Miscellaneous

With every passing season, a new audiophile-grade network switch hits the market. These products, which can cost hundreds or thousands of dollars, do the same basic thing as network switches bought at Best Buy for $30 or so (except, in some cases, slower), but their manufacturers claim they are built to a higher standard to achieve better sound.

As with all signal-conditioning devices that operate completely in the digital realm—especially those that work at packet level (more correctly referred to as "frame-level" on the local side of the router, but that's a distinction that even few experts make)—the sonic efficacy of audiophile network switches is debated, the debate being, as usual, mainly between those who insist they hear a difference and those who insist, on theoretical grounds, that no difference is possible (footnote 1).

Years ago, I almost certainly would have been on the side of those who refused to listen on theoretical grounds. Since then, several things happened. As I got older and gained experience, I learned that I'm not as smart as I thought I was when I was younger. With this came a greater acceptance of uncertainty, a willingness to give up the certainty of scientific proof and accept things with less evidence than, say, a rigorous double-blind test (footnote 2). In this I am not alone in this industry: I often ask and have found it quite rare for those working in hi-fi R&D to require such rigorous evidence. They cultivate their listening skills, then believe and accept what they hear.

The next thing that opened my mind was a series of demonstrations of the efficacy of digital-side signal conditioning that had very clear results. The first involved replacing the power supply powering a purely digital audio server with an improved, lower-noise version, which brought about an obvious sonic change in the sound emerging from the speakers via the DAC. The second demonstration involved reclocking a USB signal. USB, like Ethernet, is a packet-based technology; to reclock a USB signal is to reclock the packets, not the audio data inside. With asynchronous USB transmission, the data are reclocked by the receiving device anyway—yet, again, I heard a clear change in the musical presentation, which I never would have expected. These experiences, but especially the last, opened my mind to the possibility that signal conditioning at the packet (rather, frame) level could bring about changes (and perhaps improvements; footnote 3) in the sonic presentation, via mechanisms that aren't yet known but that need not involve angels, demons, or poultry sacrifice.

The Silent Angel Bonn NX
I chose this device for my first network-switch test because it is the most ambitious I've encountered. If you're going to judge the sonic efficacy of new technology, you might as well start with one that seems the most likely to make a difference.

The Bonn NX ($3999) comes from the same Chinese company, Silent Angel, that makes the much less expensive Bonn N8 Pro ($599), which gained audiophile attention a few years ago. Silent Angel also makes servers, streamers, linear power supplies, and a few other things.

To quote Anthony Chiarella, Silent Angel's US agent, the cheaper N8 Pro "was conceived as an entry-level product and built to a price point, whereas the NX was conceived and built as a state-of-the-art product with cost no object." The two devices utilize the same core electronics—the same switching board (about which more in a moment)—but in the NX, both the power circuit and the clocking circuit were redesigned, the former to achieve lower noise and the latter to use what designer Chorus Chang calls "direct clock" technology.

What is "direct clock" technology? The N8 Pro utilized a 10MHz clock source, which must be converted to 25MHz to drive a 10/100/1000Base-T network switch (footnote 4). The NX, in contrast, uses a 25MHz clock, so no conversion is necessary. That 25MHz clock is of the TXCO variety and promises an "extra-low phase noise output." There's a 25MHz word clock input (and Silent Angel makes an external word clock that operates at 25MHz, the Genesis GX), presumably for when you want to coordinate the network switch's operation with other devices on the network, as is sometimes done in recording studios.

Where the NX goes farthest beyond the N8 Pro, in my judgment, is in its mechanical design. The N8 Pro was basically a board in a box with rubber feet. The NX enclosure is a sophisticated design intended to reduce vibrations. The chassis is double-layer, the inner layer made of unspecified "metal," the outer of aluminum, with vibration-dampening spacers in between. The NX's feet are very wide—2.36"—and made of stainless steel with embedded, vibration-reducing rubber O-rings.

But what about that core switching circuit? How is it improved? It was, Silent Angel says, "purpose-engineered for high-resolution streaming networks" and "delivers a high-purity signal to connected components, for a performance unattainable with typical network switches." The clock has its own power source, isolated from the power source for the switching circuit. There's a choke to remove high-frequency noise from the AC line. (A choke would work both ways, also keeping noise from the NX off the AC line.)

And the switching board itself? Chorus Chang told me in an email that "the switch board design is optimized to lower the crosstalk between ports and also to lower the noise from the digital part to the analog part. (The Ethernet signal is an analog signal.) As a result, the Ethernet signal quality is better." These improvements are common to both the Bonn NX and the N8 Pro.

The NX has an external clock input, presumably superior to the internal clock. Use of an external clock raises the perpetual question of which is more important, clock accuracy or clock proximity. The NX can also utilize an external power supply. (The one built in is said to be a "radar-grade" switching supply; radar-grade seemingly implies very low noise.) There's a ground lug to further reduce noise by providing a low-impedance path to ground for the enclosure.

$4000 is a lot for a network switch, but there's a lot going on in this one. The key question is whether any of what's going on matters sonically.

---

Footnote 1: There is, I'm happy to say, a third group: those who listen carefully and hear no change but have the humility not to assume that's definitive evidence that such devices are snake oil. These few admirable souls admit that just because they can't hear it, or haven't heard it yet, that doesn't mean it can't possibly work.

Footnote 2: I'm quite happy to dismiss without listening any proposed tweak that requires a mystical or paranormal "mechanism," something beyond physics, and there have been plenty of those over the years. I'm merely saying that I need not have a clear, specific physical mechanism in mind to accept the evidence of my senses. The specific explanation can come later. JA and I have both written about this previously; see, for instance, stereophile.com/asweseeit/787/index.html, stereophile.com/content/nothing-too-wonderful-be-true, and stereophile.com/content/audio-meet-science.

Footnote 3: I'm being careful not to call any change an improvement. The first key question is whether there is any change at all—a matter of fact if not easily verifiable. Improvement or not is subjective: Some may prefer separation while others prefer continuity. Indeed, at the first demo described above, one audience member insisted that the putatively inferior power supply resulted in the best sound. I didn't hear it that way, but I was not inclined to argue. You should seek the sound that you prefer.

Footnote 4: I'm no networking expert, but it isn't clear to me why the N8 Pro uses a 10MHz clock, nor how widespread the use of 10MHz is in audiophile network switches. 10MHz seems to be common for external word clocks in digital audio sources and is often used internally. Silent Angel's literature makes cryptic reference to the satellite guidance system known as GPS; it may be that this 10MHz clock is derived from the universal "atomic clock" that GPS utilizes. None of which fully explains why the N8 Pro uses 10MHz when by default, generic network switches need 25MHz, whether they achieve it directly or by converting 10MHz.