Digital Gear for the Musical Ear
by Nicholas Bedworth

Having spent much of the last year sorting out a computer-based playback system, it became rather apparent to me that there are still quite a few "issues" in making digital really sing. Although the audio community identified most of the inherent problems back in the 1970's—thirty plus years ago—understanding the possibilities of what can go wrong, from a scientific perspective is only the beginning: Delivering a truly satisfying digitally-sourced musical experience to one's living room still seems to be somewhat of a work in progress. That being said, major changes in technology and business models have also, somewhat unexpectedly, come together to help make top-quality audio available, accessible and affordable in ways that it never was before.

To be sure, digital's "promise" of convenience and ubiquity is already largely realized: purchasing and enjoying music in the home and on the go has been completely transformed. However, getting "digital" done correctly—meaning achieving the effortlessness and immediacy that permeate the musical art form—turns out not to be so easy. It can be done. It just takes a lot patience, with the gear and with yourself, along with a willingness to start from the fundamentals, and of course, a great deal of listening.

Consumer expectations for quality and convenience are rising steadily, to the extent that technology—such as truly horrendous audio codecs—being passed off as acceptable only a few years ago, no longer makes the grade (not that it ever really did). Many audio companies "who don't do digital" will simply be left behind, as market demands exceed their more traditional engineering abilities. These days, a product has to reflect considerable expertise in user interface software, interaction with networks and computer operating systems, and advanced signal processing. One also has to be at the cutting-edge of new areas in electronic design, which ranges from exotic miniature power supplies to ultra-high-quality clocks, or else. Few of the many "mom and pop" operations that characterize much of audio's high end have all these abilities under one technical roof. The ones that survive and flourish will.

So, how can one really take advantage of "going digital"? Well, it's definitely easier than it was a few years ago, and the results will most assuredly be better. Let's first start with a look at what can go wrong, and more importantly, understand how get it right.

Dancing with jitter bugs

If your media player software is serving up bit-perfect output—easy enough to achieve after a little futzing around—and the USB interface to the S/PDIF transport or DAC is asynchronous, these trivial excuses for mediocre digital playback are out of the way. Sure, it was certainly easy to plug a computer into the DAC, but my newly-minted digitally-sourced system just doesn't sound as well as it should. Sound, as it were, familiar? What to do?

For me, usual list of supposed "digital" remedies is really just tiresome marketing hype. We've all heard the assertions that certain playback software "has" to be used; supposedly "special" USB drivers are essential; "audio grade" hard drives are critical; "exotic" cabling will do the trick; that there's "only" one really usable operating system, and so on. Although some of these suggestions were plausible, but inadequate, attempts to Band-Aid the underlying problems, it's always better to drain the swamp, etc. None of these are really an "answer", but the resulting confusion has distracted consumers from the more important—and difficult to solve—issues that do remain. Just check out the forums.

In any event, from the perspective of the near future, this merchandising palaver may well be considered to have been the swan-songs of those who didn't make it in the "brave new world" of high-end digital audio. Like it or not, it is already with us. In the meantime, the wily "mammals", so to speak, namely a new generation of designers, as well as those veterans who stay ahead of the times, have the technical chops, energy and enthusiasm to do what really needs to be done. Some of these firms are already delivering terrific "next-generation" products that obviate the need for "solutions" no one really needs.

Many audio enthusiasts are already aware that reducing jitter—namely various forms of distortion in clock signals and the consequences thereof—is an enduring obstacle for making digital finally come alive. In my experience, reducing jitter to vanishingly-low levels, or perhaps more correctly, reducing clock phase noise, is essential. It makes the difference between music that sounds convincingly real in many ways, and the frustrating, music-like sound products that many are resigned to accept as the best that can be done.

It's rather odd that given the importance this parameter to the entire digital recording chain, very few manufacturers publish meaningful or technically-correct jitter specifications for their products. The "theoretical jitter" figures in advertising copy doesn't count—although at least this shows growing awareness that consumers increasingly expect to see such numbers. However, accurately measuring the clocks that shape the S/PDIF data stream, and trigger the DACs, does. To make these results reliably comparable, they need to be expressed for a given bandwidth, which in turns gives some idea of their real-world effect.

