Sorry, this post probably vies for being one of the longest ever, but I think it�s worth dropping a few thoughts into the mix here.

Screening
Screening in Ethernet cables seem to be getting a bad rap here. In fact, a contiguous earth screen running end-to-end is generally a good thing when it comes to preventing noise ingress, and only presents problems if you have different earth potentials on respective devices. The term �ground loop� or �earth loop� is a misnomer, as there is no full loop as such. Instead, if the grounds of two pieces of equipment are at different potentials, a small AC current can flow between those two items along the screen or earth of a signal cable. The signal on the screen then induces corresponding signals onto the signal line(s) and, in analogue equipment, the impact is easily heard as �mains hum�.

As for the earth screen carrying noise, it is no more likely to carry noise than any of the positive and negative going signal lines of the balanced pairs. More on this below.

Noise gets onto Ethernet cables from several sources:

1. Airborne RFI - all of us live in very 'dirty' RFI environments these days. Your own Wi-Fi router is banging out a pretty hairy-arsed signal at 2.4GHz and, in some cases, also at 5GHz. We all live within a km or so of mobile phone masts, which push out a lot of power at anything from 900MHz to 1,800MHz, 2.1GHz, 2.4GHz etc. You could safely say we�re in a high-frequency war-zone. RF is fabulously talented at inducing small electric currents in conductors all around it, more so where there is no shielding. Our Wi-Fi routers are particularly nasty, as you have high-amplitude RF being generated and transmitted within a few cm of Ethernet connections � guaranteed heavy noise ingress there.

2. Mains noise. Most mains noise gets onto domestic circuits in the last few hundred meters to the household, as the high-tension lines used over longer distances are much more RF-resistant. I've experimented a lot with mains and one thing I've learned is that a big percentage of all total noise potentially affecting our kit is right there on the good old mains line. Ironically, a lot of it is generated locally within our own domestic mains systems. Any slightly loose connection, e.g. behind sockets in the ringmain, or just 13A plug push-fit joints, can generate a lot of broadband RF, as the small gaps act like spark-gap transmitters, like those used in early radio transmitters.

3. Power supply noise - so many household appliances, especially computer related devices, now have switch mode PSU, which all bang out RF at around 500KHz. Most of these PSUs are cheap and really leak RF for England. This noise affects both the powered device itself and it pollutes your domestic mains system. In my dedicated mains supply, I make sure to avoid using any SMPSUs for this reason.
The real-word example we all know about is Meridian switching from SMPSU to linear between 808.3 and 808v5. Of course, there the SMPSU was high quality, but there were clearly big gains to be had from going linear and Meridian knew it.

Back to earth screens and noise. RF signals induce noise onto conductors by virtue of the oscillating magnetic field that goes with them. The oscillating field induces small currents in nearby conductors. In the context of an Ethernet cable or connection point (effectively an open door to RF ingress), RF is agnostic of what kind of conductor it affects, be it an earth screen or positive or negative-going signal line in one of the balanced pairs. All are equally affected, at least in the absence of screening. In fact, by its nature, the earth screen, if connected to properly grounded devices, will deal with RF noise better than the signal lines, as a good earth offers quite a good low-impedance or �suck out� route to a lot of RF noise. Of course, once RF noise is inside an Ethernet cable, it�s there to stay unless you interfere directly with the signal to remove it.

So, as mentioned elsewhere, breaking the earth screen with the GISO or other types of LAN isolator is of very limited significance in any sonic gains. The big gains come from the filtration which actively removes high frequency RF noise from the signal lines and the earth/screen. But screening in Ethernet cables is very important, as it significantly helps prevent RF ingress. There is better screening in CAT6 than in CAT5 and, in the case of CAT7, you�ll often get two layers of outer screening, in addition to the spec requirement of each of the four balance signal pairs also being independently screened. So the gains Chris and others who have made by switching from CAT5 to CAT6 are to be expected from an engineering point of view. I�ve experimented with Ethernet cable and have heard significant gain from CAT5 to CAT6 and further gain (though not as big) going from CAT6 to CAT7.

Digital signals transmitted in analogue domain
Digital signals, when transmitted over any transmission medium (be it air, vacuum or cable), travel as electromagnetic waves in the analogue domain. Inescapably so. Let�s assume the simplest case, that the base digital signal is perfectly square (not true in practice). Let�s also say that signal moves between 0 and 1 levels at a base or �fundamental� frequency of circa 12MHz (about right in AES/EBU carrying hi-rez audio). This signal, when transmitted, becomes a series of sinusoidal waveforms, known as a Fourier series. First, there is a sine wave at the fundamental frequency and let�s say that the amplitude of this first sine wave is notionally 1. Then there are many other similar sine waves, appearing at all the odd harmonics (3rd, 5th, 7th and so on) of the fundamental frequency and at reducing amplitudes. The diagram below shows how you can build up an approximation to a square wave by adding subsequent harmonics.

