caper," test data (available at www.oakland.edu/~djcarlst/abx.htm)
[3] Richard Clark, "Amplifier Damping Factor," Autosound 2000 Tech Briefs, Nov 1994, pp 469-470.
[4] Ralph Morrison, "Solving Interference Problems in Electronics," John Wiley & Sons, 1995, pp. 61-62.
[5] Deane Jensen & Gary Sokolich, "Spectral Contamination Measurement," AES Convention Nov 1988, Reprint #2725.
