Cable Reviews

Sidebar 1: Measurements, from June 1991 (Vol.14 No.6)

Back in the February issue (Vol.14 No.2, p.158), young Dick Olsher, Stereophile's resident physicist, gave a rave review to interconnect from a new California company, Lindsay-Geyer. Their model 4-40 is different from every other interconnect in that it is constructed from a magnetically permeable material, namely "Mu-metal." (Four individually insulated 40-mil strands are used.) Normally such a material is avoided for conducting signals, due to its low conductivity. (The fact that it is permeable means that the current is squeezed into a shallow skin around the circumference, even at audio frequencies, thus increasing its resistivity.)

Why, then, would having a permeable conductor be an advantage?

In Dick's review, he paraphrased Lindsay-Geyer's white paper to explain that while the signal's electromagnetic wave propagates along an ordinary conductor at a velocity approaching that of light, "the signal sinks into the wire as an inverse function of frequency (the skin effect). The magnitude of [this current] decreases exponentially with depth of penetration because of ohmic losses. At each frequency, a skin depth may be calculated at which the attenuation is exactly 1/e, or 36.8%. The signal is also retarded in phase as it sinks into the wire because of the finite velocity of propagation [of a current] inside the wire...In copper at 1kHz, the [current] speed is a relatively pedestrian 13m/s. At that speed, a signal will sink through a 1mm wire in 77 microseconds. A 77µs delay should clearly be audible, assuming the magnitude of the delayed signal is significant—which is the case here. The skin depth at 1kHz is 2.1mm in copper, so the magnitude of the re-emergent signal for a 1mm wire is only down about 4dB."

Note that DO is not talking about the EM wave that carries the signal information, but about the associated current at a direction of 90° to the conductor axis. His point is that textbook electronics theory—see Engineering Electromagnetics, by William H. Hayt, McGraw-Hill, pp.398–402, for example—appears to indicate that the initial wave propagating along the cable will be followed by a delayed version reduced in level due to the attenuated 90° current reemerging after passing through the cable thickness. (Imagine a circle at the cable surface collapsing evenly through the conductor to a point at its center, then re-expanding back to a circle at the surface, all the while diminishing in intensity by a factor of 1/e for every skin depth.) That is, in fact, if the re-emergent current is itself associated with an electromagnetic wavefront. If it is, then although each slice of the conductor will presumably give rise to its own echo, because the speed of the EM wave is so much higher than the current speed, the effect at the end of the cable will still be of a single echo.

"Let's consider what happens to a transient waveform propagating down this 1mm copper wire," continued DO in his review. "Because the waveform is composed of many harmonics and because the 'sinking speed' is a function of frequency, a transient that sinks through the wire will be smeared out in time. The typical [copper] interconnect then propagates the original signal plus a smeared-out copy of that signal. It is possible for the smeared copy to sink through the wire again and generate another smeared copy of itself."

In other words, Lindsay-Geyer claims that using a conventional conductor such as copper will result in progressively attenuated "echoes" smearing the musical information. But by making the conductor permeable—a topology patented by Dr. David Lindsay—the skin depth will be made so small that the delayed current will be totally attenuated, the result being an absence of any such smearing, to the benefit of the music. And DO did find that the L-G interconnect was eminently musical, its sound quality, he wrote, "founded on three cornerstones: treble purity, harmonic integrity, and image cohesiveness."

The obvious correlation to draw is that between the L-G's sonic performance and its supposed absence of signal smearing. Or is it? I must point out that I regard all this talk of echoes and transient smearing as conjecture. If such an effect did exist in conventional copper conductors at audio frequencies—at radio frequencies, echoes galore occur every time there is an impedance mismatch between source, cable, and receiver—then surely someone would have noticed? I therefore decided to set up a simple experiment to look for the presence of the Lindsay-Geyer effect.

To hand was not only a 6m length of Lindsay-Geyer interconnect, but 8m of twisted-pair, solid-copper, single-conductor R232 data cable, this having a conductor diameter of 0.6mm, giving an "echo" time of 46µs for a 1kHz signal. I also had available a 5m length of a high-performance commercial interconnect, AudioQuest Lapis. At 2800pF, the measured capacitance of the L-G cable was much higher than that of either of the other two cables. (The Lapis measured 650pF, the solid-core copper 500pF.)

Cable capacitance will only be a factor if the source driving the cables offers a high output impedance. The frequency at which the system's response is down 3dB can be calculated by the formula f = 1/(2PiRC) (where C is in farads, R in ohms). With a 1k ohm source impedance, typical of the worst case when it comes to high-end tube and solid-state preamplifiers, the response will be down 3dB at 245kHz with Lapis, at 318kHz with solid-core copper, and at 57kHz with Lindsay-Geyer. All of these frequencies are above the audio band, of course, but the L-G is a little close for comfort. With a significantly higher source impedance, such as that offered by a typical passive control unit, the high frequencies will be rolled-off by this length of L-G cable. The Mod Squad's Line Drive, for example, has a maximum output impedance of 2050 ohms with the volume control set at 2 o'clock, which will give a –3dB point at 27.7kHz with L-G, the result being audibly dulled high frequencies.