Truth, or consequences

The insidious, and occasionally invidious, effects of jitter may appear as subtle, or not so subtle, distortions occurring across the entire bandwidth of your system: This type of degradation, once present, is very hard to remove, and its audibility may be increased by the fact that it isn't hidden by music's natural harmonic structure. The intensity of these artifacts depend upon both the frequency of the jitter, in Hz, and its magnitude, usually in picoseconds, or mere trillionths of a second.

While there is a continuing debate over how much jitter, and what sort, can produce audible distortion, and under what circumstances, suffice it to say, the less there is, the happier everyone is going to be.

In establishing the taxonomy of jitter, one may examine how a given clock period differs from the ideal, or determine the cycle-to-cycle variation between adjacent clock intervals. These time-domain jitter specs are useful, especially those that are refined enough to differentiate the random, unbounded and deterministic, bounded varieties. Some jitter is physically inevitable in any realized design; others are definitely artificial and subject to remedy, if they can be identified in the first place. Oddly enough, even the bit patterns of the data being converted can affect clock jitter, which makes things even more "interesting".

A more generalized and heuristic jitter measurement is phase noise, which, simply put, is a frequency-domain analysis of the excursions a clock makes from the ideal, evaluated at increasing offsets from the nominal. These values are expressed in dBc/Hz, and can be summed as an RMS jitter spec for a particular bandwidth, say 0.1 Hz to 100kHz in the case of audio applications.

All these measurement techniques help build up a complete picture of what's going on, and are highly useful to designers as they strive to improve the performance of new products.

Given the pervasiveness of jitter and its apparent importance for sound quality, why don't all manufacturers publish appropriate jitter and phase noise specifications for their products? A slightly cynical response would be, "Because the numbers generally aren't that good". Somewhat more indulgently, one needs to keep in mind that assembling hardware and software test suites accurate enough for today's 24 bit, 192 and even 384 kHz audio sampling rates can easily approach six-figures, which is way beyond the reach of most manufacturers: The Symmetricom 5120a-01 phase noise analyzer that we use here runs around $37.5K, and time-domain digital oscilloscopes from Agilent or Tektronixfor measuring period jitter easily add another $30K to the bill. There's also a steep learning curve with respect to making and interpreting measurements correctly.

Still, it can be done, even by a new-comer to the industry, such as Audiophilleo, which builds high-performance USB-S/PDIF transports and processors. And of the established firms, TAD, with its very-high-end CD player and DAC, is also on that very short list of companies that publish comprehensive phase noise plots and properly-described period jitter. Given that the basic Audiophilleo device sells for just $500, while the TAD components run over $20,000, phase noise characterization is not just the province of the high-priced spread. It will be interesting to see if anyone else picks the implied challenge: if these two companies can already provide consumers with the critical information they need to make informed comparisons, why can't everyone else?

The Good, the Bad, and the Really Ugly

Expressed in terms of period jitter, what comes out of a Mac Mini or other typical consumer laptop TOSLINK connector runs around 1100 picoseconds RMS. If you connect such a source up to even a mid-range system, the results may be pretty ugly: Dynamics are constricted, the soundstage shrivels, and massed strings have all the allure of a chorus of dental drills. The entire presentation is hard, gritty and edgy. Sure, it's easy to connect everything up, no small consideration, but the results aren't very good at all.

Measurements made by experts such as John Atkinson, editor of Stereophile, round out the picture: Several hundred picoseconds or more of jitter are not associated with the kind of experience to which most high-end designers and music-enthusiasts aspire. John certainly has done everyone a big favor by pointing out the incredible range of jitter found in many popular products: One set of components he surveyed had as much as 4570ns of jitter, and it's fair to assume that it didn't sound all that hot. Fortunately, affordable, consumer-grade digital gear is evolving rapidly, and is already approaching levels of quality associated with the most demanding artisans in the audio field.

To get an idea how good the best can be, Robert Harley's recent encomium in TAS of Reference Recordings' "Doc" Johnson mentioned Keith's long-standing reputation for having the "Fastest ears in the West". Apparently he's well-known for his ability to differentiate equipment whose jitter specs are just handfuls of picoseconds apart. Keith's listening skills are certainly exceptional, as one might expect given his roles as the co-inventor of HDCD encoding, the designer of the iconic Pacific Microsonics Model One and Two A/D and D/A converters, and a contributor to various Spectral circuits.