[img:left]I:\Documents\Hi-Fi\Hitchhikers\Square Wave Harmonics to 15th Odd Harmonic[/img]

For the integrity of a digital signal in transmission to be preserved, this series of analogue sine waves needs to be carried faithfully and accurately, out to many harmonics. In terms of cable design, materials and electrical characteristics, this brings things into perspective. Even if we go right down to the 51st harmonic, 2% of the total original signal energy still lies in that harmonic, and the frequency of this harmonic is circa 600MHz.

In practice of course, you�d get decent recreation of the digital signal at the receiving end using fewer harmonics than this, but the key point now is that, to preserve the digital waveform accurately, we need to carry a wide bandwidth of analogue signals, many of which are quite delicate. Also, being in the analogue domain, all these signals are subject to all the external and internal electrical, magnetic and mechanical influences that affect analogue cables and equipment. So you could quickly see why, in terms of bandwidth alone, CAT5 is removing a lot of important harmonics.

So, over a short run, the 100MHz bandwidth of CAT5 might be OK at a push, but it really isn�t good enough for longer runs. Even if the 100MHz bandwidth of the cable miraculously started exactly at the bottom of the wanted AES/EBU signal range (unlikely, as Ethernet is different to AES/EBU), you could only reach the 9th harmonic before hitting serious attenuation. The 9th harmonic contains over 11% of the signal energy.

Group Delay or Phase Delay
More issues here. Many will have heard of group delay in the context of speakers and speaker cables. In effect, different frequencies are carried at different speeds along a given conductor, meaning that different parts of signal arrive at different times. This is a bigger issue at higher frequencies, where wavelengths are shorter. In the context of one of the harmonics above, if one of them fell behind by 90 degrees (or a quarter of a wavelength), you�d be looking potentially at zero voltage where a notional 1 was expected. If the delay slipped out to 180 degrees, you�d have potential for a -1 instead of a +1 and, very quickly, the edges of the square wave fall apart and there is a destructive level of ripple, at which point, error correction has to resort to extrapolation. Minimising group delay ties into careful cable design and use of more expensive materials.

Skin Effect
OK, last point before I finally bugger off. When an oscillating signal travels down a conductor (i.e. as an electromagnetic wave), the magnetic currents act to �push� the electrical current toward the outer layers of the conductor. The strength of this push rises rapidly with frequency so that, at 100KHz, all the electrical energy is being carried in the very outer skin of the conductor. This is called Skin Effect and is well established in signal transmission. Given the frequencies we�re dealing with in AES/EBU signals (used between Sooloos source & endpoint), effectively 100% of the electrical energy is carried in a vanishingly thin outer layer of the signal conductors (microns). In effect, the signal is using only a tiny fraction of the total cross-sectional area of the conductor and this causes a dramatic increase in impedance, leading to heavy signal attenuation.

So the best cables are optimized around skin effect, in two basic respects. First, by increasing the ratio of conductor surface area to diameter (thicker conductors, solid core). Second, by using more costly materials such as silver, which is significantly more conductive than copper at very high frequencies. On top of that, getting really good results requires superbly smooth surfaces on the conductors, again down to micron level. I have some experience of these engineering issues, as one of my earliest jobs was in designing RF wave guides for use in airborne radar systems (albeit nearly 200 years ago). Of course, that�s a very different world to small Ethernet and audio conductors, but the same principles apply.

Find a spare short run of CAT5 cable, cut it open and look closely at the signal conductors (with a magnifying glass in the case of my shagged-out old eyes) you�ll see that the signal lines are multi-stranded, which is bad for skin effect: lots of small conductors, with low surface area to conductor size, along with other unwanted magnetic interactions. You�ll find solid core conductors in CAT7 cable, which offer a bigger proportion of skin to conductor size. This is exactly why, when it comes to the much-maligned �audiophile� Ethernet cables, some highly credible high-fi reviewers have reported quite dramatic improvements in sound quality in streaming applications. In light of this, even if very few of us would ever make the investment, the tech described by Audioquest in its Diamond Ethernet cable makes perfect sense. So I�m not surprised that Crion has heard benefits from his cables.

Of course, I�m not saying we�re all making a mistaking by not using thousands of pounds worth of high-end Ethernet cable, but there is certainly sound science behind the idea that they can improve things. That said, I can�t afford to invest in such cable myself, as my main run from MD600 to 808v5 is 20 metres. So I just went with the best industrial quality CAT7 I could find.

Please don�t think I�m under any illusion that I know even one hundredth of what the guys at Meridian know on this whole topic. I�m just drawing attention to some engineering foundations that show how all this stuff could have a very real impact on our sound.

OK time for me to bugger off. Thanks for reading :-)

Dave

808v5, MD600 & DSP7200.1
Dedicated audio mains supply