But echoes and transient smearing are the order of the day's experiment. Lindsay-Geyer's putative time-smearing is frequency-dependent in that low frequencies produce more separated, less attenuated echoes than high frequencies. It would seem appropriate, therefore, to use a single-frequency sinewave which would suffer a discrete echo. However, to look for the presence of such echoes with a sinewave wouldn't be very informative as the human ear-brain is very poor at detecting echoes with continuous waveforms. Audio-frequency echoes would also occur very close to the stimulus waveform, and therefore would be very hard to detect with a 'scope. No, as the ear-brain is superb at detecting echoes with transient stimuli, I would choose pulses as my test signal and assume that though the constituent frequencies in the pulse would suffer varying delays, an echo effect of the strength suggested by L-G would still make its presence known. I used DRA Labs' MLSSA system to generate repeated unipolar rectangular pulses (footnote 1) which I then fed through the cable under test to a Heath 8-bit digital storage 'scope. (Although the MLSSA incorporates a 12-bit 'scope, this has a built-in anti-aliasing filter which could confuse things. It also cannot cope with input peaks that would saturate its ADC.) To capture the pulse shape centered around the 0V axis, I used the 'scope's AC input connection; its input impedance was 1M ohm.

For the first experiment, I used an impulse of some 7V peak amplitude and 19µs length fed to the 'scope via the solid-core copper cable. This can be seen as the top trace in fig.1, which features two such pulses. The impulse tail between the pulses features a degree of noise, probably due to the unshielded cable picking up some RF hash from the computers in Stereophile's lab. Expanding the vertical sensitivity by a factor of 20 gives the lower trace in fig.1. (Ignore the clipped top and bottom of the waveform; this is due to the signal exceeding the ADC's dynamic range capability.) Because the LSB was still toggling a little on this magnified trace, I averaged 32 such samples, so that the noise would fall away, allowing consistent features such as echoes and wrinkles in the wave shape or other such time-smearing artifacts to be made visible. None can be seen, however.

If these echoes do exist, then this measurement indicates they would have to be 64dB or more down from the level of the pulse; ie, at or below the LSB level in the lower trace, which is 4mV.

Fig.2 shows the same traces for the Lindsay-Geyer cable. With the exception of the slightly lower level of noise, they are identical to those in fig.1, with no echoes discernible. Fig.3 shows the same traces for the AudioQuest Lapis. A shielded design, this picks up less noise than the other two cables, but apart from that, again the traces are identical.

What if this pulse is too short to generate a visible L-G effect, its wideband frequency content resulting in any echo being smeared too much in time? I therefore set up a second series of tests with a longer pulse, approximately 240µs in length and of 10.2V peak amplitude. Fig.4 shows both the complete pulse (top trace) and a 100x-magnified version of the pulse tail (bottom trace) with the solid-core copper cable, while fig.5 shows the same curves for the L-G cable. This time, just in case it was the averaging of multiple traces that was contributing to the lack of visibility of any echoes, I just captured one trace. But note that even if there appear to be artifacts in the pulse tail, I didn't find these to be repeatable.

Note also that this time, the solid-core copper curves in fig.4 offer significantly less noise overall than the L-G curves in fig.5; DO did note in his review that the L-G cable does seem to be prone to picking up RF noise. Again, there are no obvious echoes. If there are any such echoes in the magnified solid-core copper trace, they would be buried in the noise, which has a peak amplitude of 4mV, some 68dB below the level of the pulse.

But, to be honest, we are digging around at the very limit of the measuring equipment's resolution here, and it is impossible to say what is real and what is a coincidental noise artifact. I would say that these results, while not disproving the Lindsay-Geyer hypothesis, indicate that any such transient smearing in conventional copper cable is going to be buried in the noise floor with any typical recording.

One point about all three cables used here strikes me, however. All are symmetrical in that the signal and ground are carried on identical conductors. Let us hypothesize that the Lindsay-Geyer effect exists, but that with symmetrical conductors, an equal but opposite echo will be produced in the ground conductor, this canceling that produced in the signal conductor. Isn't it then possible that the cable's symmetry is more important than whether or not the conductors are made from a permeable material? A final experiment suggested itself, therefore: to repeat the test with a physically asymmetrical RF coaxial cable, where this hypothetical, screen-generated anti-echo will not be opposite and identical to that produced by the central solid-core conductor.

Fig.6 shows the curves generated by the 'scope with the single unipolar pulse of 19µs duration and 7V peak amplitude applied to 5.1m of coax. (This cable has a 1.1mm-diameter solid-copper central conductor with four ground drain wires and a foil shield 4.5mm in diameter. The 5.1m length had a total measured capacitance of 330pF.) Comparison with figs.1, 2, and 3 shows no discernible difference. With the longer pulse (fig.7), it is apparent that the shielded cable picks up less RF noise, but the waveshape is fundamentally the same as in figs.4 and 5.

So, what to conclude? Evidence of effects which, if they occur, do so at the LSB level of a digital system, is hardly convincing. Further experiments with test equipment having a greater resolving power are planned, but if the Lindsay-Geyer effect is both so hard to find and so small if it does exist, can it really be subjectively important?—John Atkinson

---

Footnote 1: The MLSSA signal generator has an output impedance of 75 ohms, so early HF rolloff due to the cable capacitance will not be a problem with any of the cables under investigation.