However, even with my own limited experience, it's not that hard to hear the differences—on highly competent but not extraordinarily expensive equipment, such as Wilson Sashas, Weiss DAC and Odyssey Kismet monoblocks—between the best digital gear, with period jitter in the 3-10 ps RMS range, and the good gear, which runs around 100ps. Much above this level isn't going to sound that well based upon my experimentation. For those devices with nanoseconds of jitter, well, caveat emptor.

If you're in the market for a new DAC, which almost without exception these days will have a USB interface (USB being, after all, pretty much universal) be sure to press gently, but firmly, for the details of its jitter performance and how it was measured. If the designer doesn't know, it's not a good sign. Typically those who design their products from the ground up will be better informed; others, who essentially package interfaces around standardized kits (which, of course, may sound very well indeed) are less likely to know the details. Most new DACs, even modestly-priced ones, now support 192kHz sample rates, so if they can deliver low jitter, the high resolution will also add significantly to your musical enjoyment.

It Ain't Nothing Like the Real Thing, Maybe?

What to listen for when trying out a new USB S/PDIF interface or DAC? My recommendation is firstly to refresh your listening skills. Attend some live concerts of unamplified music. Take notes. And ask your audio friends to go along, and then listen again to the same music at home. At a recent performance here on Maui of the 100-strong Hawaii Youth Symphony I (as in one of three such symphonies in this very small state), their stellar playing renewed my acoustic memory of many elements of live music.

First on the list is the oh-so-smooth and delicate sound of massed strings, which in real-life is often shockingly different from what one hears through audio gear. This experience also brought home why you need a rather large number of them to be heard even in a medium-sized, 1200-seat auditorium: Violins are simply not very loud, taken one at a time. Together, yes, they're a bit louder, but it's as if a very large number of butterflies decided to take up humming. That these contraptions make any kind of audible sound at all is really quite remarkable.

The eclectic repertoire, perfect for my purposes, included a wild bacchanalia from Respighi's Belkis, Regina di Saba; a flute, oboe and horn trio; patriotic tunes; and parts of John William's score for Goblet of Fire. The variety of genre demonstrated just about every possible orchestral sonic style and element. One could readily observe how even a diminutive triangle "blooms" or radiates sound across an enormous volume of space.

When one of our local pop musical heroes, Maui-born Willie Kahaiali'i—a mountain of a Hawaiian fellow with a powerful, refined and well-controlled baritone voice of exceptional range—sang, incredibly enough, Ave Maria while clutching a tiny ukulele to his chest. Few in the almost-full house failed to be moved to the core. Just thinking of the performance brings tears to my eyes. But this is what music, all kinds of music, does to one. By the way, the 90-minute-plus concert was free, in the best venue in the state. Welcome to life in the Islands.

My point of bringing up these examples is not that what you hear at home has to sound "exactly like the real thing", or it isn't any good; this simply isn't possible, or even particularly desirable. Musical performance, whether by musicians or conveyed by electronics, always involves multiple layers of interpretation and nuance. We all know, but sometimes forget, that the moment after the sound hits the microphone, everything is irrevocably changed. However, to a surprising degree, on good recordings, to say nothing of high-end productions, all the fundamental sonic and emotional qualities of a live performance can also grace your living room, even with mid-range gear. They'll never be an exact match, but that's not the point: If the music moves you, and reaches the depths of your spiritual core, all is well.

As you will gradually discover, much of the realism, presence, and life-like thrill of real music comes increasingly to the fore as the jitter bugs are eliminated. When the sound goes beyond any sense of strain, into that zone of effortless flow, then the "suspension of disbelief" starts to set in and one is lost in the experience, rather than worrying about inadequacies of the sound. In short, with the best digital gear as your source, it's now possible to experience those essential qualities of music which have made it an enduring and essential hallmark of human civilizations over the millennia.

Resistance is Futile

What we're going through right now in the audio space is a far-reaching makeover of the status quo maintained for two or three generations. Doing everything "digitally" also creates many new opportunities: Expect even more surprises in lifestyle-friendliness, architecture, cost and performance during the coming months and years.

Specific examples of how the old is being changed, and improved, by digital, can be appreciated in the context of my personal "First Law of Electronics" which posits that all connectors are bad; an obvious corollary is that all cables are bad, but at least some do less damage than others. The "Second Law" is less is more, meaning keeping it really simple is an important virtue. Already, today's digital gear eliminates layers of cabling, numerous connectors, interfaces, switches, contacts, and so forth.

Many preamplifier functions are increasingly built-in to the S/PDIF transport or DAC. What happens when one simply runs cheap CAT5 Ethernet cabling to monoblocks by the speakers, and manages the sound from a wireless tablet with all the signal processing software it in? Well, you can almost do this today, and it's clearly the way of a very positive future.

Technology once again seems to be bringing "forward" the visual and audio art-forms of the recent and relatively distant past, while simultaneously making them universally accessible in ways, and at levels of quality, that simply weren't possible just a few years ago. This transformation was all accelerated by a disruptive, bottom-up "democratization" of the way popular music is merchandised and enjoyed, which spread quickly to even the highest of high-end audio experience, and rather remarkably made the entire range of musical experience considerably better. Despite the occasional grousing of our Luddite friends, does anyone really want to return to the 1970s?

We are certainly living in interesting musical times, and it's a great blessing.

Nicholas Bedworth

Biographical sketch

Music and electronics have both played important roles in Nick's life from an early age. He takes great delight in a variety of musical genres, listening during the day to a computer-based office system, with a second reference system in the living room for evenings. Nicholas is a native New Englander, and spent countless enjoyable hours in Symphony Hall with the Boston Symphony Orchestra and Handel and Haydn Society.

However, the continuing prospect of six months a year enduring cold and darkness, plus the sale of a high-tech scientific image-processing startup, prompted relocation to Hawaii. Eventually he landed on Maui, where the music scene is quite lively: Willie Nelson lives there part-time, sometimes materializing in North Shore hang-outs unannounced. And Maui's attractions bring in the likes of Björk, Aerosmith, Eddie Vedder, Elton John and many others, who are often looking for some R&R at the end of a long tour.

As a balance to his current "real" work in knowledge-visualization software, Nick also enjoys writing for the consumer electronics press. This endeavor is focused on helping the enthusiast community understand new technologies and product categories, starting with HDTV in the original incarnation of The Perfect Vision back in 1987 ("HDTV is just around the corner.").

These efforts continued to evolve with extensive coverage of emerging digital media technologies, including the first digital music server review in the The Absolute Sound, the Linn Kivor. Featuring a computer-based product in TAS in 2004 raised quite a few eyebrows, but the editor, Robert Harley, had his eye on the future. In the past few years, TAS, TPV, and Mix have published 20+ columns, equipment reviews, articles and other contributions along these lines.

Nick also enjoys the usual Maui enthusiasms of world-class yoga, meditation, swimming, and attempting to train a pair of green-cheeked conures, and an African Gray.

Nicholas Bedworth

Representative professional research and development activities

1. High-resolution 4096 x 4096 x 24 bit analogue-digital conversion of high-energy-particle-scattering imagery, with three-dimension reconstruction; measurement of vertices and calculation of particle momentum, with computer-graphic visualization. Brookhaven National Laboratory, Upton, NY and Yale University, New Haven, CT.

2. Electrophysiological investigation of sensory neurons and cybernetic modeling of transfer functions; software-driven signal analysis. Sonderforschungs-bereich fuer Kybernetik, and Max Planck Gesellschaft, Muenchen, BRD.

3. Multichannel electroencephalographic analysis with real-time Fast Fourier Transform, University of California, Los Angeles, CA.

4. High-speed, image acquisition subsystems for computed-axial-tomographic devices for industrial and medical application, including multi-ported memory; narrow-aperture sample-and-hold amplification, with low-jitter analogue-digital sample conversion. American Science and Engineering, Cambridge, MA.

5. Microprocessor writable control store and firmware; Quadex Technology, Cambridge, MA.

6. Ultra-low-dosage X-ray digital radiography scanner systems, American Science and Engineering.

7. Narrow-band spectroscopic analysis of hyper-sonic combustion flow fields, using nanosecond-scale pulsed dye laser illumination and image intensification. University of Illinois, Champaign-Urbana, IL; Georgia Tech, Atlanta, GA; Volvo, Goteborg, Sweden; University of California, Berkeley, Berkeley, CA.

8. Real-time Fast Fourier Transform image processing software for phase-contrast microscopy, Mayo Clinic, Rochester, MN, and University of California, Los Angeles, CA.

9. Standing-acoustic-wave optical signal processing apparatus, 10 megahertz sampling rate, 24+ bit quantization. Naval Research Laboratory, Washington, D.C.

10. High-phase resolution image scanning and networked distribution technology, Bell and Howell Corporation, Wooster, OH.